University of Southern California Dissertations and Theses

On the syntactic structure of Spanish noun phrases

(USC Thesis Other)

On the syntactic structure of Spanish noun phrases

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content INFORM ATION TO U SE R S This maniiscript has been reproduced from the microfilm master. UMI films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter face, while others may be from any type of computer printer. The qn ali^ of this reproduction is dependent upon the quaJi^ of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photogrsqihs, print bleedthrough, substandard margins, and inqjroper alignment can adversefy afreet reproduction. In the unlikely event that the author did no t send UMI a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note win indicate the deletion. Oversize m aterials (e.g., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand comer and continuing from left to right in equal sections with small overlaps. Each original is also photographed in one exposure and is included in reduced form at &e back of the book. Photographs included in the original manuscript have been reproduced xerographically in this copy. Higher quality 6" x 9" black and white photographic prints are available for any photographs or illustrations q>peaiing in this copy for an additional clmrge. Contact UMI directly to order. A Bell & Hwell Informaiion Company 300 North Zeeb Road. Ann Arbor. M l 48106-1346 USA 313.'761-4700 800/521-0600 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. ON THE SYNTACTIC STRL'CTLUE OF SPANISH NOUN PHRASES by Juan Martin A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOLTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (Spanish) May 1995 Copyright 1995 Juan Martin R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. UMI Number: 9614046 Copyright 1995 by Martin, Juan All rights reserved. UMI Microform 9614046 Copyright 1996, by UMI Company. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code. 300 North Zeeb Road Ann Arbor, MI 48103 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. UNIVERSITY OF SOUTHERN CALIFORNIA THE GRADUATE SCHOOL UNIVERSITY PARK LOS ANGELES. CALIFORNIA 90007 This dissertation, written by Juan Martin under the direction of hX s. Dissertation Committee, and approved by all its members, has been presented to and accepted by The Graduate School, in partial fulfillm ent of re quirements for the degree of D O CTO R OF PHILOSOPHY Dean of Graduate Studies September 1994 DISSERTATION CO ilTTEE f c f c t w U ' .... Chairperson . . . r . . T 7 . — «................... ............... R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 11 Acknowledgements First I would like to gratefully acknowledge the help and advice of the present and past members of my committee: Ramon Araluce, Audrey Li, Mario Saltarelli, Carmen Silva-Corvalan, and Maria Luisa Zubizarreta. I am especially indebted to Mario Saltarelli, my dissertation director, for all these years of academic advice and teaching. Looking back on these last seven years of academic work if anything emerges clear and neat it is his direction of my studies and his teaching. Several times in our talks about the dissertation I had the impression that he already knew the path I should follow but was training me in my own search. He skill fully convinced and directed me with the slightest of hints. Now, those slight hints seem to me to be the milestones of my dissertation, and the many hours that we spent talking about syntax the comforting company that every traveller needs on their journey. Carmen Silva-Corvalan deserves special mention for providing me with experience in different linguistic fields, constructive criticism, and a very professional role model. I am also very grateful for all her support during these years. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Ill I wish to thank Roger Andersen, Youssef Aoun, Marla Carreira, Alicja Goreka, and Hajirae Hoji for enlightening the course of ray studies at USC with their teaching. Jon Franco and Alazne Landa have been my fellow students in many courses during these years, and, above all, great friends. It has been a privilege to share these years with them. Pablo Albizu, Alfredo Arnaiz, Gorka Elordieta, and Marta Ona have provided me also with great friendship and valuable comments on my dissertation. Special thanks to Domnita Dumitrescu for all her help with the Romanian data. What appears in the dissertation is just the tip of the iceberg of many e- mails full of comments and data. Also, I am especially indebted to Helen Franks and Daniel Garcia for providing me with a room and friendship in the last year of my dissertation. Helen helped me a lot with my English and some cooking lessons, and Daniel was invaluable with his interdisciplinary knowledge. No matter what, he always came out with a valuable and precise answer. Among other things, he gave me the name for the "raise en abyme" effects. I am very grateful to the University of Guangzhou for giving me the opportunity of teaching there and, in particular, to the professors of the Spanish Department R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. IV for their help. They provided me with all the available means to continue my research. It was there where the affectedness analysis was born and developed. Special thanks are also due to my students who taught me the little Chinese that I know. They were enthusiastic learners, amazing flamenco dancers, and great travel companions. Their hospitality is unforgettable. Many thanks to my friends in LA and Madrid for their support, some of them versed in linguistics and always eager to talk about it. I am especially grateful to my family for these years of understanding and encouragement. Finally, I would like to dedicate this dissertation to Judy Dyer. Thank you for climbing mountains in search of Stone Buddhas and for fueling me to finish my dissertation. I hope it is worthy of you. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table of Contents Acknowledgements......................................il Abstract........................................... viii Chapter 1: Introduction and Theoretical Considerations..1 1.1. Introduction: NPs and their Cross-categorial Aspects......................................... 1 1.2. The Thematic Correspondence Hypothesis..........5 1.3. The Possessivization Principle.................18 1.4. Functional Categories of NPs....................20 1.5. The DP-Hypothesis..............................24 1.6. The possessor variable.........................33 1.7. The Model of the Grammar and the Full Interpretation Principle....................... 35 1.8. X’-Theoiy...................................... 37 1.9. The ta Theory................................... 40 1.10 .Typology of Nouns.............................. 45 1.11 .Organization of the Dissertation............... 48 Chapter 2: Argument Structure of Noun Phrases: Affectedness and Possessives............... 52 2.1. Introduction................................... 52 2.2. The Lexicalist Hypothesis and the Syntactic Structure of Noun Phrases...................... 54 2.3. The ©-marking of Nominal Arguments and Case Theory................................ 67 2.3.1. Theta Marking and Inherent Case Marking 67 2.3.2. Grimshaw’s (1990) Analysis................. 68 2.3.2.1. Theta-raarking of QPs................... 68 2.3.2.2. Passive Nominals and the Affectedness Constraint............................. 70 2.4. Towards a re-interpretation of the Affectedness Constraint.................... 80 2.5. Af fectedness Effects in Spanish.................91 2.5.1. Passive Nominals and Quantifier Raising 91 2.5.2. Affectedness and Possessivization.......... 99 2.5.2.1. The Derivation of Simple Event Nominals and Result Nominals........... 99 2 . 5.2 . 2. PRO and the Af fectedness Constraint.... 120 2.6. The Lexical Representation of Nominals.........129 2.6.1. Introduction.................. 129 2.6.2. The Lexical-Syntax Interface.............. 130 2.6.3. Psych-predicates...........................142 2.6.4. An Aspectual-Referential Theory of Role Assignment.................................154 2.7. On the nature of de............................ 158 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. VI Chapter 3: Agreement within NP: the Double Articulation of DP and the "mise en abyme" Structures.............172 3.1. Introduction.................................. 172 3.2. Previous Studies of Nominal Agreement..........180 3.2.1. Valois (1991)............................. 180 3.2.2. Bernstein (1993)...........................184 3.3. Adjective Placement........................... 185 3.3.1. Introduction.............................. 185 3.3.2. Adjectives and Phrasal Stress..............188 3.3.3. Previous Proposals........................ 197 3.3.3.1. Abney (1987)...........................197 3.3.3.2. Bernstein (1993)...................... 201 3. 3. 3. 2.1. Intensional Adjectives............ 201 3.3.3.2.2. On the Difference between Intensional Adjectives and Attributive Adjectives............ 203 3. 3. 3. 3. Cinque (1993)......................... 211 3.3.3.4. Valois (1991)......................... 216 3.3. 3.5. Giorgi and Longobardi ( 1991 )..........218 3.4. N°-movement................................... 220 3.4.1. Introduction.............................. 220 3.4.2. N“-to-D° Movement......................... 221 3.4.3. N“-movement to an intermediate position....233 3.5. Agreement within NP and the "mise en abyme" Structures.................................... 244 3.5.1. Introduction.............................. 244 3.5.2. The DP system............................. 251 3.5.2.1. Introduction.......................... 251 3. 5. 2. 2. The Romanian Case..................... 253 3. 5. 2. 3. An Analysis for Spanish................259 3. 5. 2. 3.1. The Double Articulation........... 259 3. 5. 2.3. 2. Wh-in-situ................ -61 3.5.2.3.3. Affective Possessives............. 261 3. 5. 2. 3. 4. Coordination Facts................ 266 3.5.2.3.5. Nominal Elision...................270 3. 5. 2.3.6. Typological Evidence.............. 288 3.5.3. Predication and XP-movement within DP: the "mise en abyme" effects................297 3. 5. 3.1. Introduction.......................... 297 3. 5. 3. 2. 6-Theory, again....................... 304 3. 5. 3. 3. Adjective Ordering Restrictions....... 314 Chapter 4: Extraction from NPs in Spanish.............329 4.1. Introduction.................................. 329 4.2. The Data...................................... 329 4.2.1. Cinque’s Generalization................... 329 4.2.2. Specificity Effects....................... 334 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. v i l 4.3. Previous Studies.............................. 338 4.3.1. The Specifier Hypothesis: Cinque (1980) and Longobardi (1991)..................... 338 4.3.2. Towards a Syntactic Solution of Specificity: Bowers (1988) and Hudson (1989)............................. 355 4.3.3. Diessing (1992)........................... 361 4.4. On the Difference between Restrictive and Nonrestrictive Clauses........................ 366 4.5. Spec-head Agreement, the Subjacency Condition, Specificity Effects................ 372 Conclusions......................................... 379 References.......................................... 384 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Vlll Abstract This dissertation studies lexical and syntactic aspects of Noun Phrases in Spanish under the Principles and Parameters framework. In particular, it investigates the similarities and asymmetries between the verb and the noun. Under a Configurational approach, preferable in theoretical terms, my study tries to explain the different behavior of Noun Phrases shown by different scholars in previous studies. I conclude that although verbs and nominalizations share the same lexico- conceptual structure, they differ with respect to number and type of syntactic functional categories and syntactic behavior of the lexical head. While nominal expressions are a continuum towards reference, where prepositional functions are mapped onto properties, verbs undergo predication, i.e. property expressions are mapped onto prepositional functions. Within nominal expressions all the syntactic layers, apart from non-restrictive modification, are computerized by the determiner system as a result of the intersective nature of modification. On the other hand, the syntax of verbs reflects the combinatorial nature of predication. Hence, their functional categories are not a referential R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. IX continuum and can act independently. This entails that N’ can be a semantically saturated expression, but V’ cannot. This is the case of Nouns that project an affected object. I claim that affectedness is defined by the ability of a complement to semantically saturate a lexical head. In this way we can explain why affected objects can saturate the noun functional system but not the verbal one. The stacked nature of nominal modification is captured through a successive spec-head relationship between modifee and modifier. This relationship occurs at the syntactic level of representation in languages with overt nominal morphology. Otherwise it occurs at LF. This explains the word order parameter between Romance and English adjectival modification. Additionally it allows us to further define previous proposals for two types of determiners: a weak and a strong determiner position. The former enters in a spec-head relation with a modifier, whenever it is present, while the latter dominates that modifier. Finally, this analysis- and the conditions on Spec-head agreement account for the extraction facts in Spanish. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chapter 1 Introduction and Theoretical Considerations 1.1. Introduction: NPs and their cross-categorial aspects The purpose of this study is to define the structure of Noun Phrases in Spanish at the different levels of representation of grammar within the framework of the Principles and Parameters Theory. In particular, their differences and similarities with the verbal category. In this regard, there are two tendencies in recent works within the Principle and Parameters framework: the Configurational Hypothesis and the Semanticist Hypothesis. The point of departure for this debate is Chomsky's (1970) article 'Remarks on nominalizations', in which he claims that verbs and nouns share the same lexical properties and that this fact determines some syntactic similarities between them. Along these same lines, the Configurational Hypothesis (Anderson, 1979, 1984; Cinque, 1980; Torrego, 1985; Giorgi and Longobardi, 1991; Picallo, 1991; Valois, 1991) argues that the argument structure of nouns is systematically mapped onto hierarchical syntactic structures, similar to those of R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 verbs. This assumption was formalized in (1) by Giorgi and Longobardi (1991): (1) The Configurational Hypothesis A. It is possible to identify, within NPs, definite 0-(and non ©-) positions at various levels of hierarchical attachment : whenever an element of the N frame appears in a position arguably different from the one where it should be projected at D-structure, its displacement must, then, be governed by the general conditions holding on antecedent-trace relationships created by "Move a", moreover the binding of anaphors and pronouns in NPs obeys the same constraints observed in clauses. B. The ©-structure of Ns (their ©-grid and the condition on ©-assignment) strictly parallels that of Vs so that the differences appearing on the surface must be due to the intervention of other modules of Grammar which determine some systematic variation. Giorgi and Longobardi (1991) point out three of these differences: a.) NPs project an external argument optionally, since their argument status allows them not to be licensed by predication; b.) Nouns are not structural governors or case markers (Kayne 1981; Chomsky, 1986a; Rizzi, 1990); c.) Nouns do not have passive morphology. The semanticist hypothesis (Safir, 1987; Zubizarreta, 1987) posits that the semantic functions a Noun is able to assign appear to be more radically and mutually dependent than those assigned by the verb. This R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 point will be illustrated by the syntactic behavior of psych-nouns in section 2.6.3. The configurational approach is theoretically preferable to the semanticist one, provided that explanatory adequacy is attained, since special stipulations are not required for the syntax of Noun Phrases. It is also important to define which deviancies might be allowed for the Configurational Hypothesis not to be invalidated. Stowell (1989:233) defines the similarities and dissimilarities between Noun Phrases and other lexical categories in the following words: "nouns seem to have a dual nature: like adjectives and verbs, they seem to have an internal argument structure (8-grid); but unlike adjectives and verbs, they seem to be able to function as referring expressions." This difference between nouns and adjectives is probably related to another one: "nouns normally require specifiers (quantifiers, numerals, determiners, or possessive CNPs), whereas adjectives do not" (Stowell, 1989:234). Stowell describes N as uniformly predicative and D as uniformly referential. I would like to claim in this dissertation that some of the asymmetries between verbs and nouns, such as government properties , passive morphology, asymmetries on extraction can be derived from the dual nature of the R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4 noun. Nouns have a different syntactic behavior than verbs due to the fact that they are referring expressions as well. In other words, they are subjects and predicates at the same time. They can project an argument-predicate structure, but, at the same time, they are a subject of that predicate-arguraent structure, since this predicate- argument structure specifies reference to the noun. In the Romance languages, this dual relationship translates into the coindexation with an expletive subject, DP, and into movement to Spec of the immediately dominating phrase. In contrast, verbs move to the head of the immediately dominating category, as proposed by Pollock (1989) and Chomsky (1991). In sections 1.2 and 1.3 I will present the evidence that has been given in earlier literature, specifically in the work by Giorgi and Longobardi (1991), with respect to the fact that verbs and nouns share the same Argument Structure. In sections 1.4 and 1.5. I will refer to the previous proposals about functional categories within a nominal expression and the syntactic status of the determiners. In this chapter I will make reference to some of the proposals that I make in the rest of the dissertation. This is necessary because these proposals are relevant to chapter 2. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5 1.2. The Thematic Correspondence Hypothesis The Thematic Correspondence Hypothesis states that there is a strict correspondence between the mapping of nominal arguments onto their syntactic positions and the mapping of verbal arguments onto their syntactic positions internal to VP, assuming the Internal Subject Hypothesis elaborated by Kuroda (1988). This entails that at the Semantic level of representation we already have a structured list of the arguments projected by a lexical item, as proposed by Zubizarreta (1987) against Rappaport (1983). Hence, if we assume a strong version of the Lexicalist Hypothesis (Chomsky, 1970) Nouns and Verbs should share the same argument structure in terms of hierarchy. This is what Giorgi and Longobardi (1991) conclude in the first chapter of their book. Their main proposals are the following: a.) one of the arguments specified by the thematic grid of a non-ergative Noun must occupy a higher structural position, the subject position, than the other arguments within the Noun Phrase. Nouns assign this position to the same argument as their lexically related verbs do within the VP; b. ) the other arguments are assigned a position internal to N’ , an intermediate projection that defines their c- command domain, in accordance with Reinhart’s (1976) notion of c-command; c.) the prepositions ^ and a are R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6 dummy prepositions, since they do not define a c-command domain. This constitutes a problem in Rappaport’s (1383) proposal which states that all prepositions that introduce a nominal argument restrict their complement semantically, if we assume that semantically restricted prepositions define a c-command domain; d.) arguments projected by an ergative noun are assigned syntactic positions under N’. Giorgi and Longobardi base their conclusions on binding relations among the nominal arguments. Conclusions (a) and (b), and the fact that ^ does not define a c-command domain are illustrated by (2).^ (2) L’opinione di luij della madre di Gianni- è Gianni’s mother’s opinion about him is too ♦His opinion about Gianni’s mother is too troppo lusinghiera. flattering. flattering. As we can see in the English counterparts to (2), the only possible reading of (3) is when di lui is interpreted as the THEME. The explanation of the only possible interpretation of (2) is that there is an intermediate projection, N’ , that contains di lui but not della madre di Gianni, as in (3), and that this ^ The sane facts can be observed in French (Valois, 1991) R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 7 intermediate projection is a c-domain, otherwise we should expect a violation of Principle C of the Binding Theory. (3) NP !\ I N’ i\ I N° opinione Hence, Reinhart’s (1976) definition of c-command, in (4), makes the correct prediction for this particular case over Aoun and Sportiche’s (1983) definition, in (5), if we assume this structure of NP. The first branching node for the pronoun in (3) is N’, assuming that the status of di is a dummy preposition. This entails that the pronoun cannot c-command its coreferential R-expression, which would have resulted in a violation of Principle C. Aoun and Sportiche’s definition would give rise to a violation of Principle C, since the maximal projection dominating the pronoun is NP, which also dominates Gianni. as shown by (3). If we adopt a Larsonian syntactic mapping of arguments, both definitions account for (2). (4) Node A c(onstituent)-commands node B iff neither A nor B dominates the other and the first branching node which dominates A dominates B. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. (5) a c-commands B if every maximal projection dominating a also dominates B. Conclusion (c) follows from the fact that the only reading available in (2) is the one where la madre di Gianni is the subject. Otherwise, coreferentiality between the pronoun and Gianni is ruled out, since ^ does not define a c-domain because of its dummy status. The properties of the non locative preposition a seem to be the same as the preposition since in (6) there is a violation of principle C of the Binding Theory, given that the preposition a does not define a c-domain. (6) *La restituzione a leij del figlio di Maria; ’The restoration to her of Maria’s son (da parte dei rapitori) (by the kidnappers)’ As (7) illustrates, Giorgi and Longobardi’s analysis can be extended to Spanish, since the only reading possible when the pronoun and the R(eferential)- expression are coindexed is that of the pronoun as an object 2 The most usual way of expressing the intended meaning in ( 7) is as in ( i ) if we use two ^ phrases. Otherwise it would be as in (ii). Language strategies might be involved in the preference for this form. (i ) La opinion de Juan; de su; madre. 'The opinion about Juan ot his mother.' R eproduced witti permission of ttie copyrigfit owner. Furtfier reproduction profiibited witfiout permission. (7) La opinion de élj de la madre de Juanj es Juan’s mother’s opinion about him is too *His opinion about Juan’s mother is too demasiado halagadora. flattering. flattering. If we interpret the pronoun as the subject, there is a violation of Principle C, as pointed out by Giorgi and Longobardi. In the object interpretation, there is no violation of any principle, although the sentence is not easily interpretable. The sentence has an easier interpretation if we focus the pronoun by moving it to the right and assigning contrastive stress to it. Again, we obtain the same kind of facts as those presented by Giorgi and Longobardi in (2). We can observe in (8) that the only reading available for the pronoun is that of object.^ (8) La opinion de la madre de Juan, de ELj es Juan’s mother’s opinion about him is too *His opinion about Juan’s mother is too demasiado halagadora. flattering. flattering. (ii)La opinion de Juanj que tiene sUj madre. 'The opinion about Juan that his mother has,' Saltarelli (p.c.) has similar judgements for Italian as mine for Spanish, R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 0 The pronoun, after rightward movement, c-commands the R- expression in (8). However, it c-commands the R- expression from an A’-position, hence, it cannot induce a violation of Principle C. Notice that a subject interpretation of the pronoun is out since it violates Principle C. In this case, the original trace c-commands the R-expression from an A-position. Therefore, Giorgi and Longobardi’s analysis is correct after these refinements for Spanish. We can observe also that the mapping of nominal arguments onto the syntax has a strict thematic correspondence with the mapping of verbal arguments onto the syntactic positions internal to VP, as (9) illustrates. (9) Race un aho, opinaba de él la madre de ago one year, thought about him the mother of ’One year ago, Juan’s mother thought that he was Juan que era demasiado malo. Juan that (he)-was too bad too bad.’ The Lexicalist Hypothesis (Chomsky, 1970) puts forward the need for the existence of some redundancy rules in the Lexicon that can capture generalizations among sets of words with the same lexeme, but belong to different lexical categories. Rappaport (1983) and Rozwadowska (1988) claim that a verb and a noun with the same lexeme share their argument structure. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 11 Following the models of grammar proposed by Zubizarreta (1987) and Levin and Rappaport (1988), in which there are two levels of lexical representations, a semantic and a syntactic one, the Lexico Conceptual Structure is the level of representation where Nouns and Verbs share their Argument-structures. Thus, different syntactic rules, depending on the lexical category of the word, will apply in order to map the LCS onto the syntactic component of the grammar. For instance, the external argument of a verb will be projected into the syntactic lexical structure by means of Linking Rules, as proposed by Zubizarreta (1987) and Levin and Rappaport (1988). As noticed by Zubizarreta (1987), Nouns fail to undergo Predication. Since they cannot function as Lexical-predicates, they lack a representation at the syntactic lexical structure. Rappaport’s thematic constancy principle will account for their insertion in the syntactic level without a syntactic lexical structure. The parallelism of the syntactic structures between (8) and (9) supports the Lexicalist Hypothesis and also the hypothesis that claims that the Argument- structure is already structured at LCS (Zubizarreta, 1987). Giorgi and Longobardi (1991) give further evidence for the Thematic Correspondence Hypothesis, which I R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 12 summarize below. This evidence is based on different subject-object asymmetries. In (10) the anaphor can only be interpreted as the theme of the description. (10) La descrizione di se stesso, di Giannij. John’s description of himself. In Spanish we have similar facts as those shown by (8) and (9), as illustrated by the counterparts to (10) in (11). An analysis similar to that for (8) and (9) is suggested. (11) a. La descripciôn de si mismo (Theme) de Juan (Agent). b. La descripciôn de Juan de SI mismo. The next piece of evidence is based on the distribution of proprio. Proprio can behave either as a long-distance anaphor or as an anaphor bound within a clause. In (12) the phrase that contains proprio must have the theme interpretation, otherwise proprio cannot be bound. (12) La descrizione di Mario, della propria; madre. Mario’s description of his own mother. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 3 In Spanish, propio is only an emphatic particle, thus, this piece of evidence is not relevant in the case of Spanish. The distribution of epithets gives us another piece of evidence. In (13), the epithet must be interpreted as the theme in order to avoid a violation of Principle C.^ (13) La descrizione di quello stupidO; della madre John’s mother description of that stupid di Giannij. person. The clearest evidence for the structural asymmetries between an external argument and an internal argument within a nominal expression is given by the distribution of pronouns bound by a quantifier and by the sloppy reading of a pronoun. In (14) the quantified NP must bear Judgements in Spanish vary among native speakers. In this case, word order or contrastive stress do not seem to be a factor. Speakers that accept the coreferentiality between the R- expression and the epithet do not make a difference between a presubject position of the pronoun and a contrastive postsubject position. Again, I rather prefer the construction with the coreferential possessive (see fn . 2). In this case the pronoun is necessary for both positions, the presubject position and the postsubject position, as ( i) and (ii) illustrate. ( i) La descripciôn del estupido de JuaUj de SU| madre. The description of-the stupid of Juan of his mother (ii) La descripciôn de SU| madre del estùpido de Juan^. The description of his mother of-the stupid of Juan R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 14 the agent role, since the NP containing the bound pronoun must be c-commanded by the quantified antecedent in order to avoid a weak cross-over violation at LF. (14) La descrizione di ogni ragazzoj di sua^ madre. i.*His mother description of every boy. ii.Every boy’s description of his mother. Similar distribution is shown by the Spanish translation regardless of word order as (15a) and (15b) illustrate. Nevertheless, notice that when the internal argument occupies the position after the external argument, it must receive contrastive stress. (15) a. La descripciôn de cada niho; de SU| i.*His mother description of every boy. ii.Every boy’s description of his mother, MADRE/*madre. b. La descripciôn de suj madre de cada ninoj i.*His mother description of every boy. ii.Every boy’s description of his mother. Word order does not affect the sloppy reading of a pronoun either. Following Reinhart (1976), in order for a pronoun to have a sloppy interpretation it must be c- commanded by its antecedent, since it is a bound variable. Again, the bound pronoun in (16) must be contained in the di+NP that is the internal argument so as to have the sloppy interpretation. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 5 (16) Ti ho esposto I’opinione di Gianni di sua i. I revealed to you Gianni-’s opinion about ii.*I revealed to you hiSj wife’s opinion about raoglie, non quella di Franco. hisj wife, not Franco:’s opinion about his- wife Giannij not hiSj wife opinion about Francoj. Again, Spanish NPs show the same kind of asymmetries, as (17) illustrates. (17) Te he revelado la opinion de Juan de su MUJER/ i. I revealed to you JuaUj’s opinion about his: ii.*I revealed to you hiSj wife’s opinion about *mujer, no la de Pedro. wife, not Pedro-’s opinion about his- wife. JuaUj not hiSj wife opinion about Pedroj. As we see in (17), the internal argument is placed after the external argument and must receive contrastive stress to obtain the bound reading, otherwise, the only possible reading of the internal argument is the appositive one. If we have the appositive reading, then the interpretation of the internal argument of Pedro’s opinion is free. Therefore, we can have an accidental reading in this case, where Pedro’s opinion is about his own wife, but not a bound reading. Notice that, if the internal argument is preposed in relation to the external one, we can have a bound reading of the internal argument, as in (18). R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 6 (18) Te he revelado la opinion de su mujer de Juan, i. I revealed to you Juan^'s opinion about his, ii.*I revealed to you hiSj wife’s opinion about no la de Pedro. wife, not Pedro-’s opinion about hiS: wife. JuaU| not hiSj wife opinion about Pedro,. In accordance with the Thematic Correspondence Hypothesis, we should also expect subject-indirect object asymmetries. The Thematic Correspondence Hypothesis predicts that an External Argument will be projected into a higher syntactic position than the Indirect Object, a+NP. within an NP, parallel to what happens within a sentence. (19) shows that an anaphor in the NP introduced by the preposition a, can be bound by an external argument, but not the other way around. (19) a. La llaraada telefonica de Juan^ a si mismOj. The phone call of Juan to himself. b.*La llamada telefonica de si mismoj a JuaUj. The phone call of himself to Juan. c.*La llamada a Juan; de si mismoj. Furthermore, the Thematic Correspondence Hypothesis predicts that, since ergative verbs project only internal arguments, their related nominals would also only project internal arguments. (20) shows the case of a ditransitive nominal, where the direct object and the indirect object mutually c-command each other. If an anaphor and its R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 7 antecedent mutually c-command, no disjointedness effect arises, as posited by Giorgi and Longobardi (1991). (20) La restituciôn a Maria por el psicoanalista The restoration of herself to Maria by the de SI misma. psychoanalyst. The prediction that ergative Ns project their arguments into internal positions is illustrated by (21). Both arguments are able to bind an anaphor in the other argument position. Therefore, we must consider that the de+NP occupies a similar position to that of the internal arguments of transitive predicates. (21) a. La apariciôn en suenos a Mariaj de SI The appearance of herself to Maria in her misma^. dreams. b. La apariciôn en suenos de Maria^ a si The appearance of Maria to herself in her mismaj. dreams. In summary, the Binding facts show us that the argument structure within NPs strictly parallels the one within VPs. These binding relations may be altered in the case of VPs through passivization or mapping of an object onto the preverbal position, in Spanish, as (22) illustrates. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 18 (22) a. Cada estudiantej fue descrito por suj hermano. b. A cada estudiantCj lo describiô suj hermano. Each student was described by his brother. Notice that the binding relations that are altered are those of anaphors, not those of bound pronouns. This suggests that the encoding of scope relations happens at a different level of representation than that of anaphors. Passivization occurs at the Lexical Syntactic level of representation, which NPs lack, as posited by Zubizarreta (1987). In Martin (1992), basing my argument on Saltarelli’s (1990) analysis for psych-verbs, I propose that the direct object in (22b) has been mapped onto the preverbal subject position at NP-structure (van Riemsdijk and Williams, 1981). Therefore, we should consider the direct object as the subject of the predicate in (22b). This position is occupied by the internal argument at the output of the Lexical Syntactic level of representation, or NP-structure in the model of grammar proposed by van Riemsdijk and Williams. 1.3. The Possessivization Principle There is a semantic function, possession, that does not appear in VPs, but can be borne by a de+NP within a result nominal. As we have seen above, the agent is R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 9 mapped onto a higher position than the theme. The de+NP that bears the possessive function is mapped onto the highest position as shown by (23) and (24). (23) a. El cuadro de cada coleccionistaj (possessor) Each collector's picture of his favorite de sUj artista favorite, painter. b.*El cuadro de suj coleccionista (possessor) His collector’s picture of each favorite de cada pintorj favorite (agent), painter. (24) a. El cuadro de cada coleccionista, (possessor) Each collector’s picture of his favorite de su. ciudad favorita (theme), city. b.*El cuadro de suj coleccionista (possessor) Its collector’s picture of each city. de cada ciudadj (theme). The Thematic Hierarchy in (25) reflects Giorgi and Longobardi’s binding facts. (25) Possessor > agent > internal argument This Thematic Hierarchy reflects the facts regarding possessivization also, as shown by (26) and (27). Only the highest element in the hierarchy can be possessivized. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 20 (26) a. Su (agent) retrato del rey (theme). His portrait of the king. b.*Su (theme) retrato de Picasso(agent). His portrait of Picasso. (27) a. Su (poss.) retrato de Picasso (agent/theme). His portrait of Picasso. b.*Su (agent/theme) retrato de Getty (poss.). His portrait of Getty. In accordance with these facts, Giorgi and Longobardi (1991:68) propose the Possessivization Principle, in (28), based on the structural facts. (28) Possessivization Principle The only phrase allowed to appear as a possessive is the hierarchically highest genitive argument of an NP. 1.4. Functional categories of NPs Another more recent issue linked to the syntax of the Noun Phrases is the need to define their functional categories and their syntactic function and relevance. Again, their parallelism with the traditional functional categories of the verb, CP and IP, has been studied by several authors (Abney, 1987; Malien, 1989; Valois, 1991, among others). Abney (1987) argues that DPs are similar to IPs, since the _ls morpheme in D“ assigns case to the nominal subject, in a similar way to that in which subjects get case in a clause. In more recent work (Valois, 1991), DP has been considered the counterpart of R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 21 CP, and NumP the counterpart of IP. Valois considers DP to be the equivalent of CP because its specifier is the escape-hatch for extraction, while NumP is the counterpart of IP because it has the number features and in some languages there is N°-to-Num° movement. These languages are the same ones that have V°-to-I“ movement. N°-to-Num“ movement has been based on the position of adjectives (Siloni, 1991; Valois, 1991; Picallo, 1991; Cinque, 1993) and on the sloppy interpretation of pronouns (Valois, 1991). Pollock (1989) and Chomsky (1991) have argued that the position of adverbs in French and English indicates to us that in French there is V°- to-I°, while in English there is Affix-hopping. In a similar way, it has been argued that the position of adjectives tells us whether there is N°-to-Num° movement or not. As proposed by Siloni (1991) for Hebrew, adjectives are projected to the left on the Spec-of-NPs. N°-to-Num° movement leaves adjectives in the right position of Nouns at S-structure in Romance languages. On the other hand, in English, adjectives occupy the prenominal position. The second piece of evidence for N°-to-Num° has been given by Valois (1991) and it concerns the availability of a sloppy reading of a pronoun inside NP. The sloppy identity reading is available in the second conjunct in R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 22 (29), but it is not available in (30). In accordance with Valois (1991), this is an indication that the Noun has moved out of the NP shell in the French case, but not in the English case. (29) J’ai vu la photo de Paul de son chat ainsi que le portrait de Jean. (30) I saw Paul's picture of his cat as well as John’s portrait. In Spanish, it is not clear to me that we can have the sloppy identity reading. Despite some parallelism between the syntactic behavior of nominal and verbal functional categories, we must consider that the nature of the nominal functional categories and the verbal functional categories, as they are established in current literature, is quite different. Functional categories within nominal expressions seem to be interdepent among themselves and to form a continuum towards referentiality. NumP seems to be relevant to the notion of partitiveness, in accordance with Bull (1965), and DP seems to be related to the existential reading of a Noun. The notion of partitiveness is essential to the notion of existentiality. Furthermore, determiners can have a quantified reading or a cardinal reading, in the sense of Milsark (1974). Some authors, such as Bowers (1988) and R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 3 Hudson (1989), have suggested two different syntactic positions for the determiner, defined by their semantic interpretation. It might be interesting to change the direction of the research, and see if there is anything similar in the verbal functional categories. IP has some of the properties of DP, as proposed by Abney (1987), and NumP could be equated with an AGRP with aspectual properties (see Martin, 1992, among others). I will not pursue the definition of the parallelism between verbal and nominal functional categories in this dissertation. Further research is needed. Above all, we need to define the functional categories of each lexical category, and then determine their similarities. We should not forget that verbs undergo the lexical-syntactic process of predication and that this process can give rise to some syntactic positions licensed by functional categories that the noun lacks. NPs and their constituents show a considerable dependence on the DP for different syntactic processes, such as case assignment, extraction, etc. My objective is to define the syntactic and semantic properties of DP and its syntactic configuration. I will adopt the DP-Hypothesis proposed by Abney (1987) and define two syntactic positions for weak and strong determiners, in accordance with Milsark’s classification, and as proposed by Bowers (1988) within the GB framework. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 4 1.5 The DP-Hypothesis Abney (1987) proposed that Noun Phrases are selected by a Determiner Phrase.^ These functional categories have syntactic properties and semantic functions similar to IP in the clauses. Both select their corresponding lexical categories in their complement position, forming a functional category X’, and both of them define their predicate’s extension. In both cases, X° projects to a maximal projection XP, through the combination of a specifier with X’. According to Abney, two types of elements may head a Determiner Phrase: a determiner, as required by X’-theory, or a genitive case marker (J_s in English). Obviously, this equation of the genitive case marker to the Determiner weakens the X’-theory, since two elements of different nature can head the same maximal projection. I will differentiate between a Genitive Phrase and a Determiner Phrase, although the possessive element is mapped onto the determiner position when it appears in prenominal position. Bowers (1988) gives additional evidence for the DP- hypothesis, based on distributional evidence, specificity effects, and the Degree system of APs. NPs and DPs can be ^ This idea was first proposed by Brame (1981, 1982), See also Szabolsci (1983) and Fukui and Speas (1986) for earlier proposals within the GB framework. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 5 questioned by each other, as Bower’s (1988:47) examples show. (31) a. John bought a what? b. You saw Bill’s what? The second piece of evidence in relation to distributional facts, given by Bowers (1988:48), is based on the structure of English gerundive nominals. The subject receives genitive case, while the gerund and its complements have the form of VP, such as in John’ s winning the prize. If we divide them into two different categories, VP and DP, we would solve the problem, assuming that DP can select VP. Specificity effects are another piece of evidence for the DP-Hypothesis. Bowers proposes that strong determiners occupy a higher projection within the nominal expression and that this projection, DP, induces a Subjacency violation, since the extracted element has to cross two bounding nodes.® Finally, Bowers (1988:53) proposes that there are other categories parallel to DP, such as degree words. We ® I will review Bower’ s analysis of specificity effects more extensively in chapter 4 . Although, I disagree with his analysis of extraction, I follow his idea that there must be a Determiner Phrase. In fact, I propose two determiner positions in chapter 3 . R eproduced witfi permission of ttie copyrigfit owner. Furtfier reproduction profiibited witfiout permission. 2 6 can also distinguish two types of degree words, weak and strong. The first one does not block extraction, as (32a) illustrates, while the second one does, as (32b) shows. Bowers gives the same kind of analysis for (32b) as he gives for the specificity effects of strong determiners. (32) a. Who is John very fond of? b.*Who is John as fond of as Sue? An important change in the understanding of Determiner Phrases is found in Hudson (1989). Here, he proposes that strong Determiners appear in the specifier of their own projection, which was headed by a non-overt referential element, while weak Determiners appear in the specifier of NP. This allows him to account for the specificity effects on extraction from Noun Phrases by Binding Theory and Subjacency. In my dissertation, I will adapt this proposal for Spanish,^ although, contrary to Hudson, I will claim 7 Id terms of an analysis for extraction, I will argue in chapter 4 that it is not necessary to leave the Spec-of-DP position open as a escape-hatch for extraction in order to explain the constraints on extraction from other domains than the internal argument, as has been proposed in Giorgi and Longobardi (1991). According to them, the fact that extraction from within an NP is only allowed when the NP occupies the internal argument position is explained by the verb's ability to govern an intermediate trace in Spec-of-DP only in that position, '/ie can derive the R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 7 that this Is the case in Spanish but not in English. In Spanish the determiner is an operator coindexed with the highest element within the nominal expression. English lacks this operator in the overt syntax and requires the highest element within the nominal expression to move to Spec-of-DP at LF. This allows us to explain why in English only those elements that have the potential to lexically saturate NPs, such as affected objects, can appear in Spec of DP, while not elements that may do it in the syntax, such as unaffected objects, cannot. On the other hand, syntactic operations are done overtly in Spanish. For instance, affected objects are always coindexed with the operator in Spec-of-DP, while unaffected objects may or may not be coindexed with that operator, depending on the fact of whether they referentially saturate the NP or not. I will also propose in chapter 3 that the so-called weak determiners head their own projections, acting as a 1-place predicate operator. Another advantage of this proposal is that it explains the parameters observed among the major Romance same effects assuming that the highest DP does not need to appear in the case of internal arguments, in order for the internal argument to receive case, although i t does in the case of other syntactic functions. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 8 languages with respect to the structure of DP. I claim that there is a Genitive Phrase (GenP), following Picallo (1991), and a Demonstrative Phrase (DemP)and that the q NP moves to the specifier of these projections. In In case the two of them co-occur, DemP selects GenP, such as in (i). I dP l o e m P I c e n P ' ' ' ) In this way we can explain cases such as (ii) and (iii). (ii) Esta casa suya. this house his 'This house of his’ (iii) La casa suya aquella. The house his that 'That house of his.' In (ii) NP moves to Spec-of-GenP, and the demonstrative moves head to head to D*, and from D’ t o the head of Spec-of-DP. In (iii) NP moves to Spec-of-GenP, and, then, GenP moves to Spec-of-DemP. This is what I will call in chapter 3 the "mise en abyme' structures. See chapter 3 for further discussion. In chapter 3 I will also propose the existence of a second DetP that selects NP. The underlying structure of (ii) and (iii) would be as in (iv). I DP (oemP I GenP I DetP ^NP ' "III I ^ This movement assumes the AdjP proposed by Abney (1987). I n an early version of my paper on extraction from NPs (Martin, 1989), I proposed a movement of NP to Specifier of prepositional genitive phrases in order to explain why you cannot extract an internal argument in the presence of a possessor by the Freezing Principle (Wexler and Culicover,1980). At that time an unclear definition of the nominal functional categories, the syntactic behavior of nonrestrictive relative clauses with respect to extraction from NPs, and the feasibility of the Freezing Principle caused me to abandon this idea. In this dissertation, I claim that this movement must R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 9 this way we can explain the parameter between Spanish and Italian with respect to the prenominal possessive constructions. In Spanish, a possessive moves head-to- head from Gen° to D°, and from D° to the head of Spec-of- DP, as in (33) . be generalized to all the categories that can appear postnominally in Spanish, i.e. DemP, GenP, AdjP, and relative clauses, which I argue that are adjuncts to DetP and DP, and therefore coindexed with D“. I will not propose this movement for intensional prenominal adjectives, which 1 claim appear in Spec-of-NP. My motivation is based on agreement facts and on universal adjective ordering facts, and, hence, it is quite different from my first proposal. The interesting thing is that this structure can fuliy explain the extraction facts based on the Subjacency Condition and general conditions on strong agreement. See chapter 4 for further discussion. Movement to Spec-of-AdjP was first proposed by Bernstein (1993) for the case of contrastive adjectives in Romance. Independently, Sanchez (1994b) has proposed this movement t o explain the iicensing of pro in nominal elision structures by Spec-head agreement. See chapter 3 for further discussion. 1 claim that in this case there is not a violation of the Head Movement Constraint (Travis, 1984), since the trace is governed by the head of Spec. I will argue in chapter 3 that the BMC is subject to Kayne's (1984) unambiguous path condition for anaphors, i.e. you cannot move any head from Spec to a governing head position, hut you can move a head from the head position governed by a Specifier to the head of that Specifier. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 0 (33) DP / \ mi; D’ ■ ’ / \ t; GenP / \ casa- Gen’ ^ / \ ti tj On the other hand, in Italian the possessive element moves from Gen“ to D°, but does not reach Spec-of-DP, as in (34). (34) DP / \ la D’ / \ mia- GenP / \ casa; Gen’ ^ / \ ti tj Further parameters should be expected. For instance, in Toscan NP does not reach Spec-of-GenP. Instead, the possessive moves from Gen° to that position, as in (35). (35) DP / \ la; D’ / \ e; GenP / \ mi; Gen’ ^ / \ tj NP casa R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 31 Therefore, I assume that the word marker in Romance languages indicates that the word occupies a head in the main path of the nominal expression,and not a Spec- position. In this way, we can explain that it appears in the prenominal possessive element in Italian, but not in Spanish or Toscan. Once it is lost, the foot slot for the gender marker may disappear, being subject to resyllabification conditions, along the lines of the model of prosodic morphology proposed by McCarthy and Prince (1990 ) . Further evidence for this proposal is given by intensional prenominal adjectives. Intensional adjectives do not have the word marker either, as buen(*o) hombre illustrates. In chapter 3 I will argue that they are mapped onto Spec-of-NP, in order to explain word order facts and their semantic effects, that is, that they always appear in prenominal positions and that they are semantic operators of the concept denoted by the noun, but without referential import. Hence they must be I an using Harris' (1991) tcra "word marker", although, if my analysis is correct, i t would he more adequate to call it "phrasal marker." R eproduced witti permission of ttie copyrigfit owner. Furtfier reproduction profiibited witfiout permission. 3 2 projected below the lowest DetP, proposed by me in chapter 3.^^ Finally, articles are expletive elements without lexical content. They appear at PF if there is no other element occupying the Spec-of-DP position. This hypothesis entails that there is not a directionality parameter between English and Spanish, such as Head Subject Parameter (Giorgi and Longobardi, 1991), but a morphological parameter. 12 In order t o analyze the syntactic and semantic differences between strong and weak determiners (see Milsark (1974), Bowers (1988), Hudson (1989), and Vergnaud and Zubizarreta (1990, 1992) for further discussion) I propose in (i ) the underlying syntactic structure of nominal expressions in ( i). ( i) QP I \ DP / \ strong det Extensional Elements / \ DetP I \ weak det NP / \ Intensional N ' Elements This proposal is relevant to my analysis of affectedness in chapter 2 , since complex event nouns lack DP, but simple event nouns and result nouns do not. For another analysis of DP, based on its double articulation, see Cornilescu’ s (1992) analysis for Romanian. R eproduced witti permission of ttie copyrigtit owner. Furttier reproduction profiibited wittiout permission. 3 3 1.6. The possessor variable As pointed out by Zubizarreta (1987:73), the Possessor-variable is borne by N’ and covers the notion of possessor or owner and the notion of agent or creator in the case of Result Nominals. In accordance with Zubizarreta (1987), Process Nominals lack the Possessor variable, and, instead, they have an E(vent)-variable that may be identified by one or more adjuncts through the Rule of Modification, as shown by the fact that we may have the adverbial expression in (36) more than once. (36) John’s discussion of the book yesterday during the dinner. (Zubizarreta, 1987:56) Therefore, the Poss-variable is identified through Referential Predication in Result Nominals, and the E- variable is identified through the Rule of Modification in Process Nominals. This entails that Process Nominals lack the highest Functional Category that selects a nominal expression, as I will argue in chapter 2 based on double articulation of DP proposed in chapter 3. As pointed out by Zubizarreta (1987), the Biuniqueness Condition, that applies to arg-variables lexically determined, does not apply to the E-variable, since time adverbials do not belong to the Semantic Representation of a lexical item. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 4 The relationship of a genitive head and a complex- event nominal head must always be lexical and never extrinsical, in Godard’s (1992) terms. I will argue that the difference between complex-event nominals and their corresponding simple event nominals and result nominals can be defined in terms of predication. I.e., an extrinsical relation, such as predication, allows simple events and result nominals to project two genitive phrases: one licensed by a lexical variable of the head; the second one as a result of an extrinsic possessive relation, i.e. a modifier evaluates a possessive variable. In (37) descripciôn ’description’ can be interpreted as a result nominal, in (37a), and as a simple event nominal, in (37b). (37) a. Su descripciôn de Barcelona esta sobre His description of Barcelona is on the la mesa, table. b. Su descripciôn de Barcelona sera en el salôn His description of Barcelona will be in the de actos. conference room. In both cases de Barcelona evaluates the variable of the internal argument of the noun, while the prenominal possessive is the result of an extrinsic relation, that has been formalized by Godard (1992:258) in (38). R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 5 (38) Extrinsic relation of possession If the head N of the dominating NP is saturated, the subject is interpreted as an argument of the relation POSSESSION. By saturation of N“ we must understand that DP can be bound by a referential operator, and not only by an existential operator. Complex-event nominals are not by definition allowed to hold these extrinsic relations, therefore only one genitive phrase may appear, in fact "must" appear, in their syntactic structure, and a complex event nominal can be introduced only by an existential determiner. 1.7. The model of the grammar and the Full Interpretation Principle I would like to propose the following model of Grammar, based primarily on Chomsky (1981), van Riemsdijk & Williams (1981), and Zubizarreta (1987): (39) (A) Lexicon (i) Semantic Representation (ii) Lexical Representation { (A-binding operations) (B) Syntax: (i) NP-structure - > Semantic Interpretation (SI) I (A’-binding operations) (ii) S-structure --> SI (C) Interpretative components: (i) Phonological Form (i) Logical Form R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 6 The Chomskyan model contains a set of rules and principles that can be classified in different subtheories, such as bounding theory, X’-theory, goverment theory, ©-theory, binding theory, case theory, control theory. In the next sections I examine the most relevant to my dissertation, X’-theory and ©-theory. The lexicon consists of a list of words, specified by certain syntactic, semantic, and phonological properties. Lexical entries are specified by the information necessary to shape ©-theory. This information is crucial in order to understand the way in which the semantics-syntax interface occurs, in accordance with the Universal Alignment Hypothesis (Perlmutter and Postal, 1 0 1984). I will propose that this theory is based on the referential and aspectual properties of the head. The conceptual structure will predict role assignment and also the case assignment possibilities, while it is at the NP-structure level where the case parameters and binding relations are set up. The projection principle in (40) ensures that lexical information determines syntactic structure. (40) Projection Principle Lexical information is syntactically represented. ] • ) See sections 2.3 and 2,6 for further discussion. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 7 The Theta-Criterion guarantees that this lexical information is not altered in the course of the derivation, as has been stated in (41) (cf. Chomsky, 1982) . (41) Theta-Criterion Every term of LF that requires a theta role is associated with one and only one position to which theta roles are assigned, and each theta role determined by the lexical properties of a head is uniquely associated with one and only one argument. Case Theory must make these thematic relations visible for 6-role assignment. At LF the interpretation of predication relationships, Q-referring expressions and R-expressions that undergo R-predication takes place, in a model similar to the one proposed by Diessing (1992). The model of grammar, with its different principles and rules, must ensure the well-formedness of sentences. This entails that at LF every maximal projection must be licensed, in accordance with the Full Interpretation Principle. 1.8. X’-Theory I will adopt the basic configuration for X-bar theory in (42), although I suggest a variation is needed in the X-bar scheme in order to incorporate the adjunct R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 8 modifier of the head, projected to the right, into the X’ system within a larsonian syntactic model (see fn. 14). (42) XP / \ Spec X’ / \ X- (YP) Every X" must project to XP, and every XP must be headed. XPs can be terminal, and in this case they do not project a complement. Functional categories are never terminal, and, among lexical categories, we should distinguish between terminal projections, VPs, NPs, and some AdjPs, and non terminal projections, PPs and restrictive noncontrastive adjectives. Specifiers of lexical projections are A-positions, and agreement with the head is required. Specifiers of pure functional categories, i.e. DPs in the case of nominal expressions, are A ’ - positions. Coindexation with the head is required for elements occupying A’-positions. Coindexation is based on the notion of partitiveness, hence, on the identificational properties of the Specifier with respect to the head. I constrain the number of adjuncts to one per maximal projection at the input of the syntactic component. This adjunct functions as a modifier of the R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 9 head of the maximal projection, as in (43), where fiel ’loyal’ has a predicative and contrastive function. (43) DP / \ el ‘AdjP / \ AdjP a la causa / \ soldado Adj’ I fiel In a similar fashion to (43), a head may be empty but modified by an adjunct, within a nominal expression. The empty head is coindexed with the adjunct. by modification.^^ I use superscripted numbers to express this relation in (44). (44) DP / \ el "AdjP / \ , AdjP de lingiiistica / \ estudio; Ad j ’ Hence, a more appropriate X'-model is ( i ) . (i ) XP / \ •X’ I \ 1 X’ (ZP)^ /,\ X'^ (YP) R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4 0 AdjP is essentially a predicative category. I adopt Abney’s AdjP-Hypothesis for noncontrastive postnominal adjectives. According to this hypothesis, adjectives head their own projection, and select NP in their complement position. The difference between the major Romance languages and English is that NP overtly moves to Spec- of-AdjP in Romance. Notice the difference between (44) and its synonymous (45). (45) DP / \ el AdjP / \ estudiO: Ad j ’ / \ lingiiistico tj 1.9. Theta theory The syntactic import of 6-role labels as the ones in (46) has been challenged by several authors, for example Rappaport and Levin (1988), Grimshaw (1990), Zubizarreta (1987), and Tenny (1987, 1988). (46) Fillmore (1968) GIVE; [ objective case + dative case + agentive case] Stowell (1981) PUT: <agent, theme, location> Levin and Rappaport (1986) PUT: agent <theme. location> R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4 1 Higginbotham (1985), for instance has proposed that we do not need to refer to these 0-role labels, as we can observe in (47). (47) Higginbotham (1985) SEE: <1, 2, E> It is necessary to define the nature of the cognitive information visible to the syntax. The internal argument receives the 0-role from a lexical head. External arguments receive a 0-role compositionally from the projection of that head. Therefore, the internal argument intervenes in the theta-marking of other arguments. The internal argument is the closest argument to the head, to the degree that they operate together in the projection of other arguments. I will claim that there is a referentiality scale that defines the argument structure. Those arguments that are able to lexically saturate a predicate, hence the external argument, will occupy the internal position. Affected objects in nominalizations and middle constructions are a clear example of this proposal. They are able to lexically saturate their predicates, as we will see in Chapter 2. The referential scale will give us the canonical positions of the arguments. Other lexical and syntactic processes, such as R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4 2 passivization or focalization, act on this canonical structure. Therefore, the importance of ©-role labels is that they are mapped onto the universal hierarchy, and that once this is done, their position in that scale is what is grammatically relevant. In accordance with this, I will assume a layered structure for VPs and NPs, similar to the one proposed by Larson (1988) and Bowers (1993) for verbal structure and by Valois (1991) for nominal structure. The fact that agents cannot be associated with internal argumenthood either in complex-event nominal or verbal expressions will be derived from the fact that they are not associated with the sub-event STATE, along the lines of Grimshaw (1990). The syntax of predication acts over this hierarchy by mapping it onto propositional functions. Is this type of hierarchy a unique phenomenom in the grammar? It has also been suggested that some kind of semantic principle is active in the serialization of adjectives. Hetzron (1978), among others, has pointed out the universal character of adjective ordering. The semantic analyses of this phenomenom are still vague in some way, such as Hetzron’s which is based on a gradual scale defined by the objectivity of the adjective. Hetzron (1978:178) defines this principle as follows: R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4 3 "the major rule is to place the more objective and undisputable qualifations closer to the noun, and the more subjective, opinionlike ones further away". Seiler (1978) points out another more precise analysis of adjective ordering based on the referential properties of the adjective with respect to the noun. Seiler (1978:308-309) proposes the complementary generalizations in (48). (48) a. The range of head nouns for which a determiner D is potentially applicable increases with the positional distance of that determiner from the head noun. b. The potential of a determiner D for singling out the object referred to by the head noun N increases proportionally with the positional distance of D from N. Therefore, if an adjective in that scale "singles out" the noun, it would saturate the positions above it. For instance imagine that we want to single out Pedro from a group of men, where there is only one Spanish man and several tall men, Pedro having both attributes. We have to use tall in a nonrestrictive way, such as in (49). (49) Pedro es el hombre espanol, alto. Pedro is the man Spanish tall ’Pedro is the tall Spanish man.’ Now, imagine the reverse situation where there are several Spaniards but only one of them is tall. In this R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4 4 case, we can use both adjectives as restrictive, and no comma is necessary between the two adjectives. A way to avoid this rule in the first case would be to place espanol in a contrastive position. In this way both adjectives can be restrictive, given the first situation. This contrastive function in Spanish is characterized by being the last one of the postnominal adjectival string, and by receiving contrastive stress, as in (50). (50) Pedro es el hombre alto ESPAROL. To ask in what way this analysis relates to argument structure would be a fair question. The relation appears to be very vague. However, we must recall that arguments can be expressed by adjectives within nominal expressions. The ordering of adjective arguments is noun object-subject under normal intonation, as (51) shows. (51) a. el estudio lingiiistico aleman the study linguistic German 'the German linguistic study’ b.??el estudio aleman lingiiistico Again, (51b) would be fine if lingiiistico were to receive contrastive stress. This is just what my hypothesis predicts. We can conclude that syntactic dominance relations are defined by referential properties. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4 5 1.10. Typology of Nouns I will adopt Grimshaw’s (1990) typology of nouns given its syntactic relevance. She distinguishes those nominals that take arguments, "complex event nouns", from those that do not. She points out that among the latter some may take complements, elements that belong to the conceptual structure of the noun but have not become arguments given the nature of the noun, and modifiers, while others only take modifiers. The first ones are result and simple event nouns and the second ones are individual concrete nouns. The main difference between argument-taking nominals and other nominals is that the first ones project a hidden eventive argument variable. Other nominal expressions are predicated of a referential variable. There are four main tests to differentiate complex event nouns from the rest. The first one is that only complex event nominals allow ’event control’, such as in (52) . (52) La destrucciôn de la ciudad para hacer que el ’TJie destruction of the city in order to make enemigo se rinda. the enemy to surrender.’ The second one is that they admit aspectual modifiers such as frequent and constant, as in (53). R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4 6 (53) La frecuente descripciôn de la ciudad. The frequent description of the city. Agent-oriented adjectives impose a complex event reading of a noun as in (54). (54) La deliberada destrucciôn *(de la ciudad) tomô The deliberate destruction of the city took mucho tiempo. a long time. Finally, complex event nominals admit the same kind of aspectual modifiers as their related verbs, as (55) demonstrates. (55) a. La destrucciôn de la ciudad en dos boras. The destruction of the city in two hours. b. La ciudad fue destruida en dos horas. The city was destroyed in two hours. Other characteristics of complex event nominals are that they always project the internal argument, as (56) illustrates. (56) La frecuente descripciôn *(de la ciudad). The frequent description of the city. Grimshaw also states that complex event nominals do not pluralize and they do not take other determiners except definite determiners. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4 7 Finally, I would like to comment on the very interesting observation, made by Picallo (1991:309), that nouns related to unergative verbs do not have a complex event reading in Catalan. This observation can be extended to Spanish. They always denote results, and do not have nominalization affixes. Their argument always receives genitive case, and never appears introduced by the preposition por 'by', as the Spanish counterparts to Picallo's examples in the first place show in (57), (58), and (59). (57) a. CAT: el salt de I'atleta SPA: el salto del atleta the jump of the athlete b. CAT:*el salt per part de I'atleta SPA:*el salto por el atleta the jump on the part of the athlete (58) a. CAT: la lluita del gladiador SPA: la lucha del gladiador the fight of the gladiator b. CAT:*la lluita per part del gladiador SPA:*la lucha por el gladiador the fight on the part of the gladiatoi (59) a. CAT: el xiscle del nen SPA: el grito del nino the scream of the child b. CAT:*el xiscle per part del nen SPA:*el grito por el nino the scream on the part of the child The other characteristic of unergative nouns is their inability to have a nominalizer affix, such as -cion. - R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4 8 miento. etc. These facts might reflect several aspects of the syntax of complex event nominals. First, the presence of event structure, i.e. an eventive variable, triggers nominal affixation. This variable is related to only one of the two possible subevents of an event, the second one or STATE. The way to saturate this variable is by projecting the corresponding argument or arguments related to this subevent, i.e. the internal arguments. The state subevent licenses the aspectual features of the activity subevent, but not a syntactic position linked to these features, at least, not a position visible to the syntactic structure of a noun, i.e. a position that receives case. The only way to discharge this argument is through a minor predicate, i.e. a preposition with semantic content. In conclusion, this fact could indicate that subevents also have a syntactic representation and that not all the potential subevents have to be syntactically projected. 1.11. Organization of the dissertation In chapter 2 of the dissertation I study the lexical structure of NPs. The study will be concentrated on the definition of the lexical properties of affected objects. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4 9 I will argue that affected internal arguments are lexical saturators of one-place predicates, such as nominals. That is, affected objects semantically saturate a nominal head, and, therefore, this nominal head cannot undergo further restrictive modification. The presence of overt operators in Romance languages, determiners and prenominal possessives, allows us to do this operation at the syntactic level of representation. The lack of overt operators in English forces us to do this operation at the lexical level of representation or at LF, by raising the affected object either to the specifier position of QP, if it is a complex event noun, or to the specifier position of DP, if it is a result nominal. In the case of the lexical saturation of a complex event nominal, the internal argument appears at the Spec-of-GenP position, i.e. the prenominal position, by the output of the lexical level, and the agent is lexically saturated, i.e. the noun lacking a complex event interpretation. This results from the fact that in English the noun forms a closer syntactic and semantic unit with the prenominal possessive than with the rest of complements and modifiers. However, when the affected object appears in the postnominal position, of-insertion occurs in order to assign case to the internal argument, and the syntactic presence of the external argument is allowed as PRO. At R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5 0 LF the affected complement saturates the complex "PRO + noun." Complex event nominals are selected by QP and DetP, and result nominals by (QP), DP, and DetP. Finally, the properties of the preposition ^ are studied, and the semantic properties of genitive case assignment explored. With regards to the first issue, Spanish offers crosslinguistic evidence for Grimshaw’s claim that of- insertion is a ©-marking mechanism, in addition to a case-marking mechanism. With respect to the semantic properties of genitive case assignment and their relation to the determiner system, I argue that genitive case assignment can only be assigned to those nominal complements that hold a potentially referential partitive relation with respect to the head noun. In chapter 3 I propose a double articulation of the DP system in order to give an account for distributional facts, word order facts, coordination facts, and the syntactic and semantic effects of weak and strong determiners. I apply Abney’s AdjP-Hypothesis to Romance by generalizing movement to Spec-of-AdjP to DeraP, GenP, and all kinds of AdjPs, except for prenominal intensional adjectives, which I argue appear mapped onto Spec-of-NP. This proposal explains the agreement facts within nominal expressions in Romance based on the Spec-head hypothesis. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5 1 and the adjective ordering differences between English and Romance. In chapter 4, I explore the extraction facts from a nominal expression. I establish that only syntactic arguments, that is, possessive elements and sentential complements, can be extracted from a nominal expression. Otherwise, extractable elements from a nominal expression are subject to general principles of the grammar, namely, the Subjacency Condition and general conditions on strong Spec-head agreement. I claim that strong agreement is a way of ©-marking through the assignment of an index to the element in Spec. In this way, we can explain why non restrictive and D-linked questions that are formed out of a specific NP or a preverbal subject are possible in Spanish. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5 2 Chapter 2 Argument Structure of Noun Phrases: Affectedness and possessives 2.1. Introduction The general objective of this chapter is to define which principles of grammar determine the mapping of the Argument Structure of a Nominal Phrase onto its syntactic configuration. In particular, I attempt to discover which semantic properties are necessary for an argument to be mapped onto the possessive function. I claim that this mapping is properly defined by referential-aspectual properties, following Tenny’s (1987) analysis of the syntactic organization of VPs. I propose that the argument variable licensed by a complex event nominal and identified by genitive case is always an argument related to the state, or change of state, sub-event, within a phasic theory of aspect, where an event is defined by subperiods (activity and state), as proposed by Pustejovsky (1988) and Grimshaw (1990). Following an analysis based on referential-aspectual properties, I will characterize the affectedness constraint as the one shown by lexical items which have R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5 3 a saturated event variable. This saturation has a compositional nature, since it is determined by the saturation of the internal argument variable and the saturation of a delimitedness variable, in the sense of Tenny (1987) . Therefore, a nominal head which may project an affected object onto its syntactic structure would be the one that might be referentially saturated through a referential internal argument and delimitedness. This follows from the fact that nominalizations only denote properties in the sense of Chierchia (1985) and Bowers (1993). In this way the syntactic structures where there is obliteration of the external argument^^ can be explained in a more natural way. Finally, I look at the relationships between complex events and simple event nominals, in the sense of Grimshaw (1990), and at the relationships between simple event nominals and result nominals. I claim that simple events can be derived from complex event nominals by a predication process, and that result nominals are derived by lexical saturation of the event variable. The consequence of this hypothesis is that complex event nouns are selected by QP and DetP, and simple event nouns I understand obliteration of the external argument as the syntactic absence of the external argument. R eproduced witti permission of ttie copyrigfit owner. Furtfier reproduction profiibited witfiout permission. 5 4 and result nouns by QP, optionally, DP, and DetP.^^ In this way the similarities between these structures, as the presence of the same semantic type of possessive complement, are captured in a more explanatory way. Certain asymmetries that may be observed in the derivation of simple events from some complex events are explained by the affectedness constraint, as defined above. 2.2. The Lexicalist Hypothesis and the Syntactic Structure of Noun Phrases One of the most important issues raised by the projection of arguments by a lexical head is to determine which syntactic representations are related to the Argument-structure of a predicator, and to define which principles are involved in the association of an NP with a variable in the Lexical Conceptual Structure (LCS). It is necessary to define which cognitive principles act in the determination of the argument structure of a predicate. The argument structure defines the association of an NP with a variable of the predicate and the I n chapter 3 , I will propose a double articulation of the determiner system. The higher determiner, DP, is defined by a referential reading of the noun, while the lower DetP, by the existential reading of the noun. R eproduced witfi permission of ttie copyrigfit owner. Furtfier reproduction profiibited witfiout permission. 5 5 hierarchical position of that argument. The syntactic mapping is determined by this argument structure in the case of nominal expressions, and by this argument structure and the extended projection principle in the case of verbal expressions. We should not expect all the nominal and verbal syntactic mappings to be identical. Psych-predicates constitute a good example, as we will see below in examples (5) and (6). Therefore, it is crucial to determine whether a lexical item is marked in the Lexicon as a lexical category or whether its lexical category status is acquired in the course of its grammatical derivation. The Lexicalist Hypothesis (Chomsky, 1970) put forward the need for the existence of some redundancy rules in the Lexicon that can capture generalizations among sets of words with the same lexeme, but belong to different lexical categories. The Transformationalist Hypothesis , developed by Lees (1960), is rejected by Chomsky (1970) due to the fact that the relation between nouns and verbs is non productive. There is not a straightforward syntactic relation between verbs and nouns as regards all the different verb types. (1) Basically, the subcategorization features and the selection properties would go under a neutral lexical entry, unspecified for the lexical category, in addition to other redundancy rules that express regularities between different lexical entries, as nominal suffixation. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5 6 shows the case where there is a straightforward relationship, (1) a. The barbarians destroyed Rome. b. The barbarians’ destruction of Rome. c. Rome’s destruction by the barbarians. On the other hand, in the case of frighten psych-verb class this syntactic relation seems more difficult to establish, as shown in (2).^® (2) a. The ghost frightens John. b.*The ghost’s fright of John. c. John’s fright at the ghost. 1R Probably the Transformationalist Hypothesis would have more possibilities to be successful if we derive verbs from nouns instead of the other way around. Some of the conclusions drawn in this study could suggest that this is the right direction. Also, as suggested to me by Saltarelli (p.c.), the relation between some concrete nouns and their verbal counterparts could indicate, at least, that the relationship between nouns and verbs i s bidirectional. An example of a verb derived from a noun in Italian would be casare 'to house' from casa 'house'. Nevertheless, this possible derivation, as we will see in this chapter, would have a lexical nature and not a syntactic one, and in economy terms it would be more costly. Therefore I will assume a strong version of the Lexicalist Hypothesis, where related verbs and nouns are derived from the same lexical item underspecified for a syntactic category. R eproduced witti permission of ttie copyrigtit owner. Furttier reproduction profiibited wittiout permission. 5 7 As we can observe in (2), there is not a straight correlation between the syntactic forms of (1) and (2). The argument that acts as the verbal subject in (2a), cannot be mapped onto that position in the related nominal expression as (2b) shows. On the other hand, the argument that is assigned accusative case in (2a) can be mapped onto the subject position of the Noun Phrase, once the appropriate preposition is selected for the other argument, as (2c) shows. The love psych-verb class shows also other asymmetries between verbal and the nominal mappings. As the paradigm in (3) illustrates, the so-called passive form in (3c) is not allowed, in contrast with the correct passive construction of destruction in (Ic). (3) a. Mary loves God. b. Mary's love of/for God c.*God's love by Mary. The paradigms in (2) and (3) are problematic for the Transformationalist Hypothesis since they show important mismatches between the syntactic mappings of argument structure in NPs and VPs. For instance, in order to bar (2b), we would need a powerful and ad hoc mechanism to be introduced into the grammar of natural languages. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 5 8 There are two other drawbacks for the Transformationalist Hypothesis. First, the unpredictability of the relation of the meaning of the derived noun to the one of the corresponding verb, because it would force us to mark this rule lexically. Second, event nominals are much more alike, in syntactic terms, to concrete nominals than to verbs. Rappaport (1983) and Rozwadowska (1988) claim that a verb and a noun with the same stem share their argument structure. Rappaport (1983) points out that the fact that nouns and verbs share the same a-structure does not imply an isomorphism at the syntactic level. In this regard, she proposes that nominals do not use the grammatical functions subject and object, but only the possessive and oblique functions. The grammatical functions Subject and Object are semantically unrestricted, this fact entails that different semantic arguments may be mapped onto the grammatical functions Subject and Object. On the other hand, the grammatical functions oblique object and possessive are semantically restricted. In particular. Again this position assumes that the only possible derivation is from verb t o noun and not the other vay around. As I said before this is arguable. 20 Consider, for instance, descripciôn. 'description', and informacion, 'information'. Informacion can be only a concrete nominal, while descripciôn can be both of them. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5 9 Rappaport posits that the possessive grammatical function is restricted to themes. However, in addition to the difficulty for a precise definition of the term "theme", the arguments to which the possessive function is restricted do not form a semantically natural class, as pointed out by Zubizarreta (1987:46), or, at least, the semantic theory related to Argument Structure needs some refinements. Despite all these problems, the insight behind Rappaport’s proposals is essentially correct, as indicated by Zubizarreta (1987:46). The syntactic structure of NPs seems to be semantically more transparent than the one of sentences. In accordance with Zubizarreta (1987:47) the distortions that occur in the case of the syntactic mapping of the argument structure of verbs occur at the lexico-syntactic level of representation. Nouns lack this level of representation, hence, their arguments are directly mapped onto the syntax from the lexico-semantic level of representation as indirect arguments, i.e. prepositions introduce them. In the case of verbs, internal arguments are mapped onto the syntax by the Rule of Projection and external arguments by the Rule of Predication. Zubizarreta (1987:39-45) points out five cases where the syntactic structure of arguments is different, when comparing verbs and nominals. The first case that she R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6 0 indicates is that of nonstative experiencers in nominals, which was mentioned above. The basic contrast is shown in (4c) and (4d) where the "experienced" cannot occupy the subject position, i.e. the prenominal position, on the syntactic structure of nominals but it can do so on the syntactic structure of verbs (examples from Zubizarreta, 1987: p.40). (4) a. The boredom of the children with the stories. b. The children’s boredom with the stories. c.*The stories’ boredom of the children. d. The stories bore the children. Spanish reflects these distributional facts, since the object of experience cannot occupy the prenominal position, as shown by (5). (5) a. El aburrimiento de los ninos con los cuentos. b. Su (=los ninos) aburrimiento con los cuentos. c.*Su (=los cuentos) aburrimiento de los ninos. d. Los cuentos aburren a los ninos. Furthermore, experiencers are mapped always onto the possessive function of nominals in Spanish, regardless of the syntactic type of psych-verb from which they are projected, while the possessive is barred for other arguments in the presence of the experiencer. (6) R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6 1 illustrates the syntactic structure of Belletti and Rizzi’s (1988) first class of psych-words. (6) a. El amor de Maria por la lingiiistica. The love of Mary for linguistics. b.*El amor de Maria de la lingiiistica. Mary's love of linguistics. This case will be explained in the last section of this chapter. I will argue that objects of experience do not have the potential to referentially saturate the noun, and therefore cannot be assigned the possessive function. However, what is relevant is that the experiencers projected by nouns are consistently marked with the same case in contrast with experiencers projected by verbs, as we can observe in (5a) and (6a) where the experiencer projected by the two main classes of psych-nouns is assigned the possessive function. I will argue that an analysis of the mapping of argument structure onto syntax strictly based on aspectual properties would account for this phenomenon!. Also, in section 2.6.3, I will study the structure of psych-nouns and their implications for the syntactic structure of the different classes of psych-verbs. In particular, I will study the syntactic structure of psych nouns in relation to their lexical structure and whether Rappaport's (1983) hypothesis, whereby prepositions semantically restrict R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6 2 nominal arguments, can shed any light on the syntactic structure of psych-verbs. For Rappaport (1983) and Rozwadowska (1988) verbs and nouns share the same a-structure at a lexical semantic level of representation. The second case, pointed out by Zubizarreta (1987), where there is an asymmetry between the syntactic structure of nominals and verbs is when a "locative" or "goal" argument is projected by certain lexical entries. While some specific verbs may mark these arguments with the accusative case, the nominal heads related to these verbs cannot mark these arguments with the genitive case as direct syntactic arguments, being always introduced by a locative preposition or by the preposition to. This is illustrated by (7) (examples from Zubizarreta, 1987:41). (7) a. The soldiers fled from the city. b. The soldiers fled the city. c. John’s flight from the city. d. John’s flight of the city. As indicated by Zubizarreta (1987), languages vary with respect to whether a verbal head can or cannot mark a locative argument or goal argument as a direct dependent on the head. English would be an example of a language that can mark these kinds of arguments as direct R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6 3 objects. This is probably due to the "dative shift" mechanism of English, that allows a dative object to receive accusative case. On the other hand, Spanish would not allow this kind of structure. According to Zubizarreta (1987), French would represent an intermediate case. However, as pointed out by Zubizarreta (1987), what is relevant, in terms of Lexical Structure and Universal Grammar, is the fact that there is no language variation with respect to whether a locative argument or a goal argument can or cannot be mapped onto the possessive function, since they are never mapped onto the possessive function. Again, the syntactic functions projected by a nominal head seem to be more restricted than those projected by a verbal head. In addition, verbal heads seem to offer a wider range of possible syntactic positions for an argument to be mapped onto them. While nouns have only the possessive function as a semantically unrestricted function, verbs have the subject and the object positions as semantically unrestricted, i.e. more than one type of the traditional 6- roles can be mapped onto those functions. The third case, pointed out by Zubizarreta (1987), is the asymmetries observed in the mapping onto the determiner position of an internal argument of an NP in the so-called passive construction of NPs, assuming that R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6 4 the prenominal possessive in English occupies the determiner position (Abney, 1987).^^ Anderson (1979) derived these facts from the affectedness constraint. Anderson observes that the syntactic behavior of an internal object in this kind of construction is related to whether or not the action denoted by the head affects the internal object. An affected object can occupy the prenominal position, as shown in (8). (8) The city’s destruction by the Romans. An unaffected object cannot occupy this position, as (9) illustrates. (9)*The book’s discussion by John. This constraint does not arise in verbal passive constructions, as (10) illustrates. (10) The book was discussed by John. However, as pointed out by Jaeggli (1986), verbal middle constructions do show the same kind of distributional 21 Following Zubizarreta (1987), I will assume that the so-called passive construction does not involve any kind of syntactic raising from object to subject position, and that constituents that appear in that position are mapped onto the Spec-of-DP by the input of the syntactic level of representation. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6 5 facts with respect to this constraint. Therefore, it is possible that these structures are more closely related to the middle voice than to the passive voice. I will argue that the affectedness constraint must be defined in terras of lexical saturation, following Tenny’s (1987) analysis of affectedness based on aspectual properties. In a fashion similar to the affectedness constraint in NPs, the English middle voice could be explained by proposing that a generic affected object is able to saturate a verb. A generic variable is able to neutralize the referential properties of the external argument variable. The fourth case is given by the impossibility of finding prenominal subjects raised from an embedded clause, as (11) illustrates (examples from Zubizarreta 1987:44) . (11) a. The conclusion proved to be wrong. b.*The conclusion’s proof to be wrong. Zubizarreta (1987) proposed that the affectedness constraint accounts for this case, as well, since conclusion is not an object of the nominal head, it cannot be an "affected object". Another account can be based on the fact that nouns do not govern their R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 66 complement position, thus, the intermediate trace on 99 Spec-of-CP is not governed. The fifth case is the obligatory nature of the projection of external arguments in VPs but not in NPs. In fact, in a language such as Spanish, external arguments are never projected in the prominent position of a complex-event NP, as (12) illustrates. Their status is very similar to the por-phrase in the passive voice, its syntactic projection also being optional. (12)*Su destrucciôn de Roma. Their destruction of Rome. Although we have enumerated a number of differences between NPs and VPs, some of which, no matter which way they are analyzed, clearly remain different, there are also some similarities between them, as, for example, the one pointed by Rappaport (1983) and Rozwadowska (1988), i.e. verbs and their corresponding nouns share the same A-structure. If this is true, and A-structure is not only just a list of thematic roles but a hierarchical one, then we should explain the asymmetries of syntactic 99 In Chapter 3 the fact that nouns cannot govern their complement position is derived from the fact that nouns do not comply with the configurational requirements to do so. In this chapter, 1 will argue that nouns move to the specifier of the higher projection instead of moving to the head as verbs do. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6 7 mappings as the ones showed by (5b) and (5d), which are different in terms of prominence. As a result of this, a Lexical Theory may evolve which is composed of different modules, as the one proposed by Zubizarreta (1987) and Levin and Rappaport (1988), to explain the lexical differences between the different syntactic categories. Before studying the lexical mechanisms that give rise to the different mapping of argument structure onto syntactic structure across categories, I would like to define Theta theory and Case theory in relation to nominals. 2.3. The 6-marking of Nominal Arguments and Case Theory 2.3.1. Theta marking and inherent case assignment Noun heads have been traditionally considered defective structural case-markers (Kayne, 1983; Chomsky, 1986a; Rizzi, 1990). Case assignment within an NP is realized through gf-insertion. Of-insertion is an inherent type of case assignment since it is sensitive to thematic relations, as shown by (13) (example addapted from Chomsky 1986a: 190). (13)*The belief (of) John to be the winner. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 68 (13) would be ruled out by the Case filter, given the Uniformity Condition in (14).^^ (14) If A is an inherent case assigner, then A assigns case to an NP if and only if A theta- marks the NP. Some authors, as Emonds (1985) and Grimshaw (1990), have characterized prepositional insertion within NPs as the nominal heads defective faculty to theta-mark their arguments directly, as we will see in the next section. 2.3.2. Grimshaw’s (1990) analysis 2.3.2.1. Theta-marking of CPs Grimshaw (1990) states that theta-roles are assigned by Nouns through prepositions. Argument structures of prepositions that theta-mark nominal arguments are identified with the Argument Structure of the Nominal head, its argument structure being satisfied in this way. Grimshaw based her proposal on two facts. The first one is the absence of sentential arguments within NPs. According to the standard theory, CP arguments do not need to be assigned case. However, they 2 3 At this point, I will leave open the question of whether a nominal head governs its complement position or not. See Rizzi (1990: fn. 130-131) for further discussion. Also, see f n . 22 of this Chapter. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6 9 are not able to satisfy the Argument structure of a nominal head. Therefore, a nominal head that selects a CP complement without a preposition would not be a complex event nominal, as (15) illustrates (Grimshaw, 1990: 77). (15)*The constant announcement that results have been achieved should not be condoned. Since verbal counterparts to this kind of nouns also select a CP, Grimshaw proposes that CPs within NPs are complements, i.e. they are related to the Lexical Conceptual Structure of the noun but not to its Argument Structure. Spanish seems to corroborate Grimshaw’s claim, since a preposition is allowed in the complementizer position and the sentence is grammatical, as (16) shows. (16) El constante anuncio *(de) que se han conseguido los resultados esperados no se debe condonar. English complementizers that can be introduced by prepositions also seem to support Grimshaw’s claim. In Spanish complementizer gue does not allow for an appositive reading either, as (i ) shows. (i)*El anuncio, que se han conseguido los resultados esperados, no se dehe condonar. 9C The behavior of that disallowing preposition insertion should he included in the properties of this complementizer as another of its peculiarities. One might posit that the syntactic position of tW is one projection higher than that of whether, preventing of-insertion R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 7 0 (17a) and (17b), the insertion of the preposition of is necessary for the sentence to be grammatical. (17c) shows that the preposition of is optional when the noun does not have complex event reading. (17) a. The question PROj of whether the data were relevant or not in order PRO; to help the student was clever. b. The frequent question *(of) whether this happens or not is not relevant. c. The question (of) whether this happens or not is not relevant. In conclusion, Spanish seems to support Grimshaw’s claim that Nouns do not theta-mark directly their arguments with additional crosslinguistic evidence, since Spanish prepositions are able to theta-mark a CP in the complement position of an NP. 2.3.2.2. Passive nominals and the affectedness constraint The second piece of evidence presented by Grimshaw (1990) is the behavior of the so-called passive nominals. Passive nominals seem not to take arguments since their behavior parallels that of simple event nominals and result nominals with regard to aspectual properties, as illustrated by (18) (Examples from Grimshaw, 1990:83-84). and antecedent government of a trace within the clause that i s introduced by tW at S-structure. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 71 (18) a.*The politician’s frequent/constant nomination. b.*The building’s construction in three weeks. c.*The book’s translation (in order) to make it available to a wider readership. However, I claim that the unavailability of the complex event reading is related to a different aspect of the syntax of nominal phrases and not to their defective theta-marking of arguments. The impossibility of the complex event reading in (18) is due to the saturation of the complex event variable by the internal argument at the lexical output. The event variable that selects a process nominal must be bound by an existential operator on Det°, given the existence of an open external argument variable. The mapping of the internal argument onto spec- of-DP would referentially identify the event variable and disallow the projection of the external argument variable, provided that the internal argument is a referential element by S-structure, i.e. it does not have to go through referential reinterpretation at LF by a Quantifier Raising type of rule. Raising of an internal argument to Spec-of-DP is allowed when the internal Following Rizzi (1986), Roberts (1986), and Fagan (1988), I assume that the projection Principle does not apply to saturated 0-roles. In accordance with them, I also assume that undischarged 0-roles receive an arbitrary interpretation at L F, R eproduced witti permission of ttie copyrigfit owner. Furtfier reproduction profiibited witfiout permission. 7 2 argument is able to saturate referentially the internal argument slot in the Argument structure of a nominal head, i.e. when the Noun can be referentially predicated of its internal argument. This is a property that not every internal argument has, as was first noticed by Anderson (1979). Examples (19) and (20), below, illustrate this difference. (19) a. La construcciôn del edificio. The construction of the building. b. La interpretacion de los datos. The interpretation of the data. The difference between (19a) and (19b) have usually been explained in terms of affectedness (Anderson, 1979). In (19a) the event denoted by the NP affects the object. In (19b) the event denoted by the NP does not affect the object. An affected object has been usually defined as the object that undergoes change of state or condition P7 because of the event denoted by the predicate. This definition accounts for the cases in (19). In (19a), the internal argument edificio. ’building’, undergoes change of state because of the event denoted by construcciôn. ’construction’, while the internal argument in (19b), See Anderson (1979), Roberts (1985) and Hale and Keyser (1987, 1988) for a definition of affectedness based on this notion, R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 7 3 datos. ’data’, does not undergo change. However, Tenny (1987) observed that not all the cases fall into one or the other category so clearly, as we will see below. As we mentioned before, Anderson (1979) shows that affected objects may occur prenominally in English, as in (8), but unaffected objects may not occur in that position, as (9) showed. Zubizarreta (1987) proposes that predicates that project an affected object may incorporate the argument- variable of the internal argument into the head, since a predicate and an affected internal argument form a complex predicate. When the argument-variable is incorporated into the lexicon, the internal argument is not projected, evaluating the variable from the Spec-of- DP adjunct position through Modification.^® An important consequence of Zubizarreta analysis is that NP-trace relations do not appear within NPs. In contrast to Zubizarreta, Tenny (1987) analyzes affectedness as the property by which internal arguments The Rule of Modification that Zubizarreta (1987:52) proposes is the following: (i ) A modifies B in the context [g iff C immediately dominates A and B , C is a projection of B , and B is not a head. If A is an adjunct predicate which contains a variable x, then B or the head of B contains an arg-variable with the index i and x is assigned the value i . If A is an adjunct argument with index i , then B or the head of B contains a variable X and the value i is assigned to x . R eproduced witti permission of ttie copyrigfit owner. Furtfier reproduction profiibited witfiout permission. 7 4 can delimit predicates. Delimitedness is an aspectual property. An internal argument that delimits a predicate is that one that bounds the event denoted by the predicate in its temporal aspect. For instance, in eat an apple the spatial extent of the apple delimits the event denoted by eat. The internal argument provides the predicate with a logical endpoint. Therefore Tenny characterizes affectedness as an aspectual property. Other properties of the internal argument can measure out the event, as the ripeness of the object in ripen the fruit. As Tenny (1987) points out, a definition of affectedness based on delimitedness gives an account for unexplained cases under the "undergoing change of state or condition" definition. Perform, read. and record may appear in the English middle construction and their related nominals in the so-called passive nominal construction, as (20), (21), and (22) illustrate (examples from Tenny, 1987:80), sharing this property with other predicates that project affected internal arguments. It is clear that their internal arguments do not undergo any change of state or condition caused by the event denoted. Instead, we can posit that their internal arguments delimit their aspectual properties. In particular, the internal arguments of these predicates R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 7 5 delimit the lapse of time of the predicates. Tenny’s hypothesis is that aspectual properties mediate the interface between lexical semantics and syntax, instead of 0-role labels. Hence, certain aspectual properties are associated with certain argument positions. (20) a. This play performs easily. b. The play’s performance by the company. (Anderson, 1979) (21) a. This paper reads easily. It is very well written. b. The paper’s reading at the chemical society created a lot of excitement. (22) a. This magnetometer data recorded easily. The data from the echo-sounder was more of a problem, because the scale continually needed to be readjusted. b. This data’s recording was not the problem. It was the analysis that caused us so much trouble. On the other hand, it is clear that ’book’ in (23) cannot delimit the temporal quality of ’discussion’. (23) The discussion of the book by John. Tenny (1987) proposes the Argument structure in (24) for affectedness verbs, where S is a change on some scale derived from some property of the object, and t is a time interval. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 7 6 (24) (direct argument) = (ÛS, ût) Tenny derives the Affectedness Constraint from other independent properties of language that have independent motivation in the grammar: the delimitedness constraint and Case Theory, as in (25). (25) Delimitedness-Case Constraint INFL/AGR cannot assign case to the direct argument of a verb describing a non-delimited event, without the mediation of some special element appearing in the syntax. The problem with Tenny’s proposal is the lack of uniformity in the definition of delimitedness. In some cases it is the time span of the internal argument which delimits the temporal quality of the predicate, as in reading. In other cases it is a property of the internal argument which delimits the temporal quality of the event, as in explosion. In this case the property [+/- explosive] is independent of the time span of the event. The property [+/- explosive] allows the event to have a time span, but it does not delimit it. In addition, in the second case the notion of delimitedness rests on an undefined number of features, as [+/- explosive], [+/- destructible], etc., their systematic importance in linguistic terms being very dubious. Furthermore, these properties are more related to selectional restrictions R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 7 7 of a predicate than to the assignment of argument positions. In a similar fashion, we can argue that the fact that a book is subject to discussion allows the event denoted by ’discussion’ in (23) to have a time span. The logical endpoint for this event would be when all the topics related to the book have been discussed. In the memorization of the lesson, where we have another unaffected object, we have a similar case. The lesson is subject to memorization, and this allows the event to have a time span and also a logical endpoint. When the lesson is memorized the action is finished. We can argue that the limits of the time span of discussion and memorization are more vague than those of destruction and explosion, but the fact that the difference is not so systematic considerably weakens Tenny’s hypothesis. In any case, they do not offer the sharp difference that other non-delimited events show, such as push the cart. We can also argue that in some cases, actually in most cases, the internal argument does not delimit the time span of the event. A property of an internal argument allows the event to occur, but it is difficult to determine to what extent it defines its time span. For instance an apple ripens faster under the sun of the Mohave desert than in the refrigerator at home. Hence, the relationship between the aspectual properties of R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 7 8 affected internal arguments and their predicates in the lexico-syntactic level are weaker than those proposed by Tenny. In conclusion, Tenny’s hypothesis, although it is an interesting proposal, does not make a clear distinction between affected and unaffected objects and is based on a number of unspecified features. In relation to Zubizarreta (1987) and Tenny’s (1987) proposals, Grimshaw (1990) raises an interesting question. She questions whether the affectedness constraint, in the case of nominals, is only related to predicate structure properties or not. Grimshaw (1990) puts Zubizarreta (1987) and Tenny’s (1987) hypotheses into question saying that their accounts are respectively based on Argument structure and Event structure. In accordance with Grimshaw, to base the notion of Affectedness on Argument structure or Event structure apparently raises a contradiction since the so-called passive nominals do not behave as argument structure takers, as shown before in (18). In (18) we can observe that these structures do not behave as complex event nominals. However, Grimshaw points out that the Lexical Conceptual Structure must be involved in some way. As noticed by Grimshaw, prenominal genitives bearing features other than [+ human] can occur in the passive R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 7 9 nominal construction and internai arguments cannot be related to the head noun across a copula, as illustrated in (26). (26)*La construcciôn es del edificio. The construction is the building’s. This entails that passive nominal constructions cannot be a simple relation of predication. In addition, if this were the case, we would loose an important generalization with respect to verbs and middle constructions. Grimshaw concludes that the possessive in passive nominals are complements and not modifiers since they obey selection by the Lexical Conceptual Structure, although they do not act as arguments. Since in the case of nouns not only lexical Conceptual Structure is involved in obtaining reference, but a referential variable is also involved (Williams 1981, 1982), we should study the relationship between this R-variable and the Affectedness Constraint. In the next two sections we will see how the simple event status of "passive nominals" does not depend on the defectiveness of nouns to theta-mark their arguments as proposed by Grimshaw (1990), but on a theory of argument saturation. In other words, complements modify the argument variable of the internal argument, in accordance R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 80 with Zubizarreta (1987), which in turn may saturate the R-variable. I will claim that this is the case of affected internal arguments. 2.4. Towards a re-interpretation of the Affectedness Constraint Jaeggli (1986:608) suggests that the Affectedness Constraint is not an independent theoretical 90 principle, and that it can be derived from a more complex thematic theory. He observes that the thematic relation held by a predicate and an affected object seems to be much better defined than the relation established between a predicate and an unaffected object. It seems that the interpretation of the internal theta role depends on the interpretation of the external role, in the unaffected cases. To put it in another way, an unaffected internal argument may need the presence of an external role to be interpreted. This is not the case for affected objects, that may be interpreted without the presence of an external role. In an attempt to formalize Jaeggli’s insight, I propose that the Affectedness constraint is not to be formulated as an independent principle of grammar; rather Zubizarreta (1987) and Tenny (1987) follow this line of thinking, as we have seen. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 81 its effect is derived from a referential Theory of Argument Saturation. In particular, I claim that an affected object may saturate referentially a predicate, but an unaffected object may not. As noticed by Sanchez de Zavala, p.c. in Demonte (1987: 288), nominals that project an affected internal argument have the property of being non repeatable. To be more precise, they cannot be repeated at a particular point in time. Some events that show affectedness effects, such as the capture or the arrest of a person, can be repeated at different points of time. However, at the same point in time we can have only one event of the ’capture’ or ’arrest’ type with respect to the same object. In that sense, the internal argument identifies referentially an event as ’capture’ or ’arrest’ in a straightforward way. On the other hand, an event with an unaffected object as ’persecution’ is not referentially identified by its internal argument since the same individual can undergo several persecutions at the same point in time. Hence, the event will be identified not only by the internal argument but also by the external argument. At this point, we should consider that nominal morphology, unlike verbal morphology, lacks tense, hence, the event that a nominal denotes cannot be bound to a specific point in time by its own means. Hence, location in the time span R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 82 is not a function of nouns. Then, we should assume that events denoted by nouns should occur in an abstract point in time, that it could be modified by an adverbial expression, but not saturated since it does not have argument representation. Therefore, nominals that project affected objects depend absolutely on their affected arguments for their referentiality. They are introduced by an expletive article or by a possessive pronoun coindexed with the affected object, when the internal on argument is [- generic]. This determiner acts as an operator of the internal argument with respect to the E- variable, if the noun is a complex event noun, or with respect to the R-variable if the noun is a simple event noun or an individual noun. The NP in (19a) denotes an EVENT-TOKEN with respect to its internal argument. We cannot have more than one construction of the house. The NP in (19b) denotes an EVENT-TYPE with respect to its internal argument. We may have more than one interpretation of the data. Hence, affected internal arguments bear a referential index which is able to saturate the internal argument variable, while unaffected internal objects do not. In fact, I am assuming that in both cases there is an incorporated 111 We will see the effects of a generic internal argument below. R eproduced witti permission of ttie copyrigfit owner. Furtfier reproduction profiibited witfiout permission. 8 3 object and that this internal object defines the time span of the event. Therefore, if the projected argument assigns a referential index to the internal variable, it would look as if the syntactic object is determining the time span of the event. For instance, in (19a) the temporal quality of the event is defined by an internal generic variable (=construction), which is referentially bound by the syntactic internal argument (=building). Hence, the temporal aspect of the event is defined by the argument projected onto the syntax by transitivity. In (19b) the internal variable (=interpretation) defines the aspectual properties of the event, but the syntactic argument cannot assign a referential index to the variable, given the properties of the predicate, hence it fails to define the time span of the event by itself. This entails a switch with respect to Tenny’s hypothesis. What is relevant for the notion of affectedness is the referential properties of the internal argument with respect to the predicate, but not its aspectual properties. Internal arguments are a major factor in defining the aspectual properties of a predicator, but in many cases they are not the only one. In most cases the aspectual properties of a predicate are defined in a compositional way. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 8 4 As a result of the discussion above, we can draw the following generalization in (27). (27) An affected internal- argument referentially saturates a nominal predicate. My proposal is that events may license a clitic-like element in the determiner system which they are predicated of. This clitic-like element can take the form of a definite article or a possessive clitic, that act as operators of the internal argument with respect to the E- variable, that is intrinsically referential in the case of nouns that project an affected internal argument. Since events obtain reference through quantification, the clitic must be raised to Spec-of-QP at LF. In accordance with Williams (1981, 1982) and Zubizarreta (1987) among others, result NPs are predicated of a referential index. This referential index is borne by the head of DP. In the case of a simple event nominal or a result nominal, this referential variable may be saturated by one of the nominal arguments through the licensing of a clitic, i.e. definite article or a possessive element in Spanish, in Spec-of-DP. In the case of affected internal arguments, the saturation will 3 1 This property is shared by other kind of nouns, such as proper nouns, and generic nouns. See Longobardi (1992) for further discussion. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 8 5 always occur, while in the case of unaffected objects is optional. Complex eventive nouns are quantified by definition, hence, they must be selected by QP. On the other hand, QP is optional for the other kinds of nouns, as we can observe in (28c), (28d), (28e), and (28f). The R-variable of simple event nouns and result nouns is saturated by Spec-of-DP. In the case of simple event nouns and result nouns with affected nouns, DP is necessary, as we see in (28c) and (28e), since the internal argument referentially saturates the noun. In (28c) and (28d), we can observe that the R-variable binds the E-variable of simple event nouns, while in (28e) and (28f) the E- variable is saturated. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 86 (28) a. Complex event nouns with affected objects QP / \ Op- DetP <e> / \ GenP / \ GenP internal argument- / \ Î Gen’ / \ ! NP b. Complex event nouns with unaffected objects QP / \ 0p(-v DetP <e> ID / \ GenP / \ GenP internal argument,.-, / \ ' ' Î Gen’ Î / \ i________ NP c. Simple event nouns with affected objects (QP) / \ DP <R <e>> / \ Op,- DetP / \ GenP / \ GenP internal complement, / \ Î Gen’ I / \ 1 NP R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 8 7 Simple event nouns with unaffected objects (QP) / \ (DP) (<R <e>>) / \ Opc) DetP <R <e>> ID / \ GenP / \ GenP internal complement^, / \ ' ' t Gen' / \ I NP e. Result nouns with affected objects (QP) / \ DP <R> / \ Op, DetP <*e> / \ NP / \ NP appositive complement; f. Result nouns with unaffected objects (QP) / \ (DP) <R> / \ Op,,, DetP <*e> ID / \ GenP / \ GenP internal complement,,, / \ ' ' Î Gen’ 1 / \ I I _ _ _ _ _ _ _ _ _ _ _ ' NP As we have seen, unaffected internal arguments cannot saturate the R variable, although, a clitic R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 88 identifying these kind of arguments may appear in the Spec-of-DP. As (29) illustrates, coordination between a nominal that affects its internal argument and a nominal that does not affect its internal argument is possible under the node that the possessive pronoun occupies. Therefore the possessive pronoun must occupy the same position, either if the possessive pronoun identifies an affected object or an unaffected object. (29) Su persecuciôn y arresto llevô pocos minutes. His persecution and arrest took a few minutes. (29) entails that, while an affected object has to saturate the R-variable by definition, an unaffected object only optionally does this. Notice for instance that in la descripciôn de una catedral ’the description of a cathedral’ the saturation of the R-variable is optional, i.e. it can have a token reading or a type reading. In chapter 3, I will propose a double articulation of the determiner system. The highest DP is headed by strong determiners, while the lower DP, DetP, will be headed by weak determiners. The mechanism of licensing this clitic in DP is through raising the operator from Spec-of-DetP to Spec-of-DP/QP at the syntactic level of representation, or by direct mapping of the operator at the lexical level of representation. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 8 9 As a result, the external argument will be syntactically obliterated in the second case, and non-complex event reading will arise. Complex-event nominals have an existential interpretation by definition, i.e. they are introduced by DetP. However, their internal argument must saturate the E-variable, inducing the determiner to raise to QP. This mapping would be triggered by the same referential properties as those of head movement in the form proposed by Longobardi (1992) for proper names. In accordance with Longobardi (1992) proper names raise to D“ giving their specific referential properties. In a similar fashion the clitic that identifies an internal argument is able to saturate the referential variable of which the NP is predicated, hence, the mapping of the clitic onto DP is licensed. This leads to the definition of a theory of mapping of an element onto DP based on the ability of that element to saturate or modify the referential variable of the nominal. Another issue raised by this reasoning is why only possessives that identify affected internal arguments are licensed in DP in English, as (30) illustrates, while in Spanish a possessive pronoun can identify either an affected object or an unaffected object, as (31) exemplifies. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 9 0 (30) a. Rome’s destruction. b.*Rome’s description. (31) a. Su (=Roma) destruccion. Its (=Rome) destruction b. Su (=Roma) descripciôn. Its (=Rome) description I will answer this question in the next section, presenting some cases of affectedness effects within Spanish nominals unnoticed in previous literature. I will propose that an internal argument that saturates the E( vent)-variable, i.e. an "affected" internal argument, must take scope over the E-variable, either at S- structure or at LF. If it occurs at the lexical level of representation, as in (30a) the E-variable is lexically saturated, and, therefore, the external argument variable too. In the case, that there is a complex event reading, such as in the destruction of Rome, the identification of the E-variable occurs in English at LF, since the determiner in English lacks agreement features. On the other hand, the clitic possessive in Spanish shows strong agreement features, hence, the NP that saturates the E/R- variable can check features at the syntactic level of representation. 12 I will use this term when I do not sake a difference between the different types of nouns. R eproduced witti permission of ttie copyrigfit owner. Furtfier reproduction profiibited witfiout permission. 91 2.5. Affectedness effects in Spanish 2.5.1. Passive nominals and Quantifier Raising. It is well-known that passive nominals and middles in Spanish apparently do not show the same Affectedness effects that they do in English, as we mentioned at the end of last section regarding the passive nominals. In Spanish the prenominal position cannot be occupied by a full NP, Only possessives may occupy that position. In contrast with English, this prenominal position can be occupied by possessives identifying either affected internal arguments or unaffected internal arguments, as noticed by Zubizarreta (1987:70) for French and Spanish and Cinque (1980) for Italian, This has been illustrated in (30) and (31), and see also Zubizarreta’s examples in (32) and (33). 99 See Roberts (1986) for the differences between the middle construction in English and Spanish. Cinque (1980) and Zubizarreta (1987) do not explicitly characterize these nouns a s predicators of unaffected internal arguments, but these nouns obviously match quite well the group of nouns that show unaffectedness effects in English. Traduccion, ’ translation’, and lectura, ’ reading’, might be considered problematic, since Tenny (1987), for instance, includes them as predicators of affected objects. I think that these two nouns have two interpretations, one as predicators of affected objects and another one as predicators of unaffected objects. I n the case of lectura, when i t refers to the action of reading, it projects an affected object. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 9 2 (32) a. sa description (Fr.)/su descripciôn (Sp.) its description b. son interpretation (Fr.)/su interpretacion its interpretation c. sa traduction (Fr.)/su traducciôn (Sp.) its translation d. sa demonstration (Fr. )/su demostracion (Sp. ) its demonstration (33) a. sa destruction (Fr.)/su destruccion (Sp.) its destruction b. sa capture (Fr.)/su captura (Sp.) his/her capture c. son assassinat (Fr.)/su asesinato (Sp.) his/her assassination d. son execution (Fr.)/su ejecuciôn (Sp.) his/her execution These facts present some problems for a theory of Affectedness based on semantics within the Principles and Parameters Theory and the Minimalist Hypothesis (Chomsky, 1993). Since the Affectedness constraint seems to have a semantic nature, the parameter between English and the Romance languages must be morphological, in accordance with Chomsky (1993). In particular, the lack of overt $- features in the English possessive particle and the When its meaning is close to 'interpretation', it projects an unaffected object. In the case of traducciôn. if it refers to the translation of something into another language, as if there were only one translation, or an ideal translation, then i t projects an affected object. When it is considered that there can be different versions, then it projects an unaffected object. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 9 3 presence of them in the Romance possessive pronouns appear to be the clue to these facts. Zubizarreta (1987), following Zubizarreta (1979) and Aoun (1985), proposes that the Determiner can act as an A’-binder of a postnominal position. As an A’-binder of a postnominal position it should be able to bind either an affected internal argument or an unaffected internal argument. Therefore, the possessive pronoun is the element that saturates the E/R-variable. I claim that the element that saturates the E/R-variable must be in Spec-head relation with the E/R-variable, either at S- structure, as in the Romance languages, or at LF, as in English. The lack of ^-features by the possessive particle in English requires the raising of the postnominal internal argument at LF to comply with the Principle of Full Interpretation. Raising of an unaffected internal argument is lexically blocked since they are not able to saturate the R-variable until its $- features are available, at a component of the grammar intermediate between the lexical level of representation and the syntactic level of representation. At the same time, a possessive particle cannot appear in D° as a 3 5 Thus, the variable moves to the head of QP, in the case of complex event nominals, and to the head of DP, in the case of the other kind of nouns. R eproduced witti permission of ttie copyrigfit owner. Furtfier reproduction profiibited witfiout permission. 9 4 binder of a postnominal unaffected internal argument since the possessive particle lacks any ^-features to be interpreted in English. On the other hand, Romance possessive determiners and definite articles bear ^-features, being able to act as A’-binders of a postnominal NP, in the first case, for instance su^ destruccion pro^, while in the second case, la destruccion de Roma for example. Spec-head agreement will transmit the features to the noun, which, in turn, is coindexed with the determiner. In chapter 3 I will propose that DetP, the lower determiner projection moves always to Spec of the higher projection that selects it. Finally, we need to support our claim that an affected internal argument within a complex event nominal is always coindexed with Spec-of-QP, while an unaffected internal argument does not necessarily. In (34), we see that a possessive pronoun binding a postnominal unaffected internal argument is possible in the presence of a quantified adjunct-argument and the NP still denotes an event, as the controlled subject of a purpose clause confirms. (34) Su (=libro) discusiôn por cada profesor para Its (=book) discussion by each professor in clarificar el contenido fue brillante, order to clarify its content was brilliant. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 9 5 In order for the noun discusiôn to receive a quantified reading we must assume that the quantified adjunct moves to Spec-of-QP to take scope over the E- variable. Since the unaffected internal argument does not have to be raised in order to quantify the E-variable, we do not have a clash between the quantified adjunct and the internal argument in Spec-of-QP at LF, since the internal argument is in Spec-of-DetP. Notice that the overt presence of a quantifier in Spec-of-QP blocks the raising of a quantified adjunct at LF, as in (35). The sentence is ungrammatical since the quantifier cada ’each’ takes obligatory wide scope, and narrow scope interpretation is not allowed. (35)*Toda su (=libro) discusiôn por cada profesor All its (=book) discussion by each professor fue brillante, was brilliant. If we look now at the behavior of affected internal arguments, we can observe that they disallow the presence of a quantified A-adjunct, as (36) illustrates.^^ An analysis of Quantifier Raising in which the quantifier occupies an adjunct position, à la May (1985), would render the same results through the Biuniqueness Condition. Furthermore, in order to explain the presence of a prenominal adjunct in English, such as in the enemies’ destruction of the city, we need to assume the projection of QP and movement of the internal R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 9 6 (36)*Su (=ciudad) destruccion por cada ejército Its (=city) destruction by each enemy array eneraigo para ganar territories fue condenada in order to earn territories was condemned por la ONU. by the UN. As we can observe in (36), the affected internal argument and the quantified A-adjunct collide in their quantification of the E-variable at LF. We have the same effects when we have an overt quantifier in Spec-of-QP by S-structure when the noun is a complex event nominal but not when it is a simple event nominal, as (37) shows. In (37a) we cannot have an interpretation where the event of destruccion can control PRO in the purpose clause, i.e. we cannot have a complex event reading of the noun. (37) a.*Toda su destruccion para permitir el paso de 'All its destruction to let the troops de las tropas ... (event control) pass ...’ b. Filmaron toda su destruccion. 'They filmed all its destruction.' argument to Spec-of-QP at LF. The issue that remains to solve is whether the clitic that hinds the affected internal argument of a complex event noun in Romance is mapped onto Spec-of-QP, at the syntactic level or at LF. I am assuming that it occurs at the syntactic level for reasons of economy, but further research is needed. Nothing crucial hinges upon this issue for my analysis. R eproduced witti permission of ttie copyrigtit owner. Furttier reproduction profiibited witfiout permission. 9 7 In (37a), the quantifier and the possessive clash in the Spec-of-QP position. In (37b), the quantifier occupies Spec-of-QP, while the possessive occupies Spec- of-DP, as proposed in (28c). In conclusion, we have observed that affected internal arguments and unaffected internal arguments have a different syntactic behavior in Spanish also. Affected internal arguments are bound by an operator in Spec-of-QP at the syntactic component of the grammar, while unaffected internal arguments do not. The first ones, then, behave as quantified expressions. The saturation of the internal argument variable by an affected object and the binding of the E-variable by this argument variable produce a one-to-one relation between the event and the internal argument. Hence the number of events would be given by the number of referring expressions in the internal argument. In the destruction of the cities we have "a destruction event" for each city. In the interpretation of the data we have just "one interpretation event" for all the data. Therefore, the quantified interpretation of a complex event noun is given by its internal argument. Now, we can define an affected internal argument as in (38). See f n . 3 6 . R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 9 8 (38) An affected internal argument is the one that bears a one-to-one relation to its predicator. Mourelatos (1981) proposes that events are count- quantified, in contrast with states and activities that are mass-quantified. The consequence of this fact is that events can be quantified by cardinal count adverbials while states and activities or processes cannot. When we deal with nominalizations, we should consider that the complex events that project an affected object must be Op always singular. Op Zubizarreta (1987:73) made a very interesting observation in this respect. She notices that [-count] event nouns, such as destruction, always take a predicate-argument structure. As ( i) shows, these kinds of noun cannot appear without a predicate-argument structure. (i) a.*Tres destrucciones. Three destructions h.‘Varias destrucciones. Several destructions c.‘Esta destruccion. This destruction On the other hand jtcountj event nouns, as execution, may take a predicate-argument structure or not. (ii) a . Tres ejecuciones. Three executions h . Varias ejecuciones. Several executions c . Esta ejecuciôn. This execution R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 9 9 In the next section, I am going to study the relation between affectedness and the Possessivization Principle. 2.5.2. Affectedness and Possessivization. 2.5.2.1. The derivation of simple event nominals and result nominals. The proposals in sections 2.4. and 2.5.1. have important consequences for the explanation of the contrast in (39), which was first observed by Cinque (1980) for Italian, Zubizarreta (1987) for Spanish and French, and Picallo (1991) for Catalan. (39) a.*Su destruccion de la casa (llevô mucho His destruction of the house (took a long tiempo). time). b. Su descripciôn de la casa (llevô mucho His description of the house (took a long tiempo). time) . Therefore, the only way for a [-count] event noun t o be quantified i s through predicate-argument structure, unless it is quantified directly by a quantifier expression. A direct quantification of a [-count] event noun must yield a grammatical expression. This is borne out in (iii). (iii) Toda esta destruccion no nos llevara a ninguna parte. All this destruction will not take us anywhere. R eproduced witti permission of ttie copyrigtit owner. Furttier reproduction protiibited wittiout permission. 100 Cinque (1980) argues that the two nouns have a different lexical nature. The one in (39a) belongs to a class of nouns that always appears in the passive through the rule of passivization by a lexical rule. The second class is represented by the noun in (39b). This class allows both forms, an active one and a passive one. Following this analysis, Picallo (1991) posits that there is a morphological difference between non-deverbal nominals, which cannot undergo passivization, and the deverbal nominals, that can undergo passivization. Deverbal nominals are formed by a stem, a nominalization affix, with Gender and Number. Non-deverbal nominals contain a stem with Gender and Number. The nominalization affix allows a nominal to undergo passivization. However, as pointed out by Zubizarreta (1987), nominals lack the Lexical Syntactic level of representation, where passivization occurs, as do other processes that affect the syntactic structure of arguments. Hence, nominals cannot undergo passivization. Zubizarreta (1987) and Picallo (1991) claim that the interpretations of the two forms are different. The "passive form" has a complex event interpretation, while the "active form" has a result interpretation. On the other hand, nouns related to destruction, as assassination, capture, or execution, cannot denote an individual object, hence, they can only R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 101 be interpreted as events. This is not completely correct. We can find counter-examples in the two groups. In Spanish, as in English, construcciôn. ’construction’, may denote either the result of a process, as in (40a), or an event, as in (40b), where it is able to control the purpose clause. However, construcciôn cannot co-occur with the prenominal possessive pronoun, referring to an agent, and an internal argument, as shown in (41). (40) a. La construcciôn de San Pedro esta en el The building of Saint Peter’s is in the centro de la ciudad. center of the city. b. La construcciôn de este puente para The construction of this bridge in order to facilitar la comunicaciôn entre las dos make the communication between both regions comarcas ha sido una gran idea, easier has been a great idea. (41)*Su construcciôn de la torre de Babel. Their construction of the tower of Babel. The difference between construcciôn and destrucciôn is that construcciôn can denote an individual object. On the other hand, memorizaciôn ’memorization’ or persecuciôn ’persecution’ may not have a result reading, as (42) illustrates. In (42) memorizaciôn and persecuciôn are incompatible with estar. because they are not R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 102 individual objects, but events.They may only have an event reading, as in (43). Although they do not have a result reading, they may co-occur with a prenominal possessive pronoun and an internal argument, as in (44). (42) a.*La memorizaciôn esta en el despacho. The memorization is in the office. b.*La persecuciôn esta en las calles. The persecution is in the streets. (43) a. La memorizaciôn de la lecciôn para mostrar The memorization of the lesson in order to su capacidad. show his skills. b. La persecuciôn del ladrôn para asustarlo. The persecution of the thief in order to scare him. (44) a. Su lenta memorizaciôn de la lecciôn nos His slow memorization of the lesson caused causô muchos problemas. us many problems. b. Su espectacular persecuciôn de los ladrones His spectacular persecution of the thieves les valiô una recompensa, earned them a reward. on estar, 'to be’, is used to express location of individual objects in Spanish, as in ( i). Ser, 'to be', is used to express location of events, as in (ii). | i | La mesa esta/*es en la casa de Ana. 'The table is in Ana's house.' (ii)La fiesta es/*esta en casa de Ana. 'The party is Ana's house.' R eproduced witti permission of ttie copyrigtit owner. Furttier reproduction protiibited wittiout permission. 1 0 3 Furthermore, we can observe that the interpretations of (44a) and (44b) are eventive. However, this is not the complex event interpretation, as Grimshaw’s tests show in (45), (46), and (47), but the simple event one. Grimshaw (1990) argues that nouns like party, race. or trip do not have Argument Structure, but they still have an event interpretation. They differ with respect to complex event nominals, nouns that do have Argument Structure, in some syntactic aspects, as we have pointed out in section 1.10. First, "simple event nominals" do not allow event control of a purpose clause. Second, they do not allow aspectual modifiers like frequent or constant. Finally, they do not occur with the same kind of aspectual adjuncts as their related verbs do. (45) a. Su minuciosa memorizaciôn^ de la lecciôn His thorough memorization of the lesson (*para PRO; hacer que los profesores lo (in order to make the professors to respect respeten) llevô mucho tiempo. him) took a long time. b. Su espectacular persecuciôn^ de los ladrones His spectacular persecution of the thieves (*para PRO; hacer que los ladrones lo teman) (in order to make the thieves scared of acabô en tragedia. him) finished tragically. (46) a. Su (??frecuente) memorizaciôn de la lecciôn His (frequent) memorization of the lesson R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 0 4 fue criticada por el profesor porque no was criticized by the professor because facilita la comprensiôn. it did not make the comprehension of it easier, b. Su (??frecuente) persecuciôn de los gitanos Their (frequent) persecution of the gypsies fue condenada por el gobierno. was condemned by the government. (47) a. Su minuciosa memorizaciôn de la lecciôn His thorough memorization of the lesson (♦durante sôlo una hora) fue criticada por (for just one hour) was criticized by the el profesor porque no facilita la professor beacuse it did not make the comprensiôn. comprehension of it easier. b. Su espectacular persecuciôn de los ladrones His spectacular persecution of the thieves (♦durante una hora) acabô en tragedia. (for one hour) finished tragically. In (45) event control of the purpose clause, in Williams’ (1985) sense, is ruled out when the nominal has an "active form". In (46) and (47), we can observe that an "active form" does not have a complex event interpretation, since it does not admit aspectual modifiers. Therefore, eventive nominals that have two genitive phrases within their syntactic structure cannot have a complex event interpretation. Notice that all R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 0 5 these sentences become good when the agent is projected in a prepositional adjunct as (48), (49), and (50) illustrate. (48) a. La minuciosa memorizacioUj de la lecciôn The thorough memorization of the lesson por parte de él para PRO- hacer que los by him in order to make the professors profesores lo respeten llevô mucho tiempo. respect him took a long time. b. La espectacular persecuciôn^ de los ladrones The spectacular persecution of the thieves por parte de él para PRO- hacer que los by him in order to make the thieves ladrones lo teman acabô en tragedia. scared of him finished tragically. (49) a. La frecuente memorizaciôn de la lecciôn The frequent memorization of the lesson por parte de él fue criticada por el by him was criticized by the professor profesor porque no facilita la comprensiôn, because it did not make the comprehension of it easier. b. La frecuente persecuciôn de los gitanos por The frequent persecution of the gypsies by parte de ello fue condenada por el gobierno. them was condemned by the government. (50) a. La minuciosa memorizaciôn de la lecciôn The thorough memorization of the lesson por parte de él durante sôlo una hora fue by him during just one hour was criticized R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 0 6 criticada por el profesor porque no facilita by the professor beacuse it did not make the la comprensiôn. comprehension of it easier. b. La espectacular persecuciôn de los ladrones The spectacular persecution of the thieves por parte de él durante una hora acabô en for one hour finished tragically. tragedia. Complex event nominals lack an active form in the Romance languages discussed in this section.Only one argument can receive the genitive case-marking within an NP in French, Spanish, Italian, and Catalan. This entails that when we are in the presence of two genitive phrases, as in (44), one is an extrinsic element in relation to the Argument Structure of the NP, in the sense of Godard (1992). The R-variable of complex event nominals can be evaluated only by its own arguments. The presence of an extrinsic element, as in (51), produces the saturation of To be more precise, a complex event nominal lacks an active form when the external argument either is not PRO or has not been mapped as a referential adjective. These cases need further study. The point is that we cannot use the term "active form" as it has been used in previous literature, i.e. when the apparent external argument was assigned genitive case, for these Romance languages, since it does not parallel the verb case. In the case of Romance eventive nouns, when another element receives genitive case, it is an extrinsic element that does not belong to the A-structure of the Noun. R eproduced witti perm ission of ttie copyrigtit owner. Furttier reproduction protiibited wittiout permission. 1 0 7 the Referential variable, and the loss of the existential interpretation of the nominal by S-structure, hence, the loss of the complex event reading. (51) Esta invasion sera condenada por la ONU. This invasion will be condemned by the UN. The difference between (51) and (52) is that the R- variable of the nominal in (51) is evaluated by an extrinsic element, while the R-variable in (52) is evaluated by an internal argument, or not projected, allowing the complex event reading. (52) La invasion de este pais ha sido condenada The invasion of that country was condemned por la ONU. by the UN. Therefore, we do not need to posit either two different lexical entries or two different morphological affixes of invasion in order to explain its two different readings: the complex event reading and the simple event reading. Going back to (39a), I claim that this sentence is ruled out because of a class in Spec-of-DP, or the Biuniqueness Condition. The R-variable cannot be evaluated by two different elements, the possessive pronoun and the affected internal argument, hence, (39a) R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 0 8 is ruled out since coindexation of the affected internal argument with a scope marker or raising the affected internal argument to quantify and evaluate the R-variable is obligatory in natural languages. With regards to result nominals, I claim that the difference occurs at the level of affixation. Some important differences between result nominals and process nominals indicate that the nature of a noun, as either a process or a result, is decided no later than the occurrence of affixation. As pointed out by Picallo for Catalan, the agent of a result nominal can never appear in a by-phrase, as illustrated by (53). (53) La descripciôn del Mediterraneo de/*por ’The description of the Mediterranean sea by Josep Fla fue publicada por primera vez en Josep Fla was published for the first time un periôdico. in a newspaper’ In (53), the verb to be published forces the result reading of the nominal. As a consequence of the fact that the result reading is the only one available, the agent can be introduced by the possessive preposition de, ’of’, but not by the agentive por. ’by’. Zubizarreta (1987) points out that the possessor relation, only borne by a R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 0 9 result nominal or an object, includes those notions of owner and creator. Another difference between process nominals and result nominals is that a "group adjective" can refer to an owner or to a creator in the case of result nominals, while in the case of process nominals, denoting either a complex event or a simple event, it can only refer to an agent. (54) illustrates how a group adjective that modifies the possessor variable of a result nominal may refer to an owner, but not to a creator. (54) La construcciôn andaluza, erigida por los The andalucian construction, erected by the arabes en la Edad Media, es un claro ejemplo arabs in the Middle Ages, is a clear example del esplendor del Islam en aquella época. of the Islamic splendor in that period. On the other hand, when the noun denotes an event, a group adjective can only refer to the "provenance" of the event, as in (55). (55) La construcciôn alemana del puente en tres The German construction of the bridge in three dias. days. These facts entail that the selection of the process or the event denotation of a noun must occur at the R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 110 lexical level of representation, since once we enter the computational system of syntax we cannot alter lexical material. To maintain only one lexical entry for result nominals and process nominals would be the optimal solution in terms of economy, since it would be interesting to explain in some fashion the parallel between the semantic properties of the complement position of result and process nominals. In contrast with Picallo (1991), I do not think that there is enough evidence to propose that the affix type is the element that defines the dichotomy process-result. The difference could be better established by stating that a nominalization affix is required in the case of process nominals, but it is not relevant in the case of result nominals. Those result nouns that lexically derive from an eventive noun have the same affix as the eventive noun. I suggest that the nature of the eventive variable determines whether a nominal is a process or a result. If the eventive variable undergoes affixation as an open variable, i.e. [- presupposed], then we have an event that can be saturated by an extrinsic element, producing a simple event nominal in that case, or a complex event nominal if it is not saturated by an extrinsic element, in that it is only saturated by an argument. If the eventive variable appears in the syntax as a closed R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. I l l variable, i.e. [+ presupposed], then we have a result nominal. As Picallo (1991:290) herself points out, the use of different affixes by the same stem to denote a result or a process is not very productive. She only points out two, and in one of the cases there is not an absolute consensus among native speakers. The two cases are descoberta. denoting a process for some speakers, and descobriment. denoting a result, meaning ’discovery’; and crema ’burning’, and cremada ’burn’.^^ It is difficult to think of similar contrasts in Spanish. As Lang (1990:140) points out, the main deverbal nominalizing affixes in Spanish ( -miento. -cion. -dura, -a.ie ) are broadly defined as acciôn y efecto. ’action and effect’, in Spanish Dictionaries. Lang (1990:144) points out also that the one that tends more towards materialisation is - dura. This is especially true of those nouns with this kind of affix that denote a state or property, as it is in the case of rozadura ’scratch, friction’, rotura ’break, breaking’, quemadura ’burn, burning’, fractura ’fracture’, etc. Even when they appear in the Lexicon as actions and effects, they mainly define an effect. For The perfective value of the suffix is interesting in Catalan. Verbs form their past participles with this suffix. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 112 instance, we are more likely to identify rozadura with an effect than with an action, for which we usually use rozamiento. This could be related to the fact that state- affixes and action-affixes show a clearer dichotomy. - dura is a polysémie affix that denotes ’action and effect’ or a ’property or quality’. Lang (1990:136) points out that -ura is used also as a de-adjectival nominaliser, in cases such as altura ’height’, anchura ’width’, bravura ’bravery’, locura ’madness’. In diachronic perspective, its productivity, either as a deverbal or a de-adjectival nominaliser is decreasing, as noticed by Lang (1990), except for some technological terms such as acopladura ’assembly’, soldadura ’weld, welding’, cinceladura ’chiselling’, etc. These forms seem to me to be semantically closely linked to the rozadura group. The fact that the affix -ura is used for two different purposes, de-adjectivization and deverbalization, makes it the most unstable of the deverbal affixes. Otherwise, we have also inherent accomplishment nouns with -ura, such as botadura. The fact that there is not a clear dichotomy seems to entail that there is another mechanism apart from affixation involved in the choice of interpretation for nominals. This is the choice of eventive variable. If the eventive variable is lexically saturated then we have a result R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 1 3 nominal. The lexical conceptual structure may appear syntactically, but not as arguments since the eventive variable is not syntactically active. The only active cover variable is the R-variable, that may be saturated by an extrinsic variable, as in (53), or left open, as is the case for the type reading of descripciôn in (56). (56) Juan y Maria leyeron una descripciôn de Juan and Maria read a description of Barcelona. Barcelona. In the case of complex event nominals only the eventive variable is syntactically active. This variable is saturated by the internal argument in the case of affected objects. The complex event nominals lack the function TENSE, hence the external argument cannot saturate the construct EVENT + INTERNAL ARGUMENT. Thus the external argument can appear within complex event nominals as a prepositional adjunct, as a syntactic open variable (PRO), or as an adjective. A complex event nominal cannot assign double genitive case, otherwise it would result in a Case Filter violation. Simple events are defined by the syntactically active presence of the two cover variables, the eventive and the referential. Simple events can assign double genitive case. The eventive variable, as a first R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 1 4 predicate, may assign genitive case to the internal argument, if it is not lexically saturated, and the referential variable as a second predicate, may assign genitive case to the external argument. The relationships between the cover arguments and the overt arguments are defined by two different relationships, projection and saturation. The E-variable projects the internal argument and the external argument. The R-variable projects the argumentai Possessor- variable. These arguments can saturate the non-overt variables. Recall that the possessor variable covers also the notion of agent, in accordance with Zubizarreta (1987). However, what is relevant to an understanding of this analysis of affectedness is the relationship between the two non-overt variables, the eventive and the referential variables. In the case that both of them are present, the eventive variable may or may not saturate the referential variable. The eventive variable is saturated by the syntactic projection of the argument- structure. In the case of an affected internal argument, the eventive structure saturates any referential variable. This is reflected by the fact that complex event nouns that project an affected internal argument are always selected by a definite article as (57) shows. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 1 5 (57) La/*una destrucciôn de la ciudad. The/a destruction of the city. In this case, the definite article acts as a backwards anaphor or cataphor. As we will see in chapter 3, its omission is not possible when the head noun is deleted. On the other hand, this is possible with unaffected objects. Also, unaffected objects can co-occur with indefinite articles selecting the main noun, as (58) illustrates. (58) a. un constante estudio de la economia. a constant study of the economy. b. un estudio de la economia para impresionar a study of the economy in order to impress al jefe de la oposiciôn. the leader of the opposition. c. un estudio de la economia durante tres a study of the economy over three years. anos. We should recall Bull’s (1965) notions of totality and partitivity. He relates the definite article to the notion of totality and the indefinite article to the notion of partitivity. Bull (1965:215) states that "the definite article indicates totality (one unique) and the indefinite article marks partitiveness (one of many)." Hence, an eventive noun that projects an affected object R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 1 6 denotes always totality, while one that projects an unaffected object may or may not denote totality. If the correct analysis for (39a) is the one based on semantic saturation of the referential variable, proposed here, why is the English counterpart to (39a) grammatical? Why is (lb), repeated here under (59), also correct? (59) The barbarians’ destruction of Rome. Basically, the difference is that in (39a) ^ ranges over destrucciôn de la casa, while in the English counterpart, of the house ranges over his destruction when said with neutral intonation, i.e. the phrasal stress falls on house. Since his does not saturate the referential variable of destruction, his destruction can be modified by another element, in this case of Rome. As (60) illustrates, a noun that is saturated by a prenominal element cannot be further modified. (60)*This execution of Peter. In the case of a noun that does not affect its complement, such as (61), we can give a similar analysis. (61) His description of Rome. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 1 7 In (61) his description is modified by of Rome when said with neutral intonation. If we want destruction of Rome to be modified by his. we need to change the intonation of the phrase to (62). (62) HIS description of Rome. If my analysis for Spanish is correct, it would be impossible to apply the intonation pattern in (62) to a nominal expression headed by a noun that affects its complement. This would yield ungrammaticality as seen in (63) . (63)*HIS execution of Peter. In (63) of Peter does not bear the phrasal stress, hence it does not take scope over the possessive modifier his. Since it saturates the Event variable of the head noun, the head noun cannot be further modified, his being a case of vacuous modification. Therefore, (63) would be the parallel case to the Spanish (39a). (39a) cannot have a similar interpretation to (59) or his execution of Peter. Following Zubizarreta (1987:75), I propose that the possessive element is a scope marker in an A’- position. This would disallow any other element from within the nominal expression to take scope over the prenominal possessive modification at LF in Spanish. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 1 8 Hence, when we have a prenominal possessive in Spanish, scope relations are codified by S-structure. If another element from within the nominal expression took scope over the prenominal possessive at LF, it would give rise to a violation of the Relativized Minimality Condition (Rizzi, 1990), as (64) shows. (64) DP Î / \ 1 suj D’ X / \ I ■'i i I I I t; GenPl 1 / \ j GenP2j Gen'l descripciôn de B. / \ On the other hand, in English this movement at LF is allowed since the prenominal possessive occupies an A- position, i.e. a case marked position. Therefore, description of Barcelona can take scope over "his + case- marker." The prenominal possessive in English is mapped onto a Specifier position at the beginning of the syntactic derivation. The genitive case marker acts as a linker between the possessor and the possessum. This linker incorporates into the higher head or the lower head, depending on whether the possessum has taken scope over the possessor or not. Since it does not involve the agreement strategy of Romance genitive case assignment, the possessor and the possessum do not have to hold a R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 1 9 "direct" spec-head relation. By raising or lowering the genitive linker at LF we have an "indirect" or transitive spec-head relation between the possessor and the possessum. (65) shows the syntactic representation at LF for (61). (65) DP / \ GenP2; D' description of B. / \ Gen marker- GenPl ^ / \ his Gen’l / \ In terms of interpretation, the Spanish counterpart to (65) would be la descripciôn suya de Barcelona, where the possessive suya does not act as a scope marker, allowing the adpositional genitive phrase to take scope over the adjectival modification. In summary, we can extend our analysis of affectedness in nominalizations to English, once that the different strategies of genitive case assigment and their syntactic consequences have been clarified. Furthermore, the contrast between (62) and (63) gives additional empirical support to our hypothesis. Another fact that distinguishes English from Spanish must be studied. PRO is obliterated when the affected internal argument is preposed in English, but not when R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 120 the possessive pronoun appears in prenominal position in Romance Languages. 2.5.2.2. PRO and the Affectedness Constraint Several arguments in favor of the presence of PRO in nominalizations have been given. Its syntactic position within the nominal structure has been widely discussed. Chomsky (1981) defines PRO as a pronominal anaphor that must appear in an ungoverned position as a result of the PRO Theorem and the Binding Theory. The first argument that supports the existence of PRO within event nominals was given by Roeper (1983). He argues that there is subject control when a purpose clause appears within an eventive NP, as in (66). (66) El hundimiento del barco para cobrar el seguro. The sinking of the ship to collect the insurance. Roeper (1983) argues that the subject in the purpose clause is controlled by a non-overt subject of the NP, that receives a "transitive interpretation", namely PRO. As (67) illustrates, if we supress the external argument But see Bouchard (1984) fora "governed account' of PRO. The ungoverned characterization of PRO is necessary for Stowell's (1989) hypothesis for the position of PRO. I will present some independent evidence, which does not rely on the PRO theorem, in favor of Stowell's proposal. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 121 of the event noun through anticausativization, the sentence is ruled out since there is no available controller for PRO. (67)*E1 hundimiento del barco solo para cobrar The sinking of the ship alone to collect el seguro. the insurance. Chomsky (1986b) presented another argument in favor of PRO within NP. The presence of PRO would account for the contrast between (68a) and (68b). (68) a.*The boySj told a PRO story about them;, b. The boySj heard a PRO story about them.. In (68a) PRO is controlled by the subject of VP. The pronoun cannot be coindexed with the subject of the matrix sentence given Condition B of the Binding Theory that forces the disjoint reference between PRO and the pronoun. In (68b) PRO is not controlled by the subject of VP, hence, coindexing of the pronoun them with the boys is allowed since it does not result in the violation of Binding Conditions. Giorgi and Longobardi (1991) gave four further pieces of empirical support for the optional projection of PRO in the syntax of Noun Phrases. The four cases are based on the projection of an anaphor in the complement R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 122 position of an NP that must be bound by a proper antecedent. Giorgi and Longobardi have proposed that this antecedent must be PRO, following Chomsky’s (1981) suggestion. As proposed in Giorgi (1984), the Italian possessive proprio ’own’ is a long-distance anaphor, hence, it is subject-orientated, as (69) illustrates (example from Giorgi & Longobardi, 1991:177). (69) Gianni, ha informato Maria- che il proprio^j;; Gianni informed Maria that nis own lawyer would avvocato avrebbe seguito il processo. follow the trial. In (70a) the long-distance proprio is coreferential with an indirect object, but, since it is a long-distance anaphor, it cannot be directly bound by the indirect object. Giorgi and Longobardi argued that a PRO projected by the NP binds the anaphor. In fact, the only possible interpretation of the sentence is that where the external argument of indagine corefers with the anaphor, as (70b) illustrates. (70) a. Ho consigliato a Maria; un’attenta indagine I recommended to Maria a careful sui fatti che avevano portato all’arresto investigation of the facts which led to del proprij genitori (da parte della the arrest of her own parents (by the R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 2 3 polizia). police), b.*Ho consigliato a Mariaj un’attenta indagine I recommended to Maria a careful da parte del suo avvocato sui fatti che investigation by her lawyer about the facts avevano portato all’arresto del propri. which led to the arrest of self’s genitori. parents. The second argument that supports the projection of PRO as an external argument of NPs is based on the possibility for PRO to have split antecedents in contexts of obligatory control, when both antecedents c-command PRO, in Italian. (71) illustrates this point. (71) Gianni; voleva che Maria- si convincesse che Gianni wanted Maria to be convinced that it was era ora di PRO;^: liberare se stessi da time to free thdmselves from that embarrassing quell ’ imbarazzante situazione. situation. Lexical anaphors cannot have split antecedents, hence, in (72) there must be an element that mediates the binding relation between the anaphor and its antecedents. Giorgi and Longobardi propose that that element is PRO. (72) Gianni; voleva che Maria- sapesse che era Gianni wanted Mario to Know that the time of R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 2 4 arrivata finalmente I’ora della liberazione di the liberation of themselves from slavery had se stessi;^j/„K dalla schiavitù, finally ari The third piece of evidence is based on cases of "backward control", as in (73). (73) La conoscenza di se stessOj è stata molto utile The knowledge of himself has been very useful a MariOj. to Mario. In (73) the antecedent of the anaphor does not c-command the anaphor, hence, there must be an element that bridges the gap in the coreferential relation, as is the case in (74). (74) PRO: conoscere se stessoj è stato molto utile To know himself has been very useful to Mario. a MariOj. The last argument is based on the necessity for an arbitrary anaphor to be bound, as in (75).^^ (75) Mario ti convincerà dell’importanza della Mario will convince you of the importance conoscenza di se stessi^.^ of the knowledge of oneself. See Stowell (1983) for a siailar argument. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 2 5 Therefore, we can conclude that PRO is projected onto the syntax, at least in an optional way. The second problem related to PRO is to discern onto which position it is projected. Roeper (1983) argues that PRO occupies Spec-of-DP, if we adapt his proposal to the DP-Hypothesis. The fact that we cannot have a controlled reading of the purpose clause in (76) indicates that the internal argument and the subject of the eventive noun are competing for the same position. (76)*The ship's destruction in order to collect the insurance. Along the lines as Roeper, Giorgi and Longobardi point out that a prenominal possessive in Romance does not "obliterate" the external argument PRO in contrast with Germanic, as (77) illustrates. (77) Su hundimiento para cobrar el seguro. Its sinking in order to collect the insurance Giorgi and Longobardi argue that this difference is accounted by the Subject parameter. In Germanic the external argument is projected in the prenominal specifier position. Hence, the position is not free for the internal argument to be occupied, as proposed by Roeper (1983). On the other hand, the external argument R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 2 6 is projected to the right in Romance, allowing the prenominal position to be occupied by the internal argument. Stowell (1989) argues that PRO is projected onto the specifier of NP and not onto the specifier of DP, if we assume the DP Hypothesis (Abney, 1987). In (78) the verb L-marks DP, according to Chomsky (1986b), hence, it does govern the specifier position of DP. Since PRO cannot be governed in accordance with the PRO Theorem (Chomsky, 1981), PRO cannot occupy the specifier position of DP. (78) Este moviraiento filosôfico apoya el This philosophical movement supports the conocimiento de uno raismo. the knowledge of oneself. Stowell (1989) claimed that PRO is projected onto the Specifier of NP. V cannot govern PRO in that position since NP is an intrinsic barrier and DP is a barrier by inheritance, in accordance with Chomsky (1986b). This solution has the advantage that PRO occupies the natural position of the external argument within NP and allows us to rule out cases like (79a) and (79b), since the possessor 0-role is assigned to the Specifier of DP, and this position is governed by the verb disallowing the presence of PRO there (examples from Stowell, 1989:239). R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 2 7 (79) a.*Johrij bought [PROj(’s) book]. b.*Johnj bought [PRO; a/the shoe]. Valois (1991) proposes that PRO occupies an adjunct position, arguing against the Subject Parameter. This adjunct position is defined by the projection to which the noun moves head-to-head, i.e. NumP in Romance and NoP in English. In these positions PRO can remain ungoverned. The fact that control is not compatible with a prenominal genitive is derived from the Case Filter. Eventive nominals with internal arguments mapped onto the prenominal position are ruled out by the Case Filter, since the internal argument would recieve case by the noun and by the genitive morpheme. This can be the reason for the inexistence of an eventive reading but not for the supression of PRO. Control structures are allowed when the prenominal internal argument is quantified, as (80) illustrates. (80) Every city's destruction in order to win the war was condemned. Roeper, Giorgi and Longobardi, and Valois’ proposals cannot account for (80), since an internal argument in the prenominal position can co-occur with PRO under very specific circumstances, namely when the internal argument is quantified. We can propose that the internal argument R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 2 8 is mapped onto Spec-of-QP at the output of the lexical component, but why can’t we do the same thing with other affected internal arguments? A possible solution is that the quantificational interpretation of an affected internal argument is licensed by case assignment, a syntactic mechanism, hence, it cannot occupy the Spec-of- QP by the output of the lexical component, while the internal argument can occupy that position since it is introduced by a lexical quantifier expression. This hypothesis presents a problematic aspect. It supports the Head Subject Parameter. On the other hand, Stowell’s proposal works perfectly even in the case of (80). PRO is left on its canonical position, and another filter must account for the difference between (80) and (76). The nature of this filter needs more research, hut must he linked to the interaction of the Quantifier Raising rule and the interpretation of PRO. It is probable that the lexico- syntactic saturation of the noun in (76) by the affected object in Spec-of-DP induces the lexical saturation of the external argument, along the lines of Rizzi (1986). In (76) the noun is not allowed to project an arbitrary subject, because it is saturated. On the other hand, the noun is not saturated by the affected object at the output of the lexical level of representation in (80), R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 2 9 since the affected object needs to be interpreted at LF. Also, at LF the interpretation of the external argument PRO must be bound to the interpretation of the affected object. 2.6. The lexical representation of nominals 2.6.1. Introduction Following the models of grammar proposed by Zubizarreta (1987) and Levin and Rappaport (1988), in which there are two levels of lexical representations, a semantic and a syntactic one, the LCS is the level of representation where Nouns and Verbs share their Argument-structures. Thus, different syntactic rules, depending on the lexical category of the word, will apply in order to map the LCS onto the syntactic component of the grammar. For instance, the external argument of a verb will be projected onto the syntactic lexical structure by means of Linking Rules, as proposed by Zubizarreta (1987) and Levin and Rappaport (1988). As noticed by Zubizarreta (1987), Nouns fail to undergo Predication. Since they cannot function as Lexical- predicates, they lack a representation at the syntactic lexical structure. Rappaport's thematic constancy principle will account for their insertion in the syntactic level without a syntactic lexical structure. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 3 0 What remains is to determine which principles act in those kinds of predicates that show asymmetries between the argument structures of verbs and nouns, such as psych-predicates, as noticed by Amiratavalli (1980), Rappaport (1983), and Zubizarreta (1987), and whether or not there are some linking rules that could apply to the selection of genitive arguments. I will start by reviewing past proposals for defining the lexical structure of verbs in order to study their feasibility in their analysis of nominal lexical structure. 2.6.2. The lexical-syntax interface Aktionsart and its interaction with syntax is one of the aspects of syntax that linguists have started to develop within the generative framework in recent years. In previous studies we can find three different types of analyses within the generative framework that are concerned with the mapping of arguments onto syntactic structure. These hypotheses try to formalize the most general hypothesis stated by Perlmutter and Postal (1984) in (81), within the framework of Relational Grammar, and by Baker (1988) in (82), within the framework of Government and Binding. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 1 3 1 (81) Universal Alignment Hypothesis There exist principles of Universal Grammar which predict the initial relation borne by each nominal in a given clause from the meaning of the clause. (82) Universal Theta Assignment Hypothesis Identical thematic relationships between items are represented by identical structural relationships between those items at the level of D-structure. This general hypothesis establishes a general framework where the objective would be to determine what aspects of meaning are relevant for the syntactic mapping of arguments. As we said above, we can identify three trends in current theoretical literature. The first one tries to determine which thematic relationships are relevant for this mapping. Under this hypothesis, the mapping of argument structure onto the syntax would be through a set of linking rules, such as the ones proposed by Hale and Keyser (1986) in (83). (83) Unmarked theta assignment Conventions: a. The theme role is assigned to the object grammatical function. b. The agent role is assigned to the subject grammatical function. As is well-known one of the main problems for the Universal Theta Assignment Hypothesis results from the psych-verbs case. Although it would give excellent R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 3 2 results with the mapping of the possessive argument in the case of psych-nouns, since the experiencer argument would always be the one assigned genitive case, this is not the case with psych-verbs, where, in some cases, the experiencer is mapped onto the subject position, and the theme occupies that position in others. The second one posits that the mapping of argument structure onto the syntactic level of representation is mediated by aspectual constraints. This hypothesis, in its strongest version, is defined by Tenny (1987, 1988) in (84). (84) Aspectual Interface Hypothesis The mapping between cognitive structure and syntactic structure is governed by aspectual properties. Only the aspectual part of cognitive structure is visible to the syntax. Tenny’s hypothesis is based on aspectual classifications of predicates. Several classifications of aktionsart have been given in previous literature. The main differences among them are the relevant relations among the different types in linguistic groups. Vendler (1967) proposes the existence of four types of verbs : As pointed out by Mourelatos (1981), it would be better to use the term types or categories of verb predication, since a verb can have different values, depending on the context. For instance, the verb understand is related to three different interpretations: a state in ( i ) . R eproduced witti permission of ttie copyrigfit owner. Furtfier reproduction profiibited witfiout permission. 1 3 3 activities, accomplishments, achievements, and states. These four types have been considered the basic classes. However, the researchers that have studied them have grouped them in different ways, as we will see below. Activities are defined by their homogeneity. Abraham calls them monophasic since we cannot distinguished different phases in their realization or there is not a culmination. Phase 1 of running would be equal to phase 2, etc. We can use the present progressive at any time during the realization of the activity denoted by the verb and the sentence would be correct. The different parts of the process are equal to the process as a whole. On the other hand, accomplishments are not homogeneous. There is an activity that leads to the culmination of that activity. The phases that are part of an accomplishment are not all the same, as pointed out by an achievement in (ii), and an activity in (iii). (i ) He understands French. (ii) Please understand that I am only trying to help you! ( o f . Mourelatos, 1981:196) (iii) I’ m understanding more about quantum mechanics as each day goes by. (of. Comrie, 1976:36) These aspectual distinctions are marked in some languages through morphology, such as in Russian or in Greek, and not lexically as in English. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 3 4 Vendler (1967:101). If we write a letter, we have an activity that ends in a result, the letter. In accordance with Vendler (1967:103) achievements "occur at a single moment", in contrast with activities, states, and accomplishments that occur during a period of time. This "single moment" is indivisible, hence, it is homogeneous by definition. Achievements are denoted by perception verbs, among others.^ For instance, the moment that someone sees something is indivisible, i.e. it cannot be divided into phases. Finally, states are the most idyosincratic denotation possible for a verb, since their grammatical behavior, as we will see, is quite different from the other three groups. States are homogeneous, since all the parts that compose a state have the same nature as the whole. States, also, last for a period of time, in contrast with achievements. States are not actions, they are more related to qualities and relations, as pointed out by Vendler (1967:109), even when states may be caused or, as pointed out by Mourelatos (1981), when they may be the result of a change. Other examples given by Vendler (1967:107) are recognizing, realizing, spotting and identifying something. losing, finding, reaching, winning the race, crossing the border. starting, stopping, and resuming something, being born, and dying. R eproduced witli permission of ttie copyright owner. Further reproduction prohibited without permission. 1 3 5 Vendler relates achievements to states since achievement verbs cannot be formed in the present progressive, according to himStates and achievements are not processes, while accomplishments and activities are. Kenny (1963) proposes another classification. He put achievements and accomplishments together in one group: performances. Mourelatos (1981), who argues also for the integration of accomplishments and achievements in one "genus", points out that accomplishments and achievements are actions that have a culmination. Following some ideas outlined by Vendler in (85), Mourelatos (1981) proposes that accomplishments and achievements take definite time, while states and activities take an indefinite time. (85) For activities: A was running at time t means that time instant t is on a time stretch throughout which A was running. Mourelatos (1981) disagrees with Vendler's observation, and gives the following example in (ij. ( i ) He i s winning the race. However, I do not think that (i ) implies any achievement, ( i ) does not entail that, at some point in the future, we may say He won the race. Mourelatos is interpreting winning in (i ) as an accomplishment verb, ( i ) refers to the activity before the achievement, as reading does in (ii). (ii) He is reading a book. R eproduced witti permission of ttie copyrigtit owner. Furtiier reproduction profiibited witfiout permission. 1 3 6 For accomplishments: A was drawing a circle at t means that t is on the time strecht in which A drew that circle. For achievements: A won a race between tj and to means that the time instant at which A won that race is between tj and t2 . For states: A loved somebody from t| to t2 means that at any instant between t, and t2 A loved that person. (Vendler (1967:106)) Mourelatos (1981:201) gives the following classification in (86), and gives the following examples for each one in (87), (88), (89), and (90). (86) situations I I ! I states occurrences (actions) I I I I processes events (activities) (performances) I I I I developments/ punctual occurrences (accomplishments) (achievements) (87) The air smells of jasmine, (state) (88) It's snowing, (process) (89) The sun went down, (development) (90) He blinked, (punctual occurrence) R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 3 7 Mourelatos argues that states and processes are mass- quantified, while events are count-quantified, since they can be intrinsically counted. Mourelatos’ (1981) work is relevant to ours because he applies the mass-count distinction to the different categories of verb predication. Tenny uses the terms "delimited" for those predicates that have a definite time span, or that are count quantified in Mourelatos’ analysis, and "non delimited" for those predicates that have an indefinite time span or that are mass quantified. Tenny’s theory concentrates on the mapping of the internal argument. The internal argument is that one that measures out over time the event denoted by the predicate. In the case of events, it is the argument that delimits the definite time denoted by the event. We saw in section 2.3.2.2. some of the inconsistencies of this theory in relation to affected objects. The problems increase when Tenny analyzes unaffected internal objects or predicates that can be classified as activities. In an activity such as push the cart the internal argument does not measure out the time of the event but its location. A goal phrase can delimit the sentence as in push the cart to New York. Again, the location of the internal argument is measuring out the event. Therefore, in push the cart, we can say R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 3 8 that the internal argument, or one of Its properties, has the potential to measure out the event. Demonte (1992) makes a similar proposal claiming that aspectual features are implicit or potential, except where associated with a syntactic position. For instance, the potential accomplishment in (91) does not become an actual accomplishment unless the constituent between parentheses is present. (91) give a letter (to John) Tenny’s hypothesis poses important problems when attempting to account for the case of psych-verbs, and also when attempting to explain the uniform syntactic mapping of the arguments of psych-nouns, as we mentioned before. Tenny claims that when the e.xpeiciencer is mapped onto an internal position, it measures out the event, and when it is an external argument it does not, as the contrast in (92) shows. (92) a.*JohOj feared the truth into drinking^. b. The truth frightened John- into drinking-. However, notice that the delimiting expression in (93) can apply either to the external argument or to the internal argument. Thus, the test does not offer clear empirical evidence for Tenny’s claim. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 1 3 9 (93) Juan; admira a su hermana- hasta la médula:/- ’Juan admires his sister to the core.’ The other type of psych-verbs presents a similar case, as we can observe in (94) with the verb gustar ’to like’. (94) Su hermana; le gusta a Juan, hasta la His sister him likes to Juan to the Juan likes his sister to the core. médula,, • core The reason for the behavior of arguments projected by a psych-verb in relation to delimiting expressions lies in the fact that the underlying position for both arguments is the internal argument position, in accordance with Belletti and Rizzi’s (1988) proposal. However, the appeal of Tenny’s hypothesis is clear since it allows us to explain the syntactic mapping with unified principles of grammar and to constrain the relevance of the role of thematic theory in grammar. A possible solution to the psych-predicates puzzle will be studied in section 2.6.3., based on the lexical-syntactic differences between verbs and nouns. Since nouns map their arguments directly from the lexicon, with the possible exception of that argument marked with genitive case, it would be possible to define the lexical relations of psych-arguments in a nominal environment. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 4 0 The conclusion will be that Tenny’s hypothesis needs a refinement in the sense that we must incorporate sub event structure into it. The third hypothesis is based on the interaction between the aspectual dimension and the thematic hierarchy. The thematic hierarchy is specified in (95), and the aspectual hirarchy is specified in (96).^^ (95) (Agent(Experienc. (Goal/Source/Location(Theme) ) ) ) (96)(Cause(other(...))) The most prominent argument would be the most prominent role in the aspectual hierarchy, when there is a conflict, such as the frighten type of verbs. The aspectual dimension is derived from sub-event structure, based on the aspectual classes of verbs stated by Vendler (1967) and Dowty (1979) and the phasic structure given to them by Pustejovsky (1988). Grimshaw claims that we can divide events into one or more subevents. For instance, accomplishments would be divided into an activity sub event and a state sub-event, as in (97). (97) event / \ activity state 1 2 Grimshaw specifies only the role CAUSE for this tier. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 4 1 The most prominent argument in the aspectual dimension would be the one that is associated only with the first sub-event. For x breaks y. we have the activity of x breaking y, and the state in which y is broken. Thus, x only participates in the first subevent, being the most prominent argument in the aspectual dimension. Grimshaw (1990) pointed out that we are still in need of some stipulations when her analysis comes to the fear class of psych-verbs. The experiencer argument is the most prominent in the thematic hierarchy. By hypothesis, the experiencer role is the most prominent in the aspectual tier, as well. A closer analysis of this case in the next section will show that stipulations are not needed for the fear class of psych-verbs. In addition to this, the theoretical machinery needed in Grimshaw’s analysis is much bigger than the former analyses, since we have two types of principles acting, the aspectual and the thematic, and we need, as well, to posit some kind of argument fusion device, as does Jackendoff (1987), to articulate the organization of the thematic roles from the different tiers. Therefore, a syntactic organization based only on aspectual properties would have more explanatory power and would be theoretically more constrained. Again, a closer analysis of psych-verbs will R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 4 2 allow us to base our lexical analysis on aspectual properties only. 2.6.3. Psych-predicates The fact that NPs map their arguments onto the syntax through semantically unrestricted prepositions, in accordance with Rappaport (1983), would allow us to see in a better way the semantic relations borne by each argument projected by a psych-word. I would like to establish a typology of psych-nouns and to see whether this typology is related to the one established in the literature of psych-verbs. If my analysis is correct, we would be able to see in which way psych-verbs preserve the Universal Alignment Hypothesis. There are three major proposals that try to explain how the syntax of psych verbs preserves the Universal Alignment Hypothesis, as stated by Perlmutter and Postal (1984), in (81). Belletti and Rizzi (1988) propose that there are three different classes of psych-verbs considering their S-structures and the Case assigned to their Experiencer arguments as shown in (98). (98) CLASS i temere (fear): Nom Acc Exp Theme R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 4 3 CLASS II preoccupare (worry): Nom Acc Theme Exp CLASS III piacere (like): Nom Dat Theme Exp The three classes have the same D-structure. Belletti and Rizzi assume that the Experiencer argument always has a D-structure position higher than the Theme argument, although at this level it is not an external argument. The theme position is always the underlying direct object position. In the case of classes II and III, the theme argument is promoted, in order to fulfill Case requirements, to the subject position at S-structure, since the experiencer has been assigned accusative Case inherently in the case of the preoccupare class, and dative Case in the case of the piacere class. As we can see, they do not make any semantic distinction among the arguments projected by the three different classes of psych verbs. Grimshaw (1990) and Pesetsky (1988) point out that the syntactic differences of these verbs are in relation to their lexical conceptual structure. Grimshaw (1990) argues that verbs in class II and III do not have a derived subject, although they do lack an external argument. This approach is feasible since she does not equate the external argument with a D-structure subject. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 4 4 She proposes that there are two dimensions,^® maybe more, in the LCS: the thematic dimension and the aspectual dimension, as we have seen in (95) and (96). The external argument in her framework is the one that is most prominent in both dimensions. For instance, an agent is always the most prominent argument of a verb in both dimensions^® since she assumes that the subject of an agentive predicate and a cause have some property in common. Therefore, agents are always external arguments. As long as she treats the Theme in class II verbs, equivalent to her agentive frighten class, as an Agent, there is no problem to the mapping of the a-structure of these verbs into the syntactic representation. In the case of class III verbs, equivalent to her frighten psych verbs, there is no argument more prominent in both dimensions. In this case she stipulates that an argument that participates in the first sub-event, the aspectual dimension or "activity" is more prominent than an argument which participates in the second sub-event "state". Since the theme argument also bears the cause argument, this argument should be mapped onto the subject position at D-structure. These dimensions are similar to Jackendoff's (1987) tiers. Grimshaw claims that CLASS I verhs are agentive verbs. R eproduced witti permission of ttie copyrigtit owner. Furttier reproduction protiibited wittiout permission. 1 4 5 The third proposal that tries to preserve the UAH is Pesetsky's (1988) account based on his particular classification of thematic roles involved in "psych predications": (99) a. predicate (cause, experiencer) b. predicate (experiencer, target of emotion/subject matter) He has divided the theta-role Theme into Cause of emotion and Target of emotion. In my analysis. Class I in (98) is related to (99a) and (99b) and class II and III to (99a). The NP that bears the role of TARGET OF EMOTION in class I, bears also the role CAUSE. I will argue, based on a lexical analysis of psych nouns, that in Spanish you cannot derive (99a) from (99b) at any level of representation. On the other hand, in Spanish, as Franco (1990) has shown, we can derive the class II of psych verbs from the class III syntactically. Franco (1990) proposes that this derivation is possible through the incorporation of a class III verb into a causative node from which it can assign accusative case to the experiencer. Can we derive verbs of the first class from an abstract class of verbs related to class III in a model close to the one of Hale and Keyser (1993), since we assume that all Theta roles R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 4 6 in class III are present in class I? In Spanish we do not have any productive morphological evidence. Maybe, the only case that we should consider is (100). (100) Maria enamorô a Juan. Maria won the heart of Juan. In any case, I do not think the problems raised by these questions fall under the scope of this study. For the time being, one thing has been proposed, that is, that we can syntactically derive verbs of class II from those of class III in Spanish, as Franco (1990) points out. In this case we have widely productive morphological evidence. Thus, the difference between class II and III is not relevant for the lexical level of representation. Since my study is lexical in nature, from now on I will collapse class III into II. One further assumption before I go into the analysis of Spanish Psych Nouns. Since I believe that SUBJECT MATTER OF EMOTION is more related to CAUSE semantically than to TARGET OF EMOTION, I equate CAUSE with SUBJECT MATTER OF EMOTION. From now on, I will use these labels but notice that what has grammatical relevance are the prepositions that introduce each argument to refer each semantic type of argument, given the unstable definitions of these definitions, and that, as pointed out by Zubizarreta (1987), they do not have any grammatical significance. In a somewhat impressionistic way we can relate R eproduced witti permission of ttie copyrigtit owner. Furttier reproduction protiibited wittiout permission. 1 4 7 Seemingly, Spanish psych nouns do not show any difference in the syntactic representations of the arguments projected by the two lexical classes that I posit for this section, as (101) and (102)^^ show. (101) el amor/ odio/ temor/ adoraciôn de Juan por Juan’s love/ hate/ fear/ adoration of Maria. Maria (102) a. la preocupaciôn de Juan por Maria. Juan’s worry for Maria. b. el enojo de Juan por Maria. Juan’s anger at Maria. c. la molestia de Juan por Maria Juan’s annoyance at Maria. Nouns in (101) are related to Spanish verbs of Class I, and Nouns in (102) are related to those of class II and class III. In both cases, the genitive position de theta-marks and assigns case to the EXPERIENCER and the the following prepositions with these thematic roies. ( i ) a . ^ 'of EXPERIENCER b. gor CAUSE c . a 'to', back ’towards' TARGET OF EMOTION (ii) illustrates (la) and (ib), and (ill) illustrates (ia) and (ic). (11) La preocupaciôn ^ Pedro rot Maria. (ill) El amor ^ Pedro hacia Maria. diversion 'enjoyment' seems to be one of the possible exceptions. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 4 8 preposition por theta-marks and assigns case to the SUBJECT MATTER OF EMOTION. If we wished to maintain the thematic difference expressed in (99) between (101) and (102), we should find some differences between these two structures. Otherwise, we should consider that Belletti and Rizzi (1988) are right, and that there is no difference between the underlying structures of different classes of psych verbs. The paradigms in (103) and (104) show us this difference, when we use the preposition hacia 'towards' to introduce the SUBJECT MATTER OF EMOTION. (103) el amor/ odio/ temor/ adoraciôn de Juan the love/ hate/ fear/ adoration of Juan 'Juan's love/hate/fear/adoration of Maria' a/hacia Maria to/towards Maria (104)*la preocupaciôn/ enojo/ molestia de Juan the worry/ anger/ annoyance of Juan 'the worry of Juan for Maria' 'the anger/ annoyance of Juan at Maria' a/hacia Maria to/towards Maria In (101) and (102), we can observe that both classes of psych-nouns can mark their arguments with the prepositions ^ and por. while, in (103) and (104) the distributional analysis shows us that only nouns of the R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 1 4 9 lexical class I can use the prepositions ^ and hacia to map their arguments. These facts are expressed in (105). (105) a. Nouns of class I: (de: ________ (por: )/ (a/hacia: ) ) b. Nouns of class II: (de: ________ (por:_______ )) Hence, prepositions semantically restricted assign only one role from only one dimension or tier, or sub event, and are mapped directly from the LCS onto the syntactic level of representation. Predication and linking processes can subsume more than one theta role from different tiers under one syntactic position. These processes map the LCS information onto the PAS (Predicate argument structure). Nouns seems to skip this level of representation, as proposed by Zubizarreta (1987), although genitivization seems to work as a default linking rule but it is probably not connected to lexical or event structure. Nominals of the first class can express just one tier at a time as (105a) shows. Whereas verbs must subsume both tiers, given that these tiers affect Predication and the Linking Rules involved in objectivization. We probably need an Argument Fusion Principle close to the one proposed by Jackendoff (1987) , R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 5 0 in (106). This principle should explain the fusion among the lexical conceptual arguments and the positions of the subcategorization frame of a lexical item. (106) Argument Fusion Into each indexed constituent in the reading of a lexical item or preposition, fuse the reading of the syntactic constituent in the sentence that satisfies the coindexed position in the lexical item's subcategorization feature. Into the position indexed i in the reading of a verb, fuse the reading of the subject. Into the position indexed i in the reading of a noun, fuse the reading of a subject. We can use this mechanism, or just do the mapping directly from subevent structure, what would be preferable in Economy terms. In accordance with Grimshaw, the aspectual hirarchy has always more prominence than the thematic one. Hence, if the aspectual hierarchy does the correct predictions, we need to redefine the thematic hierarchy functions. Let us first see how the aspectual hierarchy works with psych-predicates. If we accept that in the frighten class the SUBJECT MATTER OF EMOTION role is mapped onto the subject position at the output of the syntactic component because it participates in the first subevent but not in the second, as proposed by Grimshaw, then we can link the preposition por to this subevent participation. Since Spanish psych-nouns do not show any other kind of R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 1 5 1 prepositional marking for this role, we can conclude that in the case of class II-III of psych-predicates the SUBJECT MATTER OF EMOTION role is linked only to the first subevent. On the other hand, class I of psych-nouns show two different prepositional markings, por and hacia. As we stated for the class II, por indicates a participation in the first sub-event, that we can identify with the cause of the state, hacia is related with the second subevent that is the state itself. In this case, the SUBJECT MATTER OF EMOTION role is not only the cause of the state but it has become also a TARGET OF EMOTION role. This TARGET OF EMOTION role is a participant of the second subevent, the state itself, and, actually, we can say that it is the last participant of the whole event, because it is the addressee of the emotion. Since it participates in the second subevent, it cannot qualify as the most prominent argument over the experiencer, and lexical structure is respected without deviations. We are assuming that subevent structure acts over lexical structure. Hence, there is already a structure formed, that is the nominal argument structure selected by the predicate. I assume that subevent structure only acts in verbal environments. The reason is that TEMPORAL ARGUMENT STRUCTURE (TAS) selects VPs but not NPs. TAS R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 5 2 selects an event, and triggers temporal relations among the subevents. The definition of the criteria that act in the determination of this structure needs further research. We can assume that lexical structure is defined by thematic information. At this level, for instance, experiencers are always more prominent than objects of experience, in accordance with Belletti and Rizzi (1988) . It is at this level that scope relations are encoded (Zubizarreta, 1992), so we can explain why the anaphor within the object of experience can be bound by the experiencer in (107). (107)?Each other's stories annoy the politicians. However, what is most relevant is to define the criteria for this thematic hierarchy. Why is the agent more prominent than the patient, or the experiencer more prominent than the object of experience? What is the grammatical principle that defines structural relations? I would like to suggest that the nature of argument structure is referential, and that it is defined by the relation of the argument to its predicate. Each argument specifies reference to the predicate. The one that is more relevant in referential terms, the internal argument, is the first one to be 0-marked. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 5 3 One question remains unanswered as yet. That is why the de-argument of class III is assigned dative case in the case of verbs. The por-argument is probably assigned nominative case because of the Default Linking Rule (Zubizarreta, 1987:15),given that there is not any argument whose prominence can qualify for the Core Linking Rule to apply. The remaining position in a universal hierarchy of cases is the dative one. The fact that Nouns related to class I of psych verbs show a different distribution in the mapping of their arguments to that of Nouns related to class II and III induces us to think that structures of Nouns and Verbs are related, and that both of them preserve the Universal Alignment Hypothesis. Predication, objectivization, and genitivization map semantic arguments through linking rules into a lexical syntactic level of representation. The rest of the arguments are Zubizarreta ( 1987: 15) states the following Liniiing Rules: (i) Core Linking Rales If a predicate P is projected onto a position H in the lexical frame, then A. Link the internal arg-variable of P to a position governed by H (Rule of Projection) B. Link the external arg-variable of P to the head of the lexical frame (Rule of L- Predicate Formation) Default Linking Rules C. If B does not apply, then copy the index of the agurment governed by the head of the lexical frame onto the head of the lexical frame. (Default Rule of L-Predicate Formation) R eproduced witti permission of ttie copyrigtit owner. Furttier reproduction protiibited wittiout permission. 1 5 4 semantically restricted by prepositions and do not suffer any change at the lexical syntactic level of representation, i.e. they are mapped directly from the Lexical Semantic level onto the Syntax. This fact allows us to define more precisely which roles are associated to the arguments of a psych-noun. Therefore, we are able to determine the different properties of the so-called Theme in the case of the different classes of psych-verbs. 2.6.4. An aspectual-referential theory of role assignment Tenny’s hypothesis points to the fact that the internal arguments are those that define the event on a scale. Agents cannot be internal arguments since they cannot measure out the event in any way. As she states: "an agent by its very nature is free to act in unspecified ways to effect something" (Tenny, 1988:12). Agents are not "necessarily consistent throughout an event." Therefore, they cannot measure out the event. This fact prevents agents from identifying an event in its referential aspect. On the other hand, an internal argument is able to identify an event, as we have seen in the case of affected objects, or, at least, to constrain the semantic type of event. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 5 5 My claim is that a role assignment analysis based on aspectual properties à la Tenny, is, in fact, an analysis based on the referential properties of arguments with respect to the predicate. In accordance with Stowell (1981), Zubizarreta (1985), and Zagona (1990), among others, I assume that lexical structure is associated with referential indices, and that argument referential indices define the referential properties of a predicate. An event is referentially defined by its aspectual properties in conjunction with its semantic content. Thus, all the arguments that can define the aspectual properties of an event will be more relevant in referential terms for that event. This fact has important syntactic consequences, since the more relevant an argument is in referential terms, the lower it will be mapped onto the tree. In other words, the more consubstantial an argument is to the predicate the lower it is mapped onto the syntactic tree. The argumentation is quite logical if we assume that external arguments are 0-marked in a compositional way. The most salient aspectual property of an event is the intrinsic result of that event. The argument that is able to assign a referential index to the intrinsic result denoted by the predicate will be the internal argument. This is what (108a) shows for an affected R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 5 6 internal argument and (108b) for an unaffected internal argument. Notice that the identificational term Barcelona in (108b) has a generic flavor, since unaffected internal arguments cannot saturate a predicate. (108) a. La construccion es San Pedro. The building is Saint Peter. b. La descripciôn es Barcelona. The description is Barcelona. This analysis has also some consequences in terms of the middle construction in English also. As observed by Jaeggli (1986) among others, the middle construction is possible when the verb projects an affected object. In the case of verbs we have the eventive variable and a temporal variable. The presence of a generic variable, as proposed by Zubizarreta (1987), has as a result that middle constructions do not depend on Infl in their temporal aspect. Hence, following Roberts (1985), we can consider them stative predicates. As stative predicates, the internal argument is able to saturate them, and the rule "Externalize Theme," proposed by Williams (cf. Roberts, 1985:358) can be applied. There are some events that do not have intrinsic results that can be referentially identified, such as destruction or push. In this case, the argument that defines the locative properties of the predicate will be R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 5 7 the internal argument, in accordance with Tenny’s proposal. States are inherent results. Again, the argument that best identifies the state is the argument mapped in the lower position, i.e. the object of experience, as (109a) illustrates for psych-nouns of class I, and (109b) for psych-nouns of class II-III. (109) a. El temor es la economia. The fear is the economy. b. La preocupaciôn es la economia. The worry is the economy. Nominal arguments respect the hierarchy in (95) in a strict way. The cognitive principles that define this hierarchy need more study, but the referential properties of the arguments that form this hierarchy with respect to the predicate seem to be on the right track for further investigation, as the case of affected objects indicates. Some asymmetries will appear between the projection of nominal arguments and verbal arguments. This is because the TIMES argument structure that selects VP, in accordance with Zagona (1990) activates event structure, as defined by Pustejovsky (1988), in the sense that, in the case of verbs, event structure occurs in a time slot. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 5 8 2.7. On the nature of ^ As we saw in Chapter 1, the Thematic Correspondence Hypothesis states that there is a strict correspondence between the mapping of nominal arguments onto their syntactic positions and the mapping of verbal arguments onto their syntactic positions internal to VP (Kuroda, 1988). This entails that at the Semantic level of representation we have already a structured list of the arguments projected by a lexical item, as proposed by Zubizarreta (1987) against Rappaport (1983). In Chapter 1, we saw that the Thematic Correspondence Hypothesis holds for transitive, ditransitive, unergative and unaccusative verbs and nouns. However, psych-verbs and psych-nouns show some asymmetries. To be more precise, the Hypothesis cannot be tested in the case of psych-predicates, since the object of experience role cannot be mapped as a possessive argument, as (110) illustrates for the two relevant types of psych-nouns. (110) a. la preocupaciôn por/*de la economia the worry for/about the economy ’the government’s worry for the economy.’ del gobierno. of-the government b. la admiraciôn por/*de su hermana de Juan the admiration for/of his sister of Juan ’Juan’s admiration of his sister.’ R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 5 9 This fact does not allow us to check whether the experiencer is mapped onto a higher position with respect to the object of experience or not, since the object of experience is always introduced by a preposition with semantic content, a minor predicate in Zubizarreta’s (1992) terms. Arguments within minor predicates always fall under the scope of the other arguments since they are not able to c-command these other arguments due to the presence of this preposition that blocks c-command. The only generalization possible is that minor predicates always fall under the scope of possessive arguments, as (111) and (112) demonstrate. (111) a. la preocupaciôn de cada padre; por su; hi jo ’the worry of every father about his son.’ b.*la preocupaciôn de suj padre por cada hijo; the worry of his father for every son. (112) a. la descripciôn de cada padre; por su; hijo the description of every father by his son. b.*la descripciôn de su; padre por cada hijo; the description of his father by every son. (Ill) illustrates the case of a psych-noun. A quantified experiencer, mapped onto the possessive function, can bind a pronoun modifying an object of experience, as (111a) shows. (111b) demonstrates that we cannot have the reverse situation, a quantified object of experience binding a pronoun that modifies the R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 6 0 experiencer. (112) shows a similar situation with an agentive noun. Again, the element mapped onto the possessive function, the theme, has scope over the element mapped onto a minor predicate, the agent. There are two possibilities, either the possessive element and the minor predicate are sisters or the possessive element c-commands the minor predicate. Picallo (1991), following Belletti and Rizzi’s (1988) analysis of psych verbs, argues that the psych nouns related to the first class of psych-verbs assign the experiencer role to an external argument position, while verbs of the second type assign that role to the internal argument. She bases her proposal on the fact that psych nouns of the first type can be modified by referential adjectives, whereas psych nouns of the second type cannot, as (113) and (114) illustrate (examples from Picallo 1991:288). (113) a. el temor polonès a l’ocupaciô del territori the Polish fear of the occupation of the nacional national territory b. la (proverbial) admiraciô anglesa per la the (proverbial) English admiration for puntualitat punctuality (114) a.*1’atracciô catalana per les questions de the Catalan attraction for the language R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 6 1 llengua questions b.*la commociô sueca amb 1'assassinat de the Swedish commotion with the Palme assassination of Palme c.*??la preocupacio americana per I'obesitat the American worry for (with) obesity d.*??1’interès suis per la rellotgeria the Swiss interest for clockmaking However, the Spanish counterparts to (114) are well- formed, as (115) shows. (115) a. la atracciôn catalana por las cuestiones de la lengua b. la conmociôn sueca por el asesinato de Palme c. la preocupaciôn americana por la obesidad d. el interés suizo por la relojeria As we saw in section 2.6.3. the different psych- predicates have a straight thematic mapping in the case of nouns. The only difference is the ability of the object of experience in class I to be introduced by the preposition hacia 'towards'. On the other hand, the experiencer is always mapped onto the possessive function, regardless of the type of psych-predicate. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 6 2 Of-insertion has traditionally been considered a mechanism to avoid a violation of the case filter in (116), as argued by Chomsky (1981:49). (116) *NP if NP has phonetic content and has no case Therefore, only empty categories, such as traces, PRO, and pro, are not subject to the case filter. Stowell (1981:127) defines the environments where the of- insertion rule applies, in (117). (117) In the environment ...b...], adjoin of to b, where (i) a is some projection of [+N] (ii) b is an inmediate constituent of a, and (iii) for some c, c the head of a, c precedes b. Genitive case is assigned inherently, as claimed by Chomsky ( 1986a : 186-204 ) . Chomsky distinguishes structural case marking, which depends only on government, from inherent case marking in terms of ©-marking. In order for an element to receive inherent case from a head, it must be ©-marked by that head, as expressed in (118). (118) Inherent case condition If A is an inherent case assigner, then A assigns case to an NP if and only if A theta- marks the NP. (Chomsky, 1986a:194) R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 6 3 We have seen that the possessive function can cover different notions across the different types of nouns, such as THEME, AGENT, OWNER, and EXPERIENCER. We have seen also that when the noun projects arguments, the ones mapped onto the possessive function are the THEME, if the noun denotes an agentive event, and the EXPERIENCER, if the noun denotes a state. Hence, the assignment of the possessive function does not obey structural criteria, because in that case the THEME is considered to be lower in the tree than the AGENT, and the OBJECT OF EXPERIENCE is lower than the EXPERIENCER. We must point out that the OBJECT OF EXPERIENCE cannot be mapped onto the possessive function in any case, as (119a) demonstrates for psych- nouns of class I, and (119b) for psych-nouns of class II- III. (119) a.*la admiraciôn de la mûsica de Juan. Juan’s admiration for the music. b.*la molestia de la mûsica de Juan Juan’s annoyance at the music. It is interesting that this correlates with the distribution of referential adjectives. Since Kayne (1984), it has been considered that referential adjectives can only encode a theta-role not syntactically linked, such as an agent, as we see in (120). R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 1 6 4 (120) la destrucciôn nazi de Guernica the nazi destruction of Guernica However, this is not completely correct. We can have syntactically linked 0-roles expressed by referential adjectives. Picallo (1991:287) points out the following example given by Grimshaw (1990). (121) la derrota rusa the Russian defeat In fact there are other examples as (122) shows. (122) a. la persecuciôn hebraica the Hebraic persecution b. la descripciôn lingiiistica the linguistic description c. la creaciôn poética the poetic creation Furthermore, the THEME argument projected by ergative nouns, in the sense of Burzio (1986), can be expressed by a referential adjective in Spanish, as (123) illustrates. Giorgi and Longobardi (1991:48-49), for Italian, and Picallo (1991:287), for Catalan, have pointed out that this is not possible. Giorgi and Longobardi give the examples in ( i). (i ) a.'l'entrata governativa in parlamento the governmental entrance in the parlamient b.?*l’ appariozione tedesca di fronte alia linea Haignot the German appereance in front of the Maignot line R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 6 5 (123) la desapariciôn espanola de la escena polîtica the Spanish disapperance from the political scene Interestingly enough, those theta-roles that cannot be expressed by a possessive cannot be expressed by a referential adjective either. This is the case with the OBJECT OF EXPERIENCE in (119). As we can observe in (124), these objects of experience cannot be expressed by a referential adjective. (124) a.*la admiraciôn musical de Juan, Juan’s musical admiration b.*la molestia musical de Juan Juan’s musical annoyance Referential adjectives are generic, hence, they cannot denote specific members of their class. This is the reason why (ia) is not well-formed. On the other hand, the Spanish counterpart to (ib) is well-formed, as (ii) illustrates. (ii) La aparicion alemana en Trente de la linea Haignoi Picallo's examples are in (iii). (iii) a.*la tornada matrimonial, a Paris the matrimonial coming back to Paris b.*la desapariciô austrohongaresa de l'escenari politic the Austro-Hungarian disappereance of the political scene (ilia) has the same kind of problem as (ia) has. The intentional reading of the referential adjective is not generic, but specific. The Spanish counterpart to (iiib) i s correct as (123) shows. R eproduced witti permission of ttie copyrigfit owner. Furtfier reproduction profiibited witfiout permission. 1 6 6 The generalization that can be drawn from these facts is that a role projected by a nominal can only be expressed by a referential adjective if it can also be mapped onto the possessive function as expressed in (125) . (125) The same constraints that apply to insertion of ^ in Spanish, as a genitive case marker, apply to referential adjectives. What is the nature of this constraint? I would like to argue that it is semantic. Bull (1965:226-227) and King (1992:207) explain the variation of word order between the adjective and the noun by a semantic analysis based on the notion of partitivity. The sequence adjective-noun, as in blancas nubes 'white clouds’, is defined by a [- partitive] relationship between the adjective and the noun. The adjective does not denote a subset of a greater whole denoted by the noun. On the other hand, the sequence noun-adjective, as in nubes blancas. is defined by a [ + partitive] relation, blancas defines a subset within the set of clouds. I claim that this is the case also for the possessive function. Possessors, themes, agents, and experiencers enter into a partitive relation with the head noun. They point out a subset or a token from a larger set of events, states, or objects. For instance. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 6 7 in (126) the possessive restricts the kind of pictures to those of Barcelona. (126) una fotografia de Barcelona a picture of Barcelona Prima facie, objects of experience hold this partitive relation with the head noun. The love for one person does not have to be the same as the love for another person. The love of a child is different to the love of a mother. Or the annoyance caused by a noise in a conversation does not have to be the same as the one caused by a silence in a conversation. The difference is that an object of experience does not have the potential to referentially saturate the head noun. A state-noun modified by an object of experience will always denote a type of state but not a definite occurrence of that state. If we say el amor por Maria 'the love of Mary’, the phrase would always denote a type of state, never an occurrence of that state. We need another element to denote a definite occurrence of that state, such as an experiencer, el amor por Marla de Juan 'Juan’s love of Mary', or a demonstrative este amor por Maria 'this love for Mary'. Notice the contrast with, for instance, la fotografia de Barcelona 'the picture of Barcelona', that denotes a token. Or the case of an unaffected object, el R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 6 8 estudio de este caso ’the study of this case’, that can denote a type or a token. Therefore, the processes of assignment of genitive case and adjectivization^^ are defined by the notion of partitivity and the notion of potential referentiality,^ as in (127). (127) Genitive case is assigned to a nominal complement when there is a potentially referential partitive relationship between them. We should look at some apparent exceptions, noticed in previous literature (Cinque, 1980; Godard, 1992; among others). There may be some objects of experience that can Notice, for instance, that in el amor naterno 'the motherly love', the adjective refers to the experiencer not to the object of experience. The notion of potential referentiality also covers the cases of the so-called bound pronouns, such as in ( i). ( i ) Nadie- vio su, fotografia. 'Nobody saw his picture.' In this case, it can be argued that the relationship between the possessives and the head is not referential. In Martin (1991:233), I argue that in this case the possessive is not a pronoun but an anaphor. Following Chomsky (1981), I consider anaphors as potentially referential elements. In chapter 4 I specify that this relationship must be reciprocal to cover material, locative, and time complements. In la silla de esta madera 'the chair of this wood', the wood i s generic with respect to the head noun. R eproduced witti permission of ttie copyrigfit owner. Furtfier reproduction profiibited witfiout permission. 1 6 9 be introduced by the preposition such as in the case of el deseo de ti ’the desire of you’. Cinque (1980:78) has shown that possessivization is not possible for an object of experience, even when it can be introduced by the preposition de/di. (128) illustrates this point for Spanish. (128)*tu deseo your desire ’the desire of you’ Cinque (1980:81) also notices that extraction is impossible in these cases, as (129) illustrates for Spanish. (129) Ana, [de qui en.] hemos recordado el deseo ’Ana, of whom we recalled the desire ...’ Grimshaw’s (1990:99-101) proposal for sentential modifiers of result nominals such as conclusion, belief. hypothesis. or proposal. is applicable here. The preposition in these cases does not assign genitive case. It is an appositive preposition that identifies the content of the head noun. In this case, deseo does not denote the state but the state as a result. The modifer does not hold any partitive relationship with respect to the noun, since it is identifying the noun itself, as the R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 7 0 fact that it can occur across copula shows in (109). In addition to this, the presence of ^ does not always entail the assignment of genitive case, even if it is not appositive. As we will see in chapter 4, in the case of material, time, and locative complements, the preposition de only theta-marks the complement, without assigning case. We can conclude that genitive case assigment in Spanish is restricted to those elements that hold a [+ partitive] relation with respect to the head noun and have the potential to referentially saturate it. The close relation between possessives and determiners results from the fact that the determiner system computerized that partitive relationship between a noun and a related element and the referential properties of that partitive relationship. In the case of affected objects, the noun is always introduced by a definite determiner, since the compound noun+affected object cannot be partitioned. Following King (1992), among others, I assume that definite determiners denote a [- partitive] noun, i.e. a noun that does not enter in a partitive relation with a larger set, while indefinite determiners denote a [+ partitive] noun, i.e. a noun that enters in a partitive relation with a larger set. Other R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 7 1 relations with other roles are expressed through minor predicates, i.e. prepositions with semantic content. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 7 2 Chapter 3 Agreement within DP: the Double Articulation of DP and the "mise en abyme" Structures 3.1.Introduction One of the most important issues related to Romance nominal phrases is the definition of the algorithm accounting for agreement among the different constituents found within DP. Articles, ordinal numerals, demonstrative elements, possessives, and prenorainal and postnominal adjectives agree in gender and number in Romance languages, as shown in (1) for Spanish. (1) Las preciosas casas blancas. The beautiful white houses. There have been some attempts to explain the agreement mechanism for some of the constituents within DPs, such as Valois (1991) for French, but there is no cross-categorial unified account for it as far as I know. This chapter is an attempt to arrive at a unified account under the Principles and Parameters framework and the Minimalist Program (Chomsky, 1993). R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 1 7 3 Within the Principles and Parameters framework, there has been a strong tendency to consider morphological categories as syntactic heads, especially after Baker (1988) and Pollock’s (1989) works, in the spirit of Barriers (Chomsky, 1986b). Chomsky’s (1986b) proposal unifies X-bar theory for lexical projections and functional projections. There is no doubt that the incorporation cases, studied by Baker, and the verbal morphology, investigated by Pollock, show syntactic effects, and therefore are subject to syntactic principles such as the Head Movement Constraint (Travis, 1984). This line of investigation has been extended to include other morphological elements as heads of functional projections, resulting in a more articulated framework. In particular, some of the features have been chosen for this function in recent literature, such as Lamontagne and Travis (1986) for case, Ritter (1990) for number, and Picallo (1991) for Gender.Within a Chomsky's proposal must be included in this list if we consider his proposed AGR^P and AGRjP for the verbal system as pure case projections. In particular, AGR^F can be subsumed under ASPP (see Franco and Landa’s 1992 paper for further discussion), and AGR^P under TP (see Pollock, 1989; Belletti, 1990; Chomsky, 1991; Kayne, 1991; Rizzi, 1991; Roberts, 1992; for further discussion). My analysis, minimalist in spirit, departs from Chomsky (1993) on the conceptualization of the Checking function in syntax, as 1 will argue below. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 7 4 theory of Functional Categories (PCs) such as the one proposed by Hudson (1989), where PCs are licensed by -0 their faculty to saturate the non-overt argument of the lexical head that they select, we should be careful to consider the theoretical ground for those proposals. The other aspect to be taken into account is the definition of the syntactic import of these categories. Clearly, we must leave aside grammatical gender in Romance languages since it does not apparently saturate the noun, and in most cases the assignment of gender is arbitrary.On the other hand, tense, for Event is the cover argument of Vs (Higginbotham, 1985), and Reference is the cover argument of Ns (Williams, 1981). Complex event nominals probably project the two of them (see Chapter 2) . ® Hudson based his concept of saturation on Higginbotham (1985). However, in Hudson's hypothesis. Functional Categories are the ones responsible for the saturation of the predicates, whilst for Higginbotham this role is borne by thematic relations. As we have seen in Chapter 2 , an argument does not always saturate a predicate, hence, I will follow Hudson’ s hypothesis. Although it would be interesting to study if arbitrariness has any connection with intensionality, or if gender can be used in an intensional way. For instance, in Spanish el mar 'the (masc.) sea' is usually introduced by a masculine determiner, but the sailors usually say la mar 'the ( f e r n . ) sea' with a clear affective connotation. Further research is needed, but the fact that other intensional particles, such as affective suffixes, work in a different morpho- syntactic way, for instance, they do not show agreement with other constituents of a nominal R eproduced witti permission of ttie copyrigtit owner. Furttier reproduction protiibited wittiout permission. 1 7 5 instance, does saturate the event variable of the verb in some way. Gender in the major Romance languages seems to be the reflection of other lexical and syntactic processes, such as word-marking and noun movement. In particular, it appears to have a very similar function to the person feature in the Romance verbs, in the sense that, while the person feature seems to trigger verb movement in Romance, Gender triggers noun movement within the nominal expressions. Napoli (1989) claims that gender-number agreement entails a predication relationship. Ritter (1993) states that there is not a one-to-one relation between morphological features and functional projections. In Ritter’s hypothesis, while number heads a functional head, gender is realized on one of the nominal syntactic heads. The choice of the syntactic head for gender to be realized is subject to crosslinguistic variation. Based on the manipulation of gender as a derivational strategy in Hebrew, Ritter proposes that gender appears attached to the noun at the lexical level of representation in Hebrew. In Hebrew, the derivation of expression induces me to think that this is the la mar case is a specific case, and we can't obviously draw any generalization from it . See also Chomsky (1993:9) R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 1 7 6 a related noun from a masculine one by attaching the R2 feminine ending is quite productive. Ritter also states that the gender feature is attached in the Num° node to Romance nouns. There is not conclusive evidence to state this, especially if we consider, for instance, that some nominal affixes are inherently feminine, such as -dad or -cion. The gender marker appears to be a cyclic morpheme that can be attached at different levels. For instance, television ’television’ is feminine since it ends in the derivational nominal affix -sion. If we attach an affective affix, the Spanish feminine marker -a appears, as in television-cill-a ’little television’. Number is a much more complicated issue. First we should distinguish between Numeral Phrases, cardinal and ordinal, which have a syntactic import, since they head their own projections, and Number, which does not. Following Chomsky (1993), I will consider the ^-features as part of the derivational morphology. They appear as part of X°s. The only syntactic import that they have is to trigger overt movements or not, i.e., to decide at which level of representation movements take place. A language with overt morphology will have to do Checking For instance, we can derive from the masculine noun magav 'wiper', magev-et 'towel', by adding the feminine ending -el. R eproduced witti permission of ttie copyrigtit owner. Furttier reproduction protiibited wittiout permission. 1 7 7 in the computational system of the syntax. A language with ^-features but lacking overt morphology will have to do more operations at LF. A language lacking overt morphology and some or all of the 0-features related to Domain D will have to do more operations in the L e x i c o n . In accordance with Chomsky (1993), derivational morphology is a small programmer of the computational system of the syntax and LF.However, the motivation for lexical, syntactic, and logical operations is based on quantification in order to obtain referentiality. 0-features are the pure reflection of the way in which these operations occur. In accordance with the general principles of Chomsky (1993), I claim that agreement within DPs occurs under the Spec-head agreement relation, as specified in (2). What I mean by overt morphology is grammaticalized morphology, ^-features not related in any way to the Domain D (Chomsky, 1981), as it could be grammatical gender. A possible example is the sino-japanese character that is prefixed t o some words to define their prototype, in accordance with Posser’ s (1991) analysis. An interesting case is Walloon, studied by Bernstein. The noun is mapped onto the syntax without the plural suffix, that is borne by the adjectives. There is no noun movement attested to in Walloon in the overt syntax, hence, in accordance with the theoretical framework of Chomsky (1993) it must move at LF. R eproduced witli permission of ttie copyrigfit owner. Furtfier reproduction profiibited witfiout permission. 1 7 8 (2) Spec-Head Agreement Hypothesis Agreement is found only if there is co-indexing between a controller in [Spec, XP] and a target which is the head, X, of XP. (Georgopoulus, 1991) My analysis is based on the following proposals: a.) What we have traditionally understood by an NP is a string of restrictors that narrow down a referent from a type to a token when possible, as we go up in the syntactic tree. b.) A more articulated version of the DP-hypothesis is needed, as the one proposed by Cornilescu (1992) for Romanian. To start with, I will propose that there are two projections of DP, DetP that selects NP, and DP that selects the adjective-like elements, as Demonstrative Phrase, Genitive Phrase, and AdjP. However if we look at the Romanian data the situation could be more complicated than this, with a determiner phrase for each category (DegP in the case of adjectives). c.) AdjP is headed by an A° and selects an NP (Abney, 1987). d.) Intensional adjectives are mapped onto the specifier position of NP. e.) AdjPs may be selected by DegP (Abney, 1987; Bowers, 1988; Bosque, 1994). In that position, they act as operators of the noun, and intensional predicate calculus takes place. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 7 9 f.) Epithets, or intensifiers, move to the head of DegP, after feature checking with the noun, and, from there, to Spec-of-DegP. In that position, they act as operators of the noun, and intensificational predicate calculus takes place. g.) Movement to Spec is triggered by Checking agreement (Chomsky, 1 9 9 3 ). h.) Extraposition is triggered by the alteration of the functions of an element in the Determination Scale. The extraposed element receives contrastive stress. 1 will start by presenting the former proposals on agreement within NPs. In section 3, 1 will study adjective placement. In section 4, 1 will present former analyses based on N°-movement and the contexts where N- to-Det is possible. Finally, 1 will present ray proposal in section 5. In UG terms this movement probably Involves some kind of presuppositional effects. Overt agreement will trigger this movement in the computational system of the syntax. Lack of this morphological agreement forces movement to Specifier to occur at LF. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 8 0 3.2. Previous Studies of Nominal Agreement 3.2.1. Valois (1991) Valois (1991) proposes two different mechanisms to account for agreement within DPs, one for the Determiner, and another one for adjectives. According to Valois (1991:87), Determiners agree with nouns through Spec-head agreement in Number Phrase. Valois argued that DPs are generated in Spec-of-NumP, and, later, cliticize onto D°, in a fashion similar to Sportiche’s (1990) analysis of clitic pronouns in French. (3) illustrates Valois’ analysis. (3) DP / \ D’ / \ D. NumP / \ DP Num’ I I Di In addition to a question of legitimacy of the movement in terms of the HMC, this clitic movement would have difficulties in accounting for a case like (4). (4) Las dos casas. The two houses. For a different mechanism of the movement hypothesis for clitics in French, see Kayne (1975). R eproduced witfi permission of ttie copyrigfit owner. Furtfier reproduction profiibited witfiout permission. 1 8 1 In (4), it is not clear which one is the position of the numeral under Valois’ analysis, since the Specifier of NumP is occupied by the trace of the Determiner and Num° is occupied by the noun that has moved there from N“. Other questions of a conceptual nature arise such as the binding status of the determiner trace in terms of the HMC (Travis, 1994) and the unambiguous path condition for anaphors (Kayne, 1984). Valois (1991) extends this analysis to the possessive pronouns with some refinements. Possessive pronouns are generated in Spec-of-NP and moved to Spec- f i f i of-NumP. In addition to the problems presented by the analysis for Determiners, we face an undesirable situation for the Italian case, if we want to give a unified account of the nominal system in Romance. In Italian the Determiner cooccurs with the possessive, as (5) illustrates. (5) La sua casa. The his house ’His house’ If we leave aside the problems that we have seen above and consider la an expletive in Spec-of-DP we can save See, for instance Zubizarreta (1987) and Godard (1992) for reasons against this generation. R eproduced witti permission of ttie copyrigtit owner. Furttier reproduction protiibited wittiout permission. 1 8 2 Valois’ analysis for Italian, we can posit possessive incorporation into the head of a DP. It is true that possessives have a behavior very similar to a Determiner, but as the Italian case demonstrates, they are not exactly of the same category. This complicates the legitimacy of head movement from Spec-to-head, where this head has a projection different of the one initially projected by the element moved, and there is not an unambiguous path between both of them. In the case of Spanish and French I will assume a similar analysis to Vergnaud and Zubizarreta’s (1992), based on the rule of morphological conflation Suppletion, as in (6a), although, in the Spanish case, I propose that the correct derivation is as in (6b), where the pronoun incorporates into the determiner position, that I claim to be Spec-of-DP, following Hudson (1989). (6) a. Ipp pronoun [p. det]] --> suppleted form b. [pp det [p. pronoun]] --> --> [pp suppleted form [p, t] ] I claim that pronominal possessives are clitics along the lines of Rivero (1986) for Spanish, and Authier go The French case needs further research. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 8 3 (1992) for French, and, therefore, they identify a pro.^ In the case of adjectives of event nouns, Valois (1991:171-173) proposes that the Noun after raising to Num° binds a PRO in the external argument position of AdjP, assuming Stowell’s (1983) analysis for adjective phrases. Under this analysis AdjPs are small clauses that project external arguments as in (7). (7) [^p PRO adjective]] This proposal entails a violation of the i-within-i 71 condition, since Valois hypothesis entails that the antecedent for PRO is the noun. Hence, a head would be binding an element within its domain. This proposal cannot become a unified account for other kinds of adjectives, such as prenominal adjectives, since the noun does not c-command them by the output of the syntax, as pointed out by Bernstein (1993:95). 7 0 As pointed out to me by Paredes (p.c.) some varieties of Andean Spanish show possessive doubling 7 1 Chomsky (1981:211-212) proposes the following filter to avoid circularity in reference. ( i ) The i-within-i Filter * [ a i • • • B j . . . 1 R eproduced witti permission of ttie copyrigtit owner. Furttier reproduction protiibited wittiout permission. 1 8 4 In conclusion, Valois proposes two mechanisms to explain agreement within DP. In the case of the determiner and the noun, agreement would be explained by Spec-head agreement, and in the case of the adjectives of event nouns by percolation of noun features. 3.2.2. Bernstein (1993) Although Bernstein (1993:93-95) does not defend a particular proposal in her dissertation, I will briefly examine the possibility that she discusses. She generalizes checking theory (Chomsky, 1993) to the nominal expressions by saying that nominal constituents come out from the Lexicon with their inflectional 7? features inserted. N°-raising to Num° is triggered by checking features with the adjective in the adjunct position of NumP. The problem here is that this is not the right configuration for Checking Agreement. The right configuration which is proposed by Chomsky (1993:9-10) is Spec-head. Cinque’s (1993) hypothesis that adjectives are Specifiers would be a better solution in this case. My hypothesis overcomes these problems by recursive movement of a nominal projection to the Specifier of AdjP, GenP, DemP, and QPs. Whenever one of these elements appears See also Bernstein (1933:248-249). R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 8 5 prenominally in Spanish, it moves head-to-head to the functional category that selects it, and later to the Spec position of that functional category. In the case of intensional adjectives, there is also Spec-head agreement, since the adjective occupies the Spec-of-NP position. 3.3. Adjective Placement 3.3.1. Introduction In the major Romance languages adjectives appear prenominally and postnominally. This fact has made it very difficult to achieve a unified analysis of adjectives in the past. Bello (1928:12) points out that the prenominal adjective explains while the postnominal adjective specifies. Salvâ (1949:119) observes that the prenominal adjective refers to an inherent property of the noun, while the postnominal one refers to an accidental quality. Lenz (1925:179) noted that the prenominal adjective has a subjective quality and gives an affective value to the noun, while the postnominal adjective specifies logically the noun. The canonical order in Romance is noun-adjective as (8) shows, where the meaning of the adjective intersects 70 Walloon is an interesting exception t o this generalization, as we pointed out before. R eproduced witti permission of ttie copyrigtit owner. Furttier reproduction protiibited wittiout permission. 1 8 6 with the meaning of the noun to create a lexico-semantic sub-set. (8) a. El coche azul. The car blue. b.?*El azul coche, The blue car ’The blue car’ When an adjective appears prenominally, it has been pointed out that it has a nonrestrictive interpretation, as we have said above (Bello, 1928; Lujan, 1980; Bernstein, 1993). In this case the adjective is not delimiting a group of the objects denoted by the noun, as in (9). (9) el simpatico cliente the nice customer However, semantic distinctions do not correlate easily with syntactic positions as pointed out by Bolinger (1972), Lujan (1980), Cinque (1993), Sanchez ( 1994a),among others. Sanchez observes that the traditional semantic classes (Sussex, 1974; Demonte, 1982) are not related to only one syntactic position. Modal ( mismo, ’ self’, simple, ’ simple’) adjectives and predicative [ nuevo, ’ new’, pobre, poor) adjectives can occupy the prenominal and the postnominal position, while relational l aleman, ’ German’, lingûistico, linguistic) adjectives only the postnominal. I argue that this distribution correlates with the distribution of the R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 8 7 As pointed out by Bolinger (1972) there is also a restrictive reading of prenominal adjectives and a nonrestrictive reading of postnominal adjectives, but they have a different stress pattern from postnominal restrictive adjectives. Bolinger (1972) and Lujan (1980:87-89) show that a restrictive adjective is possible in prenominal position with a contrastive value and contrastive stress, as in (10). (10) Hablo de las HERMOSAS casas, no de las feas. 'I am talking about the beautiful houses, not about the ugly ones.’ Bolinger (1972) notices also that some postnominal adjectives have an appositive reading as in (11).^^ (11) No perder el tiempo en calaveradas estûpidas, ’Don’t waste time with foolish stunts.’ In the same way, we can argue that when we have a nonrestrictive adjective in a postnominal position, the modification of adjectives by Degree words. If tbe word co-occurs easily with a degree word, then it can raise to DegP, given its quantificational properties. No special distinction is needed between tbese two kinds of adjectives. ^ Lujan (1980:83-84) argues t h a t t h e a djective i n ( 1 1 ) d oes n ot have an unambiguously appositive r e a d i n g , b u t th e point here i s t hat an appositive r e a d i n g i s possible. I will not analyze Lujan’ s transformationalist analysis o f adjectives. For r e a s o n s against a transformationalist analysis see Bolinger ( 1 9 6 7 ) . R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 8 8 nuclear stress is borne by the noun or a restrictive adjective, as in (12a) and (12b). (12) a. Pedro tiene un COche rojo, Pedro has a red CAR. b. Vi un coche NUEvo rojo. I saw a new car that was red The table in (13) sums up this adjectival distribution. (13) pronominal postnominal 4 * $ * * * * nonrestrictive A N N A + * * * * * restrictive A N N A As we can observe, it appears that what makes the difference between the nonrestrictive reading and the restrictive reading is the secondary stress, leaving aside the contrastive or emphatic stress. 3.3.2. Adjectives and phrasal stress I will argue that, in addition to syntactic constituency, a theory of semantic saturation is needed to determine phrasal stress. Saturation will be defined in terms of the identification of the referential properties of a lexical head (Nouns, Verbs, Adjectives and Prepositions) by another element in the discourse or R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 8 9 in the sentence. In other words, when the meaning of a constituent depends on the meaning of another one, it is this other constituent that is going to bear the phrasal stress. (14) illustrates this point. (14) a. Juan vio un COche rojo. b. Juan vio un coche ROjo. ’John saw a red car.’ In (14a) Juan saw just one car, and this car was red. The value of the adjective is explicative. In (14b), Juan could have seen more than one car, but one of the cars was red. In this case the adjective specifies the noun. The two main proposals about phrasal stress do not cover these cases. The Nuclear Stress Rule (Chomsky & Halle, 1968) predicts that the phrasal stress falls in the rightmost constituent and Cinque’s (1990) Null Hypothesis predicts that the phrasal stress is assigned to the most embedded constituent. Neither one gives an account for (14a), taking these cases as contrastive stress, which is not subject to phrasal stress rules. However, the fact that restrictiveness is a requirement for both kinds of stresses may mean that there is a strong link between phrasal stress and contrastive stress. The Nuclear Stress Rule or the Null Hypothesis R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 9 0 may act as default rules when there is no contrastive stress. My hypothesis is that phrasal stress is the phonological representation of the ^-feature [+wh], where [+wh] is a variable for a non-overt referential argument projected by a lexical head. This non-overt referential element can be saturated under the syntactic relation of c-command. It follows, then, that when there is not an identifier of such a non-overt element, the most embedded element bears the phrasal stress, as proposed by Cinque (1990). We can attempt to collapse the focus stress (new information) and the presupposition stress (old information), distinguished by Cinque (1990). The presupposition articulation of the sentence gives us information already given in the discourse. The focus is the syntactic element that saturates this old information. This hypothesis could also account for the stress pattern of interrogative and relative pronouns, and for the so-called emphatic pronouns. Stress is a suprasegmental feature of the domain of the word, and phrasal stress is a suprasegmental feature of the syntactic constituent. Halle & Vergnaud (1987) state that three representational lines at most are necessary to assign stress to words, besides the one that R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 9 1 marks all the possible stress bearers. These three degrees for the representation of stress plus the setting of universal parameters would allow us to define the word-stress patterns of natural languages. These parameters are: (15) [±HT]: whether or not the head of the constituent is adjacent to one of the constituent boundaries. (16) [±BND|: whether or not the head of the constituent is separated from its constituent boundaries by no more than one intervening element. (17) [+HT] constituents are left/right-headed. (Halle and Vergnaud, 1987:10) Spanish is binary left-headed at the foot level and unbounded right-headed at the word level. When words form syntactic constituents, their stress representation is not affected in an important way. One of the word stresses gets more prominence. In Spanish, as in English, the rightmost element is the one that receives the most prominent stress under Chomsky & Halle's (1968) Nuclear Stress Rule, incorporated by Halle & Vergnaud (1987): (18) Nuclear Stress Rule a. Parameter settings on line N (N à 3) are [- END, +HT, right] b. Interpret boundaries of syntactic constituents composed of two or more stressed words as metrical boundaries. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 9 2 c. Locate the heads of line N constituents on line N+1. (Halle & Vergnaud, 1987:264) This rule would add new lines to the metrical grid without changing the representation in former lines, with the exception of some phenomena in some languages like the cases of Stress Shift in English. The parameter settings for Spanish are the same ones that operate in English, (18a). (19) illustrates how the Nuclear Stress Rule works in Spanish. (19) (. . . *) line 6 (. . *) line 5 (. *) line 4 * * * * line 3 Juan comprô el coche de Pedro Cinque dispenses with the parameter settings of the Nuclear Stress Rule in (18a), since (18a) follows from the head-final/head-initial syntactic parameter. This parameter would be responsible for the direction of embedding. Languages that are head-final should assign the stress prominence to the leftmost element and languages that are head-initial should assign the stress prominence to the rightmost element. This parameter in interaction with the principles in (20) would allow us to determine the phrasal stress. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 9 3 (20) a. Interpret boundaries of syntactic constituents composed of two or more stressed words as metrical boundaries. b. Locate the heads of line N constituents on line N+1. c. Each rule applies to a maximal string containing no internal boundaries. d. An asterisk on line N must correspond to an asterisk on line N-1. Cinque’s proposal is a theoretical improvement over the Nuclear Stress Rule in terras of the Economy Principle since we do not have to specify any special rule for phrasal stress assignment. We can derive it from the setting of the head parameter. In the case of contrastive stress something else should be added. For instance, when a Noun projects two genitive arguments the possessor occupies a higher position than the theme, as demonstrated by Giorgi and Longobardi (1991). If the contrastive stress falls on the possessor, as in (21), then the possessor is more prominent in intonational terms than the internal argument. (21) Juan USÔ el libro de Fisica de PEdro. Juan used the book of Physics of Pedro Juan used Pedro’s book of Physics. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 9 4 Again a theory of saturation that interacts with (20) will give the correct results, assuming that the possessor saturates the NP. My proposal departs from the assumption that every syntactic constituent licensed by a lexical head bears a referential variable for a non-overt argument. This variable can be saturated syntactically or in the discourse. If it is in the discourse. Cinque’s Null Hypothesis is pertinent. If it is at the sentence level, something else should be added. My proposal is to add another line to the algorithm -line 4- where all the stresses of the syntactic constituents marked with the feature [+ wh] are copied, as in (22). (22) a. Copy in line 4 all the stresses of the syntactic constituents marked with the feature [+ wh]. b. Apply (20). Therefore if we apply (22) to example (14a) it would be as in (23). (23)(. • * .) line 6 . (. * .) line 5 • * line 4 * * * * line 3 Juan comprô el COCHE rojo If we have two focused constituents, then it would be as in (24). R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 9 5 (24) ( . . *) line 6 * (. *) line 5 * . * line 4 * * * line 3 MARIA se lo dio a JUAN If we consider that the marked positions for adjectives in (13) are restrictive prenominal and nonrestrictive postnominal, and we also consider Cinque’s Null Hypothesis, we can suggest that there is a connection between restrictiveness and phrasal stress. The Null Hypothesis is the default case of rule (20). Furthermore, the Null Hypothesis constrains the possible analyses for the adjective position. This is something that we should consider from now on. In a string of two postnominal adjectives the most restrictive one, and the one that bears the phrasal stress, is the last one. Therefore, any analysis, such as Lamarche (1991) or Sanchez (1994a), that proposes the underlying position for adjectives to the right, will clash with the Null Hypothesis, since the last adjective cannot be the most embedded one, as (25) shows. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 9 6 (25) DP / \ el NP / \ N’ / \ N’ grande / \ N° rojo coche As we can observe in (25), grande. which receives phrasal stress, is not the most embedded adjective contrary to what is predicted by the Null Hypothesis. Nevertheless, this would not be a problem for Cinque’s (1993) analysis, since each adjective is associated with the specifier position of a different functional projection and the noun moves head to head. Hence, grande would be the most embedded adjective. However, Cinque’s analysis faces some unresolved questions as we will see in section 3.3.3.3. It, furthermore, does not account for the fact that a possessor bears phrasal stress in presence of a complement. Under my proposal, section 3.5., the Null Hypothesis would give the right results for (25) and the aforementioned possessor case, assuming the X’-bar scheme from fn. 14. Before explaining my proposal, I would like to review previous proposals related to adjective placement and nominal head movement. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 9 7 3.3.3. Previous studies 3.3.3.1. Abney (1987) Abney, in accordance with Jackendoff (1977), proposes that adjectives head their own projection in accordance with X'-theory, as in (26). (26) DP / \ D’ / \ D° AP / \ A’ / \ A° NP In contrast with Jackendoff, who projects AP onto the specifier of NP, Abney posits that DP selects AP and AP selects NP. In this way, we limit the number of NP specifiers to one. Abney (1987:299-304) considers adjectives, quantifiers, and adverbs instances of the same category, [+N, +Adj], since they are gradable, in the same sense as mass/plural nouns and predicate nominals are. He proposes that these three subvarieties of adjectives are, in fact, selected by a functional category which is Degree Phrase. Abney proposes that adjectives project two types of theta roles, measure and theme. (27b) would be the interpretation of (27a). R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 9 8 (27) a. John is six feet tall. b. tall (e) & Measure (six-feet,e) & Theme (John,e) However, in Spanish, for instance, we do not have a structure parallel to (27a), as shown in (28), probably attributable to differences in morphological parameters. A definite degree would block agreement between the noun and the adjective in the output of the syntactic component of the grammar, which is where agreement takes place in Spanish, since the DegP is headed by a noun pies. (28) *Juan es seis pies alto/alto seis pies. The morphological differences cannot give rise to such semantic differences. Therefore, I consider that Degree Phrase is a functional category of the predicate denoted by the adjective that quantifies the degree of the attribute. In Spanish, there is overt movement from the XP in the complement position of the adjective to Spec- of-AdjP. In English, the noun is raised to the subject position of the Adjective Phrase from the complement 7 f i We can only have definite degrees in lexicalized adjectives, as in ( i ) (i ) el nino sietemesino the seven-Bonth baby R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 9 9 position at LF. The subject of Degree Phrase is an optional position, similar to the prenominal subject position of verbs in Spanish. Curiously enough, they have similar quantificational effects, if we adopt the VP internal hypothesis (Kuroda, 1988). In (29a) and (30a) we can have either a universal reading or an existential reading of the determiner. In (29b) and (30b) we can have only the universal reading. (29) a. Venia uno de ellos por las tardes. Came one of them in the afternoons. 'One of them came in the afternoons.’ b. Uno de ellos venia por las tardes. One of them came in the afternoons. 'One of them used to come in the afternoons' (30) a. Leia un libro interesante por las tardes. 'He used to read an interesting book in the afternoons.' b. Leia un libro interesantisimo por las 'He used to read a very interesting book in tardes. the afternoons.' In summary, when a noun is modified by a superlative adjective in Spanish, either the adjective is raised to Deg° and the noun to Spec-of-DegP, such as in (30b), or the degree particle is lowered, such as in un libro muy interesante 'a book very interesting'. In (28) pies cannot agree with the noun, so the only way to obtain agreement is to move the adjective head to head to its R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 200 adjunct position, and from there to the adjoined position to Deg’ See (i), (ii), and (iii) for the three derivations. i ) DP / \ un DegP / \ libro • Deg' / \ interesantj-isiiDo AdjP / \ t; A I \ t : t , (ii) DP I \ un DegP I \ Deg’ / \ t : AdjP ^ I \ libro- Adj' / \ muyj interesante t - (iiij IP / \ Juan- VP / \ DegP I \ t- leg' / \ Deg' de alto- I \ ’ ’ seis pies AdjP I \ tj -Adj' / \ Adj' t ; / \ ^ ‘j R eproduced witti permission of ttie copyrigfit owner. Furtfier reproduction profiibited witfiout permission. 201 3.3.3.2. Bernstein (1993) 3.3.3.2.1. Intensional adjectives While the DP-hypothesis has been widely embraced, the same cannot be said about the AdjP-hypothesis. Bernstein (1993) accepts Abney’s hypothesis only for those prenominal adjectives that appear invariably in prenominal position, for instance un mero accidente ’a 7 f i mere accident’, as we will see below. She points out that a uniform position for adjectives would not allow us to account for the different interpretations of canonical prenominal adjectives in Romance and canonical postnominal adjectives. She proposes that postnominal adjectives are in adjunct position to Noun Phrase and their postnominal position is derived by N° to Num° of the Noun. In the case of prenominal epithets, she argues 7 8 Sanchez (p.c.) is developing a similar proposal for intensional adjectives, but she derives the position of extensional adjectives by movement to Spec-of-AdjP. This is not a unified account either, since she treats intensional adjectives as functional heads, and extensional adjectives as lexical heads, in order to account for the lack of movement of the noun to Spec-of- AdjP, in the case of intensional adjectives. Nevertheless, this hypothesis has its advantages in economical terms, if we directly treat them as operators. The problem is that they still have lexical content. The advantage of Sanchez's proposal over Bernsteins’ analysis is that we can explain agreement by an operator-variable relationship, if we posit a movement of the intensional adjective to its specifier or an empty operator that occupies that position, binding the noun. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 202 that these adjectives adjoin to Number Phrase instead of Noun Phrase. In accordance with Bernstein, I will argue below that a distinction between epithets and reference- evaluating adjectives must be made, but, contrary to Bernstein, I will argue that reference-évaluating adjectives, the unmarked position of which is the prenominal position and change their meaning in the prenominal position, such as el pobre hombre ’the pitiable man’, occupy the Spec-of-NP position. The interpretation of epithets or intensifiers in Romance is obtained through movement of the adjective to the Degree Phrase, along the lines of Bosque (1994), and movement from that position to Spec-of-DegP, which allows the adjective to modify the noun at LF in a quantificational way. The interpretation of the reference-evaluating adjectives is difficult to capture in a syntactic analysis. The structure cannot be lexical since coordination of two nouns modified by the same prenominal adjective is allowed, and coordination is a clear syntactic phenomenon. (31) illustrates this point. (31) Mi viejo amigo y colega. ’My old friend and colleague.’ R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 0 3 In (31) we have two interpretations. In the first one viejo only modifies amigo. hence, it may be a lexical compound or a noun modified by an adjective in this interpretation. In the second one, viejo modifies amigo and colega. hence, viejo amigo cannot be a lexical formation in this interpretation. In these cases, we are probably facing a historical change towards a compound, as was the case of gentilhombre for instance, and, we can, therefore, expect some variation. Nevertheless, we should give an account for those cases where these adjectives can be analyzed as independent constituents. 3.3.3.2.2. On the Difference between Intensional Adjectives and Attributive Adjectives If we compare intensional adjectives to epithets, they show some relation and some differences semantically. Reference-evaluating adjectives intensify, in a positive or negative direction, the prototypical properties of the noun. For instance, in el pobre hombre. all the properties that form the noun are modified by the adjective. On the other hand, epithets specify one of the noun parameters with a prototypical value of the noun and emphasize it, as in una blanca nube. ’a white cloud’. Both of them are distinct from postnominal adjectives. As R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 0 4 stated by Cornilescu (1992) , prenominal adjectives modify "reference", while postnominal adjectives modify the referent. Syntactically, epithets and some of the reference- evaluating adjectives seem to work in a very different way. Bosque (1994) claims that indefinite NPs selected by epithets are tokens in Zubizarreta and Vergnaud sense. He equates their syntactic behavior to morphological elatives as -1simo, and to lexical elatives as espléndido. 'splendid', estupendo. 'marvelous, great', magnifico. 'magnificent'. This is shown in (32), (33), and (34) (32) Construid aqui un {edificio elegante/??elegante 'Build an elegant building here.' edificio}. (33) Juan regalo una {*interesante novela/novela 'Juan gave an interesting novel to Luis, Pedro interesante} a Luis, Pedro y Antonio. (TOKEN) and Antonio.’ (34) Una {novela interesante/??interesante novela} 'An interesting novel usually shows the 7 9 In my opinion, these sentences show a contrast between the prenominal and the postnominal position as long as we consider the postnominal adjective non gradahle and just extensional. The contrast improves if we use, for instance, modernista, ’ modernist’, for the postnominal adjective in (32), and historien, ’ historic’, for the postnominal adjective in (33) and (34). R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 0 5 suele mostrar el talento literario de su autor. literary talent of its author.’ (Bosque, 1994) In (32), a postnominal adjective allows a type reading of the indefinite NP, but a prenominal one does not, and therefore it surfaces ungrammatically since the imperative selects for a type. In (33), the postnominal adjective allows either a type reading or a token reading, whereas the prenominal adjective allows only a token reading. In (34), unselective binding is allowed only by a postnominal adjective. If we apply the same syntactic tests to reference evaluating adjectives, the results are very different from those of epithets, as (35), (36), and (37) illustrate. (35) Trae un buen amigo a la fiesta. ’Bring a good friend to the party.’ (36) El profesor les dio una nueva oportunidad a ’The profesor gave another chance to Luis, Luis, Pedro y Antonio. Pedro and Antonio.’ (37) Una nueva oportunidad suele ser aprovechada. ’A second chance is usually made good use of.’ As (35) shows these prenominal adjectives modifying an indefinite NP are compatible with a verb in the imperative. They allow for a type reading of the object NP, as (36) illustrates. Finally, we can observe in (37) R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 0 6 that unselective binding of an indefinite noun modified by these kind of adjectives is possible. Therefore, a reference-evaluating adjective does not force the token reading of an indefinite NP. Chierchia and McConnell-Ginet (1990) define these kind of adjectives as property modifiers, and argue that they can even change the first extension of the noun, creating a new one, as in el falso policia. ’the false policeman’, where its extension is not a subgroup of ’policemen’. Another linguistic phenomenon that has similar intensional semantic effects is the affective affixation. For instance, if valiente. ’bold’, undergoes affective affixation the extension of its reference may change as in valentôn. ’boastful’. In Spanish we have several lexicalizations through this process such as: fûtbol ’football’ > futbolin ’table football’; molino ’mill’ > molinillo ’coffee grinder’; etc. It is interesting that in Italian singular kinship terms allow a possessive without the definite article, as in tuo cugino. ’your cousin’, but when either a prenominal adjective or an affective affix is present the article becomes obligatory as in (38). (38) a. *(il) tuo povero cugino. your poor cousin R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 0 7 b. *(il) tuo cuginetto your little cousin As pointed out to me by Saltarelli (p.c.), constructions as mio cugino behave as R-expressions, since they cannot be used in the plural. The problem is why doesn’t the sequence "possessive + noun + affective affix" behave in the same way? Case reasons might be involved and the determiner might be a case assigner for any adjectival element.In addition to this, the determiner system might be more complex than proposed here. Another determiner that binds intensional relations might exist, this being the lowest determiner in the tree. The effects that we have with these kind of adjectives are very close to the one obtained by epistemic modal adjectives as, probable, ’probable’, and posible. ’possible’, as we can observe in (39). (39) a. Un plan posible. A plausible plan. b. Un posible plan. A possible plan. Also, in Romanian the affixed definite article has to appear in the presence of an affective affix, as pointed out to me by Dumitrescu (p.c.). This could indicate that affective affixation activates DetP. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 0 8 In (39a) the adjective does not modify the noun in an intensional way, but in (39b) it mayIn (39a) we evaluate the possibilities of a plan to be successful. In (39b) we refer to the plan as a possible alternative among others. We cannot have this alternative in the postnominal position. The postnominal position appears to be more related to circumstantial or root modality and the prenominal position to epistemic modality. The syntactic behavior of these adjectives is the same one as that exhibited by the viejo amigo class as shown in (40), (41), and (42), and the contrary of the one shown by epithets. (40) Dame una posible soluciôn. Give me a possible solution. (41) Dio una posible soluciôn al equipo A y al He gave a possible solution to team A and team equipo B. B. (42) Un posible problema suele tener las siguientes A possible problem usually has the following caracteristicas: characteristics: ... In (40) an epistemic adjective can modify an indefinite NP projected by a verb in the imperative form. 0] There is another interpretation where the 'plausible* reading is possible. Then we should consider posible as an epithet. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 2 0 9 In (41) we can have a different "possible solution" for each team, hence, the type reading is available for an indefinite noun modified by a prenominal epistemic adjective. Finally, in (42) the unselective binding is possible for this kind of NP. Bernstein (1993:51-54) points out three other differences between attributive adjectives and reference- evaluating adjectives. Regular attributive adjectives can act as predicates but not as reference-évaluating adjectives, as exemplified in (43) (Examples (43), (44), and (45) from Bernstein, 1993:51-54). (43) a. El libro es corto. The book is short. b.*El accidente es mero The accident is mere The second difference is that attributive adjectives may appear in null nominal constructions, while this other kind of adjective may not appear in that construction, as (44) shows. (44) a. He visto uno corto. I have seen a short one. b.*He visto uno mero. I have seen a mere one. Finally, an attributive adjective can be modified by a degree word in prenominal or postnominal position, but R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 210 a reference-évaluâtlong adjective cannot undergo this kind of modification, as (45) illustrates. (45) a. las muy olorosas flores/las flores muy the very fragrant flowers olorosas b.*un muy mero accidente a very mere accident Therefore, as pointed out by Bernstein, epithets seem to pattern with postnominal attributive adjectives, while reference-evaluating adjectives behave in a very different way. However, Bernstein’s solution seems to me to be a little bit ad hoc and mainly guided by word order factors without considering agreement and constituency factors, as we will see in section 3.5. Bernstein’s second point against a uniform AdjP- hypothesis concerns Abney’s proposal that A® selects NP as its complement, which would not account for the cases where we have more than one adjective. Instead, she proposes that the postnominal adjective position in Romance is adjoined, as we have said above. However, this proposal also has drawbacks, as pointed out by Cinque (1993). In particular, adjectives seem to obey some selectional restrictions among themselves, which is reflected in their serialization across languages, as we will see in the next section. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 211 3.3.3.3. Cinque (1993) Crisma (1990) and Cinque (1993) point out that the interpretation of attributive adjectives differs depending on their position with respect to the noun in Italian. In the case of eventive nouns, if the adjective appears prenominally, then its interpretation can be subject-oriented or speaker-oriented, if it appears after the noun, then, it defines the manner in which the event takes place. As we see in (46), the same kind of facts can be found in Spanish. (46) a. La deliberada destruccion brutal de la ’The intentional brutal destruction of the C i u d a d . city.’ b.*La brutal (manner) destruccion deliberada ’The brutal intentional destruction of the (subject-oriented) de la ciudad. city.’ Cinque (1993) claims that the position of adjectives within DP is Specifier of XP, and not an adjunct position. He gives several reasons for this proposal. The first one is that adjectives show an unmarked order, as we have seen in (46). Since adjunction is usually defined as a free word order phenomenon, then it would be more natural to propose the mapping of adjectives onto Spec- of-XP. The serialization for event nominals is R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 212 illustrated in (47a), and the serialization for object- denoting nominals is shown in (47b) (Cinque, 1993:5). (47) a. possessive> cardinal> speaker oriented> subject oriented> manner> thematic b. possessive> cardinal> ordinal> quality> size> sbape> color> nation The second argument for the generation of attributive adjectives in Specifier of XP is that there is a maximum number of adjectives that can modify a noun without being coordinated. According to Cinque, there cannot be more than six or seven. Adjunctions do not have any numerical constraint, hence, the generation in Specifier of XP is more suitable for the mapping of attributive adjectives, since the restricted number of possible non-coordinated adjectives would be given by the number of functional categories that are selected within DP. However, this might not constitute a straightforward argument supporting Cinque’s thesis, since some authors have tried to limit the number of adjunctions to one for the same projection. This appears to be the case with adverbs as pointed out by Jackendoff (1972) and Valois (1991:149). In Spanish, for instance, an adverb of the probably-class or of the frjquentlv-class (Jackendoff, 1972) may have scope over a proposition, an event, or an object. If both of them appear the probably-class must R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 1 3 have scope over the frequently-class. and they cannot have scope over the same constituent. If they appear adjacent, the only possible interpretation is the probably-adverb modifying the frequently-adverb as in (48) . (48)?Juan probablemente frecuentemente fue a L.A. Juan probably frequently went to L.A. (48) can only mean that it was probably frequent that Juan went to L.A. , hence, we should consider that the position of probablemente is adjoined to the Adverb Phrase headed by f recuentemente. The sentence, nevertheless, sounds quite heavy to me. In conclusion, the maximum number of adjectives might be given by the number of intermediate projections if we equate the analysis of adjectives to that of adverbs. However, what are these intermediate projections? The maximum number of restrictive postnominal adjectives in Spanish is two, in accordance with my intuitions and most of my 09 informants. When there are more than two the sentence is quite marginal, as we see in (49). (49) El coche japonés rojo (??aerodinamico grande). The big aerodynamic japanese red car. DO This is under normal intonation. See section 3.5. for the cases of contrastive stress. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 2 1 4 Following Cinque’s proposal, we should define the two projections that have an adjective on their specifier. It would be more coherent in theoretical terms to argue that these intermediate projections are in fact AdjPs, headed by an adjective, as I will do in section 3.5. In this way, we would be in a better situation to explain the serialization of adjectives in terms of projection. In the case of adverbs, they interact with other notions as modality, aspectuality, etc. Adverbs modify the linguistic elements bearing these notions, such as inflectional verbal elements for mood, aspect, tense. Adjectives seem to be the bearers of these notions in the nominal expressions. The noun does not have any special morphology to define these categories. The third argument is based on reasons of economy for the theory. In accordance with the Specifier hypothesis, the fact that attributive adjectives appear to the left of the head does not need to be specified. This specification would be necessary under the adjunct hypothesis. This is desirable but not a conclusive argument. The fourth argument is given by the impossibility of attributive adjectives to have a complement, as illustrated by Cinque’s (1993:6) sentence in (50). R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 1 5 (50) [np I suoi [»p fedeli (*alla causa) sostenitori] ’His faithful (to the cause) supporters.’ This recursion restriction is characteristic of maximal projections generated on a left branch, and Emonds (1985) stated that XPs generated in a specifier position should end at their head. As illustrated by (51) the same kinds of facts appear with postnominal attributive adjectives (example from Cinque, 1993:7). (51) * I sostenitori [fedeli alia causa] di Gianni ’The supporters [to the cause] of Gianni sono pochi. are few.’ This argument is not categorical either, since it is not clear that alia causa is a complement.®^ Despite some of its advantages. Cinque’s analysis leaves two questions unresolved. First, what is the universal principle that triggers the serialization among See section 1.8. for further discussion. According t o my hypothesis, (51) is only if Gianni receives contrastive stress. The canonical word order is I sostenetori di Gianni fedeli alia causa sono pochi. This i s to be expected in terms of my hypothesis, since Gianni is the internal argument of sostenitori, and must move with the noun to the Specifier of the adjective fedeli, a subject-oriented adjective in this case. Notice that in Spanish, if we have a possessor instead, the canonical word order would be los soldados [fieles a la causal del capitan 'the captain's soldiers faithful to the cause’. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 2 1 6 adjectives? Second, how can we explain the mirror image order produced by adjectives in Germanic and adjectives in Romance? I will come back to these questions in section 5 of this chapter. It is clear that we must posit some kind of hierarchy among adjectives, and that this hierarchy cannot be explained under an adjunction analysis. 3.3.3.4. Valois (1991) Valois (1991) proposes another approach to adjective ordering, focusing on those that modify event nominals. His proposal is based on adjunction structure and adjectival incorporation to the head noun. Valois (1991) adopts a hypothesis similar to Bernstein's with respect to adjectival postnominal positions and argues that adjective order is defined by the adjunction sites. The theoretical point of departure is to define a parallelism between adjectives and adverbs that could match the parallelism between sentential clauses and Noun Phrases which he favors. Basically, he distinguishes between three classes of adverbs that are constrained by the following structural hierarchy: probably-class > frequently-class > completelv-class. from higher to lower position in the syntactic tree. The first class is adjoined to TP, the R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 1 7 second class is adjoined to TP or VP, the third class is adjoined to VP. Adjectives show the same sort of hierarchy, as (52), (53), and (54) illustrate. (52) a. The probable complete invasion of Jupiter. b.*The complete probable invasion of Jupiter. (53) a. The frequent complete invasion of Jupiter. b.*The complete frequent invasion of Jupiter. (54) a. The probable frequent invasion of Jupiter. b.*The frequent probable invasion of Jupiter. Adjectives adjoin to Noun Phrase (a sort of Affix Phrase in Valois’ analysis) and to Number Phrase, which, as adjunct sites, play the same role as TP and VP. The postnominal position of the probable-class. as in (55), is derived by further movement of N“-to-D°. (55) L’invasion probable de Jupiter. The probable invasion of Jupiter. The prenominal positions are the result of adjective incorporation into Num°. I do not think that this can be the correct derivation since, in a case like (55), we could be violating some minimality principle. In (56) the adjective is lowered from the adjunct position to the head of Number Phrase, while the R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 1 8 determiner is raised from Spec-of-NumP to D" . Since both of them are head-movements, they cause a violation of the Head Movement Constraint, unless, in some way, we exclude these incorporation cases from being subject to the HMC, as an ad hoc solution would be. (56) La probable invasion de Jupiter. The probable invasion of Jupiter. If we move the adjective to the Determiner head direct’y, then, we need to extend the subcategorization frame of Determiners to adjectives, an undesirable move, according to Valois, and still an unconventional situation in terms of the HMC, since we would have three head antecedents in the same position and each of their traces would be in a different site. The questions raised in section 3.2.1. with respect to the determiner incorporation remain unanswered. The problem of the different interpretation of the prenominal adjectives, pointed out by Bernstein, is also unsolved. 3.3.3.5.. Giorgi and Longobardi (1991) In accordance with their head subject parameter in (57), Giorgi and Longobardi posit that adjectives in Romance are generated to the right of the Noun and they can move to the prenominal position. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 1 9 (57) Head Subject Parameter External semantic functions (of the noun) are licensed at D-structure on the right in Romance but on the left in Germanic. (Giorgi and Longobardi, 1991:162) Bernstein (1993:31) pointed out that in Walloon movement would be obligatory to the left under this analysis, since Walloon does not have postnominal adjectives. She also says that this would be an undesirable move if we relate adjectives to adverbs, which have been analyzed as remaining in their base generated position in prevailing studies (Pollock, 1989; Belletti, 1990; Cinque, 1993). This approach poses other questions in relation to the lexical projection and X’-theory. First, what external semantic function is borne by an adjective with respect to a noun? This function might be clear in (58a), for instance, where the adjective denotes a generic external role, but it is very obscure in (58b), where the adjective denotes a property of the noun. The mapping of this adjective under NP would give rise to problems with X’-theory. The adjunction solution would encounter the problems pointed out by Cinque (1993). (58) a. La investigaciôn del caso aleraana. The German investigation of the case, b. El coche bianco. The white car. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 220 3.4. N“ Movement 3.4.1. Introduction Several authors (Bernstein, 1993; Cinque, 1993; Valois, 1991; Picallo, 1991; Siloni, 1991) have proposed N° movement to a higher functional head. For instance, Siloni (1991) proposed that adjectives are projected on the left to Spec-of-NPs. In Spanish, N°-to-Num° movement leaves adjectives stranded on the right of Nouns at S- structure. This proposal raises questions similar to those raised by Giorgi and Longobardi’s proposal regarding projection and X’-theory. As pointed out by Bernstein (1993), two types of Noun movement have been proposed for Romance languages. The first one is movement from N°-to-D°, proposed for Romanian to explain the occurrence of the definite article as a suffix in that language, and for proper names to describe why they do not co-occur with the determiner in some Romance languages. All these analyses assume the DP-hypothesis proposed by Szabolcsi (1983) and Abney (1987), and most of them try to equate the nominal functional categories to the verbal ones, following Szabolsci and Abney’s idea.®^ S e e also Fukui and Speas ( 1 9 8 6 ) . R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 221 3.4.2. N*-to-D° movement N°-to-D° movement has been proposed for different language families such as Semitic (Ritter, 1988, 1990; Ouhalla, 1988; Siloni, 1991), Scandinavian (Taraldsen, 1990), Celtic (Duffield, 1991), and Bantu (Carstens, 1991). Among the Romance languages, N°-to-D° movement has been proposed for Romanian (Dobrovie-Sorin, 1987; Grosu, 1988; Cornilescu, 1992) to account for the suffixed definite articles that appear in this language; for French (Valois, 1991) to explain adjective ordering and the liaison phenomenon; and for Italian (Longobardi, 1992) to define the syntactic behavior of proper names. In Romanian, the definite article varies according to case, number, and gender and is suffixed to the noun as (59) shows (Cornilescu, 1992). (59) NOMINATIVE/ACCUSATIVE SINGULAR PLURAL masculine copacu-1 copaci-i feminine carte-a carti-le GENITIVE/DATIVE SINGULAR PLURAL masculine copacu-lui copaci-lor feminine carti-i carti-lor In order to account for the position of the definite article Dobrovie-Sorin (1987) proposes that there is head-to-head movement of the noun from the N°-to-D°, as in (60). R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 222 ( 6 0 ) DP / \ D" / \ D° NP / \ copacuj 1 N t ; Grosu (1988) follows this analysis and proposes that the enclitic determiner is also the assigner of genitive case in Romanian, as Abney does for English. Like Giorgi and Longobardi (1991), Grosu proposes a directionality parameter to explain the differences between English and Romanian. The Determiner assigns genitive case to its left in English and to its right in Romanian, as in (61). (61) Portretu-1 rege-l-ui portrait-DET king-DET-GEN ’The portrait of the king.’ Cornilescu (1992) proposes the existence of two DPs to account for the co-occurrence of a definite article and a demonstrative, as in (62). (62) [npOmj-ul U .p acesta K, t= [«p t;]]]] man-DET" this ‘ ’This man. ’ In (62) N° moves first to Det° and, then, to D°, where affixation with the definite article occurs, leaving behind the demonstrative on Spec-of-DetP. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 2 3 This analysis of Romanian poses important problems for ray proposal. I will argue that the answer to these problems must be based on the appropriate definition of the syntactic properties of the Romanian enclitic article and on the definition of a more articulated version of the DP-hypothesis, along the lines of Cornilescu (1992). The definite article not only appears suffixed to the noun but to other constituents within DP, as an adjective in (63a). (63) a. Viteazu-1 print brave-DET prince b. Printu-1 viteaz prince-DET brave c.*Viteaz print-1 brave prince-DET d.*Print viteazu-1 prince brave-DET The brave prince. Grosu states that the article may be the head of the determiner head or can be generated as a subcomponent of projections other than NP. According to the paradigm in (63), the element that bears the article must appear in first position, except when it appears in a modifier, as in (64). (64) Foarte de frumos-a printesa very of beautiful-DET princess ’The very beautiful princess.’ R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 2 4 Therefore, the definite article in Romanian is one of the features that triggers overt upward movement, as number and gender do in Spanish. A more detailed analysis is needed to define whether this movement is head-to-head or to Specifier of a higher projection. I propose that in the case of the affixed determiner -uT the movement is head-to-head to the lower D° position, and in the case of demonstratives the movement is XP-to-Spec In the case of AdjPs and GenPs, the solution is once again a mixture of these two possibilities. It depends which way is open for movement. If there is an adjective, NP moves to Spec-of-AP. If there is a genitive phrase, as in (61), the noun moves head to head, assuming that the genitive phrase is in Spec-of-Genitive Phrase in Romanian. Interestingly, the case where the article is suffixed to the noun and we have an adjective and a genitive phrase is ungrammatical as (65a) shows. A is: or The derivation that I advance for (62), in accordance with my proposal in section 3.5,2., DemP / \ DetP; Dem' I \ I \ Det' acesta t= / \ ■ ' Qj-ul] NP ^ I I t. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 2 5 prepositional genitive phrase is able to rescue this structure, as (65b) illustrates. I assume that this construction does not occupy the specifier of the genitive projection but the adjunct position, allowing AdjP+NP to raise to the Spec-of GenP. (65) a.*Protretu-l frumos rege-l-ui portrait-DET beautiful king-DET-GEN b. Portretu-1 frumos a-1 rege-l-ui portrait-DET beautiful -DET king-DET-GEN The other case of N° movement in Romance languages is the one proposed by Longobardi (1992). Longobardi (1992) proposes N°-to-D° movement to account for the distribution of proper names and determiners in Romance and the requirements on argumenthood. He posits the principle in (66) for all arguments to comply with. (66) A ’nominal expression’ is an argument only if it is introduced by a category D. fifi Longobardi dismisses the "lexically filled D positions" requirement for this principle since plural nouns or mass nouns can appear in lexically governed positions in Italian with non- overt determiners. In Spanish, we have the same case as (i ) shows for plural nouns and (ii) for mass nouns. ( i ) Compramos zapatos. 'We bought shoes.' (iij Compramos azucar. 'We bought sugar.’ R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 2 2 6 We should probably distinguish between the requirements for being a syntactic argument and the requirements for being assigned case, In the first case the revisited version of (66) in (ill), below, should account for it. In the second case we should continue with the strong version of this condition in (iii). (iii) A 'nominal expression’ is assigned case only if it i s introduced by a filled category D . Plural postverbal subjects of unaccusative and unergative verbs do not require a modifier as (iv) and (v ) illustrate. (iv) Solo vinieron estudiantes. 'Only students came.' (v ) Solo llamaron estudiantes. 'Only students called up.' Suner (1982:213-214) notices that, in the case of transitive verbs, a determinerless subject can appear in postverbal position, only when the internal argument receives case in preverbal position, as (vi.a), (vi.c), and (vi.d) show. (vi) a . La cosecha la destruyeron langostas. the harvest it destroyed locusts 'The harvest, locusts destroyed it.' b.‘Langostas detruyeron la cosecha. locusts destroyed the harvest c.‘Destruyeron langostas la cosecha. destroyed locusts the harvest d.‘Destruyeron la cosecha langostas. Destroyed the harvest locusts (vi.b) shows the case that we have noted above, where an unmodified subject cannot appear in preverbal position. Suner's (1982) Naked Noun Constraint in (vii) accounts for these cases. (vii) Naked Noun Constraint An unmodified common noun in preverbal position cannot be the surface subject of a sentence under conditions of normal stress and intonation. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 2 7 For the cases in (vi.c) and (vi.d), Suner (1982:213) proposes the disambiguating principle in (viii). (viii) Postposition of unmodified subjects of transitive sentences i s not permitted under conditions of normal stress and intonation if the outcome is the sequence V NP (NP) in which the NP(s) is/are not clearly marked as to their function. Notice that the only case where the postposed unmodified subject is possible is when the internal argument has been preposed and doubled by a clitic. Its deletion, as pointed out by Suner, or mapping onto the topic position do not improve the situation, as (ix) illustrates. (ix) a.Heen alumnos. read students 'Students read.’ b.*la cosecha, destruyeron langostas. the harvest destroyed locusts 'The harvest, locusts destroyed it.’ The Naked Noun Constraint can also he applied to objects doubled by a clitic in preverbal position, as ( x ) demonstrates. (x)*Cosechas las destruyeron las langostas. Harvests they destroyed the locusts ’ Harvests, the locusts destroyed them.’ If we consider that preverbal subjects are doubled by a default clitic, as I propose in Martin (1992), where I claim that the preverhal A-position in Spanish can he occupied by the canonical subject or any of the objects, as long as they are clitic doubled, then we can extend the Naked Noun Constraint to the objects that receive case in preverbal position, as in (xi), See also the reformulation of (66) in (111). (xi) Naked Noon Constraint An unmodified common noun in preverbal position cannot be assigned case by a clitic (or AGR) under conditions of normal stress and intonation. Furthermore, we can derive (viii) from the fact that only one NP can be presented at a time, as Suner (1982:1) stated. I assume that determinerless generics always need to be marked with the R eproduced witti permission of ttie copyrigfit owner. Furtfier reproduction profiibited witfiout permission. 2 2 8 presentative function. The verb would always mark the lowest argument in the tree with this presentative function, by a default mechanism. Once the direct object receives case in another position, the verb can assign the presentative function to another argument. Finally, I would like to point out in this footnote that two cases in Spanish seem to escape Longobardi's principle in (66). The first one was pointed out by Fernandez Ramirez (1951:316-317). Determinerless enumerations are grammatical in preverbal subject position, as (xii) shows. (xii) Hombres y mujeres trabajan aqui. 'Men and women work here.’ The second one is implicit in the Naked Noun Constraint. Modified nouns without determiners can appear in preverhal position. See the following sentences from Suner (1982:220). (xiii) Hombres de calidad estudian el asunto. 'Hen of quality are studying the matter.' (xiv) Voces prestigiosas consiguieron frenar los desmanes. 'Prestigious voices managed to stop the misbehavior.' (xv) Cosas como esta desaniman a cualquiera. 'Things like this discourage anyone.’ (xvi) Poderosas razones me obligaron a obrar asi. 'Powerful reasons obliged me to act this way.' Notice, first, that the behavior of a conjoined or modified clitic doubled object in preverbal position is parallel to that of canonical subjects, as (xvii) and (xviii) illustrate. (xvii) Armarios y sillones los vendemos bien. Cabinets and armchairs they sell well 'We sell well cabinets and armchairs.' (xviii) Cosas como esta las oimos todos los dias. Things like this they hear all the days 'We hear things like this everyday.' R eproduced witti permission of ttie copyrigfit owner. Furtfier reproduction profiibited witfiout permission. 2 2 9 In order for proper names to observe the principle stated in (66), we must propose that either there is a null D in the case of proper names or N° moves to D°. The word order distribution of proper names with possessive adjectives, in (67) (Longobardi, 1992:16), and unmarked prenominal adjectives, in (68) (Longobardi, 1992:18) suggests that the solution is the second one. (67) a. II mio Gianni ha finalraente telefonato. The my Gianni finally called up. b.*Mio Gianni ha finalmente telefonato. My Gianni finally called up. Suner (1982) argues that conjoined NPs should be excluded from the domain of the Naked Noun Constraint because they are marked with contrastive stress. I claim that this is also the case for modified subjects in (xiii), (xiv), (xv), (xvi), and (xviii). The only difference is that the contrastive stress is received by the modifier, not by the noun. Hence, the Naked Noun Constraint must be reformulated again as in (xix). (xix) Naked Noun Constraint A determinerless common noun cannot be assigned case under conditions of normal stress and intonation, unless it is assigned the presentative function by a structural governor. As can be observed, the relevant notion now is determinerless and not modified. Hence we can conclude tbat we are dealing with the same kind of Filter as Longobardi. Longobardi's filter i s applied to argumenthood, while The Naked Noun Constraint is applied to case assignment, unless the NP receives the presentative function under structural government. In the case of arguments, a non-overt determiner can comply with the Filter (but see reformulation of Longobardi's Filter in (111)). In the case of case assignment, only an overt determiner can comply with the filter. R eproduced witti permission of ttie copyrigtit owner. Furttier reproduction protiibited wittiout permission. 2 3 0 c. Gianni mio ha finalmente telefonato. Gianni my finally called up. (68) a. E’venuto 11 vecchio Cameresi. Came the older Cameresi. b.*E’venuto vecchio Cameresi. Came older Cameresi. c. E’venuto Cameresi vecchio. Came Cameresi older. The paradigm in (67) shows that the determiner position must be filled in these kind of constructions. This is accomplished in (67a) by an article and in (67c) by the proper name after NP-to-DetP movement and Gen°-to-Det° 07 movement, in accordance with my hypothesis in section 3.5. (68) is the parallel paradigm of the distribution of proper names with unmarked prenominal adjectives to (67), although it is noteworthy that while in the prenominal position the adjective can have either an intensional or extensional interpretation, while in the postnominal position it can only have an extensional interpretation. Therefore, considering that proper names can appear without a determiner in non-argument function, as in (69), we can conclude that the argumenthood principle in (66), in conjunction with the referential properties of proper names, allow NP-to-Spec-of-DetP movement in the fl7 Notice that this is an affective possessive, i.e. it has intensional properties. See section 3.5. for further discussion. R eproduced witti permission of ttie copyrigtit owner. Furttier reproduction protiibited wittiout permission. 2 3 1 case of this kind of noun in Italian, this being a variation of Longobardi’s head-to-head movement. This analysis of the data in (67) explains agreement in an elegant way. In (67a), mio occupies the head of a higher projection to NP, and the determiner its Specifier. N° is coindexed with the expletive article to obtain agreement with the possessive, as in (69). (69) XP / \ il, X’ / \ mio NP I I In (67c), NP has moved to Spec-of-XP, as in (70). (70) XP / \ Cameresi, X’ / \ mio t- The Spanish facts are slightly different to the Italian ones but they do not seem to escape from the same sort of analysis. As is well-known, the counterpart of (67a) in Spanish is ungrammatical, since Spanish does not allow a lexically filled determiner with a prenominal possessive. On the other hand, the counterpart of (67b) is good in Spanish. (71) shows the Spanish correlative sentences to (67a) and (67b). R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 3 2 (71) (*E1) mi Juan ha finalmente llamado. To give the same explanation for the Spanish facts we must assume that the possessive element occupies the Spec-of-DP position, so (71) complies with principle in (66). The fact that a possessive cannot appear after an articleless proper name in Spanish, as the counterpart to (67c) shows in (72), seems to support this hypothesis. In this case, an affective possessive in Spec-of-DetP blocks NP-to-Spec-of-DetP. (72)*Juan mio ha finalmente llamado. My Juan finally called up. Finally, it is interesting to note that postnominal possessives are contrastive in Spanish, as they are in Italian (Longobardi, 1992:17). In this case the presence of the article in Spec-of-DP is required. (73) El Juan mio ha llamado finalmente. The Juan mine has called up ’My Juan has called up.’ We have suggested two different positions for the possessive adjectives that are related to two different interpretations, one has an affective interpretation and the other a contrastive interpretation. The affective possessive is related to the DetP position, Spec-of-DetP in Spanish, while Det” in Italian. This determiner is the R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 3 3 lower determiner. Beneath this determiner, only intensional elements and NP are generated. The contrastive possessive has two options in Spanish, either it is generated in Spec-of-DP,®® when there is no affective possessive present, or in a similar position to postnominal adjectives, heading its own projection and being able to select in their complement position AdjP or DetP, as we will see in section 3.5. The first case will give us the prenominal position, the second one will give us the postnominal position, after DetP-movement to Spec- of-Genitive Phrase. In Italian, they are generated heading their own projection and later raised to D° , in the case of the prenominal position. Following this hypothesis, the definite article appears in Spec-of-DP as an expletive at PP. 3.4.3. N®-movement to an intermediate projection Several authors have pursued the idea of N°-movement to categories headed by nominal inflectional morphology, resembling the syntactic behavior of the verb. The first one proposing an intermediate category between the determiner and the noun is Szabolsci (1983). She proposes Whether it is movement or is directly mapped is not crucial at this point. The relevant fact is that Spec-of-DP forms a chain with Gen’. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 2 3 4 an INFL-like category between the determiner and the noun for Hungarian. This category would have the features [+/- possessive (AGR)]. Szabolsci’s main argument is the overt number and person agreement features that the possessor and the noun bear, in a similar fashion to what happens between the subject and the verb. In addition to this, the morphological agreement shown within a nominal expression parallels the one between the subject and the verb. The paradigms in (74) and (75) illustrate this. (74) a. az én-0 vendég-e-m the I-nom guest-poss-lsg my guest b. a te-0 vendég-e-d the thou-nom guest-poss-2sg thy guest c. (a) Mari-0 vendég-e-0 the Mary-nom guest-poss-3sg Mary’s guest (75) a. én-0 ad-ok I-nom give-PRES.Isg I give b. te-0 ad-sz you-nom give-PRES.2sg you give c. Mari-0 ad-0 Mari-nom give-PRES.3sg Mary gives Therefore, Hungarian seems to behave with a similar pattern to that of verbs, i.e. head-to-head movement. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 2 3 5 This is something that is predicted by ray proposal. In this case the complement of the possessive predicate, NP, is incorporated into the predicate, freeing the specifier, which can now be occupied by the element that is the modifier in the Romance languages. Since the possessor is identified by the head, pro-drop is expected, as in (76). (76) az pro asztal-a-0 the table-poss-3sg Another case that has been studied is the movement of N° to an intermediate category between DP and NP in Modern Hebrew. Ritter (1988, 1990 (cf. Valois, 1991:45- 50)) proposes the existence of an intermediate projection between the Determiner and the NP to explain the difference between Construct State and Free State Noun Phrases in Modern Hebrew (CSNP and FSNP respectively). In CSNPs, as we can observe in (77), the prenominal definite articles are absent. Ritter's (1988) explanation is based on the equation of DPs with INFL at the sentential level. She proposes that, in the case of CSNPs, N“ moves to D° in order to activate the case assigner, as in (78). The definite article lowers to spec-of-NP, since Ritter assumes that it is not subject to the HMC. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 3 6 (77) a. beyt ha-mora house the-teacher The teacher’s house b.*ha-beyt ha-mora the-house the-teacher (78) DP / \ D’ / \ tj beyt. NP J / \ ha:-mora N’ J I I tj In fact, the HMC, as usually stated, does not represent a problem since we can assume that D° can govern the Specifier of its complement. Therefore the trace would be properly governed. What is questionable with respect to Ritter’s analysis in (78) is that the relation between the antecedent and the article would not comply with the Unambiguous Path Condition stated by Kayne (1983:131) in (79). (79) An anaphor that falls under the binding principles must be connected to its antecedent by an unambiguous path. In (78) we have two possible paths dominated by D® at the NP node, its Specifier and N’. The unambiguous path for the trace-antecedent relation is the one that continues up to N°, in accordance with (80), R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 3 7 (80) An unambiguous path in T is a path P = (Aq...Aj, A.^j...Ajj) such that: a. if Aj immediately dominates Aj^j, then A- immediately dominates no node in T other than Aj^j, with the permisible exception Aj.j b. if Aj is immediately dominated by Aj^, , then A.is immediately dominated by no node in T other than A^, with the permissible exception of Aj.j In a movement-to-Spec-of-the-noun analysis, we can give an account for lowering that would be subject to the Unambiguous Path Condition for anaphors. In this case the article would be lowered to the governed head by the head where the trace is left using the correct path. In the case of the FSNPs, where the prenominal definite article appears as an affix to the noun, as (81) illustrates, I claim that the article has been lowered all the way down to the noun. (81) ha-bayit shel ha-mora the-house of the-teacher ’The teacher’s house.’ A further problem for the analysis of (79) is the possible minimality effects that the N°-to-D° movement induces for a lowering movement of the article. Siloni (1991) points out two other problems. First, there is no explanation for the fact that the article cliticizes to any other head than the usual one. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 3 8 Secondly, Ritter’s analysis overgenerates, since we should expect the possibility of having two articles in a row, the one from the head and the one from the complement. Furthermore, when we have a construct state embedded in another one, such as in (82), we should expect that the intermediate noun bears an article, something which does not occur. (82) gag (*ha-)beyt ha-’ish roof the-house the-man ’The roof of the house of the man.’ According to my analysis, this would not represent a problem, since we can assume that the D-features can be transmitted under a Spec-head relation in order for the two nouns to receive case. First beyt would receive case from the head, and then, in turn, beyt would transmit the case features to gag. Why does the article not appear in these cases as it does in Romanian or Somali (Lecarme, 1989 (cf. Siloni, 1991)), for instance? We must conclude either that Hebrew lacks the kind of articles that can fill a lower determiner phrase or that Hebrew lacks agreement on definiteness. As Siloni concludes, the article of the noun does not occur, hence, the article that does appear must be the one that selects the possessor. Interestingly, when the possessor is a proper name, the article does not R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 3 9 occur, as (83) illustrates. Hence, the article must be deleted according to Ritter. This could be accounted for more naturally if we consider proper names as intrinsically referential. (83) maxazot shekspir plays Shakespeare ’Shakespeare’s plays.’ A second problem noted by Siloni is that Ritter’s analysis does not offer an explanation for the agreement on definiteness between the noun and the possessive, as we will see below. In her 1990 paper (cf. Valois, 1991), Ritter proposes an intermediate category, since her 1988 analysis does not account for cases with two genitive phrases such as the one in (84). (84) ha-axila shel dan et ha-tapuax the-eating of Dan ACC the-apple Dan’s eating of the apple. These cases force Ritter to propose an intermediate category, NumP, to explain how the noun leaves the subject behind, which she assumes is on Spec-of-NP,^ In fact, there is not enough evidence here since it is possible that the subject i s projected as an adjunct to the right, ( i ) is a better case since the adjective is a better candidate for the Spec-of-NP position, as proposed by Siloni (1991) for Hebrew. R eproduced witti perm ission of ttie copyrigfit owner. Furtfier reproduction profiibited witfiout permission. 2 4 0 and the determiner moves to the position occupied by the noun, as in (85). (85) DP / \ D’ / \ t; NumP / \ Num ’ / \ hapaxila- NP J / \ shel Dan N’ / \ tj NP et ha-taouax Ritter (1988:919) proposes that the DP contains two different morphemes, definiteness and an abstract case marker that assigns genitive. Since either the genitive or the definite feature can appear without the other, as we can observe in the chart in (86) (Ritter, 1988:919), we can split them into two different categories, DP and INFL or Genitive Phrase, and unify Ritter’s analysis with Szabolsci’s. (86) GEN DEF CONSTRUCT-STATE DP DEFINITE DP + + yes yes + - yes no - + no yes - no no ( i ) ha-axilat ha-menumeset shel dan et ha-uga the-eating the-pollte of Dan ACC the-cake 'Dan's polite eating of the cake.' R eproduced witti permission of ttie copyrigfit owner. Furtfier reproduction profiibited witfiout permission. 2 4 1 With regards to the issue of whether we must derive CSNPs by head-to-head movement or by movement to Spec, further research is necessary. However, if we apply a strong version of the HMC, then, the analysis must be movement to Spec. In the CSNPs, if the definite article appears cliticized to the possessor, then, it imposes definiteness on the possessor and on the noun (Ritter, 1988:915-916). If the possessor carries an indefinite article, then the noun is indefinite (Siloni, 1991). Hence, there is agreement in definiteness between the noun and the possessor in the CSNPs. We have a similar situation with adjectives and nouns in these kind of constructions, as pointed out by Siloni (1991). If we consider the movement-to-spec analysis, then we can conclude that the definiteness of the NP in these cases is triggered by agreement of Spec with a definite genitive noun or a genitive proper noun. The alternative analysis to (78) would be as in (87). (87) DP / \ D’ / \ t- GenP / \ beyt: Gen’ ^ / \ haj-mora NPj R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 2 4 2 Interestingly enough, the main stress falls on the genitive in the CSNPs, as pointed out by Siloni (1991), this, in turn, triggers the application of phonological rules that operate with unstressed words to the noun in the CSNPs. This resembles the state of affairs described above with regards to the stress pattern in the case of a noun restricted by an adjective in Spanish. In conclusion, there is evidence for an intermediate projection, but this projection does not necessarily have to be Number Phrase, and there is no evidence that the upward movement of the noun is head-to-head movement. It could be movement to Spec. Finally, we have seen that Hebrew shows agreement, not only in number and gender, but also in definiteness. However, no overt morphological evidence of this appears in the noun. Following these proposals, Picallo (1991), Valois (1991), Cinque (1993), and Bernstein (1993) have proposed intermediate projections as landing sites for N°, to derive word order within the ’nominal expressions’ . These proposals face several problems such as the explanation of the agreement facts within ’nominal expressions’ in Romance, the account for the difference between adjectives and prepositional modifiers, the difference between restrictive and appositive adjectives, and adjective ordering. In addition to these problems, the R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 4 3 syntactic nature of some of these projections is dubious, as pointed out in section 3.1. In particular, NumP is a very interesting case. If Number Phrase headed a syntactic projection, it would be very difficult to explain, for instance, to discover how agreement is simultaneously acomplished with the constituents higher than NumP, and with those ones that are lower than Number Phrase, at the same time. Longobardi (1992:fn.l4) suggests the necessity of a condition requiring morphological agreement between the features of the determiner and those of each of the head nouns, since determiners strictly match the morphological features of the heads, as (88) demonstrates. (88) a. La mia segretaria e tua collaboratrice The my secretary and your collaborator sta/*stanno uscendo. is/are going out. (88) shows that if NumP is a syntactic category we should posit another one for determiners coindexed with that of the noun, an undesirable move, since we would be proposing a functional category that saturates another functional category, something that is obviously inconsistent on a theoretical plane. I will consider determiners as operators that bind variables borne by the Noun heads. Since operators and variables share the same R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 4 4 ^-features, we do not need an additional condition to explain agreement between determiners and nouns. 3.5. Agreement within NP and the "mise en abyme" structures 3.5.1. Introduction Several linguists from different frameworks have pointed out that the noun and its modifiers constitutes a string where word order is given by specificity. Bolinger (1954), for instance, noticed that postnominal adjectives in Spanish narrow the meaning of the noun from left to right. Dik (1989) analizes nominal modification as a string of restrictors that gradually limit the referentiality of the noun. Longobardi (1992:47) identifies the N position with "a potentially infinite set which constitutes a natural kind". The modifiers, going upwards in the tree, restrict the semantic content of the noun. As in Dik (1989), I will consider these restrictors as open predicates. The limit to these restrictions is to achieve the referentiality of the noun. Once that referentiality has been achieved, the modifiers can only be appositive. What is relevant to the achievement of referentiality is the nature of the noun and the nature of the modifiers. As we have seen in chapter 2, a deverbal noun that projects an affected R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 4 5 object would achieve referentiality through the internal argument. In the same way an inalienable possessee would achieve referentiality through the possessor. As pointed out by Longobardi (1992), proper names achieve referentiality directly. Therefore, the theta roles that can bind the Referential variable of the noun can be different, depending on the nature of the noun. Nouns are defined by their entirety and a set of properties that are represented by parameters with marked and unmarked values. These parameters are set up by the projection of variables that predicate a certain quality of the noun. In Spanish, if the value assigned is generic from the point of view of referentiality, the variable takes the form of an adjective. If it is [+/- referential] the variable projects a ©-role. In this case, the variable is modified by a prepositional modifier. The more consubstantial adjectives are in relation to the noun, the closer they are mapped to the noun. Several semantic classifications have been made. Following Chierchia and McConell-Ginet (1990), we can make a difference between intensional or nonpredicative adjectives, intersective adjectives, and subsective adjectives. Intensional adjectives modify reference, in the sense of Bolinger. They map properties to properties. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 4 6 For instance in false policeman, the property of being a policeman is mapped onto that of being a false policeman. Therefore, they combine with one-place predicates to create other one-place predicates. As we can observe, they are closely linked to the cognitive experience of the speaker. Intersective adjectives modify nouns by restricting the set of individuals denoted by the noun to those that have the property denoted by the adjective. Therefore, the intersection of the extension of possible referents denoted by the noun with the extension of the possible referents with the property denoted by the adjective will give us the new subset, as for example in pink cars. Subsective or relative adjectives are extensional as intersective adjectives. The difference is that the property denoted depends on the context. This context can be defined by different means such as the predicate itself. For instance, a big ant is smaller than a big elephant; a fast car in 1920 is a slow car nowadays; etc. Hence, this kind of modification is also defined by cognition. The cognitive source can be the speaker or a second source, as in his favorite picture. Some of the subsective adjectives can perform as intensional adjectives, as, for instance, grande ’big, great’, depending on their position with respect to the noun. The R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 4 7 intensional position is the prenotninal position, as in gran hombre 'great man’. The canonical subsective position in Spanish is the postnominal position, as in hombre grande 'big man'. Paradoxically, while an intensional adjective seems to be the closest adjective to the noun in Spanish, the subsective adjectives seem to be the most distant ones from the nouns, as (89) shows. (89) intensional > noun > intersective > subsective I I ! 1 (90) illustrates this point. (90) El viejo amigo aleman alto. The old friend German tall Finally, we must point out that nouns are restricted also by spatio-temporal predicates, or what we can call in a more general way alienable predicates, such as demonstratives and alienable possessors. I would like to argue that there is a hierarchy based on alienability that determines the syntactic position of these restrictors in Spanish. The more inalienable that a restrictor is the closer it would be to the noun. The more inalienable that a restrictor is, the more identified it would be with the denotation of the noun. The function of these restrictors is to specify R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 4 8 reference for the noun. Hence, we should expect that, going upwards in the tree, the predicates would become more and more restrictive, with respect to the set of elements denoted by the noun. Once referentiality is achieved, the next restrictors are "rejected" as such, and are assigned to an appositive modifying function. I understand referentiality as in Vergnaud and Zubizarreta (1992). For them, a referential expression is one which takes scope over the entire sentence. (91) represents the referential axis of a nominal predicate, leaving aside the quantifier system and the possibility that there are more operators than DetP and DP, but this must be left for further research. (91) /\ DP \ /\ deixis \ /\ possessor \ /\ extensional adjs \ /\ DetP \ /\ intensional adjs \ noun There is also a hierarchy among adjectives, as we have seen in (47), and also among arguments. The argument that denotes the most with respect to the predicate would be the lower in the tree, as stated in Chapter 2. This R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 4 9 hierarchy is shared by nouns and verbs, in accordance with the Lexicalist Hypothesis (Chomsky, 1970). The determiner system can bind the noun as a type or as a token. Whether there is an adjectival determiner or not is questionable. Some languages, such as Romanian, seem to indicate that this is the case, since, for instance, Romanian has the so-called adjectival demonstratives, as (92) illustrates. (92) fata cea cuminte girl-the Adj Dem good ’The good girl.’ Cornilescu (1992) points out that the function of this demonstrative is to pronominalize the adjective. There is no morphological overt evidence that there is demonstrative article such as this in Spanish. However, we can find some idiomatized expressions where the definite article seems to function as the Romanian adjectival demonstrative article. This is the case with attributive adjectives modifying proper names, as in (93) . (93) Pedro el cruel. ’Pedro the cruel.’ However, this does not happen with common nouns. In general, pronominalized adjectives in Spanish appear in R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 5 0 apposition to the nominal expression they modify. Hence, I will argue that they undergo an obligatory rule of apposition to avoid a violation of Principle B of the Binding Theory and of the i-within-i Condition. Whenever the adjective is pronominalized, it must appear in the the rightward adjoined position, as in (94), at PF as proposed by Lebeaux (1990) for nonrestrictive relative clauses. (94) 'AdjP / \ DP* AdjP / \ /, \ esta- DemP pro blanca / \ casa- Dem' ^ / \ tj In this way, we can explain agreement between the noun and the appositive adjectives without further assumptions. Therefore the complete syntactic tree of nominal expressions will be as in (95). q n For the time being, 1 will leave aside QP and Numeral Phrase. Following Sportische (1388), I consider that QP selects DP. However, notice that QP seems to operate in the DetP periphery also, since todo estudiante 'every student’, for instance, is possible. In the same way, numerals seem to appear in the two environments, and allow type and token reading, as ( i ) illustrates. | i ) Pedro les dio cinco libros a Pedro y a Maria. Pedro to-them gave five books to Pedro and to Maria Pedro gave five books to Pedro and to Maria. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 5 1 ( 9 5 ) DP \ DemP \ GenitiveP \ DegreeP \ AdjP \ DetP \ NP / \ AdjP N’ t I N° In order to obtain agreement, I will argue that there is XP-movement from the complement position to the Specifier position of the immediately dominating projection, with the exception of pure functional categories (DP and DegP). First, I will analyze the DP system, in order to determine its syntactic function. 3.5.2. The DP System 3.5.2.1. Introduction Several authors have suggested the existence of two different types of determiners.^^ Vergnaud and Zubizarreta (1990) suggest that there are two different In the token interpretation, Pedro gave five books to both of them. In the type interpretation, Pedro gave five books to each one. 01 See section 1 . 5 . R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 5 2 kinds of determiners in the Romance languages, an expletive determiner that introduces a type nominal, and a determiner that binds a token nominal. In the second case, the Determiner is said to denote a token, in accordance with the Correspondence Law (Vergnaud and Zubizarreta, 1992:612) in (96). (96) Correspondence Law When a , DP or an NP denotes, the DP denotes a token and the NP denotes a type. When a DP does not denote it is considered to be an expletive with respect to denotation, and, hence, it may not bear a token-index. As pointed out by Vergnaud and Zubizarreta (1992:605) , nominal types show distributivity effects induced by an associated plural argument. In (97), the internal argument can be construed as a token or as a type. (97) On a donné le même ordinateur someone gave SING DEF DET same computer à Sophie, à Justine, et à Cléa. to Sophie, to Justine, and to Cléa. In the token interpretation, only one physical computer was given to Sophie, Justine, and Cléa. Under the type interpretation, there is one possible reading where Sophie, Justine, and Cléa have received three different computers, all of the same type, for instance, the same R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 5 3 brand and make. In the first case, the nominal expression denotes a token instantiated by the type denoted by the NP. Hence, DP bears a denotational index. In the second case, no token is instantiated, thus, DP does not bear a denotational index. When a Determiner does not bear a denotational index, the DP is transparent to predication, giving rise to the distributivity effects that we have seen. 3.5.2.2. The Romanian Case Cornilescu (1992) proposes the existence of two DPs, a DP that selects a DetP that, in turn, selects NP, to explain the distribution of the Romanian affixed definite article in presence of a demonstrative, as we said above. In Romanian the raising of a noun that appears at PF with an affixed definite article is possible in the presence of a demonstrative that has a deictic vowel [a] attached, but not when this vowel does not appear, as (98a) illustrates. On the other hand, the noun without the article can appear after the demonstrative without the deictic vowel, as (98b) illustrates. (98) a. Omul acest*(a) man-DET this b. Acest(*a) om. This man R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 5 4 She proposes that the correct derivation of (98a) is the raising of N°-to-D° through Det°, leaving behind the demonstrative in Spec-of-DetP, as in (99) (99) DP / \ D° DetP omj-ul / \ acesta Det’ / \ Det° NP ‘‘ ' n. ti In the case of (98b), acest occupies Det°, and any movement of N°-to-D" would render an HMC violation. However, this analysis is inconsistent with her former analysis as regards which elements qualify as heads. A head allows nominal elision, as cardinals do (Rothstein, 1988). As pointed out by Cornilescu (1992:208-209), this is the case for Romanian as well. (100) illustrates this point (example from Cornilescu, 1992:209(40)). (100) ^apte [e] au lipsit ieri. Seven were absent yesterday. Romanian demonstrative pronouns bear the emphatic vowel (Cornilescu, 1992:215). In accordance with Longobardi (1992), pronouns are generated as heads of DPs. Further, in section 1.5. we establish that the emphatic vowel or phrasal marker indicates that the element occupies the R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 5 5 head. Cornilescu’s analysis posits that demonstrative pronouns transmit these features to Det° through dynamic Spec-head agreement. If we accept this analysis, we have further problems with cases such as (101b) , where neither the demonstrative without the support of the emphatic vowel, as in acestea, nor the cardinal can license an empty nominal, and, therefore, qualify as heads. As pointed out by Cornilescu (1992:209), in these cases the cardinal number is generated as specifier. (101a) shows that the sentence becomes correct with an overt lexical head. (101) a. Aceste doua eleve sinte cele mai harnice. These two pupils are the most hardworking. b.*Aceste doua [e] sint cele mai harnice. These two are the most hardworking. Therefore, acest cannot occupy a head position blocking N°-to-D° movement in (98b). The reason seems to be that in Romanian there is N°-to-Det° movement, to take the definite feature. The reason seems to be that this is the only position in which definite articles can be generated in Romanian, i.e. the definite article is never generated in DP. The question, now, is how (98a) is derived. Since acesta qualifies as a head, we can no longer derive it through head-to-head movement. I suggest that the derivation is through NP-movement to the Specifier headed R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 5 6 by acesta. as in (102a).^ (102b) is the correct derivation for (82b). (102) a. DemP / \ DetP; Dem' / \ / \ Det’ acesta t- / \ ^ [Om;-ul] NP ^ I I tj b. DP / \ acesti, D’ / \ tk DemP / \ DetP; Dem’ / \ / \ Det’ tr t; / \ ■ ' Om. NP ^ I I ti An unmarked prenominal adjective cannot be left behind by a noun that has undergone article affixation, as (103) shows (example from Cornilescu,1992:214). (103)*omul acesta biet man-the this poor In accordance with Cornilescu (1992:204), (103) should be explained by condition (104). See fn . 85 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 5 7 (104) An attributive adjective must end up in a position where it c-commands its lexical head. Notice that when these kind of adjectives appear and the noun moves, the adjective follows the noun, as illustrated in (105). (105) bietul omul acesta poor-the man-the this This poor man. Since we have concluded that acesta occupies a head c- commanding the NP at the input of the syntactic component, the only possible derivation for (105) is through DetP/NP-movement to Spec-of-DemP, as in (106). I assume that biet is generated on Spec-of-NP and moved to Spec-of-DetP to take scope over the noun. (106) DemP / \ / \ DetP- Dem’ / \ / \ AdjP; Det’ acesta t- biet-iil / \ omL-ul NP / \ t; N’ J I I tk The consequence of this analysis is that DemP does not select DetP always, otherwise we could not account for R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 5 8 the case where an adjective is left behind, as in (107) (example adapted from Cornilescu, 1992:211). (107) distrugerea aceasta brutala a oragului destruction this brutal of city-DET-GEN ’This brutal destruction of the city.’ Therefore, the underlying structure of (107) would be (108). (108) DemP / \ AdjP / \ NoP / \ DetP / \ NP DemP selects AdjP, as proposed by Abney. Following a Determination Hierarchy proposed in section 3.5.1, AdjP selects NoP where the internal 6- role is projected, in accordance with Valois (1991). Finally, the expletive DetP is projected and NP as its complement. NP is defined as a type from DetP downward. Hence, in (108) DetP moves from Spec to Spec until it reaches Spec-of-DemP. We can conclude that Cornilescu’s double DP-analysis for Romanian is viable after these revisions. In the next section I will apply this analysis to the Spanish data. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 259 3.5.2.3. An analysis for Spanish 3.5.2.3.1. The double articulation of DP Do we have anything similar in Spanish to the Romanian double articulation of DP? In the case of the definite article, it is difficult to find evidence since the definite article in Spanish does not appear affixed to the noun. Furthermore, the Spanish definite article has a deictic interpretation, its analysis being more complicated. A parallel Spanish structure to (102a) would be (109). (109) La casa esta. The house this. This house. The parallel derivation to (102a) for this sentence is in (110) . (110) DemP / \ DetP: Dem’ / \ / \ la Det’ esta t; / \ NP I I N° casa In (110), Dem° receives the DP features from DetP through dynamic agreement. This allows DemP to comply with (66) without being raised further to DP. In all three cases we R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 6 0 can give an account for agreement within the nominal expression as we will see in 3.5.3. Hence, we can rewrite (66) as (111). (Ill) A ’nominal expression’ is an argument only if it is introduced by a category with D-features. The reason why the DemP option is optimal is that we can derive esta casa ’this house’ by raising the demonstrative head-to-head to D°, as in (112) (112) DP / \ esta- D’ ^ / \ t • DemP / \ NP; Dem ’ casa / \ tj ti Once the demonstrative acts as an operator, the definite article does not need to appear at PF. If this analysis is correct, we should find some empirical evidence and syntactic effects. I will present some evidence based on distributional facts, the syntactic behavior of the affective possessive, and coordination facts. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 261 3.5.2.3.2. Wh-in-situ We can find the same kind of distributional facts as Bowers did for English based on Wh-in-situ questions. As we can observe in (113), Wh-in-situ phrases, which are not subject to specificity effects, can also appear within a strong DP in Spanish. (113) iTu que casa? 'Your what house?’ This indicates the presence of a lower determiner. 3.5.2.3.3. Affective possessives In Spanish we can have a "Demonstrative + Possessive" structure, as pointed out by the R.A.E. (1978:430). There are few occurrences of the structure "Demonstrative + possessive", probably because they are quite elaborate, but they do occur. The R.A.E. gives the following examples from modern Hispanic literature. (114) a. Esta su frase figura al pie del This his sentence is at-the foot of-the 'This sentence of his is at the foot of the monumento. monument monument.' (Unamuno, Por tierras de Portugal y Espaha: un pueblo suicida.) 09 See section 1.5., examples in (31). R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 6 2 b. Este su nuevo camino que coincidiô ... This his new path that coincided ... ’This new path of his coincided with con su entrada en la vida matrimonial, with his entrance in the life married his entrance into married life.' (Eugenio de Castro) c. Nos estamos muriendo de hambre: la We are dying of hunger: the We are dying of hunger: the daughter-in- nuera y los nietos y este daughter-in-law and the grandsons and this law and the grandsons and his very own su hijo. his son son. ’ (Juan Rulfo, El llano en llamas: Paso del Norte) d. jCuanto he aprendido en esa su obra How much I-have learned in that his play 'How much I have learned in that play of como el mismo la llama! as he himself it calls his as he calls it himself.' (Unamuno, Por tierras de Portugal y de Espaha: Desde Portugal) What is interesting in (114) is that all the possessive pronouns receive an affective interpretation. A double articulation of DP can account for these cases, as (115) shows for (114a). R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 6 3 (115) DP / \ estaj D’ / \ t- DemP / \ DetP; / Dem ’ / \ su Det’ ti t / \ NP À- casa In (115) esta moves from Dem“-10-0“ , after receiving the DP-features in these sentences. If we have these two positions available the co occurrence of the two possessive elements must be possible. This co-occurrence is favored by the presence of an adjective that expresses affectiveness, as (116) illustrates. (116) Mi fotografia (favorita) tuya esta en la My picture favorite yours is on the ’My favorite picture of you is on the mesa. table table.’ If we use a genitive phrase in this kind of sentence, as in (117), the postnominal genitive phrase identifies the possessor, the agent, and the theme, leaving aside any possessor interpretation of mi in this sentence, for the sake of our discussion. Mi does not need to be the R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 6 4 possessor in this sentence, it can be just the experiencer of the state denoted by the adjective. Another interpretation possible is that mi is the experiencer and tuya the possessor, but these interpretations are not relevant now. (117) Mi fotografia favorita de ti. ’My favorite picture of you.’ This is something to be expected, given the Possessivization Principle, and assuming that the possessive licensed by the adjective is the higher possessive. The interesting point here is that, if we consider the examples in (114) and we assume the structure in (115), the adjective binds its possessor role in the lower Determiner Phrase, since DP is occupied by the demonstrative. Further research is needed to study the semantic nature of these constructions, but the affective possessor seems to have some intensional properties that it does not share with the other possessors. It also seems to make DetP directly referential, not allowing the other possessors to c- command it at the output of the syntactic component. Whether this construction could be some kind of genitive construct construction, as the one that Hebrew, for instance, has, is something that needs further R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 6 5 research. There are two different genitive case assignment strategies in Hebrew DPs and the two strategies can co-occur, as we can observe in (118). One of them seems to parallel the affective possessive structure proposed here for Spanish, since it binds the possessor role to a position close to the noun, as the possessive affix of the noun shows. This is the double genitive construction and, as pointed out by Glinert (1989), the phrase to which the genitive is affixed also has to be definite. (118) axilat-o ha-menumeset shel dan et ha-uga eating-his the-polite of Dan ACC the-cake ’Dan’s polite eating of the cake.’ (Ritter (cf. Valois, 1991:50)) If my analysis is correct, we should expect some binding effects with respect to other genitive phrases. This is borne out in (119). (119) Su (affective/*agent/*poss) foto de cada hijo. ’Her picture of each child.’ In (119), foto can have only a distributive reading if the prenominal possessive has an affective interpretation. Otherwise the sentence is ruled out since cada in Spanish always takes wide scope. Therefore, the affective possessive behaves as a weak determiner. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 266 3.5.2.3.4. Coordination Facts Further evidence for the existence of these two DPs in Spanish seems to be given by coordination facts within DPs, where coordination must occur between parallel structures, as argued by Goodall (1987). The Theory of Coordination still has obscure parts that must be elucidated in the future. However, there are some very clear facts, such as that we cannot have two coordinated paths with only one conjunction, and one common intermediate link between both structures as in (120). (120)*Pedro y Juan seran Caligula Neron. ’Pedro and Juan will be Caligula Nero.’ Obviously, we need a second conjunction to conjoin the two predicative nouns, or to form two different conjoined sentences. This should be enough to demonstrate our point about the double articulation of DP in Spanish. In section 3.5.3 I will come back to these structures to demonstrate that the AdjP-Hypothesis and the "mise en abyme" effects that are produced in Spanish are superior to any Chomsky-adjoined theory or NP-complement theory of restrictive modifiers. Either a noun introduced by a definite article or a demonstrative can be modified by a restrictive adjective, as (121) illustrates. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 6 7 (121) Juan trajo el/este coche azul. Juan brought the/this blue car. In the case of the demonstrative, este points out one blue car among a set of blue cars. In the case of the definite article, we have two possible readings. The first one is where the adjective binds the article, and thus the noun also, as a token. In this reading, Juan brought the blue car among a set of cars. In the second reading, among the possible set of cars that he could have brought, he brought the one that belongs to the class "blue". It is in this case that we have the type reading, and we can have coordination between DetPs, as we can observe in (122a). If we assume Abney’s AdjP- Hypothesis, it would be very difficult to explain the contrast in (122) between strong determiners and weak determiners when they introduce two coordinated nouns modified by a restrictive adjective under the "single DP" Hypothesis.^ I n ( 1 2 2 a ) a nd ( 1 2 2 b ) we ca n have an appositive readin g of t h e adjective, i n which t h e determiners denote a t o k e n . H o wever, as we have s e e n , nonrestrictive adjectives behave a s adjuncts o f D P , a nd adjuncts do no t s eem t o interfere with coordinated str u c t u r e s , a s ( i ) s h o w s . ( i ) Juan escribe novelas y corrige p o e u a s en su casa de v e r a n o . 'J u a n writes novels an d corrects poems i n his summer house.' R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 6 8 (122) a. Juan trajo el coche y la raoto Juan brought the car and the motorcycle ’Juan brought the blue car and the blue azules. (Restrictive reading) blue-PLURAL motorcycle.’ b.*Juan trajo este coche y esta Juan brought this car and this Juan brought this blue car and this blue moto azules. (Restrictive reading) motorcycle blue-PLURAL motorcycle.’ (122a) is possible when the definite article introduces a type, i.e. when they are expletive from the point of view of denotation in Vergnaud and Zubizarreta’s sense. However, it is ungrammatical when it introduces a token.The same sentence with a strong determiner is not possible in the case of the restrictive reading, as (122b) shows. An appositive reading is possible in both, (122a) and (122b). As we stated in (94), appositive adjectives must receive a different analysis. A type reading of the direct object in (122a) is possible, as (123) demonstrates. In (123) we can have a distributive reading of the internal argument. In the context where Juan and Maria have a set of cars and motorcycles, each The token interpretation is fine if the adjective is a depictive, but this case is not relevant now since depictives are not considered to be projected within DP. R eproduced witti permission of ttie copyrigtit owner. Furtiier reproduction protiibited wittiout permission. 2 6 9 one of a different color, and the ones that they have brought are those belonging to the class "blue." (123) Juan y Maria trajeron el coche y Juan and Maria brought the car and la moto azules. the motorcycle blue-PLURAL If we map the strong determiners onto DP, the ungrammaticality of (122b) would be explained, since the AdjP interrupts the coordination process, as (124) illustrates. The generation of DemP is irrelevant for our discussion here. (124) DP / \ este/esta D’ / \ AdjP / \ coche/motocicleta, Adj ’ / \ azules tj For the same reason that this structure explains the ungrammaticality of (122b), it cannot explain the gramraaticality of (122a). A structure parallel to the one in (113) works better, since in this case the AdjP does not interrupt the coordination between DetPs, as we can observe in (125). R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 7 0 ( 1 2 5 ) D P / \ AdjP / \ DetP; \ / \ \ el/la Det' \ / \ \ NP \ coche/moto Adj’ / \ azules t; Notice that if in (123) we introduce the nouns with strong determiners, we cannot have the type reading, as (126) shows. (126) Juan y Maria trajeron este coche Juan and Maria brought this car y esta moto azules. and this motorcycle blue-PLURAL In the next section I will show that the deletion test gives some further evidence for our proposal, although the deletion test does not seem to support this analysis prima facie. 3.5.2.3.5. Nominal Elision We should be able to delete a Determiner Phrase that does not denote a token along with the NP, but not a DP that does denote a token, since it is obviously necessary in the second clause in order to identify the second token. However, other factors can affect the behavior of R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 7 1 the deletion structures. In accordance with Williams (1977), I consider that deletion applies only to maximal projections. Hence, we should expect to be able to delete DetP, for instance. (127) shows that this is not possible, leaving aside any adverbial interpretation q? of the adjective, which we are not concerned with now. qc Nevertheless, this is further evidence for the DP-hypothesis. It shows that the determiner is part of a projection independent from NP, I t is also evidence for distinguishing determiners that introduce types as a projection from NP. q? There i s a class of adverbial adjectives t h a t we ca n classify as object-oriented. They behave i n a similar way t o t h e more widely studied subject-oriented adverbial a d j e c t i v e s . I f we apply Lujan's ( 1 9 8 0 : 1 5 4 ) t e s t fo r t h e s ubject-oriented adverbial adjectives, based o n t h e f a c t t h a t t h e y answer questions with comp ' h o w ' , a s adverbs d o , we obtain results similar t o those of the subject-oriented c l a s s , as ( i ) an d ( i i ) i l l u strate. I n ( i ) we have a subject-oriented adverbial adjective and i n ( i i ) we have a n object-oriented adjective. ( i ) i C o f f l o llegô la maestra? La maestra llegô c a n s a d a . ' H o w did t h e teacher arrive? Sh e arrived tired.' ( i i ) i C ô f f l o comprô e l coche Juan? J uan coaprô e l coche a z u l . How did-buy t h e car Juan? Juan bought t h e c ar b l u e . ' W h a t color of car did Juan b u y ? He bought a b lue o n e . ' H e n c e , we can n o t project these adjectives within D P . Substitution b y a clitic pronoun corroborates t h i s a n a l y s i s , a s ( i i i ) illust r a t e s . ( i l l ) iComprô Juan e l c o c h e j az u l]? S i , l 0| comprô » ( a z u l ) . bought Juan t h e car blue? Y e s , i t bought blue ' D i d Juan buy a blue car? Y e s , h e bought i t b l u e . ' I n ( i i i ) , t h e clitic cannot refer t o t h e adverbial adjective since i t m ust appear overtly i n the a n s w e r . F u r thermore, we can extend Antrim's ( 1 9 9 3 ) p roposal f or suhject-oriented adverbial adjectives i n I t a l i a n t o t h e object-oriented adverbial adjectives i n S p a n i s h . I n accordance with Antrim ( 1 9 9 3 ) , t he subject i s projected o nto R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 7 2 (127)*Pedro trajo el coche azul, y Marla Pedro brought the car blue, and Marla ’Pedro brought the blue car, and Marla brought (lo) trajo amarlllo.(NON ADVERBIAL) It brought yellow, the yellow one.’ As you may notice, the presence of the clitic pronoun lo does not allow the sentence In (127) to be Interpreted In the sense relevant for our discussion. This clitic Is th e Specifier o f A d v P , a n d , t h e n , moved t o Spec-of-AGRP. I n a similar f a s h i o n , we ca n propose t h a t th e direct ob ject i s projected within AdvP and t h e n moved t o th e Specifier of A G R J ? . H e n c e , Antrim's analysis predicts t h e grammatical deletion of DP when we have th e adverbial interpretation of t h e adjectives i n ( 1 1 2 ) . Antrim also p roposes t h a t Adv’ incorporates i nto t h e verbal s y stem. I n t h a t c a s e , we should po s i t t hat subject-oriented adjectives incorporate into A G R j * , an d object-oriented into A G R ^ ' . H o wever, i t i s not clear t o me t h a t t h i s i s t h e case i n S p a n i s h , since t h e c l a s s of a d jectives t h a t behave a s subject-oriented adverbs i s more restricted t han i n I t a l i a n , as we can observe i n ( i v j and ( v ) . ( i v ) Maria camnina l e s t a / l e n t a . ( v ) ‘ . M a r i a camina rap i d a / l e n t a . Maria walks f a s t / s l o w . I f we compare ( i ) t o ( v ) , the generalization seems t o be t hat only adjectives t h a t express conditions may ac t a s subject-oriented adverbial adjectives in S p a n i s h . I n other wor d s , adjectives t h a t are usually selected b y t h e verb estar can ac t a s subject-oriented adv e r b s , bu t n o t those t h a t are selected b y s e r . To c o n c l u d e , th e class o f a djectives t hat allows subject-oriented adverbialization i n Spanish seems t o involve a coreferential r e l a t i o n , w h ilst, i n I t a l i a n , we have a f u r ther c l ass t h a t allows an anaphoric predicational relat io n i n Napoli’ s ( 1 3 8 9 ) s e n s e . R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 7 3 necessary when deletion of a direct object, introduced by a filled DP, occurs in Spanish, as (128) illustrates.^® (128) Pedro trajo el coche, y Maria *(lo) trajo Pedro brought the car, and Maria did too. también. In (129), we can observe that the presence of the article makes (127) grammatical. Deletion of NP is expected in my analysis, since we would be deleting the whole NP projection. These facts support the DP-hypothesis as well, since deletion of NP is allowed without deleting the determiner. (129) Pedro trajo el coche azul, y Maria Pedro brought the car blue, and Maria ’Pedro brought the blue car, and Maria brought trajo el amarillo. brought the yellow, the yellow one.’ My claim is that, in cases of deletion of the NP, the article is necessary to license partitive predicates that bind their theta-role on the determiner. NP-deletion has been a long-standing problem in the linguistic studies of Spanish. We should point out that the definite article does not only appear in NP-deletion structures Basque Spanish allows inanimate null objects without clitic identification. See Franco and Landa (1991) for further discussion, R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 7 4 with restrictive adjectives, but also with "de + NP" modifiers, as in (130a), and with relative clauses as in (130b) . (130) a. Juan prefiere el coche de Ana y Maria Juan prefers the car of Ana and Maria ’Juan prefers Ana’s car, and Maria *(el) de Pablo, the of Pablo Pablo’s.’ b. Juan prefiere el coche que conduce Ana Juan prefers the car that drives Ana ’Juan préférés the car that Ana drives y Maria *(el) que conduce Pablo, and Maria the that drives Pablo and Maria the one that Pablo drives.’ Traditionally, there have been two main hypotheses. The first one is based on the analysis of the determiner as a pronominal element, and the second one on the analysis of the adjective as a noun. Bello (1847) proposed that the definite article functions as an unstressed pronoun in these cases. This analysis has been followed with some variations by various Spanish grammarians and scholars such as R. Seco (1930), R.A.E. (1931), Fernandez Ramirez (1951), the morphological section of R.A.E. (1973), Alcina and Blecua (1975), and Lujan (1980), among others. The second analysis, advocated by Lenz (1920), Gili (1943), Alarcos (1961), and M. Seco (1972), proposes that in cases like (129) the adjective has undergone a process R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 7 5 of norainalization. Little attention has been paid to cases such as those in (130), as pointed out by Brucart (1987). As we will see below, Lujan puts forward some important arguments against this theory. Even when they ascribe themselves to Bello's theory, Brucart and Hernanz (1987), and Brucart (1987) point out two possible drawbacks for his hypothesis. Firstly, we should form two different paradigms, the pronominal paradigm, and a paradigm that corresponds to the article, with the same forms. The only exception would be the neuter pronoun lo, as pointed out by Lujan (1980:127- 141). It would be included only in the pronominal paradigm, since it cannot introduce an overt noun. Nouns in Spanish are assigned to the feminine or masculine gender, but not to the neuter. On the other hand, lo can be used in elliptical structures, introducing an adjective, a relative clause, or a "de + NP" structure, as (131) illustrates. (131) a. lo bueno NEUTER PART good what is good b. lo que quiero NEUTER PART that want what I want c. lo de Ana NEUTER PART of Ana what belongs to/what is related to/etc. Ana R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 7 3 Lujan’s argument is that if we demonstrate that the adjective in (131a) does not function as a noun, then we cannot consider lo as an article. Lujan (1980: 118-120) gives three contexts where adjectives in the absence of nouns do not function as nouns do. The first one is where adjectives that start with a stressed /a/ do not select the feminine form of the article el, as nouns do, but la, as (132) illustrates. (132) a. la/*el alta the tall one b. la/*el agil the agile one c. la/*el ârida the arid one d. la/*el habil the clever one This shows that these adjectives do not function as nouns, since nouns of this type select el and not la, as in el/*la ala ’the wing’. The second piece of evidence given by Lujan is that these kind of constructions, when they are direct objects, require the question word cual ’which’, as other nominal constructions with attributive adjectives do. It does not occur with gué ’what’ , the question word that is used when we inquire about the whole NP and we do not know what noun-type is. (133) illustrates the use of gué, R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 7 7 and (134) does the same with the use of cual. We should add that gué is also possible in contexts such as (134b). Nevertheless, the relevant contrast for Lujan’s analysis is the one given in (133). (133) a. îQué compraste? Corapré el material. What did you buy? I bought the material. b.*i,Qué compraste? Compré el bueno. What did you buy? I bought the good one. (134) a. iCual compraste? Compré el bueno. ’Which one did you buy? I bought the good one. ’ b. i,Cual material compraste? Compré el ’Which material did you buy? I bought the material bueno. good material.’ Finally, Lujan’s third argument against considering these adjectives as nouns is that adjectives in elliptical constructions take adverbial degree words as modifiers, in the same way that adjectives, but not nouns, usually do. This is illustrated in (135). (135) a.*el muy médico the very doctor b. el muy honrado the very honest c. el empleado muy honrado the very honest employee R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 7 8 Brucart and Hernanz (1987:195) also put forward theoretical arguments against Lenz’s hypothesis. According to this theory, we should assume that syntactic projections can change their label in the course of their derivation or that a head can project a syntactic projection of a certain category different from the one that the head belongs to. Brucart and Hernanz’s second argument against Lenz’s proposal is that this hypothesis does not account for the fact that there is an antecedent in the sentence or discourse level that allows us to give an interpretation to the elliptical structure. These arguments, in conjunction with the fact that the nominalization hypothesis of the adjective in these contexts does not have a clear analysis for the cases in (130),^^ disqualify the feasibility of this hypothesis. Brucart and Hernanz (1987:194) point out a second problem for Bello’s theory, which Lujan sees as a supportive fact for that theory. When an elliptical nominal expression is modified by a relative clause, the element that appears as antecedent of the relative clause is different depending on whether the relative clause is restrictive or nonrestrictive. If the relative clause is restrictive, the antecedent surfaces as a definite 9 9 Alarcos, fo r inst a n c e , assigns head properties t o t h e specific modifiers. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 7 9 article, an unstressed pronoun in Bello’s account, as in (136a), and as a stressed pronoun when the relative clause is nonrestrictive, as in (136b). (136) a. Ana dijo que el/*él que habla francés debe ’Ana said that the one that speaks French ser incluido en el proyecto. must be included in the project.’ b. Ana dijo que él/*el, que habla francés, ’Ana said that be, that speaks French must debe ser incluido en el proyecto. be included in the project.’ In Brucart’s account, we cannot consider the definite article as a pronoun in elliptical constructions, since it does not appear in the same contexts as the pronoun. In Lujan’s account the definite article appears in similar contexts but has other properties. Despite their criticism, Brucart (1987:326) and Brucart and Hernanz (1987b:195) acknowledge the advantages of Bello’s hypothesis, in the sense that the definite article seems to be a kind of substitute for the noun in these constructions. Up to now, we have not explained why the definite article is required in these constructions. Brucart and Hernanz (1987:195) suggest that the article is necessary in order for the nominal expression to receive case, if we consider a filter similar to the one in (xix) of fn. 86, repeated here in R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 8 0 (137). This seems to be a feasible account, although it does not fully explain the distribution of all elliptical constructions, as we will see below. (137) Naked Noun Constraint A determinerless common noun cannot be assigned case under conditions of normal stress and intonation, unless it is assigned the presentative function by a structural governor. As pointed out by Brucart and Hernanz (1987:196), the definite article does not appear with other elliptical constructions where modification occurs by a prepositional modifier introduced by a preposition other than ^ ’of’ as (138) illustrates. (138) a. El camino hacia Madrid/*E1 hacia Madrid The road to Madrid/ The to Madrid The road to Madrid/the one to Madrid b. El evangelio segun San Lucas/*E1 segun The by Saint Luke/ The by The by Saint Luke/The one by San Lucas Saint Luke Saint Luke c. El amor por Pedro/*El por Pedro The love for Pedro/ The for Pedro The love for Pedro/ The one for Pedro Brucart and Hernanz relate these facts to the special status of the preposition that can be used with different semantic contents. This results from the dummy R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 8 1 status of the preposition, pointed out by Chomsky (1981:49). This preposition can be depleted in syntactic contexts where other prepositions can appear, as long as the partitivity requirements specified in Chapter 2 are met. For instance we can say el de San Lucas ’Saint Luke’s one’, with the same meaning as in (138b), but we cannot say ’el de Pedro’ ’Pedro’s one’ with the same meaning as in (138c), since, as we pointed out in Chapter 2, NPs with the SOURCE OF EMOTION role rarely induce partitive readings of the noun that they modify. Therefore, we should check the syntactic behavior of elliptical constructions in relation to the Naked Noun Constraint in (137). We should expect that under normal conditions of stress and intonation, any DP-deletion structure modified by a restrictive modifier would be impossible when it does not receive the presentative function, and possible when they receive the presentative function. The first prediction is fulfilled by DP- deletion structures modified by a prepositional phrase other than as in (139a). When they are pronounced with normal stress and intonation, and occupy a position that cannot receive the presentative function, the sentence is ungrammatical. On the other hand, when they receive contrastive stress, the sentence is grammatical. When the modifier is an AdjP, as in (139b), a "de + NP", R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 8 2 as in (139c), or a relative clause, as in (139d), the prediction is also fulfilled when the modifier does not receive contrastive stress. Notice that pronouncing these phrases with contrastive stress does not improve judgements about the sentences. In both cases we need the presence of the definite article, as in (139b), (139c), and (139d). (139) a. El evangelio segun San Lucas The Gospel according-to Saint Luke ’The Gospel according to St. Luke produced produjo grandes bénéficiés, pero *segûn produced great profits, but according great profits, but St. John’s produced San Juan/SEGUN San Juan produjo grandes to Saint John produced great great loses.’ pérdidas. loses b.*El evangelio japonés produjo grandes The Gospel Japanese produced great ’The Japanese Gospel produced great profits bénéficiés, pero ruso/RUSO produjo grandes profits, but Russian produced great but the Russian Gospel produced great pérdidas. loses loses.’ c.*El evangelio de San Lucas produjo The Gospel of Saint Luke produced ’The Saint Luke’s Gospel produced great grandes bénéficiés, pero de San Juan/DE San great profits, but of Saint John profits, but St. John’s produced great R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 8 3 Juan produjo grandes pérdidas. produced great loses loses.' d.*E1 evangelio que escribio San Lucas The Gospel that wrote Saint Luke ’The Gospel that Saint Luke wrote produced produjo grandes bénéficiés, pero que produced great profits, but that great profits, but that that Saint John escribio San Juan/QUE escribio San Juan wrote Saint John wrote produced great loses.’ produjo grandes pérdidas. produced great loses This suggests that, in addition to the Naked Noun Constraint in (137), another condition or principle affects NP-deletion when the NP is modified by a restrictive AdjP, a restrictive "de + NP" phrase, or a restrictive relative clause, but not when NP is restricted by a prepositional phrase. With respect to the second prediction, we have seen in (129) for AdjPs, in (130a) for "de + NP" phrases, and in (130b) for relative phrases, that deletion of the determiner is not allowed in a position where the nominal expression can receive the presentative function, in this case the internal argument position. Again, prepositional modifiers other than those introduced by ^ allow determiner deletion as we can see in (140). R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 8 4 (140) Juan prefiere el evangelio segun San Juan prefers the Gospel according-to Saint ’Juan prefers the Gospel according to St. Luke Lucas y Maria segun San Juan. Luke and Maria according-to Saint John Mary the Gospel according to Saint John.’ Therefore, the Naked Noun Constraint in (137) accounts for the cases with prepositional phrases. Notice that a preverbal direct object doubled by a clitic observes the same restrictions as a preverbal subject when it is modified by a prepositional phrase, as we can observe in (141). (141) El evangelio segun San Lucas lo The Gospel according-to Saint Luke it ’Juan prefers the Gospel according to St. Luke prefiere Juan y *segûn San Juan/SEGUN San prefers Juan and according-to Saint John Mary the Gospel according to Saint John.’ Juan Maria. Maria In the case of AdjPs, "de + NP" phrases, and relative clauses a further constraint seems to be at work. First notice that when an overt noun does not need an overt determiner, it does not need it when deleted, as (142a) shows for AdjPs, (142b) for "de + NP" phrases, (142c) for relative clauses, and (142d) for prepositional phrases. (142) a. Maria prefiere libros historiens y Juan Maria prefers history books and Juan R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 8 5 (libros) cientificos. science (bocks) b. Maria prefiere libros de historia y Juan Maria prefers books of history and Juan (libros) de ciencia. (books) of science. c. Maria prefiere libros que traten de Maria prefers books that are about history historia y Juan (libros) que traten de and Juan (books) that are about science. ciencia. d. Maria prefiere libros sobre historia y Juan Maria prefers books about history and Juan (libros) sobre ciencia. (books) about science. Hence, this further constraint for AdjPs, "de + NP" phrases, and relative clauses should be internal to DP. This restriction seems to be related to the Full Interpretation Principle, which requires that each element must be licensed appropriately. If we consider 0- role assignment as some sort of referential function with respect to the predicate that assigns a 0-role, we should expect some type of feedback from the constituent that has been assigned a 0-role to the assignee. Referential elements are NPs. However, they obtain referentiality through their relation with their modifiers. We can characterize them as potentially referential elements. Therefore, the type of empty category that we have in a R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 8 6 case such as el azul ’the e blue’ is a variable, not a pronominal element. The element that identifies this empty category is an operator. Hence, an adjective or a relative clause is not able to bind the referential properties of a 0-role by itself, neither is a modified individualized noun. The 0-role of modifiers is bound by the element that bears the referential index of the predicate, in the case of nouns, the determiner.It is the determiner that codifies the referential properties of a nominal predicate. If that operator is not present at the syntactic level, that information is not recoverable at LF. In (142), the modifiers do not need to bind their theta role in the Det°, since the noun has not been individualized, in the sense of Sole and Sole (1977:288). Since the noun is a lexical element, it is recoverable at LF. Thus, the modification relation is also recoverable. If only elements with referential indexes are visible at LF, then adjectives and relative clauses would not be able to modify the NP at LF, since they lack a referential index. This would result in a case of vacuous modification. The cataphoric properties of the determiner with respect to modification have been pointed out by Quirk, Greenbaum, Leech, and Svartvik (1972) for English, and Suner (1982) for Spanish, among others. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 8 7 since determiners are not lexical elements, hence, the modifier is not licensed appropriately . This account explains the cases with AdjPs and relative clauses. It also explains the cases where DetP-deletion is possible with prepositional phrases others than since, in these cases, the 0-role is discharged onto NPs, i.e. elements with existential properties, hence, able to be referential. However, the case with "de + NP" structures remains unexplained. I claim that in the case of "de + NP" phrases, the modifier binds the article because it saturates the Referential or Eventive variable of the noun. In sections 2.4. and 2.5., I argue that affected objects saturate the R-variable of nominal predicates, while unaffected objects do not. Hence, we should expect that a determiner binds the predicative relation between a nominal head and an affected object, since the object saturates the R- variable of the noun, while unaffected objects do not. In accordance with this, DetP-deletion should not be possible with event nouns that project an affected object, while it should be possible with unaffected objects. This prediction is borne out in (143). (143) a. Juan vio la destruccion del puente, y Juan saw the destruction of-the bridge, and ’Juan saw the destruction of the bridge. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 8 8 Maria *(la) de la presa. Maria the of the reservoir and Marla the destruction of the reservoir.’ b. Maria hizo la memorizaciôn de las primeras Maria made the memorization of the first ’Maria memorized the first lessons and Juan lecciones y Juan (la) de las ultimas, lessons and Juan the of the last the last ones.’ In (143a) we can observe that DetP-deletion is not possible when the noun is eventive and projects an affected object, while DetP-deletion is possible when the noun projects an unaffected object. As long as this analysis is correct, it gives further evidence for ray account of affectedness in Chapter 2. We conclude that, whenever an operator codifies an argument-predicate relation, that operator must be overtly present in the syntax. Finally, I would like to review some typological evidence that has been registered in relation to the existence of two types of determiners. 3.5.2.3.6. Typological Evidence As pointed out by Longobardi (1992), if we have these two types of determiners, an expletive determiner and "a substantive, really definite, determiner" (Longobardi, 1992:55), we should be able to find some R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 8 9 typological evidence, in the sense that we can find articles in some languages that are not homophonous. First of all, Longobardi (1992:55) distinguishes between two kinds of expletive articles, those that introduce a proper name and those that introduce generic expressions. In terms of interpretation articles of the first kind relate an interpretively relevant position D to an irrelevant one, N. In the first case, the expletives do not bind any semantic content of the Noun, while in the second case the expletive saturates the semantic content of the Noun, without adding any special semantic content. This function can be observed in Romance in the different form of stage-level generics and kind-level generics in a lexically identified position. In (144a) (Longobardi, 1992:26), we can observe that in the first case the article does not appear since it is occupying a lexically identified position, and case requirements do not arise for its presence. The interpretation is ’some old ladies were excluded’. In (144b) the article is required in order to saturate the semantic content of the noun. The interpretation is ’all the old ladies were excluded’ Spanish shows similar contrasts a s th e counterparts of ( 1 4 4 ) s h o w , ( i ) a . He excluido solo viejas s e n o r a s . b . He excluido solo la s viejas se n o r a s . R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 9 0 (144) a. Ho escluso solo vecchie signore. b. Ho escluso solo le vecchie signore. 'I only excluded the old ladies.’ The next task is to determine the position of the expletive article that introduces kind-level generics, DP or DetP. Since it has referential importance, the direct object has a de re interpretation in (144b), hence, we should posit that the position is DP. Probably, kind- level generics obliterate the DP position by raising the expletive article from DetP to DP. This position is related to the event or spacio-temporal location argument position proposed by Kratzer (1988) for stage-level predicates. When DP is not occupied by an element with semantic content, the existential interpretation typical of stage-level predicates is rendered. This allows us to explain the contrast in (145) (examples from Kratzer, 1988:1) . (145) a. There are firemen available. (STAGE) b.*There are firemen altruistic. (INDIVIDUAL) In (145a) the DP position is available for an expletive to receive case in the preverbal position. This position is not available in (145b), since it has been obliterated. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 9 1 Therefore, the term expletive is related to the DetP position. Other candidates for this position are the Italian partitive article, that may be inserted in front of stage-level generics, as in (144a), and the expletive article of proper names. Again, the typological variation is enormous. For instance, Spanish lacks the partitive and an expletive article for proper names. Catalan has an exclusive expletive article for proper names, masculine en/feminine na. The reason for this variation is the lack of semantic support, as pointed out by Longobardi. The article that appears with proper names in Spanish does not seem to be expletive. It has a deictic connotation. The use of the article for proper names depends on stylistic considerations. Its use is impolite, like the use of demonstratives to denotate persons. Furthermore, the Spanish feminine article has two forms, el and I^. el is used in front of nouns that start with a stressed /a/, as in (146). (146) el agua DET-F water ’the water’ When we use a definite article with a feminine proper noun that starts with a stressed /a/, the form selected is la, as in (147). R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 9 2 (147) La Ana The Ann This could indicate that the article has not been generated in the immediate environment of NP, probably in Spec-head relation within NP or DetP, but in DP, or perhaps DemP. Torrego, (1985, 1988:5) gives a further argument for this article with deictic content. In contrast with the French and the Italian article, the Spanish allows empty nominal heads as el de Durero ’the (one) of Dürer’. French and Italian use the demonstrative as (148) shows. (148) a. Celui/*le de Dürer. b. Quello/*il di Dürer. Hence, only those elements that are heads by the output of the syntactic component are allowed to license empty nominals. We can conclude that Spanish does not have expletive articles for proper names. There is another interesting case where Spanish differs from French and Italian with respect to expletive articles. It is with existential singular nouns in a lexically identified position, as in (149). (149) He comprado coche. I have bought a car. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 9 3 As we can observe in (150a) and (150b), the French and the Italian counterparts require the article. (150) a. J’ai acheté *(la/une) voiture. b. Ho comprato *(la/una) macchina. Spanish avoids the use of an overt expletive article when it is mapped on DetP by PF, and the determiner does not need to bind an extensional adjective.The three types of definite article that we have in Spanish are: the one which denotes a token, mapped onto DP by PF; the one constrained by case requirements, given the Naked Noun Constraint in (137), which I claim is mapped onto the DP position also; and the existential one, which is mapped onto DetP. Romanian lacks a definite article mapped directly onto the DP position. Finally, Italian probably has the existential one that moves to Spec-of-DP when saturated or when needed because of case reasons. The most interesting case, as pointed out by Longobardi (1992:58) is Frisian, a Germanic language. The differences among its dialects are especially interesting. Longobardi gives the following description As I have indicated throughout this dissertation there are probably more than the two positions that I am positing, but, for the time being, only two are proposed. Further research is needed. R eproduced witti permission of ttie copyrigtit owner. Furtiier reproduction proiiibited wittiout permission. 2 9 4 of the dialect spoken in the island of Fohr, based on Ebert’s 1970 work (cf. Longobardi, 1992): The Fohr dialect displays two different types of non-indefinite article, one occurring only with definite specific nominals in both numbers (call it the D-article, given its morphological form, namely di. det. det.for the three genders in the singular, plural don for all genders), the other systematically found with proper names and with all types of generic phrases, i.e. plurals, mass singulars, count singulars, and apparently substantivized adjectives (call it the A-article, after its morphological form, a, ^ in the singular, a in the plural). The corresponding examples of Ebert’s are in (151). In (151a) we have a proper name, in (151b) a generic mass singular, in (151c) a generic plural, and in (151d) a generic singular. (151) a. A Türkài The Turkey b. At weeder The weather c. Me a deensken san we leewen frinjer weesen With Danes we have always been friends d. A aapel faalt ai widj faan a buum The apple does not fall far away from the tree As pointed out by Longobardi, proper names, when they are modified by a restrictive modifier, switch to the D-article, as in (152). R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 9 5 (152) Det Moskau faan di tidj The Moscow of that time This matches our hypothesis for Spanish. The DetP article does not have always a phonological matrix in Spanish, while the DP article does, as we can observe in (153). (153) a. Madrid es precioso. Madrid is beautiful. b. El Madrid de los Austrias es precioso. Some conclusions can be drawn from these observations. First, proper names do not need to enter syntactically into a second level, the referential level, to comply with (111), while generics do, even when, as we said above, they saturate the R-variable. This resembles the behavior of middle constructions, where generic objects obliterate the semantic content of the subject but not its syntactic position, to which they have to move. Interestingly enough, Longobardi (1992:fn.55) informs us about another Frisian dialect that Ebert mentions, in which the A-article covers proper names and singular generics, and the D-article specifics and plural generic common nouns. Hence, there is some fluctuation with respect to the generic articles, which may be because their syntactic function is not supported by its R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 9 6 semantic content, in contrast with those that introduce specific common nouns. I have argued for the existence of a DetP that is the syntactic position for existential interpretations of NPs, a first-order predicate. Existential interpretations of definite singular common nouns, and proper names saturate this kind of nominal predicates. This allows them to act at a third level, as case assignment, lacking a representation with respect to countability. DP is licensed by a second-order predicate, that, in turn, is licensed by countability. Therefore, DPs select predicates that range over sets, such as spatio-temporal location and alienable possession. I have based my proposal on the distribution of wh- in-situ questions, the existence of expletive possessives in relation to referentiality, the co-occurrence of these possessives in prenominal position with other elements that can occupy DP in Spanish, the coordination facts with respect to type nouns, the possibility of deletion of a lower DetP, and finally the syntactic behavior of stage-level predicates and individual-level predicates in existential sentences. This proposal has been supported by the facts from Romanian and other typological data. The connection between N, Det, and D has been established through the behavior of expletive articles. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 9 7 The proposal yields a structure where we have a DP within another DP, a formation that resembles a "mise en abyme" structure. 3.5.3. Predication and XP-movement within DP: the "mise en abyme" effects. 3.5.3.1. Introduction In this section, I will present more evidence that supports XP-movement within DP, where Spec-of-DP is not involved. This movement is triggered by the agreement Special thanks to Daniel Garcia for suggesting this name to me and for discussing the linguistic and interdisciplinary aspects of this proposal. The label 'mise en abyme" i s used in literature to describe a smaller replica of a text embedded within that text. That text reflects one or more aspects of the textual whole. An example would be when Cervantes refers to the text by Cid Hamete Benengeli as the source for his story in Don Quixote. In this case we would say that Cid Hamete Benengeli's story is 'mise en abyme' in Don Quixote. The term comes from heraldry. A figure in an escutcheon is said to be 'en abyme' when i t constitutes a miniature of that escutcheon. In the same sense, I would like to use this term to reflect the 'stacking' effects of nominal modification. Syntactically the noun is inside of a the constituent 'noun + modifier’, and, if we have another modifier, this constituent is inside a larger one ('[noun + modifier] t modifer'), and is a syntactic replica of the larger constituent. Semantically it is the other way around, and it is the noun that has the wider denotation. This denotation is successively constrained by the modifiers, starting by the closest to the head. This reflects the universal setting of adjective ordering in natural languages. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 9 8 features of Romance languages, as I said in section 3.1. In the case of English, this movement occurs at LF. In this way, we can account for the stacked interpretation of adjectives in both langages. That is, when there are two restrictive adjectives, the one closer to the noun modifies only the noun, while the one furthest from it modifies the "noun + adjective" complex. Hence, the correct derivation for el coche rojo grande ’the big red car’ will be as in (154). (154) DP / \ el D’ / \ AdjP / \ AdjP: Adj ’ / \ / A DetP; Adj ’ grande t- 1 ^ / \ coche rojo tj In this way, with one adjective within the other, we can derive the stacked interpretation of adjectives, and agreement through the Spec-head relation. Bernstein (1993:46-47) was the first to propose this kind of movement. She proposes this derivation for what Cinque (1993) calls (predicative) adjectives, those adjectives that occur in the most external position, for instance hello in (155) (example from Bernstein, 1993:46). R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 9 9 (155) un [libro vecchio di ricette] bello ’a nice old book of recipes’ Bernstein (1993:47) proposes the following structure in (156). (156) DP / \ un AdjP / \ Spec Adj’ / I XP bello / \ libro- YP / \ vecchio YP / \ Y NP I / \ tj N PP I di ricette ti She suggests that XP has moved up from lower down. The logical position would be the complement position of AdjP. These adjectives are different from attributive adjectives. In Spanish they should be assigned contrastive stress. Bernsteins’ analysis is problematic because she does not treat the prepositional complements as attributive. Notice that we can have three adjectives in a row, if the last one receives contrastive stress, such as in (157). R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 0 0 (157) un libro ruse viejo BELLO a book Russian old nice 'a nice old Russian book’ As you can notice the adjective ordering in Spanish is the mirror image of English. We cannot derive their positions by N“-movement because it would not yield the correct order. The only way to derive the correct order from parallel underlying structures between Spanish and English would be through recursive movement, as in (156). We will come back to this in section 3.5.3.3. The other proposal of XP-movement within DP is by Sanchez (1994b) with the purpose of licensing elliptical nominal constructions. The Spec-head relation licenses null nominals, in accordance with Sanchez. Although the situation is more complex as we have seen in section 3.5.2.3.5., it is obvious that the licensing of an empty nominal is related to the Spec-head relation, but we should consider the licensing of an empty nominal as a consequence of that relation, and not the motivation for that relation. The null nominal is also subject to Identification Conditions, such as the Morphological Uniformity Condition proposed by Jaeggli and Safir (1989). As we have seen in section 3.5.2.3.5., in many cases the null nominal is only allowed if it is identified by an operator. Notice also that short R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 0 1 prenominal possessive forms are not able to identify the empty nominal, as (158) illustrates. (158) *Su blanca His white In Spanish, the paradigm of prenominal possessive forms is not morphologically uniform, since first and second person plural show gender agreement, but the other forms do not. Notice also that nominal elision is possible with a full possessive form, as (159) shows. (159) La suya blanca. The his white What is relevant in this example is that the order of the possessive with respect to the adjective is the reverse of the one with the overt nominal, as (160) shows. (160) La camisa blanca suya. The shirt white his ’His white shirt.’ If we reverse the order in (160), maintaining the restrictive reading of both modifiers, blanca must receive contrastive stress, something that does not happen in (159). I claim that in (159) the possessive occupies the prenominal position, but it has risen only to the D° position, as Italian possessives do. (161) illustrates this derivation. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 0 2 (161) DP / \ la D' / \ suya; GenP / \ AdjPk Gen’ / \ / \ Gj Adj t; t% blanca tj Hence, prenominal possessives also identify null nominals, but they have to be morphologically uniform. In order to obtain this, the possessive only reaches We can establish that once the nominal expression complies with the Licensing Condition, Spec-head relation, as proposed by Sanchez (1994b), and the Identification Condition, binding by an operator morphologically uniform, a null nominal can occur. Therefore, the licensing of a null variable is another piece of evidence for a Spec-relation between the noun and the adjective. Sanchez is also developing this analysis in her dissertation in order to explain agreement within nominal expressions by a Spec-head relation. She considers adjectives as heads. The Spec-head configuration is achieved by moving NP through all the specifiers of the Notice that la blanca suya is also correct. In this case the possessive is not raised t o O'. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 0 3 AdjPs that surface postnorainally. Under this proposal the referential status of NP is unclear, as well as its semantic and syntactic and semantic relation with its modifiers. It does not account for how a second adjective modifies a complex "noun + adjective" structure, since in her proposal the lowest adjective in the syntactic tree would be the one that appears at the end of the string of adjectives in the surface in Spanish. Furthermore, the adjective ordering parameter between English and Romance, pointed out by Lamarche, is left unexplained. These problems are overcome in my analysis by successive movement of the complement to the specifier. In summary, in order to account for agreement, the stacking interpretation of adjectives, adjective ordering parameters, and null nominals, we should argue that there is movement to Spec-of-AdjP of the XP in the immediate complement position of the AdjP. In the next section I will study the 0-relations within a DP, and, in section 3.5.3.3. I will motivate in more detail the XP-movement to Spec-of-AdjP based on the adjective ordering parameters among natural languages. This piece of evidence will allow us to confirm that the Spec-head relation between nouns and adjectives within DPs must be derived by movement from the immediate complement position of the adjective. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 0 4 3.5.3.2. 0-theory, again In previous literature, such as in Napoli (1989), AdjPs have usually been considered modifiers . To my knowledge, the only person that has defined the syntactic and semantic implications of modification is Higginbotham (1985). In accordance with him, I would like to argue that modifiers are predicative structures, with a more sophisticated semantic and syntactic structure than other predicative structures. Higginbotham distinguishes four ways of 0- assignment: 0-marking, 0-binding, ©-identification, and autonymous 0-marking. We have seen in Chapter 2 that Nouns in Spanish do not 0-raark elements that belong to their 0-grid, as verbs do. However, we are still in need of describing how projection occurs within nominal expressions. 0-binding is the mechanism that binds the referential argument of a noun as in (163) , where dog has the thematic grid in (162) (Higinbothara, 1985:560). The asterisk indicates that the position is closed at the NP level. (162) dog, -V +N, <1> R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 0 5 ( 1 6 3 ) ( N P , < * ! > ) / \ Spec (N’, <1>) the (N. <1>) I dog This can be easily translated into the DP-hypothesis as in (164). (164) (DetP, <*1>) / \ Spec (NP, <1>) (N’, <1>) the (N, <1>) I I dog Therefore the referential argument of a noun is accessible to DetP, which is able to bind it. This mechanism causes the noun to discharge its 0-role, complying with the ©-Criterion, Therefore hidden arguments such as R-arguments and E-arguments are bound by functional elements such as determiners and tense. Higginbotham (1985:564) introduces two other mechanisms of ©-assignment to explain how modification works. A subsective adjective autonymously ©-marks a noun, and ©-identifies the referential argument of the noun. Hence, an adjective would have two positions in its 0-grid, as in (165). R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 3 0 6 (165) big, +V +N, <1,2> In the usual cases of ©-marking, the reference of the element ©-marked becomes the value of an open position in the ©-marker in logical representations. In the cases of autonymous ©-marking, the ©-marked element is itself the value. (166) would be the graphic representation of the two ©-assignments described above. (166) (N', <1>) / \ / \ / /-\ (A,<1,2>) (N<1>) I________ Î The semantic representation of (165) will be (169) in a case like a big butterfly, with (167) describing point A, and (168) point N. (167) v(<x,y>,A) <--> big(x,y) (168) v(x,N) <--> butterfly(x) (169) v(x,N’) <--> v(x,N) & v(<x,'N>,A) “N is the attribute that acts as an argument for the Adj, assuming that we should make a distinction between the phrase marker and the attribute, as Higginbotham does. Notice that Higginbotham assumes that Adjs are sisters to Ns. Under the AdjP-Hypothesis we would have (170). R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 0 7 (170) v(x,A’’) <--> v(x,N’’) & v(<x,"N>,A’’) This is the case for subsective modification, where the butterfly is big for a butterfly. Conjunction defines the extensional modification. In the case of intersective modification, Higginbotham claims that autonymous 0- marking does not occur and therefore the N attribute is not incorporated into the Adj value, as in (171). (171) v(x,A” ) <-->v(x,N” ) & v(x,A” ) Finally, in the case of intensional adjectives, we should claim that the existential variable is closed in the argument representation of the adjective, as in (172). (172) pobre, +N +V, <*1, 2> Higginbotham argues that in the case of intensional adjectives only autonymous ©-marking occurs. The semantic representation of intensionality would be as in (173). (173) v(x, N’) <--> pobre('N(x)) Higginbotham distinguishes between intensional adjectives, pointing out two other cases, privative adjectives, such as fake pistol, and spatio-temporal adjectives, such as former congressman. In the first case, he suggests that N does not discharges its ©-role. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 0 8 and only the attribute has semantic significance. I do not believe that these cases need to he treated in a different way to pobre hombre or alleged communist, his example. Basically in intensional cases adjectives and nouns share the same variable, in terms of position, to identify their 0-roles. Whether we obliterate this position in the adjective argument structure, as in (172), or in the noun argument structure, it is irrelevant for our discussion. Since ©-identification does not take place, one of the two lexical constituents has its position obliterated. This position must be the one of the adjective, as in (172), since N extends its 0- grid to DetP, where the determiner binds its R-variable, but there is no evidence that the adjective extends its 0-grid to AdjP, where ©-identification takes place. As for spatio-temporal adjectives they do not constitute a case of conjunction as argued by Chierchia and McConnell-Ginet (1990). The other alternative, mentioned by Higginbotham, is Larson’s (cf. Higginbotham, 1985) proposal that in these cases neither the adjective nor the noun build these positions in their thematic grids. Overt DetPs tell us that the noun at least needs the presence of an expletive to bind its 0-role. Nevertheless, nothing especially relevant hinges on these two options at this moment. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 3 0 9 Higginbotham’s (1985) proposal accounts for the 0- assignment properties of the three main classes of adjectives that we have defined at the beginning of this chapter. In accordance with Higginbotham, there is a clear distinction between the three types of adjectives in terras of 0-assignment. Intersective adjectives use 0- binding to assign nouns its 0-role. Subsective adjectives use 0-binding and autonymous 0-marking. Finally, intensional adjectives employ autonymous 0-marking. Nevertheless, there are some points that must be clarified in terms of 0-theory and in terms of the AdjP- Hypothesis. The first one is that the attribute that nouns have has 0-relevance, but nothing is said in this theory about the adjective attribute. Also, the propositions employed in autonymous 0-marking by subsective adjectives and intensional adjectives are different. In the case of subsective adjectives the proposition is “N for the argument position <2>, while in the case of intensional adjectives the proposition is 'N(x). In addition, intersective adjectives seem to have only one argument position, something that would create many problems in terms of projection, since NP is not ©-marked. It is important to observe that closure of <1>, as in (172), should entail that the adjective cannot include R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 1 0 any existential variable in its value. Hence the correct representation for intensionality would be as in (174), assuming the DetP-Hypothesis, where the existence of the determiner entails the existential individualization of the noun. (174) v(x, D” ) <--> *A('N(x)) The value of N’’ would be as in (175). (175) v(N’') <--> "A(*N) This representation is not essentially different with respect to (173), however it is necessary to clarify it. Intersective and subsective adjectives need a further refinement in their formulations. If we reduce autonymous 0-marking to intensional and subsective adjectives we would not have a principled way to explain projection. Hence, I claim that autonymous 0-marking also occurs in the case of intersective adjectives. Then, what is the difference between the two kinds of adjective phrases? The difference is that the attribute denoted by the noun ranges over the attribute denoted by the adjective in the case of subsective adjectives, in the sense of (176). If we compare subsective adjectives to intensional ones, we can say that subsectives adjectives are a case of reverse R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 1 1 intensionality, where the noun modifies the reference of the adjective. (176) For a type A(x), x is big. Therefore, (169) and (170) may be accurate in order to represent subsective modification, since the reference denoted by the noun fills a thematic position in the adjective 0-grid. The question is whether this notation is sufficient. Assuming that 0-marking occurs under government, (177) would be problematic for the definition of intersective modification. (177) Un coche americano pequeno. A small American car. This sentence means that "for being an American car, the car is small." Therefore, the relevant attributes for pequeno are the ones denoted by the noun and the relational adjective. This would create problems for any theory based on N“-movement, since the attribute denoted by americano would be lost. The noun would be 0-marked only once, in accordance with the 0-Criterion, by pequeno through autonymous 0-marking. However, since pequeno cannot 0-mark americano because it does not govern it, its attribute cannot fill a position in the 0-grid of pequeno. In a similar fashion, the AdjP-Hypothesis has R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 1 2 the same problems, given the underlying structure in (178). (178) DP / \ un AdjP / \ pequeno AdjP / \ americano NP coche In the case of the AdjP-Hypothesis, the attribute of the Noun is the one that cannot fill a thematic position in the 0-grid of pequeno. However, the AdjP presents an easier solution to this problem, if we propose that an intersective adjective also autonymously 0-marks the noun, which in terms of projection seems to be a feasible solution. Then, the attributive denoted by the noun can fill a thematic position in pequeno by transitivity. The problem that remains is how to derive the difference in interpretation between intersective and subsective adjectives. I propose that there is a merging between the two positions. Therefore, (179) would be the value of an intersective adjective, and (180) would be the representation of a subsective adjective. (179) v(<x, y> A’’) <--> v(<"A(x), "N> A’’) (180) v(<x, y> A’’) <--> v(<"A(x), *N> A’’) --> (merging) --> v(<"N('A(x))> A’’) R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 1 3 In this case we can dispense also with 0-identification for attributive adjectives and claim that there is only one existential variable, shared by the noun and its attributive adjectives. The cases where merging does not occur, i.e. intersective adjectives, must be interpreted as a conjunction between the two attributes or more that range over the existential variable. The problem of the existential variable is connected to the problem of the extension of the 6-grid of the noun. The 0-grid of a noun is extended through autonymous ©-marking and by head-to-head movement of the existential variable. This allows the adjective to absorb the ©-grid of the noun, i.e. the existential variable and the attribute, as in (179) and (180). I claim that phrasal stress is the linguistic manifestation of the existential variable, and it is borne by the most restrictive element. Therefore, through extension of the ©-grid of the predicate, ©-roles are assigned compositionally while phrase markers build up, as proposed by Bowers (1993). PP modifiers, as we have seen in Chapter 2, receive their ©- role through the preposition since the noun is not able to assign them. The noun does not govern them, given that the noun, unlike the verb, does not move head-to-head in Spanish. A process of ©-identification must occur between the PP modifier and the head of the maximal projection R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 1 4 under which it is projected, at least, in the case of those that bind the R-variable of the noun as affected arguments. This head is adjectival, hence, it has the features [+V, +N]. 0-merging explains why subsective adjectives appear higher in the tree than intersective adjectives. Otherwise, the attribute of intersective adjectives could not be absorbed by the 8-grid of subsective adjectives and 8-merging would not occur. This account allows us to give only one lexical entry to an adjective that can have an intensional, an intersective, and a subsective interpretation, for example grande ’big, great’. In the case of the intensional interpretation, position 1 is closed. In the case of the subsective interpretation, 6- merging occurs. In conclusion, I take the cases of modification to be cases of referential predication in the sense of Fiengo and May (1934). Thus, there is only one index relevant to the Binding Theory. 3.5.3.3. Adjective ordering restrictions Hetzron (1978) and Sproat and Shih (1988) notice that there is the same relative ordering of the different kinds of adjectives across languages in relation to N°. This ordering is given in terms of head-proximity. Sproat R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 3 1 5 and Shih (1988:568) propose the following generalization in (181). (181) Restrictions on the ordering of multiple adjectival modifiers obtain iff the adjectives involved are hierarchical direct modifiers. Sproat and Shih (1988:571) define direct modification as the process by which the adjective 0-marks the noun directly through autonymous 0-marking or 0-identification and occurs within the scope of DP. If my analysis is correct, only 0-marking and 0-binding can occur within DP. The indirect modifiers may modify DPs. This difference corresponds to the more traditional one between attributive adjectives and predicative adjectives. Waugh (1977:84-85) characterizes the difference in terms of old versus new information. Predicative adjectives give new information, while attributive adjectives give old information. Waugh (1977) also noticed that predicative adjectives seem to be more independent than attributive adjectives. This insight is captured by Sproat and Shih’s analysis, where predicative adjectives are external to attributive adjectives, and modify DPs. Their analysis is based on Mandarin Chinese modification facts. Modifiers in Mandarin Chinese can appear with or without the particle Modifiers without de are attributive while with ^ they are predicative. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 1 6 Therefore, we expect that prototypical intensional adjectives, such as ’former’ or ’fake’, cannot appear with This is borne out in (182). (182) a. qian (*de) zongtong former DE president b. wei (*de) yao fake DE medicine The scale that they propose is similar to the one mentioned above in (47b): quality > size > shape > color > provenance. They observe that there is a distinction between absolute adjectives (shape, color, and provenance) and relative adjectives (quality and size). Any alteration in the natural order of two adjectives belonging to different classes is much worse than a change in the order of two absolute adjectives or two relative adjectives, as we can observe in (183). The absolute class matches the class of intersective adjectives and the relative class fits the subsective class. In 3.5.3.1. we relate this difference to different theta-marking properties. (183) a. El coche rojo bonito The car red beautiful The beautiful red car b.*El coche bonito rojo R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 1 7 (184) a. El coche japonés rojo The car Japanese red The red Japanese car b.??El coche rojo Japonés^^ In (183), we can observe that a color adjective belonging to the absolute class must be closer to the noun than a quality adjective belonging to the relative or subsective class. Otherwise, if the order is reversed, the sentence is ruled out, unless we give an appositive or predicative interpretation to rojo. On the other hand, when the two adjectives belong to the same class and we reverse the order, as in (184b), the sentence is better. Probably, what we have in (183b) is a violation of 6-marking conditions and selectional restrictions, while in (184b) only selectional restrictions are violated. In (183b) roJo cannot be interpreted as intersective since 0- merging has occured before, in order to obtain the interpretation of bonito. In (184b), only selectional restrictions seem to be violated, but the intersective reading for both adjectives is obtained. As we pointed out before, these selectional restrictions seem to be related to the alienability of the property. The most inalienable property should be closer to the noun. noticed by Lamarche (1991) for French, (182b) and (183b) are ruled out with neutral intonation. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 1 8 Lamarche (1991) points out several problems for an analysis of adjective placement based on N°-movement, as the one proposed by Cinque (1993), Valois (1991), and Bernstein (1993) for Romance languages. If the serialization of the attributive adjectives given by Sproat and Shih is correct, and, as we have seen, the generalization can be extended to Spanish, then we expect that the order of the adjectives from left to right would be the same if the noun moves head-to-head and the underlying position is the same one for English and Romance, as (185) shows. (185) a. English (no N°-movement): Adjj Adj% N b. Romance (N°-movement): Adjj Adj% N t : ___ In contrast, the surface order in Romance is as in (186), as we can observe in examples (183), (184), and (187). (186) N Adjg Adjj (187) a. A big red car. b. Un coche rojo grande. Cinque (1993:8-9) answers to Lamarche’s observation noticing that two adjectives cannot precede a PP complement after the noun, as we can observe in (188a). He assumes that attributive adjectives precede PP R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 3 1 9 complements and predicative adjectives follow them, as in (188b). (188) a.*Una macchina rossa bellissima da A car red very-beautiful for A very beautiful red car for racing. corsa. racing Una macchina rossa da corsa bellissima. A car red for racing very- beautiful Una macchina da corsa rossa bellissima A car for racing red very- beautiful Una macchina da corsa bellissima A car for racing very-beautiful rossa red However, these facts need a closer analysis, bellissima is an adjective which is more naturally interpreted as predicative. Using Spanish facts and an adjective that does not show any tendency to be interpreted either as attributive or predicative, such as grande ’big’, we will check the interpretations of the two possible orders given in (189), with and without neutral intonation. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 2 0 (189) a. Tienes que lavar el coche rojo grande. You have to wash the car red big A. You have to wash the big red car. B. You have to wash the red car that is big. b. Tienes que lavar el coche grande rojo. You have to wash the car big red A.*You have to wash the big red car. B. You have to wash the big car that is red. In (189a), the interpretation of grande can be attributive or contrastive. In the first case, the addressee knows the car and its attributes. In this case, grande does not need to range over a group of red cars. In the second interpretation, grande has a commitative function in the sense of Waugh, even though we cannot consider this adjective as a pure predicative adjective, since it occurs within DP. The addressee must make the connection between the noun and the adjective. In this case, grande ranges over a group of red cars. On the other hand, in (189b), only a contrastive interpretation is possible. The speaker is giving the addressee enough information so he can find out which car he should wash. The speaker is identifying the unknown car for the addressee. In this case, the only interpretation is that rojo ranges over a group of big cars. Lamarche (1991;fn.5) argues that the order is reversed when the speaker wants to focus on the last R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 2 1 canonical adjective. In (189b), it would be the size adjective. He gives the following example: "in a roadblock where the police are looking for two cars, one white, and the other red (without knowing their origin), one policeman, reporting to his superior, could say something like..." (190). (190) Nous avons intercepté une voiture rouge américaine, une voiture rouge italienne, et deux voitures blanches allemandes. The English counterpart would be (191), and the Spanish counterpart would be (192). As we can observe, Spanish parallels the French order, while English mirrors the Romance order. (191) We intercepted one American red car, one Italian red car, and two German white cars. (192) Hemos interceptado un coche rojo americano, un coche rojo italiano, y dos coches blancos alemanes. These examples support both Lamarche's and my own analysis. However, I maintain that the relevant characterization of the adjective ordering in (190), (191), and (192) is that the outer adjective in relation to the noun carries new information for the addressee, in this case the policeman’s superiority. Hence, it is the one that is on focus. Lamarche himself gives the same R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 2 2 kind of explanation to explain the difference between (193a) and (193b) for English, and (194a) and (194b) for French. (193) a. a handicapped elderly person b. an elderly handicapped person (194) a. une personne âgée handicapée b. une personne handicapée âgée The interpretation of (193a) and (194a) is that the person is an elderly person who has a handicap, while (193b) and (194b) express that the person is a handicapped person who is elderly. Hence, in addition to predicative adjectives, which we can also call adverbial,we have attributive adjectives that can be contrastive or non-contrastive. The point here is that the order between English and Romance seems to be reversed in both cases, giving rise to the so-called "mirror image". I claim that the way to derive the order in (186) is by sucessive movement to Spec-of-AdjP, as in (154), repeated here in (195). For instance they can be left behind as In (1). ( 1 ) El coche lo compro precloso. The car It be-bought beautiful He bought a car that Is beautiful. R eproduced witti permission of ttie copyrigtit owner. Furtiier reproduction protiibited wittiout permission. 3 2 3 (195) DP / \ el AdjP, / \ AdjP,,j Adj ’ / I \ coche: Adj ' grande t- J / \ rojo tj As we can observe in (195), first the NP moves to Spec- of-AdjP^ and, then, the whole AdjP% moves to Spec-of- AdjPj. In this way, we obtain the correct word order, and we can also explain the "stacked" modification of the two adjectives, where rojo modifies the noun, and grande modifies rojo and the noun. We can call this the "mise en abyme" effect. In this way we can explain also agreement since, except for DP, everything enters in a Spec-head relation. I claim that the same kind of movement occurs at LF in English to comply with the requirements on predication. Now I will explain the nature of the configuration for the contrastive modification structure. Sproat and Shih’s (1988:567) proposal for what they call indirect modification is that the adjective assigns its 0-role to a variable coindexed with the noun, as in (196), assuming a relative clause type of structure. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 2 4 ( 1 9 6 ) / \ , CP N;’ ' / \ ’ ’ IP 0; / \ 0-role However, if we assumed this structure it would be difficult to explain how agreement occurs in Spanish, without positing the existence of an anaphor or a pronoun in Spec-of-IP which would give rise to an i-within-i violation. In addition to this, there are some differences between the Chinese indirect modifiers and the Romance contrastive modifiers. As we have seen above, intensional adjectives, mapped onto the canonical pronominal position in Romance, cannot occur under the indirect modifiers structure. On the other hand, as pointed out by Klein-Andreu (1983) these kind of adjectives can occur postnominally in Spanish, as we will see below. Further, if we want to keep the explanation for the mirror image parameter between Romance and English, we should argue that, when a focused modifier appears, the different order results from different base generated positions. English shows the underlyingly order, while Romance shows a different ordering because R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 2 5 movement to Spec-of-AdjP has occurred, as in (154), repeated here under (197). (197) DP / \ el AdjP, / \ AdjP,,I Adj ’ / \ / \ coche- Adj’ rojo t- J / \ grande tj Nevertheless, the interpretative difference between (189a) and (189b) still remains unexplained. We should consider also that there is an intonational difference. Notice that the last adjective in both sentences occupies the most embedded position, hence, the unmarked focus position, in accordance with Cinque (1993). Phrasal stress is assigned by the phrasal stress algorithm. Nevertheless, in (189a) this stress does no need to be contrastive. The adjectives are distinctive but not contrastive in this case. We should understand contrastive in the sense of Klein-Andreu (1983:154), where contrast does not always entails distinction between referents. In Klein-Andreu contrast is understood either as distinction between referents and also "as distinguishing the characterization itself, from what we might call ’possibilities to the contrary’." Klein-Andreu argues that this property allows prototypical prenominal R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 2 6 adjectives to appear postnominally with the canonical prenominal meaning, as we can observe in Klein-Andreu’s example in (198). (198) Trata, ante todo, del propio Malraux, hasta el Above all it deals with Malraux himself, to punto de que Espoir. como pelicula, como the point that Espoir. as a film, as pure and relato puro y simple, es, incluso en sus simple narration (NAA), even as regards its aspectos argumentâtes, indisociable de la vida plot, is inseparable from the life and the y personalidad del escritor francés. (Cambio the personality of the French author. 16, 345:52) We should notice, as pointed out to me by Saltarelli (p.c.), that the postposition of these kind of adjectives with [+ animate] nouns is more restricted. For instance, we cannot postpone pobre with the sense of 'pitiable'. This adjective is normally associated with [+ animate] nouns. Further, the canonical position of falso 'fake' is different depending on whether the noun is animate or not, el falso policia 'the fake policeman' against el oro falso 'the fake gold'. In this case, we can postpone falso but only if it receives contrastive stress, el policia FALSO. as argued by Klein-Andreu, otherwise falso would mean 'untruthful'. The difference must be related R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 2 7 to agentivity. The existential variable might have some features like agentivity that allows it to appear in two different dimensions with respect to modification. One as an agent, as is the case of falso ’untruthful’, and the other one with respect to the attribute denoted by the noun, as in ’the fake policeman’. In the case of pobre, the modification seems to be restricted only to the attribute denoted by the noun. Why can pobre not appear in the contrastive postnominal position? I believe that the reason is related to the fact that pobre functions more as a degree word of the attribute denoted by the noun than as a real modifier. Hence, the analysis by Klein-Andreu is basically correct. The problem pointed out by Saltarelli remains. Why is the canonical position of falso postponed to the noun with [- animate] nouns, if it also modifies also the attribute denoted by the noun, as falso does in el falso policia? The answer must be related to a different way of mapping between the attribute and the existential variable in contrast with the agentive nouns. This case needs further research. In summary, in order to account for the adjective ordering parameters we need to posit that the spec-head relation between DetP and AdjP must be derived by DetP- movement to Spec-of-AdjP from the complement position. Following this analysis, I have tried to capture R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 2 8 Bolinger’s generalization that states that the adjective that is furthest from the noun is the most restrictive one, and Lamarche’s observation that there are also adjective ordering restrictions in Romance Languages. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 2 9 Chapter 4 Extraction from NPs in Spanish 4.1. Introduction Several studies within the Government and Binding framework have taken as their object the extraction from NPs in recent years. It is clear from this literature that the different asymmetries observed in extractions from NPs must be explained in accordance with the way in which the Determiner Phrases and the Noun Phrases are internally structured and in accordance with the position that they occupy in the matrix sentence. 4.2. The data 4.2.1. Cinque’s generalization In this section, we will consider the apparently bizarre asymmetries observed in extractions from VPs as opposed to extractions from NPs. In particular, we attempt to identify which module of grammar accounts for extraction from VPs and extraction from NPs as implied in Cinque’s (1980) descriptive generalization based on evidence from Italian (1): R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 3 0 (1) One can only extract a di+NP phrase from an NP if the phrase can also surface as a genitive subject (a possessive pronoun). The genitive subject is the hierarchically highest element projected in the syntax according to (28) in Chapter 1, as proposed by Giorgi and Longobardi (1991:68), repeated here under (2). (2) Possessivization Principle The only phrase allowed to appear as a possessive is the hierarchically highest genitive argument of an NP. This syntactic hierarchy is expressed in (25) in Chapter 1, and repeated here in (3). (3) Possessor > agent > internal argument Cinque’s generalization holds also in Spanish as sentences (4) to (8) show. As demonstrated in the literature on Spanish (Torrego 1985; Demonte 1987), the internal argument cannot be extracted in the presence of a possessive constituent, as we can observe in (4). (4) *&De que tema. viste varios of which subject(object) did-you-see several ’Of which subject did you see several pictures cuadros t- del coleccionista? pictures t of-the art collector(possessor) of the art collector?’ R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 3 1 On the other hand, extraction of the internal argument is possible when a possessive constituent is not present, as (5) illustrates: (5) iDe gué tema; viste varios of which subject(object) did-you-see several ’Of which subject did you see several cuadros t;? pictures t pictures?’ Extraction of an agent is possible either in the presence or in the absence of an internal argument, as illustrated by (6) and (7). (6) iDe gué pintorj viste varios by which artist(agent) did-you-see several ’By which artist did you see several pictures?' cuadros t.? pictures t (7) iDe gué pintor. viste varios by which artist(agent) did-you-see several ’Ey which artist did you several pictures of cuadros de Londres t;? pictures of London(object) t London?’ However, extraction of an object is less acceptable when the agent is present, as observed in (8):^**^ 107 There is some variation with regard to (8 ) among native speakers. As pointed out to me hy Zubizarreta (p.c.), this could be due to the reinterpretation of the preposition^ 'of that introduces the agent as por 'by'. In Spanish por cannot introduce agents projected by result R eproduced witti permission of ttie copyrigtit owner. Furttier reproduction protiibited wittiout permission. 3 3 2 (8) ??i,De que temaj viste varios of which subject(object) did-you-see several ’Of which subject did you see several cuadros tj de Monet? pictures t by Monet(agent) pictures by Monet?’ Finally, we should note, as pointed out in Giorgi and Longobardi (1991:61) for Italian, that an adjective denoting an agent blocks extraction and possessivization of the THEME, as (9) illustrates. (9) a. Criticaron la investigaciôn araericana de este ’They critized the American investigation of tema. this subject.’ b.*Criticaron su investigaciôn americana. they-criticized its investigation American ’They criticized its American investigation. ’ c.*i,De gué tema criticaron la of which subject they-critized the ’Of which subject did they criticize the investigaciôn americana? investigation American? American investigation?’ Dominais, as (i) Illustrates. (i)*el cuadrc per Monet the picture by Monet Therefore, we can assume that in this context, may have the two interpretations. Under the interpretation of 'by' the agent would not prevent extraction of the theme. R eproduced witti permission of ttie copyrigtit owner. Furttier reproduction protiibited wittiout permission. 3 3 3 (9b) shows how the possessivization of the THEME is not possible in the presence of a referential adjective, and (9c) shows that extraction is also impossible in the same kind of context. On the other hand, if the agent is expressed by a por-phrase neither possessivization nor extraction is blocked, as (10) illustrates. (10) a. Criticaron la investigaciôn de este tema por ’They critized the investigation of this los americanos. subject by the Americans.’ b. Criticaron su investigaciôn por los they-criticized its investigation by the ’They criticized its investigation by the americanos. Americans Americans.’ c. i,De que tema criticaron la of which subject they-critized the ’Of which subject did they criticize the investigaciôn por los americanos? investigation by the Americans? investigation by the Americans?’ Therefore the generalization in (1) also holds for the cases where the agent is expressed by a bare adjective or by a por-phrase. In the first case, neither possessivization or extraction of the internal argument are allowed. In the second case, both are permitted. If we compare NP extraction with VP extraction, it is well known that any constituent can be extracted from R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 3 4 a VP in the presence of a hierarchically higher constituent without violation of any Principle of Grammar or Condition. One might wonder, for example, why it is possible to extract an object from within a VP in the presence of a subject, but not possible to do the same thing from within an NP. It seems that some coindexation or agreement in certain features with DP is necessary in the case of NPs. The fact that VPs have a special functional category where wh-phrases land to trigger wh- agreement (CP in standard analyses, AGRP2 in Martin (1992)), while NPs need to use the coindexation of the structurally highest argument projected by them with D° to trigger wh-agreement with the predicator. 4.2.2. Specificity effects Another well-known factor that intervenes in extractions from Spanish NPs is the so-called "specificity effects".^**® "Specificity effects" appear to block extraction in wh-questions and restrictive 1 will define specificity in terms of predication. If the NP is a saturated predicate, it implies that its semantic value is unique; then, the NP is specific. If the NP is an open predicate, i . e . has not been saturated and, hence, its semantic value is variable, then, the NP is non specific. R eproduced witii permission of ttie copyrigtit owner. Furttier reproduction protiibited wittiout permission. 3 3 5 relative clauses in (11) through (14),^^ but not in nonrestrictive relative clauses, as observed by Giorgi and Longobardi (1991) and Torrego (1988), in (15). (11) a.*&De que autor leiste este libro? by which author did-you-read this book? ’By which author did you read this book?’ b. iDe que autor leiste varios by which author did-you-read several ’By which author did you read several libros? books? books?’ (12) a. *i,De que ciudad viste estas fotos? of which city did-you-see these pictures? ’Which city did you see these pictures of?’ b. i,De que ciudad viste varias of which city did-you-see several ’Of which city did you see several fotos? pictures? pictures?’ (13) a. *&De quién lavaste estos pantalones? of whom did-you-wash these pants? ’Of whom did you wash these pants?’ b. i,De quién lavaste varios pantalones? of whom did-you-wash several pants? Fiengo and Higginbotham ( 1 9 8 1 ) state the following in order to account for the Speclclflclty Effects in English. ( 1 6 ) » . . . X . . . if X is free in a specific NP. As we will see in section 4 . 3 . 2 . Specificity effects can be explained by more general principles of the grammar, such as Principle A of the Binding Theory (Hudson, 1 9 8 9 ) . R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 3 6 (14) *Describirâ una ciudad de la que lea he-will-describe a city of which he-reads ’He will describe a city of which he will read este libro. this book this book.’ (15) Cervantes, de quien, por cierto, hemos leido Cervantes, by whom, by the way, we-have read ’Cervantes, by whom, by the way, we have read este articule durante el semestre, muriô this article during the semester, died this article during the semester, died poor.’ pobre. poor. As Torrego (1988) points out wh extraction is constrained by the definite nature of the determiner. This means that a determiner which is high in definiteness is able to block any Wh movement from within NPs. (11a), (12a), (13a), and (14) show how the extraction of a subject or an object is blocked because of "specificity effects". In (11a), (12a), (13a), and (14) the Determiner Phrase is headed by a demonstrative, este, estas. estos. Demonstrative terms are indexical elements high in specificity. In contrast, varios ’several’ is low in specificity and allows extraction as (lib), (12b) and (13b) show. Can we establish any connection between the data in this section and those of the former section? Notice that both, a possessive element and a deictic element high in R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 3 7 specificity can appear postponed to the noun. The natural ordering between them is the possessive before the deictic element, under normal conditions of intonation, as (16) illustrates. (16) a. Juan leyô el libro tuyo aquel. Juan read the book yours that ’Juan read that book of yours.’ b. Juan leyô el libro aquel *tuyo/TUYO. Juan read the book that yours/YOURS Therefore, we can consider that demonstrative elements are higher than possessive elements in the determination scale. This observation for Spanish matches other universal scales proposed, such as Seiler’s (1978:318). Notice also that, as pointed out in Torrego (1988:39), while the demonstrative can appear in the prenominal position in the presence of the possessive adjective, as (17a) illustrates, the reverse situation is ruled out in (17b).^^ See chapter 3 for an analysis of the exceptional case of the affective possessive that can appear in prenominal position after a demonstrative, as in ( i ) . ( i ) esta su casa this his house Therefore, structures such as tu libro este are fine when the possessive receives an affective interpretation. In this case, tu libro is in spec-head relation with the demonstrative. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 3 8 (17) a. Este libre tuyo. this book yours ’This book of yours.’ b.*Tu libro este/este tu libro/tu este your book this/this your book/ your this libro book As noticed by Torrego (1988:fn. 22), possessivization is possible with a determiner that allows extraction, as her examples in (18) show. (18) sus muchos/varios/tres retratos ’his many/several/three portraits’ Therefore the generalization in (1) also takes care of the specificity effects. A "^+NP" phrase cannot be either possessivized or extracted unless it is the highest element in the scale of determination. Since the presence of a subject and specificity induce the same effects, that is, blocking of extraction and possessivization, it would be desirable to give a unified account for them. 4.3. Previous studies 4.3.1. The Specifier Hypothesis: Cinque (1980) and Giorgi and Longobardi (1991) As we have seen in section 4.2.1., the generalization proposed by Cinque (1980) states that an R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 3 9 element from within a nominal expression cannot be extracted unless it counts as a subject. Two requirements define the term subject in this proposal. The first one is to receive genitive case. This is illustrated by Cinque’s (1980:47-48) examples in (19). (19) a. Una persona [ppdi cui] apprezziamo [%pla ’A person of whom we appreciate the great grande generosità t] (è Giorgio), generosity (is Giorgio).’ b.*Il paese [ppa cui] ricordiamo [|^pun/l ’attacco ’A country on which we remember an/the t] (è la Polonia). attack (is Poland). In (19a) an element that has received genitive case through the preposition 41 Is allowed to be extracted. (19b) shows the case of an element that has not received genitive case, meaning that it is not a PP of the form di-NP. Since that element has not received genitive case, it does not qualify as a subject and the sentence is ruled out in accordance with the descriptive generalization in (1). The second requirement for a subject is to be the highest element in structural terms within the nominal expression. Examples (4) to (8) take care of this point in section 4.2.1. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 4 0 Cinque accounts for the generalization in (1) stating that extraction from within nominals is subject to the Specified Subject Condition. The SSC states the following : (20) Specified Subject Constraint No rule may relate X and Y in the structure . . .X. . . [ . . . Z ... W. Y W, . . . ] where Z is the subject of Wj Y Wg. However, wh-traces are not subject to principle A of the Binding Theory within the OB-framework^^ (Chomsky, 1980), as pointed out by Rizzi (1982) in his analysis of the violations of the Wh-Island Constraint in Italian. Longobardi in his 1987 paper, later included in his book with Giorgi, tries to redefine this hypothesis. His proposal is based on the definition of Spec of NP, Spec of DP under Abney’s hypothesis, as an escape-hatch for extraction, and on the necessity of every trace to be properly governed. Therefore, the intermediate trace in 1 1 0 Spec-of-NP must be governed, and extraction from NPs But see Aoun ( 1 9 8 5 ) for an account of extraction from NPs within the Generalized Binding framework. 112 The definition of head government proposed by Giorgi and Longobardi ( 1 9 9 1 : 1 0 8 ) is the following: (i ) a . a head a governs B iff R eproduced witti permission of ttie copyrigfit owner. Furtfier reproduction profiibited witfiout permission. 341 11 n must show some Condition on Extraction Domains effects. This constraint blocks extraction from subjects and from adjuncts, i.e. out of a category that is not lexically governed. As we have observed in (5), (6), and (7), extraction from an internal argument is allowed. (21) illustrates the case of extraction of an NP governed by an ergative verb. Since the NP is lexically governed, the prediction is that the sentence will be wellformed, as it is indeed the case. (21) i,De quién desapareciô la foto? of whom disappeared the picture? On the other hand, extraction is not allowed from either postverbal subjects or adjuncts, as (22) shows. Relevant to the first step of extraction from NPs is extraction ( a ) i . a is intrinsically a structural governor | | + V | universally, also P I n English) or II. a s-selects B or I I I . a and B are coindexed and (b) a m-commands B and (c) (Minimality Condition) There Is no | i , | i Including B and excluding a , such that I . 1 1 Is a non-internal argument of a or II. B Is canonically contained or lexically C-marked In | i Therefore, a verb Is able to govern Into the Specifier position of Its complement. Cf. Huang ( 1 9 8 2 ) . R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 342 out of a postverbal subject, that is ruled out either in the case of the subject of a transitive, as in (22a), or in the case of an unergative verb, as in (22b). (22) a.*&De quiénj ha ganado el premio Planeta [un ’Of whom has a book won the Planeta prize. ’ libro tj]? b.*&De qué| asesino ha llamado [un complice ’Of which murderer has phoned an til? accomplice.’ Head government of the intermediate trace in Spec of DP/NP is required. Thus, Longobardi’s proposal is based on a two-step process, assuming that the only element that can occupy the Specifier is the element that has been assigned the feature [Poss], in accordance with the Condition on Possessivization in (2). Condition in (2) is superior to the opacity account in order to explain cases such as (24a), in accordance with Giorgi and Longobardi (1991:72) . (23) a. La probabilité di una guerra appare molto ’The probability of a war appears very low. ’ bassa. b. La probabilité che ci sia una guerra appare ’The probability that there will be a war molto bassa. appears very low.’ R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 4 3 (24) a. Una guerra, di cui non so valutare la ’A war, of which I cannot evaluate the probabilité, sarebbe catastrofica. probability, would be catastrophic.' b.*Una guerra, che ci sia la quale non so ’A war, that there will be which I cannot valutare la probabilité... evaluate the probability...’ In (24a), as predicted by the Condition on Posessivization in (2) and the Opacity-Hypothesis, extraction is possible. Since di una guerra is the only ©-role projected by probabilité, it is the highest element able to be possessivized, in accordance with (2). Hence, it can occupy the Spec-of-NP/DP position and be extracted. With regard to the Opacity-Hypothesis, the fact that di una guerra is the only role assigned allows us to predict that no subject can induce opacity. In contrast and according to Giorgi and Longobardi (1991:72), the Opacity-Hypothesis predicts that (24b) should not be ruled out since there is no subject blocking extraction. On the other hand, (2) correctly predicts (24b) to be ruled out, since a sentential complement cannot be possessivized. This fact is demonstrated by the impossibility of a possessive relative anaphor cui to express a sentential complement, as shown in (25) (Giorgi and Longobardi, 1991:72). R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 4 4 (25) a. Una guerra, la cui probabilité è molto ’A war, whose probability is very low, would bassa, sorprenderebbe tutti, surprise everyone.’ b.*Che ci fosse una guerra in Italia, la cui ’That there could be a war in Italy, whose probabilité è molto bassa, sorprenderebbe probability is very low, would surprise tutti. everyone.’ However, we must recall that Cinque (1980:49) explicitly says that "only PPs of the form + NP] can be extracted." Thus, in this case. Cinque’s analysis accounts for (24b). Actually, Longobardi’s analysis might run into problems if we consider the Spanish data. Recall that in section 2.3.2.1. we conclude that some sentential complements need to be 6-marked in Spanish by the preposition ^ ’of’, as shown in (26a).As in (26b), extraction of a sentential complement is possible in Spanish. I believe that a possible explanation of these facts must be b a s e d on the different referential properties of the complementizer aue in Spanish, in contrast with other languages such as Italian. In Spanish the complementizer can be introduced by a definite article, as in ( i ) . (i ) E l que lloviera seria catastrofico. The that it-could-rain would-be catastrophic 'That it could rain would be catastrophic.’ R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 4 5 (26) a. La posibilidad *(de) que haya una guerra es ’The possibility that there will be a war is muy baja. very low.’ b. Que haya una guerra, de lo que no sé valorar ’That there will be a war, which I cannot la posibilidad, seria catastrofico. evaluate the possibility, would be catastrophic.’ Nevertheless, the possessivization of a sentential complement is ruled out, as the impossibility of coindexing the sentence with a possessive relative anaphor shows in (27). (27) [Que haya una guerra, ]*:, cuya,,*, posibilidad no ’That there will be a war, whose possibility I sé valorar, seria catastrofico. cannot evaluate, would be catastrophic.’ The judgements in this case are very subtle, but notice that an extracted clause without a participant involved in the event would not be possible, as (28) shows. (28)* [Que lluevajj, cuya, posibilidad no sé ’That it will rain, whose possibility I cannot valorar, evaluate, ...’ These facts confirm our analysis In section 2 . 2 , 2 . 1 . Spanish nouns theta-iark sentential complements, but do not assign genitive case t o them. R eproduced witli permission of ttie copyright owner. Further reproduction prohibited without permission. 3 4 6 If this analysis of the data is correct, assignment of genitive case is not a requirement for an element extracted from NP, contrary to Longobardi’s proposal (Giorgi and Longobardi, 1991:73). The possessivization condition for extraction in Spanish is too strong, and to propose 6-marking only as a pre-requirement on extraction would overgenerate, since in Spanish there are the same constraints as in Italian for NPs introduced by prepostions other than de/di. as the Spanish counterpart to (19b) shows. (29)*E1 pais al que recordamos un ataque es Polonia. ’A country on which we remember an/the attack is Poland. Taking into consideration these and the specificity facts, (30) would be a more accurate generalization for Spanish. (30) One can only extract a ^+NP from an NP if it is a syntactic argument and is the highest element in the determination scale, (to be revised) The notion argument is crucial in the definition of the generalization in order to bar some extractions, as pointed out by Godard (1992:fn. 6). As noticed by Ruwet (1972) for French and Cinque (1980) for Italian, only j^+NP elements can be extracted in Spanish. However, this R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 4 7 is not completely explanatory, since some ^+NPs cannot be extracted, as pointed out by Godard (1992:238-239) for French. The first case is N complements indicating time, location, cause, means, measurements, material or quality. I will center my analysis on the material complements, but let us first see Godard’s examples in (31) . (31) a.*Je ne connais pas le bois clair dont cette I do not know the light-colored wood table égayait la pièce. of-which this table brightened the room b.*Il arrive de Marseille dont/d’où il a he is coming from Marseilles of/from-which recontré un médecin, he met a doctor c.*C’est l’église près de laquelle le café est it is the church near which the café is toujours plein, always full As we have seen in (22), (31a) is supposed to be ungrammatical since it involves extraction from a subject position. In section 4.4 we will see that extraction out of a subject is fine in the case of nonrestrictive clauses. Nevertheless, extraction of a material complement is disallowed as shown in (31a) and the Spanish sentence in (32a). However, notice that (32b) and R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 4 8 (32c) are marginally acceptable under an interpretation where the wood of the chair is seen, somehow, as independent from the wood that we are referring to. A possible acceptable pragmatic context for these sentences would be in front of a display of different types of wood. In this case, we are relating two different entities. On the other hand, in (32a) the demonstrative does not allow this reading. (32) a.*Esta raadera, de la que compré una silla, . . . ’This wood, of which I bought a chair, . . . ’ b.?La madera de la que compré una silla es ’The wood of which I bought a chair is this esta, one. ’ c. ?iDe que madera corapraste una silla? ’Of which wood did you buy a chair?’ The Spanish version of (31b) is also bad, but some pragmatic effects could be at work. (33) shows that extraction of a locative complement is allowed in other contexts. (33) Marsella, de donde conozco un medico, ’Marseilles, from which I know a doctor, . . . ’ Nevertheless, in (33) the interpretation of the relation between the noun and its complement is different. In this case the only possible relation is one of provenance. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 4 9 Notice that the doctor is from Marseilles, but this does not mean that he has to be part of Marseilles anymore. He can be living anywhere else. Notice that if we neutralize the pragmatics effects in (31b) the sentence becomes good as in (34). In this case, the relation is one of provenance. (34) El llegara de Marsella, de donde he is coming from Marseilles from where conociô un medico antes de salir, he-met a doctor before leaving In (34), it is clear that he did not meet the doctor in Marseilles. Hence, extraction of a location N complement is not licit, but extraction of a provenance complement is correct. Finally, the element extracted in (31c) is introduced by the preposition près and not by the preposition 1^2 , but as we have seen in (35) extraction of a locative complement is disallowed also, but not one with the relation "part-of". Obviously, we cannot have the provenance relation in this case. (35) La plaza, [de la que]/*[de donde] estân The plaza, [of which]/[of where] they-are pintando el bar, ... painting the bar. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 5 0 Finally, let us check the facts with another kind of complement, the time complement. As we can observe in (36), extraction is also barred. (36)*iDe cuando ano compraste una revista? ’Of which year did you buy a magazine.’ None of them can undergo possessivization as shown in (37). " V u/7 3. • 5 Ü lîîâQc r âr) s jrxxct ’its (=wood) chair’ b.*su (=Marsella) medico ’its (=Marseilles) doctor’ c.*su (=ano) revista ’its (=year) magazine’ On the other hand, when the relation involves provenance from space or time, as in (38) and (39) respectively, or "part-of", as in (40), not only possessivization is correct but adjectivization also, in accordance with (125) in chapter 2. (38) a. Sus (= La Rioja) vinos son los mejores Its (= La Rioja) wines are the best del pais, of-the country b. Los vinos riojanos. The wines from-La Rioja (39) a. Sus (= siglo XX) avances han sido Its (= 20th century) advances have been R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 3 5 1 impresionantes. impressive b. Las novelas decimononicas the novels nineteenth-century ’The nineteenth-century novels.’ (40) a. Su (=universidad) comedor Its (=university) dining hall b. El comedor universitario the dining hall of-university ’The university dining hall.’ Therefore, the generalization that only NPs that can be possessivized can be extracted, appears to be correct, with the exception of sentential complements, that by their nature cannot be possessivized. Material, locative, and time complements cannot be either possessivized or extracted. Nevertheless, material, locative, time N complements comply with the requirements proposed on genitive case assignment in section 2.7. They hold a partitive relation with the head noun and they are potentially referential with respect to it, as the fact that they can individualize a token shows. Among a group of chairs, where only one is a wooden chair, we can refer to that chair as la silla de madera ’the chair of wood’. The problem is the interpretation of the definition of the potentially referential relationship. The complement itself is not referential, but generic with respect to R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 5 2 the head noun. Even when we use an introductory demonstrative for the complement noun, such as in la silla de esta madera 'the chair of this wood', the complement is generic with respect to the head. Therefore, genitive case is assigned when there is a partitive and a potentially referential relationship between the N and its complement, that relationship being reciprocal. By a referential relationship I mean when the head and the noun are referential with respect to each other. In the case of psych-nouns and objects of experience, the head is not referential with respect to the complement. In the case of material, locative, or time complements, the complement is the one that is generic with respect to the head. In this case, the preposition ^ theta-marks the complement in order to comply with the Full Interpretation Principle. This analysis is supported by the syntactic behavior of partitive constructions and pseudopartitive constructions. As noticed by Selkirk (1977), partitive constructions allow extraction, while pseudopartitive constructions do not, as illustrated by (41).^^^ Notice that in (40b), the relevant reading is the existential one, and not the generic one. The generic reading allows extraction, but it would be derived from a partitive structure, un montén de los lingüistas. R eproduced witti permission of ttie copyrigtit owner. Furttier reproduction protiibited wittiout permission. 3 5 3 (41) a. Vinieron estos lingüistas, de los que ’These linguists, of whom I know a bunch, conozco un punado. came.’ b.*Vinieron lingüistas de los que conozco un ’Linguists of whom I know a bunch came.’ punado. Abney (1987, 296-297) claims that pseudopartitive constructions violate the ECP, since they are not referential and, therefore, thay cannot be theta-marked. As we have seen in the case of the sentential complements, nonreferential elements can be theta-marked but cannot receive case. Therefore, it would be better to consider that the relevant notion here is case. In theoretical terms, it would also be a better solution in relation to the Full Interpretation Principle. Pseudopartitive complements cannot be assigned genitive case because they are always generic in relation to the head. They are not potentially referential with respect to the head. It can be argued that the same issue arises with respect to the partitive constructions, but notice that if we modify the head, as in (42), it is clear that the nature of the relationship between the head and the complement is not generic anymore. (42) La parte desértica de Espana. ’The desertic part of Spain.’ R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 5 4 Notice also, that in this case, possessivization and referential adjectivization is allowed. (43) a. Su (=Espana) parte desértica. Its (=Spain) part desertic ’Its desertic part’ b. La parte desértica espanola. The part desertic Spanish ’The Spanish desertic part.’ In conclusion, Longobardi’s analysis is correct except for the sentential complements. Therefore, it would be more correct to define the generalization for extraction in terms of syntactic argumenthood, where a syntactic argument is an NP that has been assigned case or a theta-marked XP other than NP, and not in terms of case assignment only. Notice that sentential complements introduced by an appositive i.e. a prepositional ^ that cannot theta-mark its complement, cannot be extracted, as (44) illustrates. (44)*Que los complementos apositivos no se pueden ’That the appositive complements cannot be extraer, de lo que la conclusion es extracted, of which the conclusion is correcta ... correct. ..’ 117 Godard (1992:242-243) claies that possessivization is not allowed in the case of partitive constructions, but she does not consider these cases. R eproduced witti permission of ttie copyrigtit owner. Furttier reproduction protiibited wittiout permission. 3 5 5 We need to check now whether the second part of the generalization in (30), i.e. where the element extracted is the highest element in the determination scale, or in Godard’s (1992) terms "the first argument", is correct. We will see that nonrestrictive relative clauses escape this part of the generalization. This fact suggests that another more general principle is involved in extraction of relative clauses and question formation. I will claim that this condition is subjacency, and that nonrestrictive clauses do not fall under Subjacency by means of referential coindexation. Thus, we should shorten the generalization in (30) to (45). (45) One can only extract a ^+NP from an NP if it is a syntactic argument. The relevance of the relation of the extracted element to the determiner system remains untouched. Other ungrammatical extractions result from more general principles of the grammar. 4.3.2. Towards a Syntactic Solution of Specificity: Bowers (1988) and Hudson (1989) Bowers (1988) and Hudson (1989) propose that the "specificity effects" are accounted for by a structural distinction between Determiners. Both of them make a R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 5 6 distinction between strong and weak quantifiers based on classifications of previous literature (Jackendoff, 1968, 1977; Milsark, 1974). Basically, weak quantifiers are those that can appear after a definite article, and strong quantifiers are those that cannot co-occur with the definite article, there being a linear order between them. The strong quantifiers are mapped onto the first position, and the weak quantifiers onto the second position, as shown in (46) (Hudson, 1989:19). (46) * Ç / Ç Q / Ç that each many two the two those all three many John's two Bowers and Hudson map the strong quantifiers onto DP, and the cardinal quantifiers onto Spec-of-NP, as in (47) (Jowers, 1988:49). (47) a. [pp each [ j j p picture of manatees]] b. [ f i p [^p many] [ f j , pictures of manatees]] Bowers claims that this structural distinction explains the extraction facts in terms of Subjacency.Within 118 Chomsky (1986b) defines Subjacency as a locality condition on the formation of a chain, as in 1 . (i) a. If (Q|, q .^j ) is a link of a chain, then a.+, is l-subjacent to Op b. B is n-suDjacent to q iff there are fewer tnan n+1 barriers for B that exclude 0. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 5 7 the framework of Barriers (Chomsky, 1986b), Bowers argues that you cannot extract the complement from the object position in (48b) since the verb cannot L-mark NP. NP is a Blocking Category and DP is a barrier by inheritance. Therefore the element extracted crosses two barriers, hence, there is Subjacency violation. On the other hand, in (48a), the NP is not a BC because it is L-marked by the verb. (48) a. Who did you buy pictures of? b.*Who did you buy those pictures of? The Subjacency account is problematic when we deal with multiple wh-questions, since they show the same specificity effects, as the contrast in (49) shows (sentences from Fiengo and Higginbotham, 1981). The index of the wh-phrase that remains in its underlying position is considered to be absorbed by the wh-phrase in initial position (van Riemsdijk and Williams, 1981; Aoun, Hornstein, and Sportiche, 1981). Hence, either we consider that Subjacency is a Condition on representations and not on derivations, or we assume that Subjacency applies at LF, and that wh-in-situ phrases in multiple wh-questions structures move at LF, as Pesetsky (1987) does. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 5 8 (49) a. Who saw pictures of who? b.*Who saw those pictures of who? Hudson (1989:21) explains the same facts based on the Generalized Binding Theory to the A’-system proposed by Aoun (1986).^^ That is, in (48b), the sentence is barred because the trace left by the wh-eleraent is free in the domain of an A’-subject, AGR in D°, hence, there is a violation of Generalized Principle A. In (48a), Hudson argues that the structure is similar to that of a bare noun. There is no subject in NP, since there is no AGR, and the trace can be bound from outside. In addition, the AGR of the sentence is coindexed with an argument in A-position, therefore, it is not an accessible subject, and no Principle C violation arises. Aoun (1986) extends the A-binding conditions to X-binding, including A ’-binding. Therefore, Binding Conditions are reformulated as in (i). (i) A. An anaphor must be X-bound in its governing category. B. A pronominal must be X-free in its governing category. C. A name must be A-free. Governing category is defined as in (ii). (ii) 8 is a governing category for a iff 8 is the minimal category containing a, a governor for a, and a SUBJECT accessible to q. The definition of accessibility is in (iii). (iii) 8 is accessible to o iff 8 c-commands o and coindexing of q and 8 does not violate any grammatical principles. R eproduced witti permission of ttie copyrigtit owner. Furttier reproduction protiibited wittiout permission. 3 5 9 Hudson (1989:21) gives the following representations in (50) for similar cases to (48a) and (48b). (50) a. DP b. NP / \ / \ that D’ a N’ / \ / \ [AGR, D] NP picture PP \ of X N' / \ N PP picture of x This account can be extended to the cases of LF movement, such as in (51), since the traces left by QR are considered to be A'-anaphors (Aoun, 1986). (51) a. Juan vio una foto de todo el mundo. ’Juan saw a picture of everyone.’ b. Juan vio esta foto de todo el mundo. ’Juan saw this picture of everyone.’ In (51a) there can be a wide scope reading and a narrow scope reading for todo el mundo ’everyone’. In (51b) we have only a narrow scope reading, since QR is barred by Generalized Principle A. As has been indicated by Torrego (1988), the presence of a possessor, a genitive agent or a restrictive relative clause, contributes to the definiteness of nominals. They seem to induce the same R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 6 0 specificity effects as quantificational determiners, as (52) shows: (52) Juan vio una fotografia de todo el mundo de John saw a picture of everyone of ’John saw Peter’s picture of everyone.’ Pedro. Peter. In the same manner as in (51b), in (52) todo el mundo ’everyone’ has only a group reading if Pedro is the possessor, and a very marginal wide scope reading, if Pedro is the agent. These judgements match the ones in (4) and (8) for the extraction of the object in presence of a possessor and an agent. In this case, the presence of the genitive phrase produces the mapping of the indefinite article onto the DP position. Since the article acts as an operator, it acts as a potential A’- subject. Wide scope of the quantified expression would be ruled out, since the trace would not be bound in its governing category. An account for the Specificity facts based on Generalized Principle A would result in the loss of important generalizations in relation to non-restrictive clauses and D-linked questions. I claim that the Specificity effects on extraction must be considered as CED effects. Non-restrictive relative clauses and D- R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 6 1 linked questions show CED effects when extracted from adjuncts, postverbal subjects. Complex NPs, and wh- islands. On the other hand, they do not show CED effects when extracted from a specific NP or a preverbal subject. The relevant generalization here is that when the maximal projection containing the trace is found in Spec-head relation and bears a referential index, that maximal projection does not induce Subjacency effects. I claim that Spec-head relations 0-mark the element on Spec in Spanish, by assigning a referential index to it. This is the case of preverbal subjects in Spanish and of GENPs that enter into a Spec-head relation with another GENP or a DEMP. 4.3.3. Diessing (1992) Diessing (1992:103) gives the following generalization in (53) in order to explain the extraction facts. (53) Presuppositional NP Constraint Extraction cannot take place out of a presuppositional NP. Diessing argues that presuppositional NPs undergo a rule of QR at LF to form a restrictive clause, in accordance with her Mapping Hypothesis, in (54). R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 6 2 (54) Mapping Hypothesis Material from VP is mapped into the nuclear scope. Material from IP is mapped into a restrictive clause. Diessing proposes that some condition similar to the Freezing Principle Wexler and Culicover (1980:119) is at work in these cases. The Freezing Principle states that nothing can be moved out of a node that has already been moved, or, the reverse case, no node can be moved if some constituent has already been moved from within it. Diessing (1992: 128-129) argues that this condition is Subjacency. Diessing revises her Extraction Constraint as in (55) . (55) Revised Extraction Constraint Extraction cannot take place out of an NP that must raise out of VP before tree splitting. One major problem with her hypothesis is that neither nonrestrictive clauses nor D-linked questions are subject to this Constraint, while they are subject to Subjacency, as we will see in section 4.4. A possible solution would be that elements mapped into a restrictive clause are 0-marked through assignment of a referential index. A DP containing a variable resulting from the structure of a non D-linked question or a restrictive relative clause cannot receive a referential index by R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 6 3 definition. Hence it would not be 0-marked and Subjacency effects would appear. Obviously, this argumentation is a tautology since the presuppositionality of a subject mapped onto the Spec-of-IP position cannot be optional in German (Diessing 1992:32), and therefore, neither can assignment of a referential index. The problem can be solved by positing a clash of features between the DP mapped into the restrictive clause and the variable within that DP, along the lines of Martin (1993). Therefore two independent violations can arise in these contexts, a Subjacency violation and a clash of agreement features. Another issue to be clarified in this possible solution for Diessing’s account is how the DP mapped into the restrictive clause receives a referential index. The elements mapped into the restrictive clause are not an homogeneous class in Diessing’s account. For instance, presuppositional objects are adjoined to IP at LF, while presuppositional subjects are mapped onto Spec of IP by S-structure. This needs further investigation in the case of Diessing’s analysis. In the next section, I claim that the correct configuration for assignment of a referential index is a spec-head configuration. Another problem with Diessing’s analysis and its application to Spanish is that we find the same kind of R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 6 4 facts with quantified NPs. Not only can a D-linked wh- phrase but also a non D-linked phrase be extracted out of a directly quantified phrase as both (56a) and (56b) show. (56) a. i,De que tema lelste cada libro? of which topic did-you-read each book? ’On which topic were each of the books you read?’ b. ^De que diablos leiste cada libro? of what the-hell did-you-read each book? ’What the hell were each of the books on, that you read?’ Quantified NPs should show the same effects as presuppositional DPs, since they are also raised at LF. However, this is not the case in Spanish, as (56) shows. The possible application of Diessing’s analysis of the extraction facts to Spanish needs further research. Although the Mapping Hypothesis may account for different syntactic phenomena in Spanish, such as the different interpretation of preverbal and postverbal subjects, its relevance in the case of extraction seems to me to be less important than Diessing argues. In particular, it is still unclear to me that the Quantifier Raising rule constrains extraction in any way. I claim that an explanation based on the internal structure of DPs, as proposed in Chapter 3, and on the syntactic behavior and semantic interpretation of preverbal subjects will R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 6 5 account much more neatly for the extraction facts. In some way, my analysis extends the Mapping Hypothesis to the internal structure of NPs, although I will state the dichotomy between restrictive clause and nuclear scope in other terms. I claim that a spec-head relation between a nominal expression and a [+ pronominal] unbound head assigns a referential index to the element in Spec- position. This is the case of preverbal subjects, preverbal DOs and lOs doubled by a clitic, and nominal expressions introduced by a strong determiner in 1 21) Spanish. If a referential index is assigned under this configuration, the element in Spec is ©-governed and it is not a barrier for extraction from within it. A non referential element cannot be in Spec-relation with a [+ pronominal] element. Hence, we have a clash of agreement features and a Subjacency violation. In the next section, we will study in more detail the differences between restrictive clauses and regular wh-questions, on the one hand, and nonrestrictive This is the case, for instance, where a nominal expression is introduced by a demonstrative. Recall that in chapter 3 we propose that the nominal expression enters into a spec-relation with the demonstrative before the demonstrative raises to D*. See the syntactic representation of this case in (112) of Chapter 3. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 6 6 relative clauses and D-linked questions, on the other. I will base ray proposal on these data. 4.4. On the Difference between Restrictive and Nonrestrictive Clauses In Martin (1993), I raised the issue of the nonrestrictive clauses that escape the specificity effects, as shown in (15). Also, D(iscourse)-linked questions do not show specificity effects, as shown in (57). (57) i,De que teraa has escrito esta of which topic have-you written this 'On which topic have you written this composicion? coraposition? coraposition?’ These facts are a problem for previous proposals. As pointed out by Torrego (1988), Safir’s (1986) proposal of a second level of representation at LF can account for these facts, although, Torrego does not consider these cases as genuine wh-movements. However, as pointed out by Safir (1986), binding by a relative head cannot be vacuous at the relevant level of representation, which follows from the ©-Criterion and the Full Interpretation Principle (Chomsky 1981, 1982). Additionally, nonpronominal empty categories must be properly governed. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 3 6 7 in accordance with Giorgi and Longobardi (1991), therefore traces left by wh-movement in a nonrestrictive relative clause are subject to at least some of the same constraints as other wh-movements, as (58) illustrates (Giorgi and Longobardi, 1991:74). (58) *Gianni, al qualej sapevo molte cose anche Gianni, to whom I-knew many things even prima [pp e, di parlare e^], before talking, ... As we can observe in (58), we cannot form a nonrestrictive clause out of an adjunct, hence, in this case, nonrestrictive clauses are subject to the same constraints as questions or restrictive clauses. Safir (1986) claims that the rule in (59) would apply to what he calls "orphan constituents", that can be defined as predicative elements that have not received a P-index. (59) Attach a. Tellier (1991) states that there is a difference between predication indices and the indices used for binding or resulting from movement. She posits that predication indexing takes place at the earliest level possible, once a maximal projection has acquired the necessary semantic content. I claim that there are also some locality R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 6 8 structural requirements in accordance with Williams’ (1980) Condition on Predication in (60). (60) A predicate must be c-subjacent to its subject, or antecedent. B is c-subjacent to A iff A is dominated by at most one branching node which does not dominate B. Hence, wh-words in nonrestrictive relative clauses cannot receive their P-index until LF, where they comply with the Subjacent Condition. In Martin (1993), I account for these facts based on the Empty Category Principle, trying to give a unified account for them. I consider NPs as variables that can have an index assigned from outside or inside their own structure. From inside, they receive an index through coindexation of the subject with the head of the DP (Aoun, 1985). Thus, in this process, D and the subject would have the same type of index. The type of index is going to depend on the fact of whether the subject is a variable or an essential term. I claim in that paper, that a trace within DP must comply with the constraints on Spec-Head Agreement, which, according to Rizzi (1990), will trigger proper government of the trace with index i=l by the Noun, this being a requirement of every trace for proper licensing. Spec-Head Agreement will be fulfilled when the extracted R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 6 9 element matches all the ^-features, including the [Wh] 101 feature, that the head bears. This includes wh- agreement between the head and the trace. This is probably a requirement also for wh-traces within VPs. This agreement is accomplished through dynamic agreement (Rizzi, 1991) between the wh-phrase and the agreement system of the verb, as proposed in Martin (1992). Hence wh-agreement is a condition on proper head government of a trace. For instance, in (4) the Determiner has been assigned the [-wh-] feature by the subject, while the trace left within DP bears a [+wh] feature. Hence, Wh- agreement is not achieved, resulting in a violation of the ECP, stated in (61), since the head-government requirement on traces is not fulfilled, according to Rizzi (1990). (61) ECP (Rizzi, 1990) A nonpronominal empty category a must be properly head-governed (Formal licensing) 101 Chomsky (1981) suggests including the feature [Wh] in the set of 4-features, This feature represents whether a referential expression has been saturated or not. Hence, the set of 4- features would be as in (i). (i) 4-features: Person, Number, Gender, Case, [Wh] R eproduced witti permission of ttie copyrigfit owner. Furtfier reproduction profiibited witfiout permission. 3 7 0 In (lia), (12a), (13a), and (14) the Determiner is inherently [-wh-], disallowing the extraction of a [+wh-] element. On the other hand, in (15) the trace is licensed by Wh-agreement since the trace in Spec of DP, bears a [-wh-] feature at LF’. We get the same results on extraction of an element with a [-wh-] feature in the presence of a subject with the same feature, [-wh], as (62) shows. (62) El rey, del que, por cierto, vimos un retrato ’The king, of whom, by the way, we saw a de Dali en el museo,... portrait by Dali in the museum,...’ Since the attachment operation of attach a in (59) does not apply to restrictive clauses, I predict that this kind of extraction would not be allowed in the presence of a [-wh-] D because the chain remains with the [+wh-], as seen in (14). As predicted, extraction of a [+wh-] element from an NP headed by a [+wh-] D is allowed as in Torrego’s sentences from Barriers (Chomsky, 1986b) in (63). (63) a. iDe gué autora no sabes gué by what author don’t you-know what ’By what author don’t you know what traducciones ban ganado premios translations have won awards translations have won international R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 7 1 internacionales? international? awards?’ b. Este es el autor [del que]j no sabemos [gp this is the author [by whom]j we-don’ t-know ’This is the author by whom we don’t know [g que libres tj ] leer] [ what books t^] to-read, what books to read.’ Obviously, this is a problem for Hudson’s analysis if we map the wh-word onto DP, because it must act as a potential A’-binder. However, we should map this wh-word in the lower DetP, where it cannot act as a potential A’- binder. (64) is the underlying representation for the NP in (63a). (64) GenP / \ DetP; 'Gen’ / \ / \ que NP Gen’ de que autora traducciones } \ Gen° tj In (64), we can observe that gué does not act as a potential subject for the element extracted. Further, nonrestrictive clauses are allowed from variable NPs as (65) shows. (65) Cervantes, del que, por cierto, que libro vas ’Cervantes, by whom, by the way, which book are a leer, ... you going to read, ...’ R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 7 2 In Martin (1993), I propose that this allows us to conclude that the ECP can be complied with at LF or at LF’, i.e. traces in nonrestrictive clauses can behave in two different ways, as variables or as referential expressions. Under the analysis presented in this dissertation, as (64) shows, we do not need to say anything with respect to (63) and (65), assuming Chomsky’s (1986b:25-26) analysis for the conditions that are not related to the internal structure of nominal expressions. In section 4.5 I will give an account for the lack of Specificity effects in the case of D-linked wh-questions and non restrictive clauses, and their immunity to Subjacency effects when they are formed out of a preverbal subject. The analysis presented in Chapter 3 will allow us to explain these facts by the Subjacency Condition and the conditions on Spec-head agreement. 4.5. Spec head agreement, the Subjacency Condition, and Specificity effects As (15) and (57) show, the formation of non restrictive clauses and D-linked wh-questions do not show specificity effects. On first observation, it would be difficult to account for these facts by the Subjacency Condition if we consider that non restrictive clauses and D-linked questions are subject to this condition, as (58) R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 7 3 illustrates. However, non restrictive clauses do not show Subjacency effects when they are formed out of a subject as (66) shows. (66) Cela, de quien, por cierto, una novela ganô Cela, by whom, by the-way, a novel won el premio Planeta, the prize Planeta, There is a sharp contrast with extraction of a non D-linked question, as (67) illustrates. (67)*&De quien diablos piensas que una of whom the-hell do-you-think that a novela sera publicada? novel will-be published? (67) can never have a list of writers as an answer. On the other hand, (68) , where the extraction occurs from the object position, can have a list of writers as a possible answer. (68) ^De quien diablos piensas que of whom the-hell do-you-think that publicaran una novela? they-will-publish a novel? Therefore, extraction out of a subject does not show Subjacency effects in the case of non restrictive clauses and D-linked questions. I claim that whenever a referential index is assigned through a spec-head R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 7 4 relation, the element in spec is ©-governed and, hence, it is not a barrier for extraction from within it. As stated by Chomsky (1993: 27-32), ©-marking is not restricted to sisterhood. An element can be ©-marked by a head with lexical features under the Spe-head configuration. This is the case of the preverbal subject position in Spanish.The element on that position is in the checking domain of a head with lexical features, i.e. the strong V-features of 1° in Spanish. In contrast with Chomsky (1986b:25), I believe that case-marking plays a role in the definition of barrierhood, or, at least, it is a mechanism to achieve ©-marking. I claim that whenever an element is case-marked through a Spec- head relation directly, i.e. without mediation of any element, such as a preposition, and it is assigned a referential index by an agreement with strong NP- features, in the sense of Chomsky (1993), ©-marking of that element is obtained. Therefore, it will not be a barrier anymore for an element extracted from within it. In conclusion then, there are two requirements for ©- marking under the spec-head configuration, that the head has lexical features and that the head has strong NP- features. The Subjacency Condition is stated in (69). See Martin (1992) for a definition of this position. R eproduced witti permission of ttie copyrigfit owner. Furtfier reproduction profiibited witfiout permission. 3 7 5 (69) Subjacency Condition a. If (a;, is a link of a chain, then is 1-subjacent to Q|. b. 6 is n-subjacent to a iff there are fewer than n+1 barriers for B that exclude a. (70), (71), (72), and (73) state the relevant notions for barrierhood, where I maintain the notion of government, following a model similar to Rizzi (1990). (70) Barrier U is a barrier for 6 iff (a) or (b): a. n immediately dominates 6, ô a blocking category (EC) for B; b. n is a EC for B, and p is not IP. (71) Blocking Category p is a EC for B iff p is not L-marked and p dominates B (72) L-marking a L-marks B iff a is a category with L-features and a ©-governs B (73) Theta-marking a ©-governs B iff (a) or (b): a. a ©-marks B, and a and B are sisters b. a ©-marks B, and a identifies B through strong agreement features. Therefore (66) and (67) are accounted for. In (66) the subject is identified and ©-marked by I“, a head with V-features and strong NP features, and, hence, a referential index is assigned to the subject. In (67) this is impossible since the subject contains a variable, and this variable cannot bind agreement. It cannot receive a referential index, and 1° cannot identify, and. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 7 6 therefore, 0-mark the subject. A Subjacency violation arises since two barriers, DP and IP, are crossed, and, as we will see below, the derivation does not converge because of a clash of features between the subject and I". In the case of Specificity effects, we have a similar case. I propose that all the heads within a nominal expression have nominal lexical features that are spelled out as the phrasal marker that we studied in section 1.5. They have also strong NP-features, as gender and number agreement shows. Therefore, whenever the head is pronominal, it can assign a referential index to the nominal expression in Spec-head relation and 0-mark it. In these cases, such as (15), no Subjacency violation arises. On the other hand, in (11a), (12a), (13a), and (14), Dem° cannot 0-mark the phrase on its specifier, in 1 9 1 accordance with the structure proposed in Chapter 3. Since the GenP on Spec-of-DemP does not receive a referential index, it becomes a barrier and extraction from within is ruled out. A final question remains. The question is whether utterances such as (11a), (12a), (13a), (14), and (67) involve a clash of agreement features. In Martin (1993) See (112) in Chapter 3, for instance. R eproduced witti permission of ttie copyrigtit owner. Furttier reproduction protiibited wittiout permission. 3 7 7 I argue that this is the case for (11a), (12a), (13a), and (14). In the case of preverbal subjects, the question of whether they always receive a referential index or not in Spanish is difficult to answer. A good test are CPs in that position. It is assumed that CPs do not bear a referential index. However, they can be introduced by a definite article in Spanish, as (74) shows. (74) a. Que telafoneara causé muchos problemas that he-phoned caused many problems ’That he left caused many problems.’ b. El que telefoneara causé muchos the that he-phoned caused many problemas. problems There is a slight difference between (74a) and (74b) with respect to intonation and interpretation. While in (74a) the embedded CP is a topic and the rest of the sentence is the comment, in (74b) the topic interpretation is only optional. Additionally, a pause after the embedded CP is necessary in (74a), while it is not in (74b). I claim that while the embedded CP can occupy only the topic position in (74a), the embedded CP can occupy either the topic position or the subject position in (74b). I claim that in (74b) the definite article bears a referential index that allows the embedded CP to occupy the subject position. Therefore, it R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 7 8 is a requirement for the element on the preverbal subject position in Spanish to receive a referential index. This entails that the ungrammaticality of (67) results from a clash of agreement features in addition to Subjacency, which is a well attested phenomenon in the same contexts in other languages. In summary, I have proposed that extraction out of a preverbal subject and extraction out of a specific NP in Spanish are subject to similar constraints. In particular, these structures must comply with the Subjacency condition and with the conditions on strong agreement. Further evidence is given by wultiple question structures. In (i), where the wh-in- situ phrase occupies the internal subject position, we can have a distributive reading of thewh- in-situ phrase. On the other hand, this reading is not available in (ii), where the wh-in-situ occupies the preverbal position, and only a D-linked reading is available. (i) îQuién dijo que telefoneô quién? who said that phoned who? (ü)iQuién dijo que quién telefoneô? who said that who phoned? R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 7 9 Conclusions In this study I have given further evidence for the hypothesis that notninalizations and verbs share the same argument-structure, along the lines of the Lexicalist Hypothesis (Chomsky, 1970). Following Zubizarreta (1987), the optional projection of the external arguments by an NP is explained by the fact that NPs do not undergo predication but modification. The different mechanisms of ©-assignment and case assignment at the nominal level and at the sentential level are explained by the lack of adjacency between the head and the complement within a nominal expression. Assuming an X’-theory where two complement positions are projected to the right, as 1 have suggested in fn. 14, verbal heads move head to head, probably from an affix phrase to the head of the maximal projection where the object is projected. This movement results in the right configuration for structural government, i.e. the head is adjacent to the complement, as (1) shows (labeling irrelevant at this point). R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 3 8 0 (1) VP2 / \ "V2’ / \ V’ Roma / \ destruir- VPl / \ VI ' I I ti On the other hand, nouns move as maximal projections from the lower complement position to the Spec of the immediate dominating projection. Under this configuration, the adjacency requirement on government is not fulfilled, as (2) shows. (2) GenP / \ destrucciôn; "Gen" / \ 1 Gen’ de Roma Finally, in accordance with Zubizarreta (1987) nouns do not have passive morphology because they lack a lexico-syntactic representation. Therefore, as stated by Stowell (1989:233), nouns have an argument structure but function as referring expressions. The latter aspect determines the syntactic behavior of nominal functional categories. We can define the functional system of Noun Phrases as unified. This R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 8 1 definition is based on the fact that functional categories that select NPs act as operators of the referential properties of the noun. Taking this definition into consideration, I have proposed that there are two pure functional categories, DP and DetP. DP introduces nominal expressions referentially saturated, i.e. the nominal expression denotes a token instantiated by the type denoted by the NP, while DetP does not. DP dominates NP and all its restrictive modifiers, while DetP only selects NP, although it can enter into a spec- head relation with a restrictive modifier that does not referentially saturate the nominal expression. DP and DetP correspond to the strong and weak distinction among determiners defined in previous literature. Evidence for this proposal has been given by the distribution of wh- in-situ questions, the existence of affective possessives with no referential import, the co-occurrence of these possessives in prenominal position with other elements that can occupy DP in Spanish, the coordination facts with respect to type nouns, the possibility of deletion of a lower DetP, typological data, such as the Romanian data, and finally the syntactic behavior of stage-level predicates and individual-level predicates in existential sentences. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 8 2 Following Abney (1987), adjectives are considered to be heads that select DetP or NP. Adjective ordering is defined by a universal scale of determinaLion. There is a universal underlying string of restrictors that gradually limit the referentiality of the noun when moving upward in the syntactic tree. This order surfaces in English in its underlying format, but is reversed in Romance languages by successive movement of the complement of an AdjP to its Spec at the syntactic component of the grammar. This explains the mirror effect of adjective ordering between Spanish and English, the agreement facts of Spanish nominal expressions, and the fact that adjectives in Spanish narrow the meaning of the noun from left to right, as pointed out by Bolinger (1954). (3a) shows the abstract underlying representation of a noun with two restrictive modifiers, and (3b) the abstract surface representation. (3) a. DP / \ D' / \ AdjPl / \ Adjl’ / \ AdjP2 / \ Adj2' / \ NP R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 8 3 b. DP / \ D' / \ AdjPl / \ AdjP2; Adjl' / \ / \ NP. Adj2’ t; J / \ tj Intensional adjectives are analyzed as Specifiers of NP since they do not have referential relevance. This analysis accounts for the prenominal position of intensional adjectives at the output of the syntactic component, since they maintain that position through the whole derivation. Whenever the nominal expression has been referentially saturated, there is no further restrictive modification allowed. I have argued that affected objects are lexical saturators of nominalizations. This explains why nouns that project affected objects do not allow prenominal possessives in Spanish. Another interesting consequence of this analysis is that which accounts for the fact that an element cannot be extracted when a higher modifier is present since it will give rise to a Subjacency violation. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 8 4 References Abney, Steven. 1987. The English noun phrase in its sentential aspect. Doctoral dissertation, MIT, Cambridge, Mass. Alarcos, Emilio. 1961. Los pronombres personales en espanol. Archivum XI:5-16. Alcina, Juan, and José M. Blecua. 1975. Gramatica espanola. Barcelona: Ariel. Amiratavalli, R. 1980. Expressing cross-categorial selectional correspondences: An alternative to the X-bar approach. Linguistic Analysis 6:305-343. Anderson, Mona. 1979. Noun phrase structure. Doctoral dissertation. University of Connecticut, Storrs, Conn. . . 1984. Prenominal genitive NPs. The Linguistic Review 3:1-24 Antrim, Nancy Mae. 1993. Italian adverbial agreement. In Issues and theory in Romance linguistics: Selected papers from the Linguistic Symposium on Romance Languages XXIII. ed. Michael L. Mazzola. Washington, D.C.: Georgetown University Press. Aoun, Joseph. 1985. A grammar of anaphora. Cambridge, Mass.: MIT Press. 1986. Generalized binding: The syntax and logical form of wh-interrogatives. Dordrecht: Foris. Aoun, Joseph, Norbert Hornstein and Dominique Sportiche. 1981. Some aspects of wide scope quantification. Journal of Linguistic Research 1:69-95. Aoun, Joseph and Dominique Sportiche. 1983. On the formal theory of government. Linguistic Review 2:211-36. Authier, J.-Marc. 1988. The syntax of unselective binding. Doctoral dissertation. University of Southern California, Los Angeles. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 8 5 . . 1992. Is French a null subject language in the DP? Probus 4:1-16. Baker, Mark. 1988. Incorporation: A theory of grammatical function changing. Chicago: University of Chicago Press. Belletti, Adrianna. 1990. Generalized verb movement. Aspects of verb syntax. Turino: Rosenberg & Sellier. Belletti, Adriana and Luigi Rizzi 1988. Psych verbs and Theta Theory. Natural Language and Linguistic Theory 6:291-352. Bello, Andrés. 1847. Gramatica de la lengua castellana destinada al uso de los americanos. Santiago de Chile: Imprenta del Progreso. ________ . 1928. Gramatica de la lengua castellana. Paris: R. Roger and F. Chernoviz. Bernstein, Judy B. 1993. Topics in the syntax of nominal structure across Romance. Doctoral dissertation, CUNY, New York. Bolinger, Dwight. 1954. Meaningful word order in Spanish. Boletin de Filologia (Universidad de Chile) 8:45-56. . . 1967. Adjectives in English: Attribution and predication. Lingua 18:1-34. . . 1972. Adjective position again. Hispania 55:91- 94. Bosque, Ignacio. 1994. Degree quantification and modal operators in Spanish. Paper presented at the 24th Linguistic Symposium on Romance Languages, University of Southern California, and Univesity of California, Los Angeles, March 10-13, 1994. Bouchard, Dennis. 1984. On the content of empty categories. Dordrecht: Foris. Bowers, John. 1988. Extended X-bar theory, the ECP, and the Left Branch Condition. In Proceedings of the West Coast Conference on Formal Linguistics 7, 47- 62. Stanford Linguistics Association, Stanford University. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 8 6 . 1993. The syntax of predication. Linguistic Inquiry 24:591-656, Brame, Michael. 1981. The general theory of binding and fusion. Linguistic Analysis 7:277-325. ________. 1982. The head-selector theory of lexical specifications and the nonexistence of coarse categories. Linguistic Analysis 10:321-325. Bull, William E. Spanish for teachers: Applied linguistics. New York: Ronald Press. Burzio, Luigi. 1986. Italian syntax: A government and binding approach. Dordrecht: Reidel. Carstens, V.M. 1991. The morphology and syntax of Determiner Phrases in Kiswahili. Doctoral dissertation, UCLA. Chierchia, Gennaro. 1985. Formal semantics and the grammar of predication. Linguistic Inquiry 16:417- 443. Chierchia, Gennaro, and Sally McConnell-Ginet. 1990. Meaning and grammar: An introduction to semantics. Cambridge, Mass.: MIT Press. Chomsky, Noam. 1970. Remarks on nominalization. In Readings in English transformational grammar, ed. Roderick A. Jacobs, and Peter S. Rosenbaum, 184-221. Waltham, Mass.: Ginn. ________ . 1980. On binding. Linguistic Inquiry 11:1-46. . 1981. Lectures on government and binding. Dordrecht: Foris. _ . 1982. Some concepts and consequences of the theory of government and binding. Cambridge, Mass.: MIT Press. . 1986a. Knowledege of language: Its nature. origin and use. New York: Praeger. . 1986b. Barriers. Cambridge, Mass.: MIT Press R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 8 7 ________. 1991. Some notes on economy of derivation and representation. In Principles and parameters in comparative grammar, ed. Robert Friedin, 417-454. Cambridge, Mass.: MIT Press. ________. 1993. A minimalist program for linguistic theory. In The view from building 20: Essays in honor of Sylvain Bromberger. ed Kenneth Hale, and Samuel Jay Keyser, 1-52. Cambridge, Mass.: MIT Press. Chomsky, Noam, and Morris Halle. 1968. The sound pattern of English. New York: Harper and Row. Chomsky, Noam, and Howard Lasnik. 1993. Principles and Parameters Theory. In Syntax: An International Handbook of Contemporary Research, ed. J. Jacobs, A. van Stechow, W. Sternefeld, and T. Vennemann. Berlin: Walter de Gruyter. Cinque, Guglielmo. 1980. On extraction from NP in Italian. Journal of Italian Linguistics 5:47-99. 1990. A null theory of phrasal stress. Ms., University of Venice. . . 1993. On the evidence for partial N movement in the Romance DP. Ms., Université di Venezia. Comrie, Bernard. 1976. Aspect. Cambridge: Cambridge University Press. Cornilescu, Alexandra. Remarks on the determiner system of Rumanian: The demonstratives ^ and cel. Probus 4:189-260. Crisma, Paola. 1990. Functional categories inside the noun phrase: A study of the distribution of nominal modif iers. Doctoral dissertation. Université di Venezia. Demonte, Violeta. 1982. El falso problema de la posiciôn del adjetivo: dos anélisis semânticos. Boletin de la Real Academia Espanola LXI1:453-485. ________. 1987. Reccion y minimidad en el sintagma nominal. In Sintaxis de las lenguas romanicas. ed. Violeta Demonte, and Marina Fernandez Lagunilla. Madrid: El Arquero. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 8 8 ________. 1992. Temporal and aspectual constraints on predicative AP’s. In Currents Studies in Spanish Linguistics. ed. Hector Campos, and Fernando Martinez-Gil. Georgetown: Georgetown University Press. Diessing, Molly. 1992. Indefinites. Cambridge, Mass.: MIT Press. Dik, Simon C. 1989. The theory of functional grammar, part 1: The structure of the clause. Dordrecht: Foris. Dobrovie-Sorin, Carmen. 1987. A propos de la structure du groupe nominal en roumain. Rivista di grammatica generativa 12:123-152. Dowty, David R. 1979. Word meaning and Montague grammar. Dordrecht: Reidel. Duffield, Nigel G. 1991. Particles and projections. Doctoral dissertation, University of Southern California, Los Angeles. Eraonds, Joseph E. 1985. A unified theory of syntactic categories. Dordrecht: Foris. Fagan, Sarah M.B. 1988. The English Middle, Linguistic Inquiry. 19:181-203. Fernandez Ramirez, Salvador. Gramatica espanola: Los sonidos, el nombre y el pronombre. Madrid: Revista de Occidente. Fiengo, Robert, and James Higginbotham. 1981. Opacity in NP. Linguistic Analysis 7:395-421. Fiengo, Robert, and Robert May. 1994. Indices and identity. Cambridge, Mass.: MIT Press. Fillmore, Charles. 1968. The case for case. In Universels in linguistic theory, ed. E. Bach and R. Harms, 1-88. New York: Holt, Rinehart and Winston. Franco, Jon. 1990. Towards a typology of psych verbs: evidence from Spanish. Paper read at the II SCIL at MIT, March 24, 1990. University of Southern California. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 8 9 Franco, Jon and Alazne Landa. 1991. The syntax and semantics of null objects in Basque Spanish. In Proceedings of The Western Conference on Linguistics WECOL 91. ed. Katharine Hunt, Thomas A. Perry, and Vida Samiian, 117-128. Department of Linguistics, California State University, Fresno. Fukui, Naoki, and Margaret Speas. 1986. Specifiers and projections. In MIT Working Papers in Linguistics 6, ed. Naoki Fukui, T. Rapoport, and E. Sagey, 128- 172. MIT, Cambridge, Mass. Georgepoulus, Carol. 1991. On A- and A’-agreement. Lingua 85:135-169. Gili Gaya, Samuel. 1943. Curso superior de sintaxis espanola. México: Minerva. Giorgi, Alessandra. 1984. Towards a theory of long distance anaphors: a GB approach. The Linguistic Review 3: 307-361. Giorgi, Alessandra, and Giuseppe Longobardi. 1991. The syntax of noun phrases: Configuration, parameters, and empty categories. Cambridge: Cambridge University Press. Glinert, Lewis. 1989. The grammar of Modern Hebrew. New York: Cambridge University Press. Godard, Daniêle Godard. 1992. Extraction out of NP in French. Natural Language and Linguistic Theory 10:233-277. Goodall, Grant. 1987. Parallel structures in syntax. Cambridge: Cambridge Uniyersity Press. Grimshaw, Jane. 1990. Argument Structure. Cambridge, Mass.: MIT Press. Grosu, Alexander. 1988. On the distribution of Genitive Phrases in Rumanian. Linguistics 26:931-949. Hale, Kenneth, and Samuel J. Keyser. 1986. Some transitivity alternations in English. Lexicon Project Working Papers 7, Center for Cognitive Science, MIT. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 9 0 ________ . 1987. A view from the middle. Lexicon Project Working Papers 10, Center for Cognitive Science, MIT. ________ . 1988. Explaining and constraining the English middle. In Studies in generative approaches to aspect. ed. Carol Tenny. Lexicon Project Working Papers 24, Center for Cognitive science, MIT. ________ . 1993. On Argument Structure and the lexical expression of syntactic relations. In The view from building 20: Essays in linguistics in honor of Sylvain Bromberger. ed. Kenneth Hale and Samuel J. Keyser, 53-109. Cambridge, Mass.: MIT Press. Halle, Morris, and Jean-Roger Vergnaud. 1987. An essay on stress. Cambridge, Mass.: MIT Press. Harris, James. 1983. Syllable structure and stress in Spanish: A nonlinear analysis. Cambridge, Mass.: MIT Press. ________ . 1991. The exponence of gender in Spanish. Linguistic Inquiry 22:27-62. Hetzron, Robert. 1978. On the relative order of adjectives. In Language universels: Papers from the Conference held at Gummersbach/Cologne. Germany. October 3-8, 1976. ed. Hansjakob Seiler, 165-184. Tübingen: Gunter Narr. Higginbotham, James. 1985. On Semantics. Linguistic Inquiry 16:547-594. Huang, C.-T. James. 1982. Logical relations in Chinese and the theory of grammar. Doctoral dissertation, MIT. Hudson, Wesley. 1989. The Structure of Noun Phrases in English: Specificity Revisited. Ms. USC. Jackendoff, Ray. 1968. Quantifiers in English. Foundations of language 4:422-442. . 1972. Semantic interpretation in generative grammar. Cambridge Mass.: MIT Press. . . 1977. X' syntax: A study of phrase structure. Cambridge, Mass.: MIT Press. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 9 1 . 1983. Semantics and cognition. Cambridge, Mass.: MIT Press. . 1987. The status of thematic relations in linguistic theory. Linguistic Inquiry 18:369-411. Jaeggli, Osvaldo. 1986. Passive. Linguistic Inquiry 17:587-622. Jaeggli, Osvaldo, and Kenneth J. Safir. 1989. The null subject parameter and parametric theory. In The null subject parameter, eds. Osvaldo Jaeggli, and Kenneth J. Safir, 1-44. Dordrecht: Kluwer. Kayne, Richard. 1975. French Syntax: The Transformational Cycle. Cambridge, Mass.: MIT Press. . . 1981. On certain differences between English and French. Linguistic Inquiry 12:349-371. . 1984. Connectedness and binary branching. Dordrecht: Foris. . 1991.Romance clitics, verb movement and PRO. Linguistic Inquiry 22:47-686. Kenny, Anthony. 1963. Action, emotion and will. London: Routledge and Kegan Paul. King, Larry D. 1992. The semantic structure of Spanish: Meaning and grammatical form. Amsterdam: John Benjamins. Klein-Andreu, Flora. 1983. Grammar in style: Spanish adjective placement. In Discourse perspectives on syntax. ed.. Flora Klein-Andreu, 143-179. New York: Academic Press. Kratzer, Anne. 1988. Stage-level and individual-level predicates. Ms., University of Massachusetts, Amherst. Kuroda, S-Yuki. 1988. Whether we agree or not: a comparative syntax of English and Japanese, Linguisticae Investigationes 12:1-47. Lamarche, Jacques. 1991. problems for N-movement to NumP. Probus 3:215-236. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 9 2 Lamontagne, G., and L. Travis. 1986. The syntax of adjacency. McGill Working Papers in Linguistics 4. Lang, Mervyn F. 1990. Spanish word formation: Productive derivational morphology in the modern lexis. Groom Helm Romance Linguistic Series. London: Routledge. Larson, Richard K. 1988. On the Double Object Construction. Linguistic Inquiry 19:335-391. Lasnik, Howard and Mamoru Saito. 1984. On the Nature of Proper Government Linguistic Inquiry 15:235-89. Lebeaux, David. 1990. Relative clauses, licensing, and the nature of the derivation. Ms., University of Maryland. Lees, Robert B. 1960. The grammar of English nominalizations. Doctoral dissertation, MIT, Cambridge Mass. Lenz, Rodolfo. 1920. La oraciôn y sus partes. Revista de Filologia Espanola, Madrid. Levin, Beth, and Malka Rappaport. 1988. What to do with 8-roles? In Thematic relations. ed. Wendy Wilkins. San Diego, CA: Academic Press. Longobardi, Giuseppe. 1992. Proper names and the theory of N-movement in syntax and logical form. Ms., Université di Venezia. Lujan, Marta. 1980. Sintaxis y semântica del adjetivo. Madrid: Catedra. McCarthy, John J., and Alan S. Prince. 1990. Foot and word in prosodic morphology: the Arabic broken plural. Natural Language and Linguistic Theory 8:209-283. Malien, Enrique. 1989. The internal structure of Determiner phrases. Doctoral dissertation, Cornell University. Mallinson, Graham. 1986. Rumanian. London: Croom Helm. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 9 3 Martin, Juan. 1991. Subject anaphors, bound pronouns and the Binding Theory. In Proceedings of The Western Conference on Linguistics WECOL 91, eds. Katharine Hunt, Thomas A. Ferry, and Vida Samiian, 227-238. Department of Linguistics, California State University, Fresno. . . 1992. Wh-agreement in Spanish. Paper presented at the 22nd Linguistic Symposium on Romance Languages, University of Texas, El Paso. 1993. On extraction from NPs in Spanish. In Linguistic perspectives on the Romance languages: Selected papers from the XXI Linguistic Symposium on Romance Languages. ed. William J. Ashby, Marianne Mithun, Giorgio Perissinotto, and Eduardo Raposo, 303-313. Amsterdam: John Benjamins. 1993. Wh-movement in Spanish: structural analysis and theoretical implications. In International Journal of Basque Linguistics and Philology 27. May, Robert. 1985. Logical Form: Its structure and its derivation. Cambridge, Mass.: MIT Press. Milsark, G. 1974. Existential sentences in English. Doctoral dissertation. MIT, Cambridge, Mass. Montalbetti, Mario. 1984. After Binding: On the Interpretation of Pronouns. Ph.D. dissertation, MIT. Mourelatos, Alexander P.D. 1981. Events, processes, and states. In Syntax and semantics 14: Tense and aspect. ed. Philip J. Tedeschi and Annie Zaenen, 191-212. New York: academic Press. Napoli, Donna Jo. 1989. Predication theory: A case study for indexing theory. Cambridge: Cambridge University Press. Ouhalla, Jarmal. 1988. The syntax of head movement. A study of Berber. Doctoral dissertation. University College, London. Perlmutter, David, and Paul Postal. 1984. The 1- Advancement Exclusiveness Rule. In Studies in relational grammar 2. ed. David Perlmutter and Carol Rosen, 81-125. Chicago: University of Chicago Press. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 9 4 Pesetsky, David. 1987. WH-in situ: movement and unselective binding. In The representation of (in)definiteness. ed. E.J. Reuland and A. ter Meulen, 98-129. Cambridge, Mass.: MIT Press. . . 1988. Psych predicates, universal alignment and lexical decomposition. Paper presented at the UCLA Conference on the Lexicon and Syntax. Picallo, M. Carme. 1991. Nominals and nominalizations in Catalan. Probus 3:279-316. Pollock, Jean Yves. 1989. Verb movement, UG and the structure of IP. Linguistic Inquiry 20:365-424. Posser, William J. 1990. Ma. In Interdisciplinary approaches to language. Essays in honor of S.-Y. Kuroda. ed. Carol Georgepoulus and Roberta Ishihara, 565-593. Dordrecht: Kluwer. Pustejovsky, James. 1988. The syntax of Event structure. Ms., Brandeis University, Waltham, Massachussets. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, and Jan Svartvik. 1972. A grammar of contemporary English. New York: Seminar Press. Rappaport, Malka. 1983. On the nature of deriyed nominals. In Papers in Lexical Functional Grammar, ed. L. Levin, Malka Rappaport, and Annie Zaenen, 113-142. Bloomington, Indiana: Indiana University Linguistics Club. Real Academia Espanola (RAE). 1931. Gramatica de la lengua espanola. Madrid: Espasa-Calpe. 1978. Esbozo de una nueva gramatica de la lengua espanola. Madrid: Espasa-Calpe. Reinhart, Tanya. 1976. The Syntactic Domain of Anaphora. Doctoral dissertation, MIT, Cambridge, Mass. Riemsdijk, Henk van, and Edwin Williams. 1981. NP- structure. The Linguistic Review 1:171-217. Ritter, Elizabeth. 1988. A head-movement approach to construct-state noun phrases. Linguistics 26:909- 929. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 9 5 . 1990. Two functional categories in Noun Phrases: Evidence from Modern Hebrew. Ms., UQAM. 1993. Where’s gender? Linguistic Inquiry 24:795-803. Rivero, Maria Luisa. 1986. Binding in NPs. In Generative studies in Spanish syntax, ed. Yvonne Bordelois, Heles Contreras, and Karen Zagona. Dordrecht: Foris. Rizzi, Luigi. 1982. Issues in Italian syntax. Dordrecht: Foris. . . 1986. Null objects in Italian and the theory of pro. Linguistic Inquiry 17:501-557. . 1990. Relativized Minimality. Cambridge, Mass.: MIT Press. . 1991. Residual verb second and the Wh- Criterion. Ms., University of Geneva. Roberts, Ian. 1985. Implicit and Dethematized Subjects, doctoral dissertation. University of Southern California. 1986. The Representation of Implicit and Dethematized Subjects. Dordrecht: Foris. 1992. Verbs and diachronic syntax. Dordrecht: Kluwer. Roeper, Tom. 1983. Implicit arguments. Ms., University of Massachusetts at Amherst. Rothstein, Susan. 1988. Conservativity and the syntax of determiners. Linguistics 26:999-1019. Rozwadowska, Bozena. 1988. Thematic restrictions on derived nominals. In Thematic Relations, ed, Wendy Wilkins. San Diego, CA: Academic Press. Ruwet, Nicolas. 1972. Théorie syntaxique et syntaxe du francais. Paris: Seuil. Safir, Kenneth. 1984. Syntactic chains. Cambridge: Cambridge University Press. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 9 6 ... 1986. Relative clauses in a theory of binding and levels. Linguistic Inquiry 17:663-690. 1987. The syntactic projection of lexical thematic structure. Natural Language and Linguistic Theory 5:561-601. Salva. 1949. Gramatica de la lengua espanola. Paris: Hermes. Saltarelli, Mario. 1990. The subject of psych-verbs and case theory. Ms. , University of Southern California. Sanchez, Liliana. 1994a. Climbing up the tree: from modifiers. Paper presented in the Fourth Colloquium on Generative Grammar. ________. 1994b. Aspectual adjectives and the Event Structure of DP and VP. Ms. , University of Southern California, Los Angeles. Sciarone, A.G. 1970. La place de 1*adjectif en italien moderne. The Hague: Mouton. Seco, Manuel. 1972. Gramatica esencial del espanol: Introducciôn al estudio de la lengua. Madrid: Aguilar. Seco, Rafael. 1930. Manual de gramatica espanola. Madrid: Compania Iberoamericana de Publicaciones. Seiler, Hansjakob. 1978. Determination: A functional dimension for interlangue comparison. In Language universels: Papers from the Conference held at Gummersbach/Cologne. Germany, October 3-8, 1976, ed. Hansjakob Seiler, 301-328. Tübingen: Gunter Narr. Selkirk, Elizabeth. 1977. Some remarks on Noun Phrase structure. In Formal syntax, ed. P. Culicover, T. Wasow, and A. Akmajian. New York: Academic Press. Siloni, Tali. 1991. Noun raising and the structure of Noun Phrases. Ms., University of Geneva. Sole, Yolanda, and Carlos A. Sole. 1977, Modern Spanish syntax: A study in contrast. Lexington, Mass.: D.C. Heath. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 9 7 Sportiche, Dominique. 1988. A theory of floating quantifiers and its corollaries for constituent structure. Linguistic Inquiry 19:425-449. ________ . 1990. Movement, agreement, and case. Ms., UCLA. Sproat, Richard, and Chilin Shih. 1990. The cross- linguistic distribution of adjective ordering restrictions. In Interdisciplinary approaches to language. Essays in honor of S.-Y. Kuroda. ed. Carol Georgepoulus and Roberta Ishihara, 565-593. Dordrecht: Kluwer. Stowell, Timothy. 1981. Origins of Phrase Structure. Doctoral dissertation, MIT, Cambridge, Mass. ________ . 1983. Subject across categories. The Linguistic Review 2:285-312. . 1989. Subjects, Specifiers, and X-bar Theory. In Alternative conceptions of Phrase Structure, ed. Mark R. Baltin, and Anthony S. Kroch. Chicago, Illinois: University of Chicago Press. Suner, Margarita. 1982. Syntax and semantics of Spanish presentational sentence-types. Washington D.C.: Georgetown University Press. Szabolcsi, Anna. 1983. The possessor that run away from home. The Linguistic Review 3:89-102. Taraldsen, Knut Tarald. 1990. D-projections and N- projections in Norwegian. In Grammar in progress. Glow essays for Henk van Riemsdijk. in Joan Mascaro and Marina Nespor, 419-431. Dordrecht: Foris. Tellier, Christine. 1991. Licensing theory and French parasitic gaps. Dordrecht: Kluwer Academic Publishers. Tenny, Carol. 1987. Grammaticalizing Aspect and Affectedness. Ph.D. dissertation, MIT, ________ . 1988. The Aspectual Interface Hypothesis: The connection between syntax and lexical semantics. In Studies in generative approaches to aspect, ed. Carol Tenny. Lexicon Project Working Papers 24, Center for Cognitive science, MIT. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 9 8 Torrego, Ester. 1985. On empty categories in nominals. Ms., UMASS, Boston. ________ . 1988. Evidence for Determiner Phrases. Ms., UMASS Travis, Lisa. 1984. Parameters and Effects of Word Order Variation. Doctoral dissertation. MIT, Cambridge, Mass. Valois, Daniel. 1991. The internal syntax of DP. Doctoral dissertation. UCLA, Los Angeles, California. Vendler, Zeno. 1967. Linguistics in philosophy. Ithaca, N.Y.: Cornell University Press. Vergnaud, Jean-Roger. 1985. Dépendances et niveaux de représentation en syntaxe. Amsterdam: John Benjamins. Vergnaud, Jean-Roger, and Maria Luisa Zubizarreta. 1990. On the form and interpretation of definite Noun Phrases. Ms. University of Southern California, Los Angeles, California. ________. 1992. The definite determiner and the inalienable constructions in French and in English. Linguistic Inquiry 23:595-652. Waugh, Linda R. 1977. A semantic analysis of word order. Leiden: E. J. Brill. Wexler, Kenneth and Peter W. Culicover. 1980. Formal principles of language acquisition. Cambridge, Mass.: MIT Press. Williams, Edwin. 1977. Discourse and logic form. Linguistic Inquiry 8:101-139. 1980. Predication. Linguistic Inquiry 11:203- 238. . 1981. Argument structure and morphology. The Linguistic Review 1:81-114. . 1982. NP cycle. Linguistic Inquiry 13:277-295. . 1985. PRO and subject of NP. Natural Language and Linguistic Theory 3: 297-315. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 9 9 Zagona, Karen. 1990. Times as temporal argument structure. Ms., University of Washington, Zubizarreta, Maria Luisa. 1979. Extraction from NP and a reformulation of subjacency. Ms., Cambridge, Mass.: MIT. 1985. The relation between morphophonology and morphosyntax: The case of Romance causatives. Linguistic Inquiry 16:247-289. . 1987. Levels of representation in the Lexicon and in the Syntax. Dordrecht: Foris. . 1992. The lexical encoding of scope relations among arguments. In Syntax and Semantics 26: Syntax and the Lexicon, ed. Tim Stowell and Eric Wehrli. San Diego: Academic Press. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

The grammaticalization of object agreement in Spanish

PDF

Phonological and stylistic variables in Puerto Rican Spanish

PDF

The distribution of lexical and null subjects in the acquisition of L2 Spanish by adult speakers of English

PDF

Social deixis in a Los Angeles Spanish-English bilingual community: Tu and Usted patterns of address

PDF

Definiteness types in Spanish: A study of natural discourse

PDF

Encoding of politeness in Spanish and Polish: a cross-linguistic study

PDF

Syntactic structures in nominals: A comparative study of Spanish and Southern Quechua

PDF

Conditions on null objects in Basque Spanish and their relation to leismo and clitic doubling

PDF

The Spanish continuum in Peruvian bilingual speakers: A study of verbal clitics

PDF

The syntax of focus and topic in Mandarin Chinese

PDF

The maternal theme in Garcia Lorca’s folk tragedies

PDF

On the status of possessives

PDF

Intra-lexical noun-verb dissociations: Evidence from Chinese aphasia

PDF

Nation, race, and heterogeneity in the literature of the Americas

PDF

Social networks and phonetic variation: A family network in Santiago de Chile

PDF

A critical edition with a study of the style of "la relacion" by Alvar Nunez Cabeza de Vaca

PDF

Isotope shift in the atomic spectrum of silicon

PDF

The syntax of clitic doubling in modern Greek

PDF

The liberal muse: Juvenal's sexual persona and the purpose of satire

PDF

A critical edition of the Escorial manuscript of 'historia de los indios de la neuva espana' of fray toribio de benavente (motolinia) (spanish text)

Asset Metadata

Creator Martin, Juan (author)

Core Title On the syntactic structure of Spanish noun phrases

Contributor Digitized by ProQuest (provenance)

Degree Doctor of Philosophy

Degree Program Spanish

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

Tag Language, General,language, linguistics,Language, Modern,OAI-PMH Harvest

Language English

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c17-471555

Unique identifier UC11352202

Identifier 9614046.pdf (filename),usctheses-c17-471555 (legacy record id)

Legacy Identifier 9614046-0.pdf

Dmrecord 471555

Document Type Dissertation

Rights Martin, Juan

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 au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA