ABSTRACT Title of Dissertation: IMPOVERISHED MORPHOLOGY AND A-MOVEMENT OUT OF CASE DOMAINS Cilene Aparecida Nunes Rodrigues, Doctor of Philosophy, 2004 Dissertation directed by: Professor Norbert Hornstein Department of Linguistics This dissertation suggests that referential 3 rd P null subjects in Modern Brazilian Portuguese (BP) and Finnish are residues of A-movement, rather than null pronouns. These grammars exhibit weak 3 rd P verbal and possessive agreement morphology, and do not obey the Avoid Pronoun Principle, allowing non-emphatic overt pronouns in subject position. This state of affairs has affected the licensing of referential null subjects, which are licensed only within embedded domains. I correlate the loss of agreement with this peculiar behavior of null subjects and advance the hypothesis that BP and Finnish are not pro-drop grammars, arguing on empirical grounds that in BP and Finnish null subject inside the embedded clauses and possessive DPs are residues of A-movement. Putting it boldly, these null subjects are salient copies of their antecedents. The arguments I present in favor of a movement analysis are: (i) Finnish and BP null subjects have an anaphoric behavior, requiring a sentential antecedent, which is the closest c-commanding DP. (ii) They cannot occur within relative clauses if the head of the relative clause intervenes between them and their antecedents. (iii) They display all the diagnostics used to characterize obligatory control as formed by movement; (iv) They do not occur inside paratactic constructions. (v) Inside coordinated DPs, they must occur in an across-the-board fashion. (vi) Floating quantifiers and participial forms within their c-command domains agree with their antecedents in f-features. Presupposing that in pro-drop languages pro is the verbal agreement morpheme (Agr) itself, I suggest in that in BP and Finnish Agr underwent f-degradation and was lexically reanalyzed as part of the verb. However, it is hypothesized that in these languages Agr retained a D-feature, and, consequently, it can satisfy the EPP feature of a Case-checking functional projection. As result, in these grammatical systems, a DP can undergo A-movement out of a Case-domain. IMPOVERISHED MORPHOLOGY AND A-MOVEMENT OUT OF CASE DOMAINS by Cilene Aparecida Nunes Rodrigues Dissertation submitted to the faculty of the Graduate School of the University of Maryland, College Park in partial fulfillment of the requirements for the degree of Doctor of Philosophy 2004 Advisory Committee: Professor Norbert Hornstein, Chair Professor Cedric Boeckx Professor Jeffrey Bub Professor Howard Lasnik Professor David Lightfoot Professor Paul Pietroski Professor Juan Uriagereka ?Copyright by Cilene Aparecida Nunes Rodrigues 2004 ii ACKOWLEDGEMENTS I deeptly thankful to all of those who, directly or inderectly, helped me on the completion of this reaserch. Particularly, I would like to express gratitude to the following group of people: Ticho Spotachi & Tet? Paviani, Hornstein, Lu, Shong, Gisa, Caeiro, Claudicare, Glaines, Sibelica, Lightfoot, Dazin, Marc?o, Renatinho, Uriagereka, Fofolete, Alexandros, a Vendedora de Chinelos, Pirilampo, Funilinho, O da Terceira Margem do Rio, Malfatti, Tina, Rutilica, Cora Coralina, Lia, Miss. Hodges, Quitiquinho, Kazanina, Pietroski, Lunguinho, Edu, Dani, Bub, Riobaldo, Holando, Serena do Vale, Adonias Filho, Nadorica, Kathi, Tor, Guta, Guchi, Lobato, Diadorim, Dona Fia, Lasnik, Lutfi, Saramago, Urpo, Klea, Fernando Pessoa, Coronel Ponciano de Azeredo Furtado, Herm?genes, Magee, Chomsky, Senhor do Pe-leve, Manu, Luiz, Tistu, Boeckx, Juca.net, Ros?lia e seu Pequeno Pr?ncepe, Jukka, Quincas Borba, Meiroca, P?o, Hale, Soledad, Paulinho da Viola, Jaeggli, Salles, Kathi, The Great Pretender and Irazinho. From them, I learned a great deal of what I know. iii TABLE OF CONTENTS Acknowledgementsii Chapter 1:Stating the Problem1 1The leading question1 2The Answer to be provided5 3The Outline of the Thesis6 Chapter 2:Exploring Movement 10 1Preliminaries 10 2Movement 12 2.1The Composition of Movement15 2.2Sideward Movement17 3Copy Deletion and Chains 20 4Obligatory Control: Substituting Construals for Movement25 4.1PRO Climbing at LF27 4.2Feature of Movement29 4.3Overt NP Movement 31 5Movement into Theta-Positions 35 5.1The Ban on Movement into Theta-Positions 39 5.2Moving into Theta-Positions 51 6Conclusions 63 Chapter 3:Verbal Movement and Null Subjects 65 1Preliminaries 65 2Verbal Agreement Morphology66 3Null Subjects 70 3.1Expletives and Generic Null Subjects71 3.2Referential Null Subjects76 4Matrix Null Subjects: Cases of Topic Deletion82 5Embedded Null Subjects94 5.1Absence of Logophoric Properties 94 5.2Previous Analysis98 iv 6Conclusions 112 Chapter 4:A-Movement out of Finite Clauses114 1Preliminaries 114 2Implementing the Proposal119 2.1 Verbal Agreement Morphology and Null Subject Pronouns119 2.2 Deriving Embedded Referential Null Subjects by Movement123 2.3CP as a Phase and Intermediate Movement130 3Evidence for a Movement Analysis133 3.1Anaphoric Behavior134 3.2Locality Matters 136 3.3Obligatory Control Properties144 3.4The Connective Como in Brazilian Portuguese159 3.5Referential Null Subject Inside Relative Clauses168 4Controlling Gender Agreement180 4.1Feminine Gender Agreement 180 4.2 Controlling Feminine Gender Agreement in Non-Finite Control Configurations185 4.3 Controlling Feminine Gender Agreement in Finite Control Configurations194 5Referential Null Subjects Inside Finite Adjunct Control199 5.1Right Adjuncts 199 5.2Left-Dislocated Adjuncts211 5.3Revisiting Modesto?s Argument 217 6Conclusions 229 Chapter 5:Loss of Nominal Morphology and Null Possessors240 1Preliminaries 240 2The Nominal Structure241 2.1The DP Hypothesis241 2.2What?s Between DP and NP244 2.3The DP Structure Adopted in This Thesis 248 3Possessive Descriptions 251 3.1The Possessive Thematic Relation253 3.2The Positions of the Possessor258 3.3Null Possessors 262 4Possessives in Brazilian Portuguese267 4.1Loss of Pronominal Forms267 4.2Brazilian?s Null Possessor272 4.31 st P Null Possessors: Cases of Topic Deletion273 5Possessives in Finnish 276 5.1 Possessive Agreement Morphology and Availability of Null Possessors276 5.2Speculations on Finnish Possessive Suffixes281 v 6Conclusions 287 Chapter 6:A-Movement out of Nominals 289 1Preliminaries 289 2Implementing the Proposal299 2.1Loss of Nominal Morphology and Null Possessors 299 2.2Deriving Null Possessors from Movement302 2.3DP as a Phase and Intermediate Movement 304 3Evidence for a Movement Analysis307 3.1Anaphoric Behavior 307 3.2Locality Matters309 3.3Obligatory Control Properties314 3.4Null Possessors and Specificity 323 3.5Null Possessors Inside Relative Clauses329 4Null Possessors Inside Adjuncts332 5Null Possessors Inside Coordinate Structures337 6Conclusion 343 Chapter 7: Conclusion 352 1A Summary of the Content 352 2Some Theoretical Implications of This Research355 3Possible Ramifications of the Proposed Analysis 355 References 360 1 CHAPTER 1 STATING THE PROBLEM 1.1The Leading Question The research projection of this dissertation is an investigation ? within the Generative Grammar framework ? of sentential and nominal null subjects in two of the so-called partial pro-drop languages: Modern Colloquial Brazilian Portuguese (BP, henceforth) and Finnish. In short, what follows is an answer to the question in (1). (1)What is the syntactic nature of referential null subjects in BP and Finnish? The possibility of dropping any pronominal subject is considered to be a manifestation of the pro-drop parameter (see, among others, Chomsky 1981, Rizzi 1982, 1986, Jaeggli and Safir 1989). Thus, Romance grammars like Italian, in which expletives as well as referential subjects can be null (cf. (2)), are taken to be bona fide examples of a positive setting of this parameter. 2 (2)a.Piove rain-3Sg ?It rains? b.Ho telefonato have-3Sg called ?S/he has called? Partial pro-drop languages also display null expletives, as (3) illustrates. 1 However, in matrix clauses, only 1 st and 2 nd Person referential null subjects are allowed to be null (cf (4)-(5). 2 (3)a.Chove (BP) rain-3Sg b. Sataa(Finnish) rain-3Sg ?It rains? (4)a.Falei com o Jo?o ontem a noite (BP) spoke-1Sg with the Jo?o yesterday at night ?I spoke with Jo?o yesterday night? 1 The term partial pro-drop has been ambiguously use to characterize two types of grammars: (a) grammars like German where expletive subjects can be omitted, even though referential null subjects cannot; and (b) grammars like BP and Finnish which can drop referential subjects, but only in a restricted set of configurations. In this thesis, putting aside grammars of type (a), I shall employ the term partial pro-drop unambiguously in reference to grammars of type (b), which are also called semi null subject languages. 2 BP allows only 1 st P null subjects in matrix clauses, see chapter 3. 3 (5)a. Istun si? (Finnish) sit-1Sg here b.Istut si? sit-2Sg here ?I/you/ am/are sitting here? 3 rd P referential null subjects are licensed only in embedded clauses: (6)a.* Embarcou no trem(BP) boarded -3Sg in.the train b.* Nousin junaan (Finnish) stepped-3Sg train-into ?S/he boarded the train? (7)a.H?n 1 kertoi ett? e 1/*2 nousin junaan he-Nom said-3Sg that stepped-3Sg train-into b.Ele 1 disse que e 1/*2 embarcou no trem he said-3Sg that boarded-3Sg in.the train ?He said that he boarded the train? Nevertheless the licensing of these subjects is unexpected. In pro-drop grammars (e.g. Italian and Spanish) 3 rd P referential null subjects pattern like pronouns in being free in 4 reference. Conversely, in partial pro-drop grammars, 3 rd P null subjects behave like anaphors, being referentially dependent on a nominal phrase projected within the sentence in which they occur. Thus, as the assigned indexes show, the null subjects in (7) must be interpreted as referring back to be matrix subject. This anaphoric behavior is also observed in null possessors, i.e., 3 rd P null subjects of possessive nominals expressions. The Spanish datum in (8) suggests that in pro-drop languages a null possessor may receive a deictic interpretation, referring to an entity previously mentioned in the discourse. In BP and Finnish, however, this is not possible. As shown in (9), BP and Finnish 3 rd P null possessors have an anaphoric behavior, requiring a syntactic antecedent. (8) ? Parece que [el padre e] se muri? ayer seem-3Sg that the father SE died yesterday ?It seems that her/his father died yesterday? (9)a.[a Julia] 1 lavou [as m?os e 1/*2 ] . the Julia washed-3Sg the-Pl hands ?Julia washed her hands? b.Maija 1 pit?? [ kirjastansa e 1/*2 ] Maija-Nom like-3Sg book-Ela-3 ?Maija like her book? 5 In conclusion, the state of affairs is as follows: in pro-drop grammars, nominal or clausal 3 rd P referential null subjects are pronominal null categories (pro), whereas, in partial pro-drop languages, they are anaphoric. This raises a concern for the pro-drop parameter. Shall we take BP and Finnish to be pro-drop grammars? If the answer is positive, then we need to understand how a fixed parameter can have different outcomes. While a positive setting of the pro-drop parameter in Italian and Spanish opens up the availability of pronominal null subjects, in BP and Finnish the fixation of this parameter results in anaphoric referential null subjects. On the other hand, analyzing the grammars under consideration as non pro-drop, one needs to explain the data in (4), (5), (7), and (9) and explain the nature of the empty category observed there. At any rate, the question posed in (1) is inevitable, and any theory about the pro-drop nature of BP and Finnish has to be built upon an answer for that question. 1.2The Answer To Be Provided Romance pro-drop grammars are known to have rich agreement morphology. This is often analyzed as correlated with the licensing of pro (cf. Taraldsen, 1980, Rizzi 1986, Jaeggli and Safir 1989). Interestingly, as discussed in chapters 3 and 6, BP and Finnish display weak agreement morphology inside clauses and nominals. In addition to that, these languages differ from pro-drop languages in that they are insensitive to the Avoid Pronoun Principle (cf. Chomsky, 1981), allowing non-emphatic overt subject pronouns. This has led me to hypothesize that we are leading with non pro-drop 6 languages. Under this hypothesis, however, these languages are expected to ban null subjects altogether. On the other hand, as shown above this expectation fails because 1 st and 2 nd Person null subjects seem to be licensed matrix clauses, and 3 rd Person null subjects inside embedded clauses. However, as the present research shows, these null subjects should not be taken as counterevidence for the proposed hypothesis. First, matrix null subjects result from a topic deletion process, along the lines proposed by Huang (1984), while anaphoric 3 rd P null subjects are the result of A-movement. That is, the empty subjects in (7) and (9) are traces of their antecedents. Thence, I answer question (1) by suggesting that the empty subjects in BP and Finnish are residues of movement, rather than genuine instances of pro. Note however that this analysis has two important features. First, it involves movement into theta-positions. Second, it allows A-movement out of Case domains. As discussed in chapter 2, I advocate in favor of a theory that sanctions movement into theta- positions. As for the allowance of A-movement out of Case domains, I suggest that this is related to the fact that the verbal and possessive agreement systems of BP and Finnish are deficient in f-features, being able to license null expletives, but not referential null arguments. 3 1.3The Outline of The Thesis Chapter two is a layout of the theoretical background, focusing mainly on minimalist views of Move, a syntactic operation of the computational system. As will be shown in chapters four and six, the null subjects studied here display all the properties 7 that are used to characterize obligatory control as movement. Thus, in chapter two, I will introduce these properties vis-?-vis analyses of obligatory control as resultant from movement. I will follow Hornstein?s (1999, 2001) NP-movement, and, as consequence, I will commit myself to the following ideas: (a) sideward movement (Nunes 1995, 2004, among others) is applied by the computational system, which also performs (b) movement into theta-positions. In chapter three, I hypothesize that BP and Finnish are not null subject languages due to their weak verbal agreement morphology. In accordance with this hypothesis, it is suggested that 1 st and 2 nd Person null subjects in matrix are not genuine null subjects, being rather the result of topic deletion. Moreover, I will revise previous analyses for embedded 3 rd P null subjects, paving the way for the movement analysis I will argue for in chapter four. 3 rd P referential null subjects inside embedded clauses are the topic of chapter four. Mainly, I will defend the A-movement analysis discussed in 1.1.2, suggesting that BP and Finnish allow this sort of movement because their weak verbal agreement system are still able to license null expletives. As evidence for movement, I will show the following: (i) the subjects under consideration are anaphoric, requiring a syntactic antecedent (cf. (7); (ii) their DP is the closest c-commanding DP, in accordance with the Minimal Link Condition, taken to be a condition on movement (cf. Chomsky, 1995). (iii) In BP, these subjects fail the resumption test, being disallowed inside relative clauses. Finnish relative clauses might have 3 rd P referential null subjects, but only because the nominal head of the relative 3 The relation between movement and f-defectiveness of agreement has its antecedents in other works;.see Boeckx (2003) and references therein. 8 clause undergoes object shift. (iv) In Finnish and BP, these subjects display all the diagnostics used to characterize obligatory control as the residue of movement; (v) In BP, clauses embedded under the connective como cannot host a 3 rd P referential null subjects, arguably because these clauses are left branches inside a DP. The last source of evidence that I discuss is related to the valuation of the gender feature of past participles and floating quantifiers. It will be observed that in Romance non-finite obligatory control configurations, the antecedent of PRO controls the value of the gender feature of a past participle form or of a floating quantifier embedded under the c-command domain of PRO. In BP, the same phenomenon is observed in structures involving a 3 rd P referential null subject. As discussed, this follows straightforwardly from a movement analysis. In the last section of chapter 4, I consider 3 rd P referential null subjects inside finite adjunct clauses, suggesting a sideward-movement analysis. Since the structure of nominal descriptions is less studied than the structure of clauses, I open chapter five discussing this topic. Then, I show that BP and Finnish both display weak possessive agreement morphology; thus, under the hypothesis that loss of morphology (f-features degradation) is responsible for loss of referential null subjects, I conclude that the 3 rd null possessors found in these languages (cf. (6)) are not genuine null pronominal categories. In chapter six, I extend the movement analysis proposed in chapter four to the 3 rd P null possessors illustrated in (9). As will be shown, these possessors are akin to the null subjects discussed in chapter four, in that: (i) they are anaphoric, requiring a 9 syntactic antecedent, which is the closest c-commanding DP; (ii) they fail the resumption test, being disallowed inside relative clauses, and (iii) they are obligatory controlled categories. The last section of the chapter is dedicated to the occurrence of null possessors inside adjuncts and in coordinate structures. Following the analysis proposed in 4.5, these possessors will be analyzed as gaps formed via sideward movement. In Chapter seven, I summarize of the content the dissertation, provide remarks on its contribution to our understanding of the theory of grammar, and discuss possible ramifications of the movement analysis suggested here. 10 CHAPTER 2 EXPLORING MOVEMENT 2.1.1 Preliminaries As already stated, this research seeks an answer to the question in (1): (1) What is the syntactic nature of referential null subjects in Brazilian Portuguese and Finnish? To answer this question, I will make some theoretical commitments that will be introduced in the present chapter. First, my answer accords to the methodological guidelines of Minimalist Program, the most recent version of the Principles & Parameters model (cf. Chomsky 1993, 1995, 2000, 2001a. 2001b). 1 The Minimalist Program is methodologically devoted to the development of well-designed linguistic theory, where goodness is based upon the notions of elegance, parsimony and simplicity. Thus, a minimalist grammar is a device that employs the 11 minimum necessary to deliver a pairing of sound & meaning. Therefore, given that sound and meaning are drawn from the interface levels (PF and LF), a particular linguistic expression is just a formal object that satisfies the requirement (bare output conditions) of these interfaces in an optimal way. The Government-Binding theory was a four-level model, containing, in addition to PF and LF, two representational levels internal to the grammar: D-Structure and S- Structure. D-structure was the level in which lexical properties like subcategorization and thematic relations were satisfied. S-Structure was composed by series of conditions and principles, including principles of the Binding theory, the subjacency condition, Case theory, etc. D-structure was also the starting point of a derivation, a single structure formed by combination of lexical items in accordance with the X-bar theory (see Chomsky 1970, 1981, 1986a, Jackendoff 1977 and Stowell 1981). Thus, D-structure was the ?base?, the feeder of the transformational component. Minimalism rejects this four- level model. It dispenses with D- and S-Structure by proposing that derivations are evaluated only by the bare output conditions of the interfaces PF and LF. It also gives up the idea that a derivation starts with a single structure that feeds the transformational component. Rather, it is assumed that derivations are built through a generalized transformation process that might assemble a tree by combining already assembled trees. The Minimalist Program also departs from the Government-Binding theory in the way that movement is conceived. Inside this program, movement does not apply freely. It is a last resort operation, trigged only by the necessary of checking a feature 1 For expositions and discussions on the minimalist goals, see Uriagereka (1998), Martin and Uriagereka (2000), Jenkins (2000), Freidin and Vergnaud (2001), Lasnik (2002), among others. 12 uninterpretable at the interfaces. This new view of movement is the topic of the next section. Here is the arrangement of the sections: 2.2 presents the Copy theory of movement as proposed in Chomsky (1993, 1995a) and also sideward movement (Nunes 1995, Bobaljik and Brown 1997 and Uriagereka 1998). 2.3 lays out some proposals for copy deletion at PF. The focus is on the role of chains in the process of deletion. 2.4 is dedicated to the presentation of some minimalist proposals that relate obligatory control to movement. A discussion on movement into theta-positions is offered in 2.5, and 2.6 summarizes the theoretical commitments of this dissertation. 2.2Movement Given everything we know about natural language, one of its indisputable attributes is displacement: an expression can be pronounced in a position, but interpreted in another. The grammar proposed by Chomsky at the mid of the last century (cf. Chomsky 1957, 1965) captures displacement via grammatical transformations. 2 Take as an example the passive sentence in (2). (2) and (3) are akin sentences because (2) is derived from (3) by application of the passive transformation in (4). In the model laid out Syntactic Structure, grammatical transformations have two parts: a structural analysis (4a) that indicate the type of structure to which the transformation applies, and a structural change (4b) that indicates the type of alteration the target structure will undergo. Therefore, (2) is result of the application of (4) to (3). The structural change in 2 See Lasnik (2000) for a review of this theory. 13 (4b) applies to the structure of (3), changing, among other things, the subject and object positions. (2) The wine was drunk by the guests (3) The guests drunk the wine (4) Passive Transformation (Tp) a. NP ? Aux ?V- NP b.X 1 ? X 2 ? X 3 ? X 4 ?X 4 ? X 2 + be+ en ? X 3 ? by + X 1 In the classical Government-Binding theory (cf. Chomsky 1981, 1986a, among others) displacement is taken to be a consequence of move-a, a rule that constitutes the transformational component, as originally suggested in Chomsky and Lasnik (1977). Move-a displaces phrasal constituents, leaving co-indexed traces behind. For concreteness, consider the application of move-a in (5). (5a) is the deep structure of (5b). By application of move-a, the NP the man is moved from the embedded subject position to the subject position of the matrix clause, and a trace co-indexed with the man is placed in the base position. (5) a.[ seems [ [ NP the man] to be sick]] b.[[ NP the man] 1 seems [t 1 to be sick]] 14 There are no constraints on Move-a. 3 Constituents move freely, moving any anything to anywhere. Thus, constituents move for no reason, and if the results of their movement do not violate any principle, the outcome derivations are grammatical. In the Minimalist Program, movement (Move) is taken to be a last resort operation. Its application is legitimate only if it is necessary for convergence. Technically movement is licensed only when it creates a checking configuration, where a feature that is uninterpretable at the interfaces, PF and LF, is checked off. Thus, there is no grammatical derivation in which a constituent is moved for no reason. In other words, vacuous application of Move is not Move for it falls outside the definition of Move, a last resort operation. The Minimalist program also opposes traces as part of the grammar. Traces as defined by the Government-Binding theory are representational empty categories that are introduced during the course of the derivation as a by-product of movement. Chomsky?s (1995a) implementation of the Minimalist Program requires that any structure is constituted only of elements already presented in the lexical items selected from the lexicon (Inclusiveness Condition ? Chomsky 1995a: 228). Hence, the notion of traces introduced above is incompatible with this implementation. 3 Lasnik and Saito (1992) defend a more radical version of move-a, Affect-a which does anything to anything. 15 2.2.1 The Composition of Movement Chomsky (1993:202) suggests the copy theory of movement to account for reconstruction effects. According this theory, the launching site of movement is occupied by a copy of the moved expression. This copy is deleted at PF, but remains at LF, providing material for reconstruction. The introduction of copies into the computational system made feasible a theory in which movement is not a primitive operation, but the interaction of two basic operations: Copy and Merge. Copy doubles a syntactic object by copying all of its features. 4 The syntactic object thus copied is then merged with the target of movement. Chomsky (1995a: 226) defines Merge as an operation of the computational system that takes two syntactic objects and replaces them by a new combined one, as illustrated in (4) 5 . Note that (6) is not in accordance with the standard X-theory, but with the bare version of the X?-theory proposed in Chomsky (1995a, 1995b), where the distinction between heads and terminal nodes is eliminated. 6 (6)a. [ VP [love Mary]] b. John Merge c. [ VP John [ V? [love Mary]]] 4 But see Groat and O?Neils (1996) proposal, which is presented in section 2.3. 5 (6) illustrates a case of substitution. In cases of adjunction, merge creates a two-segment category (cf. Chomsky 1995a:248). 6 This discussion is eliminated because vacuous (non-breaching) is redundant since both the head and the terminal node encode the same featural information. But see Guimar?es (2000) for a defense of vacuous projection inside the Minimalist Program. 16 Another innovation of this framework is the suggestion that Move, contrary to Select and Merge, 7 has some derivational costs (cf. Chomsky 1995a, 2000). The reasoning of this suggestion is the following: if Select and Merge do not apply, no derivation is generated. Therefore, since economy issues are raised for convergent derivation only, Select and Merge do not fall within the domain of these issues. As for Move, a derivation is generated without its application. The derivation thus formed may not converge, but it is generated. That is, application of Move is to guarantee convergence. Hence, Move is subject to economy considerations. The structure (7) illustrates well the role of economy in constraining Move. (7)[ g there seems [ b to be [ a a man in the room] Chomsky (1995a) assumes that, at stage b, merger of a category into spec of TP is required to satisfy the properties of the infinitival T. Thus, if the numeration contains an expletive, there are two optional ways to fill up the spec of the infinitival TP: (a) merger of the expletive it, (b) copy and merger (Move) of the DP a man. Both options lead to a convergent route. Hence, application of Move is not necessary for convergence, thereby economy considerations regarding the cost of the derivation blocks movement of the DP the man, and there is merged in spec of the infinitival TP. At stage g, the numeration has already been exhausted, and the requirements of the matrix T can be satisfied only if 7 Select is the operation that takes a lexical item from numeration, reduces its index by 1 and introduces it into the derivation (cf. Chomsky 1995a: 226). A Numeration is defined by Chomsky (1995a: 225) as a set of pairs formed by a lexical item and its index, 17 Move applies copying and merging the expletive with matrix TP. Thus, since at this stage Merge is not an option, Move is allowed to apply. Hornstein (2001) also reasons that the compositionality of Move makes its application more expensive than application of Merge. Since Move composed by Merge and Copy, it is arguably the case that application of just Merge is cheaper than application of Merge and Copy. In conclusion, Merge has preference over Move. This preference is known as the Merge-over-Move Condition (cf. Castillo et al. 1999 and Hornstein 2001). It is worth emphasizing that this condition is an economy metric that applies to convergent derivations. This amounts to saying that Move is allowed to supercede Merge if it is necessary to ensure the convergence. I will adopt the minimalist framework outlined above. But, for expository purpose, I will use the trace notation when convenient. Thus, henceforth, the term ?trace? is to be understood as a shorthand form for ?silent copies formed by copy-and-deletion.? 8 2.2.2 Sideward Movement Chomsky (1993,1995) proposes that movement obeys the Extension condition, extending the structure to which it applies. Thus, when movement applies to a phrase marker K, K becomes L, ?which includes K as a proper part? (Chomsky 1995: 190). However, head movement, an adjunct process, violates the extension condition in that it 8 On deletion of copies, see 2.3. 18 does not extend the target phrase marker. This result forced Chomsky to propose a weak version of the extension condition, which holds only for substitution operations. 9 Yet, Nunes (1995), Bobaljik and Brown (1997), exploring an idea originally developed by Juan Uriagereka in the early nineties (cf. Uriagereka 1993), argued that head adjunction are in accordance with the Extension Condition in a system that allows sideward movement. 10 Their line of reasoning is the following: The VP in (8) conforms to Extension, if its derivation unfolds as shown in (9). The DP this dog (9a) and the VP likes me (9b) are first built in different sub-structures. After that, the DP merges with (9b), forming the VP in (9c). (8) [ VP this dog [ V? likes me]] (9)a. DPb. VPc. VP 2 2 3 this dog likes me DP V? 2 2 this dog likes me Therefore, a derivation might proceed by building parallel structures (9a,b) and storing them in the derivational workspace. Nunes (1995), Bobaljik and Brown (1997) claim that availability of such derivational process makes viable a theory in which transformational operations affect parallel structures. Thus, sideward movement is a possibility to be tested. 9 For a discussion on Extension, see Kitahara (1995). 10 This type of movement has received many names: Nunes named it sidewards movement, Bobaljik and Brown proposed the term interarboreal movement, and in Uriagereka (1998) it is called parallel movement. 19 Sideward movement proceeds as shown in (10). A constituent (ZP) of the sub- structure K is copied and merged with the sub-structure L. Note that movement of ZP extends the structure of L as required by the Extension Condition. (10) K L 2 2 K Y ZP L Y ZP L M Head movement is carried on in the same way, as (11), a case of V-to-T movement, exemplifies. First, the VP is built (11a). T is selected and placed in the derivational workspace as a separate sub-structure (11b). Next, the verb undergoes sidewards movement and adjoins to T, forming the complex head in (11c). Finally the complex head T merges with the VP, forming the TP in (11d): (11)a. VP b.T ? c. T ? d . TP 2 2 3 V ? T V T VP 2 2 T V 1 t 1 ? The sidewards movement of V to T enlarges the structure of T, as shown in (10c). Thus, likewise movement of maximal categories (substitution), movement of heads (adjunction) also obeys the Extension Condition. Nunes (1995, 2004) extends the application of sidewards movement to 20 parasitic gap constructions, proposing that parasitic gaps are the result of movement between sub-structures. For him, (12) is derived as in (13). The prepositional phrase and the VP are assembled in parallel and the wh-phrase which paper undergoes movement from the object position inside the prepositional phrase to the object position inside the VP. 1 After that, the remnant of prepositional phrase adjoins to the VP, and, when the matrix CP is built, which paper moves to spec of CP, where it checks the relevant wh- feature. (12) Which paper did you file without reading (13)a. [ PP without reading [which paper]]b. [ VP file [which paper]] c. [ CP [which paper] 1 did [ TP you [ VP [ VP file t 1 ] [ PP without reading t 1 ]]]] The way Nunes implemented sideward movement is more complex than I have presented so far, as he assumes that movement is an interaction of the independent operations of Copy, Merge, Form Chain and Chain Reduction. 2.3Copy Deletion and Chains Inside the copy theory of movement, an important issue concerns non-pronounced copies. For instance, consider the passive structure in (14). Only the highest copy of 1 One issue about the derivation in (13) is that it imposes a direction on movement: the wh-phrase moves from the adjunction clause to a verbal complement position inside the main clause, rather than the opposite. Nunes and Uriagereka (2000) and Hornstein (2001) suggest that this reflects the fact movement always takes place from more to less embedded domains. 21 John, i.e., the copy in spec of TP, is pronounced, the lower copy is silent. Chomsky (1993, 1995a) proposes that this is the case because copies are subject to deletion at PF. But the question is why does deletion target the lower copies? (14) [ TP John 1 [ T? was [ VP seen John 1 ]]] Nunes (op. cit.) addresses the deletion issue by arguing that chains are subject to linearization at PF. Elimination of all members of chain but one is forced by the LCA (Linear Correspondence Axiom ? Kayne 1994), which maps the terminals of a phrase marker into a linear order. 12 Thence, Chains have to have their members reduced to one because a linear order is asymmetric (if a precedes b, b does not proceed a) and irreflexive (if a precedes b, a ? b). Chain reduction (15) is taken to be an operation responsible for deletion of the links of a chain. (15) Chain Reduction (Nunes, 1995) Delete the minimal number of constituents of a nontrivial chain CH which suffices for CH to be mapped into a linear order in accordance with the LCA. Following this proposal, in (14), linearization prevents the phonetic realization of the two members of the chain formed by the movement of John. Thus, one of these members is 12 The definition of the LCA is given in (i). (i) Let X, Y be nonterminals and x, y terminals such X dominates x and Y dominates y. Then if X asymmetrically c-commands Y, x precedes y. (Kayne, 1994:33) 22 deleted in accordance with Chain Reduction. But still the question is why has it to be the lower one? Nunes suggests that unintepretable features that reach PF are eliminated at that level by an operation called Formal Feature (FF)-Elimination. Furthermore, the links of non-trivial chains are not identical because a feature checked in the head of the chain remains unchecked in the tail. 13 Therefore, economy considerations regarding application of FF-elimination, dictates that deletion of the tail is the optimal choice. Like Nunes, Groat and O? Neil (1996) also assume the operation Form Chain. However, they depart from Nunes in proposing that ?forming a chain results in copying all syntactic features of the category moved, but does not copy the category?s phonological matrix. It either moves it to the new position or fails to move it.? (cf. 1996: 127). Thus, Copy is instructed to copy all but the phonological features of a category. Hence, only one link of a given chain contains phonological features. It is proposed that the computational system decides which link will carry the phonological features based on the principles in (16). (16)a.Strong features may be checked only in a checking relation with a node specified for phonological features b. Moving phonological features to the head of a chain is more costly than leaving them in the tail of the chain (Groat and O? Neil, 1996:124) 13 Note that this is contrary to Chomsky?s (1995a) assumption that if a link of chain checks a feature then all links of that chain have that feature checked. 23 Accordingly, in (15) movement of John to spec of TP is to check a strong feature. Thus, in accordance to (16a), the phonological features of John are moved. In this model, all movement, including ?covert movement? occurs before spell-out. Thus, the principle in (16b) substitutes for the Procrastinate Principle proposed by Chomsky (1995). All non-trivial chains are formed before spell-out and the upper link is pronounced only if the feature being checked is a strong one. Hornstein (1998, 2001) questions the assumption that chains are syntactic objects. 14 . He observes that Nunes? proposal for copy deletion is redundant. If there is no choice in terms of which copies survives deletion, then the deletion process does not piggyback on chains, but on the features of members of the chain. Hence, beside Nunes? stipulation that deletion targets chains, chains seem to play no role on the deletion process. In addition, Horsntein points out that there is a technical argument for eliminating chains. In a strong reading of the Inclusiveness Condition (cf. Chomsky, 1995a:228), chains cannot be syntactic objects because they are neither lexical objects nor rearrangements of lexical items. In Hornstein?s (2001) proposal, it is not chains that are linearized, but the lexical array, i. e. the array of items selected from the lexicon. All the items present in the initial lexical array have to be mapped into a unique linear order at PF. Therefore, since a linear order is asymmetric and irreflexive, copies have to be reduced to one before the mapping. Another feature of Hornstein?s theory is that deletion applies deterministically. Copies with unchecked unintepretable features have to be eliminated at PF, otherwise the derivation crashes at this level. Therefore, since movement creates a configuration in 14 See also Epstein and Seely (1999) and Kiguchi (2002). 24 which unintepretable features are checked, it is arguably the case that the highest link of a chain is free of unintepretable features, whereas the lower links are not. That?s why lower copies are the targets of deletion. To see how this proposal works, consider the passive sentence in (14) again. (14) is repeated here as (17b) and (17a) is the lexical array from which the (17b) is built. (17)a.{John, was, T, seen} b. [ CP [ TP John 1 [ T? was [ VP seen John 1 ]]] If all the items present in the lexical array in (17a) are to be mapped into a linear order, in (17b) one and only one copy of John should not be deleted. The lower copy has an uninterpretable Case feature, which has been checked in the higher copy. Hence, the lower copy is the target of deletion because its presence leads the derivation to crash at PF. Consequently, only the copy of John in spec of TP is mapped into the linear order. My account for Brazilian Portuguese and Finnish null subjects does not require chains to be part of the computational system. Hence, I will follow Hornstein?s approach of copy deletion. However, it is right to mention that in principle my analysis is compatible with any of the proposals mentioned here. 25 2.4 Obligatory Control: Movement Control configurations has been the topic of many debates. For instance, Chomsky and Lasnik (1977) and Chomsky (1977, 1981, 1986a), among others, propose that non-nominal complements of control verbs are sentences (cf. (18a). Chierchia (1984) and Dowty (1985), among others, suggest a different analysis; for them the complement of a control verb is a bare VP as illustrated in (18b). (18) a.[John hopes [ S to win the game] b. [John hopes [ VP to win the game] The standard Government-Binding theory (cf. Chomsky 1981, 1986a) assumes the sentential analysis in (18a), and assumes that the subject position of the embedded clause is occupied by an empty category. The Government-Binding reasons for postulating an empty category in (18a ) is the following: (18) contains two verbs (hope and win) that assign an external q-role. Since q-roles are assigned at D-Structure, in (18) John is lexically inserted as the external argument of hope and an empty category as the external argument of win. This alleged empty category was analyzed as being PRO, a [+anaphor, +pronominal] null lexical item that does not have a governing category and receives a null case as proposed by Chomsky and Lasnik (1993). 15 Therefore, the structure underlying (18a) is the one in (19): 15 But, see Koster (1984), who argues that in Dutch PRO is governed by the matrix verb, Manzini (1983), for whom PRO lacks Case and Zwart (1988) and San Martin (2002), who suggest that PRO receives a nominative Case. 26 (19) [ John tried [ S PRO to dance]] Since PRO is anaphoric, it does not have an inherent reference. Thus, its referent is either free (arbitrary) or determined by its antecedent (e.g. by John in (19)). Inside the Government and Binding model, coreference between PRO and its antecedent result from application of the Index rule, a construal rule responsible for indexing one expression to another (cf. Chomsky 1981, 1986a). 16 In (19), for instance, PRO and John have the same referentthey share the same same index. In sum, the Government and Binding assumption that theta-positions are filled at D-structure leaded to an analysis of obligatory control according to which the subject of the infinitival embedded clause is a [+anaphoric, + pronominal] category, which is co- indexed with a NP by a construal rule of indexation. In what follows I present three minimalist approaches of obligatory control that dispense with the construal rule of indexation by suggesting that obligatory control is the result of movement. 17 16 On coreference, see also Lasnik (1976). 17 Since I am focusing on Movement, I will leave aside Landau?s (1999, 2000) proposal, according to which obligatory control is an instance of Agree (cf. Chomsky 2000). Landau observes there are two types of obligatory control: exhaustive control involving implicative, aspectual or modal verbs, and Partial control involving factive, propositional, desiderative or interrogative verbs. In cases of exhaustive obligatory control, the functional category F that Agrees with the controller (T 0 in subject control and v 0 in object control) also agree with PRO, as shown in (i). In Partial Control, the embedded T 0 is tensed and moves to C 0 , blocking, thus, Agree between a higher Functional category and PRO. (i)[? F ? DP ? [ CP [ IP PRO T-Agr [ VP T PRO ]]]] Agree Agree For a discussion of Landau?s proposal, see Hornstein (2003), see also Barrie (2004) for an analysis of partial control under movement. 27 2.4.1 PRO Climbing at LF Martin (1996) accepts the premise that the subject of an infinitival complement of a control predicate is PRO and assumes that PRO receives a null Case. He ties the latter assumption to the fact that, in control configurations, the event of the infinitival clause is disjoint temporally from the event denoted by the matrix clause. 18 Following Stowell (1982), he takes this temporal dissociation to mean that the non-finite T embedded under a control predicate is [+tense], and proposes that a [+Tense, - Finite] T checks null Case. The novelty of Martin?s proposal is the unification of obligatory control and clitic climbing. 19 PRO is taken to be an anaphoric clitic similar to the reflexive impersonal clitic SE/SI found in the Romance languages (cf. (20). 20 Both PRO and SE/SI are understood to be morphologically weak. (20)Se levanto (a si mismo) (Spanish ? Martin 1996:132) Se raised-3Sg to the same ? He raised himself? Building on Uriagereka (1995), Martin suggests that, due to its morphological weakness, SE/SI overtly adjoins to T, winding up in the same checking domain as the subject. If the feature content of SE is not rich enough to distinguish SE from the subject, 18 This fact was originally observed by Stowell (1981, 1982). 19 On clitic climbing, v. Kayne (1989), Roberts (1997), Lightfoot and Rodrigues (2003), among others. 20 Martin presents a series of arguments showing that obligatory control and clitic climbing have the same syntactic behavior. However, it is important to notice that these two phenomena do not behave alike with respect to adjuncts. While obligatory Control is not sensitive to the island-hood of adjunct clauses, clitic climbing is. 28 the chain formed by SE collapse with the subject?s chains, forming a single chain with two q-roles. 21 Thus, the LF structure of (20) is (21): (21)[ TP pro 1 [ T? [ T se 3 , [ T V 2 , T]] [ VP t 1 [ V? t 2 t 3 ]]]] In obligatory control configurations, the same happens to PRO at LF. PRO adjoins to the matrix T and since its features are impoverished, PRO?s chain fuses with the matrix subject?s chains, thus forming a super chain with multiple q-roles. 2 The structure in (22b) illustrates Martins? proposal. 23 (The dotted lines represent covert movement.) (22)a. Rom?rio tried to score a goal b. TP 3 Rom?rio 1 T? [ T PRO 3 T] VP t 1 VP tried VP 3 t 3 to score a goal 21 This author assumes (I) as a condition on chain fusion, as proposed in Uriagereka (1995): (i)Chain Fusion Situation Where a and b are different Chains, if a?s head is non-distinct from b?s head within a given checking domain (containing within the same X 0MAX ), and the tail of a-commands the tail of b, then a and b can fuse into an integrated Chain g, subsuming the properties of a and b. 2 San Martin (2002) develops a similar analysis. 23 Here I am simplifying Martin?s proposal. Following Uriagereka (1988, 1995), he assumes that SE/SI and PRO have the following internal structure: (i)[ DP [ D? [ D PRO/SE] [ NP pro]]] 29 In conclusion, Martin's proposal can be summarized in the following way: It holds the Government-Binding biuniqueness condition on the assignment of q-roles and assumes that PRO is inserted as the external argument of the embedded predicate. However, it dispenses with the construal rule of indexing by proposing that PRO adjoins to the matrix T at LF, and, as a consequence, its chain is collapsed with its antecedent?s chain, forming a single chain. 2.4.2 Feature Movement Manzini and Roussou?s (2000) account for obligatory control departs from standard assumptions on movement, for it is argued that there is no A-movement for categories. Arguments are base generated in their spell-out positions. Subjects are lexically inserted in spec of finite Ts in order to lexicalize a D-feature of T. Non-finite Ts do not have a D-feature, hence they do not trigger lexical insertion of a DP. These authors argue that this proposal provides a neat explanation for (a) lack of reconstruction in argumental sites; (b) the fact that empty categories in A-positions do not block phonological process, as such as contraction and (c) obligatory control. 24 In (23), for instance, it is explained that them cannot bind each other because reconstruction inside the infinitival clause is impossible given that each other is based generated in the spec of matrix clause. 24 Though the authors did not comment on the issue, it is worth noticing that their proposal leaves open questions about sentential idioms embedded under raising predicates (cf (i)). Manzini and Roussou?s idea that raising subjects are based generated in their surface position is incompatible with the assumption that idioms are confined to certain syntactic domains. For instance, Marantz (1997) proposes that idiomatic readings are confined to the domain of little v. (i) a. The shit seems to have hit the fan 30 (23)[ TP each other [seem to them [to work]]] In (24), taken to be a raising construction, the phonological rule of contraction can apply because have and to are structurally adjacent to each other. That is, there is no trace (copy) of I between have and to because I is based generated in its surface position. (24) I hafta leave To explain how a subject is thematically marked by a verb without being part of the VP-shell, Manzini and Roussou propose a generalized F-attraction operations. 25 From its surface position, a DP attracts all the predicates in its scope. 26 Manzini and Roussou define attaction as an operation that forms an order pair of elements. In (24), for instance, when I attracts the predicate leave, the pair (I, leave) is formed. In obligatory control configuration in (25), there is not an empty category inside the embedded clause. The DP John is lexically inserted as the spec of the matrix TP and attracts both the matrix and embedded predicate, as shown in (26). (25) John tried to leave b. All hell seems to have broken loose 25 On attraction, see Chomsky (1995a). 26 However, it is unclear which feature of a predicate is attracted. 31 (26) TP 2 John T? T VP tried TP 2 to leave In a nutshell: Manzini and Roussou offer a theory of A-movement whereby only F-movement is involved. In obligatory control configurations, a DP lexically inserted in the spec of the matrix finite TP attracts all the predicates in its scope. 2.4.3 Overt NP Movement The raising analysis of (27) has been accepted without controversies (see, for instance, the arguments given by Burzio 1981). The DP Ira receives its q-role inside the infinitival clause and then moves to the spec of the matrix TP, where it receives/ checks a Case feature. (27)Ira seems to be playing his harmonica The Government-Binding theory did not extended this analysis to cases of obligatory control because control verbs theta-mark their external arguments. Hence, in (28) the spec of the matrix VP has to be filled at D-Structure. Consequently, Ira could not 32 have been base generated inside the embedded domain and then moved to the matrix clause. (27) Ira tried to play his harmonica The main trend of the Minimalism dispensed with D-structure, but kept the ban on movement into theta-positions (cf. 2.5). Therefore, it grants the difference between (27) and (28). This minimalist view is disputed by Hornstein (1999, 2000), who argues against the ban on movement into theta-positions proposing that q-roles are features on verbs and that there is no upper bound on the number of q-roles an argument can have. Thus, for this author, obligatory control and raising analysis are alike structures in that both of them involve overt A-movement out of infinitival clauses. 27 To see how this proposal works for obligatory control, consider the derivation sketched in (29). The DP Ira started the derivation in the spec of the embedded VP, where it checks the external q-role of play. Then it moves to the spec of infinitival TP arguably to check the EPP feature of T. After that, the matrix verb tried is merged with the embedded clause and a copy of Ira is made and merged with the matrix VP, checking the q-role of tried. When the matrix TP is built, Paul moves to spec of TP, where it has its Case feature checked. (29)[ TP Ira 1 [ VP t 1 [ V? tried [ TP t 1 to [ VP t 1 [ V? play his harmonica] 27 O? Neil (1995) also defends a movement analysis. 33 Notice that, according to this proposal, obligatory control differs from raising only in that in the DP under movement visits the spec of the matrix VP in latter type of construction, not in the former. This difference is due to the fact that control verbs assign an external q-role, whereas raising verbs do not. Hornstein claims that his analysis provides a straightforward explanation for the obligatory control properties observed by Williams (1980). A controlled PRO must have an antecedent (cf. (30a) because it is the residue of movement. The antecedent is the closest c-commanding DP (cf. (30b) because movement is to the closest c-commanding position. Split antecedents are not allowed (30c) because movement cannot target more than one position at a time. PRO requires a sloppy reading (30d) because it is a trace. A de se interpretation (30e) is forced because antecedent saturates more than one argumental position forming then a semantic compound monadic predicate. In addition, when the antecedent has the format only-NP, the whole phrase, not the contained NP, is the antecedent because movement is to a c-commanding position. (30) a. It was expected PRO to shave himself b. John 1 said [Paul 2 ?s bother] 3 expects PRO *1/*2/3 to shave himself c. John 1 told Mary 2 to PRO *1+2 leave together d. John expected PRO to win and Bill does too (= Bill win) e. The unfortunate expects PRO to get the medal f. Only Churchill remembers PRO giving the BST speech 34 To account for the licensing of control inside non-finite adjunct clauses, Hornstein?s analysis relies on sidewards movement (cf. section 2.2.2). Thus, (31) is derived as shown in (32). The non-finite clause and the matrix VP are built as parallel structures. The DP John starts the derivation inside the gerundive clause, where it checks the q-role of leaving and EPP feature of the non-finite T, and undergoes sidewards movement to the spec of the matrix VP, where it checks the q-role of saw. When the matrix VP is complete, the gerundive clause adjoins to VP, forming a VP with two- segments. 28 In the final stage of the derivation, John moves to the spec of matrix TP to check its Case feature and the EPP feature of T. (31) John saw Mary before leaving the party (32)a. [before [ TP t 1 [ VP t [ V? leaving the party]]]]]b. [ VP John [ V? saw Mary]] b. [ TP John 1 [ T? [ VP [ VP t 1 [ V? saw Mary]] [before [ TP t 1 [ VP t [ V? leaving the party]]]]]]] In sum, Horntein?s proposes that obligatory controlled PRO is a residue of category movement. This analysis dispenses with the assumption that PRO receives a special Case (null Case) and also dispenses with the construal rule of indexing. PRO is just a silent copy of this antecedent. 29 28 For a discussion on the island behavior of adjuncts, see section 4.4. 29 Hornstein's movement analysis has been extended to other domains of the grammar (cf. Grohmann 2003, Hornstein 2001 and Kiguchi 2002) and explored crosslinguistically. See, for example, Aoshima (2000) and 35 As I will show in chapter 4, 3 rd Person referential null subjects in BP and Finnish display all the obligatory control properties listed in (30). I account for this fact by proposing a movement analysis along the lines suggested by Hornstein (op. cit.). That is, I suggest that these null subjects are traces of their antecedents. Since I am assuming a movement analysis of obligatory control, I am theoretically committed to sideward movement and movement into theta-positions. Therefore, for the completeness of this dissertation, the next section provides a detailed discussion on movement into theta-positions. 2.5 Movement into Theta-Positions One of the questions that have recently gained some importance in the discussion about the best theory for the language faculty is the one in (33), and at least three different answers have already been offered. (33) How do q-roles behave in syntax? Jackendoff has consistently argued that q-roles are just a convenient way to describe particular configurations at the level of conceptual structure, an autonomous level with its own set of formation rules to generate well-formed objects (cf. Jackendoff Matsuya (2000) for control in infinitival and gerundive clauses in Japanese, Pires (2000) for control and inflected infinitives in Portuguese, Ferreira (2000) and Rodrigues (2000, 2002) for an analysis of thematic null subjects in Brazilian Portuguese, Hornstein and San Martin (2000) for obviation effects in Basque, and Hornstein and Motomura (2002) for Object/Experiencer psych predicates in English and Japanese. 36 1983, 1990 and 1997). 30 Thus, I think that it is fair to say that in his view questions about the syntactic behavior of q-roles (i.e. (33)) are simply ill-formed given that he takes q- roles to be terminological conveniences. In their theory of verbal decomposition, Hale and Keyser (1993, 1998, 2002) defend q-roles as syntactic configurations. That is, they take q-roles to be part of semantic relations that are defined over structural relations, configurations formed by the lexical categories (V, N. P, A) and their projections. 31 Hence, for them, q-roles behaves syntactically insofar as configurations are defined in syntactic terms. 32 The main trend of the Principle & Parameters approach, mainly Chomsky?s proposals, does not deny the syntactic import of q-roles. However, it is claimed that q- roles have a special status inside the computational system in the sense that they are assigned in a stage of the derivation in which all the objects involved in the assignment have not yet being manipulated by transformational operations. Chomsky (1995b) mixes this view with the configurational view of Hale and Keyser, suggesting that thematic relations are configurations formed by non-transformed objects. This special syntactic status of q-roles is disputed by Boskovic (1994), who argues that moved objects, or chains, are perfectly able to saturate thematic positions. 30 Jackendoff understands the grammar as having three independent levels: Phonology, Syntax (which corresponds to Chomsky?s notion of broad syntax) and conceptual structure. These levels are connected by correspondence rules, and syntax is connected to conceptual structure via projection rules. I will not discuss Jackendoff?s proposal any further because it poses deep questions about properties of the Conceptual?Intentional (C-I) interface, which lie far beyond the scope of this thesis. 31 These structural relations are the so-called argument structure, and they are formed in the lexicon (l- syntax in Hale and Keyser?s terminology). 32 I refrain from going into details here, but for arguments against the configurational view of q-roles, see Boeckx (1998). See also Kuroda (1999) who, working on head-internal relative clauses in Japanese, claims that q-role assignment, being blind to syntactic barriers, does not require government. That is, q-roles do not need to be assigned locally . 37 Boskovic and Takahashi (1998), in their account of scrambling in Japanese, raise the possibility of treating q-roles in a simpler way: as syntactic features. As shown in the last section, Hornstein (1999, 2001) also take q-roles to be features. For them, q-roles are syntactic features in the sense that they are able to trigger movement. His argument is that the featural view makes feasible a minimalist movement analysis of obligatory control, eliminating the Government-Binding assumption that there is construal rule of indexing, and also avoiding certain complications related to the distribution of PRO, mainly the necessity of null Case. On his account for pseudogapping (34a), Lasnik (1999a) also assumes q-roles to be features. He proposes that these gaps are the result of VP deletion at PF according to the derivation sketched in (34b). The verb dated has a strong q-feature that must be checked against the external argument prior to spell-out in order to avoid PF crash. But, since the external argument is base generated in the spec of the upper V and dated does not move from inside the inner VP, the derivation reaches PF with an unchecked strong feature on dated. Thus, to salvage the derivation at PF, deletion is applied to what remained inside VP, eliminating the offensive q-feature. 3 (34) a. Mary hasn?t dated Bill, but she has Harry 3 He follows Koizumi?s (1993) split VP hypothesis. 38 b. [...[ AgrsP she 2 [ Agrs? [ TP has [ VP t 2 [ V? [ AgroP harry 1 [ Agro? [ VP dated t 1 ]]]]]]]]] [strong q-feature] Working on the lack of reconstruction in argumental positions, Lasnik (2003) shows that treating q-roles as features has the advantage of allowing elimination of argumental traces (A-traces) from the Movement theory. A-traces serve the purpose of giving instructions to the C-I interface about the thematic relations created during the derivation. Hence, if a q-role is checked by an NP, that NP carries the information related to that q-role throughout the derivation. Therefore, given that the NP itself contains the thematic information required by the interface, there is no need for leaving a trace in the thematic position, if later on the NP moves to an A-position. If there are no A-traces, the fact that there is no reconstruction in argumental sites follows without further assumptions. Taking into consideration the debate presented so far, the goal of this section is twofold. The first goal (section 2.5.1) is to offer a discussion of the theoretical motivations within Chomsky?s proposals for imposing a restriction on the type of objects that are eligible to participate in thematic relations. The purpose of this discussion is to show that this restriction hinges on assumptions and stipulations for which the conceptual motivations are not firm. The second goal (section 2.5.2) is to present some empirical evidence that for movement into thematic positions. The conclusion to be offered is that, no matter how we understand the syntactic behavior of q-roles, either as configurations or as features, the mainstream of Principle & Parameters approach needs revision with 39 respect to its prohibition against movement into theta-positions. 2.5.1 The Ban on Movement into Theta-Positions On Lectures on Government and Binding (Chomsky 1981), the Theta-Criterion is proposed as a condition of adequacy at the level of D-Structure: 34 (35) Theta-Criterion (Chomsky 1981:36) Each argument bears one and only one q-role and each q-role is assigned to one and only one argument. However, the Theta-Criterion is too strong. As noticed by Chomsky (1981, 1986a), in sentences involving secondary predication, an argument receives more than one q-role. 35 In (36), for instance, Mary receives one q-role from left and a second one from sad. (36) Mary left sad Thus, to avoid problems with constructions like (36), Chomsky proposes that the Theta- Criterion is built into the system as a property of chains: 34 It is assumed that this criterion also holds at S-Structure and LF. In (1981, chapter 6), as well as in (1982), Chomsky formulates (35) in terms of chains. 35 For other arguments against having the Theta-Criterion at the D-Structure, see Boskovic (1994) and references therein. 40 (37) Chain Condition (Chomsky 1986a: 137) If C = (a 1 ,?, a n ) is a maximal CHAIN, then a n occupies its unique q-position and a 1 its unique Case-marked position. 36 According to (37), there is nothing wrong with an argument receiving more than one q- role, as long as it does not involve movement into a theta-position. To derive the correct thematic relations of (36) without movement from one theta- position to the other, one could argue in favor of using an analysis similar to that proposed by Williams (1994), among others, in which the adjective phrase adjoins to the VP, forming a complex predicate that discharges two q-roles to the subject. A small clause account, along the lines proposed by (Chomsky 1981) and Stowell (1983) could also be defended, taking the adjective phrase to be a small clause containing an empty category PRO. Though I will defer the discussion of these two possibilities to section 2.1, 37 note that if the small clause account is adopted, (36) is no longer a reason for formulating the Theta-Criterion in terms of Chain Condition. If there is a PRO in the structure of (36), (35) is obeyed: PRO is assigned the q-role from sad and John the q-role from left. In the Minimalist Program, D-Structure is abandoned and, as a consequence, the Theta-Criterion lost its primary role, and the operation satisfy (selection of an array from lexicon and formation of a structure in accordance with the X-bar theory before Transformation) is also dispensed with. Therefore, lexical insertion/pure merge can intermingle with movement/second merge. But, despite this shift in the theoretical 36 CHAINS are defined to include chains and also expletive-associate pairs (cf. 1986a:132). 41 perspectives, the incompatibility between q-role assignment and transformed objects is nevertheless part of the theory in the minimalist framework proposed by Chomsky: theta- positions are not to be filled via movement. In Chomsky (1993) and (1995a), an LF version of the Theta-Criterion is not adopted, and the Chain Condition is not mentioned as the way to ensure the prohibition against movement into theta-positions. The suggestion there is that this prohibition is derived from economy considerations, mainly by the Last Resort principle, which is constrained by Greed, the ?self-serving? basis of movement: (38) Move raises a to a position b only if morphological properties of a itself would otherwise not be satisfied in the derivation. (Chomsky 1995a: 400,7) Putting it plainly, (38), a strong version of Greed, dictates that movement is for feature checking purposes only, and a phrase a is allowed to move to a position b in g?s checking domain only if a itself has a formal feature F that g is able to check and F wouldn?t be checked otherwise. Thus, crucially, a feature of g cannot drive movement of a to b. According to Chomsky, Greed provides a rationale for why a verbal item like HIT with the thematic structure of hit, but without a Case feature, does not exist. 38 The existence of such a verb would presuppose a derivation like the one in (39) (1995a: 401, 37 In section 2.1, I will show that neither the complex-predicate approach nor the control analysis is able to explain properties of secondary predication in double object constructions. 38 Chomsky (1981, 1986) uses the prohibition against movement into theta-positions to rule out (39). Brody (1993) argues that the ban on (39) follows from the projection principle if D-structure is defined as a level in which all and only chain root positions are present. Taking the projection principle to be satisfied in all levels, including D-Structure, it follows, then, that only chain roots can have a q-role. 42 9a). John receives the internal q-role in its base position, moves to spec of VP to pick up the external q-role, and then moves to Infl to check its Case. Since the definition of Move accords to Greed, movement of John to spec of VP does not take place because such step is not motivated by any feature of John that needs to be checked. (39) John 1 [ VP t 1 [ VP HIT t 1 ]] Lasnik (1995) argues that the system of Chomsky (1993) is redundant. Putting (39) aside, all the cases in which Greed is used to block a bad derivation can be explained independently by other principles. 39 Thus, the mechanics of the system are not in accordance with its minimalist spirit. As for Chomsky?s explanation for the ill- formedness of (39), Lasnik?s observation is that its success depends on our assumptions about the nature of q-roles. If we take them to be similar to Case feature, being formal features of DPs that can checked only once, then Greed is not necessary to rule out (39). Moreover, Lasnik points out that it is not clear that Greed can block (39), for it is assumed in Chomsky (1993) that a phrase can move cyclically, through intermediate sites as long as the final landing site is a checking site for that phrase. 40 Boskovic (1994), building on Saito and Murasugi?s (1993) condition on the length of chain links (each chain link must be at least of length 1, and a chain link from a to b is 39 See also Marantz (1995). 40 Reflexive Predicates (ia), according to Lasnik (1995), might be like HIT, having the derivation in (ib), the surface subject starts as the logical object and moves to the sentential subject position, passing trough the external argument position: (i) a. John washed/shaved/dressed (= John washed/shaved/dressed himself) b. John [ VP t? [ VP washed t]] 43 of length n, if there are n XPs that cover a but not b), argues that (39) is ruled out because the chain link between the traces inside the VP is of length zero, given that there is no maximal projection that covers one but not other. Thus given Lasnik?s and Boskovic?s counter arguments, Chomsky?s (1993) proposal is not convincing. It does not prove either that Greed is a necessary principle independently of (39) or that its use is necessary to rule out (39). In addition, considering the proposal in total, (39) is predicted to be a possible derivation, the movement into the theta-position being an intermediate step towards the feature checking position. In Categories and Transformations (Chomsky 1995b), Greed is not formulated as a principle, and a configurational view of q-roles ? la Hale and Keyser is adopted. 41 There, the argument against movement into theta positions is the following: Suppose b [a q-role assigner, CR] raises, forming the CH = (b,?, t). The trace t remains in the structural configuration that determines a q-role and can therefore function as a q-role assigner; but the chain CH is not in a configuration at all, so cannot assign a q-role. In its raised position, b can function insofar as it has internal formal features: as a Case assigner or a binder. But in a configurational theory of q-relations, it makes little sense to think of the head of a chain as assigning a q-role. With regard to receipt of q-roles, similar reasoning applies. If a raises to a q-position Th, forming the chain CH = (a, t), the argument that must bear a q-role is CH, not a. But CH is not in any configuration, and a is not an argument that can receive a q-role. Other conditions too are violated under earlier assumptions or other like them, but I will not spell out the problem further. (Chomsky, 1995b: 313) As L?pez (2001) puts it, 42 the ban on having a chain checking or assigning a q- role depends on what we take chains to be. 43 If movement creates identical copies, as 41 Notice that, by adopting Hale and Keyser?s configurational view, Chomsky departs from the GB assumption that ?if an argument is in a chain, it gets its q-role only by virtue of its membership in the chain, not by virtue of the position it occupies?. (Chomsky 1981: 338). 44 suggested by Chomsky, then the claim that the head of a chain cannot receive or assign a q-role does not follow straightforwardly. As pointed out by Hornstein (1998), there is no reason to assume that chains themselves are assigned q-roles. In virtue of its first member being assigned a q-role, a chain is interpreted as having a q-role. Therefore, if movement into a theta-position takes place, it is not the chain itself that will be in a thematic configuration, but one of its members. Had this been the only problem, we could minimize our worries by assuming with Chomsky that the Chain Condition (cf (37)) is an LF condition on the well-formedness of a structure. If the head of a chain receives a q-role, the Chain Condition rules out the structure. However, even if we do so, it seems to me that there is nothing in Chomsky 1995b, besides a stipulation on chain formation, preventing movement into theta- positions. The claim that a moved head does not assign a q-role in its derived position might be right. 4 But it fails to be a counterargument for movement into theta-positions if we consider the developments of the Minimalist Program. In an Agr-less checking theory, 42 L?pez does not defend q-roles as features. His point is that theta-domains and checking domains may be the same module, contrary to Chomsky?s (1995) assumption that they are two complementary modules. But, as L?pez himself notices, his criticism may not be valid for the phase theory, given that in this theory, an argument in situ, besides being assigned a q-role, can also check its Case Feature via Agree with a functional category. 43 Note that this argument hinges on the existence of chains as real objects formed by the computational system, but as already discussed in section (2.3) chains may be not be real at all. 4 But see Larson (1988), according to whom, ditransitive verbs like give assign the external q-role only after moving to the upper head of the VP shell. See also Saito and Hoshi (2000). On their account of the Japanese light verb construction illustrated in (i), they suggest that q-role assignment happens only at LF after kekkon ?marriage? has raised covertly to the light verb su. (i) Mary-ga John-to (kyonen) [ NP kekkon]-o sita (sita = su + ta (past)) Mary-Nom John-with last year marriage-acc did ?Mary married John last year? 45 which is adopted by the end of Chapter 4 (Chomsky 1995b) and thereafter, this claim is valid for formal features in general: 45 Heads do not move in order to check a feature of an upstairs DP. Therefore, Chomsky?s argument using head movement might, contrary to his intention, count as an argument for treating q-roles as features. In addition, notice that Chomsky (1995) (see also Boeckx and Stjepanovic, 2001) suggests that head movement might be a PF phenomenon. Thus, if he maintains this suggestion, the assumption that a moved head does not assign a q-role in its derived position shows nothing about the nature of q-roles and the (im)possibility of movement into theta-positions. Since PF does not feed LF, and heads move only at PF, a moved head cannot possibly be a q-role assigner given the semantic import of q-roles. In other words, by opening up the possibility of analyzing head-movement as occurring at PF, Chomsky himself provides an independent explanation for why moved heads does not assign q-roles. It is important to remark that I do not follow this idea in this thesis, rather I assume with Larson (1988) and Saito Hoshi (2000) (cf fn. 51) that head movement occurs at syntax. The discussion about argumental chains is more complex requires a revision of Chomsky?s proposal for chain formation and chain interpretation at LF. To do so, consider the LF representation sketched in (40), which corresponds to number (88) of chapter 4. 46 45 By the end of Chomsky (1995b), it is assumed that V is the Accusative Case carrier, and this Case is checked only after V adjunction to little v. However, in the latter developments of the theory (cf. for instance, Chomsky 2000) little v is the carrier of Accusative Case feature. 46 Chomsky?s original example (cf. 1995b: 300, 88) is given in (i). However as Howard Lasnik (personal communication) pointed out to me (i) is a misanalysis because it treats ask as an ECM verb. Thus, in (40) I replaced asked with expected, a bona fide ECM verb. (i) We are likely [t 3 to be asked [t 2 to t 1 build airplanes]]] 46 (40) we are likely [t 3 to be expected [t 2 to t 1 build airplanes]]] According to Chomsky, the successive cyclic movement of we forms the following chains. (41)CH 1 = (t 2 , t 1 ) CH 2 = (t 3 , t 1 ) CH 3 = (we, t 1 ) Since the Chain Condition is an LF condition, the only way to guarantee the convergence of (41) is via elimination of CH 1 and CH 2 for they violate the Chain Condition. To solve this problem, Chomsky proposes that traces that are not considered for interpretation can be eliminated, that is, ?marked as invisible at LF?. 47 t 1 cannot be eliminated because it is assigned a q-role, hence it is important for the interpretation of the structure. t 2 , and t 3, on the other hand, do not have an effect on the interpretation, and are, therefore marked as invisible. Consequently, chains CH 1 and CH 2 are eliminated. Hence, CH 3 is the only chain visible for interpretative purposes at LF. Though it works, there is a stipulation hidden in this mechanism. The trace in the theta-position is always the second member of each of the chains formed. If, instead of 47 Technically it is not the trace (i.e., a term) that is eliminated, but its formal features. And to use appropriate terminology, they are erased. 47 (41), we had (42), (40) would not converge because all the chains formed by we would violate the Chain Condition. 48 (42)CH 1 = (t 2 , t 1 ) CH 2 = (t 3 , t 2 ) CH 3 = (we, t 3 ) There is another way of computing chains that was not considered by Chomsky. We can stipulate that each chain contains, as one of its members, the topmost copy of the moved element, which is free from uninterpretable features. Assuming this stipulation, from (40) we form (43). Interestingly, (41) and (43) achieve the same results: Only CH 3 is interpreted at LF. CH 1 and CH 2 violate the Chain Condition, but can be eliminated because they contain LF vacuous intermediate traces. (43) CH 1 = (we, t 3 ) CH 2 = (we, t 2 ) CH 3 = (we, t 1 ) Now, lets us see how this mechanism works if movement into theta-positions takes place. For illustrative purposes, I will use Hornstein?s (1999, 2001) analysis of 48 Notice that apparently there is a redundancy in this system. It seems that if we have elimination of chains, the Chain Condition can be dispensed with given that only chains that satisfy the Chain Condition survive elimination. One possible way to solve this problem is by assuming that movement into intermediate vacuous positions does not take place. Castillo at al. (1999) and Hornstein (2001) embrace this assumption, denying the existence of EPP. 48 obligatory control. According to this analysis, (44a) has the LF in (44b), in which t 1 and t 3 are copies of John in theta-positions. (44) a.John wanted to eat a bagel b. [ TP John [ VP t 3 wanted [ TP t 2 to [ VP t 1 eat a bagel]]]] Following Chomsky, we have the following chains: (45)CH 1 = (t 2 , t 1 ) CH 2 = (t 3 , t 1 ) CH 3 = (John, t 1 ) Everything is fine with CH 3 . CH 1 violates the Chain Condition, but is eliminated. CH 2 violates the Chain Condition, but cannot be eliminated because its members occupy theta-positions, so they must be visible at LF. Hence, (44b) is either a non-convergent derivation or it converges receiving a deviant interpretation. However, we can conversely assume what was suggested in (43); and doing so from (44b) we form the chains in (46): (46) CH 1 = (John, t 3 ) CH 2 = (John, t 2 ) CH 3 = (John, t 1 ) 49 CH 3 is fine. CH 2 violates the Chain Condition, but is eliminated. Now, importantly, CH 1 is also fine. Thus, (44b) converges, with two chains: CH 1 and CH 3 , which are structurally disconnected since t 2 is invisible at LF. Moreover, although the members of these chains are copies of the same source and identical in constitution, they are distinct terms. In essence, Chomsky?s argument that chains are incompatible with assignment of q-roles is as strong as his stipulation about chain formation. Of course, his stipulation can be maintained blocking movement into theta-positions. But, it is important to keep in mind that a slightly revised version of the system allows this type of movement; and neither the original nor the revised version seems conceptually superior. Hence, the choice between them must be empirical. To close the discussion on Chomsky (1995b), let me bring forward Lasnik?s (1999b) observation. If chains are not in any configuration, and as a result are unable to participate in thematic relations, we are lead straightforwardly to the following radical conclusion: A-movement of an argument must be disallowed altogether, given that movement from a theta-position should create a chain just as much as movement to a theta-position. Thus, the chain (a 1 ?a n ), with a n in a theta-position and a 1 in a Case position, should not exist to begin with. 49 49 Watanabe (1999) and L?pez (2001) argue that there is also a tension between the configurational view of q-roles and Chomsky?s idea that little v is the locus of the external q-role and accusative Case. If the object moves overtly to spec of vP, creating a second specifier (the first being occupied by the external argument), at LF both the subject and object will be configurationally in the right position to be interpreted as the ?agent? of the event. The assumption that only one specifier is interpreted as the ?agent? does not make the problem go away because it does not tell us whichever is first merged. As pointed out by Usama Soltan (personal communication), Watanabe?s and Lopez?s argument might not be a strong one. If LF does not allow the head of a multimembered chain to be in a theta-position, as suggested by Chomsky, then a shifted object should be prevented from being interpreted as the external argument. 50 In recent developments of the Minimalist Program (cf. Chomsky 1999, 2001a, 2001b), Chomsky suggests that the ban on movement into theta-positions is actually a principle of the Grammar: (47)Pure merge in theta-positions is required of and (restricted to) arguments But here we should question the use of the term ?argument?. How do we define arguments? A DP is defined as an argument because it saturates a q-role; thus lexical items are defined as arguments or adjuncts only after being merged. As observed by Cedric Boeckx (personal communication), yesterday is an argument in (48a), but an adjunct in (48b). Therefore, (47) is problematic in the sense that it requires a lexical item to be somehow marked as an argument or an adjunct prior to merge. 50 (48) a. Yesterday was a fine day b.John arrived yesterday Chomsky (1981) defines arguments as NPs with some sort of ?referential function?, such as names, variables, anaphors and pronouns. However, this definition does not make (48) compatible with (47). Moreover, if arguments are so defined and (47) maintained, a sentence involving left-dislocated NPs as in (49a) is arguably a violation of 50 Moreover, as pointed out by Cedric Boeckx, a companion of (47) is Chomsky?s assumption that adjuncts can be inserted later in the derivation. However, recently (cf. 2001b), Chomsky has dropped this assumption. Thus, once the companion of (47) is gone, (47) is even less justified. 51 (47). 51 One could follow Williams (1994) and say that in (49a) John is in a thematic relationship with the proposition Mary likes him, John being the topic of the comment. However, if this is so and topicalization is achieved via movement, (49b) should be prohibited since it would involve movement/second merge of an argument into a theta- position. (49)a.John, Mary likes him b. Fish, I don?t like Therefore, I close this section concluding that there is no strong conceptual argument supporting the idea that q-roles have a special syntactic status, being assigned only in a stage of the derivation in which the objects involved in the assignment are not yet complex objects expanded by movement. In the next section, I will present some empirical evidence to support the opposite conclusion: certain thematic relations are in fact formed only after NP movement. 2.5.2Moving into Theta-Positions In section 2.5.2.1, I attempt to show that Chomsky?s (1986) motive (i.e.; secondary predication) to formulate the Theta-Criterion in terms of the Chain Condition 51 Ross (1967) differentiates left dislocation (47a) from topicalization (47b). Chomsky (1977) proposes that left-dislocated NPs are base generated in their surface position, whereas topicalization involves movement. However, as Masaya Yoshida (personal communication) observed, (49a) could in principle be treated as a case of movement, leaving behind a resumptive pronoun `a la Boeckx (2001). Anyway, the right analysis of left-dislocation is not important for the present discussion. On the distinction between left dislocation and topicalization, see also Lasnik and Uriagereka (1988). 52 (cf. (37)) might, conversely, be an argument against the Chain Condition. I will focus on the thematic relationship between secondary predicates and indirect objects in an attempt to show that it may not be explained if a moved argument is not allowed to receive a q- role. In section 2.2, I will review Boskovic?s (1994) argument for assuming that Romance restructuring configurations involve movement into theta-positions, and in section 2.3, I will discuss Pesetsky?s (1992) observation that some ECM subjects receive a second q-role from the matrix verb. 2.5.2.1Secondary Predicates and Indirect Objects Recall from section 1 that, for Chomsky (1986), a single NP is interpreted as receiving two q-roles in cases of secondary predication. In (50), for instance, John is predicated by both the adjective sad and the verb arrived. (50) John arrived sad. Chomsky?s response to this type of construction was the formulation of the Theta- Criterion (35) as part of the Chain Condition (37). Thus, following the restriction of the Chain Condition, the analysis of (50) must guarantee that the two q-roles are assigned to John prior to John?s movement to the sentential subject position. That is, the two q-roles must be assigned when John is still in the object position. One problem for this analysis 53 is that it predicts that (51) is acceptable, though it is not. 52 (51)* There arrived a man sad Notice that in general depictive adjectives can predicate with subjects and direct objects, as illustrated in (52). (52)a. I 1 wrote my confession drunk 1 b. I put the food 1 on the table hot 1 This might suggest there is something special about presentational there constructions that make them incompatible with secondary predication. 53 Thus, being cautious, I will 52 Brazilian Portuguese displays the same phenomenon. (ia) allows only a reading in which falastrona ?talkative? is a simple adjective, describing a permanent property of the entity Maria. In (ib), on the other hand, falastrona is interpreted only as depictive predicate, denoting the state in which Maria was when she arrived. Hence, if (ia) has a structure in which the logical object stays in situ, while a expletive pronoun is inserted in the subject position, as suggested by Rizzi (1982), among others, then the conclusion is that depictive adjectives can only predicate with the logical object if it moves to the sentential subject position. Therefore, in (ib) the predicative relationship between Maria em falastrona cannot be established prior to the movement of the DP a Maria to the sentential subject position. (i) a. Chegou a Maria falastrona arrived-3Sg the Maria talkative ?There arrived the talkative Maria? b. A Maria chegou falastrona the Maria arrived-3Sg talkative ?Maria arrived talkative? Max Guimar?es (personal communication) observed that in special contexts (for instance, iteration and special prosody in (ii)) inverted subjects tend to accept secondary predication. (ii) Chegou a Maria falastrona, falastrona arrived-3Sg the Maria talkative, talkative ?There arrived Mary very talkative? 53 As Hornstein pointed out to me, the unacceptability of (51) might be due to a definite effect created by sad. S, if replaced sad is replaced by singing, for instance, the sentences is better. 54 put (51) aside. But this is not the end of the discussion yet. In Some languages, including English, 54 secondary predicates fail to predicate with indirect objects, as observed by Baker (1997) and Romero (1997): 5 (53)a.* I gave the meat to Mary 1 hungry 1 b. * I gave Mary 1 the meat hungry 1 54 See Pylkk?nen (2002) for a crosslinguistic survey. 5 Hale and Keyser (2002) discuss (i), where the indirect object does not accept a secondary predicate. (i) I gave the bottle 1 to the baby 2 full 1/*2 Though unclear, their explanation seems to be that the secondary predicate is inserted at D-structure as an adjunct to the inner verbal projection that composes the argument structure of give, as represented in (ii), the representation of (i) at D-structure (cf p. 162, 11). They take D-structure to be the level in which predicative relations are formed. Thus, in (ii), since DP 2 does not c-command the adjective (cf. Williams? (1980:206) C-Command Condition on Predication), there can be no predicative relation between the two of them: (ii) V 2 V 1 V DP V 2 V 2 V DP 1 V [bottle] V AP 2 [full] V 3 DP 2 [baby] What is missing from their account is an explanation for why the adjective is allowed within the argument structure of the verb, even though it is not selected by the verb. Moreover, we also need an explanation for why this type of adjunct is allowed within a verbal argument structure, whereas other adjuncts are not, given Fodor?s (1970) objections against lexical decomposition. 55 Interestingly though, as pointed out in Koizumi (1994), a logical indirect object can be the understood argument of a depictive adjective if passivization takes place and the indirect object is displaced to the sentential subject position, as in (54): (54)The patients 1 were given the drugs drunk 1 (cf. * The drugs were given to the patients 1 drunk 1 ) The underlying indirect object of (54) must first be merged in the complement domain of given in order to be interpreted as the goal of the giving event. But, from its VP internal position, the patients cannot be interpreted as being in a theta-relation with drunk given (53). Hence, the theta-relation between the NP the patient and drunk is licensed only after movement of the NP into the sentential subject position. This clearly shows that the head of a chain can be interpreted as being in a thematic relation, contrary to Chomsky?s restriction. One could disagree with the conclusion above, arguing that the phenomenon in question might be explained by appealing to the control analysis sketched in (55) and defended by Chomsky (1981) and Stowell (1983). According to this analysis, the conclusion that the NP movement in (54) licenses the thematic relationship between the underlying indirect object and drunk is just a misinterpretation of the datum. What the movement really does is put the indirect object in a position from which it c-commands and, therefore, controls PRO, the argument of the adjunct small clause. 56 (55) [ IP the patients 1 were [ VP [ VP given t 1 the drugs [ SC PRO 1 drunk]]]] However, there is evidence that something different from control is responsible for secondary predication. An indirect object within a VP can be a controller (56) (cf. Koizumi, 1994) and this clearly contrasts with (53). 56 (56) I wrote him 1 a letter [PRO 1 to show his mother] Taking into consideration languages like Russian, where secondary predicates are morphologically case marked, Schein (1995) provides another type of evidence against a control analysis for secondary predication. The Russian facts are the following: If a depictive adjective predicates with the subject, it is marked with nominative case, as (57a) shows, but if it is a predicate of the object, then it is marked with instrumental case (57b). If the sentential subject is an underlying object, as in unaccusative constructions (57c), the secondary predicate can be marked either with nominative or instrumental case. (57)a. Marik ubil los&ad? pjannyj (Nom) Marik killed horse drunk b. Ja kupil mjaso zamoro-enym (Instr) I bought meat frozen 56 Besides that, observe the contrast in (i): depictive adjectives cannot be predicates of prepositional objects (Williams 1980), though prepositional objects can be controllers (Schein, 1995). (i) a. *John ate at the meat 1 raw 1 57 c. Mas&a pris&la veseloj (Nom)/ veselaja (Intr) Mas&a arrived cheerful Schein?s observation is that (57) cannot be a case of control because in a bona fide control configuration (58), a secondary predicate cannot bear nominative case. According to his analysis, the presence of PRO in the embedded subject position of (58) blocks agreement between the nominative matrix subject and the adjective. Therefore, if there were a small clause containing a PRO in (57), nominative case agreement would be blocked. (58) Vanja xoc&et byt?vernym (Instr)/*vernyj (Nom) partii ?Vanya wants to be faithful to the party? Therefore, based on these facts, we can exclude the possibility of having a control analysis for secondary predication. A complex predicate analysis has been supported by Roberts (1986), Williams (1994) and Geuder (2000), among others. Putting aside differences in technical implementation, the proposal made by these authors is that depictive adjectives are predicates that combine with another predicate, forming a complex that predicated with an argument under the right c-command condition: the argument c-commanding the complex predicate. This might be the right analysis for secondary predication, however it does not make the secondary predication in (54) compatible with Chain Condition. Given the facts b. John pleaded with Bill 1 [PRO 1 to leave] 58 in (53), if a complex predicate is formed in (54), it does not predicate with the logical indirect object prior to its raising to the sentential subject position. Therefore, (54) should be seen as a counter argument for Chain Condition. The head of chain is able to participate in theta-relations. 2.5.2.2Prepositional Subjects in Restructuring Configurations Consider the following pair of Spanish sentences: (59) a.Marta le quiere gustar a Juan Marta clitic wants please-inf to Juan ?Marta wants for Juan to like her? b.A Juan le quiere gustar Marta to Juan clitic wants please-inf Marta ?Juan wants to please Marta? According to Gonz?lez (1988), (59a) and (59b) differ in meaning. In (59b), Marta is receiving the external q-role of querer 'want', whereas in (59b) this very same q-role is assigned to Juan. This contrast in meaning does not show up if the matrix verb is a raising verb, as illustrated in (60). This is explained by the fact that querer assigns an external theta-role, while the raising verb in (60) does not. 59 (60) a.Las estudiantes le empezaron a gustar al professor the students-fem clitic began to like to-the professor b.Al professor le empezaron a gustar las estudiantes to-the professor clitic began to like the students-fem ?The professors began to like the students? Boskovic (1994) argues that (59b) cannot be derived via control. The preposition a preceding Juan indicates that at some stage of the derivation, the DP Juan is an argument of the embedded verb gustar, being marked with inherent case, as in (59a). Hence, if (59b) were an obligatory control configuration, it would involve the derivation represented in (61), in which both PRO and Juan (t 1 ) are taken by the embedded verb as its external argument. (61) A Juan 1 le quiere [PRO 1 gustar Marta t 1 ] Hence, as suggested by Boskovic, (59b) seems to be a strong argument for assuming that an NP can receive a q-role in its derived position. 57 57 L?pez (2001) presents a similar argument, claiming that in some Spanish causative constructions a DP receives a q-role after undergoing movement. 60 2.5.2.3Exceptional Theta-Marking Originally discussed by Postal (1994), the sentence in (62) is taken by Pesetsky (1992) (also by Boscovic 1997 and L?pez 2001) as a case in which an ECM subject receives a second q-role from an agentive matrix verb. (62) Sue estimated Bill?s weight to be 150 lbs The argument is the following: verbs like estimated select, as their complements, NPs denoting measurement. In (63a), for instance, the NP Bill?s weight matches this selectional requirement and the sentence is acceptable. (63b), on the other hand, is unacceptable because the NP Bill does not denote measurement. (63)a. Sue estimated Bill?s weight b. * Sue estimated Bill Thus, if in (62) we substitute Bill?s weight for Bill, it results in an unacceptable sentence. (64) * Sue estimated Bill to weigh 150 lbs It seems then that the main verb in (62) selects as its complement the NP Bill?s weight. Hence, considering selection restrictions to be an indication of q-role assignment, it is reasonable to assume that in (62) Bill?s weight is assigned a q-role from estimated. 61 And, taking thematic relations to be local, as defended by Hale and Keyser, the q-role in question cannot be assigned before the embedded subject moves from its based generated position, otherwise locality wouldn't be satisfied. Therefore, it is plausible to assume that the LF representation of (62) has a nontrivial chain bearing more than one q-role. 58 For Pesetsky, this case might be analyzed as involving a q-role assignment across a clause boundary, similarly to the GB account for exceptional Case marking, as shown in (65a). Boscovic suggests that the embedded subject raises to the matrix Agro projection in order to check its Case feature, and its second q-role is assigned when the matrix verb adjoins to the head of AgroP, as represented in (65b): (65) a.[?[ VP estimated [ IP Bill?s weight 1 [ T? [ VP t 1 [to be 150 lbs]]]]]] b.[?[ AgroP Bill?s weight 1 [ Agro? [ Agr estimated+Agr][ VP t 2 [ IP t 1? ]]]]] As Norbert Hornstein pointed out to me, (62) can also be analyzed as a case of backwards control (cf. Farrell 1995, Polinsky & Potsdam 2002 and Hornstein 2002), where the controlled gap c-commands its antecedent. Following, an analysis of Control 58 This type of exceptional theta-marking might also be involved in ECM constructions like (i) in which the verb declared is interpreted as affecting the embedded subject in the sense that Bill?s status is changed by Mary?s declaration (Cf. Pesetsky, 1992) (i) Mary declared Bill to be dead. 62 where PRO is lexically inserted in the gap position, one could suggest that the structure of (62) is (66): (66) [?[ VP estimated PRO 1 [ IP [Bill?s weight] 1 [ I? [ VP t 1 [to be 150 lbs]]]]]] However, the analysis above is not free of problems. For instance, it violates principle C of Binding Theory. On the other hand, if Polinsky & Potsdam (2002) and Hornstein (2002) are right, backwards control is also the result of movement. Particularly, under this account, in (62) the embedded subject moves to the matrix object position, receiving, thus, its second q-role, which is assigned by estimated. (67) [?[ VP estimated [Bill?s weight] 1 [ IP t 1 [I? I [ VP t 1 [to be 150 lbs]]]] 2.5.2.4q-Roles in a Nutshell Upon closer inspection, the ban on movement into thetapositions reveals itself as conceptually unjustified. The developments of the Principle and Parameter approach do not provide any well-motivated argument against having q-roles affecting complex phrasal markers. Therefore, the right answer about the syntactic behavior of q-roles is to be decided on empirical grounds; and, as presented here, there is some empirical evidence that q-roles can be saturated by derived objects. 63 The theoretical implications of this conclusion is the following: if q-roles are understood as configurations, we need to redefine the operation Move such that it can be applied for purposes other than that of feature checking. On the other hand, if we hold our current understanding of Move, then q-roles are to be defined as features, as defended by the authors cited in the introduction. Therefore, mainstream Principle and Parameters assumptions about how theta-relations are formed need to be revised. Having said that, an important addendum to this discussion is the following: in the present work, I will adopt movement into theta-positions, and, since I am assuming that Movement is for feature checking purposes, I will implement movement into theta- positions by taking q-roles to be features. 2.6Conclusions In order to provide an answer to question (1), I will assume the minimalist ideas in (66): (1) What is the syntactic nature of referential null subjects in Brazilian Portuguese and Finnish? (66) a.The checking theory b.The Copy & Merge theory of movement c.Sideward movement 64 d.Deterministic deletion of copies e.q-roles are features f.Obligatory control is a residue of movement. 65 CHAPTER 3 VERBAL MORPHOLOGY AND NULL SUBEJCTS 3.1 Preliminaries This chapter is a transition piece. Its main purpose is to set up the empirical background necessary for the discussion of the nature of referential 3 rd Person null subjects in Colloquial Brazilian Portuguese (BP, henceforth) and Finnish. 1 Section 2 offers an overview of the verbal agreement morphology properties of BP, Standard Finnish (SF) and Colloquial Finnish (CF). Section 3 is an overview of the type of null subjects licensed in these grammars. Concentrating on referential null subjects, in section 4, it is suggested that 1 st P and 2 nd P null subjects in matrix clauses should not be taken as evidence that Modern BP and Finnish license referential null subjects. These null subjects are not genuine null subjects, being rather the result of topic deletion ? la Huang (1984). Section 5 paves the way for a new analysis of embedded 3 rd P referential null subjects in the grammars under consideration: the possibility of this category being treated as a logophoric anaphor will be dismissed and the previous accounts reviewed. 1 I will use the term Finnish to refer to both colloquial and standard Finnish. And to avoid an overflow of data, I will distinguish colloquial Finnish and standard Finnish only when necessary. 66 3.2Verbal Agreement Morphology Duarte (1995, 1996) shows that from 1845 to 1992 Brazilian Portuguese underwent a simplification in its verbal inflection. In figure 1, a paradigm with six forms (paradigm 1) was reduced to a paradigm with four forms (paradigm 2), and then to a paradigm with only three forms (paradigm 3). From paradigm 1 to paradigm 2, the indirect 2 nd Sg replaced the direct 2 nd Sg agreement, which is phonologically identical to 3 rd Sg. From paradigm 2 to paradigm 3, the 1 st Pl was replaced by 3 rd Sg, since the pronoun a gente (?we, literally, the people?), which triggers 3 rd Sg agreement, is used instead of the 1 st Pl pronoun n?s ?we?. 2 Fig. 1. Falar Finland?s ?to speak Finnish? Person and NumberParadigm 1Paradigm 2Paradigm 3 1 st Psg Fal-o Finland?sFal-o Finland?sFal-o Finland?s 2 nd PSg directFala-s------------------ 2 nd PSg indirectFala-0Falo-0Fala-0 3 rd PSg Fala-0Fala-0Fala-0 1 st PPl Fala-mosFala-mosFala-0 2 nd PPl direct Fala-is-------------------- 2 nd PPl indirectFala-mFala-mFala-m 3 rd PPl Fala-mFala-mFala-m Source: Duarte (1996) The verbal agreement morphology of Standard Finnish seems to be quite rich when compared to that of BP. For every combination of person and number, there is a different ending on the verb, as figure 2 illustrates: 2 Standard BP still has a parallel agreement system with the 1 st Pl. It is also worth mentioning that in some dialects, the direct 2 nd Sg pronoun is still in use. 67 Fig. 2. Puhu Portugalia ?to speak Portuguese? Person and NumberStandard Finnish 1 st Psg Puhu- n Portugalia 2 nd PSg Puhu-t 3 rd PSg Puhu-u 1 st PPl Puhu-mme 2 nd PPl Puhu-tte 3 rd PPl Puhu-vat Note that the richness of this paradigm becomes questionable if we consider the 3 rd PSg in detail. As apparent on figure 2, for 1stPSg/Pl, 2ndPSg/Pl and 3rdPPl the verbal inflection contains a consonant, but for the 3rdPSg, the verbal inflection consists of lengthening of the last vowel of the stem. This lengthening process occurs in the present tense only. In the past and conditional tense (cf. (1)), the process is neutralized and there is no overt agreement suffix for the 3rdPSg. This gap in verbal paradigm is reported by Vainikka (1989) and Holmberg and Nikkane (1993). 3 (1) a.H?n puhui 4 s/he spoke-3Sg Portuguese ?S/he spoke Portuguese? b. H?n puhuisi Portugalia s/he speak-3Sg-Cond Portuguese ?S/he would speak Portuguese? 3 In Finnish, there is no verbal suffix for the future tense. 4 This is SF. In CF, the past tense is puhu since the past tense marker i was lost. 68 Vainikka and Levi (1999), citing work done by Hakulinen (1979), present further evidence that in Finnish the 3 rd P verbal agreement morpheme is weak. According to them, all the agreement suffixes are phonologically related to the corresponding pronoun, except for the 3 rd P suffixes (cf. figure 3). The 2 nd PSg agreement suffix t is diachronically related to the pronoun sin?, since this pronoun is reconstructed as *tin?. 5 But between the 3 rd Sg/Pl agreement suffixes and the corresponding pronouns there is not such historical connection. Fig. 3. 1 st Sg2 nd Sg 3 rd Sg1 st Pl 2 nd Pl3 rd Pl Agreement Suffix Pronoun -n min? -t sin? -V h?n -mme me -tte te -vat he Source: Vainikka and Levy (1999) This leads us to consider that, at least in the past and conditional tense, the verbal inflection paradigm of SF is not uniform under the Jaeggli and Safir (1989) definition of a morphological uniform verbal inflection paradigm. It is a mixed paradigm, containing an underived inflectional form, mainly the form 3rdPSg. 6 Therefore, given Jaeggli and Safir?s theory of null subjects, SF is predicted not to be a null subject language. The null subject properties of this variety of Finnish will be discussed in the next section. 5 Synchronically and diachronically Finnish follows a general /ti/ ? /si/ rule. 6 Jaeggli and Safir (p. 30) states morphological uniformity in the following way: (i) Morphological Uniformity An inflectional paradigm P in a language L is morphologically uniform iff P has either only underived inflectional forms or only derived inflectional forms. 69 Similarly to BP, CF has a simplified verbal inflection paradigm. The 3rdPSg replaced the 3rdPPl and the 1 st PPl was substituted by a new form that is phonologically identical to the impersonal passive morpheme (2) 7 Fig. 4. Puhu Portugalia ?to speak Portuguese? Person and NumberColloquial Finnnish 1 st Psg Puhu- n Portugalia 2 nd PSg Puhu-t 3 rd PSg Puhu-u 1 st PPl Puhu-taan 2 nd PPl Puhu-tte 3 rd PPl Puhu-u (2)Koulussa opetetaan lapsia school-Ine teach-Pass children-Par ?Children are taught at school? Thus, the grammars under investigation share the property of having a weak verbal morphology for the 3 rd P. If this is so, given the fruitful literature on the correlation between rich morphology and licensing of null subjects (cf. Taraldsen 1980, Chomsky, 1981, 1982, Rizzi, 1982, Borer 1986, 1989, Jaeggli and Safir 1989, among others), the content of the next section is predictable. 7 The weak verbal morphology in CF seems to be correlated with a reduction in the morphological form of the pronouns, see (i) which should be compared with figure 3: (i) M?/s?/se/me/te/ne(se also means ?it? and ne also means ?those?) I / you/(s)he/we/you/they 70 3.3 Null Subjects This section starts with a brief description of overt/null expletives and generic null subjects in the languages under investigation. However, there will be no in depth discussion of these issues. Rather, I concentrate on referential null subjects. Section 3.3.2 is dedicated to referential null subjects, especially to 1 st P and 2 nd P null subjects in matrix clauses. After that, the scope of the investigation will be reduced, and I take up the question of embedded 3 rd P null subjects. 3.3.1Expletives and Generic Null Subjects The data in (3a,b) is evidence that BP allows null expletives. 8 (3c) shows that that this language also allows generic null subjects. (3)a. e tem muita coisa aqui nessa loja have-3Sg many things here in.this store ?There are many things in this store? b. e j? est? tarde demais already be-3Sg late too.much ?It is already too late? c.Aqui conserta sapato (Kato, 1999) here repair-3Sg shoe ?One repair shoes here? 8 For a discussion on null expletive in BP, see Viotti (1999). 71 It is unclear, however, that the missing external argument of sentences like (3c) is syntactic realized. If this argument were present in the structure, we would expect it to be able to bind anaphors and control inside infinitival clauses. Allowance of secondary predicates oriented to the subject should also be possible. But, as shown in (4), the alleged external argument of generic sentences like (4c) does not display any of the properties mentioned: 9 (4)a. *Aqui n?o vende nada de si mesmo here not sell-3Sg nothing of oneself ?Here one sells nothing from oneself? b.?? Aqui vende sapato caro [para PRO evitar cliente pobre] here sell-3Sg shoe expensive to avoid-Inf customer poor ?Here one sells expensive shoes in order to avoid having poor customers? c. * Aqui conserta sapato b?bado/irritado here repair-3Sg shoe drunk/irritated ?Here one repair shoes drunk/irritated? Interestingly, though, in the presence of a modal verb, control is possible and secondary predicates oriented to the subject are allowed: 9 On this matter, see Cavalcante (2003) 72 (5)a. Nesse hotel pode entrar na piscina [sem PRO in.this hotel can-3Sg enter-Inf in.the swimming pool without tirar a roupa] take.off-Inf the clothes ?In this hotel one can enter into the swimming pool without taking his or her clothes off? b.Nesse hotel n?o pode entrar na piscina in.this hotel not can-3Sg enter-Inf in.the swimming pool b?bado drunk ?n this hotel one can enter into the swimming pool drunk? This can be explained, if modality involves embedding clauses, the modal verb selecting an infinitival clause, as in the representation sketched in (6), where I am assuming that the subject of the root clause is a null expletive. (6)[Adv Loc [ Infl Pro expl [ VP Verb Modal [ Infl PRO 1 [ VP t 1 V-inf.]]]]] Guimar?es and Rodrigues (2002) suggest that the locative adverb function as the sentential subject of these generic sentences. This would explain the obligatory presence of this adverb and also the fact that it has to precede the verb. 10 I will not develop this 10 According to Postal (2002), fronted locative adverbs present some of canonical the subject properties. For instance, they cannot link to a standard floating quantifier like both or to a non-standard floating quantifier like each. See Soltan (2003) for a review of Postal?s arguments. 73 idea, though in chapter 4 I will suggest that in BP fronted locative adverbs do block sub- extraction from embedded clauses. Finnish is a SVO language that avoids verb initial in declarative sentences (cf. Vilkuna 1989 and Vainikka 1989). Thus, in SF, existential and generic constructions are acceptable as long as the locative adverb or the indefinite NP is fronted: 11,12 (7)a.* leikkii lapsia kadula (Holmberg and Nikkane, 2002) play-3Sg children in.street b. Kadulla leikkii lapsia in.street play-3Sg children c.Lapsia leikkii kadula children play-3Sg in.street ?There are children playing in the street/Children are playing in the street? (8)a.*?Voi anoa lainaa pankista (Vainikka, 1989) can-3Sg apply loan-Par bank-Ela b. Pankista voi anoa lainaa) bank-Ela can-3Sg apply loan-Par 11 Holmberg (op. cit.) points out that only referential phrases can be fronted. Thus, locative, instrumental and temporal adverbs can be fronted, whereas manner and reason adverbs cannot. 12 For a discussion on generic null subjects in Finnish, see Hakulinen and Karttunen (1973). Given the observation in BP, it is worth mentioning that according to these authors these verbs that are best suited for this generic sentences are modals or verbs that imply modality. 74 c. Lainaa voi anoa pankista) loan-Par can-3Sg apply bank-Ela ? One can apply for a loan from the bank? Verb initial is allowed if there is no category that could be fronted, as shown in (9): 13 (9)a. Sataa (Nikanne and Holmberg, 2002) rain-3Sg ?It rains/it is raining? b. N?ytt?? tuvevan sade (Sulkala and Karjalainen, 1992) look-3Sg come-lpart-acc rain-ACC ?It looks as if it is going to rain? CF differs from SF and BP in that it has overt expletives (Holmberg and Nikanne 1994, 2002 and Holmberg 2003). Therefore, in order to avoid verb initial declarative sentences, either a constituent is fronted (7) or an expletive is inserted (10). If the 13 Verb initial is also allowed if the verb is in the imperative mood (ia), in yes/no questions (ib) and also if the verb is focalized (ic). These cases are analyzed as involving V-to-C movement (cf. Vainikka 1989, Vilkuna 1989, Nikanne 1994, among others). (i)a. k?velk?? taloon(Nikanne, 1994) walk-2Pl house-Ill ?Walk into the house!? b. K?velik? ihmisi? taloon(Q-Cl stands for ?question clitic?) walked-Q-Cl people-Par house-Ill ?Were there people walking into the house?? (Foc-Cl stands for ?Focus clitic) c. K?veli(+p?s) ihmisi? taloon walked-Foc-Cl people-Par house-Ill ?There WERE people walking into the house? 75 sentence contains no constituent that could be fronted, insertion of the expletive is optional, as exemplified in (11): 14 (10)Sit? leikkii lapsia kadulla (Homlberg, 2003) expl played children in-street ?There were children playing the street? (11)a.(Se) sataa expl rain-3Sg ?It rains/it is raining? b.(Se) n?ytt?? tuvevan sade expl look-3Sg come-lpart-acc rain-ACC ?It looks as if it is going to rain? The position occupied by the fronted categories in (7) is taken to be the sentential subject position by Vainikka (1989), Vilkuna (1989), Holmberg and Nikanne (2002) and Holmberg (2003). Holmberg and Nikanne assume that the movement is driven by an EPP. For them, Finnish subjects move to spec of FP (FiniteP), a projection that immediately dominates TP. 15 Thus, in (7), the EPP feature of FP is checked by the 14 Sit? is the partitive form of se. Though Holmberg and Nikanne (1994) say that constructions with this expletive are extremely common, I acknowledge that some speakers do not accept them, characterizing them as substandard. 15 Vainikka and Vilkuna argue that this movement is to spec of IP. 76 fronted category, whereas in (10) it is checked by the overt expletives. 16 Notice, however, that none of these analyses explain the lack of agreement between the verb and the fronted indefinite NP lapsia ?children? in (7c). 3.3.2 Referential Null Subjects Present-day data from BP and Finnish clearly show that these grammars do not behave like the Romance Pro-drop languages. 17 For instance, while the Romance Pro- Drop languages obey the Avoid Pronoun Principle (12), accepting overt subject pronouns only when they are emphatic, BP and Finnish allow overt subject pronouns that do not carry any emphatic force. 16 Holmberg and Nikanne (1994) observe that CF has transitive expletive constructions (i). In this paper, instead of postulating an F Projection, the authors assumed that TP is dominated by an AgrSP whose specifier is occupied by the expletive in (i), whereas the DP subject stays in spec of TP. This analysis is built around central ideas developed by Bobalijk and Jonas (1996) for Icelandic transitive expletive constructions. (i) sit? ovat n?m? lapset jo oppineet lukemaan expl have-3Pls those children already learnt read-Inf ?These children have already learned to read? 17 In this thesis, I do not discuss all the properties related to the pro-drop parameter (cf. Chapter 1), rather I will focus only on referential null subjects, but see Kato and Tarallo (1988), and Kato (2000) on loss of free inversion in BP. For a discussion of violation of the [*that-t] filter in Portuguese, see Zubizarreta (1982). Finnish does not have free subject inversion. It has only triggered inversion (using Shlonsky?s (1997) terminology) which occurs when the verb moves to C, v. fn 9. With respect to the [that-t] filter, in Finnish the complementizer ett? ?that? is optionally dropped when wh-movement occurs,as shown in (ii) (ii)a. Kuka 1 Jukka sanoi ett? t 1 oli myynyt auton who-Nom Jukka-Nom said-3Sg that was-3Sg sold car b. Kuka 1 Jukka sanoi t 1 oli myynyt auton who-Nom Jukka-Nom said-3Sg was-3Sg sold car ?Who did Jukka say had sold the car? 77 (12)Avoid pronoun Empty categories have preference over overt pronoun (cf. Chomsky, 1981:65) In fact, Modern BP and CF seem to prefer overt subject pronouns rather than null subjects. 18 Duarte (1995, 1996) presents historical facts from BP indicating that in this 18 In BP there are particular cases in which this preference either does not arise or is blocked. (a) In object position, null pronouns are allowed in a broad range of configurations. See Cyrino (1996) for BP. (b) accordingly, Montalbetti?s (1984) Overt Pronoun Constraint, configurations involving co-indexing with an A-bar antecedent block overt pronouns. In (ia,b), for instance, the overt pronoun ele ?he? must be free in reference. For a discussion of Montalbetti?s constraint in BP, see Negr?o (1999): It is worth mentioning, however, that in some dialects a quantifier expressions with the format todo-NP ?every NP? can bind an overt antecedent, as in (ic) for example. (i)a. Ningu?m 1 acha que ele *1/2 ? um g?nio nobody think-3Sgs that he is-3Sg a genius ?Nobody thinks that he is a genius? b. Quem 1 acha que ele *1/2 ? um g?nio who think-3Sg that he is-3Sg a genius ?Who thinks that he is genius? c. Toda crian?a acha que ela ? um g?nio every child think 3Sg that she is a genius ?Every child thinks that she is a genius? (c) Co-indexing with an overt pronoun is also blocked in coordinated structures as (ii), though these structures arguably involve an A-antecedent (the matrix subject). (ii)O Jo?o 1 cantou e ele *1/2 dan?ou the Jo?o sang-3Sg and he danced-3Sg ?John sang and he danced? By analyzing coordinate clauses as a coordination of VPs (Johnson 1994 and Zoerner and Agbayani 2001), we can take the obligatory disjoint reference in (ii) as a Principle B effect: The pronoun cannot be co- indexed with the DP o Jo?o because they are clause mates. Notice that co-indexation is possible in (iii), in which the second conjunct is a clause as the presence of the complementizer indicates. (iii)Eu acho que o Jo?o vai voltar para l? e que ele vai ser feliz ?I think that Jo?o will come back to Brazil and that he will be happy there? If (ii) is a case of VP coordination, we can also explain the temporal dependency between the time of each of the two conjoined structures. As shown in (iv), unless adverbs are inserted to delimit the time scope of each coordinate structure, time independence is impossible: 78 language the use of referential null subjects has decreased considerably. As represented in figure 5, there was a certain stability in the preference for null subjects until 1918, but around 1937 this preference started falling, reaching less than 30% of the occurrence in 1992. % 0 10 20 30 40 50 60 70 80 1845 1882 1918 1937 1955 1975 1992 Fig. 5 (adapted from Duarte, 1996) Interestingly, it seems that BP loss of null subjects is correlated with its loss of verbal morphology. Duarte?s work shows that the stable period in Figure 5 corresponds to a period in which the first paradigm of figure 1 was in use. Hence, while the verbal inflection paradigm was rich, the null subject properties of BP were intact. When the (iv) O Jo?o cantou *(ontem) e vai dan?ar *(hoje) the Jo?o sang-3Sg yesterday and wil-3Sg dance-Inf today ?Jo?o sang yesterday and will dance tomorrow? 79 verbal inflection became impoverished (paradigm 2 of figure 1), referential null subjects became less common, yielding a preference for overt subject pronouns. Hoinenen (1995), studying data from CF, shows that in this variety of Finnish the 1 st PSg subject pronoun min?/m? ?I? is omitted only 12% of the time. She recorded 197 cases of 1 st PSg subjects. Out of these 197 cases, only 23 were cases of subject omission, 5 of them were replies to answers. Also related to the inapplicability of (13) is the fact that, contrary to what happens in the Romance pro-drop languages, in the Grammars under discussion there is no preference for a non-coreferential reading of the embedded subject pronoun. 19 (13)a.O Jo?o 1 disse que ele 1,2 gosta da Maria (BP) the Jo?o said-3Sg that he like-3Sg of.the Maria ?Jo?o said that he likes Maria? b.Jukka 1 kertoi, ett? h?n 1/2 oli myynyt auton (SF) Jukka-Nom said-3Sg that he-Nom had-3Sg sold car-Acc ?Jukka said that he had sold the car? c.Jukka 1 kertoi, et se 1/2 oli myynyt auton (CF) Jukka-Nom said-3Sg that he-Nom had-3Sg sold car-Acc ?Jukka said that he had sold the car? 19 Holmberg (2003) however reports that in some dialects of Finnish the embedded subject pronouns are forced to be disjoint in reference. As for Brazilian Portuguese, coreference seems to be accepted in all the dialects. 80 Besides that, these languages do not license referential 3 rd Person null subjects in matrix clauses. As observed in Tarallo (1996), the null subject in (14a) can receive a referential, specific reading in European Portuguese (14b), but in BP only an undetermined, non-specific reading (14c) is available. The data in (15) illustrate the unacceptability of matrix clauses containing a referential 3 rd P null subject in Finnish: (14)a.e n?o usa mais saia not wear-3Sg more skirts b. ?He/she does no wear skirts any more? (EP/*BP) c. ?Skirts are not worn any more? (BP) (15)a. *(H?n) oli v?synyt. (SF) he was-3Sg tired b. *(Se) oli v?synyt(CF) he was-3Sg tired ?He was tired? In sum, the facts presented so far suggest that neither BP nor Finnish are not null subject grammars. Here I will argue that this incapability of licensing null subjects is to be correlated to the fact that these grammars have weak verbal morphology at least for the 3 rd P. Hence, one might expect referential null subjects to be disallowed altogether in these grammars. However, this expectation fails because, as shown in (16), 1 st P and 2 nd P 81 null subjects are allowed in main clauses, and 3 rd Person referential null subjects in embedded clauses are also possible, as illustrated in (17): (16) a.e falei com o Jo?o ontem a noite (BP) spoke-1Sg with the Jo?o yesterday at night ?I spoke with Jo?o yesterday night? b.e istun si? (Finnish) sit-1Sg here c.e istut si? sit-2Sg here d. e istumme si? sit-1Pl here f. e istutte si? sit-2Pl here ?I/you(Sg)/we you(Pl) am/are sitting here? (17)a.Jo?o 1 disse que e 1 vem amanh? (BP) the Jo?o said-3Sg that come-3Sg tomorrow ?Jo?o said that he will come tomorrow? b. Veljeni 1 oli niin iloinen ettei e 1 voinut nukkua (SF) brother-1 st Px be-3SgPres so happy that-not-3Sg could sleep-Inf 82 c. [Mun veli] 1 oli ninn iolnen ettei e 1 voinu nukkua (CF) my-Gen brother-Nom be-3SgPres so happy that-not-3Sg could sleep-Inf ?My brother was so happy that he couldn?t sleep? Clarifications about the data in (17) are in order. First, in matrix clauses, BP allows only 1 st PSg null subjects. Second, according to Vainikka (1989) and Vanikka and Levi (1999), in matrix clauses, CF does not license referential null subjects at all. However, Hoinonen?s work shows that 1 st P null subjects are allowed in special contexts, as replies to answers. All in all, in the remainder of this chapter and in chapter 4, I will argue that (16) and (17) should not be taken as evidence that Finnish and BP still license referential null subjects. I will suggest that these empty subjects should be analyzed as traces rather than as null pronouns. 3.4 Matrix Null Subjects: Cases of Topic Deletion 20 Duarte (1999) observed that, although 1 st Sg empty subjects are still allowed in Modern BP, there is an evident preference for the use of an overt pronoun in configurations like (18), in which Spec CP is filled: 20 The discussion presented here on BP was first published in Rodrigues (2002). 83 (18) Eu nao sei [ CP quanto tempo [eu moro l?]] (Duarte 1995) I not know-1Sgs how-many time I live-1Sg there ?I don?t know how long I have been living there? The correlation between filled Spec CP and overt subject pronouns seems to be stronger than pointed out by Duarte. It is not the case that there is a preference for the insertion of an overt subject pronoun in configurations with filled Spec CP. In these configurations, the presence of the overt pronoun is obligatory, as exemplified by the unacceptability of (19): (19) a. *? Quem e esqueci de citar who forgot-1St of quote (cf. Quem eu esqueci de citar) ?Whom did I forgot to quote? b. *? O que e fiz (cf. O que eu fiz ) what I did-1Sg ?What did I do? c. ?*Por que fui despedida why was-1Sgt fired (cf. Por que eu fui despedida) ?Why was I fired? 84 Another interesting fact is the unacceptability of (20), which suggests that BP main clauses with a topicalized phrase do not allow 1 st PSg null subjects. As illustrated, in this case, the overt subject pronoun eu ?I? is required: (20) * O Jo?o, e acho que vai ser promovido (BP) the Jo?o think-1Sg that will be-Inf promoted (cf. O Jo?o, eu acho que vai ser promovido) ?As for Jo?o, I think he is going to be promoted? These restrictions on the use of the 1 st Sg null subject are a characteristic of Modern BP. Constructions like in (21) and (22) can be found in 19 th -century BP popular plays, though they are not abundant: (21)E o que e direi da menina and what say-1SgFut of-the girl ?And what am I going to say about the girl? (Martins Pena 1844. O Judas no S?bado de Aleluia (22)Mas, olha, o meu vestido est? quase pronto, e o teu, e n?o sei quando but look the my dress is-3Sg almost done and the yours not know-1Sgwhen 85 estar? be-3SgFut ?But, look, my dress is almost done, and yours, I don?t know when it will be done? (Martins Pena 1844. O Judas no S?bado de Aleluia) Another contrast between 19 th -century and modern BP is that 1 st Sg null subjects could occur inside relative clause in the 19 th century (23), but nowadays they cannot, as shown in (24): (23) Que se case, e quanto antes, com a noiva [que e lhe that SE marry-3Sg and as soon as possible with the financ?e that to.you dou] give-1Sg ?As soon as possible, get married with the fianc?e that I give you? (Martins Pena 1844. O Judas no S?bado de Aleluia) (24)*? Eu comprei aquele vestido [que e vi ontem] I buy-1SgPast that dress that saw-1Sg yesterday ?I bought that dress that I saw yesterday? As expected, (24) becomes acceptable if the overt pronoun eu is inserted as the subject of the relative clause: 86 (25)Eu comprei aquele vestido que eu vi ontem I bought-1Sg that dress that I saw-1Sg yesterday ?I bought that dress that I saw? In sum, the licensing conditions of 1stPSg null subjects in BP have changed: during the 19 th century, these subjects were licensed in more configurations than they are now. This leads us to speculate about the nature of these subjects in current BP. If they are real null subjects, they should not be subject to the restrictions shown above; they would behave as in the 19 th century. To understand the real nature of these empty subjects, it might be interesting to compare modern BP with spoken German, which, despite being a non pro-drop language, allows omission of the 1stPSg subject pronoun in sentences like (26): (26) (Ich) hab? ihn schon gesehen (Ross, 1982) ?I saw him already? Like the BP pronoun eu ? I?, ich cannot be dropped in all configurations. Ross (1982) remarked that ich can be omitted only in the sentence-initial position, i.e. when it is a topic. In (27), for instance, since the topic position is occupied by the object pronoun ihn ?him?, the subject ich remains in the third position and cannot be omitted: 87 (27) Ihn hab *(ich) schon gesehen (Ross ,1982) ?I saw him already? In addition to the similarity in topic-position dependency, 1 st PSg empty subjects in spoken German and in modern BP pattern alike in that they are both prohibited in clauses with a fronted wh-phrase and inside relative clauses. The data in (28) make the case for spoken German (for BP, see examples (19) and (24)). (28) a. * Was machte ?What did I make? b. *Ich kenne das m?dchen, da? gestern getroffen habe ?I know the girl that I met yesterday? In essence, 1stPSg null subjects in modern BP behave exactly as ich omission in spoken German. Thus, given that German is not a pro-drop language, an analysis accounting for this parallelism should not assume that those subjects are null pronouns in modern BP. Ross (1982) analyzes these gaps in German as formed by deletion of the topicalized subject pronoun (?Pronoun Zap? in his terminology). Thus, in (27) ich cannot be deleted because it is not in a topic position. Huang (1984) unifies German null subjects with Chinese null objects, arguing that they are both variables. According to his proposal, these empty arguments are formed in 88 the transformational component via movement of an overt pronoun to a topic position, which turns out to be a target for deletion. Therefore, the gap created by the movement is a variable bound by a zero topic. Modesto (2000) already extended Huang?s analysis for BP assuming that the possibility of having 3rdPSg null subjects in question-answer contexts like (98) is due to a topic-deletion operation: (29) A: O Jo?o telefonou the Jo?o caled3Sg ?Did Jo?o call? B: e telefonou ontem de manh? calledSg yesterday of morming ?He called yesterday morning? Following a suggestion of Jairo Nunes (personal communication), I propose we take topic deletion as responsible also for the allowance of 1 st PSg empty subjects in modern BP matrix clauses. I believe this proposal is motivated by the similarity with spoken German. As in spoken German, the BP 1 st PSg pronoun can be deleted only if it is in a topic position. Thus, if this position is already filled by another constituent, the deletion cannot take place (cf. (20)). The unacceptability of those empty subjects in configurations containing a fronted wh-phrase (cf (19)) and inside relative clauses (cf. (20) can be explained by considering that movement of another constituent towards Spec 89 of CP blocks the movement of the subject pronoun to the topic position, a Minimality effect (Rizzi, 1990). I will not work out the technical details of the topic-deletion operation (see Huang 1984). Here, the main point is that modern BP 1stPSg null subjects are not null pronouns. They do not have the freedom of null pronouns, being allowed only in configurations in which spoken German can drop the pronoun ich. The disallowance of those subjects in structures with either a filled Spec of CP or a topicalized phrase indicates that the topic deletion analysis might be right. The application of this topic deletion or pronoun zap might be possible only when the verbal morphology expressing agreement is rich enough to recover the features of the deleted subject pronoun. Now let me turn to Finnish. There is empirical evidence that this analysis can be extended for matrix null subjects in this language. First, notice that Finnish referential 3 rd P subject pronouns can be dropped in question-answer contexts (cf. (30). 21 Topic deletion analysis suggested above explains (3) in a straightforward manner. (30)a.A:Puhuuko Liisa englantia (S/CF) Speak-3Sg-Qcl Liisa-Nom English-Acc ?Does Liisa speak English? 21 For an analysis of Yes/No questions in Finnish, see Holmberg (2001) 90 B: e puhuu speak-3Sg ?Yes, she does? As for 1 st P and 2 nd P null subjects, Vainikka and Levi (1999) remark that they are not allowed in matrix clauses with a topicalized constituent. That?s why the sentences in (31) are unacceptable. (31)a.* kun soititte, kaupassa e olimme juuri ostamassa takkia when called-2Pl store-Ine were-1Pl just buy-Inf coat-Part ?When you called, we were just at the store buying a coat? b.* Palkankorotusta e pyysin heti raise-Par asked-1Sg immediately ? I asked for a raise immediately? It is worth noting that if the subject is a full DP or a pronoun, topicalization occurs without difficulty: (32)a. Jos kaupalta soittaa asiakas, Pekan on if store-Abl call-3Sg costumer-Nom Pekka-Gen have-3Sg 91 l?hdett?v? sinne Leave-PastPart there ? If a customer calls from the store, Pekka has to go there? b. Palkankorotusta pyysi heti. Liisa. raise-Par asked-1Sg immediately Lissa-Nom ?Liisa asked for a raise immediately? c.Pariisissa mina olen k?ynyt(Holmberg, 2003) Paris-Ine I-Nom have-1Sg visited ?It is Paris that I?ve been to? Since Finnish allows only one fronted topic per sentence (cf. Holmberg and Nikanne 2002), it follows that in (31)-(32) topicalization of another constituent bars topicalization of the subject, which, being forced to stay in a lower external-VP position, arguably in spec of TP, cannot be the target of deletion. Holmberg and Nikanne (2002) and Holmberg (2003) cite the fact, originally observed by Auli Hakuninen, that 1 st P and 2 nd P null subjects cannot co-occur with an expletive, as exemplified in (33): (33)* Sit? uskon vallankomousksen Expl believe+1 st Sg revolutions+Ill ?I belive in Revolution? 92 This datum leads Holmberg (2003) to propose the very same analysis I am suggesting here: null subjects in Finnish involve pronoun deletion. 22 The idea here is that the expletive occurs in the topic position (spec of FP in Holmberg and Nikanne?s analysis), blocking, thus, deletion of the subject pronoun. One could object to extending the topic deletion analysis to Finnish since in this language 1 st P and 2 nd P null subjects can co-occur with a wh-phrase in spec of CP: (34)Miss? e voimme lukea sanomalehti? Where can-1Pl read-Inf newspapers-Par ?Where can we read newspapers? If (34) involves topicalization followed by deletion of the subject pronouns, we expect this topicalization process to interfere with the wh-movement. Therefore, given that (34) is acceptable, a objection to a topic deletion analysis of Finnish should be in principle granted. Nevertheless, I maintain that this objection might be too hasty. As the data in (35), extracted from Holmberg and Nikanne (2002), show, Finnish has an uncommon property: It does not display Minimality effects in structures with both wh- movement and topicalization. (35)Kuka t?m?n kirjan on kirjoittanut who this book-Acc have-3SgPres write-PastPart ?(What about this book:) Who has written this book? 22 Basing his argument on Finnish, Holmberg generalizes and propose that null subjects are universally deleted pronouns. 93 Still, as pointed out by Holmberg and Nikanne, it could be that the derivation of (35) does not make use of topicalization, rather it involves scrambling the DP t?m?n kirjan ?this book? to a position outside the VP shell, but lower than the topic site. Holmberg and Nikanne disregard, correctly, I think, a scrambling-analysis for (35) because there are cases like (36), where the displaced constituent surfaces in a position higher than the negation, which is higher than the verb. The verb, in turn, is adjoined to T 0 . Putting it a more direct way: the surface position of minusta in (36) is too distant for scrambling, being reasonably a topic position. (36) Kuka minusta ei pid? 23 who me not-3Sg like ?Who doesn?t like me? As already said, Holmberg and Nikanne take this topic position to be spec of FP. They try to explain the absence of interaction between movement to spec of CP and movement to spec of FP as following from the movement theory stated in Chomsky (1995b), where attraction is defined with respect to features. Their line of reasoning is the following: if the phrase occupying the spec of FP does not contain a wh-feature, it does not block attraction between C 0 and a lower wh-feature. I will not offer an alternative analysis for this lack of Minimality effects. For the purpose of the present discussion, it is sufficient to observe that whatever explains (35) also explains the topicalization of the deleted subject pronoun in (34). However, it is 23 In Finnish, the negation word behaves like a verb, agreeing with the subject in f-features. When it is present, it carries the f-feature agreement, and the verb is marked only for tense. 94 important to observe that Holmberg and Nikanne?s solution is not a desirable one. They explain Finnish leaving uncovered all the cross-linguistic data that shows interaction between wh-movement and topicalization. 3.5 Embedded 3 rd P Null Subjects In this section, I will prepare the way for a new analysis of embedded 3 rd P referential null subjects in the languages under investigation. To do so, I will first dismiss the possibility of being dealing with logophoric anaphors, and then will revise the previous proposals, trying to justify the necessity of a new one. 3.5.1 Absence of Logophoric Properties Sells (1987) and Reinhart and Reuland (1991) show that relations between antecedents and pronouns or anaphors might be logophoric. According to them, a logophoric relationship has two distinct features: (a) the relationship between the pronoun/anaphor and its antecedent does not obey locality conditions; (b) the antecedent is the logophoric center of the sentence. 24 That is, the antecedent is the thinker, the perceiver or the entity whose consciousness or feelings are reported by the sentence. In (37), for instance, the relationship between Bill and PRO is logophoric: Even though Bill 24 Sells identifies three types of logophoric antecedents: (a) the source of the report, (b) the Self, the entity whose mental states the report describes, and (c) the Pivot, which provides the center of deixis or perspective the report. 95 does not c-command PRO, Bill is the antecedent of PRO and the entity whose point of view is being reported. (37)PRO 1 having arrived in town, the main hotel seemed to Bill 1 to be the best place to stay (37) contrasts with (38), a BP sentence, with a referential 3 rd P null subject. (38) is ungrammatical because the antecedent of the null subject does not c-command it, and, as I will show in chapter 4 (4.3.1), BP null subjects are anaphoric elements being dependent on the presence of a sentential antecedent. (38)?* Nada mais fazia sentido para o Jo?o 1 quando e 1 ficou depressivo nothing more made-3sg sense to the Jo?o when got-3sg depressed ?Everything lost its sense to Jo?o when he got depressed? Therefore, in BP locality matters for the anaphoric relationship between a 3 rd P null subject and a DP. In (38), even though Jo?o is the entity whose point of view is being reported, it fails to be the antecedent because it does not c-command the null subject. Observe that if an overt pronoun replaces the null subject, the sentence becomes acceptable: 96 (39) Nada mais fazia sentido para o Jo?o 1 quando ele ficou depressivo nothing more made-3sg sense to the Jo?o when he got-3sg depressed ?Everything lost its sense to Jo?o when he got depressed? Another property of a logophoric relationship is that the antecedent may not be overtly realized, remaining implicit as (40). (40) Having just arrived in town, the new hotel seemed like a good place to stop The unacceptability of (41) is another piece of evidence that referential 3 rd P null subjects in BP are not logophors. The antecedent of these null subjects cannot be implicit. (41)* A Ana foi tida como a mo?a mais bonita da regi?o, quando e viu the Ana was-3Sg taken as the girl most beautiful of.the region when saw-3Sg ela andando sozinha pela praia she walking-3Sg alone on.the beach ?Ana was taken as the most beautiful girl of the region when he saw her talking alone along the beach? One again notice that the presence of an overt subject pronoun improves the acceptability of the sentence: 97 (42)? A Ana foi tida como a mo?a mais bonita da regi?o, quando ele viu the Ana was-3Sg taken as the girl most beautiful of.the region when he saw-3Sg ela andando sozinha pela praia she walking-3Sg alone on.the beach ?Ana was taken as the most beautiful girl of the region when he saw her talking alone along the beach? Reinhart and Reuland (1993) mention that another property of logophoric anaphors is that they can be inside a subjectless NP, as (43) exemplifies: (43)Lucie 1 saw a picture of herself 1 If this is right, the unacceptability of (44) is extra evidence that the referential null subjects we are studying are not logophoric anaphors. If they were they would be license inside subjectless NPs. (44)?? A Leticia viu as fotos de quando e foi visitar a Ana the Leticia saw-3Sg the pictures of when went-3Sg visit-Inf the Ana (cf. A Leticia viu as fotos de quando ela foi visitar a Ana) ? Leticia saw the picture of the visit she paid Ana? In sum, the evidence above suggests that the null subjects under discussion are not logophoric anaphors. 98 3.5.2 Previous Analysis 3.5.2.1Referential 3 rd P Null Subjects as PRO Pursuing the old idea (Taraldsen, 1980) that allowance of null subjects is correlated to rich morphology, and building on Giv?n?s (1978) insight that verbal agreement (Agr, hereafter) is the result of grammaticalization of subject pronouns, 25 Kato (1999, 2000) eliminates pro from the list of lexical pronouns. She proposes that in Romance what is named as pro is actually Agr itself, which is taken to be an independent lexical D item furnished with Case and f-features. Agr enters the derivation as the external argument of VP. Being a maximal minimal projection, Agr adjoins to T 0 , where it checks the EPP features of T and its own Case feature. Thus, a projection of a specifier inside the TP is prevented. 26 Accordingly, the Spanish sentence (45a) has the structure in (45b), in which Agr attaches to the verbs after V-to-T movement: (45) a. viaj?. traveled-3sg ?S/he traveled? b. [ TP [ T? [ T viaj 2 -[ Agr o] 1 + T][ VP t 1 t 2 ]]] 25 On this grammaticalization process, see also Speas (1994) and Roberts and Roussou (1999). 26 Rohrbacher (1992), Galves (1993), Speas (1994), Alexiadou and Anagnostopoulou (1998), Manzini and Savoia (2002) and Platzack (1995). 99 Moreover, as represented in (45), in this proposal overt subjects are taken to be instances of doubled subjects in the sense that they double Agr, being generated in spec of SP, a left periphery functional projection proposed as proposed in Martins (1994). 27 (46) a.El viaj? he traveled-3sg ?He traveled? b.[ S P el [ TP [ T? [ T viaj 2 -[ Agr -o] 1 + T][ VP t 2 t 1 ]] Taking Agr to be in complementary distribution/competition with weak pronouns, Kato proposes that the existence of weak pronouns is the trigger for the loss of null subjects. Children take the existence of such pronouns in the Primarily Linguistic Data (PLD) to mean that Agr is not an independent lexical item in their target grammar. Thus, since BP has developed a series of nominative weak pronouns, the author concludes that this language is on its way to become non Pro-Drop. Her hypothesis is that BP Agr has lost its syntactic independence and may enter the derivation already as a verbal suffix. This hypothesis posits a question about the licensing of the embedded 3 rd P null subjects in sentence like (47). (47)O Pedro 1 disse que e 1 conserta sapato the Pedro say-3SgPast that repair-3SgPres shoe ?Pedro said that he repairs shoes? 27 For a defense of SP in Brazilian Portuguese Grammars , see also Galves (2001) and Britto (1998). Britto argues that, since the Brazilian Portuguese grammars lost subject inversion, they encode categorial 100 Kato (1999) argues that these null subjects are license via a traditional obligatory control configuration. As shown in (48), in her proposal, (47) involves the presence of a PRO in spec of spec of SP, doubling Agr. It is assumed by the author that PRO is deficient with respect to f-features and, as a result, it is dependent on external control. Hence in (48) PRO is obligatorily controlled by the matrix subject. (48) a.[O Pedro 1 disse que [ S P PRO 1 [ TP conserta-Agr 1 sapato]]] the Pedro said-3Sg that repair-3Sg shoe ?Pedro said that he repairs shoes? It is also assumed that in a structure in which no antecedent is provided, the f- features of PRO remains unspecified, causing PRO to receive a generic or arbitrary reading, as in (49): (49)[ S P PRO 1 [ TP aqui conserta-Agr 1 sapato]] here repair-3Sg shoe ?One repairs shoes here? Thus, I think it is fair to present Kato?s achievements in the following way: she succeeded in eliminating pro as a lexical category by proposing that embedded referential null subjects in BP are cases of obligatory control ? la Borer (1989). I concur with the elimination of pro as an independent lexical item (cf. section 4.2), however I think we jugdements (Kuroda 1972) via NP-movement to SP. 101 can accomplish a more elegant analysis for embedded null subjects in BP if we treat them as the residue of movement. A new analysis for these null subjects seems to be motivated empirically. As discussed below, the analysis shown in (48) and (49) is not free of problems. First, though Kato does not discuss the mechanism behind her analysis for generic null subjects (49), it should be noticed that the presence of a possible controller does not ensure obligatory control. In (50), for instance, although the matrix DP would be an appropriate antecedent, the presupposed PRO remains unbound receiving a generic interpretation. (50) a.[O Pedro 1 disse que [ S P PRO 1 [ TP conserta-Agr 1 sapato]]] the Pedro said-3Sg that rapair-3Sg shoe ?Pedro said that one repairs shoes here? Moreover, it seems to me that the presence of PRO in (48) and (49) is not well motivated. For Kato, both the 3 rd P Agr and PRO are f-deficient lacking person, 28 hence it is unclear what the empirical or conceptual gain is in assuming that a f-deficient item is doubled by another f-deficient item. Actually, we have a reason to believe that this doubling does suit the analysis in its total. As Kato acknowledged, placing a PRO, a deficient empty category, in spec of SP is an unnatural move, since she follows Cardinaletti and Starke (1994) in reserving spec of SP for strong pronouns and Full DPs. 28 Cf Kato (1999:27) for a discussion on the f-features of Agr. 102 A third issue omitted in Kato?s discussion was brought out by Modesto (2000). As he observed, the properties of Agr in BP are not well defined by Kato. From the fact that spec of TP is not projected in (48)-(49), 29 we can conclude that Agr is still able to satisfy the requirement of T, and able to receive the verbal q-role. 30 However, since Kato assumes that weak pronouns are spelled-out in spec of TP, in accounting for the acceptability of (51) we now, contrary to what we have just concluded, presuppose that BP Agr neither satisfy the requirements of T nor receives the theta assigned by the verb. These tasks are done by the weak pronoun. (51)Eu, eu (phonetically [o]) sinto demais isso, neh (Kato, 1999:14) I, I feel-1sg too-much this, right ?I feel this too much, right? 3.5.2.23 rd P Null Subjects as LF Bound Variables Modesto (2000), building on Vergnaud and Zubizarreta (1992), assumes that an argument needs a denotational index in order to be visible for q-role assignment at LF. His further assumption is that pro is universally generated without this index. 31 As a consequence, pro is licensed only when it borrows an index from an identifier. The 29 In the original these examples are numbered as (46a,b). 30 Kato (1999:28) asserts that ?though weak pronouns can now appear in Spec of TP like in English, third person Agr retains its argument status and can be bound by an external element.? 31 Modesto (1999) defended a variant of his (2000) analysis, which pro lack f-features. I will make comments only on Modesto (2000) because, despite the different view on the nature of pro, the two analyses are pretty much the same. 103 identifier can be either the verbal agreement morpheme (Agr), which is taken to be generated as an adjunction to the functional head AGR, or an antecedent in an A?- position. Identification via Agr occurs in languages with rich agreement morphology like Spanish, Italian and European Portuguese. In these languages, pro acquires its index from Agr in a Spec-Head relation. In BP, the agreement morphology is too meager to identify pro; hence pro is identified by an antecedent in an A?-position and interpreted as a bound variable at LF. 32 To account for the fact that in BP subjects can be the binder of pro (cf. (52a)), the author argues that in this language the subject occupies an A?-position which is taken to be spec of AgrsP. Consequently, (52a) has the structure in (52b), where non-relevant projections have been omitted. (52) a.Jo?o 1 disse que pro 1 nadou ontem Jo?o said-3Sg that swan-3Sg yesterday ?John said that he swam yesterday? b.[ AGRP Jo?o 2 [ AGR? [ TP t 2 [ T? [ VP t 2 disse[ CP que [ AGRP pro 2 [ AGR? [ TP t 2 [ T? [ VP t 2 nadou ontem ]]]]]]]]]] Inside the embedded clause of (52b), pro is generated in spec of VP, moves to spec of TP, where it has its Case feature checked, and then, driven by the EPP feature of Agr, pro moves to spec of AgrsP. In the matrix clause, the same happens to the DP subject: it starts in spec of VP, moves to spec of TP and then to spec of AgrP. At LF, pro is 32 Since indices are assigned at LF, this analysis tacitly denies the idea the inclusiveness condition (Chomsky, 1995:228). holds of the computation from the numeration to LF 104 interpreted as a variable bound by the matrix subject, which shares its denotational index with pro. The advantage of this analysis is that it neatly explains the pair of sentences in (53). 33 In (53a), the in situ object cannot be the binder of pro because it is not in an A?- position. Therefore, if it undergoes movement to spec of CP, it becomes the binder, as shown in (53b). (53)a. A Maria 1 convenceu o Pedro 2 que pro 1,*2 tinha de sair the Maria convinced-3Sg the Pedro that had-3Sg of leave-INF ?Maria convinced Pedro that she had to leave? b.Quem 2 que a Maria 1 convenceu t 2 que pro ??1/2 tinha de sair who that the Maria convince-3Sg that have-3Sg of leave-INF ?Who did Maria convince that he had to leave? Modesto explains that in (53b) the DP subject cannot be the binder because it is in spec of TP. On its way to spec of CP, the wh-object passes through spec of AgrP, checking the EPP feature of Agr, blocking, thus, further movement of the subject, which remains in spec of TP throughout the derivation. Now that I have briefly shown how the analysis works, let me go through some of its problems. The assignment of indices mentioned above is not particularly easy to understand. By adopting Zubizarreta and Vergnaud?s (1992) framework, Modesto includes as part of 33 In 4.5.3, I will revisit these sentences. 105 his theory a level of representation of sentence grammar named L-Structure. This level functions as an indexing of nominal categories. Vergnaud and Zubizarreta define it as a set of indices (denotata) that are mapped into the nominal expressions of the sentence. 34 Though Modesto doesn?t discuss it, in his analysis this mapping should occur before the binding process of pro, such that the binder (a nominal expression) is already furnished with the index to be transmitted to pro during the process of binding. Another possibility is that the binding process itself consists of mapping pro and its antecedent into the same index/element in the set of denotata. Either way, there is a question about the interaction between L-structure and the grammar. Should L-structure be considered a level of representation within the grammar or not? If yes, the analysis poses a question about the number of levels of representation that compose the grammar. If not, the question is about the interaction between the grammar and this mapping process. Apart from questions about its mechanics, the analysis has some unsolved empirical issues as well. Modesto explains (53b) by assuming that Minimality imposes a derivation in which the wh-movement forces the subject to stay in spec of TP. However in (54), despite the occurrence of wh-movement, pro is bound by the subject. (54)a. O que o Pedro 1 disse depois que pro 1 saiu da festa what the Pedro said-3Sg after that left-3Sg the party ?What did Pedro say after having left the party? 34 It is formalized in the following way: (i) L-Structure is a set of elements D = {i,j,k,?} such that each element of D is associated with some Nominal in the sentence. (Vergnaud and Zubizarreta 1992:610) 106 b.Para quem o Pedro 1 mentiu que pro 1 tinha de sair to whom the Pedro lied-3Sg that have-3Sg of leave-INF ?To whom did Pedro lie that he had to leave? Modesto?s solution to this problem is to assume that the derivations of (54a&b) are allowed to violate Minimality in order to converge. Despite occurrence wh- movement, the subject is raised to spec of AgrP because otherwise there wouldn?t be a convergent derivation. Pro would reach LF without being bound. This is not the most desirable solution since usually one is not allowed to violate Minimality in order to ensure the mapping of a lexical array into a pair of sound and meaning. Moreover, this solution raises a difficult question. Let me start by setting up the context needed in order to make the question understandable. Wh-movement is optional in Brazilian Portuguese and the distinction between overt and covert wh-movement has no impact on the meaning conveyed by the sentence. For instance, (54a&b) means essentially the same as (55a&b). The only difference between them is that (54a&b) display overt wh-movement, whereas (55a&b) do not. (55)a. O Pedro 1 disse o que depois que pro 1 saiu da festa the Pedro said-3Sg what after that left-3Sg the party ?What did Pedro say after having left the party? 107 b.O Pedro 1 mentiu para quem que pro 1 tinha de sair the Pedro lied-3sg to whom that have-3sgPast of leave-INF ?To whom did Pedro lie that he had to leave? Thus, (55a &b) are similar to (56) and, as noticed by Modesto, (56) is unambiguous with respect to the binding of pro. Only a reading in which the matrix subject is the binder is possible. Hence, Modesto needs to assume that covert wh-movement does not interact with subject movement to AGRP. (56)O Pedro 1 convenceu quem 2 que e 1/*2 tinha de sair? the peter convinced-3Sg who that had-3Sg of leave-Inf ?Who did Peter convince that he had to leave?? Now, one could imagine a grammar that is almost exactly like BP, call it BP?. The only difference between BP and BP? would be that, when the wh-element is not a proper binder for pro, instead of using minimality violation as the strategy to guarantee convergence, BP? uses wh-in-situ as the strategy. Putting it in different words, In BP? if a wh-phrase is not a good semantic binder for pro, the wh-phrase is forced to stay in situ. Thus, the difficult question is: why does BP? not exist? Or why is it that BP is not BP?? The absence of an answer to these questions in Modesto?s analysis creates an impasse. This impasse dissolves if we drop the assumption that BP embedded 3 rd P null subjects are bound variables. 108 It seems to me that the datum in (57) is also problematic for this analysis. (57) O Jo?o 1 for?ou a Maria 2 a PRO *1/2 dizer que pro *1/2 tinha de sair the Jo?o forced-3sg the Maria to say-Inf that had-3sg of leave-Inf ?Jo?o forced Maria to say that she had to leave.? Modesto argues that only subjects can be binders of pro in BP. But, this is not an accurate description since an object can be the ultimate binder as long as PRO intermediates the relation, as in (57). Hence, PRO must be in an A-position, otherwise condition C would be violated. It follows, then, that pro cannot be an A?-bound variable, as proposed by Modesto. Note that if pro can be A-bound this problem just disappears. Modesto actually argues that this problem is circumvented because, following Borer (1989) and Landau (1999), he takes the AGR of infinitive clauses to be anaphoric. In (57) AGR is bound and assigned its denotational index by the matrix object. After that, AGR transfers its index to PRO. Therefore, PRO gets the index of the matrix object in a roundabout way, via AGR. It remains, however, unclear to me how this mechanism avoids the problem. First, if AGR is the head of an A?-projection, AGR itself is arguably an A?-element. Hence, like PRO, AGR should be subject to principle C. Second, the assignment of indices via binding as proposed by Modesto seems to be very similar to the indexing mechanism used in the Government and Binding Theory (GB) (cf. for instance Chomsky 1981). A property of GB indexing is transitivity: if a is coindexed with b and b 109 is coindexed with g, then a is coindexed with g. 35 There is nothing in Modesto?s analysis saying that assignment of indices is not transitive. If we take it to be transitive, the result is clear: the binding of PRO in (57) violates condition C. Therefore I conclude that A?-bar binding is not clearly involved in the licensing of Brazilian Portuguese 3 rd P null subjects. 3.5.2.33 rd P Null Subjects and The Principle of Obligatory Occupant Licensing 36 Vainikka?s (1989) proposes that verbal agreement suffixes in SF are anaphors. The 1 st P and 2 nd P agreement suffixes are bound by an implicit speaker-NP and by an implicit hearer-NP respectively. Since there is no corresponding implicit NP that could bind the 3 rd P agreement suffixes, 3 rd P null subjects are not allowed in matrix clauses. Vainikka and Levy (1999), basing on Vainikka (1999), entertain an analysis for the mixed pattern of null subject in Finnish and Hebrew, assuming a system of binary features, according to which 1 st P is coded as [+speaker], the 2 nd P as [+ hearer] and the 3 rd P as [-speaker, - hearer]. 36 They suggest that these features are carried by pronouns, by inflected verbs and by the head of the AGR Projection. The essence of their proposal can be summarized in the representation (58) and (59). The features [+speaker] and [+ hearer] can be independently base generated in spec of VP, instead of being base generated 35 That?s why the coindexation in (i) violates the i-within-i filter, even if himself is coindexed AGR. AGR is coindexed with the whole NP by virtue of its person and number agreement. Therefore, if himself is coindexed with AGR, it is by transitivity coindexed with the NP. (i) [[ NPi a picture of [ NP i himself] AGR i will be on display. 36 This is only syntactic account I am aware of for Null Subjects in Finnish. Gutman (1999) provides an account within a theory of discourse. Since our frameworks are different, I will not revise her proposal. 36 They take these features to be N-features. 110 directly in AGR. For the authors, these features can be inserted as the VP external argument because they are able to restrict the set of potential referents that are relevant for the conversational situation (cf. p.623). Since the 3 rd P features [-speaker, - hearer] are not able to restrict the set of potential referents, these features can not be independently inserted as the external argument. Therefore, they are base generated directly as part of AGR. Thus, the deficiency of the 3rdP features forces the derivation in (59b), where an overt pronoun is inserted as the external argument of the verb: 37 (58)a. Tulen (SF) come-1 st sgPres ? I come.? b. [ AGRP [ AGR? [ VP [+speaker] [ V? tulen]]]] (59)a.*(H?n) tulee. (S/CF) he/she come-3SgPres ?He she comes? b. [ AGRP [ AGR? [ AGR0 [-speaker] ?hearer]] [ VP h?n [-speak] [-hearer] [ V? tulen]]]] They adopt as part of the Universal Grammar the principle in (60), which is a modified version of Speas? (1994) principle of licensing: 38 37 I am being faithful to the authors? structures (cf. their examples (7) and (8), p. 626-27), in which no projection intervenes between AGRP and VP. 111 (60)Principle of Obligatory Occupant Licensing In order to be licensed, both the head and the specifier of a syntactic projection must be filled by syntactic material at some level of representation. (Vainikka and Levy, 1999:627) Since AGRP is subject to (60), both its head and its specifier must be filled at some point of the derivation. Hence, the derivations in (58)-(59) are completed when the functional projection AGRP is licensed. For this to happen, the element inserted in spec of VP, (the [+speaker] features in (58), the pronoun h?n in (59), moves to spec of AGRP and V adjoins to AGR 0 . As for the allowance of 3 rd Person subjects inside embedded clauses (61a), it is proposed that in this case, the 3 rd P features [-speaker, - hearer] are generated in spec of VP and then moves to spec of AGRP to satisfy the principle in (60), as shown in representation sketched in (61b): (61)a.Jukka 1 kertoi ett? e 1 oli v?synyt jukka-Nom said-3Sg that was--3Sg tired ?Jukka said that he was tired? b.[ ?Jukka?[ CP ett? [ AGP [-speaker,-hearer] 1 [ AGR? [ AGr0 oli 2 ] [?[ VP t 1 [ V? t 2 v?synyt ]]]]]] 38 Speas? principle of licensing is defined as in (i), where ? having content? means ?having a distinction of itself phonological or semantic matrix?: Project XP only if XP has content. 112 The authors contend that this derivation is possible because the referent of embedded 3 rd P features is determined indirectly via the derived context of matrix clause. That is, the embedded subject does not refer by itself, hence it has to borrow the referent of a DP in the matrix clause. It means that in (61), the embedded null subject is semantically dependent on the matrix subject, Jukka. Though this proposal is mechanically doable, it appears only to encode the facts. Moreover, as notice by the authors, this analysis provides no way of accounting for the conditions under which embedded subject omission is possible. Thus, the embedded referential null subjects in Finnish require further research, as the authors acknowledged, and I will provide in chapter 4. 3.6 Conclusions The big picture of the present chapter can be summarized as follows: Modern BP and Finnish are not genuine null subject languages. This is arguably related to the fact that in these languages the verbal agreement morphology is weak at least for the 3 rd P. I suggested that cases of matrix referential null subjects do not constitute evidence that referential subjects are licensed. These subjects seem to be cases of topic deletion. As for the embedded 3 rd Person null subjects, I showed that these subjects cannot be treated as logophors. I also show that none of previous analysis provides a satisfactory explanation for the partial pro-drop character of BP and Finnish. In the next chapter, I suggest a new analysis for the possibility of embedded 113 3rdPerson null subjects, according to which these subjects are the result of hyper-raising movement. The allowance of this type movement will be correlated to the weakness of the verbal agreement morphology. 114 CHAPTER 4 A-MOVEMENT OUT OF FINITE CLAUSES 4.1 Preliminaries The bulk of this chapter is devoted to the analysis of the embedded 3 rd P null subjects introduced in the previous chapters. It contributes to the understanding of the grammars under investigation by suggestion that these empty subjects are gaps formed by the operation Move, i.e, they are silent copies of their antecedents. Thus, I will present for consideration the idea that A-movement from one thematic subject position to another is to be allowed even in finite configurations. This amounts to saying that the content of this chapter can be seen as an extension of Hornstein?s analysis for obligatory control in the sense that it also defends NP-movement as a creator of ?traces? in subject positions. However, despite the similarity, there are important differences to be considered. First, Hornstein analyzed non-finite configurations, whereas the configurations I am dealing with are finite. Second, the present research differs from that of Horsntein?s in that the phenomenon I am investigating is not universally attested. Therefore, it must be correlated with a specific feature of the grammars in which it is allowed. I will suggest 115 that this feature is the weak verbal agreement morphology discussed in the last chapter. The sections are organized as follows: In 4.2, I lay out the details of the technical implementation of the movement analysis adopted. The purpose of this section is to make explicit the correlation between allowance of subject-to-subject movement and weak verbal agreement morphology. 4.3 presents a series of evidence of movement. 4.4 shows a new argument for a movement analysis of obligatory control, viz. in Romance, the antecedent of PRO controls the gender agreement features of embedded universal quantifiers and past participle forms. It will be shown that the same is observed in BP constructions involving embedded 3 rd P referential null subjects. 4.5 deals with a potential counterargument for the analysis defended in this thesis: allowance of null subjects inside finite adjuncts clauses. The solution to be offered hinges on two assumptions: (a) islands emerge derivationally and adjuncts becomes islands only after being integrated into the main spine of the structure (cf. Uriagereka 1999, Hornstein 2001); (b) movement might target positions in disconnected substructures (Sideward movement ? cf. 2.2.2). Section 4.5.1 looks into another case of null subject inside adjuncts. The novelty of this case is that the null subject, being within an adjunct clause that surfaces in a clause initial position, is not c-commanded by its antecedent at the surface structure. I will treat this as a case of adjunct remnant movement. Modesto?s argument against movement (sentences involving double object verbs like convencer ?to convince?) is revisited in 4.5.6. Section 4.6 offers a summary of the chapter and there will be an appendix, where I will consider null subjects within adnominal adjunct clauses. A movement analysis was previously defended in Rodrigues (2000, 2002) and 116 much of what follows is built on ideas introduced there. Ferreira (2000) also argues for a movement analysis. 1 According to him, a finite f-defective tense has emerged in the BP grammars that, as result, have now two types of finite tenses: one that is f-complete and another one that is f-incomplete. To see how it works, consider the sentence in (1) and its derivation in (2). The derivation starts with one token of each type of tense being ?copied? from the lexicon and place into the numeration (2a). The defective tense is inserted first into the derivation and consequently the DP o Jo?o does not check its Case feature inside the embedded clause (2b). Thus, when the derivation reaches the matrix vP, the DP moves to spec of vP to check the relevant q-role (2c) and from there it is raised to the spec of the matrix TP, where it checks its Case feature against the f-complete T. (1) O Jo?o 1 disse que e 1 viu a Maria the Jo?o say-3SgPast that see-3sgPast the Maria ?John said that he saw Mary? (2)a. Num = {o 1 , Jo?o 1 , n 1 , Tf-def 1 , Tf 1 , disse 1 , que 1 , viu 1 , a 1 , Maria 1 } b.[CP que [C? [ TP [ DP o Jo?o Case ] 1 [ T? [T f -def ][ n P t 1 ] [ n ? [ VP viu a Maria]]]]]]] c. [ TP [ DP o Jo?o case ] 1 [T-f] [ T? [ n P t 1 [n?[ VP disse [ CP que[ C? [ TP t 1 [ T? [T f -def ] [ n P t 1 ][ n ? [ VP viu a Maria ]]]]]]] 1 Rodrigues (2000) and Ferreira (2000) analysis developed simultaneously, but independent of each other. 117 The existence of two types of finite tense is crucial for this analysis, and the order in which they are merged defines the success of the derivation. If the defective tense is merged first, the derivation converges because the DP does not check its Case feature inside the embedded domain being, thus, available for computations at the matrix level. Conversely, if the f-complete tense is the first one to be merged, the derivation crashes. The DP checks its Case feature too early, becoming unavailable for computations at the matrix domain. Although it works, this analysis encounters some difficulties. First, the author offers no independent motivation for the existence of two types of tense. He says it is morphologically based, but he doesn?t show how. The distinction cannot simply be a function of loss of verbal morphology because French lost verbal morphology like BP but did not develop the alleged defective tense. Second, in Rodrigues (2000), I observed that if children are cue-based learners that scan only root domains, as argued by Lightfoot (1989, 1991, 1994 ? the Degree-0 Hypothesis), they cannot acquire the alleged defective tense for it occurs only in embedded domain, i.e.; out of reach of children. Ferreira addressed this issue by bringing forward a fact originally observed by Moreira da Silvia?s (1983), according to which Brazilian Portuguese allows constructions like (3), called hyper-raising in Ura?s (1994) terminology. 2 According to Ferreira, since this type of construction is available on the primarily linguistic data (PLD), it serves as the necessary cue for the existence of a f-defective tense. As the raising verbs do not assign a q-role, children are led to conclude that in (3) the matrix subject comes from the embedded 2 According to Ura (1994), Finnish also allows hyper-raising. 118 domain and its nominative Case is checked only at the matrix clause because the embedded tense is f-defective. (3) A Maria 1 parece que t 1 est? doente the Maria seem-3Sg that is-3sg-Pres sick ?Maria seems to be sick? This is a legitimate way to address the problem raised. However, it leaves open the question about the robustness of (3) as a cue. 3 Is (3) a cue robust enough to trigger the emergence of the defective tense? Occurrences of sentences like (3) in the Primarily linguistic (PLD) may not achieve the threshold necessary to change the grammar, specially because (3) has a twin sentence, mainly (4), which is also available in the PLD and cannot be a cue because it does not involve hyper-raising. (4) e parece que a Maria est? doente seem-3Sg that the Maria is-3Sg sick ?It seems that Maria is sick? (Lit. * He seems that is sick) Notice that there is a twist in the way the author sees the processes of language change and language acquisition. Saying that the emergence of a defective tense is morphologically based amounts to saying that at certain point in the history of BP, children piggybacked on the loss of morphology to cognize a grammar with a defective 119 tense. In a cue-based theory, this means that the loss of verbal morphology was the cue that triggered the system with the defective tense. Therefore, allowance of hyper-raising (3) is the consequence of the new system, not its cause. 4 A remaining question about this analysis is how we can test it, given that the alleged defective tense is present in a derivation only when subject-to-subject movement occurs. In order words, a defective finite tense does not occur in any other cases besides that it purports to explain. This means that the proposed analysis cannot be tested, as there is no conceivable linguistic phenomenon that would refute it. Therefore, it does match the criterion of falsifiability, which is required from scientific hypothesis (cf. Popper, 1963). 4.1 Implementing the Proposal 4.1.1 Verbal Agreement Morphology and Null Subject Pronouns Since Taraldsen (1980) there has been the intuitive idea that the licensing of referential null subject is somehow related to verbal agreement morphology . Recently it has been proposed that that in null subject languages with a rich agreement morphology, spec of TP/AgrP might not be realized by pro or any other material because the verbal agreement morphology carries a D feature, being, thus, able to satisfy the EPP features of these non-substantive categories. For developments of this idea, see Rohrbacher (1992), 3 For a discussion about robustness, see Clark (1992) and Lightfoot (1999). 120 Speas (1994), Platzack (1995), Alexiadou and Anagnostopoulou (1997, 1998), Vainikka and Levy (1998), Benamamoun (1999) and Kato (1999, 2000), among others. Alexiadou and Anagnostopoulou distinguish a two types of verbal agreement systems a weak and a strong one. In weak system, verbal agreement affixes (Agr, henceforth) are not impendent units the computational component, entering derivations already attaches to their host. A weak, therefore, is arguably not manipulatable by syntactic operations. In a strong system, instances of Agr are listed in the lexicon as separate lexical items, and enter the computational component as independent syntactic units that can be used by syntactic operations. It is particularly suggested that a strong Agr enters a derivation furnished with D-feature, f-features and perhaps with Case feature too. Hence syntactically, it is expected to function as a prononimal category. The authors were concerned mainly with null expletive constructions, and their suggestion is that in grammars with a strong agreement system, a null expletive pronoun is Agr itself. The assumption is that a strong Agr might be merged directly on the head of the Agreement phrase (AGRP), satsisfying, thus, the EPP feature of AGR. Hence, when the verb adjoins to AGR, AGR projects, but, since its EPP feature had already being satisfied, it does not project a specifier. Though they did not explore it, Alexiadou and Anagnostopoulou raised the possibility of replacing referential pro by Agr, presupposing that Agr counts as theta bearing argument in Null Subjects Grammars. As already shown in chapter 3 (cf. 3.5.2), Kato (1999, 2000) puts this possibility forward, suggesting that in Romance Pro-Drop languages, a sentence like (5a) is derived as shown in (5b), Agr enters the derivation as 4 Cf Rodrigues (2000), where I suggested that Children use the presence of null expletives in matrix clauses as a cue to the availability of A-movement. This suggestion accords to theory I offer here. 121 the external argument of VP. Being a maximal minimal projection, Agr adjoins to T 0 , where it checks the EPP features of T and its own Case feature. Thus, a projection of a specifier inside the TP is prevented. (5) a. viaj? traveled-3sg ?S/he traveled? b. [ TP [ T? [ T viaj 2 -[ Agr o] 1 + T][ VP t 1 t 2 ]]] As Kato observes, this analysis presupposes that DPs overt pronouns in subject positions are instances of doubled subjects in the sense that they double Agr, being generated at the left periphery of the clause, possibility receving a default Case. That is, according to this treatment overt subjects are left-dislocated DP, and Agr a resumptive subject clitic pronoun. 5 Assuming this analysis to be on right track and considering the correlation between loss of referential null subjects and weak verbal agreement morphology in BP and Finnish (cf. chapter 3), 6 I hypotheze that in these grammars Agr became f-defective, being unable to instanciate person&number distinctions (for discussion on BP, seeGalves 1996 and Kato 1999) and, as a consequence, was reanalyzed as part of verb, loosing its syntactic independence. However, I pressupose that BP and Finnish Agrs still have a D- 5 For evidence that Romance and Greek preverbal subjects surface at the left periphbery of the clause, see Barbosa (1995), Ord?nez (1997) and Alexiadou and Anagnostopoulou (1997), among others. 6 It is important to remark, though, that in my proposal, the f-features of Agr might not have phonological realization. Hence, Agr is an empty pronimal category. This will be proposed in chapter 6, where I will consider null possessors in Romance. 122 feature (or whatever turns out to be necessary to satisfy the EPP-feature of T). Hence, in these grammars, when V adjoins to T 0 , carrying Agr, Agr satisfies the EPP feature of T, but, being f-defective, is unable to delete the f-features of T. 7 The assumption that Agr is able to satify the EPP feature of T is motivated by the fact that null expletives, null quasi-arguments and null subjects with arbitrary reading are still allowed in present-day BP and Finnish, as presented in chapter 3. Thus, if there is no pro, in contructions like (6), the EPP feature of T ought to be satisfied by some other element, which I take to be Agr. (6)a. Tem muita coisa aqui nessa loja (BP) have-3Sg many things here in.this store ?There are many things in this store? b. N?ytt?? tuvevan sade (Finnish) look-3Sg come-lpart-acc rain-ACC ?It looks as if it is going to rain? If this is the right interpretation of the correlation between the loss of morphology and the loss of referential null subjects in BP and Finnish, it follows that in a finite clause with a referential subject, a f-complete item must be inserted in the complement domain of T, such that verbal q-role can be checked and the the f-features of T can be matched and deleted. But, the [spec, TP] need not to be projected, since Agr is able to satisfy the EPP feature. If this right, in the derivation of sentences like (7), the DP O 7 On the f-features of T, see Chomsky (2000). 123 Jo?o/Jukka needs to be first merged within the embedded clause, and then moved to the matrix clause. Moreover, this movement may not involve spec of TP. (7)a.Jo?o 1 disse que e 1 comprou um carro. (BP) the Jo?o said-3Sg that brought-3Sg a car ?Jo?o said that he will come tomorrow? b. Veljeni 1 oli niin iloinen ettei e 1 voinut nukkua (Finnish) brother-1 st Px be-3SgPres so happy that-not-3Sg could sleep-Inf my-Gen brother-Nom be-3SgPres so happy that-not-3Sg could sleep-Inf ?My brother was so happy that he couldn?t sleep? 4.2.2 Deriving Embedded Null Subjects by Movement If the empty subjects in (7) are the result of movement, then two questions arise: (i) How can a DP receive more than one q-role? (ii) How can a DP check more than one structural Case? The first question can be answered by assuming that q-roles are features and that there is no upper bound on number of q-roles a Dp can have. The second reason is more complex since it is standardly assume that a DP checks structural Case only once. I provide an answer to this question without overgenerating, I assume that Case- checking accords to (8): (8) a.A sructural Case feature is checked in a spec-head relation 124 b. Agrement in f-features prompts movement to check Case (8a) revamps Chomsky?s (1993) proposal, according to which, a DP could check its Nominative Case against a fuctional Category only if it had moved to the spec of that functional category, checking thus the EPP feature related to the functional head. For example, a DP checks Nominative Case against T only after having moved to spec of TP, checking the EPP feature of T. In his recent proposals (2000 and thereafter) however, Chomsky disassociates Case checking from EPP checking, and correlates Case with agreement in f-features. This correlation is expressed through the following technical mechanism: if a functional head H (a probe) has uninterpretable f-features, H probes its domain looking for a goal, an element with an identical (identical in the choice of the features, not in their values) set of f-features. As soon as the goal is found, the operation Agree takes place, deleting the f-features of H and the Case feature of the goal. If H also has an EPP feature, the goal is pied-piped and merged with HP. I maintain the correlation between agreement in f-features and Case checking, but I retain the ?old? idea that structural Case is checkable only in a spec-head configuration. To do so, I assume (8b): a goal can move to the specifier of a probe and check its Case feature only if they both agree in f-features. Intuitively, the idea is that Case is a feature that allows an element to enter a functional projection. What f-feature agreement does is to identify the element before ?opening the door?. To give a concrete example of how this works, consider the sentence in (10). The relevant step in the derivation of (9) is represented in (10a). John was merged with up, 125 checking the theta role of v. When T was inserted, T established agreement with John, and the f-features of T was deleted. But the Case feature of John was not deleted because it requires John to enter TP. The movement per se has already been authorized since T and John agreed in f-features. Note that the raising of John is also demanded by the EPP feature of T. When John moves to [spec TP], its Case feature and the EPP feature of T is checked, forming a convergent phrase marker (10b): (9) John left (10)a. [ TP [T f F, EPP ][ vP John q ,Case, f -F [ v ? left q ]]] b. [ TP John q , Case, f -F [ T? [T f F, EPP ][ vP t [ v ? left]]]] If the grammar works like this, the null subjects (7) are derived as exemplied below. The Deveriation of (7a), repetead here as (11), starts with the numeration in (12). (11) Jo?o 1 disse que e 1 comprou o carro Jo?o said-3 Sg that bought-3Sg the car ?John said the he bought the car? (12) Num = {T 2 , o 1 , Jo?o 1 , disse 1 , que 1 comprou 1 , o 1, carro 1 , v 1, } 126 The verb comprou merges with its internal argument, the DP um carro, which checks the q-role of the verb. This merge forms the embedded VP. When v is selected, V adjoins to v 0 . Next, the DP o Jo?o is built and merged with vP, checking the q-role of v. At this step of derivation of embedded vP (cf. (13a)) is formed. After that, T is selected and the complex [comprou+v] adjoins to T 0 , carrying Agr that has a defective set of f-features and a D-feature. Thus, Agr checks the EPP feature of T, but not the f-features, which probe the complement domain of T, looking for a goal. The DP o Jo?o in [Spec, VP] is localized and the agree operation takes place, deleting the f-features of T (13b): (13)a. [ vP [ DP o Jo?o] q , Case, f [ v? [ v [comprou f -def, D ] 1 +v] [ VP t 1 [ DP o carro]]]] b. [ TP [ T [ v [comprou f def,D ] 1 +v] 2 +T - f EEP ] [ vP [ DP o Jo?o] Case, f [ v? t 2 [ VP t 1 [ DP o carro]]]]] At this point of the derivation, the system can either move the DP o Jo?o to [spec, TP] and consequently check its Case feature or continue the derivation by merging the next item in the numeration, namely the complementizer que. The Case of the DP demands the movement, but, if the DP is moved, it becomes inactive for further computation, and the derivation will crash at the matrix level because there will be no item available to check the external theta role of the matrix v and the f-features of T. On the other hand, if the system continues the derivation by selecting and merging the complementizer que, delaying thus the deletion of the Case feature of o Jo?o, the derivation converges. At the 127 stage represented in (14), the numeration contains only a T. Hence, to check the matrix external theta role, the system must apply move, copying the o Jo?o and merging the copy with the matrix vP: (14) vP 3 t 3 v? t 5 VP t 4 CP 3 que TP 3 [[comprou f def,D ] 1 +v] 2 +T - f EEP ] vP [ v t 3 t 2 [VP t 1 [ DP o carro]]]] When the matrix T is inserted, the complex [disse+v] adjoins to T 0 , Agr checks the EPP feature of T, but fails to delete the f-features of T. These agree with the copy of o Jo?o in the matrix [spec, vP]. The DP o Jo?o is now pied-piped and merged with TP, checking its Case feature. The covergent phrase marker in (15) is sent to the interfaces. 128 (15) TP 3 [ DP o Jo?o Case, f ] 3 T? [v [disse f -def,D ] 4 +v] 5 +T f - EEP ] vP t 3 v? 3 t 5 VP t 4 CP que [ C? [ TP [ T? [ v [comprou f def,D ] 1 +v] 2 +T - f EEP ] [ vP t 3 [ v? t 2 [ VP t 1 [ DP o carro]]]]] Consider now a case in which an overt pronoun is inserted as the subject of the embedded clause, as in (16). (16) a. O Jo?o 1 disse que ele 1/2 comprou o carro Jo?o said-3 Sg that he bought-3Sg the car ?John said that he bought o car? The first relevant step of the derivation of (16) is the one in (17a). The embedded clause has been constructed and the pronoun ele was selected from the numeration and inserted as the external argument of v. When T entered into the derivation, the complex [ v [comprou + v]] adjoins to T 0 and Agr checks the EPP features of T, but not the f- features. Hence, T agrees f-features with ele and, as a result, the pronoun can move to [spec, TP] and check its Case. If Move does not apply at this point, the derivation would not converge. The Case feature of ele would remain unchecked since at the matrix level, 129 the DP o Jo?o is inserted as the subject. Therefore, the only way to ensure convergence is by moving the pronoun ele to [spec, TP], forming the CP represented in (17a). At the matrix level, o Jo?o is inserted in [spec, vP], getting the theta role assigned by v, and then moving to [spec, TP] to check its own Case feature, as depicted in (17b): (17)a. CP 3 que TP [ele q ,Case, f ] 3 T? [[comprou f def,D ] 1 +v] 2 +T - f EEP ] vP [ v t 3 t 2 [VP t 1 [ DP o carro]]]] b. TP 3 [ DP o Jo?o Case, f ] 3 T? [v [disse f -def,D ] 4 +v] 5 +T f - EEP ] vP t 3 v? 3 t 5 VP t 4 CP que [ C? [ TP [ele q ,Case, f ] 3 [ T? [ v [comprou f def,D ] 1 +v] 2 +T - f EEP ] [ vP t 3 [ v? t 2 [ VP t 1 [ DP o carro]]]]] In brief, the analysis proposes here accounts for the presence of null subjects in BP and Finnish by taking them to be formed via category movement. The loss of 130 referential referential null subjects in these grammars is due to the fact that Agr lost its status as an independent D item, being demoted to a verbal affix, conserving its D- feature, but having undergone degradation of f-features. Movement of the external argument of an embedded VP to the subject position of an immediately higher clause is possible because, in the current stage of BP and Finnish, V moves to T, carrying Agr, which checks the EPP-feature of T. Thecnically, it was suggested that that the system can delay the movement of a DP to the spec of a functional projection to check Case. 4.1.2 CP as a Phase and Intermediate Movement In the analysis proposed above, a DP can move directly from an embedded q- domain to a non-embedded one. But this movement is rather suspicious because it is too long and finite clauses are standardly taken to be opaque domains for extraction unless the movement proceds cyclically through spec of CP, as in wh-movement. Putting it in recent Minimalist terms, in (14), the DP-movement to the spec of the matrix vP should be considered illicit because the embedded CP is a phase (cf. Chomsky 2000,2001a, 2000b) and only the head and the border (specs) of a phrase are visible for computations outside that phase, in accordance with the Phrase Impenetrability Condition. (18) Phase-Impenetrability Condition(Chomsky, 2000: 108) In phase a with head H, the domain of H is not accessible to operations outiside a, only H and its edge are accessible to such computations. 131 One way to make the movement analysis proposed here compatible with the phase approach is to do the following: modify the left periphery of the embedded structure in (14) such that it can host an intermediate A-movement of the extracted DP. Second, refine the phrase-impenetrability condition with respect to extraction out of a left periphery Uriagereka (1988, 1995, and 1997) proposes that in Romance western languages TP is dominated by a functional projection FP. I will extend this proposal to BP and Finnish, suggesting that in these grammars subjects might move to spec of FP. 8 Hence, the left periphery of the embedded clause in (14) contains a FP and it is plausible that the DP o Jo?o moves to its specifier before moving to matrix clause, as represented in (19). (19) ? FP 3 [ DP o Jo?o Case, f ] 3 TP t 3 T? [v [disse f -def,D ] 4 +v] 5 +T f - EEP ] vP t 3 v? t 5 VP 3 t 4 CP que [ C? [FP t 3 [F? [ TP [ T? [ v [comprou f def,D ] 1 +v] 2 +T - f EEP ] [ vP t 3 [ v? t 2 [ VP t 1 [ DP o carro]]]]]]] 8 Raposo and Uriagereka (1995, 2002) assume that spec of FP is spell-out position of Romance preverbal subjects. 132 Chomsky?s formulation of phases is neutral on the possibility of having extra projections inside the CP domain. He assumes the abstract structure in (20), and suggests CP and vP are strong phases, whreas TP and VP are weak phrase (cf. 2000b). In addition, he proposes that a strong phase is evaluated and spelt-out when the derivation reaches the head of the next strong phrase. Therefore, in (20) when the computational system reaches the embedded vP, the material below C 0 (i.e. the TP phrase) is evaluated and spelled-out, being, therefore, not available for computations outside CP. That is, nothing below C 0 can be used to build the upper VP shell. (20) [vP [VP [CP [TP [vP [VP?]]]] In (19) above the embedded CP domain contains an extra projection and material immediately contained by this projection is available for computations outside the CP, so we need further clarification. As proposed in Lightfoot and Rodrigues (20003), the phrase-impenetrability condition can be defined in such way that inside the domain of a strong phase (HP) only sub-domains that are themselves phases are not accessible to operations outside HP. That means that when spell-out applies, it applies to phases, leaving behind elements of the structure that are not dominated by the head of the phase that is being spelled-out. If this is correct, in (20) the embedded TP is part of the domain of the head of the strong phase CP , and it is a phase; therefore, when the system reaches vP, CP is evaluated and TP, being a phase, is spelled-out. 133 If the spell-out process of phases is understood in this way, in (19), the movement of the DP o Jo?o from spec of FP to the spec of the matrix vP is not an issue. When the derivation reaches the matrix vP, CP is evaluated, and spell-out applies, sending TP, which is phase, to the interfaces, but not FP, because FP is not phase. This means that the material in FP is accessible for computation at the vP level. In summary, in this section I suggested that the A-movement under consideration in this thesis might procede in a cyclically way, though spec of FP, and left periphery Functional Category. This suggestion couple with Lightfoot and Rodrigues? interpretation of the spell-out process of phases suggests that the movement analysis defended here is ompatible with the concept of phases. In what follows, for the sake of the space, I will put this this intermediate movement aside, omitting the FP functional projection on the representations. Thus, from now on, the A-movement under consideration will be represented as if it were directly from one q-domain to another. 4.3 Evidence for a Movement Analysis This section provides evidence of movement. The main arguments are: (i) the null subjects under consideration have anaphoric behavior, requiring a sentential antecedent; (ii) the relationship between a null subject and its antecedent obeys the Minimal Link Condition, the antecedent being the closest c-commanding DP; (iii) In BP null subjects fail the resumption test, being unable to occur inside relative clauses. In Finnish null 134 subjects can occur inside relative clauses but only because the nominal head of the relative clause may undergo object shift; (iv) BP and Finnish null subjects display all the diagnostics used to characterize obligatory control as the residue of movement. Moreover in BP null subjects are not allowed in clauses after the connective como which is arguably a left branch, occupying a specifier positions inside of a DP (Torrego and Uriagereka 1995). 4.3.1 Anaphoric Behavior A crucial difference between the sentences below is that the null subjects are provided with an antecedent in (21), but not in (22): (21)*e estava cansado was-3Sg tired-Sg (22)O Jo?o 1 disse que e 1 estava cansado the Jo?o said-3Sg that was-3Sg tired-Sg ?Jo?o said that he was tired? Notice that in (22), the empty subjects must refer back to the matrix subject and cannot take a discourse referent. This anaphoric behavior is confirmed by the unacceptability of (23) in which a proper antecedent is not available: 135 (23)* pro expl parece que e tinha telefonado seem-3Sg that had-3Sg called ?It seems that s/he had called? In the Pro-drop Romance languages, represented here by European Portuguese, null subjects are free in reference. Thus, they can appear in matrix clauses, being interpreted as referring to an entity mentioned in the discourse. When they appear in embedded clauses (24), they are ambiguous in reference, being interpreted either as coreferential with a DP in a higher clause or as related to some person identifiable in the discourse. (24)O Pedro 1 disse que pro 1/2 ganhou na loto the Pedro said-3Sg that won-3Sg in.the lottery ?Pedro said that s/he won the lottery? (EP - Modesto, 2000) The 19th-century BP seems to behave like European Portuguese in that embedded null subjects are free in reference: (25)e creio que e teve aviso believe-1Sg that had-3Sg warning ?I believe that he was informed? (Martins Pena 1844, O Judas no S?bado de Aleluia) 136 Chomsky and Lasnik (1993) remark that a [+anaphor] empty category, i.e. NP- trace, gets its reference from an antecedent, while a [+pronoun] empty category, pro/PRO, can have reference on its own, although its reference may also be determined by an antecedent. Thus, in the Pro-drop Romance languages (24) and in 19th-century BP (25), null subjects behave as a [+pronoun] empty category (pro). In modern BP and Finnish, on the other hand, null subjects are identified as [+anaphor] empty categories, therefore as NP-traces, according to Chomsky and Lasnik?s classification. 4.3.2 Locality Matters Chomsky (1995b: 311) suggests that the definition of Move incorporates the property in (26), where closeness can be defined as in (27): (26)Minimal Link Condition (MLC) K attracts a only of there is no b, ? closer to K then a, such that K attracts b. (27) Closeness b is closer to K than a if b c-commands a, and b is not in the same minimal domain as t or a, where t is the target of raising. (Chomsky, 1995b: 335-6) These definitions were included because in BP the antecedence relationship 137 between a 3rdP null subjects and a DP obeys the MLC, the antecedent being the closest c-commanding DP. In (28a,b) for instance, the DP o Paulo cannot be the antecedents because they do not c-command the null subjects. In (29a,b), the matrix subjects fail to be the antecedents because the subjects of the higher embedded clause intervene: (28) [o pai do Paulo 1 ] 2 disse que e *1/ 2 vai ser promovido the father of.the Paulo said-3Sg that will be-Inf promoted ?Jo?o said to Paul that he will be promoted? (29)a. O Paulo 1 me contou que o Jo?o 2 disse que e *1/2 vai ser promovido the Paulo me told-3Sg that the Jo?o said-3Sg thar will-3Sg promoted ?Paulo told me that Jo?o said that he (Jo?o) will be promoted? Again, in the Pro-drop Romance grammars, when a null subject is interpreted as co- referential with a higher DP, the relation is not subject to the MLC. In (30a) and (30b), for instance, the null subject can take the DP o Pedro as its antecedent: 9 (30) a. [O amigo do Pedro 2 ] disse que pro 2 ganhou na loto the friend of.the Pedro said-3Sg that won-3Sg in.the lottery ?Pedro?s friend said that s/he won the lottery? 9 In European Portuguese, there is a strong preference for the readings in which the null subject takes the whole DP o amigo do Pedro in (30a) and the DP o Paulo in (30b) as its antecedent (Jo?o Costa, personal communication). As discussed in Calabrese (1986), this strong preference is also observed in Italian. Thus, the real contrast between Romance null subject languages and BP and Finnish seems to be the fact that in the former the co-indexation in (30a) and (30b) is possible, though dispreferred; while in the latter it is impossible. 138 b. O Pedro 1 disse que o Paulo acredita que pro 1 ganhou na loto the Pedro said-3Sg that the Paulo believe-3Sg that won-3Sg in.the lottery ?Paulo said that Pedro believes that s/he won the lottery? (EP - Modesto, 2000) The 19th-century BP is like European Portuguese in that the co-indexation between a null subject and a DP does not obey the MLC. 10 In (31), for instance, the null subject inside the adjunct clause is not c-commanded by its antecedent senhor, and the closest potential antecedent is the 1stP null subject of the matrix clause: (31) E do senhor pro queixo me, porque da primeira vez pro and of.the sir complain-1Sg me because of-the first time abusou da minha posi??o. abused-2Sg of.the mine position ?And I complain about you, sir, because the first time you took advantage of my position.? (Martins Pena 1844, O Judas no S?bado de Aleluia) The restrictions in (28) and (29) follow straightforwardly if the null subjects are formed by movement. The antecedent of the subject must be the closest c-commanding DP because the MLC is a locality condition on movement. (29a,b) is parallel to the super- 10 It may be worthy to mention that data from 19 th -century BP suggest that at that time there are at least two competing grammars in the sense of Kroch (1988): one was pro-drop and the other one was non-pro-drop. The non-pro-drop replaced the pro-drop at the beginning of the 20 th century. On this topic, see Lightfoot and Rodrigues (2003). 139 raising construction in (32). In (29a,b), the DPs o Paulo and Jukka cannot be the antecedents because movement to the matrix subject position would cross over the subjects of the intermediate clauses. (32) is ungrammatical for the same reason: the subject of the intermediate clause intervenes, blocking the movement of the DP John towards the matrix subject clause. (32) *John 1 seems that it is likely t 1 to win 4.3.2.2 Potential Interveners 4.3.2.2.1 Crossing Over a Subjectless Clause The comparison made with (32) above brings forward an intriguing property of the null subjects under discussion. As show in (33), a clause containing a raising predicate may in fact intervene between the null subject and its antecedent: (33)a.A Maria 1 me disse que parece que e 1 vai ser promovida the Maria me told-3Sg that seems-3Sg that will-3sg be-Inf promoted ? Maria told me that it seems that she will be promoted? b.Jukka 1 sanoi, ett? oli onni, ett? e 1 oli arpajaisissa Jukka 1 -Nom said-3sg that be-3sg fortune that pro 1 was-3sg won 140 voittanut lottery-In ?Jukka said that it was fortunate that he had won the lottery? Moreover, the raising verb might not agree in j-features with the matrix subject (cf (34)). This suggests that that the movement from the inner to the upper subject position may not involve the sentential subject position of intermediate clause. (34) *? Os meninos 1 me falaram que parece que e 1 v?o ser promovidos the boys-Pl me told-3Pl that seem-3Sg that will-3Pl be-Inf promoted ?The boys told me that it seems that they will be promoted? If the movement analysis I am proposing is on the right track, the structure of (2) involves a long movement, as represented in (3). The DP os meninos is base generated inside the lowest clause and moves directly to the matrix clause, crossing over the upper embedded clause. (35) [[os meninos] me falaram [ CP que parece [ CP que t v?o ser promovidos]]] the boys-Pl me told-3Pl that seem-3Sg that will-3Pl be-Inf promoted This analysis poses a fundamental question about the sentential subject position of raising predicates. If there is a phonetically null expletive occupying this position, in 141 (35) the movement of os meninos ?the boys? would violate Minimality. The expletive occupies a Case Position, hence a position in which the DP under movement could have checked its Case feature had it not been taken by the expletive. On the other hand, if null expletives in BP and Finnish (as well in grammars like Romance and Greek) does not exist as independent lexical items, as we hypothezed in 4.2.1, then there is no intervening Case position in (35) and, consequently, the long movement represented does not violate Minimality. According our hypothesis, Agr are BP and Finnish are able to satify the EPP feature of TP. Thus, in (35) the spec of the intermediate TP is not projected, and the DP under movement does not cross over a specified subject. The sentential EPP feature of intermediate TP is satisfied by Agr, which is assumed to have a D-feature. 4.3.2.2.2.Referential Expressions Blocking Movement As I briefly showed at chapter 3 (3.5.2.1), in Brazilian Portuguese, a preverbal locative adverb has the effect of blocking coreference between a null subject and a higher DP. In (36), for example, since the locative adverb na praia ? on the beach? intervenes, the embedded clause is understood as a generic statement with a missing subject. Notice that if we demote the locative adverb to a lower adjunct position, then coreference becomes possible, as shown in (37): 142 (36) Jo?o me contou que na praia *(e 1 ) vende cachorro quente Jo?o me told that at.the beach sell-3Sg dog hot ?Jo?o told me that hot dogs are sold at the beach ? # ?Jo?o told me that he sells hot dogs at the beach? (37)Jo?o me contou que e 1 vende cachorro quente na praia Jo?o me told that sell-3Sg dog hot at.the beach ?Jo?o told me that he sells hot dog at the beach? # ?Jo?o told me that hot dogs are sold at the beach? Vainikka ad Levi (1999), citing Kakulinen (1976), show that Finnish displays the same blocking effect. While in (38) the null subject is coreferent with the matrix subject, in (4) the coreference is blocked and the embedded clause is interpreted as a generic statement. The only difference between (38) and (39) is the left dislocation of the direct object in (39): (38)Oppilas 1 tiet?? ettei e 1 pysty ratkaisemaan teht?v?? student-Nom know-3Sg that.not-3Sg can-3Sg solve-Inf assignment-Par ?The student knows that he cannot solve the assignment? # ?The student knows that one cannot solve the assignment? 143 (39) Oppilas 1 tiet?? ettei teht?v?? pysty ratkaisemaan student-Nom know-3Sg that.not-3Sg assignment-Par can-3Sg solve-Inf ?The student knows that one cannot solve the assignment? # ?The student knows that he cannot solve the assignment? The contrast in (36) and (37) indicates that locative adverbs, like a displaced verbal complement (38) and (38) can block A-movement out an embedded finite clause. I will take these elements to be in spec FP. Therefore, if CP is a phase, in the examples above, the matrix DP subject couldn?t be extracted from inside the embedded clauses because the scape hatch from the CP domain, viz. spec of FP, is already occupied by another consitituent. Note that only referential expression is able to create the blocking effect under discussion. In (40), for instance, a temporal adverb is placed in a preverbal position, but its does interact with movement of the DP o Jo?o. (40)O Jo?o 1 me contou que todos os dias e 1 vende cachorro quente na praia the Jo?o me told that every the days sell-3Sg dog hot at.the beach ?Jo?o told me that he sells hot dogs at the beach every day? 144 4.3.3 Obligatory Control Properties As shown in section 2.4, Hornstein (1999, 2001) proposes that obligatory control configurations (OC) are formed by movement. For expository purposes, I repeat here the core of his proposal. The sentence in (41a) has the structure in (41b). The DP John is first merged as the external argument of the verb win. In this position, John receives the winner q-role. After that, John moves to the spec of the embedded infinitival TP to check the relevant EPP feature. When the matrix VP is built, the same DP moves to spec of the matrix clause to check the external q-role of hope. Finally, a TP is projected on the top of the structure and John moves spec of TP, where it checks its Case feature. (41) a. John hopes to win b. [ TP John q , q Nom [ T? [ VP John q , q Case] [hopes [ TP [John q ,Case to [ T ? [ VP John q ,Case win]]]]]]]] I will not go into details here (see 2.4.3), but it is worth recapitulating Horsntein?s evidence for a movement analysis. In cases of obligatory control, (a) the controlled element is anaphoric, taking the closest c-commanding NP as its antecedent, (b) split antecedents are not allowed, (c) a de se reading is forced and (d) if the closest c- commanding NP has the format only-NP, then only a covariant interpretation is possible. In what follows, I show that modern BP and Finnish 3rdP embedded null subjects display all these properties: 11 11 For a discussion about OC and the loss of inflected infinitives in Modern BP, see Pires (2001). 145 First, as already shown, these null subjects require a sentential antecedent, which must be the closest c-commanding DP. (42) a. * Parece que e saiu seem-3Sg that left-3Sg ?It seems that he left? b. [O pai do Paulo 1 ] 2 disse que e *1/2 vai ser promovido the father of.the Paulo said-3Sg that will-3Sg be-Inf promotted ?Paulo?s father said that he will be promoted? c. O Paulo 1 sabe que o Jo?o 2 disse que e *1/2 tinha de sair the Paulo know-3Sg that the Jo?o said-3Sg had-3Sg of leave-Inf ?Paulo knows that Jo?o said that he had to leave? (43)a.* Vaikuttaa silt? ett? e oli oittanut seem-3Sg it that had-3Sg called ?It seems that s/he had called? b.[Veljeni 1 vaimo] 2 oli ninn iloinen, ettei e *1/2 voinut brother-Gen1stPx wife-Nom was-3Sg so happy that-not-3Sg could nukkua sleep-Inf ?My brother?s wife was so happy that she couldn?t sleep? ( Vainikka and Levi, 1999) 146 c.Jukka 1 sanoi ett? Pekka 2 ajattelee etta e *1/2 oli voittanut Jukka-Nom said-3Sg that Pekka-Nom think-3Sg that had-3Sg won arpajaisissa lottery-In ?Jukka1 said that Pekka2 thinks that he (Pekka) had won the lottery? Figueiredo Silva (1996) reports a prohibition against split antecedents in modern BP (44), and Gutman (1999) and Vainikka and Levi (1999) report the same prohibition in Finnish (45). (44)*A Maria 1 disse que o Jo?o 2 acredita que e 1+2 v?o morar juntos the Maria said-3Sg that the Jo?o believe-3Sg that will-3Pl live-Inf together ?Maria said that Jo?o believes they will live together? (45)*? Pekka 1 kysyi vaimoltaan 2 voiovatko e 1+2 menn? Espanjaan Pekka-Nom asked-3Sg wife-Abl-3Px can-3Pl-Q-Cl go-Inf Sapin-Ill Lomalle vacation-All ?Pekka asked his wife if they can go to Spain for a vacation? Notice that the unacceptability of (44) and (45) contrasts with the acceptability of (46) and (47), in which an overt pronoun was inserted in the embedded subject position. 147 Hence, overt pronouns allow split antecedents, but null subjects do not: (46) A Maria 1 disse que o Jo?o 2 acredita que eles 1+2 v?o morar juntos the Maria said-3Sg that the Jo?o think-3Sg that they will-Pl live-Inf together ?Maria said that Jo?o believes they will live together? (47)Pekka 1 kysyi vaimoltaan 2 , viovatko he 1+2 menn? Espanjaan Pekka-Nom asked-3Sg wife-Abl-3Px can-3Pl-Q-Cl they-Nom go-Inf Sapin-Ill lomalle vaction-All ?Pekka asked his wife if they can go to Spain for a vacation? Negr?o (1999) shows in BP, that under VP ellipsis, only a sloppy reading is available for an embedded null subject: (48) A Maria 1 encucou que e 1 estava gr?vida e o Paulo tamb?m the Maria worried-3Sg that was-3Sg pregnant and the Paulo too (sloppy/*strict/*deictic) ?Maria got worried that she was pregnant and Paulo did too? (48) is anomalous since the elided part is interpreted as Paulo1 got worried that he1 was pregnant. Note that if an overt pronoun is inserted, the strict reading becomes available. 148 In (49), for example, the elided part can receive a strict reading, being interpreted as Paulo got worried that she, Maria, was pregnant. The pronoun ela ?she? can also be interpreted as referring to a person not mentioned in the sentence (the deictic reading): (49)A Maria encucou que ela estava gr?vida e o Paulo tamb?m the Maria got.worried-3Sg that she was-3Sg pregnant and the Paulo too ?Mary got worried that she was pregnant and Paulo did too? The anomaly observed in (48) is also observed in raising constructions as shown below: (50) A Maria 1 parece e 1 estar gr?vida e o Paulo tamb?m the Maria seem-3Sg be-Inf pregnant and the Paulo Paulo too ?Maria seems to be pregnant and Paulo does too? The same interpretative restriction seems to hold in Finnish. In (51a) only a sloppy reading is possible, whereas (51b) allows a strict or a deictic interpretation: (51)a. Jukka 1 sanoi ett? e 1 oli voittanut arpajaisissa, ja niin Jukka-Nom said-3sg that had-3sg won lottery-IN and so Pekkakin Pekka-also ?Jukka said that he (Jukka) had won the lottery and so did Pekka? b.Jukka sanoi ett? h?n oli voittanut arpajaisissa, ja niin 149 Jukka-Nom said-3sg that s/he had-3sg won lottery-IN and so Pekkakin Pekka-also ?Jukka said that s/he had won the lottery and so did Pekka? With respect to de se interpretation, the embedded null subjects of BP and Finnish behave exactly like an obligatory controlled empty category. For instance, in the context given in (52), the BP statement in (53) is false. For (53) to be true, it demands a de se belief (cf. Casta?eda 1966, Salmon 1986 and Chierchia 1990). That is, Ronald Reagan must be aware that it is himself that he believes to be the fortieth president of the United States. Using Cole?s at al. (2001) way to put it, we say that under a de se reading the protagonist of the event actually ascribes, or is disposed to ascribe, to himself/herself the property denoted by the predicate containing the empty category or anaphoric element. (52) Because of his Alzheimer?s disease, Ronald Reagan cannot remember who he was. One day, reading the newspaper, he read the headline ?Reagan was the fortieth President of the United States?. After having finished reading the article, Ronald Reagan comes to know that the person called Reagan was the fortieth President of the United States. (53)O Ronald Reagan 1 sabe que e 1 foi o quadrag?ssimo presidente the Ronald Reagan know-3Sg that was-3Sg the fortieth president 150 dos Estados Unidos of.the States United ?Ronald Reagan knows that he himself was the fortieth president of the United Stated ? Although subtle, there is a contrast between (53) and (54) below. (54) is ambiguous. The pronouns ele ?he? might refer back to Ronald Reagan, forcing us to attribute a de se knowing to this entity, and judge the statement as false given the context in (52). However, the pronoun might refer not to Ronald Reagan, but to Reagan, the name in the headline. Following Higginbotham (1992), I indicate the concept of Reagan as [?the person whom Ronald Reagan is reading about?]. If ele refers to Reagan, a de se reading does not emerge, and (54) is true even in the context given in (52). (54)O Ronald Reagan sabe que ele foi o quadrag?ssimo presidente the Ronal Reagan know-3Sg that he was-3Sg the fortieth president dos Estados Unidos of.the States United ?Ronald Reagan knows that he was the fortieth president of the United Stated ? In this sense, null subjects patterns not with overt subject pronouns, but rather with emphatic reflexive forms like ele mesmo ?he himself?. In (55), for instance, the subject of the embedded clause is filled out with the emphatic reflexive form ele mesmo 151 and the statement is unambiguously false if (52) is taken into consideration. (55) Ronald Reagan sabe que ele mesmo foi o quadrag?ssimo the Ronald Reagan know-3Sg that he himself was-3Sg thefortieth presidente dos Estados Unidos president of.the States United ?Ronald Reagan knows that he himself was the fortieth president of the United Stated ? The contrast between a null and an overt subject pronoun becomes acute in (56) and (57) (cf. Chierchia 1990). (56) is a contradiction, whereas (57) is not. In (56) the null subject induces a de se reading, which is contradicted by the content of the conjoined sentence. The contradiction disappears in (57) because the overt subject pronoun does not force a de se reading. (56)# O Ronald Reagan sabe que e foi o quadrag?ssimo presidente the Ronald Reagan know-3Sg that was-3Sg the fortieth president dos Estados Unidos, mas o Ronald Reagan n?o sabe que ele mesmo of.the States United. But the Ronald Reagan not know-3Sg that he himself 152 foi o quadrag?ssimo presidente dos Estados Unidos was-3Sg the fortieth president of.the States United ?Ronald Reagan knows that he himself was the fortieth president of the United States, but Ronald Reagan does not know that he himself was the fortieth president of the United States? (57) O Ronald Reagan sabe que ele foi o quadrag?ssimo presidente the Ronald Reagan know-3Sg that he was-3Sg the fortieth president dos Estados Unidos, mas o Ronald Reagan n?o sabe que ele mesmo of.the States United But the Ronald Reagan not know-3Sg that he himself foi o quadrag?ssimo president dos estados Unidos was-3Sg the fortieth president of.the States United ?Ronald Reagan knows that he was the fortieth president of the United States, but Ronald Reagan does not know that he himself was the fortieth president of the United States? Traditionally attitude report verbs are taken to express relations between agents and propositions (cf. Salmon 1986). 12 Hence, in the sentence above, the matrix verb saber ?know? express a knowing relation between Ronald Reagan and the proposition ?x was the fortieth president of the United States?. Therefore, if an overt pronoun is inserted as the subject of the embedded sentence, the x in the proposition can be valued either as 12 For a different view about propositional attitudes, see Moltmann (2003), who argues that these types of verbs do not express a 2-place relation between an agent and a proposition, but rather an n-place relation between an agent and the constituents of the proposition. 153 Ronald Reagan or as Reagan. If it is valued as Ronald Reagan, then Ronald Reagan is in a knowing relation with a proposition about himself; therefore, this entity is in a de se knowing relation. Conversely, if x is value as Reagan, Ronald Reagan is not in a de se knowing relation for he knows a proposition that is about somebody else, mainly the person whom he read about. Having said that, the question is about (53). Why is it that the presence of a null subject forces us to attribute a de se knowing to Ronald Reagan? The movement analysis proposed here gives us a straightforward answer. In the structure of (53), the embedded sentence is Ronald Reagan foi o quadrag?ssimo presidente dos Estados Unidos ?Ronald Reagan was the fortieth president of the United States?, where Ronald Reagan is a copy of the subject of the matrix clause. Therefore, (53) asserts that Ronald Reagan is a knowing relation with the proposition ?Ronald Reagan was the fortieth president of the United States?. In other words, Ronald Reagan is a knowing relation with a proposition about himself. 13 Let me turn now to another property of obligatory control configurations: the absence of an invariant interpretation for sentences containing only-NPs. Higginbotham (1992) observes that the co-indexation of the pronoun he in (58) is ambiguous, it either takes John as its antecedent (59a), or the antecedent is the whole DP only John, as in (19b). (58) Only John expects [that he will win] (59) a.[Only [John] 1 ] 2 expects [that he 2 will win] 154 b. [Only [John 1 ]] 2 expects [that he1 will win] The author remarks that the given co-indexations lead to two different readings. (59a) is assigned the covariant interpretation in (60a), whereas (60b) gets the covariant interpretation in (60b) (60) a.Only John is an x such that x expects x will win (Covariant interpretation - (59a)) b.Only John is an x such that x expects he, John, will win (Invariant interpretation - (59b)) Negr?o (1999) pointed out that the BP sentences in (61) receive different interpretations. While (61a) is assigned the covariant interpretation in (62a), (61b) is assigned the invariant interpretation in (62b): (61) a. S? o Maluf acha que e vai ganhar as elei??es only the Maluf think-3Sg that will-3Sg win-Inf the elections b.S? o Maluf acha que ele vai ganhar as elei??es only the Maluf think-3Sg that he will-3Sg win-Inf the elections ?Only Maluf thinks that he win the elections? (62) a.Only Maluf is an x such that x thinks that x will win the elections 13 I failed to test the de se reading in Finnish. 155 b. Only Maluf is an x such that x thinks that he, Maluf, will win the elections Therefore, we reach the conclusion that in BP a 3rdPerson null subject cannot be co-indexed with the NP contained in the only-NP phrase. It must be co-indexed with the whole only-NP phrase, as represented in (63a). Conversely, an overt pronoun must be co- indexed with the NP contained in the only-NP phrase, as in (63b). According to Higginbotham the restrained co-indexation in (63a) is responsible for the covariant reading in (62a), whereas the restrained co-indexation in (63b) leads to the invariant reading in (62b). (63)a.[s? [o Maluf] 1 ] 2 acha que e *1/2 vai ganhar as elei??es (62a) only the Maluf think-3Sg that will-3Sg win-Inf the elections b.[s? [o Maluf] 1 ] 2 acha que ele 1/*2 vai ganhar as elei??es (62b) only the Maluf think-3Sg that he will-3Sg win-Inf the elections As already pointed out, (cf. fn. 18, section 3.3.2) BP grammars generally obey Montalbetti?s constraint, prohibiting co-indexation between an overt pronoun and a quantifier-like expression. Therefore, if the phrase only-NP is a quantified NP, the impossible co-indexation in (63b) might fall within the realm of the Montalbetti?s Constraint. The interaction between Montalbetti?s Constraint and impossibility of a covariant 156 interpretation for a full pronoun in only-NPs contexts can be tested. As I mentioned earlier (fn. 18, section 3.3.2), in some dialects of BP the quantifier expression todo-NP ?every-NP? allows co-indexation with an overt pronoun. Therefore, if todo-NP can violate Montalbetti?s constraint, it can be co-indexed with an overt pronoun, triggering, thus, a covariant interpretation. The ambiguity of (64) shows that this is indeed true. (64), similarly to the English sentence in (58), can receive either the covariant (65a) or invariant (65b) interpretation. (64) [toda f? da [Carla Peres] 1 ] 2 acha que ela 1/2 deve agir como esposa do Xande every fan of.the Carla Peres think-3Sg that she have-3Sg act-Inf as wife of Xande ?Every fan of Carla Peres thinks that she has to behave as Xande?s wife? (65)a."x (x = a fan of Carla Peres) x thinks that x has to behave like Xande?s wife b. "x (x = a fan of Carla Peres) x thinks that she, Carla Peres, has to behave as Xande?s wife In sum, with respect to the co-indexing relations between an overt subject pronoun and the terms of a only-NP, BP is different from English only in that the former, but not the latter, is subject to Montalbetti?s constraint. Having discussed the non-occurrence of a covariant interpretation in (61b), we should now ask ourselves about the impossibility of an invariant interpretation for (61a). 157 (61a), repeated here (66a) allows only the co-indexation in (66b) (= (63a)). (66)a. S? o Maluf acha que e vai ganhar as elei??es only the Maluf think-3Sg that will-3Sg win-Inf the elections ?Only Maluf thinks that he win the elections? b. [s? [o Maluf] 1 ] 2 acha que e *1/2 vai ganhar as elei??es According to the proposal I am defending here, the gap in (66) is not a null pronoun, but a salient copy of its antecedent. Therefore, the antecedent cannot be the DP o Maluf because this DP does not c-command anything outside the containing phrase s? o Maluf. Conclusion, the antecedent, i.e. the moved element in (66), must be the whole quantifier phrase s? o Maluf. Finnish is not subject to Montalbetti?s constraint, as shown in (67). 13 (67)a. Kukaan 1 ei luule, ett? h?n 1 on fiksu anybody-Nom not-3S believe that he is-3Sg smart ?Nobody thinks that he is smart? 13 Tor Ashan (personal communication) pointed out me that the sentence (67a) is unusual in Finnish. Instead of a full finite embedded clause, native speakers prefer a participial embedded clause, as exemplified in (i). (i) Kukaan1 ei luule olevansa1 fiksu anybody-Nom not-3Sg believe be-Part.3Sg smart ?Nobody thinks that he is smart? 158 b.Joka tytt? 1 luulee, ett? on totta, ett? h?n 1 on every-Nom girl-Nom think-3Sg that is-3Sg truth-Part that she is-3Sg t?ydellinen perfect ?Every girl thinks that it is true that she is perfect? Therefore, (68a) can receive either an invariant or a covariant interpretation; the pronoun h?n/se ?he? being co-indexed either with Jukka or with vain Jukka. In (68b), as predicted by the movement analysis proposed, only a covariant interpretaiton is possible: the gap in the embedded subject position must take the whole quantifer phrase vain Jukka as its antecedent. (68) a. [vain [Jukka 1 ]] 2 ajatteli ett? h?n(se) 1/2 oli voittanut arpajaisissa only Jukka-Nom though-3sg that he was-3sg won lottery-In (?Covariant/?Invariant) b.[vain [Jukka 1 ]] 2 ajatteli ett? e *1/2 oli voittanut arpajaisissa only Jukka-Nom though-3sg that be-PAST-3sg won lottery-In (?Covariant/ *Invariant) ? Only Jukka thinks that he had won the lottery? 159 4.3.4The Connective Como in Brazilian Portuguese Davidson (1968) suggests that in (69a) the embedded sentence is only paratactically dependent on the matrix clauses. For him, (69b) is the underneath structure of (1a). The clause the earth moves, is an independent clause cataphorically related to that. (69) a.Galileo said that the earth is round b. Galileo said that 1 [the earth moves] 1 In Chomskyan syntactic theory, however, paratactic dependencies are generally not considered. (69), for example, is commonly treated as a case of hypotactic dependency; the clauses the earth moves being part of the syntactic complement of the verb said. Torrego and Uriagereka (1995) defend that both types of sentential dependencies are realized by UG. Their argument is that there are two types of finite connectives in Romance languages that are actually instances of hypotaxis and parataxis. (70) illustrates occurrences of these two connectives in Spanish: (70) a.No veras que diga la verdad jamas (Hypotaxis) not see-Fut-2Sg that say-3Sg the truth never ?You will see that s/he never says the truth? 160 b. Veras como tu madre llevaba razon (Parataxis) see-Fut-2Sg how your mother took-3Sg right ?You will see how your mother was right? Their analysis predicts that, while the computational system can produce syntactic dependencies across que, syntactic dependencies across como are not possible because the clause after como is not syntactically connected to the chief phrase marker, remaining as a separate text throughout the derivation. This prediction is borne out, as the authors show. Dependencies such as wh-movement (71a), predicate raising (71b), neg-raising (71c), licensing of polarity items (71d), and bound variable binding (71e) are fine with the connective que, but not with the connective como. (71)a. Qu? os ense?? que/*como estaba escribiendo what you-cl showed-3Sg that/ how was writing ?What did s/he show to you that/* how s/he was writing? b.A punto de llorar vieron que/*como estaba up of cry-Inf saw-3Pl that/how was-3Sg ?Up to crying, they saw that/*how s/he was? c. No ver?s que/*como diga la verdad jam?s not see-Fut-3Sg that how say-3Sg the truth never ?You will see that/*how she never says the truth? 161 d. No ver?s que/*como venga bicho viviente not see-2Sg that how arrive-3Sg bug living ?You won?t see a soul coming? e. Nadie ve como es tonto nobody see-3Sg how is-2Sg stupid ?Nobody sees how he is stupid? The behavior of the BP como is similar to its Spanish counterpart. As shown in (72), all the restrictions illustrated in (71) are observed in Brazilian Portuguese (cf. (72), expect for predicate and neg-raising, which, according to my own judgements, are unacceptable independently of the connective chosen. (72) a. [o que ] 1 a Maria n?o ? capaz de ver que/*como ela est? fazendo what the Maria not is-3sg able of see-Inf that how she is-3Sg doing errado t 1 wrong ?What is Mary not able to realize that she is doing wrong? b. Ele nunca vai ver que/*?como nada est? acontecendo na he never will-3Sg see-Inf that how nothing is-3Sg happening in-the vida dele life of-he ?He will never see that nothing is going on in his life 162 c.Ningu?m v? que/*como ? parte do sistema nobody see that how is-3sg part of.the system ?Nobody sees that he is part of the system? But, there are two striking differences between BP and Spanish: First, while in Spanish the sentence after como cannot be introduced by an overt complementizer, in BP the presence of an overt complementizer does not alter the acceptability of the sentence. (73)a. *Juan explic? como que la tierra es redonda (Sp.) Juan explained-3Sg that the the earth is-3Sg round (Torrego and Uriagereka, 1995) b.Voc? vai ver como que a Maria tinha raz?o (BP) you will-3Sg see how that the Maria had-3Sg right ?You going to see that Mary was right? Second, in Spanish the clause following como can contain a 3rdP null subject that might be co-indexed with the subject of the main clause (74a). This is impossible in BP. (74b) can have an overt subject pronoun, co-indexed or not with the matrix subject, but not a null subject. (74)a.Con el tiempo, Mar?a 1 ver? como pro 1/2 tiene raz?n (Sp.) with the time, Mar?a see-3SgFut how has-3Sg right 163 b.Com o tempo, a Maria 1 vai ver como ela 1/2 /*e 1 tem raz?o (BP) with the time, the Maria will-3Sg see-Inf how she has-3Sg right ?With time, Maria will see how s/he is right? The difference concerning the presence of an overt complementizer can be easily explained if we accept Torrego and Uriagereka?s analysis for the connective como and also grant the hypothesis that BP does not license referential null subjects anymore. So, it is necessary to get acquainted with the details of Torrego and Uriagereka?s proposal. Since como is etymologically derived from the Latin quod mod, Torrego and Uriagereka entertain the possibility of treating como as a bimorphemic category, composed by a D item, instantiated by the morpheme co-, and a predicate represented by ?mo. Therefore, como is taken to be a DP. Particularly, it is proposed that the internal structure of como is similar to the structure of possessive DPs suggested by Kayne (1993) and Szabolcsi (1983, 1994), proposed a layered DP in the modes of (75a). The representation below capitalize the suggested similarity between these types of DPs. (75b) is an enriched version of (75a) because of the predicational nature of ?mo: 14 14 On this matter, see also Hornstein at al. (1994). 164 (75) a.Possessive DPb. ver?s como tu madre llevaba raz?n ?you will see how your mother was right? DP ? DP 3 3 D? D? 3 3 D AgrP co- AgrP 3 possessor Agr? Agr? 3 Agr Possessed Ag XP CP XP -mo [tu madre llevaba raz?n] Torrego and Uriagereka take (75b) to be an LF representation however. For them, in the overt syntax (75b) looks rather like (76), in which the clause tu madre llevaba raz?n is not merged, remaining as an independent matrix clause. This is possible because a null pronoun (pro), cataphorically related to the non-merged clause is inserted in the spec of XP, moving later to spec of AgrsP. (76) ? DP 3 D? co-mo AgrP pro 1 Agr? Agr XP 3 t 1 XP -mo [ CP tu madre llevaba raz?n] 165 At LF the separated clause reconstructs into the pro-site, as in (74b). Thus, the clause after como is associated with the main structure in a paratactic way. This explains why syntactic dependencies across como are not permitted. According to the authors, the impossibility of inserting an overt complementizer on the top of the separate clause follows from Chomsky?s (1995b) conjecture that an overt complementizer cannot be inserted on the top of a matrix clause because lexical insertion after spell-out violates the Extension Condition. It is also suggested that insertion of pro in the overt syntax is necessary on grounds of economy. If the separate clause in (76) is not overtly merged, insertion of an overt complementizer is avoided. Hence, insertion of pro is less expensive than insertion of the clause. 15 If this is right, we need now an explanation for why an overt complementizer is possible in BP. I will defer the discussion about null possessors until chapters 5 and 6, where I suggest that in BP referential empty pronouns are not licensed as the external argument of possessive description. For now, let me present it as hypothesis. This hypothesis explains the allowance of an overt complementizer after como in (73b) in the following way: If null pronouns are banned altogether in this language, it follows that the external argument of the predicate ?mo cannot be null. That is, this language cannot use the strategy used in (76) in order to avoid overt insertion of a full clause headed by an overt complementizer. Therefore, in BP the clause that is represented as a separate clause in (76) is actually overtly merged within XP. It amounts to saying that in this language the connective como may not create a paratactic dependency. The clause after como is connected to the main 15 This falls under the Avoid Pronoun Principle (cf. 3.3.2) if we take this principle to be restriction on the insertation of material with phonological content. 166 structure in a hypotactic fashion. As a consequence, an overt complementizer is allowed. Now, if this is right, we need an account for the the facts in (77), which show that, despite the present of an overt complementizer, syntactic dependencies across como are not allowed in BP: (77) a. * [o que ] 1 a Maria n?o ? capaz de ver como que ela est? fazendo what the Maria not is-3sg able of see-Inf how that she is-3Sg doing errado t 1 wrong b. * Ele nunca vai ver como que nada est? acontecendo na he never will-3Sg see-Inf how that nothing is-3Sg happening in-the vida dele life of-he c. ?? Ningu?m v? como que sua mem?ria est? ficando fraca nobody see-3Sg how that his memory is-3Sg getting weak The wh-extraction movement in (77a) is occluded because the embedded clause occupies an specifier position (spec of AgrP) within a DP. Hence it is within an island which is itself embedded within another island. As for the non-licensing of polarity items and the impossibility of a bound variable binding (77b,c), they might as well be a reflex of the fact that the clause containing these items are embedded within a DPs. (78a) shows that a negative polarity item embedded within a complex noun phrase is not licensed by 167 the matrix negation. The same seems to be valid for bound variables. If they occur inside a complex nominal, they can hardly be bound by an external, as (78b) suggests. (78)a.* Eu n?o ouvi o boato de que o Jo?o comprou nada I not heard-3Sg the rumor of that the Jo?o bought-3Sg nothing ?I did not hear the rumor that Jo?o bought nothing? b.?? Ningu?m ouviu o boato de que a sua av? morreu nobody heard-3Sg the rumor of that the his grandma died-3Sg ?Nobody heard the rumor that his grandma died? Let me know return to the contrast in (74), repeated here as (79). This shows that whereas Spanish allows the subject of the independent clause to be null, in BP it is not possible, even though the clause after como is syntactically part of the main structure. (79)a.Con el tiempo, Mar?a1 ver? como pro1/2 tiene raz?n with the time, Mar?a see-3SgFut how has-3Sg right b.Com o tempo, a Maria 1 vai ver como que ela 1/2 /*e 1 tem with the time, the Maria will-3Sg see-Inf how that she has-3Sg raz?o right ?With time, Maria is going to see how s/he is right ? 168 (79a) is not surprising, since Spanish is a pro-drop language and, as such, allows referential null subject in matrix clauses. (79b) also is expected because 3rdP null subjects in BP are the result of NP-movement, and the embedded domain in (79) is a island for extraction (cf. (77a). Summarizing, the behavior of constructions involving the connective como in BP provides two arguments for the proposal of this thesis. First, Modern BP is not a pro-drop language, therefore the paratactic structure proposed by Torrego and Uriagereka may not be possible. Second, BP 3rdP referential null subjects are formed by movement; as a result, they are not licensed inside a clause that follows the connective because this clause is an opaque domain for subextractions. 4.3.5Referential Null Subjects Inside Relative Clauses The unacceptability of the Spanish sentence in (80), in which the direct object was extracted, shows that relative clauses are strong islands in Romance languages. But as observed in Rizzi (1982), Chomsky (1981) and Jaeggli and Safir (1989), these languages have a resumption strategy that salvages subject extraction from relative clauses via insertion of a resumptive pro, as illustrated in (81): 16 (80)* Qu? 1 bes? Juan a la chica 2 que t 2 compr? t 1 what kissed-3Sg Juan to the girl that bough-3Sg ?What did Juan kiss the girl that bought ? 16 Since Italian null objects receive only an arbitrary reading (Rizzi, 1986), the analogue of (81) is also unacceptable. 169 (81) a. Ese es el tipo que 1 Mar?a conoce a la mujer [con quien pro 1 se cas?]] this is the guy that Maria know-3Sg to the woman with whomSE maried-3Sg ?That is the guy that Mar?a knows the woman who he married? (Jaeggli and Safir, 1989) b. Ecco la ragazza [che 1 me domando [chi crede [che pro 1 possa VP]]] this the girl that me ask-1Sg who thinks-3Sg that can-3Sg ?This is the girl who I wonder who thinks that she may ?? (Chomsky, 1981) By comparing Spanish and Italian with BP, we can have extra proof that null subjects in modern BP do not behave like null subject pronouns. In (82), the empty subject fails the resumption test. As in English, only an overt pronoun can be used in a resumption strategy in BP: (82) Esse ? o rapaz que 1 a Maria conhece a garota 2 que 2 ele 1 /*e 1 beijou t 2 this is the guy that the Maria know-3Sg the girl that he kissed-3Sg ?This is the guy that Maria knows the girl that he kissed? The contrast in (83a) and (83b) shows that in Spanish non-resumptive null pronouns can freely occur inside a relative clause, whereas in BP, only an overt pronoun is allowed as the subject of a relative clause: 170 (83) a. Juan 1 vi? a la chica que ?l/pro bes? anoche Juan saw-3Sg to the girl that kissed-3Sg last.night ?Juan saw the girl that he kissed last night? b. O Jo?o encontrou a carteira que ele/*e perdeu the Jo?o met-3Sg the wallet that he lost-3Sg ?Jo?o found the wallet that he lost? Clearly, embedded null subjects in modern BP pattern like wh-traces (cf. 84) in being prohibited inside relative clauses: (84) * Quem 1 que o Jo?o encontrou a carteira que t 1 perdeu who that the Jo?o met-3Sg the wallet that lost-3Sg ?Who did Jo?o find the wallet that he lost? The traditional view of restrictive relative clauses involves adjunction; the embedded CP is an adjunct to the maximal projection of the NP, which is taken as the nominal head of the relative clause. In this adjunction analysis, the head of relative clause is generated outside of the relative clause and the relative pronoun inside. The relative pronoun, being an operator, moves to spec of the relative CP, becoming adjacent to the nominal head of the relative clause (cf. Safir 1986, Browning 1987 and Fabb 1990). Kayne (1994), concerned with linearization and its relation to c-command, concluded that the adjunct analysis is not right. According to him, if the relative clause is 171 a CP that is adjoined to the maximal projection of the noun phrase (NP), the CP would asymmetrically c-command the NP, therefore, the CP should linearly precede the NP. Based on that, Kayne advocates in favor of the raising/promotion analysis originally suggested in Vergnaud (1974). In the raising analysis, the whole relative CP is taken as a complement of D 0 . In Kayne?s proposal, the noun, which is usallly taken to be complement of the D, forms a constituent with the relative pronoun. This constituent is taken to be a DP headed by the relative pronoun and is first merged in the extraction site inside the relative clause as in (85a). Then it moves to the spec of the relative CP (85b). At this stage, the noun splits off from the relative pronoun, raising to the spec of the upper DP, as in (85c). (85)a.the [C 0 Bill saw [ DP which picture]] b. the [ CP [ DP which picture] 1 [ C? [Bill saw t 1 ]]] c. the [ CP [ DP picture 2 [ D? which t 2 ] 1 [ C? [Bill saw t 1 ]]] There are some empirical facts that favor Kayne? raising analysis. 17 One of them is that in sentences like the Paris that I love or the three books of John?s that I read the presence of the article is dependent on the presence of the relative clause. Hence, it seems that there is a selectional relation between the determiner and the relative clause. Moreover, Bianchi (1999) shows evidence that the determiner is not part of the head of the relative clause. For example, there are certain idiomatic expressions that require an indefinite NP as the complement of the verb, as made in (86). Thus, if reconstruction 17 For an elaborate discussion and presentation of the raising analysis, see Bianchi (1999). For evidence that the nominal head escapes the relative CP, see Hornstein (2001). 172 takes place in (86b), putting the complement of made back into its original place, the determiner the that precedes fun, cannot be part of the reconstructed constituent, otherwise the indefinite restriction shown in (86a) would be violated. (86)a.They made (*the) fun of me b.The fun 1 that he made t 1 of me As originally presented in Schachter (1973) and Vergnaud (1974), idiomatic expression also provides evidence that relative clauses involve raising of the NP head. In French, idiomatic expressions allow the object to be relativized only if the idiomatic verb is inside the relative clause. If it is external, relativization is not possible. This is shown by the contrast in (87), extracted from Vergnaud (1974): (87) a.Peu de gens ont parl? du [parti qu?il a tir? t few of people have-3Pl sopken of.the advantages that.he have-3Sg taken des difficult?s ?conomiques] of.the difficulties economic ?Few people have spoken of the advantages that he has taken from the economical difficulties? b. * Il a tir? (des difficult?s ?conomiques [um parti don?t peu de he has-3Sg of.the difficulties economic an advantage of.which few of 173 gens ont parl? t] people have-3Sg spoken At any rate, the raising analysis of restrictive relative clause appears to be well motivated, and I will adopt it, returning now to the disallowance of null subjects inside relatives in BP. I will make use of Kayne?s implementation in terms of raising, even though my analysis does not crucially depend on it. We may resume the discussion on the unacceptability of null subjects inside relative clauses in BP by arguing that A-movement cannot cross over the nominal head of relative clause. To see this point, consider the sentence in (83b), repeated here as (88): (88) * O Jo?o1 encontrou a carteira que e1 perdeu the Jo?o kissed-3Sg the wallet that lost-3Sg ?The Jo?o found the wallet that he lost? The derivation of (88) starts by building the relative clause. The DP que carteira ?that wallet?, formed by merging the relative pronoun and the noun carteira ?wallet?, is first inserted in the complement position of perdeu ?lost? and the DP o Jo?o is taken as the external argument of the same verb, receiving, thus, the agent q-role, as shown in (89a). When the complementizer is merged on the top of the object built in (89a), the DP que carteira moves to spec of CP, and the noun splits off from the relative pronoun as shown in (89b). The CP thus formed is taken as the complement of a ?the?, and the derterminer 174 projects forming a DP, as represented in (89c). When the matrix verb is pulled from the numeration and merged with its internal argument, the DP constructed in (89c), it needs to check its external q-role (89d). At this point of derivation, the numeration does not contain any element that could check the verbal q-role. Hence, movement has to take place. Assuming the proposal presented in section 4.2, in (89d) the DP o Jo?o hasn?t checked its Case feature yet, hence it is available for further computation. Thus, this DP could in principle be copied and merged in spec of the matrix VP. However, this movement violates the Miminal Link Condition. The head of the relative clause, the DP que carteira ?that wallet?, occupying spec of CP, is closer to the matrix verb than the DP o Jo?o is. That is, the DP in the spec of the relative CP creates an intervention effect, blocking the movement of the subject of the relative clause towards spec of VP, as represented in (89e): (89)a.[ TP [ DP o Jo?o] 1 [ T? [ VP t 1 [ V? perdeu [ DP que carteira]]]] b.[ CP [ DP [carteira 1 [que t 1 ]] 2 [ C? [ TP [ DP o Jo?o] 1 [ T? [ VP t 1 [ V? perdeu t 2 ]]]]]] c.[ DP a [ DP [carteira 1 [que t 1 ]] 2 [ C? [ TP [ DP o Jo?o] 1 [ T? [ VP t 1 [ V? perdeu t 2 ]]]]]]] d. [ VP encontrou [ DP a [ CP [ DP [carteira 1 [que t 1 ]] 2 [ C? [ TP [ DP o Jo?o] 1 [ T? [ VP t 1 d. [ V? perdeu t 2 ]]]]]]]] x e.[ VP [ V? encontrou [ DP a [ CP [ DP [carteira 1 [que t 1 ]] 2 [ C? [ TP [ DP o Jo?o] 1 [ T? [ VP t 1 [ V? perdeu t 2 ]]]]]]]] 175 Note that null subjects are not licensed inside relative clauses even when the antecedent is a quantifier expression, which in accordance with Montalbetti?s constraint (cf. 3.3.2, fn. 18), cannot be co-indexed with an overt pronoun. Hence the sentence (90) is ungrammatical independently of the phonetic realization of the subject pronoun. This shows that BP 3rdP null subjects cannot be used in a resumptive strategy even it is necessary for convergence. (90)* Ninguem 1 encontrou a carteira que e 1 /ele 1 perdeu nobody found-3Sg the wallet that he lost-3Sg ?Nobody found the wallet that he lost? Finnish, differently from BP, allows null subjects inside restrictive relative clauses. Witness the data in (91). (91)a.Siet? tulee se tytt?, josta puhuin there come-3Sg it girl which-Ela talked-1Sg ?Here come the girl I was telling you about? b.Pekka 1 n?ki tyt?n, jota e 1 oli suudellut Pekka-Nom saw-3sg girl-Acc which-Part have-3sg kissed-3Sg juhlissa party-In ?Pekka saw the girl that he had kissed at the party? 176 Notice that in this language restrictive relative clauses behave like islands with respect to wh-movement (cf. (92)). Therefore, Finnish null subjects do not pattern like wh-traces, in that they can be occur inside relative clauses. This is rather unexpected under the treatment I am suggesting in this thesis. (92)* Kuka 1 Pekka n?ki tyt?n, jota t 1 oli suudellut juhlissa who-Nom Pekka-Nom saw3Sg girl-Acc which-Part have-3sg kissed party-In ?Who did Pekka see the girl that he kissed? As already remarked on chapter 3 (section 3.4), an important property of this language is that it allows topicalization to co-occur with wh-movement in both free and relative clauses. This is exemplified below: (93) a.Kuka t?m?n kirjan on kirjoittanut who this book-Acc have-3SgPres written ?(What about this book) Who has written this book? b.Iti antaa kirjan [lapselle 1 , joka sit? 2 t 1 odotaa t 2 ] mother give-3Sg book-Acc child-All which-Acc it-Part wait-3Sg ?The mother is giving the book to the child who is expecting it? (Sulkala and Karjalainen, 1992) 177 Therefore, we may assume that in Finnish the structure inside a restrictive relative clause is the one in (94). The DP heading the relative clause moves to spec of CP, and the subject also moves to the left periphery, arguably to spec of FP, which according to the proposal we presented in 4.2, is right below CP. (94) [ CP NP 1 [ FP NP 2 [ TP t 2 V t 1 ]]] If this is right, (91a) is compatible with the topic deletion I proposed in 3.4 for 1stP null subjects. During the derivation of this sentence, the 1stP overt subject pronouns min? moves to spec of FP, becoming thus the target of deletion. (95)Siet? tulee [ DP se [ CP tytt? 2 josta [ FP min? 1 [ t 1 ? puhuin ?t 2 ?.]]]] there come-3Sg it girl which-Ela (I) talked-1Sg The allowance of a null subject in (91b) seems to suggest that in Finnish, the antecedence relationship between a null subject and a DP does not obey MLC. However, this conclusion is unlikely to the correct because as I have shown before this relationship displays locality restrictions, the antecedent being the closest c-commanding DP (see 4.3.1 and 43.2). Therefore, it must be the case in (91b) the nominal head of the relative clause is not actually intervening between the null subject and the DP Pekka. One interesting aspect of restrictive relative clauses in Finnish is that the relative pronoun and the noun that heads the relative clause are morphologically marked with 178 different cases. The relative pronoun is marked with the Case assigned by the embedded verb (partitive in (91b)), whereas the nominal head realizes the Case assigned by the matrix verb (accusative in (91b). Moreover, as shown in (96) the head of relative clause might surface separated from the relative CP (Cf. Sulkala and Karjalainen 1992 and Helasvuo 1994): (96)ja sil oli sitten p??ryn?korit puun alla joihin se and he-Ade had-3Sg then pear.baskets-Nom tree-gen under whichIll he-Nom tyhjensi ne p??ryn?t. 18,19 emptied-3Sg those pear-Acc ?And he had then the pear baskets under the tree into which he emptied the pears? (Helasvuo, 1994:164) Therefore, it is plausible that nominal head moves from the relative CP to a Case position position in the matrix VP shell, as represented in (97). The question is about the empty subject inside the relative clause. Can it be a trace of the DP Pekka? If the noun is the element that matters for thematic relations, in (97) the presupposed movement of tyt?n to spec of vP might remove the blockage on the extraction the DP Pekka. Representationally, Pekka?s movement towards spec of vP crosses a trace (ie. a non- pronounced copy). 18 I will put the particle se aside. Its syntactic and semantic functions are unclear to me. 19 This is recorded spoken Finnish. To avoid an overflow of irrelevant I cut from the original sentence the hesitations and repetitions. 179 (97) [?[vP tyt?n [vP [v? n?ki 4 [ VP t 4 [ DP [ CP [ DP t 3 jota] 1 [C? [FP Pekka [F?[TP girl-Acc saw-3Sg which-Part Pekka t 2 oli suudelut t 1 ]]]]]]]]]]]] have-3Sg kissed-3Sg This raises an issue about the Minimal Link Condition. From (97) one can conclude that a trace (ie. unpronounced copy) does not block extraction of a lower constituent. That is, the Minimal Link Condition does not count copies when defining closeness. This accords to Chomsky?s (2001a) idea that only the head of an A-chain (i.e., the whole chain) blocks matching under the Minimal Link Condition. Notice, however, that the movement of Pekka to spec of vP might happen before the movement tyt?n. Therefore, it is possible that Pekka?s movement does violate Miminal Link Condition derivationally, though not representationally. This leads to the conclusion that the Minimal Link Condition is a condition on representations, as already suggested elsewhere (cf. Chomksy (2001a) for instance). 180 4.4 Controlling Gender Agreement There is an interesting aspect of control in Romance that has hitherto remained unnoticed: in obligatory control configurations the antecedent of PRO controls gender agreement inside the embedded domain. I will investigate this phenomenon here, suggesting that this accords to an NP-movement analysis of obligatory control. In addition, I will show that the same happens with 3rdP referential null subjects in BP. 4.4.1Feminine Agreement There are certain nouns in Romance that are invariably marked as [+feminine], though semantically they can refer to either male or female entities. One of these nouns is the Romance counterpart of victim. Thus, given the property just mentioned, in Romance, when victim is combined with the auxiliary verb to be followed by a participial form, the participial form records feminine gender agreement, as Italian, Spanish, European and Brazilian Portuguese illustrate. (98)a.La vittima fu aggredita/*aggredito dai fascisti (It.) the vicitms-Fem was.3sg attacked-Fem/Masc by fascists b.La v?ctima 1 fue atacada 1 /??atacado 1 en la calle (Sp.) the victim-Fem was-3Sg attacked-Fem/-Masc in the steet c. A v?tima 1 foi atacada 1 /??atacado 1 na rua (E/BP) the victim-Fem was-3Sg attacked-Fem/-Masc in.the street ?The victim was attacked by (the) fascists/on the street? 181 (99) shows that feminine gender is also morphologically recorded by an universal quantifier all that is related to the [+feminine] noun. (99)a.Tutte/*tutti le vittime arrivarano nello stesso momento (It.) all-Fem/-Masc the victims-Fem arrived-3Pl in-the same moment ?All the victims arrived at the same time? b. Todas/*todos las v?ctimas llegaron al mismo tiempo (Sp.) all-Fem/-Masc the victims-FEM arrived-3Pl at.the same time c. Todas/*todos as v?timas chegaram no mesmo hor?rio (E/BP) all-Fem/-Masc the victims-FEM arrived-3Pl at.the same time ?All the victims arrived at the same time? It does not matter whether the quantifier is floating is not, it is always [+feminine]: (100)a.Le vittime arrivarano tutte/*tutti nello stesso momento (It.) the victims-Fem arrived-3Pl all-Fem/-Masc in-the same moment b. Las v?ctimas llegaron todas/*todos (Sp.) the victims-Fem arrived-3Sg all-Fem/-Mas c. As v?timas chegaram todas/*todos (E/BP) the victims-Fem arrived-3Sg all-Fem/-Masc ?All the victims arrived (at the same time)? 182 The agreement in (100) is predictable in a theory that treats floating quantifiers as a type of antecedent-trace, like the one suggested by Sportiche (1988). As well-known, Sportiche takes the syntactic dependency between a floating quantifier and an NP to be the result of NP-movement. 20 The quantifier is base generated as an adjunct to the NP, but the NP undergoes leftwards movement leaving the quantifier behind. Thus, as represented in (101), it is arguably the case that in (100) the agreement in gender between the [feminine] noun and the quantifier happens locally, i.e. inside the NP, prior to the movement of as v?timas ?the victims? to spec of TP. (101) TP 4 [ NP as v?timas] 2 T? 3 chegaram 1 VP NP t 1 2 todas t 2 It is assumed that agreement between a noun and a past participle form also obeys locality. Actually, it is taken to be the result of a spec-head relation. That?s why past participle agreement has been used as a diagnosis of movement in clitic climbing configurations (cf. Kayne 1989, Rouveret 1989, Belletti 2000, Sportiche 1996, 1998, among others). The essence of the argument is the following: in French, a direct object 20 For a different analysis of floating quantifiers, see Boskovic (2000), among others. Boskovic suggests that floating quantifiers are inserted acyclically. After NP-movement, the quantifier adjoins to the copy left behind. 183 agrees with a participial form only if the former precedes the latter. The data in (102) exemplifies this restriction. (102)a.Jean a peint(*e) la porte(Sportiche, 1998) Jean has-3Sg painted-(*Fem) the door ?Jean painted the door? b. La porte 1 que Jean a peint(e) t 1 the door that Jean has-3Sg painted-(Fem) ?The door that Jean painted? Sportiche (1996) takes this fact to mean that, in order to trigger f-feature agreement on a participial form, an NP has to move to/through the specifier of the maximal projection that hosts the past participle. Hence, past participle agreement is a reflection of a relation between a specifier and a head. It follows, then, that in (103a) the clitic moved to its spell-out position, passing through the specifier of the maximal projection hosting the participial form, as represented in (103b). This is exactly what Sportiche (1996) defends. (103)a.Jean l?a peint(e) Jean it has-3Sg painted(Fem) ?Jean has painted it? b. Jean [le 1 a [t 1 [peinte ? t 1 ]]] 184 Under a base-generated analysis for clitic climbing (v. Rivas 1977, Jaeggli 1982, Borer 1983 and Sportiche 1983) in (103a), the clitic is lexically inserted into its surface position, and a null pronoun (pro) is inserted in the object position. Therefore, there is no explanation for the agreement between the object and participial form, given that there is no evidence that pro moves to a position higher than that occupied by the object in (102a). 21 Chomsky (2001a) drops the assumption that agreement is licensed under a spec- head relation. However, he maintains that past participle agreement is local. Accordingly, the derivation of (104a) has a stage a (104b) in which the f-features of the past participle match and agree with the f-features of the direct object. That is, the past participle agrees in f-features with the closest NP within its c-command domain. 22 (104)a.There were believed to have been caught several fish b.[ a caught f F [ DO several fish ] f F ] Agree At any rate, it is arguably the case that f-agreement with a universal quantifier and a past participle form involves locality. Interestingly though, in what follows, I show that in obligatory control configurations gender agreement between a [+ feminine] noun and a universal quantifier 21 Uriagereka (1995) provides an analysis for clitic placement in Western Romance that also invokes moveme7unt of the clitic to its spell-out position. He, however, proposes that the clitic is first merged with a null pronoun forming the DP in (ia). The DP thus formed is merged in the argument position of the predicated, then the clitic moves to higher functional projecting - FP, leaving pro stranded in the argument position (ib): (i)a. [ VP V [ DP clitic [ NP pro]]] b. [ FP clitic 1 F [[ VP V [ DP t 1 [ NP pro]]]] 185 or a past participle form is at a first sight a syntactic phenomenon that does not require locality. If the so-called controller is [+feminine] noun, it obligatorily agrees in gender with a universal quantifier or a past participle form inside the infinitival clause. Thus, at the surface of obligatory control configurations, the gender agreement under discussion looks like a long-distance process. However, as I will show, by assuming a NP- movement analysis of obligatory control, we can get rid of this unexpected lack of locality. The agreement is local since the controller is base-generated within the infinitival clause. Unsurprisingly, non-obligatory control configurations, which are not formed by movement according to Hornstein, do not display this ?long-distance? agreement. I will also show that in BP sentences containing an embedded finite null subject clause behave like obligatory control structures in that a [+feminine] noun in the subject position of the matrix clause controls the gender agreement of an embedded universal quantifier or past participle form. This fact corroborates the idea that BP null subjects are obligatory controlled empty categories, and bolsters the movement analysis I am defending in this thesis. 4.4.2 Controlling Gender Agreement in Non-Finite Control Configurations In Romance obligatory control configuration, when a [+feminine] noun is the antecedent of PRO, a universal quantifier or a past participle form inside the embedded infinitival clause must be morphologically marked for feminine gender. 22 Note that this proposal does not explain the French facts in (102) and (103). 186 (105) a.La vittima 1 ha cercato di essere trasferita 1 /??trasferido 1 (It.) the victim-Fem had-3Sg tried of be-inf transferred-Fem/-Masc alla stazione di polizia de College Park to.the station of police of College Park b. La v?ctima intent? ser transferida /?? transferido(Sp.) la victim-Fem tried-3Sg be-Inf transferred-Fem/-Masc c.A v?tima 1 tentou ser transferida 1 /??transferido 1 (E/BP) the victim-Fem tried be-inf transferred-Fem/-Masc para a delegacia de pol?cia de College Park to the station of police of College Park ?The victim tried to be transferred (to the police station at College Park)? (106)a. Le vittime 1 hanno cercato di testimoniare tutte 1 /*tuttti 1 (It.) the victims-Fem have-3Sg tried of testify-Inf all-Fem/-Masc lo stesso giorno the same day b.Las victimas 1 intentaron testificar todas 1 /*?todos 1 (Sp.) the victims-Fem tried-3 rd Pl testify-Inf all-Fem/-Masc el mismo d?a the same day c.As v?timas 1 tentaram testemunhar todas 1 /*todos 1 (E/BP) the victims-Fem tried-3 rd Pl testify-Inf all-Fem/-Masc 187 no mesmo dia in.the same day ?All the victims tried to testify the same day? It is worth noticing that this gender agreement is obligatory even in a context in which all the victims are known to be males. This means that the agreement is defined syntactically, not semantically. Moreover, notice that this obligatory agreement is also observed in obligatory control inside non-finite adjunct clauses. (107)a.La vittima 1 mori? dopo essere stata trasportata 1 /??stato transportato 1 (It.) the victim-Fem died-3 rd Sg after brought-Fem/-Masc all? ospedale to.the hospital b.La v?ctima 1 muri? despu?s de ser tra?da 1 /?(?)tra?do 1 (Sp.) the victim died after of be-inf brought-Fem/-Masc al hospital to.the hospital c.A v?tima 1 morreu depois de ser trazida 1 /??trazido 1 (BP) 23 the victim died after of be-inf brought-Fem/-Masc para o hospital to the hospital ?The victim died after being brought to the hospital? 23 According to Jo?o Costa in European Portuguese the agreement is optional. 188 (108)a.Le vittime 1 lasciarono l?aula subito dopo aver (It.) the victims-FEM left-3 rd Sg the-courtroom right after have-Inf testimoniato tutte 1 /*tutti 1 insieme contro il mafioso testified all-FEM /-Masc together against the gangster b. Las v?ctimas 1 se fueron despu?s de testificar todas 1 /*todos 1 (Sp.) the victims SE left-3Pl after of testify-Inf all-Fem/-Masc contra el mafioso against the gangster c.As v?timas 1 foram embora logo depois de testemunhar (BP) 24 the vicimts-Fem left-3 rd pl right after of testify-Inf todas 1 /*todos 1 contra o mafioso all-Fem/-Masc together against the gangster ?The victims left after (the courtoroom) having testified all (together) against the gangster? In sum, if we were to assume a Government-Binding theory of control, we could state the phenomenon under discussion in the following way: in obligatory control configurations, an embedded quantifier or a past participle form syntactically dependent on PRO agrees in gender with the controller of PRO, as represented in (109). This suggests that PRO does not have f-features and, as result, it allows the long distance agreement relation indicated below. 189 (109) Obligatory Control:[? DP?[ CP [ TP PRO [?Quant/Past Participle?]]] Consider now, cases of non-obligatory control. As exemplified in (110) and (111), in these constructions, agreement with the [+ feminine] noun is not obligatory. In fact, it is rejected in Italian. In Portuguese, it is quite acceptable if the embedded element is a universal quantifier, but, if the embedded element is a past participle, the agreement is at most marginal. In Spanish it seems that the past participle and the quantifier can be either [+masculine] or [+feminine]. (110)a. La vittima 1 ha detto che essere *portata 1 /portato (It.) the victim-Fem has-3Sg said that be-Inf brought-Fem/-Masc alla stazione di polizia non era una bona idea to.the station of police not was-3Sg a good idea ?The victim said that being brought to the police station was not a good idea? b. La v?ctima 1 dijo que ser ?tranferida 1 /transferido a (Sp.) the victim-Fem said-3Sg that be-Inf transferred-Fem/-Masc to otra ciudad no es buena idea another city not is-3Sg good idea c.A v?tima 1 disse que ser ??tranferida 1 /trasferido 1 (E/BP) the victim-Fem said-3Sg that be-Inf transferred-Fem/-Masc 24 Here I am not considering European Portuguese. In this language, the adjunct clause would contain an inflected infinitival verb, allowing, then, non-obligatory control, as discussed by Pires (2001). As for BP, I am considering the colloquial dialect, which has lost inflected infinitives. 190 para outra cidade n?o ? uma boa id?ia to another city not is-3Sg a good idea ?The victim said that being transferred to another city is not good idea? (111)a. Le vittimme 1 hanno detto che testiomoniare *tutte 1 /tutti (It.) the victim-Fem have-3Pl said that testify-Inf all-Fem/-Masc lo stesso giorno puo? non essere una bona idea the same day may-3Pl not be-ing a good idea b. Las v?ctimas 1 piesan que testificar todas 1 /?todos (Sp.) the vicitms-Fem think-3Pl that testify-Inf all-Fem/-Masc el mismo d?a puede no ser una buena idea at.the same day may-3Sg not be-inf a good idea c. As v?timas 1 acham-3Sg que testemunhar ?todas 1 /todos(BP) 25 the vicitms-Fem think-3Pl that testify-Inf all-Fem/-Masc no mesmo dia pode n?o ser uma boa id?ia at.the same day may-3Sg not be-inf a good idea ?The victims said/think that all testifying at the same day may not be a good idea? Therefore, differently from obligatorily controlled PRO, a non-obligatorily controlled PRO has its own set of f-features. Therefore, an embedded universal 25 In European Portuguese, this infinitival clause would also contain an inflected infinitive. Hence, I am not considering European Portuguese. 191 quantifier and past participle dependent on PRO agree with PRO itself, not with a possible antecedent of PRO that is outside the infinitival clause. (112)Non-Obligatory control: [? DP?[ CP [ TP PRO [?Quant/Pparticiple?]]] * This difference between obligatory and non-obligatory control hasn?t been investigated yet, and to mantain an adequate linguistic theory we need to explain it. If we assume the standard Government-Binding theory of control, we are forced to say that in obligatory control the agreement between an NP and a past participle or a universal quantifier can be non-local, being rather a case long distance agreement, as the representation in (112) suggests. Notice however that this is an undesirable solution. If a past participle form is a head with non-interpretable f-features, how can it get its f- features checked by an NP that is base generated in a higher clause? A theory of obligatory control based on Agree, along the lines proposed by Landau (1999, 2000), fares better than the Government and Binding theory does without invoking movement of the controller. In this theory, a universal quantifier or a past form within the infinitival clause has its f-features checked not by the controller, but by PRO, which, in its turn, agrees with its antecedent. Thus, in the obligatory control configurations presented above, the quantifier or the past participle agrees with the controller only in an indirect way, via PRO. Though mechanically doable, note that this analysis requires an elaborate feature transmission process, according to which PRO is 192 able to check the f-features of the past participle form/quantifier only when it agrees with its antecedent. Under a movement analysis, obligatory control configurations are akin to raising constructions in that both of them involve subject-to-subject movement. There is no obligatory controlled PRO. The empty subject of an infinitival clause embedded under a control verb is actually a trace (i.e. a silent copy created by the operation of copy and deletion) of the controller (cf. 2.4). Therefore, in this theory, the quantifier or the past form gets its f-features checked by the [+feminine] controller locally, i. e, prior to the movement of the controller to matrix clause, as shown in (113) and (114). 26 The sentences are Portuguese and, according to my own judgements, (114) is better if the quantifier follows the infinitive verb. Since this a V-to-T grammar, in (114) I am assuming that the quantifier is stranded in spec of VP. (113)As v?timas 1 tentaram ser transferidas 1 para outra cidade the victims-Fem tried-3Pl be-inf transferidas-Fem to another city ?The victims tried to be transferred to another city? (113?) [ TP [ NP the victims] [ VP t [ V? tried [ CP [ TP t [ T? ser [ PastPP t [ PastPP? tranferida?]]]]]]]] (114)As v?timas 1 tentaram testumunhar todas 1 no mesmo dia the victims-Fem tried-3Sg testify-Inf all-Fem at the same day ?All the victims tried to testify at the same day? 26 (113?) is a simplified structure, where I am assuming that the past participle form heads its own maximalprojection. 193 (114?)[ TP [ NP as v?timas] [ VP t [ V? tentaram [ CP [ TP t [ T? testumunhar 1 [ VP [ NP todas [NP t ]] [v ? t 1 ? ]]]]]]]] This analysis provides us with an elegant explanation for obligatory gender agreement observed. In addition, notice that under a movement analysis obligatory control configurations are parallel to raising constructions. This predicts correctly (cf. (115)-(116)) that raising constructions also display the gender agreement under discussion. 27 (115)a.La vittima sembra essere ferita/*ferito (It.) the victim-Fem seems be-inf injured-Fem/Masc b.La v?ctima parec?a estar herida/*?herido (Sp.) the victim seemed-3 rd Sg be-inf injured-Fem/-Masc c. A v?tima pareceu estar ferida/*?ferido (E/BP) the victim seemed-3 rd Sh be-Inf injured-Fem/-Masc ? The victim seemed to be injured? (116)a.Le vittime sembrano essere tutte/*tutti malate/*malati (It.) the victims-Fem seem-3Pl be-Inf all-Fem/-Masc sick-Fem/-Masc 27 Raising constructions with sembrare ?seem? are are not perfect in Italian. In (115) and (116), for instance, speakers prefer to omit the copula or use a finite complement clause. However, despite this, they report that there is a clear contrast between the sentences with the gender and the sentences without it. 194 b.Las v?ctimas parecen estar todas/?*todos en el juzgado (Sp.) the victims-FEM seem-3 rd Pl be-Inf all-FEM/-Masc in the courtroom ? The victims seems all to be in the courtroom? c.As v?timas parecem estar todas/*todos doentes (E/BP) the victims seem-3 rd Pl be-inf all-Fem/-Masc sick ?The victims seem all to be sick/ in the courtroom? Let us turn now to non-obligatory control. According to Hornstein?s analysis, in non-obligatory control, the empty subject of the embedded clause is not PRO, but pro. Hence, assuming that pro has its own ramdom set of f-features in (115) and (116) the embedded quantifier and the participle form get their f-features checked by pro. Putting it baldly, in non-obligatory control, the presence of an empty pronominal category inside the embedded clause blocks the agreement between a [+feminine] noun in the matrix clause and an embedded quantifier or past participle form. As the next subsection shows, in Romance null subject languages, pro also blocks the agreement under consideration. 4.4.3Controlling Gender Agreement in Finite Control Configurations In the Romance null subject languages there is no obligatory gender agreement between the antecedent of pro and a universal quantifier or past participle form related to pro. For instance, in (117) and (118), gender agreement between the [+feminine] noun 195 and the universal quantifier or the past participle is not required. 28 In fact, while in European Portuguese and Spanish, the agreement seems to be optional, in Italian it is prohibited. (117) a.La vittima 1 ha detto che pro 1 era *stata (It.) the victims-Fem has-3 rd Sg said-3Sg that was-3Sg been-Fem aggredita 1 /stato aggredito 1 in strata attacked-Fem/-Masc in street b.La v?ctima 1 dice que pro 1 fue atacada 1 /atacado 1 en la calle (Sp.) the victim said-3Sg that was-3Sg attacked-Fem/Masc in the street c. A v?tima 1 disse que pro 1 foi attacada 1 / attacado 1 (EP) the victims said-3Sg that was-3Sg attacked-Fem/Masc na rua in.the street ?The victim said that he was attacked on the street? (118) a.Le vittime 1 hanno detto che pro 1 faranno ricorso *tutte 1 /tutti 1 (It.) the victims has-3Pl said that will.do-3Pl appeal all-Fem/-Masc insieme together ?The victims have said that they will testify all together? 28 Again, the speakers were asked to judge all the sentences taking into consideration a situation in which all the victims are males. 196 b. Las v?ctimas 1 dijeron que van a testificar todas 1 /todos 1 (Sp.) the victims said-3Pl that will-3Pl to testify-Inf all-Fem/-Masc al mismo tiempo the same time c. As v?timas 1 falaram que v?o todas 1 /todos 1 depor (EP.) the victims said-3Pl that will-3Pl all-Fem/-Masc testify-Inf no mesmo dia at.the same day ?The victims said that they will all testify at same time? This is an expected fact since in these languages pro is syntactically independent of its antecedent, having its own features. Thus, in (117) and (118), the past participle and the quantifier arguably agree with pro in gender. In conclusion, Romance null subjects behave as a non-obligatorily controlled PRO. (119)Romance null subjects:[? DP?[ CP [ TP pro [?Quant/Pparticiple?]]] * Colloquial Brazilian Portuguese (BP) differs from Romance null subject languages in that a participle form or a universal quantifier embedded under a null subject finite clause obligatorily agrees with the antecedent of the null subject. Thus, in (120), the quantifier and the past participial must record feminine gender. 29 29 Some speakers do allow a [+masculine] quantifer. It take it to be an interference of the standard grammar. 197 (120)a.A v?tima 1 disse que e 1 foi atacada 1 /??atacado 1 the victim-Fem said-3Sg that was-3Sg attacked-Fem/-Masc na rua in.the street ?The victim said that he was attacked on the street? b.As v?timas 1 falaram que e 1 todas 1 /??todos 1 v?o depor the victims-Fem said-3Pl that all-Fem/-Masc will-3Pl testify-Inf no mesmo dia at.the same day ?The victims said that they will all testify at the same day? This clearly shows that BP null subjects have a different nature from their Romance counterpart. The f-feature is syntactically dependent on the f-features of the antecedent. In other words, null subjects in Brazilian Portuguese behave like an obligatory controlled PRO: (121) Brazilian null subjects: [? DP?[ CP [ TP e [?Quant/PastParticiple?]]] The agreement in (121) falls under the movement analysis I am defending in this dissertation. In (120), the embedded null subject is just a trace of this antecedent, the [+ feminine] noun. As in (113) and (114), in (120) the [+feminine] noun is base generated as the external argument of the embedded predicate and moved to the matrix clause as the 198 derivation unfolds. Thus, the universal quantifier and the past participle form agree with it prior to its movement to the matrix clause. Thus, we can say that the quantifier in (120b) is floating in the sense of Sportiche (1988). Unsurprisingly, in BP the gender agreement we are investigating is obviated if an overt pronoun is inserted in the subject position of the embedded clause. 30 (122)a.A v?tima 1 morreu depois que e 1 foi transferida 1 /?? transferido 1 the victims died-3Sg after that was-3Sg transferred-Fem/-Masc para um hospital em S?o Paulo to a hospital in S?o Paulo ?The victim died after he had being transferred to a hospital in S?o Paulo? b. As v?timas 1 ganharam o caso depois que e 1 todas 1 /??todos 1 concordaram the victims won-3Pl the case after that all-Fem/-Masc decided-3Pl em testemunhar in testify-Inf ?The victims won the case after they all agreed on testifying? Here, as in (117) and (118), since the [+feminine] noun is base generated in the matrix clause, it is unable to trigger gender agreement inside the embedded clause. As final remark, notice that this obligatory agreement is also observed if the universal quantifier and/or the past participle is embedded under a null subject finite adjunct clause: 30 The possibility of inserting an overt pronoun in the embedded subject position of sentences like (122a) was first reported by Menuzzi (1999), who also observes that this is not possible in European Portuguese. 199 (123)a.A v?tima 1 morreu depois que e 1 foi transferida 1 /?? transferido 1 the victims died-3Sg after that was-3Sg transferred-Fem/-Masc para um hospital em S?o Paulo to a hospital in S?o Paulo ?The victim died after he had being transferred to a hospital in S?o Paulo? c. As v?timas 1 ganharam o caso depois que e 1 todas 1 /??todos 1 concordaram the victims won-3Pl the case after that all-Fem/-Masc decided-3Pl em testemunhar in testify-Inf ?The victims won the case after they all agreed on testifying? 4.5Referential Null Subjects Inside Finite Adjunct Clauses 4.5.1 Right Adjuncts Adjuncts are islands for wh-extraction, as exemplified in (124) (v. Huang 1982, Lasnik and Saito 1984, 1992, Chomsky 1986, among others). This fact led Figueiredo Silva (1996) to argue that a movement analysis for BP 3rdP null subjects is out of the question given that these subjects can occur inside adjuncts, as shown in (125): 200 (124)* Quem 1 o Jo?o saiu depois que t 1 jantou who the Jo?o left-3Sg after that had.dinner-3Sg ? Which x is an x such that Jo?o left after x having had dinner? (125) O Jo?o 1 saiu depois que e 1 jantou the Jo?o left-3Sg after that had.dinner-3Sg ?Jo?o1 left after having had dinner? The very same contrast between wh-traces and null subjects is observed in Finnish: (126)a.* Kuka-Nom Heikki 1 ostaa auton kun t 1 muuttaa Suomeen Kuka-Nom Heikki-Nom buy-3sg car-Acc when move-3Sg Finland ?Who is the x such that Heikki will buy a car when x moves to Finland? b.Heikki1 ostaa auton kun e1 muuttaa Suomeen Heikki-Nom buy-3sg car-Acc when move-3Sg Finland ? Heikki will buy a car when he moves to Finland? In what follows, I argue that the movement analysis I am proposing is not incompatible with the data above. This section examines only adjuncts that surface to the right of the VP. Adjuncts that appear to the left of matrix clause are considered in the next section. As already shown in Chapter 2, some recent work has challenged the idea that 201 movement out of adjuncts is to be disallowed altogether. Assuming that the computational system of natural language has sidewards movement, Nunes (1995, 2004) proposes that parasitic gaps are cases of A-movement out of adjuncts. Hornstein (1999, 2001), building on Nunes? work, proposes that obligatory controlled empty subjects inside infinitival adjunct clauses are also derived by A-movement. I will adopt the core of Hornstein?s analysis. But, before laying out the details of his proposal, I should demonstrate that the null subjects in (125) and (126b) reproduce all the proprieties of obligatory control discussed in section 4.3.4. Here are the tags for the examples below: (127) and (128) show that these null subjects are anaphoric, requiring a proper antecedent. (129) and (130) demonstrate that the relationship between one of them and an antecedent obeys the MLC in that the antecedent is the closest c-commanding DP. (131) exemplifies the ban on split antecedents. (132), (133) and (134) show the semantic restrictions on sentences containing null subjects: only a sloppy reading is permitted under ellipsis (132); with only-NPs only a covariant interpretation is allowed (133), and, a de se belief is required (134). (127) a. * Chove quando e fala com a Maria rain-3 rd Sg when speak-3Sg with the Maria ?It rains when she/he speaks with Maria? b.*? Sataa kun e tulee t?nne rain-3Sg when come-3Sg here ? It rains when he/she comes here? 202 (128) a.O Jo?o 1 chora quando e 1 fala com a Maria the Jo?o crie-3 rd Sg when speak-3Sg with the Maria ?Jo?o cries when he speaks with Maria? b.Jukka 1 meni suihkuun, kun e 1 saapui kotiin Jukka-Nom went-3Sg shower-Ill when arrived-3Sg home-Ill ?Jukka took a shower when he arrived at home? (129)a.A Ana 1 disse que a Maria 2 olha para o ch?o quando e *1/2 fala the Ana said-3Sg that the Maria look-3Sg at the ground when speak-3Sg com o Paulo with the Paulo ?Ana said that Maria looks at the ground when she speaks with Paulo? b.Pekka 1 sanoi, ett? Jukka 2 itkee, kun e *1/2 puhuu Suomesta ?Pekka said that Jukka cries when he talks about Finland? (130)a.[A amiga da Ana 1 ] 2 olha para o ch?o quando e *1/2 fala com the friend of.the Ana look-3Sg at the ground when speak-3Sg with o Paulo the Paulo ?Maria?s mother looks at the ground every time she speaks with Paulo? 203 b.[Pekan 1 yst?v?] 2 itkee, kun e *1/2 puhuu Suomesta Pekka-Gen friend-Nom cry-3Sg when speak-3Sg Finland-Ela ?Pekka?s friend cries when he/she talks about Finland? (131) a.* O Luca 1 disse que a Ana 2 chorou pra caramba depois que e 1+2 deixaram the Luca said-3Sg that Ana cired-3Sg a lot after that left-3Pl o Brasil the Brazil ?Luca said that Ana cried a lot after they left Brazil? b.*? Pekka 1 sanoi, ett? Jukka 2 itki, kun e 1+2 l?htiv?t Pekka-Nom said-3Sg that Jukka-Nom cried-3Sg when departed-3Pl Suomesta Finland-Ela ?Pekka said that Jukka cried when they left Finland? (132)a.A Ana 1 voltou para o Rio depois que e 1 ficou gr?vida e o Luca tamb?m. the Ana wnet-3Sg back to Rio after that got-3Sg pregnant and the Luca to ?Maria went back to Rio after having got pregnant and Jo?o went too? (= and Luca also went back to Rio after he himself having got pregnant) b. Maija 1 meni Suomeen, kun e 1 tuli raskaaksi, Maija-Nom went-3Sg Finland-Illa when came-3Sg pregnant-Trans 204 ja niin Jukkakin 31 and so Jukka-Nom ?Maija went to Finland when she got pregnant and so did Jukka? (= and Jukka also went to Finland when he himself got pregnant) (133)a.S? o Maluf 1 ficou chateado depois que e 1 perdeu as elei??es only Maluf got-3Sg upset after that lost-3Sg the elections ?Only Maluf is an x such that x got upset x after having lost the elections? b.[vain [Jukka] 1 ] 2 oli surullinen, kun e *1/2 menetti pelin only Jukka-Nom was-3Sg sad-Nom when lost-3Sg game-Acc ?Only Jukka is an x such that x was sad when x lost the/a game? (134) O Ronald Reagan 1 acredita que e 1 ficou doente depois que e 1 foi the Ronald Reagan believe-3Sg that got-3Sg sick after that was-3Sg presidente dos Estados Unidos President of.the States United ?Reagan believes that he himself got sick after he himself had been the president of the United States? In addition, recall that in BP if a participle or unverisal quantifier is inserted inside a null subject adjunct clause, its gender feature is determined by the antecedent of the null subject (cf. (123). 31 Kin is an inclusive particle, and Trans = Translative 205 As shown in 2.4.3, according to the movement analysis of obligatory control, the derivation of (135) is built as shown in (136). First, from the lexical array in (136a), the adjunct (136b) is constructed. Next, the matrix vP is assembled by merging saw with the DP Mary forming a VP which in turn merges with the functional category v. Since v has a q-role to check, John is moved sideways from the adjunct clause to the spec of vP, creating the vP in (136c). When the matrix vP is assembled, the adjunct adjoins to vP forming the two-segment category in (136d). After that, a finite tense is merged on the top of the matrix vP and John is moved again to spec of TP (136e) to check its Case feature and satisfy the EPP feature of T. (135) John saw Mary before leaving the party (136)a. {John 1 , saw 1 , Mary 1 , before 1 , leaving 1 , the 1 , party 1 , assorted functional categories} b. before [ IP John 1 [I [ vP t 1 v [ VP leaving the party]]]] c.[ vP John 1 v [ VP saw Mary]] ? Sidewards movement from (b) to (c) d. [ vP [ vP John 1 v [ VP saw Mary]] before [ IP t 1 [I [ vP t 1 v [ VP leaving the party]]]]] e.[ TP John T [ vP [ vP t 1 v [ VP saw Mary]] before [ IP t 1 [I [ vP t 1 v [ VP leaving the party]]]]]]] This proposal has two features that I will preserve in my treatment of Finnish and 206 BP referential null subjects: First, the adjunct is merged with the matrix vP in a strictly cyclic way, i.e, obeying the Extension Condition (Chomsky 1993 ? cf. 2.2.2). Second, the merge of the adjunct happens only after the sidewards A-movement illustrated in (136c). This feature follows from the assumption that adjuncts became islands only after having been merged onto the main spine of the tree. This assumption carries over in a theory in which an adjunct is not an adjunct per se, but becomes one after being adjoined to the main spine of the tree. Hornstein implements this theory by revamping the concept of barrierhood proposed in Chomksy (1986b), according to which an adjunct clause is a barrier because it is not q-marked. Crucially, however, the non q-marked status of a constituent is defined only upon insertion of that constituent into the structure. Hence, an adjunct is defined as barrier only after being adjoined to a maximal projection. It follows then, that the sidewards movement in (136c) does not cross a barrier. In other words, in Hornstein?s proposal, the availability of sidewards movement is predicted by the dymanics of the computational system. Uriagereka?s (1999) multiple spell-out system, as Hornstein observed, also assents to the idea that adjuncts are islands only after being integrated to the main phrase marker. For Uriagereka, when two sub-structures (or c-command units, using the author?s terminology) are about to be merged, one of them has to be spelled-out in order to ensure the linearization of the resultant phrase marker. Thus, it is plausible that in (136) the adjunct is spelled-out prior to its merge with VP for the same reason. The adjunct clause (136b) and the matrix vP (136c) are built in different substructures and there is no c- command relation between the terminals inside the adjunct and the terminals inside the 207 VP. As result, there is no way to impose a linear order among them and the derivation is doomed to crash for violating the LCA (cf. Kayne 1994 and Uriagereka 1999). Thus, the system saves the derivation by applying the spell-out operation to the adjunct, transforming it into a frozen unit, akin to giant word. In what follows, I will assume the idea that an adjunct becomes an island only after being adjoined to the maximal projection that hosts it. However, I will not adopt any particular implementation. Rather, I leave the implementation issue open. The derivation below provides a concrete example of how the movement analysis proposed here encompasses cases of null subjects inside of finite adjunct clauses: (137) O Jo?o 1 saiu depois que e 1 jantou the Jo?o left-3Sg after that had.dinner-3Sg ?John left after having had dinner ? The derivation of (137) is built upon the numeration in (138a). By putting together tokens from (138a), the adjunct clause in (138b) is constructed. Note the DP o Jo?o is first merged inside the adjunct clause, as the external argument of predicate headed by jantou ?had dinner?. When the derivation reaches the matrix vP, the same DP is moved sideways to the matrix vP to check the q-role of v (138c). Notice that at this step of the derivation, merge of a DP in spec of vP is not a possibility given that the numeration contains only functional categories. 208 (138)a. Num = {o1, Jo?o1, jantou1, saiu1, depois1, que1, v2, T2} b. depois [ CP que [ TP [T, jantou1] [ vP [ DP o Jo?o] [ v? t1 ]]]] c. [ vP [ DP o Jo?o] [ v? [ saiu ]]] ? sideward movement from (b) to (c) Obeying Extension, in the next step, the adjunct clause needs to be merged with the matrix vP. Thus, this constituent is first spelled-out and then merged with vP forming the object in (139a). Next, the matrix T is selected, and the DP o Jo?o moves to spec of TP to checks its Case feature. (139b) is thus a convergent phrase marker. (139)a.[ vP [ vP [ DP o Jo?o] 2 [ v? [ v saiu ]]] #[depois [ CP que [ TP [ T jantou 1 ] [ vP t 2 [v? t 1 ]]]]#] ? (138b) is spelled-out and merged with (138c) b. TP 3 [ DP o Jo?o] T? saiu vP vPdepois [ CP que [ TP [ T jantou] [ vP t [v? t ]]]]]# t t All in all, the possibility of having 3rdP null subjects inside adjuncts can be accounted for if sideward movement is allowed, and the island status of adjuncts emerges derivationally, from the necessity of satisfying the LCA or from the barrier status of 209 adjuncts as proposed by Hornstein. According to this system, A-bar extraction from inside adjuncts (cf. (124) and (126a)) is not allowed because by the time the derivation reaches the stage in which the matrix C is inserted, the adjunct clause has already being adjoined to matrix vP and sub- extraction is impossible. This conclusion makes an interesting prediction about the licensing of 3 rd P referential null subjects inside finite adjunct clauses: their antecedents must the subject of the clause that hosts the adjunct. Assuming that our movement analysis is correct, a DP in a higher clause cannot be interpreted as the antecedent for the same reason that a fronted wh-phrase cannot be paired with a trace inside an adjunct: movement out of an adjunct must procede sideways; it cannot occur after the merge of the adjunction with vP. Urpo Nikanne (personal communication) showed that there is a compelling piece of evidence for this locality restriction. In 4.3.2.2.1, it was observed that in sentences like (140), the null subject of most embedded clauses can take the subject of matrix subject as its antecedent, despite the presence of a raising predicate in between. (140)a.A Maria 1 me disse que parece que e 1 vai ser promovida. the Maria me told-3Sg that seems-3Sg that will-3sg be-Inf promoted ? Maria told me that it seems that she will be promoted.? b.Jukka 1 sanoi, ett? oli onni, ett? e 1 oli arpajaisissa Jukka 1 -Nom said-3sg that be-3sg fortune that pro 1 was-3sg won 210 voittanut lottery-In ?Jukka said that it was fortunate that he had won the lottery? Now, note that if in BP and Finnish 3rdP null subjects inside adjunct clauses are the result of sidewards movement, we expect the long distance antecedence relationship observed in (140) to be disallowed if the clause containing the empty subject is an adjunct to the raising predicate. This expectation in fulfilled by the unacceptability of (141). 32 (141)a.*O Jo?o 1 me disse [ CP que choveu [quando e 1 chegou em casa]] the Jo?o me told-3Sg that rained-3Sg when arrived-3SG at home ?Jo?o told me that it rained when he arrived at home? b. * Jukka 1 sanoi [ CP ett? satoi [kun e 1 saapui kotiin ]] Jukka-Nom said-3Sg that rained/snowed-3Sg when arrived-3Sg home ?Jukka said that it was raining/snowing when he arrived at home? Here the adjunct modifies the embedded clause, in which spec of TP is arguably not projected (cf. 4.3.2.2). Therefore, when compared with (140), (141) is an argument for the sideward movement analysis proposed here. Concluding thus, the contrast between (124) & (125) and (126a) & (126b) is due to their different derivational histories. In (125) and (126b), the matrix subject starts as 32 These sentences are fine if the adjunct is interpreted as modifying the matrix clause. However, this reading is irrelevant for what we are testing here. 211 the external argument of the adjunct clause, and moves sideways to spec of the matrix vP, and then to spec of the matrix TP to check its Case. In (124) & (126a) the wh-phrase starts as the external argument of the adjunct clause, checks its Case locally and remains inside the adjunct. Thence, when the matrix CP is built, it is too late to extract the wh- phrase because the adjunct is already an island for extraction. 4.5.2 Left-Dislocated Adjuncts 33 At the last section, we took into consideration adjuncts that surface adjoined at the right of the VP shell. However there are a host of cases of finite adjunct clauses that surface to the left of the matrix clause. The data below show that BP referential 3rdP null subjects are also licensed inside these adjuncts: (142)Depois que e 1 tomou banho, [o Jo?o] 1 saiu after that took-3Sg bath the Jo?o left-3Sg ?After having taken a bath, Jo?o left? As in cases involving right adjuncts, the null subject above obligatorily co-refers with the matrix subject and all the obligatory control properties discussed in section 4.3.3 hold between them except for the local c-command requirement on the antecedent. Thus, (142) is similar to cases of backward control (cf. Farrell 1995, Polinsky and Potsdam 2002 and Hornstein 2003) in that the antecedent does not c-command the null subject. 33 The analysis presented was first delivered as Rodrigues (2000), but it derives great benefit from Pires and Rodrigues (2002) who analyzed left-dislocated non-finite adjuncts. 212 However, it differs from backward control in that the null subject, by virtue of being inside the left-dislocated adjunct, does not c-command the antecedent either. The adjunct clauses in (142) can also appear to the right of the VP shell, as (143) exemplifies. The only evident difference between (142) and (143) in that the adjunct clause receives a focused reading in (142), but not (143). (143)[o Jo?o] 1 saiu depois que e 1 tomou banho the Jo?o left-3Sg after that took-3Sg bath ?Jo?o left after having taken a bath? Note that the lack of interaction between left-dislocation of adjuncts and binding properties is quite general. Take as an example (144), in which the matrix subject binds the reciprocal phrase despite the lack of c-command. (144) After each other?s trial, they were in shock This suggests that left dislocated adjuncts are generated in a lower position. There are at least two pieces of evidence favoring this conclusion. The first one comes from binding relations in ECM constructions involving left dislocated adjuncts; the second one is the interaction between left dislocated adjuncts and the inner island constraint proposed by Ross (1984). Concerned with binding relations in ECM constructions, Lasnik and Saito (1992) 213 observed the contrast in (145). They (as well as Branigan 1992 and Lasnik 1999) proposed that ECM subjects move in the overt syntax to a position high enough to c- command the adjunct clause and bind the reciprocal each other. In Branigan?s and Lasnik?s analysis this movement is to the spec of matrix AgroP, as represented in (145c): (145) a.? The DA proved the defendants to be guilty during each other?s trials b. ?* The DA proved that the defendants were guilty during each other?s trials c. The DA proved 2 [ AgroP the defendants 1 [ VP [ VP t 2 t 1 to be guilty] [during each other trials]]] With (144) in mind, note that the adjunct during each other?s trial of (145a) may appear displaced, at the left periphery of the clause, as in (145). In addition, observe that the considerations on the binding relations shown in (145) are also valid for (146). That is, DP the defendants is able to bind of each other inside the left dislocated adjunct only if it moves to spec of the matrix AgroP, according to the analysis in (144c). (146)a.? During each other's trials, the DA proved the defendants to be guilty b. ?* During each other?s trials, the DP proved that the defendants were guilty This suggests that (145c) is also a step of the derivation of (146a). That?s why the 214 DP the defendants can bind the reciprocal each other in (146a), but not in (146b). If this conclusion is right, difference between (145a) and (145a) is that (146a) involves leftwards movement of the adjunct, whereas (145a) does not. Hence, (145a) has the spelt- out structure in (147), where the copy of the adjunct adjoined to VP, represented as t 3 , is c-commanded by the ECM subject: (147)[ CP [during each other trials] 3 [ TP the DA [proved 2 [ AgroP the defendants 1 [ VP [ VP t 2 t 1 to be guilty] t 3 ] Let us consider now the interaction between left dislocated adjuncts and the inner island constraint. Ross (1984) showed that wh-movement of certain adverbs (manner, degree, instruments and frequency) produces bad results if a negation is present. The negation creates an island for extraction of adverbs (the inner island constraint, in Ross?terms). 34 (148) How did (*n?t) you find a solution The very same effect is observed in (149), which contains a left-dislocated adjunct clause. The presence of the negation bars the presence of left dislocated adjunct. Hence, to the extent that Ross was right in suggesting that negation blocks extraction of certain adjuncts, the sentence in (149) is evidence the left-dislocated adjuncts are moved to their surface position. 34 For an explanation in terms of relativazed minimality, see Rizzi (1990). 215 (149) ?? [e1 dirigindo o BMW], [o Rafael]1 n?o chegou em casa driving the BMW the Rafael not arrived-3Sg at home ?Driving the BMW, Rafael did not arrive at home? (cf. O Rafael n?o chegou em casa dirigindo o BMW the Rafael not arrived-3Sg at home driving the BMW ?Rafael did not arrive at home driving the BMW?) I will assume, therefore, that (142) also involves movement of the adjunct clause. Thus, (142) differs from (143) only in that (142) involves remnant movement of the adjunct. More precisely, I account for licensing of the null subjects in (142) by suggesting that left-dislocated adjuncts are not base generated in their spell-out position. Rather they start the derivation as adjuncts to the matrix VP shell as in (143), and then move to the left periphery of the clasue. This means that in (142) the DP o Jo?o ?the Jo?o? is moved sideways from inside the adjunct clause before the adjunction of this clause to the matrix VP shell. After that, when the derivation reaches the left periphery of the matrix clause, the remnant ( cf. M?ller 1998, Nunes 2003) of the adjunct clause moves to a position in the matrix CP domain, receiving thus a focused reading. I will assume here that this movement is to check an EPP feature of the functional projection that hosts the adjunct. This is concretely shown in (151), the derivation of (150). 35 The derivation starts with construction of the adjunct (stage1). After that (stage 2), the DP o Jo?o ?the Jo?o? moves to the specifier of the matrix VP, and the adjunct clause is spelled-out and merged with VP. Stage (2) is completed when the DP o Jo?o undergoes movement to spec of TP, 35 In (151), I am not representing irrelevant functional projections 216 where it checks its Case feature. In stage (3), a complementizer is merged on the top of the TP and the remnant of the adjunct clause moves to spec of CP. (150) Quando e 1 chegou, o Jo?o 1 tomou banho when arrived-3Sg the Jo?o took-3Sg bath ?When he arrived (at home), Jo?o took a bath? (151) Stage 1 Stage 2 Stage 3 CP 3 quando TP 2 chegou VP 2 t [ DP o Jo?o] TP 3 [ DP o Jo?o] T? tomou VP VP 2 t V? t banho CP 3 TP [ DP o Jo?o] T? 2 omou VP 2 VP t 2 t V? 2 t banho In a nutshell, the analysis proposed in this section subsumes 3rdP referential null subjects inside left dislocated adjuncts under the movement analysis I am arguing for. 217 4.5.3 Null Subjects Inside Clausal Complements of Double Object Constructions Modesto (1999, 2000) proposes that BP 3rdP null subjects are interpreted as variables at LF in the sense that they are bound by a DP in A?-position. This proposal was driven mainly by the contrast in (152). In effect, the DP a M?aria in (152a) cannot be the antecedent of the null subject because it is in an A-position. In (152b), on the other hand, the underlying direct object is a legitimate antecedent because it has being raised to A?- position, spec of CP. 36 (152) a. O Max 1 convenceu a Maria 2 que e 1/*2 tinha de sair the Max convinced-3Sg the Maria that had-3Sg of leave-Inf ?Max convinced Mary that he had to leave? b. Quem 2 o Max 1 convenceu t 2 que e ?1/2 tinha de sair who the Max convinced-3Sg that had-3Sg of leave-Inf ?Who did Max convince that had he had to leave? In 3.5.2.2, after having reviewed Modesto?s proposal I suggest we abandon it, for it presents problems in its implementation and empirical coverage. However, if we do so, we will need an alternative explanation for (152). That is purpose of this section. My goal here is to show that a movement analysis might give us this alternative explanation. 36 For Modesto only the wh-phrase can be the antecedent in (152b). Indeed some speakers do not allow co- indexation with the matrix subject. However, when confronted with the sentence in (i), the majority of the speakers take the matrix subject as the antecedent. Hence, there might be nothing within the grammar preventing the matrix subject in (152b) to be the antecedent. (i) Quem 1 a Maria 1 convenceu t 1 que e 1 estava gr?vida. ?Who did Maria convince that she was pregnant? 218 Modesto took for granted the assumption that in (152) the direct object of the verb convencer ?to convince? c-commands the embedded clause. However, as I discuss below, it is unclear that this assumption is correct. Ferreira (2000) observed that the subject of sentences embedded under the verb convencer ?to convince? can be an epithet referring back to the matrix object: (153) O Ira 1 convenceu o Diogo 2 que a bob?o *1/2 n?o deveria comprar o carro the Ira convinced-3Sg the Diogo that the silly should-3Sg buy-Inf the car da Cilene of.the Cilene ?Ira convinced Diogo that the silly shouldn?t buy Cilene?s car? If epithets cannot be coindexed with a c-commanding nominal in argument position (cf. Lasnik 1976, 1989, Hornstein and Weinberg 1990 and Higginbotham 1992), then in (153), the fact that co-indexation between the epithet a bobona ?the silly? and the DP a Maria is allowed suggests that the matrix direct object does not c-command the embedded subject. 37 37 Juan Uriagereka (personal communication) observed that this analysis wrongly predicts that the pronoun can be co-indexed with the DP a Maria in (i): (i)O Jo?o convenceu ela* 1,2 que a Maria 1 est? errada the Jo?o convinced-3Sg her that the Maria is-3Sg wrong ?Jo?o convinced her that Maria was wrong? The point is well taken, but it is not entirely persuasive as couterevidence to the lack of c-command assigned to (152). BP seems to disallow co-indexation of DP with pronoun that precedes it even in a configuration like (iii), where it is clear that the pronoun does not c-command the DP which it is co- indexed with. 219 Thus, if the matrix object does not c-command the embedded clause, the question is about the structural position occupied by embedded clause. In Rodrigues (2000), I suggest that this is an adjunct position. That?s why these clauses resist wh-extraction, as (154) illustrates: (154) a. ?? Quem 1 o Jo?o convenceu a Maria que t 1 vem amanh? who the Jo?o convinced-3Sg the Maria that come-3Sg tomorrow ?Who is x such that Jo?o convinced Maria that x will come tomorrow? b. ??O que 1 o Jo?o convenceu a Maria que o Pedro precisa comprar t 1 what the Jo?o convinced-3Sg the Maria that the Pedro need-3Sg buy-Inf ?What is x such that Jo?o convince Maria that Pedro needs to buy x ? c. ?*Como/porque 1 o Jo?o 2 convenceu a Maria que e 2 tinha consertado how/why the Jo?o convinced the Maria that had-3Sg fixed o carro t 1 the car ?How/why did Jo?o convinced Maria that he had fixed the car? (iii) O Jo?o vai acreditar [que ele *1/2 est? falando a verdade] muito mais r?pido do que a the Jo?o will-3Sg believe-Inf that he is-3Sg speaking the truth much more fast of.the that the m?e do Paulo 1 (vai) mother of.the Paulo will-3Sg ?Jo?o will believe that he is speaking the truth much more fast than Paulo?s mother will? This clearly contrasts with English, where a non-commanding pronoun can freely corefer with an R-expression, as Lasnik and Saito (1991) show with (iv): (iv) Joan believes that he 1/2 is a genius even more fervently than Bob 1 ?s mother does 220 This adjunction status of the finite clause following the verb convencer can be seen as demotion of an argument in the sense of Larson (1991). 38 I will assume that it is an adjunction to the vP shell. Therefore, given the movement analysis I am arguing for, (152a) is to be analyzed as an instance of movement out of an adjunct clause. The sentence in (152a), repeated here as (155), is derived as shown in (156). The derivation starts in the numeration in (156a). First the adjunct clause in (156b) is assembled, with the DP o Jo?o being inserted as the external argument of the leaving predicate. After that, the matrix vP (156c) is constructed. The DP a Maria is merged as the internal argument of the matrix verb, being assigned the theme q-role. Since convencer ?to convince? is a transitive predicate, it requires an external argument. At this stage of the derivation, there is no item in the numeration that can hold a q-role feature, therefore the DP o Jo?o is moves sideways to the matrix spec of vP, forming the vP in (156d). (155) O Jo?o 1 convenceu a Maria 2 que e 1/*2 tinha de sair the Jo?o convinced-3Sg the Maria that had-3Sg of leave-Inf ?John convinced Mary that he had to leave? (156)a.N= {o 1 , Jo?o 1 , a 1 , Maria 1 , convenceu 1 , v 1 , T 1 , C 1, tinha 1 , de 1 , sair 1 , que 1 } b. [ CP que [ TP [ T? [ T tinha 1 ] [ vP [o Jo?o] [ v? t 1 tinha de sair]]]] c. [ vP [ V ? [ v convenceu] [ VP t 2 a Maria]] 38 Based on the VP shell structure proposed by Larson (1988), Larson (1991) suggests that in the VP shell headed by promise, the inner VP-subject position is dethematized, as a consequence, the notional direct object ? the embedded clause in (i) is structurally realized as V? adjunct. In as shown in (i), in this proposal, the indirect object, Mary, moves to the empty specifier of the inner VP for Case reasons. (i)[ VP John [ V? promise 2 [ VP Mary 1 [V? t 2 t 1 [V? to return home by 5:00 p.m]]]]] 221 d. [ vP [o Jo?o] 3 [ v? [ v convenceu 2 ] [ VP t 2 [a Maria]]]] ? Sidewards movement from (b) to (c) In order to obey extension, at this step of the derivation the adjunct clause in (156b) needs to adjoin to vP. Given that (156b) and (156c) are different c-command units, for linearization purposes (156b) is spelled-out before being merged with vP. The object in (157a) is thus formed. Finally, a finite Tense is merged on the top of the matrix vP and the DP o Jo?o moves to spec of TP where it has its Case feature checked, and the convergent structure in (157b) is sent to the interfaces. (157)a. [ vP [ vP [o Jo?o] 3 [ v? [ v convenceu 2 ] [ VP t 2 [a Maria]]]] #[ CP que [ TP [ T? [ T tinha 1 ] [ vP [o Jo?o] [ v? t 1 tinha de sair]]]]#] ? (b) is spelled-out and merged with (d) b. TP 3 [ DP o Jo?o] T? [convenceu] 2 vP vP [ CP que [ TP [ T? [ T tinha 1 ] [ vP t [ v? t tinha de sair]]]] t v? 3 t VP t [ DP a Maria] 222 In sum, the movement analysis can account for (152a). The movement of the DP o Jo?o is taken to be a case of intra-arboreal (sidewards) movement from the adjunct to matrix clause. Note that Merge-over-Move as an economy condition (cf. 2.2.1) blocks the matrix objects from being the antecedent. Had the derivation in (156)-(157) taken a different path, with the DP a Maria being merged as the subject of the adjunct clause and then moved to the matrix object position, Merge-over-Move would have been violated. The DP o Jo?o could have been assembled and merged in the matrix object position, blocking, therefore, the movement of the DP a Maria. Consider now (152b). The derivation of this sentence diverges from economy conditions because movement of the wh-phrase quem ?who? from subject position of the adjunct clause to matrix direct object position supercedes merge of the DP o Jo?o. I will return to this issue below. Norbert Hornstein (personal communication) pointed out to me that (152b) resembles parasitic gap constructions (cf. (158)) in that the gap inside the embedded clause is parasitic of the gap in the matrix object position. (Following the terminology suggested in Culicover (2001), hereafter, I will call these two gaps the parasitic gap and the true gap, respectively). To see the similarity under discussion compare (158) with (159). Among the sentences in (159), only (159c) (=152b) contains a true gap, therefore only in this sentence the parasitic gap (the null subject) is licensed. 39 (158) Which report1 did you file t1 without reading PG. 39 On the properties of parastic gaps, see Engdahl (1983), among others. 223 (159)a. * O Jo?o convenceu a Maria 1 que e 1 tinha de sair the Jo?o convinced-3Sg the Maria that had-3Sg of leave-Inf ?Jo?o convinced Maria that she had to leave? b.?? Quem 1 o Jo?o convenceu a Maria que t 1 tinha de sair who the Jo?o convinced-3Sg the Maria that had-3Sg of leave-Inf ?Who did John convinced Maria that he had to leave? c.Quem 1 o Jo?o convenceu t 1 que e 1 tinha de sair who the Jo?o convinced-3Sg that had-3Sg of leave-Inf ?Who did Jo?o convince that he had to leave? Another similarity between the cases under discussion and parasitic gap constructions is the prohibition against replacing the true gap with a resumptive pronoun. The Spanish sentence in (160), for instance, is unacceptable with the resumptive pronoun lo as discussed in Chomsky (1982). (160) El reloj de que me hablaste, que (*lo) han conseguido arreglar t sin mover PG, ha quedado muy bien ?The clock you spoke to me about, which they got to fix (it) without moving, now works very well? In (152a), as noted by Modesto, the DP a Maria can be the antecedent if it is topicalized, as in (161a). However, if a resumptive pronoun replaces the true gap, the 224 antecedent relationship between a Maria and the parasitic gap (the null subject) is blocked: (161)a.A Maria 1 , o Jo?o convenceu t 1 que e 1 tinha de sair the Maria, the Jo?o convinced-3Sg that had-3Sg of leave-Inf b. * A Maria 1 , o Jo?o convenceu ela 1 que e 1 tinha de sair the Maria, the Jo?o convinced-3Sg her that had-3Sg of leave-Inf ?As for Maria, Jo?o told her that she had to leave? (162) shows another similarity between (152b) and parasitic gap constructions: the parasitic gaps in (162a) and (162b) cannot occur within an island within another island. (This condition on parasitic gaps was discussed in Kayne 1984, Contreras 1998, and Chomsky 1986a). (162) a.* Which book did read t [before Fred asked Bill [whether you review PG]] b. *Quem 1 o Jo?o convenceu t 1 [que choveu [assim que e 1 saiu]] who the Jo?o convinced-3Sg that rained-3Sg as.soon.as that left-3g ?Who did Jo?o convince that it rained as soon as he left? In short, the null subject in (152b) displays the hallmarks of parasitic gaps. Therefore, I proposed a unified analysis, treating (152a) as a parasitic gap construction. To do so, I will adopt Hornstein?s (2001) account of parasitic gaps. The derivation of 225 (163a) gives us an overview of Hornstein?s proposal. (163) a.Which book did you read before Fred reviewed b. {which, book, you read, before, Fred, reviewed, assorted functional categories} c. before [which book 1 [ IP Fred I [ VP Fred [V? reviewed t 1 ]]]] d. [ VP you [ V? read which book]] e.[ CP which book 1 [ IP you I [ VP [ VP you [ V? read t 1 ]] [before [t 1 [ IP Fred I [ VP Fred [V? reviewed t 1 ]]]]]] Using lexical items from the numeration in (163b), the adjunct in (163c) is built. Notice that inside the adjunct clause the wh-phrase which book moves to A?-bar position, which Hornstein takes to be either an outer spec of IP or spec of CP. After having built (163c), the matrix verb is pulled out from the numeration. To check the theme q-role of read, which book is moved sideways from (163c) to the object position of read. This step of the derivation violates Merge over Move. The pronoun you, which is still available in the lexical array could have been merged with read, checking theme q-role feature. But, the wh-phrase is moved anyway and you is merged later in spec of VP, being assigned the agent q-role of read (cf. (163d). As represented in (163e), when the matrix C is inserted, the wh-phrase moves to spec of CP to check its wh-feature. Thus, building on Nunes (1995, 2003), Hornstein?s adopts a movement analysis for parasitic gaps. However, differently from Nunes, this author proposes that the wh- 226 phrase moves to an A?-position inside the adjunct clause. This A?-movement according to him is the reason for why the derivation of (162a) is allowed to violate merge over Move. This violation of economy is a strategy to satisfy Principle C. The A?-movement of which book in (163c) created a variable, defined by Hornstein (2000:84) as a ?Case-marked trace?. Since variables are to be A-free, in (163d), movement to the matrix object position, a non-commanding position, is the only way to guarantee that the variable inside the adjunct will not be A-bound. Consequently, Move supervenes upon Merge in order to satisfy Principle C. Note that Merge over Move is understood as an economic evaluation on convergent derivations, whereas condition C is condition on convergence. 40 Now consider sentence (152b), repeated here as (164): (164) Quem 1 que o Jo?o convenceu t 1 que e1 tinha de sair who that o Jo?o convinced-3Sg that had-3Sg of leave-3Sg ?Who did Jo?o convinced that he had to leave? I propose that the derivation of (164) is on a par with the derivation in (163). From the numeration in (165a), the adjunct clause in (165b) is constructed. Here, the wh- phrase quem checks the q-role of the leaving predicate and after that it moves to spec of TP here it checks its Case feature, and then it moves to the specifier of the local CP. Thus, the copy of the wh-phrase is spec of TP has its Case feature checked and it is also bound by an operator. Hence, a variable is created. When the matrix verb is pulled out from the numeration, a copy of the wh-phrase is merged with it, creating the VP in 40 Hornstein (2001) attempts to derived principle C from the restriction imposed upon the mapping of einitial array into linear scope order at LF. 227 (165c). The functional category v is merged on the top of the matrix VP and the DP o Jo?o is assembled and merged with vP, checking the external q-role of the matrix predicate (165d). The adjunct clause is spelled-out and merged with (165c), creating (165e). After that, a finite T is selected from the numeration and merged with (165d), and the DP o Jo?o moves to spec of TP in order to check its Case feature. When the matrix CP is built, the wh-phrase moves to spec of CP, where it has its wh-feature checked. The convergent structure in (165f) is, thus, formed. (165)a.N= {o 1 , Jo?o 1 , quem 1 , convenceu 1 , v 1 , T 1 , C 1, tinha 1 , de 1 , sair 1 , que 1 } b.[ CP quem 2 [ CP que [ TP t 2 [ T? [ T tinha 1 ] [ vP t 2 [ v? t 1 tinha de sair]]]] c. [ vP [ V ? [ v convenceu] [ VP t 2 quem 1 ] ? Sidewards movement from (b) to (c) d.[ vP [o Jo?o] [ v? [ v convenceu 2 ] [ VP t 2 quem 2 ]]] e. [ vP [ vP [o Jo?o] [ v? [ v convenceu 3 ] [ VP t 3 quem 2 ]]] [ CP 2 [ CP t 2 que [ TP t 2 [ T? [ T tinha 1 ] [ vP t 2 [ v? t 1 tinha de sair]]]]] ? (b) is spelt-out and merged with (c) f. CP 3 quem 2 C? C TP [ DP o Jo?o] T? [convenceu] vP 3 vP CP t v? 2 [ t que [ TP t [ T? [ T tinha 1 ] [ vP t [ v? t de sair]]]] t VP t t 228 The violation of economy in (165c), where movement of the wh-phrase quem takes precedence over merge of the DP o Jo?o, is requiered for convergence. Inside the adjunct clause, the trace of the wh-phrase in spec of TP is a variable, hence it cannot be A-bound. Therefore, as in (164d), in (165c) violation Merge over Move is allowed because there is no other convergent route. If the wh-phrase had moved through spec of the matrix vP, principle C, which is a condition on convergence, would have been violated. An important remark concerns the availability of the ?parasitic gap? movement proposed in (165) in structures with other types of adjuncts. Since we are assuming that the embedded clause in (152) is an adjunct, we are predicting that the derivational conspiracy that forced sidewards movement through the object of the matrix in (165) is also observed in structures involving bona fide adjunct clauses. This prediction is indeed borne out, as (166) shows. 41 (166) Quem 1 a Maria visitou t 1 [quando e 1 foi para Bras?lia]] who the Maria visited -3Sg when went-3Sg to Brasilia ?What is the x such that Mary visited x when x went to Brasilia.? Interestingly, speakers that accept (166) do not accept (167), where an overt pronoun is inserted in matrix object position. This shows that bona-fide adjuncts behave exactly like the case we discussed in this section. 41 It is important thought to observe that (166) is allowed only in the right pragmatic context. According to the speakers, this sentence is marginal in an out of the blue context. 229 (167) * O Pedro 1 , a Maria visitou ele 1 [quando e 1 foi para Bras?lia] the Pedro, the Maria visited-3Sg he when was-3Sg to Brasilia ?Pedro, Maria visited him when he went to Brasilia? I will leave skirted the question as to how movement from one Case position to another is possible. In (165c), the wh-phrase quem moves a Case position (the object position of the matrix clause) after having checked its Case feature against the embedded finite T. Although, I will leave this issue unsolved, notice that it is resolved if we follow Nunes and Uriagereka (2000) in assuming that, when one copy ?migrates? from a command unit to another, its already checked uninterpretable features become active again. In (165) the sidewards movement of the wh-phrase quem from the adjunct clause to matrix object position resultes in the reactivation of the Case feature of wh-phrase, which is thus checked again against the matrix v. 4.6Conclusions In this chapter, I analyzed 3rdP referential null subjects in Finnish and Modern BP vis-?-vis NP-movement. Mainly, it was proposed that in these grammars A-movement out of finite clauses occurs and correlates with the weakening of the 3rdP verbal agreement morphology (Agr). Presupposing that in pro-drop languages pro is the verbal agreement morpheme (Agr) itself, I suggested in that the grammars under investigation are not null subject 230 grammars, because Agr underwent f-degradation and was lexically reanalyzed as part of the verb. However, it was hypothesized that Agr retained a D-feature that can satisfy the EPP feature of T when the verb moves to T. As result, in this grammatical system, a DP can move from a finite embedded clause without visiting spec of TP, the Nominative Case checking position. The following arguments were presented in favor of a movement analysis: (i) the null subjects under consideration have an anaphoric behavior, requiring a sentential antecedent. (ii) Their antecedence relationship with a DP obeys the Minimal Link Condition, the antecedent being the closest c-commanding DP. (iii) In BP these subjects fail the resumption test, being disallowed inside relative clauses. Finnish relative clauses might have 3rdP referential null subjects but only because the nominal head of the relative clause undergo object shift. (iv) In Finnish and BP, these subjects display all the diagnostics used to characterize obligatory control as the residue of movement; (v) In BP, clauses embedded under the connective como cannot host a 3rdP referential null subjects arguably because these clauses are left branches inside a DP. The last source of evidence that I discussed was related to gender feature of past participles and floating quantifiers. It was observed that in Romance non-finite obligatory control configurations, the antecendent of PRO controls the value of the gender feature of a past participle form or a floating quantifier embedded under the c-command domain of PRO. In BP, the same phenomenon is observed in structures involving an embedded 3rdP referential null subjects. As discussed, this follows straightforwardly if the considered empty subjects are residues of movement. 231 I also analysed the occurrence of 3rdP referential null subjects inside finite adjunct clauses and suggested that they are formed via sideward movemt. 232 I Appendix More on Adjuncts I turn now to another potential case of A-movement out of a finite adjunct clause. However, this time I will be considering the DP domain, as the examples in (167) indicate: (168)a.[ DP o susto do Jo?o 1 [quando e 1 chegou em casa]] foi grande the shock of.the Jo?o when arrived-3Sg at home was-3Sg big ?Jo?o?s shock when she arrived at home was huge? b. Voc? perdeu [a cara do Jo?o 1 [quando e 1 viu a Maria chegando]] you missed-3Sg the face of.the Jo?o when saw-3Sg the Maria arriving ?You missed John?s face when he saw Maria arriving? In (168), the possessor is the antecedent of the null subject and the sentences are fine. But, as illustrated in (169), the complement of NP cannot be the antecedent. Note that the sentences in (169) are acceptable with a full pronoun replaces the null subject: (169) a.*[aquela foto da Gisele 1 quando e 1 desfilou em S?o that picture of Gisele when waked-3Sg down the cat way in S?o Paulo] ? linda Paulo is-3Sg beautiful (cf. Aquela foto da Gisele1 quando ela desfilou em S?o Paulo ? linda) ?That picture of Gisele in the S?o Paulo Fashion show in very beautiful? 233 b. *A pol?cia n?o falou das condi??es do cad?ver quando e foi encontrado (cf. A pol?cia n?o falou das condi??es do cad?ver quando ele foi encontrado) ?The police did not talk about cadaver?s condition when it was found? In addition, a DP inside the possessor phrase also fails to be the antecedent of the null subject (cf. (170). Again, (170) are acceptable with an overt pronoun replaces the null subject: (170)a.[O susto [da m?e [da Jo?o 1 ]] 2 quando e *1/2 chegou em casa] the shock of.the mother of.the Jo?o when arrived-3Sg at home foi grande was-3Sg big (cf. O susto [da m?e [do Jo?o 1 ]] 2 quando ela 1/2 chegou em casa] foi grande) ?Jo?o?s mother?s shock when she arrived at home was big? b. Voc? perdeu [a cara [do pai [do Jo?o] 1 ] 2 quando e *1/2 estava you missed-3Sg the face of.the father of.the Jo?o when was-3Sg 234 conversando] com a Maria talking with the Maria (cf. Voc? perdeu a cara [da m?e [do Jo?o] 1 ] 2 quando ele 1/2 estava conversando com a Maria) Assuming that in BP nominal complements stays in situ throughout the derivation and that nominal adjuncts are adjoined to the NP, (169) and (170) corroborate the c- command requirement on the antecedent. In these sentences, the null subject is not license because it is not c-commanded by its antecedent. The occurrence of a null subject in (168) poses two questions about propositioned possessors: (171)a. What is their position inside DP? That is, are they high enough to c- command the adjunct clause that is arguably adjoined to NP? b. What is the role of a proposition phrase headed by de ?of? in defining a c- command domain? A possessor phrase is arguably generated inside de NP as an argument of possessed NP (cf. chapter 5 for a discussion), but it has being largely assumed that it moves to a position outside the NP (cf. Szabolcsi 1983, 1994, Abney 1986, Giorgi and Longobardi 1991, Ritter 1991, Cinque 1994, Kayne 1994, Wit 1997, Longobardi 2001 among other). For instance, in (172) the NP is elided, but not the possessor phrase. I take 235 it to be evidence that in BP the possessor phrase moves inside the NP. If the possessor phrase survives NP ellipsis, it is because it has been raised to a position outside the NP. 42 (172)a.Eu li o livro do Chomsky, mas o do Lakoff n?o I read the book of Chomsky, but the of.the Lakoff not ?I read Chomsky?s book, but Lakoff?s book? Moreover, notice that in general the possessor precedes NP adjuncts: (173) a.Eu encontrei aquela filha do Jo?o [de cabelo comprido] I met-1Sg that daughter of.the Jo?o of hair long ?I met John?s daughter with long hair? I will not dicuss the details of this movement here. Based on (172) and (173) I will simpy assume that it occurs. Thence, I will answer question (171a) with the following hypothesis: in (168) the possessor phrase moves to position higher than the adjunct clause. Let me now address (171b). If the c-command requirement on the antecedent is real, then in (168), the DP o Jo?o ?the Jo?o? contained inside prepositional phrase do Jo?o ?of Jo?o? must be able to c-command outside the prepositional phrase. This falls in place if the preposition de ?of? is a dummy preposition, as proposed by Giorgi and 42 On movement of the possessor phrase, see Schoorlemmer (1998), ad Raposo (1999). I am following Raposo (1998) in assuming that one substitution (Hornstein and Lightfoot 1981) corresponds to ellipsis is Portuguese. 236 Longobardi (1993). As these authors show, the complement of the proposition de (di in Italian) ?of? is able to command outside the prepositional phrase containing it. In (174a), for instance of DP Gianni is able to bind the reflexive se stessso, and in (174b), the Gianni binds the pronouns, which, as result is not free in its minimal domain. (174)a.L?opinione [di [se stesso] 1 ] [di [Gianni] 1 ] ? troppo lunsinghiera the opionion of himself of Gianni is too flattering ?Gianni?s opinion about himself is too flattering? b. *L?opinione [di [lui] 1 ] [di [Gianni] 1 ] ? troppo lusinghiera The opinion of him of Gianni is too flattering ?Gianni?s opinion about him is too flattering? Mart?n (1995) shows that Spanish behave in the same way: (175)a.La opini?n [de [s? mismo] 1 ] [de [Juan] 1 ] es demasiado halagadora the opionion of himself of Juan is too flattering ?Juan?s opinion about himself is too flattering? c. *L?opini?n [de [?l] 1 ] [de [Juan] 1 ] es demasiado halagadora the opinion of him of Juan is too flattering ?Juan?s opinion about him is too flattering? And Brazilian Portuguese does too, as the data in (176) show: 237 (176)a.A vis?o [d[a Maria] 1 ] [de [si mesma] 1 ] ? esquisita demais the view of.the Maria of herself is weird too ?Maria?s view of herself is too weird? b.A vis?o [d[a Maria] 1 ] [d[ela] 1 ] ? esquisita demais the view of.the Maria of.her is weird too ?Maria?s view of herself is too weird? Thus, following Giorgi and Longobardi (1993) and Mart?n (1995), I will assume that de ?of? does not define a c-command domain. This answers (171b). One could propose an alternative anwer to (171b) basing on Anderson (1979). Anderson shows that in English a complement of a thematic preposition inside a VP can c-command other object inside the VP. Some of her examples are reproduced in (177). The negative polarity item anything in (177a), the reciprocal each other in (177b) and the bound variable pronoun he in (177c) are licensed by a element inside a prepositional phrase. (177)a.He spoke with to very few people about anything important b. Jake likes to talk to his daughter about herself c. The shift boss talked to every blaster before he left the job Given that a possessor is arguably an argument of the noun (cf. chapter 5), (168) could in principle be grouped with Anderson?s case. This would explain the why the DP 238 o Jo?o inside an of phrase is allowed to be the antecedent of the null subject. Notice, however, that this explanation raises questions about the unacceptalibity of (178), where ningu?m ?nobody? inside a thematic preposition phrase fails to license the negative polarity item nada ?nothing?. 43 (178)*A Maria conversou com ningu?m sobre nada the Maria talked-3Sg with nobody about anything ?Maria talked to nobody about anything? In addition, assuming an analysis ? la Anderson. an explanation for (179) is also required. In these sentence, the DP o Jo?o ?the Jo?o?cannot be the antecedent of the null subject, even though is occurs inside a prepositional phrase that is an argument of the bridge verb disse ?say?: (179) [a Maria] 1 disse para [o Jo?o] 2 que e 1/*2 tinha de sair the Maria said-3Sg to the Jo?o that had-3Sg of leave-Inf ?Maria told Jo?o that she had to leave? Anderson tied the facts in (177) to the availability of preposition stranding in English. For her, the c-command relations in (177) are possible because the preposition is 43 Notice that (178) becomes acceptable if the negative polarity item is properly licensed, as in (i), where nada ?nothing? is licensed by the negative item nunca ?never?. (i) A Maria nunca conversou com nigu?m sobre nada da vida dela. the Maria never talked-3Sg with nobody about nothing of.the life of-she ?Maria never talked to anybody about any aspect of her life? 239 reanalyzed, being adjoined to the adjacent verb. 44 Branigan (1992) suggests a minimalist analysis for the cases, proposing that at LF the DP inside the prepositional phrase moves to spec of AgroP, leaving the preposition stranded. If the grammatically (177) is due to preposition stranding, then the fact that BP does not allow preposition stranding explains why the sentences in (178) and (179) are unacceptable. It also indicates that (168) is not be related to Anderson?s findings. Therefore, we can dismiss an analysis along the lines proposed by Anderson. 44 On reanalysis of prepositions, see also Hornstein and Weinberg (1981). 240 CHAPTER 5 LOSS OF NOMINAL MORPHOLOGY AND NULL POSSESSORS 5.1 Preliminaries Like chapter 3, this chapter is a transition piece. Its main purpose is the set up the background necessary for the movement analysis I will defend in chapter 6 for BP null possessors. Differently from chapter 3, however, this chapter starts with a brief overview of the syntax of nominals because they less studied. In this overview I will provide neither a deep discussion of the matter nor analyze new data; rather my purpose is to make explicit the theoretical apparatus necessary to understand the technical implementation of the movement analysis I will offer in chapter 6. In the second part of the chapter, I discuss loss of morphology inside the nominals and also the licensing of null possessors in BP and Finnish. Here is the division of the sections: 5.2 is dedicated to the structure of nominal descriptions. It starts presenting the ?DP hypothesis? and ends with the DP structure I am 241 assuming this thesis. Section 5.3 is about possessive phrases. In particular, following Abney (1987), and Vergnaud and Zubizarreta (1992), among others, I will assume that relational nouns are argument-taking categories that q-mark the possessor phrase. Section 5.4 discusses the loss of morphology inside nominals in BP grammars, where the pronominal possessive system was simplified due to the loss of the 2 nd P possessive form tua ?your?, which was replaced by sua, originally a 3 rd P pronoun. In section 5.5, I take on null possessors in Standard Finnish, showing that this grammar also displays correlation between agreement morphology and the allowance of null possessors. Section 5.6 is dedicated to final remarks. 5.2 The Nominal Structure 5.2.1The DP Hypothesis The traditional view of nominal descriptions takes them to be NPs whose specifier is occupied by a determiner. This view became obsolete after Szabolcsi?s and Abney?s work showing that nominals are structurally similar to clauses. Szabolcsi (1981, 1983) claimed that nominals, likewise clauses, have an expanded syntactic structure composed by lexical and functional projections. 1 She was taking into consideration Hungarian possessive descriptions. As shown above, in Hungarian the 1 A parallelism between clauses and nominals was first suggested by Lees (1960), who noticed that both clauses and nominals can be taken as arguments and can undergo passivization. Lees? suggestion was that embedded sentences are dominated by an NP node. 242 possessor phrase agrees in f-features with the possessive noun and displays Nominative Case. Thus, syntactically it behaves like a subject. (1)a.az ?n-? kar-ja-i-m (Szabolcsi, 1981) the I-Nom arm-Poss-Pl-1Sg ?my arms? b. (a) P?ter- ? kar-ja-i- ? the Peter-Nom arm-Poss-Pl-3Sg ?Peter?s arms? However, as exemplified in (2), Hungarian possessors can also be marked with Dative Case, and, when it happens, the possessor obligatorily precedes the article. (2) a.?n-nek-e-m a kar-ja-i-m (Szabolcsi, 1981) I-Dat-Poss-1Sg the arm-Poss-Pl-1Sg ?My arms? b. P?ter- nek a kar-ja-i- ? Peter-Dat the arm-Poss-Pl-3Sg ?Peter?s arms? In addition, a dative-marked possessor can appear detached from the possessive description, surfacing in a position outside the nominal domain. 243 (3)[ S?? [ S? P?ter-nek [ So hossz?-ak a karja-i- ?- ?]]](Szabolcsi, 1981) Peter-Dat long-pl the arm-poss-pl-3Sg-Nom ?It is Peter whose arms are long? These facts leaded Szabolcsi to a nominal structure composed by two functional categories: [+Poss] Infl and KomP, KomP being the nominal counterpart of CP, headed by lexical categories of the determiner class. Whereas nominative possessor phrases move to spec of [+Poss] Infl, 2 dative possessors raised to spec of Komp, which serves as an escape-hatch for movement out of the nominal domain. Hence, dative-marked possessors can move away from the nominal domain, as shown in (3). Abney (1987) put forward the so-called ?DP hypothesis?. Defending an analysis similar to that of Szabolcsi, he argued that determiners head the D-N constituent, forming the structure in (4): (4) DP 2 D' D NP Coupled with the assumption that non-lexical categories also participate in X?bar schema, the DP-hypothesis unifies the clausal and the nominal domains by considering that determiners, similar to complementizers and modals, are a functional category heading its own phrase. 3 2 In her (1994) paper, Szabolcsi, building on Kayne (1993), proposes that the nominal structure is composed by an AgrP projection, where Nominative Case is assigned. 3 Abney (1987:64) lists the following reasons for classifying D as a functional element: D categories constitute a close lexical class, lack descriptive content and can be phonologically dependent. 244 In short, work by Szabolcsi and Abney gained a broad consensus mainly because it permits a conceptual unification between clauses and nominals, and, as a result, accounts for the cross-linguistic similarities between these two domains. 4,5 5.2.2What?s Between DP and NP Even though the DP-hypothesis is well accepted, the actual structure of the DP remains an open issue. Researchers have argued that the structure in (4) needs to be enriched to accommodate the empirical evidence for the existence of structural material between the DP and the NP. In particular, the spell-out position of Romance nouns has been taken as strong evidence for a more elaborate structure. It is commonly assumed that the underlying order inside nominals is Adjective- Noun (cf. Valois 1991, Picallo 1991, Bernstein 1993, Zamparelli 1995, Cinque 1994, Crisma 1996 and Longobardi 2001), among others. However, as the examples in (5) and (6) show, depending on the grammar, the noun surfaces in a position lower than the determiner and higher than some adjectives. In Romance, the noun appears sandwiched between the article and restrictive manner adjectives. Conversely, in Germanic, nouns follow restrictive manner adjectives, as shown in (6). (These examples are from Longobardi, 2001.) 4 For a summary and discussion of the empirical evidence in favor of the DP hypothesis, see Bernstein (2001). See also Ogawa (2001) for additional arguments in favor of a structural unification between clauses and nominals. 5 Larson (2004) challenges the correlation between DP and CP (i.e., Szabolcsi?s insight) or between DP and Infl (Abney?s theory arguing that D is similar to a transitive predicate). For him, the clausal correlate of a 245 (5) a.La probabile reazione ostile tedesca (Italian (Most of Romance)) the probable reaction hostile German b.The probable hostile German reaction (English (Germanic)) (6)a. Un bel vestito azzurro tedesco (Italian (Most of Romance)) a nice dress blue german b. A nice blue German dress (English (Germanic)) According to the authors cited in the last paragraph, the word order in (5a) and (6a) is derived by noun raising. Even though there is no consensus with respect to the identity of the functional projection hosting the movement, the fact that an adjective can intervene between the determiner and the noun suggests that this projection is not the DP. Therefore, granting that restrictive manner adjectives precede the noun in the underlying structure, the word order in Romance is evidence for an extra projection inside the DP. Noun raising is not confined to Romance. In considering a different type of nominal construction, Ritter (1991) claims that Hebrew nouns move to an intermediate projection too. Her argument is the following: in Hebrew, N-to-D happens only if D is null, as in the construct state illustrated in (7a). In (7b), a free state construction, the presence of an overt determiner blocks N-to-D movement, but the order Noun-Possessor signifies that the noun has left the NP. DP is a VP shell in the sense of Larson (1988, 1990, 1991) and Chomsky (1995b). I will put this possibility aside. 246 (7)a.ha-axila shel Dan et ha-tapuax the-eating of Dan of the-apple ?Dan?s eating of the apple? b.parat ikar cow farmer ?A farmer?s cow? For Ritter, in (7b) the noun moves to Num(ber)P, the base position of numerals and quantifiers. 6 Brug??s (1996) and Bernstein?s (1997, 2001) research indicates that the left periphery of nominals is also more elaborated than the structure in (4) suggests. It contains at least a functional projection in which reinforcer elements are base generated and to the left of which focused nouns are moved. (8)a.Ce livre ?ci (French)(Bernstein, 1997) 7 this book here b. Questo livro qui(Italian) this book here ?This book? 6 Bernstein (1993,2001) suggests that the plural marker that appears attached to pre-nominal adjectives below is the phonological realization of the head of the NumP projection. (i) d?s n?urs-ouy (Walloon; Bernstein 2001) some black eyes ?Some black eyes? 7 The emphasis is mine (CR). 247 Brug? and Bernstein are not alone in suggesting a more elaborated nominal left periphery. Ormazabal (1991), for example, proposes that the DP is dominated by a K(om)P, whose spec serves as an escape hatch for extraction from DPs. Ticio (2003) also assumes that an extra projection is involved in wh-extraction from DPs, however, following Grohmann & Haegeman (2002), she labels it TopP. There is much to be said about the structure of DP, but I not going to enter into that debate here because it would take us far beyond the scope of this dissertation. Suffice it to observe the existence of extra functional projections inside the DP domain, even though their identity is still an open issue. Another open issue is their number. Longobardi (2001), for instance, based on a cross-linguistic study, concluded that UG makes available two Case positions between DP and NP. Zamparelli (1995) defended the hypothesis that the DP layer splits into three projections: SDP - Strong Determiner Phrase, an individual denoting phrase, PDP - Predicative determiner Phrase, which denotes predicates, and KID - Kind determiner Phrase, which is a Kind-denoting phrase in the sense of Carlson (1977). Crisma (1990) and Cinque (1994) argued that each prenominal adjective occupies the specifier of a different functional projection. Hence, for Crisma and Cinque, the number of functional projections corresponds to the number of prenominal adjectives. 248 5.2.3 The DP Structure Adopted in This Thesis From the exposition above, one can see that the syntax of the nominal domain is not yet fully understood. Thus, since I am specifically interested in the syntactic properties of null possessors in BP, I will put aside many of the issues concerning the actual format of DPs, and assume a simplified and generic structure. First, I will take for grant the existence of an Agr-like projection whose specifier is a Case position for possessors (cf. Kayne 1993 and Szabolcsi 1994). The identity of this category does not matter for the purposes of this thesis and I will use the generic XP label. 8 Building on Bernstein (1997, 2001), I will also assume that the nominal left periphery contains an extra functional projection - FP. Therefore, the DP structure I am considering mirrors the clausal structure I used in chapter 4 (section 4.2). XP is the nominal counterpart of TP, whereas DP is an analogue of CP. Between CP/DP and TP/XP there is an FP projection. (9)a.Clausal structure 9 b. Nominal structure CP DP 2 2 FP FP 2 2 TP XP 2 2 VP NP 8 Based on clausal structure proposed by Chomsky (1993), Kayne and Szabolcsi labeled this projection AgrP. De Wit (1997) and Schoorlemmer (1998) used Ritter?s Num(ber)P label. 9 I am omitting the vP projection. 249 In Bernstein?s (1997, 2001) proposal, the nominal FP is the locus of reinforcers. Reinforcers are the nominal counterpart of sentential locative adverbs which, according to my analysis (cf. 4.3.2.2), surface in the spec of FP. Hence, across domains, FP is arguably a host for referential locative expressions. For Bernstein, reinforcers are the heads of FP, but she offers no evidence for that. Here I will explore the possibility of taking them to be in the spec of FP, rather than in the head. The reason for doing so is related to the licensing of null arguments in BP. As I showed in chapter 4, locative adverbs in the spec of an embedded FP block an antecedence relationship between an embedded null subject and a higher subject DP. The relevant data is repeated in (10). In (10a), the locative adverb aqui ?here? is in spec of FP, and, as a consequence, the embedded null subject can receive only an arbitrary interpretation. In contrast, (10b), where the locative adverb aqui is placed at the end of sentence, allows a coreferential reading; in fact, this is the only interpretation available. (10) a.[a Maria 1 disse [ CP que [ FP aqui [ F? e arb/*1 vende sapatos]]]] the Maria said-3Sg that here sell-3Sg shoes ?Maria said that shoes are sold here? b. A Maria 1 disse que e ??arb/1 vende sapatos aqui the Maria said-3Sg that sell-3Sg shoes here ?Maria said that she sell shoes in the market? I haven?t introduced BP null possessors yet, but as shown below, these possessors 250 pattern like sentential referential null subjects in that they require a sentential antecedent. (11) Maria 1 encontrou com [um e 1/*2 primo] Maria met-3Sg with a cousin ?Maria met with her cousin? Interestingly, the antecedence relationship in (11) is blocked if a reinforcer is inserted inside the DP: 10 (12)?? A Maria encontrou com [ DP um primo l? e 1 ] the Maria met -3Sg with a cousin there ?Maria met a cousin of hers? Note that (12) is fine if the null possessor is replaced by an overt pronoun: (13) A Maria 1 encontrou com [um primo l? dela 1 ] 1 the Maria met-3Sg with a cousin there of.she ?Mary met a cousin of hers? 10 I am putting aside questions about the spell-out position of the noun in (12). 1 The possessor phrase can also precede the reinforcer, as in (i). Arguably (i) involves XP movement to the left of the reinforcer, whereas (13) involves only head movement. (i) A Maria encontrou com [um primo dela 1 l?] the Maria met ?3Sg with a cousin of.she there ?Mary met her cousin? 251 The parallelism between (10) and (12) will be addressed in 6.3.2 in connection with the locality requirements on the antecedent-null-possessor relationship. But let me observe at this point that reinforcers behave like preverbal locative adverbs in being able to prevent an antecedence relationship between a DP and null argument. In view of that, I will consider that the position of reinforcers inside the nominal domain duplicates the position of preverbal locative adverbs inside the clausal domain. 5.3 Possessive Descriptions Like verbs, nouns admit satellite phrases, some of them being arguments, rather than modifiers. 12 In (14), for instance, it is arguably the case that the phrases Scipio and Carthage are arguments of the deverbal noun destruction, as argued in Chomsky (1970). (14) Scipio?s destruction of Carthage That non-deverbal nouns can take arguments is disputable (cf. Anderson 1983, among others). 13 Nevertheless, (15) is similar to (14) in that the two phrases in construction with the noun are interpreted as an agent (Rembrandt) and a theme (Saskia). Thus, some authors (cf. Longobardi, 2001) have assumed that nouns like portrait are argument-taking categories. 12 However, nouns optionally take argument, whereas verbs take them obligatorily. For a discussion, see Grimshaw (1990). 252 (15)Rembrandt?s portraits of Saskia Aside considerations on pragmatics, Rembrandt in (15) can be interpreted as the possessor of the portraits of Saskia, rather than its agent/author. 14 Hence, nouns may have another satellite: the possessor or the R-related phrases in the sense of Higginbotham (1983). A more convincing example of possessor is given in (16), where Paul might be interpreted as being in a possessive relation with the noun eyes, eyes being an inalienable possession of John. (16)John?s eyes The syntactic properties of the possessor phrase are the topic of this section. The goal is to present a short summary of the literature discussion on the possessor-possessum semantic relation. Presupposing that possessors, agents and themes are always nominal arguments, I will show the evidence we already have for assuming that the possessor is the highest argument inside the DP. This review of the literature will allow me to expose the groundwork of my explanation of possessor raising in BP. 13 In fact, Grimshaw argues that the satellites of nouns like destructions are argument adjuncts. Like arguments, these satellites satisfy the argument structure of the noun, but, like adjuncts, they are not q- marked. 253 5.3.1The Possessive Thematic Relation The existence of a semantic relation between a possessor and a possessum seems to conform to our intuitions. However, the nature of this relation is not self-evident. As Williams (1982) and Szabolcsi (1994), among others, pointed out, it can be massively vague. Take, for instance, Williams? example in (17), where the noun cat could refer to the cat that John owns, the cat that is sitting on John?s lap or even to the cat that John stepped on. The same observation is valid for (17b), Szabolcsi?s example. (17)a.John?s cat b.My train Nevertheless, Williams?s conclusion that a possessive relation can be any relation at all is misleading for it ignores the fact that there are constraints on what counts as possessive relation. Revamping Chomsky (1970), Barker (1995) and Uriagereka (1996) observed that, in possessive part-whole relations, the possessor must be the entity representing the whole, and the possessum, the entity representing the part. (18) a. The table?s leg(Barker, 1995) b. * The leg?s table c.The city?s poor neighborhoods (Uriagereka,1996) d. *The/a poor neighborhood?s city 254 It is worth noticing, however, that the nouns in (17) and (18) may not have the same semantic type (cf. Partee 1987, Barker 1995 and Partee and Borschev 2000). For Barker, there are two classes of possessive nouns, extrinsic possessives (cf. (19), which contains nouns like the ones in (17), and lexical possessives which are formed by nouns expressing part-whole relations (cf. (20)). (19)Extrinsic Possessives a. John?s cat b. John?s yogurt c. John?s firetruck (20) Lexical Possessives a.John?s purchase (Derived nominals) b.John?s child (Kinship terms) c.John?s nose (Body part terms) d.The table?s top (Generalized part/whole relations) e.The woman?s pen pal (Arbitrary relational nouns) Lexical nouns, contrary to extrinsic noun, are relational. That is, they denote relations over a pair of entities, the possessor and the possessum. This translates into a semantic approach in which lexical nouns are two-place predicates, whereas extrinsic nouns are plain one-place predicates: 255 (21)a. [ child] = lxly[child (x,y)] b. [firetruck] = ly[firetruck (y)] Barker remarks that this distinction has some syntactic consequences: with relational nouns, a possessor can show up as a postnominal of-phrase, as in (22a), but with non-relational nouns this is not allowed. (22) a. A child of John b. *A firetruck of John In Barker?s analysis, the contrast in (22) follows from the fact that relational nouns are able to take a possessor phrase as their argument, but non-relational are not. Szabolcsi (1994), not considering the difference between relational and non relational nouns, proposes that the possessor in general is assigned a q-role in the spec of PossP, a functional category that immediately dominates the NP. Particularly, it is suggested that N adjoins to the head of PossP, and the syntactic unit thus formed assigns the possessor?s q-role. While Poss has the formal ability to assign a q-role in accordance with the Theta-Criterion, N has the ability of specify the content of the q-role assigned. Uriagereka (1996), trying to subsume the possessive relations in (23) under the same syntactic analysis, suggests that these sentences share an underlying small clause structure, consisting of the possessor and the possessum. 15 256 (23) a. John?s sister b. John has a sister In this analysis, there is not inherently relational noun and the possessor is not assigned a q-role. Rather, it is proposed that a possessive relation is an abstract semantic relation R that holds within the small clause that underlies the structure of possessive DPs. Den Dikken (1998) also explores an analysis that involves an underlying small clause. However, differently from Uriagereka, he suggests that the possessor phrase is contained inside a prepositional phrase that predicates of the possessum. Hence, for him, the insertion of possessor is not thematic licensed by the possessum, and a possessive relation results from the fact that the possessor is contained by the prepositional phrase that modifies the possessum. In conclusion, there are various proposals for the semantic relation between a possessor phrase and a possessum noun. Arbitrating between them is not my goal here; rather, as have already stated, I will assume a simplified structure. Following Abney (1987), Authier (1988), Tellier (1991) and Vergnaud & Zubizarreta (1992), I will consider that inalienable possessed nouns (kinship and body part terms) are inherently relational, being argument-taking categories that q-mark the possessor. Being more specific, I will entertain the possibility of having a q-role checking relation between the possessor and the possessum: when merged with a relational noun, a possessor phrase checks a q-role feature of that noun. 15 This analysis stems from Freeze?s (1992), Kayne?s (1993) and Hornstein et al. (1994). For developments of this idea, see Muromatsu (1994) and Castilllo (2001). 257 For the time being, I will put non-relational possessions aside, assuming that possessors of relational and non-relational nouns does not differ in terms of height inside the DP. This allows me to talk about possessor of non-relational nouns as arguments, even though it has been argued that they are adjuncts (cf. Tellier 1990 and Partee and Borschev 2000). Tellier offers an interesting argument for this view. Observe the contrast in (24), which are French relativizations in which the genitive relative marker dont corresponds to two empty possessor phrases. (24)a.La fille dont 1 [ DP le e 1 p?re] ne parle plus avec [ DP la e 1 mere] the daughter of whom the father not talk-3Sg more with the mother ?the girl whose the father no longer speaks with her mother? b. * Une employ?e dont 1 [ DP la e 1 patronne] ne parle plus avec one employee of who the boss not speak-3Sg more with [ DP la e secr?taire] the secretary ?An employee whose boss no longer speaks with the secretary? Tellier analyzes this type of relativization as parasitic gap constructions because they contain a gap inside a domain that is inaccessible for extraction. In (24a), for example, the second gap (i.e. the gap inside the DP la mere) is inside a prepositional phrase that is opaque for extraction. Moreover, this gap is obligatorily interpreted as coreferent with the first gap (i.e the gap inside the DP le p?re); and this anaphoric 258 dependency is licensed only under the presence of a c-commanding A?-binder. Given that both (24a) and (24b) match the properties above, the fact that only (24a) is acceptable is quite mysterious. There is, however, a crucial difference between these two sentences: the gaps in (24a) are inside relational nouns, whereas the gaps in (24b) are inside non-relational nouns. This difference relates in an interesting way with Stowell?s (1985) observation that a parasitic gap corresponds to an argument: If only possessors of relational nouns are arguments, as Tellier proposes, then (24b) in which a parasitic gap is in an adjunct position is ruled out. 16 5.3.2 The Position of The Possessor The structural hierarchy in (25) has been proposed in the literature: (25)Possessor > Agent > Theme The evidence for this hierarchy comes from binding relations. Among the nominal arguments, if one of them is an anaphoric expression or contains one, and the other is the antecedent, then it is always the case that the possessor will be the antecedent and the agent or the theme will be (or will contain) the anaphor. If the possessor is absent, then the agent is the antecedent and theme the anaphor, as the scheme in (26) illustrates, and 16 If Tellier?s reasoning about the contrast in (24) is right, then my suggestion that the possessor is a q-role is risky because even adjuncts can be interpreted as possessors, as in (24b). This is a problem and I acknowledge it. However, it is also important to notice that, the potential problem just raised is everybody?s problem since it is traditionally assumed that arguments are q-marked, but adjuncts are not. Note further that a small-clause account of possessive relations does not fare better for it does not predict the contrast in (24). 259 (27) and (28) from BP exemplify: 17,18 (26)[Poss ?[ Agent ?[ Theme]]] (27)a. ? Uma scultura [do Rodin] agent/poss [de si mesmo] theme a sculpture of.the Rodin of himself ? Rodin?s sculpture of himself? b. A insatisfa??o do Jo?o consigo mesmo the dissatisfaction of.the Jo?o with himself ? Jo?o?s dissatisfaction of Jo?o with himself? (28)a.* Uma scultura [de si mesmo] agent/poss [do Rodin 1 ] theme the poem of himself of Rodin * ?Himself?s sculpture of Rodin? b. * A insatisfa??o de si mesmo com o Jo?o the dissatisfaction of himself with Jo?o ?Himself?s dissatisfaction with Jo?o? 17 On the relation between the hierarchy in (26) and binding relation discussed here, see Giorgi and Longobardi (1994) and Ticio (2003), among others. 18 In Colloquial BP, the reflexive form si mesmo is not frequently used. Thus, the sentence in (27a) is not totally accepted. 260 Another source of evidence in favor of the structure in (26) is wh-extraction (cf. Cinque 1980, Ormazabal 1991, Giorgi and Longobardi 1991 and Ticio 2003). As shown in (29), a theme can undergo wh-movement only if it is the only argument of the noun: (29) a.[de quem ] 1 que voc? viu [uma foto t 1 ] ( BP) of who that you saw-3Sg take-inf a picture ?Of whom did you see a picture? b.?? [sobre o que] 1 que j? voc? leu [ DP v?rios livros [do Chomsky] agent t 1 ] about what that already you copied- 3Sg several book of Chomsky ?*About what did you already copy several books of Chomsky? c.* [sobre o que] 1 que voc? j? leu [v?rios livros [da biblioteca ] poss t 1 ] about what that you already read-3Sg several books of.the library About what did you already read several books of the library? As for agents, they can be extracted in the presence of a theme but not in the presence of a possessor: (30)a. De que autor voc? leu v?rios livros [de ling??stica] theme (BP) of which author you read-3Sg several books of ling??stica ?*Of which author did you read several books of linguistics? 261 b.* De que autor voc? leu [ DP v?rios livros [da biblioteca central] Poss t 1 ] of which author you read-3Sg several books of library central ? Of which author did you read several books of the central library? Possessors, on the other hand, can be extracted in the presence of agents and themes, as the BP examples below illustrate: (31)a.? [de qual shopping] 1 voc? viu [t1 a decora??o t 1 [de natal] theme Of which shopping you saw-3Sg the decoration of christmas ? *Of which shopping center did you saw the Christmas decoration? b. De que museu voc? viu [as esculturas t 1 [da Camille Claudel] agent ] of which museum you saw-3Sg the sculptures of.the Camille Claudel ?*Of which museum did you see the sculptures of Camille Claudel? In summary, inside the DP, possessors are higher than agents and themes. As a consequence, the syntactic realization of a possessor blocks extraction of an agent or a theme, but not vice-versa. Thus, given this lack of interaction between extraction of a possessor and presence of other nominal arguments, I will ignore the presence of other nominal arguments when analyzing possessor raising. 262 5.3.3 Null Possessors In Romance, a possessor can be syntactically realized as an of-phrase (30a), but if it is a co-indexed pronoun, it is either realized as genitive pronominal form (30b) or as a null category (30c). It is noteworthy that, for some speakers, in (30c) the null possessor can be interpreted as a deictic pronoun, referring to somebody else rather than to Juan. (31) a. Juan se encontr? con [ DP el primo de Mar?a] (Spanish) Juan SE met-3Sg with the cousin of Maria ?Juan met with Maria?s cousin? b.Juan se encontr? con [ DP su primo] Juan SE met-3Sg with his cousin c.Juan se encontr? con [el e primo] Juan SE met-3Sg with cousin ?Juan met his cousin? As illustrated in (32), emphatic or contrastive co-indexed pronouns can also occur as of-phrases in Romance. But crucially, this possibility is restricted to emphatic or contrastive pronouns, as the unacceptability of (33) reveals: (32)a.Maria 1 gosta do namorado DELA 1 (Galician and Eur. Port.) the Maria like-3Sg of.the boyfriend of.she ?Maria likes HER boyfriend? 263 b. Mar?a 1 ama alnovio de ELLA 1 , no al de Ana. (Spanish) Maria love-3Sg to.the boyfriend of her, not to.the the Ana ?Maria loves HER boyfriend, not Ana?s boyfriend? (33)a.?? Maria 1 gosta do namorado dela 1 (Galician and Eur. Port.) the Maria like-3Sg of.the boyfriend of.she b. * Mar?a 1 ama al novio de ella 1 (Spanish) Maria love-3Sg to.the boyfriend of her ?Maria loves her boyfriend? Cardinaletti (1998) argues that possessive pronouns fall into the weak/strong partition that she and Starke proposed for personal pronouns in (1994). According to her, prenominal possessive pronouns are weak forms. That?s why they cannot receive contrastive stress, be coordinated or modified: 19 (34)a.* la SUA casa, non tua(Italian - Cardinaletti, 1998) the his/her house, not yours b. * la sua e tua/ sua e di Maria Casa the his(her)/his(his) and of Maria house c.la solo/proprio sua casa the only/own his/her house 19 The same is true for Spanish, cf. Picallo (1994). 264 Thus, assuming that the empty category in (31c) is a null pronoun, which is a weak form for Cardinaletti and Starke, we can speculate that Romance has a preference for weak possessive pronominal forms and that the unacceptability of (33) follows from this preference. Since in (33) the possessive is not emphatic or contrastive; a weak form can be used, blocking the insertion of a strong pronoun. This state for affairs resembles the restrictions on overt subject pronouns we discussed in chapter three. According to the Avoid Pronoun Principle, an overt subject pronoun is allowed only if it is emphatic or contrastive. Hence, if the Avoid Pronoun Principle is now taken to be a principle that regulates the use of weak and strong pronominal forms, as stated in (35), we can determine that the unacceptability of (33) follows from the Avoid Pronoun Principle. 20 (35)Avoid Pronoun Principle Give preference to weak pronominal forms In the next section, I will examine possessive DPs in BP. But, before focusing on other language, let me make a digression and consider Castillo?s (2000) analysis for (35). 20 Or from whatever deeper principle that may turns out to be responsible for its effects. For Cardinaletti and Starke (1994) this deeper principle is their economy of representations principle. Since, weak forms are structurally smaller than strong forms, they are preferred in accordance with (i). I will not discuss this issue here. (i)Economy of Representations (Cardinaletti and Starke, 1994:40) Minimize structure 265 (36)Juan levant? la mano. Juan raised-3Sg the hand ?Juan raised his hand? According to Castillo, (36) involves possessor raising to the sentential subject position, as the derivation in (37) shows: (37) [ TP Juan 2 [ T? [ vP [ v? [ VP levant? [ DP t 2 [ D? [ D 0 la] [ AgrP mano 1 [ Agrs? [ SC t 2 t 1 ]]]]]]]]] The assumption behind the derivation in (37) is that the movement of Juan is Case driven. Considering that DPs comes from the numeration with their Case feature already valued, Castillo suggests that in (37) Juan enters the derivation with a nominative Case feature. Hence, since its Nominative cannot be checked inside the DP, Juan is forced to move to spec of TP. Another assumption of Castillo?s proposal is that there is no movement into theta- positions. Following Parsons (1990), the author claims that the verb levantar ?lift? is a direct-motion type of verb, and, as such, it does not involve a causative subevent as bona- fide transitive verbs do. That is, levantar behaves like an unaccusative verb in that it does not assign an external q-role. This is the reason whereby in (37) Juan moves directly from inside the possessive DP to spec of TP. I will not question whether a movement analysis is right or not, but I would like to review Castillo?s assumption that (37) does not involve movement into theta-positions. 266 I take it that the author is assuming a configurational view of q-roles, otherwise it is unclear why movement into a theta-position is banned. However, if this is the case, Castillo needs to address the following issue: by assumption, in a theory that takes q-roles to be the result of configurations, the v-VP configuration expresses the causative/agentive q-role of the external argument (cf. Chomsky 1995b:315). Therefore, it is not obvious why this very same configuration fails to express a cause or an agent in (37). Apart from the unclear nature of its assumptions, this analysis has some unsolved empirical issues. If levantar ?lift/raise? is an unaccusative verb, why is (38) unacceptable? (38)* La mano de Juan levant? the hand of Juan raised ?John?s hand raised? Notice further that in a language like BP, where the counterparts of (36) and (38) are both acceptable, (36) does not mean the same as (38). (36) implies a volitional action from Juan?s part, whereas (38) does not. 21 This difference in interpretation is unexpected if levantar does not assign an external q-role. 21 According to my own judgments, in BP (37) is fine only in specific contexts like the one in (i), which clearly does not involve a volitional action: (i) Por causa do Parkinson, o bra?o da vov? levanta toda hora because of.the Parkinson, the arm of.the grandma raise-3Sg every time ?Because of her Parkinson?s disease, grandma?s arm raises all the time? 267 5.4 Possessives in Brazilian Portuguese I will start this section by showing the change that BP underwent in its weak possessive pronominal system. Next, I will focus on null possessive pronouns in BP grammars. 5.4.1Loss of Pronominal Possessive Forms Silva (1984) and Cerqueira (1996) show that the change in the pronominal system of BP (cf. section 3.2) also affected weak possessive forms. As show in figure 1, a paradigm with 6 genitive possessive pronominal forms (paradigm 1) was reduced to a paradigm with only three forms (paradigm 2 ? Colloquial BP). The 2 nd P possessive form tua ?your? was replaced by sua, which was originally a 3 rd P possessive pronoun. As consequence of this replacement, 3 rd P possessors are now realized by analytic genitive forms dele/deles ?of he/of them?. Notice in addition that the 2 nd Pl pronominal form vosso was also replaced by an analytic form de voc?s ?of you?. 2 Fig. 1. Change in the Possessive forms in BP Person and NumberParadigm 1Paradigm 2 1 st Psg Meu my meu my 2 nd PSg directteu your seu your 3 rd PSg Seu his dele of.he 1 st PPl Nosso our nosso our 2 nd PPl direct Vosso yourde voc?s of.you 3 rd PPl seu themdeles of.they The data below illustrate the simplified possessive system of Modern BP: 2 A 1 st Pl possessor can also be expressed by analytic form ?da gente? of we, literally, the folks?. 268 (39)a. Eu nunca vi o meu pai I never saw-3Sg the my father b. Voc? nunca viu o seu pai you never saw-2Sg the your father c. Ele nunca viu o pai dele he never saw-3Sg the father of.he d. N?s nunca vimos o nosso pai we never saw-1Pl the our father e. Voc?s nunca viram o pai de voc?s you-Pl never saw-3Pl the father of you-pl f. Eles nunca viram o pai deles they never saw-3Pl the father of.they ?I/you/he/we/they never saw my/your/his/our/their father? The data below from the 19 th -century BP suggest that at that time the pronoun sua was already used a 2 nd P pronoun: (40) a.D? c? o seu chapeu give here the your hat ?Give me your hat? 269 b. Concede-me a m?o de sua fillha give-me the hand of your daughter Give-me your daughter hand (in marriage)? (Martins Pena, O Judas no S?bado de Aleluia, 1844) Notice, however, that sua could also be used as 3 rd P possessive pronoun: (41) a.?mas prouve a Deus que eles me servissem para descobrir a sua but wanted-3 Sg to God that they me served-3Sg to find-out-Inf the her perf?dia e ouvir a tua ing?nua confiss?o disloyalty and hear-Inf the your na?ve confession But thanks God, they helped me realizing about her disloyalty and hearing you na?ve confession.? b.Estava com meu amigo Ant?nio falando de seus neg?cios?. was-1Sg with my friend Ant?nio talking of his business ?I was with my friend Ant?nio talking about his business? (Martins Pena, O Judas no S?bado de Aleluia, 1844:139) Following Kroch?s (1989) theory of coexisting grammars, I presuppose that during the 19-century there were at least two competing grammars, one with Paradigm 1 of figure 1, and another one with simplified paradigm, where sua was already a 2 nd Person pronoun. 270 Silva?s (1984) research shows that in Modern spoken Portuguese the 3 rd P analytic form dele/dela ?of.he/of.she? occurs 75% of the time, whereas sua as third person pronoun occurs only 14.1% of the time. 23 Cerqueira (1996) interprets this simplification in the possessive pronominal system as correlated with the impoverishment of Agr in BP. Assuming a layered DP ? la Kayne (1993) and Szabolcsi (1987), he argues that possessive pronominal forms are licensed only in languages in which Agr contains a person feature, which is taken to be crucial for the licensing of the possessive pronoun. In Romance, Agr has a person feature, thus Romance possessive pronouns moves to spec of AgrP and have genitive Case feature checked. (42) Romance DP 2 D AgrP 2 seu Agr? your Agr +person NP pai t 1 father 23 It is worth mentioning Negr?o and M?ller?s observation (citing Almeida 1993), that in Modern BP, when the antecedent is non-specific or generic, the pronoun sua can be used as 3 rd P pronoun, as the data in (i) show. I will put this observation aside, waiting for future research. (i)a. ?um rapaz que paga seus estudos com sacrif?cios ele n?o pode ter dinheiro a guy that pay-3Sg his studies with sacrifice he not can-3Sg have-inf money nunca para ir ao teatro never to go-ing to.the theater ??A guy that pays his studies with sacrifice cannot have money to go to the theater? b. Acho que a televis?o Brasileira (?) ir? encontrar seu caminho think-1Sg that a television brazilian will-3Sg find-inf its way ?I think that the Brazilian television will find its way? 271 Assuming that an Agr lacking a person feature is unable to assign Case, Cerqueira suggests that Agr in Modern BP lost its person feature, being, consequently, unable to assign genitive Case. As a result, the possessive pronoun stays in its base generated position, receiving genitive Case from the proposition de ?of?. 24 (43) Brazilian Portuguese DP 2 D AgrP Agr -person NP pai dele father of.he In Cerqueira?s analysis, BP loss of referential null subjects is related to the loss of possessive form because both the nominal and the sentential Agr became weak, being, consequently unable to license weak pronominal forms. All in all, leaving aside the details of Cerqueira?s analysis, the relevant issue here is the correlation between the effects of loss of morphology inside the nominal and the clausal domain. Thus, given the content of this thesis, main question is about the availability of null possessors in BP Colloquial Grammars. (Negr?o and M?ller 1996) 24 Since, the second paradigm in fig. 1 retained a possessive pronoun form the 1 st and 2 nd Person, Cerqueira is forced to assume that Agr related to 1 st and 2 nd persons still has number feature in BP. 272 5.4.2Brazilian?s Null Possessor A crucial difference between Brazilians and speakers of Romance pro-drop languages is that Brazilians allow strong possessive forms that do not carry a contrastive stress or emphasis (cf. (44), whereas speakers of pro-drop languages do not, as illustrated in (33). 25 (44) a.A Maria 1 gosta muito do pai dela 1 the Maria like-3Sg a.lot of.the father of.she ?Maria likes her father a great deal? b.A Maria 1 adora pentear o cabelo dela 1 the Maria love-3Sg comb-3Sg the hair of.she ?Maria loves to comb her hair? This supports the hypothesis that BP colloquial grammars have become insensitive to the Avoid Pronoun Principle as a result of losing its pro-drop properties. This is the hypothesis I defended in chapter 3, and there I face two issues: (i) 1 st P null subjects are allowed in matrix clauses; and (ii) 3 rd P referential null subjects are allowed inside embedded clauses if provided with a sentential antecedent. In the course of chapters three and four, I argued that (i) and (ii) are not counter-evidence to the hypothesis I proposed 25 However, as the unacceptability of the sentences in (i) show, a strong pronoun is not allowed if it?s co- indexed to a quantifier expression. This follow from Montalbetti?s (1984) constraint (cf. 3.3.2, fn.16): (i)a. ?? [nenhuma mulher interesseira ] 1 ama o marido dela 1 no woman selfish like-3Sg the husband of.her ?No selfish woman loves her husband? b. ???mas [qualquer mulher s?] 1 ama o filho dela 1 but any woman sane love-3Sg the son of.her ?But any sane woman loves her son? 273 since the allowed null subjects are residue of movement, rather than null pronouns. Interestingly, the investigation of possessive DPs in BP brings back the same difficulties: (i?) for some speakers, a 1 st P null possessor is licensed even in the absence of a sentential antecedent, as illustrated in (45a); (ii?) 3 rd P null possessors are allowed if provided with a sentential antecedent as in (45b): (45)a.[ DP o pai e] ? uma pessoa muito querida the father is-3 rd P a person very lovely ?My father is an lovely person? b. As meninas 1 adoram [ DP o pai e 1 ] the girls adore-3 rd Pl the father ?The girls love their father? In the remainder of this dissertation, I will show that, likewise to (i) and (ii), (i?) and (ii?) should not be taken as evidence that Modern BP still licenses null pronouns. Similar to sentential null subjects analyzed in the previous chapters, the null categories in (44) should be analyzed as residues of movement, rather than as null pronouns. 5.4.31 st P Null Possessors: Cases of Pronoun Deletion In section 3.4, I suggested that 1 st P referential null subjects in BP and Finnish are the result of pronoun zap (Ross 1982). Following Huang?s (1984) topic deletion 274 operation, I suggested that 1 st P in BP and Finnish results from topic deletion. A 1 st P pronoun moves to a topic position, which is a target of deletion. Assuming this on the right track, here I want to consider the possibility of analyzing 1 st P null possessors along the same lines. That is, inside the DP, 1 st possessive pronouns can be topicalized, and, as a consequence, deleted. While, in the clausal domain, there are direct ways of testing subject movement to a topic position, it is not easy to test topicalization of possessor phrases inside the DP. For instance, in chapter 3 (3.4), it was shown that wh-movement of any other constituent blocks topicalization of the subject. Thus, in chapter 3 (section 3.4), I took the complementary distribution of 1 st P null subjects and wh-movement as evidence that 1 st P null subjects are the result of topic deletion in BP and Finnish. This test cannot be used to verify a topic-deletion analysis of 1 st P null possessors. As shown in 5.3.2, the presence of an overt possessor blocks wh-movement of other arguments of the noun. That is, even overt possessor phrases are incompatible with wh-movement. Therefore the wh- movement test says nothing about the possibility of topicalization and consequent deletion of a 1 st P possessive pronoun. 26 There is, however, indirect evidence for a topic deletion analysis of BP 1 st P null possessors. In BP, in possessive descriptions containing a possessive pronominal form like (44a), a chunk of the DP containing the noun and the possessive pronoun might be topicalized, surfacing in position that precedes the reinforcer, as shown in (46). If possessive pronouns move to spec of XP as I claimed in 5.2.3., then in (46), the XP node is the constituent under movement, as the structure sketched in (46b) shows: 26 Notice further that body-part and kinship nouns take only an argument: the possessor. This make the wh- movement test even harder (if not impossible) to be done. 275 (46) a. Esse meu pai aqui ? uma pessoa muito querida. this my father here is-3Sg a person very lovely ?My father here is lovely a person.? b.[[ DP esse [ XP meu 1 [ NP t 1 pai]] 2 [ FP aqui [ F? t 2 ]]] ? uma pessoa muito querida] Interestingly, in (46), the possessive pronoun must be overt. Witness the unacceptability of (47): (47) * [ DP Esse e pai aqui] ? uma pessoa muito querida this father here is-3Sg a person very lovely ?My father here is lovely a person? If 1 st P null possessors are residues of topic deletion, the unacceptability of (47) follows: The topicalization of the whole XP blocks topicalization of the pronoun and, consequently, deletion does not apply to the pronoun. At any rate, this suggestion clearly requires further investigation, but the fact that 1 st P null possessors are incompatible with topicalization of other constituents is a hint towards an analysis along the lines I am proposing. 276 5.5 Possessives in Finnish 5.5.1 Possessive Agreement Morphology and Availability of Null Possessors Some languages display overt morphological agreement between the possessor and the possessum and unsurprisingly license null possessors. Consider for instance the Turkish data in (48), extracted from Kornfilt (1984): (48)a.ben-?m/pro ?stakoz-um I-Gen lobster-1Sg b.biz-?m/pro ?stakoz-umuz we-Gen lobster-1Pl ?My/our lobster? Standard Finnish also exhibits overt agreement between nouns and possessors, the so-called possessive suffixes. Conversely, Colloquial Finnish does not display morphological agreement of this sort (cf. Vainikka (1989). 27 The contrast between the two dialects is given in figure 2. 28 27 I am putting aside the colloquial dialect spoken in Tampere, where, as reported by Vainikka, possessive suffixes exist for the singular, but not the plural. 28 As apparent in figure 2, in CF the possessive pronouns underwent morphological reduction too. I believe this is to be correlated with the loss of possessive suffixes, though I will not investigate it here. 277 Fig. 2. ?my/your/his/her/our/your/their cat? Person and Number Standard Finnish (SF) Colloquial Finnish (CF) 1 st Psg Minun kissa-nimun kissa-0 2 nd PSg directSinun kissa-sisun kissa-0 3 rd PSg H?nen kissa-nsa 29 se kissa-0 1 st PPl Meid?n kissa-mmeMei?n kissa-0 2 nd PPl direct Teid?n kissa-nneTe?an kissa-0 3 rd PPl Heid?n kissa-nsaniiten kissa-0 Another difference between SF and CF is the availability of null possessors. In SF, for all combinations of person & number, except for 3 rd P Singular and plural, the possessive pronoun can be dropped (cf (49)). In CF, this is not possible, as shown in (50). (49)a. (minun) kirjani(SF ? Vainikka, 1989) my book-1 st Sg b. (sinun) Kirjasi your book-2 nd Sg c. *(h?nen) kirjansa his/her book-3 rd Sg d. (meid?n) kirjamme our book-1 st Pl 29 The third person possessive suffix has the allomorph ?Vn, where V is a vowel copied from the stem, see examples in (i): (i)a.h?nen veljen he-Gen brother-Part-3 rd Sg ?His brother? b. h?nen ksivarttaan he-Gen arm-Part-3 rd Sg ?His arm? 278 e. (teid?n) kirjanne your book-2 nd Pl f. *(heid?n) kirjansa their book-3 rd Sg (50)a. *(mun) kirja. (CF ? Vainikka, 1989) I-Gen book b. *(sun) kirjasi you-Gen book c. *(sen) kirja his/her book d. *(mei?n) kirja our book e. *(tei?n) kirja your book f. *(niiten) kirja their book Under the hypothesis that licensing of null pronouns are correlated with the presence of rich morphology, the ungrammatically of (50) is arguably due to the loss of possessive agreement morphology in CF. Vainikka (1989) reports a third difference between Standard and Colloquial 279 dialects of Finnish. In SF, but not in CF, overt genitive possessive pronouns obligatorily obviate in reference from local DPs. That is, in standard grammars, overt genitive pronouns are interpreted only as pronouns, never as anaphors that refer to a syntactic local antecedent. Contrastively, colloquial grammars seem to accept an anaphoric interpretation. (51)a.Min? l?ysin ??(minun) rahakukkaroni (SF) I-Nom found-1 st Sg my wallet-1 st Sg ?I found my wallet? b.Maija 1 hukkasi h?nen 1/*2 kasettinsa Maija-Nom lost-3 rd Sg her-Gen tape-3 rd Sg ?Maija lost her tape? (52)a.M? k?velytin mun koira (CF) I-Nom walked-1 st Psg my-Gen dog ?I walked my dog? b. Jukka k?velytti se koira Jukka-Nom walked his-Gen dog ?Jukka walked his dog? This can be seen as a reflex of the Avoid Pronoun Principle (cf. (35)). In SF the availability of null possessive pronominal forms blocks the use overt pronouns. CF does 280 not license null possessive pronouns, hence the Avoid Pronoun Principle is arguably not operative in these grammars. Note that this conclusion is not straightforward, though. It predicts lack of obviation effects for 3 rd P possessive pronouns in SF. As illustrated in (49), 3 rd P null possessors are not allowed in SF. Hence an overt 3 rd P pronoun should be insensitive to the Avoid Pronoun Principle, but, as shown in (51b), they are not. Figure 2, SF 3 rd Sg and 3 rd Pl agreement is morphologically indistinguishable. That is, this particular agreement marker is morphologically weak, not instantiating number distinctions. Therefore, unavailability of the 3 rd P null pronouns in (49) is expected under the hypothesis I am arguing for. But interestingly, a 3 rd P null possessor is licensed in SF if provided with a sentential antecedent, as (53) illustrates. That is, 3 rd P null possessors pattern like anaphors, differently from its overt analogue (cf. (51)). (53) a. Jukka 1 k?velytti kassaansa 1/*2 . (Vainikka, 1989) Jukka-Nom walked-3 rd Sg dog-3 rd Sg ?Jukka walked his own dog.? b. He 1 tulevat autollaan 1/*2 they-Nom coming-3 rd Pl car-3 rd Sg ?They are coming in their own car.? This state of affairs is quite similar to that described in chapter 3 with respect to sentential null subjects: when the agreement morphology becomes impoverished, referential null pronouns turn into anaphoric gaps. 281 I will put aside SF 3 rd P null possessors for a while, returning to them only in appendix of chapter 6, where I will discuss evidence both in favor of and against analyzing these element as residues of NP-movement. In the next section, I will rather offer some speculations on the nature of possessive suffixes in Finnish. Sketching an analysis along the lines suggested in 4.2 for verbal agreement morphemes, I will take SF 1 st and 2 nd P possessive markers to be clitic-like categories at the syntax level. This amounts to saying that the structure of the relevant DPs in (49) does not involve the presence of a null pronoun. The possessive agreement marker itself is the element that saturates the possessor argument position. 5.5.2 Speculations on Finnish Possessive Suffixes Possessive agreement markers are not an exclusive property of nominals, they occur in a variety of syntactic environments. For example, they appear with past participle forms, which can, as a result, function as the main verb of an embedded clause as shown in (54a). They can also show up with infinitival forms inside adjunct clauses (54b), adjectives (54c), and also with pospositions (54d): (54)a.Jorma kertoo tuleva-nsa Jorma say-3 rd Sg come.Past part-3 rd SgPoss it ?Jorma said he will come? 282 b.Kompastuin joustessa-ni stumble-past-1 st Sg run-Inf-Iness-1Sg ?I stumbled while running? c.kaltaisekse-en Jumala loi ihmisen like-Trans-3 rd Sg God made-3 rd Sg men ?God made men like himself? d. Tulet kanssa-mme come-2 nd Sg with-1 st Pl ?You will come with us, won?t you?? These many functions of possessive markers have leaded scholars to debate about their morpho-syntactic nature. Pierrehumbert (1980) and Nevis (1984, 1985), for example, have argued that these suffixes are clitics. They observe that similarly to clitics, possessive markers always follow inflectional and derivational suffixes. Kanerva (1987) represents the structural order in which these elements appear as follows. (55)a.Nouns: stem>(number)>(Case)>(Possessive suffix)>(clitics) b. Verbs: stem>non-finite suffix 30 >(case)>(possessive suffix)>(clitics) 30 These suffixes are: (i) -ta first infinitive -te second infinitive -ma third infinitive -va (non-part) participle -nee/-nutpast participle -ttu passive past participle 283 Nevis (1984) also notices that possessive suffixes do not trigger consonant gradation, a word-internal phonological rule that lenites obstruents in the onset of a closed syllable. Hence, these suffixes are arguably outside the word domain, being thus clitics, rather than suffixes. Kanerva (1987) discusses the possessive markers under discussion at length and, contrastively, concluded that are better analyzed as suffixes. First, she remarks that the word order in (55) does not decide between a clitic and a suffix analysis. Possessive markers do pattern like clitics in that they follow suffixes; but, on the other hand, they are like suffixes in that they precede clitics. Kanerva also offers several pieces of empirical evidence showing that these markers behave as suffixes at PF. I will not reproduce all of her arguments here, but for the benefit of the discussion I will bring some of them forward. First, the phonology of Finnish rules out words that begin with consonant clusters like mm and nn. Nevertheless, the 1 st P and 2 nd P possessive suffixes (-mme and - nne) begin with these clusters. Thence, analyzing them as clitics requires stipulating that phonology treats them as exceptions. Second, the morphological behavior of possessive markers resembles that of suffixes in that they can condition allomorphy in the stem that they attach to. (56), for instance, illustrated that the word nainen ?woman? exhibits stem- final allomorphy: -nen becomes ?se before derivational and inflected suffixes. This change also occurs before possessive suffixes, but not after clitics, as shown in (57). (56)a.naise-llinen woman-like (or feminine) 284 b.naise-lla woman- Adess c.naise-n woman-Gen (57)a. naise-nsa /*naine-nsa woman-3 rd Sg b.*naise-o/ nainen-ko woman-question-clitic Additional morphological evidence comes from compound-formation. Finnish compounds might be formed from both inflected and uninflected stems, and the example in (58) suggests that they can also be formed by words containing a possessive suffix. (58) Jorma n?ytt?? hyvin itse-e-ns?/tyytyv?ise-it? miehelt? 31 Jorma-Nom look-3 rd Sg very self-Illa-3 rd P/satisfied-Ablat man-Ablat ?Jorma looks like a very self-satisfied man? Clitics, on the other hand, do not condition allomorphy processes and do not appear within compounds. Therefore at the word-formation level, possessive markers are treated as suffixes, not as clitics. 31 The deverbal adjective tyytyv?inen ?satisfied? case marks its complement with Illative case, explaining why itse ?self? is marked with illative Case. 285 At any rate, Kanerva is arguably right in claiming that Finnish possessive markers are suffixes at PF. However, she does not present any evidence that these markers are not clitics at the syntax level. Vainikka (1989) actually suggests that these markers are syntactic units that do not depend on their morphological hosts. She takes into consideration possessive markers inside the DP domain, and assumes that they are lexically inserted in spec of NP, where genitive Case is assigned. That the implementation of her analysis is more complex and less clear than just described for her suggests that possessive markers is co-indexed with a pronoun already in the lexicon, and, for these reason, in the syntax these two items are base generated together forming a complex NP that is inserted in spec of NP. For concreteness, she argues that the possessive phrase h?nen tuolinsa ?his/her chair? has structure in (59), where the possessive marker attaches to the possessum via affix hopping. (59) NP (Vanikka, 1989:209) 2 NP N g ? ? touli-nsa g g h?nen As for 1 st and 2 nd Person possessive markers, Vainikka stipulates that they are generated inside a DP containing an implicit binder, which can be taken to be a null pronoun. This is very much in the spirit of our proposal for verbal agreement morphology 286 (cf. 4.2) and I take it to be on the right track, though I recommend we drop the stipulation that possessive markers are generated inside a complex NP containing a pronominal form. Taking seriously the idea that agreement markers can function as syntactic clitics, we can have a simplified version of Vanikka?s analysis. Putting it in a bold way, I consider Finnish nominals containing 1 st P and 2 nd P possessive markers to be structurally similar to Romance finite clauses (cf 4.2). The possessive marker is a minimal maximal projection that starts the derivation as an independent lexical item and is arguably furnished with Case feature. It is first merged in the possessor theta-position, and then adjoins to X, the nominal counterpart of Infl, in order to have its Case feature checked. To illustrate, a possessive DP like kirjani ?my book? would have the structure in (60), where I am taking for grant the N(oun)-to-X movement. (60) DP 2 D XP 4 X NP 2 kirja 2 X t 1 t 2 book 2 ni 1 X 1 st Sg This analysis raised an issue concerning the Case feature of the possessive marker. Finnish displays overt Case morphology; therefore, the assumption that possessive markers bear Genitive Case features faces a question about the lack of morphological realization of this Case. I will not explore this issue here, however it might 287 be related to the fact that the possessive markers under consideration are word-internal units (suffixes) at PF. Hence, they might be invisible for the process of Case realization. At any rate, the analysis sketched above provides an account for the pronominal behavior of 1 st and 2 nd P possessive suffixes in Finnish without invoking the existence of explicit binders. As for 3 rd P possessive marker, I will suggest in the chapter 6 (cf. appendix) that similarly to what happens to the 3 rd P verbal agreement morpheme in BP and Finnish, this agreement marker, being morphologically weak, became lexically reanalyzed as nominal suffix, losing, thus, its syntactic independency. 5.6Conclusions The core of this chapter was dedicated to following aspects of the nominal domain: (61)a.The structure of nominal is parallel to that of clauses, and possessors are arguably the nominal counterpart of clausal subjects. b. Relational nouns are argument-taking categories furnished with a q-role that is checked by the possessor phrase, which might be syntactic realized by null pronoun. c. The correlation between weak morphology and loss of null pronominal categories is also attested inside the nominal domain. Weak morphology is observed inside DPs in BP and SF and, as a consequence, in these 288 grammars 3 rd P null possessors behave as anaphors, rather than as pronouns. In the Next chapter, I will argue that these anaphoric null possessors are formed by A-movement, similarly to the sentential null subjects discussed in chapter 4. 289 CHAPTER 6 290 A-MOVEMENT OUT OF NOMINALS 6.1Preliminaries All sentences in (1) display the possessor phrase outside the nominal vicinity. That is, the possessor phrase appears as a verbal satellite, rather than as a satellite of the noun. For that reason, these are often called external possessor constructions. (1)a.Gil sataf le-Rina et ha-panim (Hebrew ? Landau, 1999) Gil washed-3Sg to-Rina Acc the-face ?Gil washed Rina?s face? b. Julie l?a frapp? dans le ventre (French ? Tellier, 1990) Julie him has-3Sg hit-3Sg in the stomach ?Julie hit his stomach? 291 c.Juan le levant? la mano a Mar?a (Spanish ? Castillo, 2000) Juan cl lifted the hand to Mar?a ?Juan lifted Maria? hand? d.John-i Mary-lul phal-lul ttayryessta (Korean ? Yoon, 1990) John-Mom Mary-Acc arm-Acc hit ?John hit Mary?s arm? Different accounts for these constructions have been proposed, many of them assuming a non-movement analysis. For example, Guer?n (1985) for French, and Borer and Grodzinsky (1986) for Hebrew postulate the existence of a bound anaphoric null category inside the possessive DP. The essence of this proposal is also shared by Cheng & Ritter (1987) and Tellier (1990, 1991). Tellier, for instance, argues that this anaphoric category is a null operator in spec of DP, as represented in (2). Zubizarreta and Vergnaud (1992) also defend a similar analysis, suggesting that the external possessor binds (? la Williams, 1980) the empty argument position of the possessum noun. (2) Elle lui 1 a coup? [ DP Op 1 [ D? les cheveux t 1 ]] (Tellier, 1991) she him had-3Sg cut-3Sg the hair ?She cut her hair? Contrastively, Landau (1999) and Castillo (2001) advocate in favor of movement. Landau?s arguments bear on the fact that external possessor constructions ? particularly 292 in Hebrew ? are characterized by (i) having the external possessor c-commanding the possessum, as in (3); (ii) not allowing the possessum to be the external argument of the verb, as in (4) and (iii) requiring the possessor-possessum relationship to be strictly local. (5), for instance, illustrate that the possessor and the possessum must be clause mates and may not be separated by more than one DP: (3)a.Gil lixlex [le-Rina] 1 et [ha-xulca e 1 ] Gil dirtied-3Sg to-Rina Acc the-shirt b.* Gil lixlex et [ha-xulca e 1 ] [le-Rina] 1 Gil dirtied-3Sg Acc the-shirt to-Rina ?Gil dirtied Rina?s shirt? (4) * [ha-kelev e 1 ] hitrocec [le-Rina] 1 the-dog ran.around-3Sg to-Rina ?Rina?s dog ran around? (5)a.Jean lui 1 semble avoir lav? [les cheveux e 1 ] Jean him-Dat seem-3Sg have-Inf washed-3Sg the hair ?Jeans seems to have washed his hair? b. Gil ripe [le-Rina] 1 et [ha-ima e 1 s&el [ha-gur e 2 ]] Gil cured-3Sg to-Rina Acc the-mother of the-puppy ?Gil cured the puppy?s mother which belongs to Rina? 293 Landau assumes that these possessors move to spec of VP where they have their Dative Case feature checked. Hence, (4) is ruled out by a prohibition against lowering. Since Spec of VP is lower than spec of vP (the position occupied by the external argument of the verb), its occupant (the external possessor) cannot originate inside the verbal external argument, which is arguably in spec of vP, a position higher than spec of VP. The explanation for (5) is straightforward: movement is strictly local; therefore these are impossible derivations. In Castillo?s and Landau?s analyses there is no movement into theta-positions; hence the displaced possessors in (1) are not q-marked by the verb. 1 Landau presupposes that a possessor of a non-relational noun is base generated in spec of DP, where it is assigned a q-role by D, and then moves to spec of VP in order to check its Dative Case against V. 2 Hence, the chain thus formed conforms to the Chain Condition: its tail is q- marked and its head is Case-marked. Given this set of assumptions, a question for Landau and Castillo is about external possessors in double object constructions. Is it the case that a possessor cannot surface as the indirect object of a ditransitive verb? Neither Landau nor Castillo address this question directly, but Landau presents an empirical argument against having a goal or a benefactive argument accumulating a possessor q-role. The argument is mainly tied to the availability of bound variable readings for (6). Whereas (6a) favors a bound variable reading, (6b) does not. That is, in (6a) values of girls determine the values of tokens of ?the prettiest shirts?, but in (6b) ?the prettiest shirt? is interpreted as a unique token. 1 But see Kayne (1975), among others for arguments that dative external possessors are to be analyzed as a malefactive/benefactive argument of the verb. Orhle and Nishio (1981) defend a different view, arguing that the malefactive reading comes from our knowledge of the word. 294 (6)a. Gil saraf le-kol yalda et [ha-xulca haxi yafa] (Landau, 1999) ?Gil burned every girl?s prettiest shirt? b. # Gil natan le-kol yalda et [ha-xulca haxi yafa] Gil gave-3Sg to-every girl Acc the-shirt the-most pretty ?Gil gave to every girl the prettiest shirt? Thus, since empty categories behave semantically as variables, allowing a bound variable reading when the antecedent is a quantificational DP, it seems that (6a) contains an empty category inside the DP ?the prettiest shirt?, but (6b) does not. There is, however, a problem with this argument. The test in (6) is not quite accurate because the author did not control the type of ditransitive verb he used. Due to its lexical semantics, the verb give is incompatible with a possessive reading. A giving event results in possession; therefore the existence of a prior possession is pragmatically awkward. (7) shows that this is a relevant observation. When give is replaced by another verb (return in (7)), speakers do allow a bound variable reading. 3 (7) Ha-mora hexzira lekol yalda et ha-meil ha-xam beyoter the teacher-Fem returned-3Sg to.every gir Acc the-coat warm most ?The teacher returned to every girl her warmest coat? 2 For a brief summary and comments on Castillo?s analysis, see section 5.3.3. 3 The same observation is valid for Spanish, as (i) shows. Some of the speakers report that the presence of a comparative adjective makes the bound variable reading harder. (i) La profesora devolvi? a cada alumna la chaqueta m?s abrigada the teacher-Fem returned-3Sg to every student-Fem the coat most warm ?The teacher returned to every student her warmest coat? 295 In other words, (7) suggests that double object constructions do allow a possessive relation with the external possessor being the indirect object of the verb, the goal participant of the given event. If this is right, a movement analysis of external possessors is forced either to admit movement into theta-positions or to assume that the possessive relation in (7) is the result of a different mechanism, say, an anaphoric binding relation. This last solution has the drawback of not explaining why possessive relations in double object constructions like (7) display the restrictions given in (3)-(5) which Landau?s takes to be evidence for movement. In (8a), for example, the indirect object does not c-command the direct object and, as a consequence, a bound variable reading is not possible. This shows that the goal cannot be interpreted as the possessor of the entity expressed by the theme if structurally it does not c-command the theme. (8b) is parallel to (4): the possessum cannot be the external argument of the verb. And (8c) shows that locality is also crucial in licensing a possessive reading in double object constructions. According to the speaker?s judgments, one cannot interpret every girl as the possessor of the mother without interpreting it as the possessor of the pictures too. That is, the possessor and the possessum cannot be separated by more than one DP. (8)a.# Ha-mora hexzira et ha-meil ha-xam beyoter lekol the teacher-Fem returned-3Sg Acc the-coat warm most to.every yalda girl ?The teacher returned to every girl her warmest coat? 296 b.# Ha-kelev hexzir lekol yeled et hakadur the-dog returned-3Sg to-every boy Acc the-ball ?The dog returned to every boy the ball? c. Gil hexzir lekol yalda et hatmuna [s&el haem] Gil returned-3Sg to-every girl Acc the.picture of.the mother ?Gil returned to every girl the/her mother?s picture? In sum, if (3)-(5) serve as evidence for movement, so does (8). The discussion above indicates that the ban on possessor raising into a theta- position is at odds with the empirical data. More evidence for possessor raising into theta- positions is the empty 3 rd P null possessor in BP which is the topic of this chapter. The chapter is parallel to chapter four and its sections are organized as follows. 6.2 shows the technical details of the movement analysis I am presenting for consideration. 6.3 contains empirical support for this analysis and 6.4 discusses the licensing of null possessors inside adjuncts. Similarly to what was proposed for null subject in 4.5, in 6.4.1, I will suggest that 3 rd P null possessors inside adjuncts are derived via sidewards movement. In 6.4.2, null possessors in across-the-board configurations will discussed. 6.5 presents the outline of the semantic effects of possessor raising. Exploring the idea of context-dependency ? la Higginbotham (1988) and Raposo and Uriagereka (2002), in 6.5 I will offer some remarks on how to proceed from the syntax to the semantics. 6.6 contains the interim conclusions. Before showing the technical details of my analysis, let me briefly show that BP 297 grammars allow possessor-raising, independently of our assumptions on movement into theta-positions. First, consider the examples in (9), discussed by Galves (1999) Lobato (2000) and Lunguinho (2003), among others: (9) a.O rel?gio quebrou o ponteiro the watch broke-3Sg the hand ?The watch, its hand broke? b.A geladeira estragou o termostato the fridge damaged-3Sg the thermostat ?The fridge, its thermostat damaged? In these cases, the possessor is spelled-out in the sentential subject position. Evidence for that comes from verbal agreement. As shown in (10), the external possessor controls the verbal agreement as sentential subjects do. 4 4 Another evidence is given in (i) where the possessor is a quantifier phrase that cannot be topicalized: (i) a.Alguns rel?gios quebraram o ponteiro some watches broke-3Pl the hand ?Some watches had their hands broke? b.* Alguns rel?gios, a Maria quebrou some watches the Maria broke-3Sg ?Maria broke some watches? 298 (10)a.Esses rel?gios quebram sempre o ponteiro menor these watches break-3Pl always the hand smaller ?These watches, their smallest hands always break? b. * Esse rel?gio quebraram os dois ponteiros this watch broke-3Pl the-pl two hand-Pl ?This watch, their hands broke? Another important observation is that (9) and (11) have the same propositional content. That is, apart from considerations on topic-hood, these sentences seem to have the same semantic interpretation. (11)a. O ponteiro do rel?gio quebrou the hand of.the watch broke-3Sg ?The watch, its hand broke? b. A porta da geladeira quebrou? the door of.the fridge broke-3Sg ?The fridge, it?s door broke? As Lunguinho (2003) argues, this paradigm suggests that these are unaccusative constructions with two possible derivational outcomes: (12a) in which the possessor is sub-extracted and moved to spec of TP, and (10b) where the whole possessive DP is raised to spec of TP. 299 (12)a.[ TP [o rel?gio] 1 [ T? [ T0 quebrou 1 , T 0 ] [ VP t 1 [ DP o ponteiro t 1 ]]]] (=(9a)) the watch broke-3Sg the hand b. [ TP [ DP o ponteiro do rel?gio] 1 [ T? [ T0 quebrou 1 , T 0 ] [ VP t 1 ]]] (=(10a)) the hand of.the watch broke-3Sg Mechanical details apart, it is important to consider closely that (10a) matches standard cases of possessor raising: the possessor phrase is spelled-out in a Case position outside the DP domain and linked to a single theta-position inside the DP. 5,6 Of course, this possibility raises question about the nominal domain in BP. Since the DP is a domain where a possessor can have its Case feature checked, any analyses defending (10) needs to explain why possessors in BP can escape the DP domain and have their Case feature checked by a sentential Case checker. 7 In the next section, I will provide an answer to this question, suggesting that this is to be correlated with the loss of morphology discussed in chapter 5. 5 (If 10)) is an unaccusative structure, explaining how the possessive DP has its Case checked is not a trivial matter. I will not discuss the issue here, but it is worth noticing that this might be case of double Nominative Case checking relation, similarly to what happens in Japanese (cf. (i)) and Korean. (i) Kono konpyuutaa-ga monitaa-ga kowareta (koto) this computer-nom monitor-nom broke (fact) ? (the fact that) this computer?s monitor broke ? 6 Notice that, this analysis predicts that possessor-raising of this sort is possible in passive construction where the possessive DP surfaces in the object position and the possessor in spec of TP. This prediction is borne out, as the acceptability of (i) suggests. Notice that in (ib), the past participle agrees with the possessor. On past participle agreement and object movement, see 4.4. (i) a. Esse carros foram trocados o motor these car-Pl were changed-Pl the engine ?These car, their engines were changed? b. Esse computador foi concertado a tela this computer was-3Sg fixed the screen ?This computer, its screen was fixed? 7 Of course the same question is relevant for the analysis of (1), which I am putting aside here. 300 6.2 Implementing the Proposal 6.2.1Loss of Nominal Morphology and Null Possessors As already shown in chapter 5, in languages like Turkish, nouns exhibit overt agreement with the possessor phrase and null possessors are licensed. (13)a.ben-?m/pro ?stakoz-um I-Gen lobster-1Sg b.biz-?m/pro ?stakoz-umuz we-Gen lobster-1Pl ?My/our lobster? In Romance, nouns display only overt number agreement with the possessor (cf. (14), but possessors can still be null, as we discussed in chapter 5. (14) a. nossa filha(EP) our-Sg daughter-Sg b. nossas filhas/*filha our-Pl daughter-Pl/ daughter-Sg ?Your daughter? 301 Given that, I will postulate that Romance has a possessive agreement marker (possessive Agr, henceforth) similar to that of Turkish, the difference being that only the plural number feature of Agr is overtly realized in Romance. That is, in Romance the possessive Agr is phonetically null (pro) when its number feature is valued as singular. It is possible that this Agr is a clitic form in the sense of Cardinaletti and Starke (1994). This would explain why it requires an antecedent prominent in the discourse. The data in (15) illustrate this point. Though they prefer an overt possessive pronoun, Spanish speakers allow the possessor in (15) to be null as long as it refers to the topic of the conversation. 8 . (15) ? Parece que [el padre e] se muri? ayer seem-3Sg that the father SE died yesterday ?It seems that her father died yesterday? We can hypothesize that the possessive Agr under consideration is like the verbal Agr discussed in 4.2 in that it may enter the derivation as an independent item. Being able to check the possessor q-role, it is first merged with the relational noun, adjoining then to X (cf.5.2.3), which also hosts the noun. 9 Consider now Modern Colloquial BP, which has undergone a reduction in its possessive pronominal forms (cf 5.4). As exemplified in (16), these grammars have also lost number agreement between a noun and a possessor: 8 Though a preference of an overt possessive pronoun is reported. 9 For evidence that Romance nouns move inside the DP, see 5.2.2 302 (16)a.minhas coisa My-pl stuff-Sg ?my stuff? b. minhas filha my-pl daughter-Sg ?my daughters? According to Cardinaletti and Starke (1994), pronominal forms form a tripartite paradigm from which the hierarchy in (17) is obtained. (17) Clitic ? weak pronouns ? strong pronouns The essential idea is that clitics are morphologically and structurally deficient versions of weak pronouns, which, in turn, are morphologically and structurally deficient versions of strong pronouns. That is to say, clitics are a proper subset of weak pronouns, which are a proper subset of strong pronouns. Suppose, then, that such subset relation is correct. This leads to the expectation that if a historical change targets weak pronouns, reducing their number, the clitic forms will also be affected. I believe that this is what happened in BP. BP grammars underwent a loss of weak possessive pronominal forms and, consequently, the possessive Agr, which is a clitic, was lost. 10 But, if this is the correct conclusion about the effects of loss of morphology inside the nominal domain in these languages, we must now return to the 10 It important to remark that clitics in general are disappearing in BP; see Pagotto (1992) and Cyrino (1996), among others. 303 problem I left open in chapter 5. That is, we want to understand the availability of null possessors in sentences like (18). (18) a.[o Jo?o] 1 encontrou [a e 1 irm? ] the Jo?o met-3Sg the sister ?Jo?o met his sister? b. [ o Jo?o] 1 lavou [as e 1 m?os] the Jo?o washed-3Sg the hands ?Jo?o washed his hands? If the possessive Agr is not generated as an independent clitic form in Modern BP, then it might be that the gaps in (17) are NP-traces, similar to the anaphoric null subjects discussed in chapter 4. This amounts to saying that (18a&b) involve movement of the DP o Jo?o from the possessor q-position to the verbal external q-position. 6.2.2 Deriving Null Possessors from Movement A movement analysis of (18) requires a theory compatible with movement into q- positions. In addition, it also requires a discussion about Case checking relations inside the DP in the grammars under investigation. It is standardly assumed that a possessor checks genitive Case inside the DP domain; hence we must explain why in BP a possessor phrase is allowed to check its Case feature outside the DP, as in (18). 304 In BP, it is arguably the case that nouns move to a position outside the NP. Witness the data in (19), where the noun precedes restrictive manner adjectives, which are considered to be outside the NP domain (see 5.2.2) (19)Uma blusa vermelha italiana a blouse red Italian ?An Italian red blouse? Therefore, taking into consideration the DP structure I proposed in 5.2.3, BP nouns move at least to the XP domain. Moreover, if X is a category that syntactically requires a nominal expression inside its projection (something equivalent to the sentential EPP), in BP the syntactic requirements of X are satisfied via N-raising. Thus, accordingly to our understanding of the Case checking procedure (cf. 4.2), we may say that in this language X optionally projects a specifier. Once N adjoins to X, spec of XP will be projected only if there is an element inside the NP that checks its Case feature against XP. Recall that it was suggested that the computational system can delay the movement of a DP to a Case position. More concretely, the suggestion is that sentences in (18) are derived as shown in (20). The DP o Jo?o is merged with the relational noun, checking the possessor q-role, and then it moves to spec of vP, where it checks the q-role assigned by v. After that, it moves to spec of TP, where it has its Case feature finally checked. 1 1 With Chomsky (2000 and thereafter), I am assuming that Case features receive a value only after being checked. 305 (20) TP 4 [ DP o Jo?o Case, f ] 3 T? 3 [ T [v [lavou] 2 +v] 4 +T]] vP t 3 v? t 4 VP 3 t 2 DP as XP m?os 1 NP 3 t 3 t 1 In summary, the gist of this implementation is that it allows movement to a Case position (i e. creation of a spec-head configuration) to be optional as long it is not forced by features of the checker. 6.2.3. DP as a Phase and Intermediate Movement It has been assumed that A-bar extraction out of DPs in Romance involves movement to spec of DP (cf. Cinque 1980, Giorgi and Longobardi 1991, Tellier and Valois 1995, Gavruseva 2000 and Ticio 2003, among others). This is supported by what is often called Cinque?s generalization, which states that an extractable nominal argument must appear as a prenominal possessive (ie. as a weak pronominal possessive form), which surfaces in spec of DP according the authors cited above. In a phase-based approach to linguistic derivations, this condition is captured by the assumption that DPs, 306 like CPs, are strong phases (cf. Chomsky 2001a, Matushansky 2003, among others). Thus, only the determiner and the border (specs) of the DP are visible for computations outside the DP domain, in accordance with the Phase Impenetrability Condition (cf. 4.2.3). Therefore, in a structure with the format in (21), nothing below D 0 can be used to build the vP shell. (21)[ vP [ VP [ DP D 0 [ XP [ NP ]]]]] Consequently, assuming that this is the right way to analyze the island-like behavior of DPs, we need to rethink the movement analysis suggested in the last section. In (20), the described movement of the possessor from the q-position inside the DP to spec of vP crosses over the DP, a strong phrase. This is the same isue I considered in 4.2.3 with respect to A-movement out of finite CPs and the solution I proposed there was that the clausal left periphery contains a functional category FP whose spec hosts an intermediate trace of the extracted DP. That is, A-movement out of finite embeded clauses proceds through spec of FP. Since the FP projection itself does not constitute a phase level, it is not speled-out when the derivation reaches the next strong phase (i.e. vP). Consequently, movement from spec of FP to spec of vP is alowed by the computational system. Now taking seriously the idea that XP is nominal counterpart of TP, one can posit that XP is a weak phase in the sense of Chomsky (2000a). Therefore, if we understand the phase spell-out process as formulated in 4.2.3 and grant the DP structure proposed in 307 5.3.3, repeated here as (22), we can entertain the possibility of analyzing possessor A-mo vement in BP as proceeding through spec of FP, as represented in (23): (22) [ vP [ VP [ DP D 0 [ FP [ XP [ NP ]]]]] (23) TP 4 [ DP o Jo?o Case, f ] 2 T? 3 [ T [v [lavou] 2 +v] 4 +T]] vP 3 t 2 v? 3 t 4 VP 3 t 2 DP [as [ FP t 2 [ F? [ XP [ X? [ X m?os 1 + X] [ NP t 2 t 1 ]]]]] In summary, Possessor raising in BP might proceed in a cyclic way, though spec of FP, a left peripheral Functional Category. This possibility, when coupled with the assumption that FP is not spelled-out as part of the complement of DP/CP within a phase- based approach to locality, suggests that the movement analysis defended here is compatible with the concept of phases. In what follows, for the sake of space, I will put this intermediate movement aside. Omitting some functional projections, I will represent the nominal domain as being a DP dominating an NP. Thus, the possessor movement under consideration will be represented as if it were directly from one q-domain to another. 308 6.3Evidence for a Movement Analysis 6.3.1 Anaphoric Behavior As already discussed in chapter 5, a BP 3 rd P null possessor is allowed only if provided with a sentential antecedent. Thus, (24) is perfectly acceptable as long as the DP in the subject position is taken to be the antecedent of the empty possessor. 12 (24)a.[o Jo?o ] 1 machucou [o bra?o e 1/*2 ] ontem(BP) the Jo?o cut-3Sg the arm yesterday ?Jo?o hurt his arm yesterday? b. [o Jo?o] 1 encontrou [o irm?o e 1/*2 ] no supermercado the Jo?o met-3Sg the brother in.the supermarket ?Joao met his brother in the supermarket? The requirement of a sentential antecedent is reinforced by the data in (25) in which the absence of an appropriate antecedent renders the sentence unacceptable. (25) a. * Parece que [o bra?o e] quebrou(BP) seem-3Sg that the arm broke ?It seems that his/her arm broke? 309 b. * Parece que [o irm?o e ] foi encontrado seem-3Sg that the arm was-3Sg found ?It seems that his/her brother was found? This contrasts with the behavior of null possessors in Romance pro-drop languages in which a null possessor can find its antecedent outside the sentence, as already shown. (26) ? Parece que [el padre e] se muri? ayer seem-3Sg that the father SE died yesterday ?It seems that her father died yesterday? Now, for the completeness of the paradigm, notice that in BP, constructions like a deictic interpretation are reserved for overt pronominal possessors: (27)a. Parece que [o braco dele] quebrou seem-3Sg that the arm of.he broke ?It seems that his arm broke? b. Parece que [o irm?o dele ] foi encontrado seem-3Sg that the arm of.he was-3Sg found ?It seems that his brother was found? In other words, contrasting with null possessors in Romance Pro-drop languages, BP and Finnish null possessors do not pattern like pronouns. Instead, they pattern rather 12 Notice that I am disregarding non-relational readings of kinship and body-part nouns. 310 similarly with traces, in that they are dependent on the presence of a suitable sentential antecedent. 6.3.2Locality Matters 6.3.2.1 The Basic Facts The anaphoric behavior of null possessors in BP and Finnish is more complex than depicted above for one reason: their antecedent has to be the closest c-commanding DP. The sentences in (28), for example, are unacceptable because the closest c- commanding DPs are semantically/pragmatically inappropriate antecedents. In our world, snakes do not have arms, and doctors who are male do not have uteruses. 13 The unacceptability of the following sentences, then, shows that non-local DPs are not interpreted as antecedents, even when this interpretation is pragmatically biased in favor of. This suggests that an antecedence relationship between a null possessor and a DP is restricted by syntactic principles. 13 Although in (28), I am using body-part terms, note that the same observation is valid for kinship terms: (i) a.* [ a cobra [ do Jo?o] 1 ] 2 mordeu [a irm? e 1 ] the snake of.the Jo?o bit-3Sg the sister ?Jo?o?s snake bit his sister? ?? O Jo?o 1 disse que a cobra mordeu [ a irm? e 1 ] the Jo?o said-3Sg that the snake bit-3Sg the sister ?Jo?o said that the snake bit the sister? 311 (28) a.* [a cobra [do Jo?o] 1 ] mordeu [ o bra?o e 1 ] the snake of.the Jo?o bit-3Sg the arm ?Jo?o?s snake bit his arm? b.* [a Maria] 1 falou que [aquele m?dico de Bras?lia] vai the Maria said-3Sg that that doctor-Masc of Brasilia will-3Sg do fazer a cirurgia [no ?tero e 1 ] do-Inf the surgery in.the uterus ?Maria said that that doctor from Brasilia will do the surgery in her uterus? As expected, these sentences become acceptable if an overt genitive pronoun replaces the null possessor. (29) a.[a cobra [do Jo?o] 1 ] mordeu [ o bra?o dele] the snake of.the Jo?o bit-3Sg the arm of.he ?Jo?o?s snake bit his arm? b.[a Maria] falou que [aquele m?dico de Bras?lia] vai fazer a the Maria said-3Sg that that doctor of Brasilia will-3S do-Inf the cirurgia [no ?tero dela] surgery in.the uterus of.she ?Maria said that a doctor from Brasilia will do the surgery in her uterus? 312 These findings are on par with the BP sentential null subject data discussed in chapter four. In both cases, the relationship between the empty category and its antecedent obeys the Minimal Link Condition (MLC). Thus, to the extent that MLC is a condition on the operation Move (see 4.3.2), the data above corroborates a movement analysis. The locality restriction is also illustrated by cases like (30). (30) is similar to (28b), the difference being that (28b) involves clausal embedding, whereas (30) displays nominal embedding. In (30), the matrix direct object is a possessive DP containing a noun that takes as its complement a prepositional phrase containing another possessive DP. That is, in these sentences, the matrix direct object is a complex DP containing two possessors. In (30a), both possessors are null, but in (30b) only the most embedded one is null. (30)a.[a Maria] 1 nunca assumiu [ DP o relacionamento e 1/*2 [ PP com the Maria never assumed-3Sg the relationship with [ DP o namorado e 1/*2 ]]] the boyfriend ?Maria never assumed her relationship with her boyfriend? b. [a Maria] 1 acha engra?ado [ DP o relacionamento [da Ana] 2 the Maria think-3Sg funny the relationship of.the Ana 313 [ PP com [o namorado e *1/2 ]]] with the boyfriend ?Maria thinks that Ana?s relationship with her boyfriend is funny? As the indexes show, in (30a), the DP a Maria ?Maria? is obligatorily interpreted as the antecedent of both possessors. Contrastively, in (30b), a Maria is prevented from being interpreted as the antecedent of the null possessor that co-refers with a Ana, the possessor of the highest DP. According to the analysis I am suggesting, (30a) is generated by successive movements of the DP a Maria into q-positions. This DP is first merged inside the lowest possessive DP, where it checks the q-role of the relational noun namorado ?boyfriend?, from which it moves to the spec of the NP headed by the noun relacionamento ?relationship?, and in this position, it checks the possessor q-role of this noun. Later on, when the matrix vP is built, the same DP moves to spec of vP, thereby checking the agent q-role of the verb assumiu ?assumed?. After that, it moves to spec of TP where it has its Case feature checked. In (30b), the successive movement just described is blocked by the possessor DP a Ana. This possessor DP is closer to spec of vP than the lowest possessor. Therefore, if the DP a Maria starts as the lower possessor, its movement towards spec of vP crosses over the DP a Ana and infringes the MLC. 14 Notice that the blocking effect observed in (30b) brings back the discussion of the role of the preposition de ?of? in defining the c-command domain of a genitive possessor. 14 Similarly to what happens in external dative possessors; see example (5b). 314 The DP a Ana in (30b) is inside a prepositional phrase, hence it should in principle be unable to c-command the lowest possessor position and should not be, contrary to fact, an intervener for the movement of the DP a Maria from the embedded possessive DP to spec of vP. I will not repeat the content of the appendix of chapter 4 here, but as presented there, there is evidence that a prepositional phrase headed by de ?of? does not define the c-command domain of a genitive possessor. Thus, the data in (30) shows once again that in BP the relationship between a 3 rd P null possessor and its antecedent is subject to locality constraints. 6.3.2.2 Potential Interveners: Reinforcers As already discussed in chapter 3, In BP whenever a reinforcer is present inside a possessive DP, a null possessor is not licensed. (31)* A Maria 1 encontrou com [um irm?o l? e 1 ] the Maria met-3Sg with a brother there ?Maria met one of her brothers? Assuming that (a) BP null possessors are formed by the operation Move; and (b) reinforcers, similar to preverbal locative adverbs (see 4.3.2.2), are overt expletive-like categories that surfaces in spec of FP, we can provide an explanation for the interaction between reinforcers and null possessors. The presence of a reinforcer blocks movement 315 of the possessor to the left periphery of the nominal domain. (32) * De quem 1 que a Maria encontrou [um irm?o l? t 1 ] of whom that a Maria met-3Sg a brother there ?Of whom did Maria meet a brother? (30) becomes acceptable when the null possessor is substituted by an overt pronoun, as in (33). This follows from the analysis proposed here. (31) but not (33) involves movement. Thus, in this respect (31) is syntactically similar to (32), not to (33). (33) A Maria 1 encontrou com [um irm?o l? dela 1 ] the Maria met-3Sg with a cousin there of.she ?Maria met one of her brothers ? In 6.3.4, I will revisit reinforcers vis-?-vis the matter of specificity. 6.3.3 Obligatory Control Properties Following Hornstein (1999, 2001), I?ve been taking obligatory control as the result of movement. Thus, we expect BP configurations containing 3 rd P null possessors to display obligatory control properties. As shown below, this expectation is fulfilled. First, as discussed in the previous sections, these elements are anaphoric, taking the closest c-commanding DPs as their antecedents. Furthermore, the unacceptability of 316 (34) reveals that split antecedents are not allowed if the possessor is null. Compare (34) with (35) which contains an overt possessive pronoun: (34)a.* A Maria 1 disse que o Paulo 2 encontrou [o amigo e 1+2 ] the Maria said-3Sg that the Paulo found-3Sg the friend ?A Maria said that Paulo found their friend? b.* A Maria 1 prometeu para a Ana 2 que ia cortar the Maria promised-3Sg to the Ana that would-3Sg cut-inf [ o cabelo e 1+2 ] the hair ?Maria promised to Ana that would cut their hair? (35)a.A Maria 1 disse que o Paulo 2 encontrou [o amigo deles 1+2 ] the Maria said-3Sg that the Paulo found-3Sg the friend 0f.they ?A Maria said that Paulo found their friend? b.A Maria 1 prometeu para a Ana 2 que ia cortar the Maria promised-3Sg to the Ana that would-3Sg cut-inf [o cabelo delas 1+2 ] the hair of.they ?Maria promised to Ana that would cut their hair? Under VP ellipsis, only a sloppy reading of obligatory controlled gap is possible. 317 Thus, (36) confirms that BP null possessors are obligatory controlled gaps. (36)a. O vov? 1 n?o corta [as unhas e 1 ] e a vov? tamb?m n?o ?Grandpa does not cut his nails and grandma doesn?t either? b. A Maria bateu no irm?o e a Ana tamb?m the Maria hit-3Sg in.the brother and the Ana too ?Maria hit her brother and Ana did too? Unsurprisingly, strict and deictic readings are allowed if the possessor is realized by an overt pronoun. (37)a. O vov? n?o corta [as unhas dele] e a vov? tamb?m n?o ?Grandpa does not cut his nails and grandma doesn?t either? b. A Maria bateu no irm?o dela e a Ana tamb?m the Maria hit-3Sg in.the brother of.she and the Ana too ? Maria hit her brother and Ana did too? As discussed in the literature, a gap resulting from obligatory control forces a de se reading. In chapter 4, I have shown that this is a property of 3 rd P referential null sentential subjects in BP. This seems to be true of 3 rd P null possessors as well. For 318 example, given the context in (38), the statement in (39) is false, as Ronald Reagan does not remember his wife, but Nancy Reagan, Reagan?s wife. (38) Because of his Alzheimer disease, Ronald Reagan does not have memories about who he was or about who his kin is. He remembers Nancy Reagan, which he takes to be the wife of President Reagan. However, he fails to remember that he himself is former President Reagan and Nancy Reagan his wife. (39)O Ronald Reagan lembra d[a esposa] the Ronald Reagan remember-3Sg of.the wife ?Ronald Reagan remember his own wife? Though subtle, there is a contrast between (39) and (40) where an overt pronoun is inserted. (39) accepts only a de se reading. That is, the gap in the possessor position must refer back to Ronald Reagan, forcing us to attribute de se memories to Ronald Reagan. Conversely, (40) is ambiguous between a de se and a non-de se reading. The overt pronoun can either refer back to Ronald Reagan, which results in a de se reading, or refer to Reagan, the former president. If the pronoun co-refers with Ronald Reagan, (40) is false, but if it refers to Reagan, a de se reading does not emerge, and the statement is judged as being true. 319 (40)O Ronald Reagan lembra d[a esposa dele] the Ronald Reagan remember-3Sg of.the wife of.he ?Ronald Reagan remembers Reagan?s wife very well? Similar to BP null subjects (cf. 4.3.3), BP null possessors pattern like the emphatic reflexive form dele mesmo ?of he himself?, which also triggers a de se reading. For instance, (41) is false, if (38) is taken as the contextual background. (41) O Ronald Reagan lembra d[a esposa dele mesmo] the Ronald Reagan remember-3Sg of.the wife of.he himself ?Ronald Reagan remember his own wife? To make this paradigm self-evident, let me apply Chierchia?s (1990) diagnosis for de se readings (cf. 4.4.3). When (39) is conjoined with a negative version of (41), as in (42), the expressed thought is incoherent. This incoherence is avoided if an overt genitive possessive pronoun is inserted and understood as referring to Reagan, rather than to Ronald Reagan, as in (43): (42)# O Ronald Reagan lembra d[a esposa e], mas Ronald Reagan n?o the Ronald Reagan remember-3Sg of.the wife, but Ronald Reagan not 320 lembra a esposa dele mesmo remember-3Sg the wife of.he himself ?Ronald Reagan remember his own wife, but Ronald Reagan does not remember his own wife? (43) O Ronald Reagan lembra d[a esposa dele Reagan ], mas o Ronald Reagan the Ronald Reagan remember-3Sg of.the wife of.he but the Ronald Reagan n?o lembra a esposa dele mesmo not remember-3Sg the wife of.he himself ?Ronald Reagan remembers Reagan?s wife, but Ronald Reagan does not remember his own wife? The intimate relation between a null possessor and the de se reading falls under our movement analysis. If the empty category in (39) is a silent copy of the DP Ronald Reagan, this sentence asserts that Ronald Reagan remembers Ronald Reagan?s wife, i.e., Ronald Reagan remembers one of his own kin. Conversely, movement is not involved in the derivation of (40) and, if the possessive genitive pronoun is interpreted as referring to Reagan, then, the assertion is about Ronald Reagan remembering Reagan?s wife. Consider now sentences with only-NP phrases. As presented in the previous chapters (Chapters 2 and 4), in obligatory control configurations if the antecedent of PRO has the format only-NP, then only a covariant reading is possible. Thus, if BP 3 rd P null possessors are also obligatory controlled elements, they are expected to trigger a 321 covariant interpretation when the antecedent is an only-NP constituent. The data below illustrate that this indeed happens. (44a) contains a null possessor and is assigned the covariant interpretation (44a). In (44b), the null possessor is replaced by an overt pronoun and, as a result, the sentence receives the invariant interpretation in (45b). (44) a. S? a Maria respeita [o marido e] only the Maria respect-3Sg the husband b. S? a Maria respeita [o marido dela] only the Maria respect-3Sg the husband of.she ?Only Maria respects her husband? (45)a.Only Maria is an x such that x respects x?s husband (Covariant interpretation ? (44a)) b.Only Maria is an x such that x respects her, Maria?s, husband (Invariant interpretation ? (44b)) This can be paraphrased in the following way: (44a) asserts that only Maria is a good wife because only she has the husband-respecting property. (44b), on the other hand, asserts that Maria?s husband is an unpopular guy, since nobody but Maria has respect for him. Consider now sentences containing body-part terms, as in (46). According to (46a), Maria is the only person that combs their hair (covariant interpretation). (46b), on 322 the hand, is compatible with a context in which Maria is the sort of person that doesn?t like to be touched, thus, only her, and nobody else, is allowed to comb her own hair (invariant interpretation). (46)a.S? a Maria penteia [o cabelo e ] only the Maria comb-3Sg the hair (Only Maria is an x such that x combs x?s hair) b.S? a Maria penteia [o cabelo dela ] only the Maria comb-3Sg the hair of.she (Only Maria is an x such that x combs Maria?s hair) ?Only Maria combs her hair? As presented in 4.3.3, a covariant interpretation is forced upon structures containing obligatory control gaps because these gaps, being the result of movement, require a c-commanding antecedent. Hence, when the closest DP has the format only-NP, the possessor gap must be co-indexed with the whole only-NP constituent; it cannot be co-indexed with the NP contained inside the only-NP as represented in (47a). On the other hand, an overt possessive pronoun, which is not formed by movement, allows co- indexation with the NP, as illustrated in (47b) (47) a.[ ? [ DP only [NP ] 1 ] 2 ? [DP ? e poss*1/2 ?]] b.[ ? [ DP only [NP ] 1 ] 2 ? [DP ? Pronoun poss1/*2 ?]] 323 Suppose that only-NPs are quantified phrases, as already suggested in this thesis. Then, by virtue of being subject to Montalbetti?s Constraint, a BP overt possessive pronoun cannot be co-indexed to the entire only-NP phrase. Hence, a structure with the abstract schema in (47b) cannot receive a covariant interpretation. Notice again that in the dialects of BP where quantified phrases with the format todo-NP ?every-NP? can violate Montalbetti?s constraint, an overt possessive pronoun can be bound by the todo-NP phrase, triggering, thus, an covariant t interpretation. For example, the interpretation of (48a) is unambiguously covariant (49a); but (48b) accepts both a covariant and an invariant (49b) interpretation. This corroborates our hypothesis that there is an interaction between Montalbetti?s constraint and the restricted interpretation of (46b). (48) a.[toda f? [da Carla Peres] 1 ] 2 gosta de falar [do marido e *1/2 ] every fan of.the Carlas Peres like-3Sg of speak of.the husband b. [toda f? [da Carla Peres ] 1 ] 2 gosta de falar [do marido every fan of.the Carlas Peres like-3Sg of speak-Inf of.the husband dela 1/2 ] of.she (49)a. "x (x = a fan of Carla Peres) x likes to talk about x?s husband b. "x (x = a fan of Carla Peres) x likes to about her, Carla Pere?s, husband 324 Summarizing, in this section I have shown that BP 3 rd P null possessors display all the properties that have been used as diagnosis of obligatory control. Hence, to the extent that obligatory controlled gaps fit a movement analysis, the discussion above supports the suggestion I am presenting in this chapter: BP 3 rd P null possessors are NP-traces. 6.3.4 Null Possessors and Specificity As well known, specificity induces a barrier for movement out of DPs. 15 Thus, in BP a wh-phrase might escape a definite (50), but not a specific DP (51): (50) a. [de qual dos pacientes] 1 voc? obturou [o dente t 1 ] of which of.the patients you fill-3S the tooth ?Which of the patient did you fill the tooth of? b.? [de quem] 1 que voc? viu [a irm? t 1 ] of who that you saw-3Sg the sister ?*Who did see the sister of ? (51)a.?? [de qual dos paciente] 1 voc? obturou [aquele dente t 1 ] of which of.the patients you fill-3Sg that tooth ?* Which of the patients did you fill that tooth of ? 15 For a discussion on specificity, see, among others, Chomsky (1973, 1977), Fiengo and Higginbotham (1981), Fodor and Sag (1982), Hintikka (1986), En? (1991) and Diesing (1992). I will not offer or adopt any definition for this concept; rather, I am interested in exploring the syntactic effects of specificity on extraction. 325 b.* [de quem] 1 que voc? viu [ aquele irm? t 1 ] of who that you saw-3Sg that sister ?* Who did you see that sister of ? The islandhood property of specific DPs is also observed in the semantic behavior of quantifiers. Quantifier phrases embedded within specific DPs cannot have a wide- scope reading. For example, Ticio (2003) notes that while (52a) is ambiguous, meaning either that Ivan saw a picture of a group (narrow scope of the quantifier) or that Ivan saw several pictures (wide scope), (52b) unambiguously means that Ivan saw a picture of a group. (52)a.Ivan vio la foto de todo el mundo (wide/narrow) Ivan saw-3Sg the picture of everybody ?Ivan saw the picture of everybody? b. Ivan vio esta foto de todo el mundo (*wide/narrow) Ivan saw-3Sg this picture of everybody Ivan saw the picture of everybody? If wide scope results from syntactic movement at LF (Quantifier Raising; May 1985), then the conclusion is that specific DPs, contrary to definite DPs, behave like islands, blocking overt and covert movement. 16 16 Unsurprisingly, LF wh-movement is also blocked: (i)* Who wants to see these pictures of what? (Uriagereka, 1993) 326 If this is right, our analysis for BP 3 rd P null possessors predicts that these possessors are not licensed inside specific DPs. The data below confirm this prediction. A 3 rd P null possessor is allowed inside a definite DP, but prohibited inside a specific one. In order words, BP 3 rd P null possessors behave similarly to wh-traces and quantifier phases with wide scope, being disallowed inside specific DPs. 17 17 I am omitting from this discussion examples like (i), which is interpreted as involving a possessor relation between the entity Z? do Caix?o and the relational noun unhas ?nails? (i)O Z? do Caix?o podia nos fazer o favor de cortar aquelas unhas the Z? do Caix?o could us-cl do the favor of cut-Inf those nails ?Z? do Caix?o could do us the favor of cutting those nails? Building on Authier (1988) and Vergnaud and Zubizarreta (1992), I will assume that these cases involve a deictic use of the relational nouns and, consequently, differ from the cases discussed in this chapter. Evidence for positing this differentiation comes from appositive adjectives. Appositive adjectives cannot co-occur with external possessor (cf. Kayne 1975, Authier 1988 and Vergnaud & Zubizarreta 1992); that?s why (ia), which contains a dative external possessor, is illicit. (i)* Pierre lui a lav? les mains sales Pierre to.him washed-3Sg the hands dirty ?Pierre washed his dirty hands? These adjectives are also blocked in the BP structures we are studying in this chapter. Consider, for example, the paradigm in (ii), where BP and Spanish are contrasted in order to emphasize the different syntactic behavior of BP null possessors. (ii)a. * A Maria cortou o cabelo tingido the Maria cut-3Sg the hair painted b. Mar?a se corto el pelo te?ido the Maria Cl cut-3Sg the hair painted ?Maria cut her painted hair/ had her painted hair cut? Now, notice that an appositive adjective can occur in (i). This suggests that (iii) should not be grouped with the cases considered here. (iii)O Z? do Caix?o podia nos fazer o favor de cortar aquelas unhas bizarras the Z? do Caix?o could us-cl do the favor of cut-Inf those nails bizarre ?Z? do Caix?o could do us the favor of cutting those bizarre nails? 327 (53) a.A Maria 1 cortou as [unhas e 1 ] the Maria cut-3Sg the nails ?Maria cut her nails? b. A Maria 1 encontrou [ o irm?o e 1 ]. the Maria met-3Sg the brother ?Maria met her brother? (54) a.* A Maria 1 n?o penteia [aquele cabelo e 1 ] the Maria not comb-3Sg that hair ?Maria does not comb that hair of hers? b. * A Maria 1 encontrou [aquele irm?o e 1 ] the Maria met-3Sg that brother ?Maria found that brother of hers? Brug? (1996) proposes that demonstratives are base-generated in a functional category between the NP and the DP and that they might move cyclically to spec of DP, where they are spelled-out. 18 This proposal provides a rationale for why BP 3 rd P null possessors, as well as wh-traces, are prevented from occurring inside a specific DP 18 For Brugg? the functional projection under consideration is NumberP, and she suggests that the demonstrative movement to spec of DP is optional in some grammars. In Spanish, for instance, the demonstrative can precede or follow the noun. Thus, if in Romance nouns move to higher position inside the DP, then (ib) is evidence that demonstratives in Spanish might not move. (i)a. este/ese/aquel libro this/that/that book b. el livro este/ese/aquel the book this/that/that ?This/that book? 328 containing a demonstrative. Being the specifier of a functional projection, a demonstrative interacts with movement within the DP domain, arguably creating a blocking effect for extraction of an NP argument. Uriagereka (1995), investigating clitic placement in western Romance languages suggested that third person clitics are specific determiners and as such that they adjoin to the head of the sentential FP, the locus of specificity. If this is right, it is arguably the case the nominal FP is also a site for specific items. Hence, we can hypothesize that demonstratives are either base-generated in FP, or moved there during the course of the derivation. This suggests that demonstratives interfere with extraction out of the DP domain in the same way that reinforcers do: by blocking spec of FP. If DPs are phases as discussed in 4.2.3, it means that demonstratives and reinforcers block extraction because they occupy the position that nominal arguments would use as escape hatch from the phase otherwise. Perhaps the hypothesis that FP is related to specificity can give substance to the assumption that reinforcers surface in spec of the nominal FP. The fact is that reinforcers mark an indefinite DP as specific. According to En? (1991), specific DPs are restricted discursively in the sense that they are linked (perhaps in a inclusive way) to previously established discourse referents. Particularly important for our discussion, En?, building on Hintikka (1986), also observes that an indefinite DP referring to an unfamiliar entity (i.e, an entity not previously mentioned in the domain of the discourse) is specific if it contains adjectives like certain and specific. (55), for instance, is acceptable even if the set of relevant tasks has not been previously introduced into the domain of discourse. 329 (55) The teacher gave each child a certain task to work on during the afternoon. Tavares (2001) shows that reinforcers are like these type of adjectives in that they also induce an indefinite unfamiliar nominal to be specific. In (56) the DP uma atleta a? ?an athlete there? is understood as referring to a specific athlete, even though it does refer to an entity that had already been introduced discursively. (56) A C?tia deve vencer uma atheta a? se quiser ser a primeira the Catia need-3Sg win-inf a athlete there if wants be-inf the first do ranking(BP - Tavares, 2001) of.the ranking ?C?tia needs to win over a certain athlete if she wants to be the first in the ranking? Given this feature of reinforcers, it is rather natural that they surface in spec of FP, assuming, of course, that Uriagereka is correct in suggesting that FP encodes specificity. As the hypothesized correlation between specificity and spec of FP predicts, specific adjectives also block extraction from inside nominals. (57), for example, illustrated that under the presence of this adjective in BP, wh-extraction is ruled out as well as the type of possessor raising we are investigating here. Thus, perhaps these adjectives are in competition for spec of FP. 330 (57)a.* De quem 1 que voc? vai encontrar com [uma certa tia t 1 ] of whom that you will-3Sg meet-inf with a certain aunt ?* Whom will you meet a certain aunt of? b.* O Jo?o 1 encontrou [com uma certa tia e 1 ] the Jo?o met-3Sg with a certain aunt ?Jo?o met with a certain aunt of his? At any rate, the facts above corroborate the movement analysis I am arguing for in this chapter. 6.3.5 Null Possessors inside Relative Clauses The fact that in (58) the null possessor can refer back to Maria illustrates that in European Portuguese, Spanish, and Galician null possessors are pronominal categories, as opposed to traces. As such, they are able to occur inside a relative clause, taking a non- local DP as the antecedent. (58)a.Maria 1 bateu na [menina 2 que beijou [o e 1/2 namorado]] (EP) Maria hit-3Sg in.the girl that kissed-3Sg the boyfriend b. Mar?a 1 peg? a la chica que bes? al [e 1/2 novio] (Sp.) Maria hit-3Sg to the girl that kissed-3Sg to.the boyfriend 331 c. Maria bateu a rapaza que beixou [ ? e 1/2 mozo] (Gal.) Maria hit-3Sg the girl that kissed-3Sg the boyfriend ?Maria hit the girl that kissed her boyfriend.? Compare (58) with its Brazilian counterpart: (59)A Maria 1 bateu na [menina 2 que beijou [o namorado e *1/2 ]] (BP) a Maria hit-3Sg in.the girl that kissed-3Sg the boyfriend ?Maria hit the girl that kissed her boyfriend? In (59), the empty possessor cannot refer back to the DP a Maria. This suggests that it is not a null pronoun. It patterns rather as the wh-trace in (60): (60)* De quem que voc? viu a menina que bejou o namorado of whom that you saw-3Sg the girl that kissed-3Sg the boyfriend ?* Whom did you saw the girl that kissed the boyfriend of? This is fully parallel to the behavior of null subjects discussed in chapter 4. In both cases, the 3 rd P empty category is prohibited inside relative clauses. If a raising analysis of relative clauses (cf. 4.3.5) is adopted, this prohibition follows from the fact that movement out of the relative clause results in violation of the MLC. The derivation of (59), repeated below as (61) unfolds as sketched in (62). The 332 possessor a Maria is first merged with the relational noun namorado ?boyfriend? where it gets the possessor theta-role, after which the CP in (62a) is built. This CP is taken as a complement of the determiner a ?the?, and the determiner projects forming the complex DP represented in (62b). Next, the verb bateu ?hit? merges with (62b), as in (62c). This verb has an external q-role to be checked. Since the possessor phrase a Maria hasn?t checked its Case feature yet (cf. 6.2), at this step of the derivation it is still available for further computations. Therefore, it could in principle be copied and merged as the specifier of the matrix verb. Yet, this move is not possible, for it violates the MLC. The head of the relative clause (i.e. the NP menina ?girl?) is closer to the matrix verb than the possessor phrase is. Hence, it prevents the movement of the possessor towards spec of VP, as shown in (38d). (61) * A Maria 1 bateu na menina que beijou [o namorado e 1 ] . the Maria hit-3Sg in.the girl that kissed-3Sg the boyfriend ?Maria hit the girl that kissed the boyfriend? (62)a.[ CP [menina] 1 [ TP t 1 [ T? [ VP t 1 [ V? beijou [ DP [ NP [a Maria] poss namorado]]]]]]] b. [ DP a [ CP [menina] 1 [ TP t 1 [ T? [ VP t 1 [ V? beijou [ DP o [ NP [a Maria] poss ] o namorado]]]]]]]] c.[ VP bateu [ DP a [ CP [menina] 1 [ TP t 1 [ T? [ VP t 1 [ V? beijou [ DP o [ NP [a Maria] poss namorado ]]]]]]]]] 333 * d. [ VP [ V? bateu [ DP a [ CP [menina] 1 ?[ DP o[ NP [a Maria] poss namorado ? ]]]]]] In conclusion, as the analysis I am arguing for predicts, the possessors under consideration cannot occur inside relative clauses. 19 6.4Null Possessors inside Adjuncts Adjuncts are also opaque domains for extraction. Hence, in accordance with the analysis I am arguing for, in BP and Finnish, the occurrence of null possessors inside adjuncts is unexpected. Yet, as (63) shows, it may occur: (63) a.A Maria 1 viajou com [a irm? e 1 ] the Maria traveled-3Sg with the sister ?Maria traveled with her sister? 19 Although I judge sentences like (i) to be marginal, for some of my informants they are acceptable. (i) [a Maria] 1 viu [o cara que [a filha e 1 ] beijou na festa] the Maria saw-3Sg the guy that the daughter kissed at.the party ?Maria saw the guy that her daughter kissed at the party? Obviously this is not amenable to a movement analysis, but I will not comment on the matter here, especially because it is not clear what this shows us. The speakers that accept (i) do not accept (37). Hence, it seems that, inside a relative clause, they allow a 3 rd P null possessor inside a subject DP, but not inside an object DP. Moreover, they also reject sentences in which the antecedent of a null possessor is not a local or c-commanding DP. Therefore, the reasons for why their grammars allow (i) are not obvious. If these grammars could resort to a null possessive pronoun, we would expect them to generate (61) as well as sentences in which the antecedence relationship between a null possessor and a DP does not obey the MLC. 334 b.A Maria 1 dormiu com [o cabelo e 1 ] molhado the Maria slept -3Sg with the hair wet ?Maria slept with her hair wet? The impossibility of wh-extraction from the prepositional phrases above indicates that these phrases are adjuncts. 20 (64) a.* De quem 1 que a Maria viajou com [a irm? t 1 ] ? of whom that the Maria traveled-3Sg with a sister ?* Whom did Maria traveled with the sister of? b.* De quem 1 a Maria odeia sair com [o cabelo t 1 ] molhado of whom the Maria hates-Inf go.out with the hair wet ? * Of whom does Maria hate to go out with the hair wet? In chapter four, I analyzed occurrences of sentential null subjects inside adjuncts by suggesting that they are to be derived by sideward movement. The main idea behind this mechanism is that a category is able to undergo a lateral movement from inside a phrase that is in a posterior stage of the derivation, and will be adjoined to the main spine of the tree, becoming thus an adjunct (i.e. an island for extraction). That amounts to 20 Notice that if movement does not occur, (63b) is perfectly fine. See, for instance, (i) which means that Maria hates to go out with Tatiana when Tatiana?s hair is wet. (i) A Maria odeia sair com [o cabelo da Tatiana] molhado the Maria hate-3Sg go.out with the hair of Tatiana wet 335 saying that movement out of an adjunct depends on which step of the derivation it occurs. It succeeds only if it happens prior to the adjunction process. Bearing this in mind, consider again the contrast in (63) and (64). (63) involves A-movement into a theta-position, whereas (64) are attempts at A?-movement to the spec of the matrix CP. Thus, suppose that in these sentences the prepositional phrase is adjoined to vP. In derivations of the sentences in (63), the possessor is raised to spec of vP before the merge of the prepositional phrase with vP. This is possible if movement between parallel phrase markers (sideward movement) is allowed. In (64), contrarily to what happens in (63), the possessor phrase tries to move to spec of CP after the prepositional phrase has been adjoined to vP. That is, the movement is from an island and arguably ruled out by the computational system. To see the difference between (63) and (64), compare the derivations in (65) (= (63a) and (66) (= 64a). In first relevant stage of (65), the prepositional phrase and the verb phrase are built as separate sub-structures and the possessor moves sideways from its original position inside the relational DP to spec of vP, where it checks the sleeper q-role. After that, in stage 2, in accordance with the Extension Condition, the prepositional phrase adjoins to vP and, consequently, becomes an island for extraction. When T is merged with the vP (stage 3), the possessor moves to spec of TP, where it checks its Nominative Case feature. 21 21 For space reasons, the representations of the DP structure are simplified. 336 (65) Stage 1 Stage 2 Stage 3 PP 2 com DP 2 D? 2 a NP 2 [ DP a Maria] irm? vP 2 v? 2 viajou 1 t 1 vP 4 vP [ PP com [ DP a t 2 irm?]] 5 [ DP a Maria] 2 v? 3 viajou 1 t 1 ... TP 3 T? 3 vP 4 vP [ PP com [ DP a 3 t 2 irm?]] [ DP a Maria] 2 v? 3 viajou 1 t 1 In the derivation of (64) (cf (66)) there is no parallel movement. The prepositional phrase is built and merged with matrix vP with the wh-possessor in situ (cf. Stage 1). After that, the matrix TP is assembled with the matrix subject, which is moved to spec of TP, as shown in stage 2. At the end of the derivation (stage 3), when the matrix C is inserted, the wh-possessor, which is still inside the prepositional phrase, tries to move to spec of CP. But, under the assumption that a phrase becomes an island after having been adjoined with a maximal projection, the computational system cannot perform the wh-movement illustrated in stage 3 of (65). 337 (66) Stage 1 Stage 2 vP 4 vP [ PP com [ DP a irm? [de quem]]] 3 [a Maria ] v? 2 viajou 1 t 1 TP 3 [a Maria] 2 T? 3 T vP 4 vP [ PP com [ DP a irm? [de quem]]] 3 t 2 v? 3 viajou 1 t 1 Stage 3 CP 3 * C? 2 TP [a Maria] 2 T? 3 T vP 4 vP [ PP com [ DP a irm? [de quem]]] 3 t 2 v? 3 viajou 1 t 1 In sum, one important outcome of the theoretical framework presented in chapter 2 is that it permits A-movement out of adjuncts by taking sideward movement to be performed by the computational system and by assuming that islands emerge derivationally, as a result of the adjunction process. In chapter 4, I showed this is of 338 particular relevance in the analysis of 3 rd P null subjects in BP and Finnish and here I have suggested that it is also relevant in explaining BP 3 rd P null possessors. 6.5Null Possessors inside Coordinate Structures Ross (1967) observed that in structures involving coordination, the conjoined nodes and the material contained by them are restricted from being moved: (67)a.* What sofa will he put the chair between some table and b. * What table will he put the chair between and some sofa (68) a.* The lute which Henry plays and sings madrigal is warped b. * The madrigals which Henry plays the lute and sings sound lousy He captured this restriction by adding the following constraint to the Grammar: (69) The Coordinate Structure Constraint (Ross, 1967:161) In a coordinate structure, no conjunct maybe be moved, nor may any element contained in a conjunct be moved out of that conjunct However, as he noticed, (69) can voided if movement applies in an across the board (ATB) fashion, moving a single element from all the conjoined phrases at the same time. 339 (70)a.Which lute did John play and Paul win b.Which film did the critics hate and the audience love Interestingly, this restriction on movement also governs the licensing of BP 3 rd P null possessors. As illustrated in (71) and (72), these null categories are not allowed inside coordinated DPs, unless they occur across the board. 2 (71)a.* O Pedro 1 encontrou [[ DP a irm? do Jo?o] e [ DP a m?e e 1 ]] the Pedro met-3Sg the sister of.the Jo?o and the mother ?Pedro met Jo?o?s sister and his mother? b.* A Maria 1 fez [[ DP o meu cabelo] e [ DP as unhas e 1 ]] the Maria did-3Sg the my hair and the nails ?Maria did my hair and her nails? (72)a.O Pedro 1 encontrou [[ DP a irm? e 1 ] e [ DP a m?e e 1 ]] the Pedro met-3Sg the sister and the mother ?Pedro met his sister and mother? 2 Some speakers accept (ia), even though they reject (71) and (ib). I have no explanation for this. (i) a. * O Pedro 1 encontrou [[ DP a irm? e 1 ] e [ DP a m?e do Jo?o]] the Pedro met-3Sg the sister and the mother of.the Jo?o ?Pedro met his sister and Jo?o?s mother? b. * A Maria 1 fez [[ DP o cabelo e 1 ] e [ DP as minhas unhas ]] the Maria did-3Sg the my hair and the nails ?Maria did her hair and my nails? 340 b.A Maria 1 fez [[ DP o cabelo e 1 ] e [ DP as unhas e 1 ]]. the Maria did-3Sg the hair and the nails ?Maria did her hair and nails? Thus, if ATB is a requirement on movement out as originally proposed by Ross, then the contrast between (71) and (72) indicates that BP 3 rd null possessors are formed by movement. However, this conclusion does not immediately grant the analysis we are suggesting here. After Ross the ATB phenomenon received many accounts (cf. Ha?k 1985, Munn 1993, Goodall 1994, Nunes 1995, Hornstein 2001, Hornstein and Nunes 2002, Boskovic and Franks 2001, Citko 2002) and among the frameworks that assume movement, there is disagreement about which type of movement is in fact involved. For example, Munn argues that coordinate structures are Boolean phrases head by a conjunction (as in (73) where XP and YP are conjoined phrases, hierarchically organized as the specifier and the complement of and respectively) and that ATB involves movement of a null operator along the lines proposed by Chomsky (1986b) for parasitic gaps. (73) [ BP XP [ B? and [YP]]] It is easy to see that a null operator account is not amenable to the treatment I am offering for 3 rd P null possessors in BP. In exchange for admitting that ATB involves null operators, I would be forced either to change, or weaken my analysis, and assume that BP 341 3 rd P null possessors can either traces of NPs or null operators depending on the configuration. It is unclear, however, that we need such a radical move. A null operator account for ATB is not free of problems to begin with. As observed in Hornstein and Nunes (2002), it accommodates extraction of DPs, but requires non-standard assumptions in accounting for extraction of other constituents, including the auxiliary in (70). Moreover, there are alternative analyses. For example, Nunes (1995), and later Hornstein and Nunes accept the Boolean structure in (73); but suggest subsuming ATB extractions under Nunes? analysis of parasitic gaps, proposing that this type of extraction also instantiates sideward movement application. To see the relevant details of their proposal, consider (74), which is the structure of (70a) in accordance with their proposal. (74) [ CP which lute [ C? did [ TP John did play which lute] [ B? and [ TP Paul did win which lute]]]] (74) is derived by moving the auxiliary did and the wh-phrase which lute from one conjoined phrase to another. This derivation starts by assembling the second conjoined phrase, viz. the clause Paul did win which lute, which is then merged with and. Next, the first conjoined phrase is built as separate phrase maker, and did and which lute are copied and sideward moved to inside this phrase marker, forming the clause John did play which lute. To satisfy the feature of the complementizer, the auxiliary and the wh-phrase are finally moved into the CP domain. 342 This analysis of ATB, besides fitting the theoretical approach adopted in this thesis, allow us to explain the null possessor in (72) in accordance with the movement analysis I am arguing for. The relevant derivation of (72a), repeated here as (75), is the one shown in (76). The DP o Pedro is first merged with the noun m?e ?mother?, where it checks the possessor q-role assigned by this noun. At the next relevant step of the derivation (76c), the possessor phrase moves sideways, being merged with the noun irm? ?sister? and checking the possessor q-role assigned by this noun too. When the verbal phrase is built the DP o Pedro moves to a third theta-position (spec of VP) (cf. (76f). Finally, when T is inserted into the derivation, o Pedro is raised to spec of TP, where it has its Case feature checked. (75) O Pedro 1 encontrou [[ DP a irm? e 1 ] e [ DP a m?e e 1 ]] the Pedro met-3Sg the sister and the mother ?Pedro met his sister and mother? (76)a.[ DP a [ NP [o Pedro] poss m?e ]] ? (building the second conjunction) b. [ BP e [ DP a [ NP [o Pedro] poss m?e ]]] ? (Merging the second conjunct with the conjunction) c. [ DP a [ NP [o Pedro] poss3 m?e]] ? (sideward movement of the possessor) d. [ BP [ DP a [ NP [o Pedro] poss irm?]] [ B? e [ DP a [ NP t poss m?e]]]] ? 343 (merging the second conjunct with the BP) e.[ VP encontrou [ BP [ DP a [ NP [o Pedro] poss irm?]] [ B? e [ DP a [ NP [t poss m?e]]]]] ? (merging BP with the verb) f. [ VP [o Pedro] poss [ V? encontrou [ BP [ DP a [ NP t poss irm?]] [ B? e [ DP a [ NP [ t poss m?e]]]]]]] ? (moving the possessor to spec of VP) g. [ TP [o Pedro]poss [ T? [ VP t poss [ V? encontrou [ BP [ DP a [ NP t poss irm?]] [ B? e [ DP a [ NP [ t poss m?e ]]]]]]]]]] ? (moving the possessor to spec of TP) To sum up, since sideward movement is already part of the framework adopted in this thesis, I adopt here Nunes and Hornstein?s treatment of ATB and maintain a movement analysis for BP 3 rd P null possessor. However, let me observe that the present analysis does not provide an account for the parallelism requirement on ATB extractions, according to which movement from coordinated structures has to affect all the conjuncts. There still is no syntactic account for this requirement, 23 and Munn (1993) assumes it to be a semantic condition. For some as yet unclear reason, coordinate conjunctions have to be semantically similar with respect to the presence of variables. Nunes and Hornstein offer a minimalist version of Munn?s suggestion by taking the parallelism requirement to be a bare output condition. I 23 Citko (2002) argues that ATB are structures with multi-dominance. The shared constituent is merged simultaneously with the relevant element inside the conjuncts. (In (70a), for example, which lute is merged simultaneously with play and win,) Hence, when the given constituent is moved, a gap is concurrently created inside each conjunct. Note, however, that there is nothing syntactic in her system forcing multi- dominance on coordinate structures. Hence, structures involving single dominance and, consequently, a single gap (as (69)) are fine from a syntactic point of view. 344 will not discuss this issue any further; rather I will presuppose that an approach along the lines proposed above is correct. 6.6 Conclusions In this chapter, I analyzed BP 3 rd P null possessors as formed by movement. Presupposing that null possessors are possessive agreement morphemes (possessive Agr), I suggested that in the grammars under investigation possessive Agr was lost. Thus, null possessors are not licensed anymore. However, since nouns move to X, the movement of a possessor phrase to spec of XP, a genitive Case checking position, is optional; therefore, in this grammatical system, a possessive phrase can be sub-extracted, checking its feature against a sentential functional clause The following arguments were presented as evidence of movement: (i) the null possessors under consideration have an anaphoric behavior, requiring a sentential antecedent; (ii) they require their antecedent to be closest c-commanding DP in accordance with MLC; (iii) they fail to occur inside specific DPs (and relative clauses) which are known to be island for extractions. The occurrence of 3 rd P null possessors inside adjuncts and in coordinate structures in an across-the-board fashion were taken to instances of gaps formed via sideward movement. 345 I Appendix 3 rd P Null Possessors in Finnish In chapter five, I introduced the idea of treating the possessive agreement markers as having an argument status. In particular, at 5.5, I considered Standard Finnish possessive agreement markers, and suggested that that in this grammar 1 st and 2 nd Person agreement morphemes enter the computational system as independent lexical items, being first merged in the possessor position, i.e. as an argument of the possessum. This analysis was motivated by the fact that these morphemes behave syntactically as clitics. However, having said that, the issue is about 3 rd P possessive markers. As already discussed, these markers have an anaphoric behavior. Inside the possessive DP, the genitive possessive pronoun is obligatory inside the possessive DP (cf. (77)), unless the 3 rd P possessive agreement marker finds its antecedent within the sentence in which it occurs (cf. (78)). (77)a.*(h?nen) kirjansa his/her book-3 b. *(heid?n) kirjansa their book-3 (78) a.Maija 1 pit?? Kirjastansa 1/*2 Maija-Nom like-3Sg book-Ela-3 ?Maija like her book? 346 b.he 1 pit?vat kirjastansa 1/*2 they-Nom like-3 rd Pl book-Ela-3 ?They like their book? This is quite similar to the behavior of 3 rd P null possessors in BP, and we can actually treat the 3 rd P possessive marker in parallel to the Finnish verbal inflection, given that this morpheme is morphologically weak, since there is only one morphological form for 3 rd Singular and plural. Under the proposal of this dissertation, this means that this 3 rd P possessive marker became f-defective, being unable to instantiate number distinctions and, as a consequence, it may have been reanalyzed as part of the noun, losing its syntactic independence. However, as I have already suggested, in Finnish, similarly to what happens in BP, nouns adjoins to X, satisfy, thus, the nominal requirements of the head X. This hypothetical analysis provide a rationale for why (77) is not generated in Standard Finnish grammars, and also suggests that (78) is formed by movement along the lines proposed in 5.2 for BP null possessors. In (78a) for instance, the DP Maija was first merged inside the possessive DP, then moved to spec of the matrix vP, and after that to spec of TP, where it had its Case feature checked. There are some pieces of evidence supporting a movement analysis for (78). First, as already shown, a 3 rd P possessive agreement marker in these grammars is anaphoric, requiring the presence of suitable antecedent within the sentence in which it occurs. This is confirmed by the data in (79). In (79a), there is not possessive marker, and thus the 347 relational noun is interpreted as non-relational, and the sentence is acceptable. In (79b), which contains a possessive marker, the possessor is realized by genitive pronoun. (79c), also contains a 3 rd P possessor marker, but it is not provided with an antecedent; thence the sentence is not grammatical. (79)a. Nytt silt, ett veli lytyi show-3SG it-Abl that brother-Nom was.found-3Sg ? It seems that the brother was found? b. Nytt silt, ett h?nen veljens lytyi show-3Sg it-Abl that his brother-Nom-3 was.found-3Sg ?It seems that his brother was found? c. * Nytt silt, ett veljens lytyi show-3Sg it-Abl that brother-Nom-3 was.found-3Sg ?It seems that his brother was found? (80) shows that the antecedence relationship between a possessive marker and its antecedent obeys the Minimal Link condition, the antecedent being the most local c- commanding DP. (80)a.* [H?nen 1 k?rmeens] 2 puri ksivarttaan 1 his/her snake-Nom-Poss.3sg bit-3Sg arm-Part-3 ?His snake bit his arm? 348 b.* Maija 1 sanoi, ett k?rme 2 puri veljens 1 Maija-Nom said-3Sg that snake-Nom bit-3Sg brother-3 ?Maija said that the snake bit her brother? In addition to this, the possessive agreement under consideration displays other obligatory control properties. Split antecedents are not allowed (cf. (81)); under VP ellipsis, only a sloppy reading is available (82). Moreover, when the antecedent has the form only-NP, only a covariant is interpretation is allowed, as shown in (83). 24, 25 (81) Maija 1 lupasi Pekalle 2 , ett h?n 1 leikkaisi tukkansa *1+2 Maija-Nom promised-3Sg Pekka-All that she cut-Cond-3Sg hair-3 ?Maija promised Pekka that she would cut her hair? (82) Isois 1 ei leikkaa kynsin 1 , eik grandfather-Nom not-3Sg cut-3Sg nails-Part-3Sg, not-inclusive.clitic isoitikn grandmother-Nom-inclusive.Clitic ?Grandpa does not cut his nails and grandma doesn?t either? (= Grandma does not cut her own nails either) 24 Though some speakers allow an invariant interpretation, see Kayser (1998). 25 I failed to test the de se reading. 349 (83)Liisa uskoo, ett? [vain h?n] kunnioittaa miest??] Liisa-Nom think-3Sg that only she-Nom respect-3Sg husband-Part-3 ?Liisa thinks that only she is an x such that x respects x?s husband? In addition to that, it is worthy mentioning that an overt possessive genitive pronoun is always disjoint in reference from a local c-commanding DP. (84)a.Maija 1 hukkasi [h?nen *1/2 kasettinsa] Maija-Nom lost-3Sg her tape-3 ?Maija lost her tape? b. [Maijan 1 sisar] 2 hukkasi [h?nen 1/*2 kasettinsa] Maija?s sister lost-3Sg her tape-3 ?Maija?s sister lost her tape? c.Maija 1 sanoi, ett? k?rme puri [hnen 1 ksivarttaan 1 ] Maija-NOM said-3Sg that snake-NOM bit-3Sg her arm-Part-3 ?Maija said that a snake bit her arm? This can be analyzed in the following way: in configurations that favor obligatory control, insertion of a pronoun is allowed to suppress movement only if the pronoun obviates in reference. 26 This accords to Hornstein?s (1999) economy principle of ?Move First?, which asserts that pronominal insertion is an expensive operation that applies only 26 Of course, this analysis poses a question about lack of obviation effects in BP. 350 when movement is unable to take place. For a discussion on the relation between control and ?Move first? see also San Martin and Hornstein (2001). 27 All in all, the data discussed above support a movement analysis for external possessors in Finnish. However, before concluding that this is the right analysis, the following facts have to be sorted out. First, as shown in (85), definite possessive DPs might have an external possessor, even though they block wh-extraction. (85)a.* Kenen 2 Jukka 1 tapasi tuon [t 2 velje(ns)] whose Jukka-Nom met-3Sg that-Acc brother-Acc-3 ?Whose did Jukka met that brother of? b. Jukka 1 tapasi [tuon veljens 1 ] Jukka-Nom met-3Sg that-Acc brother-Acc-3 ?Jukka met that brother of his Second, possessive DPs inside a relative clause are not allowed. This is rather unexpected given that null subject data suggest that relative clauses do allow A-extraction (cf. 4.5). It seems, then, that in this language A-extraction out of relative clause is allowed unless the constituent is inside a DP. That is, movement out of an island is fine, but movement out of an island within an island is not. 27 Richards (1997) also proposes that disjoint reference follows from economy. Sorting out binding relations in Norwegian and Japanese, he suggested that disjoint reference results from the principle of structural economy, according to which, given a choice between two elements, the one with fewer specified features is preferred. This can be seen as reformulation of the Avoid Pronoun Principle as an economy principle. 351 (86)a. * Jukka 1 n?ki [jtkn, jota tyttrens 1 suuteli] Juka-Nom saw-3Sg uy-Ac who-Part daughter-Nom-3 kised-3Sg ?Jukka saw the guy that his daughter kissed? b. * Jukka 1 li [jtk, joka suuteli tytrtn 1 ] Juka-Nom hit-3Sg guy-Part who-Nom kised-3Sg daughter-Part-3 ?Jukka hit the guy that kissed his daughter? I will not discuss these issues any further, rather I will leave them aside, awaiting more fine-grained research on Finnish DPs. 352 CHAPTER 7 CONCLUSION 7.1A Summary of the Content The presented research was driven by the following question: (1) What is the syntactic nature of clausal and nominal referential null subjects in BP and Finnish? I have answered (1) by correlating the licensing of referential null subjects with rich verbal agreement morphology. Thus, since BP and Finnish are languages with poor possessive and verbal agreement morphology, I hypothesized that these grammars are non pro-drop. That amounts to saying that the null subjects under investigation are not null pronouns (pro). In fact, these gaps are formed by movement, rather than lexically empty categories. As I have shown, 1 st P null subjects are formed via movement to a topic position and deleted later, in accordance to a topic-deletion analysis. Embedded 3 rd P null subjects 353 are formed via A-movement: a DP inserted as an argument of an embedded finite clause or possessive DP undergoes A-movement to a higher syntactic domain, in which it Checks its Case feature. The evidence that the 3 rd P null subjects under consideration are formed by movement is listed in (3): (3)a.They are anaphoric, requiring a syntactic antecedent, which is the closest c-commanding DP. b.They pass all the diagnostics used to characterize obligatory control as formed by category movement. c. They are allowed inside relative clauses only if the head of the relative clause does not intervene between them and their antecedents. d. They are not allowed inside paratactic configurations. e. Their antecedents control the value of the gender feature of a past participle form or a floating quantifier embedded under their c-command domain. Nominal 3 rd P null subjects (i.e.; null possessors) appear to have the same properties. They require an antecedent, which is the closest c-commanding NP; are not allowed inside relative clauses and display all the characteristics of obligatory controlled gaps. Moreover, inside coordinated DPs, they must occur across-the-board. Hence, it is arguably the case that in BP and Finnish, 3 rd P null possessors are also residues of 354 movement. Presupposing that in pro-drop languages pro is the verbal agreement morpheme (Agr) itself, I have suggested that in BP and Finnish 3 rd P verbal agreement morphemes (Agr) underwent a degradation in j-feature and, as a consequence, was reanalyzed as part of the verb. Nevertheless, it still contains a D-feature. Hence, assuming that an EPP T is checked by a D-feature, it plausible that in these grammars when V adjoins to T, the D feature of Agr checks the EPP features of T. Within DPs, a similar state of affairs is observed: the EPP feature X, the nominal counterpart of T, is checked via N-to-X movement. Thus, in these languages, the EPP feature of the functional categories responsible for checking the Case of subjects is satisfied by other means, rather than by projection of a specifier/subject. It is the result of this morphological change which gives raise to the unexpected, yet observed, A-movement out of a Case Domains. A DP merged as the external argument of a predicate inside a finite clausal or a nominal domain can move out of that domain without engaging into spec-head relationship with a Case- checking functional category (T or X) because The EPP of feature of the Case-checking category is checked by Agr. Thus, assuming that Case is checked in Spec-Head relation, in BP and Finnish, a DP can move out of a Case domain before checking its Case feature. The schemes in (3) summarizes the proposed analysis: 355 (3) a. A-movement out of clauses b. A-movement out of nominals ? ?. CP DP 3 3 C TP D XP 3 3 [T EPP ,V D ] VP [X EPP , N] NP 3 3 DP Case V? DP Case N? 2 2 t ? t ? 7.2Some Theoretical Implications of This Research At this point it is worth pointing out some broad implication of the research reported in this thesis. First, if q-roles are the result of configurations, it is difficult to see how one could analyze the phenomenon depicted here. Moreover, if the Chain Condition is part of UG (Chomsky and Lasnik 1993), that would present some problems for the movement analysis I suggested. Finally, sidewards movement ought to be applied by the computational system, making available A-movement out of adjunct clauses. To that extent, this research is evidence for a featural view of q-roles and for sidewards movement. 7.3Possible Ramifications of the Proposed Analysis Before finalizing this thesis, let me briefly show that the A-movement analysis suggested above is extendable to languages other than BP and Finnish. Recently it has been observed that, in certain Balkan languages and Hebrew, 356 obligatory control is possible in subjunctive clauses (cf. Borer 1989, Varlokosta 1993, Kapetangianni and Seely, 2004 and Landau 2004). Using Landau?s term, we might refer to this type of obligatory control as finite control. The sentences in (4) are examples of finite control. 28 (4)a.Hem 1 kivu se& e 1/*2 yelxu ha-bayta mukdam they hoped-3Pl that go-3 rd PlFut home early ?They hoped that they would go home earlier? (Hebrew - Landau 2004) b. O Akrisio theli na kalesi tus filus sta the-Nom Acrisio-Nom wants-3Sg na invites-3Sg the-Acc friends-Acc tu sta genethlia tu his-Gen to birthday-Acc his-Gen ?Acrisio wants to invite his friends to his birthday party? (Greek - Kapetangianni and Seely 2004) As the Greek data in (5) show, these constructions display all the properties of obligatory control: (i) the empty subject is anaphoric, being co-referent with a c- commanding DP (5a&b); (ii) split antecedents are not allowed (5c); (iii) under VP ellipsis only a sloppy reading is possible (5d); and (iv) a de-se interpretation is obligatory (5d). (the data in (5) were extracted from Kapetangianni and Seely 2004.) 28 Na is Greek subjective marker. 357 (5) a. [o Akrisio] 1 kseri e 1/*2 na horepsi the-Nom Acrisio-Nom knows-3sg na dances-3Sg ?Acrisio 1 knows (how) to dance? b. [o filos tu Akrisio 1 ] 2 kseri e *1/2 na the-Nom friend-Nom the-gen Acrisio-gen knows-3Sg na horevi dances-3Sg ?Acrisio?s friend knows how to dance? c. *[o Akrisio] 1 nomizi oti [i Maria] 2 the-Nom Acrisio-Nom thinks-3Sg that the-Nom Maria-Nom kseri e 1+2 na voithisun o enas ton allo sta genethlia tu knows-3Sg na help-3Pl each other to birthday-acc his-Gen ?Acrisio 1 thinks that Mary 2 knows (how) to help each other at his birthday party? d.[o Akrisio] 1 kseri na e 1 horevi,. the-Nom Acrisio-Nom knows-3Sg na dances-3Sg to idhio ke o Ntanie the same and the-Nom Daniel-Nom ?Acrisio knows how to dance, so does Daniel? (?Sloppy/*strict reading) 358 e.[o Atichis] 1 kseri e 1 na ektimisi to metalio the-Nom unfortunate-Nom knows-3Sg na appreciate-3Sg the-acc medal-acc ?The Unfortunate knows to appreciate the medal? (de se reading) Landau observes that finite control is found in other grammars. His list includes Albanian, Romanian, Spanish and Kannada. Interestingly, as he points out, these are all bona-fide pro-drop languages, with the exception of Hebrew, which is arguably another partial pro-drop grammar. Hebrew allows referential null subjects only in past and future tense clauses and, like BP and Finnish, licenses referential 3 rd P null subjects only within embedded clauses. 29 Putting the partial pro-drop nature of Hebrew aside for a moment, it is important to notice that the languages listed by Landau are similar to BP and Finnish in that they also have null expletives. Thus, if the licensing null expletives means that the EPP features of T is satisfied via V-to-T movement, we can in principle propose that the sentences (4) are derived by a mechanism similar to the one represented in (3a). One might in fact suggest that Hebrew sentences like (4a) is derived by (3a). As in BP and Finnish, the verbal agreement in Hebrew has a D-feature, but it is not syntactically independent, being rather an affix to the verb. Thus, (4a) is derived as shown in (6). The pronoun hem is first inserted in the spec of the embedded VP, where it checks the external q-role of the embedded verb. When the embedded T is inserted into the derivation, V adjoins to T, and Agr checks the EPP feature of T. Assuming that F is an active functional projection in Hebrew (cf. Rodrigues 2003 for independent evidence), 29 . It is worth mentioning however that not all speakers allow 3 rd P referential null subjects within past-tense embedded clauses. 359 the pronoun hem in (6) moves to the spec of the embedded FP before moving to the spec of the matrix VP where it checks the external q-role of kivu ?hope?. When the matrix T is inserted, the pronoun moves to spec of TP and checks its Case feature. (6) [?[ TP [hem] 1Case [ T? [ T kivu 3 , T] [ VP t 1 [ VP t 3 [ CP s&e [ FP t 1 [ TP [ T yelxu 2 , T] [ VP t 1 they hope-3Pl that go-3PlFut [ V? t 2 ha-bayta mukdam]]]]]]]]]] home earlier Assuming the analysis proposed in chapter 4 for pro-drop languages, (4b) cannot be derived from (3a). In Greek verbal agreement morphemes are syntactically independent, being base generated as an argument of the verb. Thus, in (4b), the DP o Akrisio ?the Acrisio? must be lexically inserted at the left periphery of the embedded clause and from there it moves to the matrix clauses. Following the proposal laid out in chapter 4, I will assume here that in pro-drop languages with rich agreement morphology overt subjects are lexically inserted in spec of FP, being doubled by Agr, which is inserted in the argument position. I will not go through the details of the analysis, but it is important to note that the data presented in this section suggest that A-movement out of finite clauses is possible in languages in which the verbal agreement morpheme carries a D-feature, which facilitates the type of A-movement discussed in this thesis. 360 References Abney, S. 1986. The English Noun Phrase in Its Sentential Aspect. Doctoral dissertation, MIT. Alexiadou, A. and E. Anagnostopoulou. 1998. Parametrizing Agr: Word Order, V- Movement, and EPP-Checking. Natural Language and Linguistic Theory 16: 491- 539. Almeida, A. B. 1993. Pronomes Possessivos de 3 a Pessoa no Portugu?s Falado de S?o Paulo. Ms, University of S?o Paulo. Anderson, M. 1979. Noun Phrase Structure. Doctoral dissertation, University of Connecticut, Storrs. Anderson, M. 1983. Prenominal Genitive NPs. The Linguistic Review 3: 1-24. Aoshima, S. 2000. Mono-Clausal Control Constructions in Japanese. Proceedings of the Seoul International Workshop on Generative Grammar, 33-49. Authier, J-.M. 1988. The Syntax of Unselective Binding. Doctoral dissertation, University of Southern California, Los Angeles. Baker, M. 1997. Thematic Roles and Syntactic Structure. In L. Haegeman (ed.). Elements of Grammar. Dordrecht: Kluwer. 361 Barbosa, P. 1995. Null Subjects. Doctoral dissertation, MIT. Barker, C. 1995. Possessive Descriptions. Stanford: CSLI.. Barrie, M. 2004. Moving Towards Partial Control. Proceedings of the North Eastern Linguistics Society 34. Belletti, A. 2000. Agreement Projections. In: M. Baltin and C. Collins (eds.) The Handbook of Syntactic Theory. London: Blackwell. Benamamoun, E. 2000. The Feature Structure of Functional Categories: A Comparative Study of Arabic Dialects. Oxford: Oxford University Press. Bernstein, J. B. 1993. Topics in the Syntax of Nominal Structure across Romance. Doctoral dissertation, CUNY, New York. Bernstein, J. B. 1997. Demonstratives and Reinforcers in Romance and Germanic Languages. Lingua 102: 87?113. Bernstein, J. B. 2001. The DP Hypothesis: Identifying Clausal Properties in the Nominal Domain. In M. Baltin and C. Collins (eds.) The Handbook of Contemporary Syntactic Theory. London: Blackwell. Bianchi, V. 1999. Consequences of Antisymmetry: Headed Relative Clauses. Berlin: Mouton de Gruyter. Bobaljik, J. D and S. Brown. 1997. Interarboreal Operations: Head Movement and the Extension Requirement. Linguistic Inquiry 28: 345-357. Boeckx, C. 1998. Traces of Argument Structure. Ms, University of Connecticut, Storrs. 362 Boeckx, C. 2001. Mechanisms of Chain Formation. Doctoral dissertation, University of Connecticut, Storrs. Boeckx, C. 2003. Islands and Chains. Amsterdam: John Benjamins. Boeckx, C. and S. Stjepanovic. 2001. Head-ing toward PF. Linguistic Inquiry 32: 345-55. Borer, H. 1983. Parametric Syntax. Dordrecht: Foris. Borer, H. 1986. I-Subjects. Linguistic Inquiry 17: 375-416. Borer, H. 1989. Anaphoric Agr. In O. Jaeggli and K. Safir (eds.) The Null Subject Parameter. Dordrecht: Kluwer. Boskovic, Z. 2001. Floating Quantifiers and the q-Role Assignment. Proceedings of the North Eastern Linguistics Society 30: 83-107. Boskovic, Z. 1994. D-structure, Theta Criterion, and Movement into Theta-Positions. Linguistic Analysis 24: 247-86. Boskovic, Z. 1997 The Syntax of Nonfinite Complementation: An Economy Approach. Cambridge, Mass.: MIT Press. Boskovic, Z. and D. Takahashi. 1998. Scrambling and Last Resort. Linguistic Inquiry 29: 347-66. Boskovic, Z., and S. Franks. 2000. Across-the-Board Movement and LF. Syntax 3: 107 28. Borer, H. and Y. Grodzinsky. 1986. Lexical Cliticization vs. Syntactic Cliticization. In H. Borer (ed.) The Syntax of Pronominal Clitics, Syntax and Semantics Vol. 19. New York: Academic Press. 363 Branigan, P. 1992. Subjects and Complementizers. Doctoral dissertation, MIT. Britto, H. 1998. Deslocamento a Esquerda, Resumptivo-Sujeito, Ordem SV e a Codifica??o Sint?tica de Ju?zos Categ?ricos e T?tico no Portugu?s do Brasil. Doctoral dissertation, University of Campinas. Brody, M. 1993. q-Theory and Arguments. Linguistic Inquiry 24: 1-23. Browning, M. 1987. Null Operator Constructions. Doctoral dissertation, MIT. Brug?, L. 1996. Demonstrative Movement in Spanish: a Comparative Approach. University of Venice Working Papers in Linguistics 6.1. University of Venice, Venice. Burzio, L. 1981. Intransitive Verbs and Italian Auxiliaries. Doctoral dissertation, MIT. Calabrese, A. 1986. Pronomina. MIT Working Papers in Linguistics 8. MIT, Cambridge, Mass. Cardinaletti, A. 1998. On the Deficient/Strong Opposition in Possessive Systems. In A. Alexiadou and C. Wilder (eds.) Possessors, Predicates, and Movement in the Determiner Phrase. Amsterdam: John Benjamins. Cardinaletti, A. and Starke, M. 1994. The Typology of Structural Deficiency. Ms, University of Venice and University of Geneva. Carlson, G. N. 1977. A Unified Analysis of the English Bare Plural. Linguistics and Philosophy 1: 413?56. Casta?eda, H.-N. 1966. ?He?: A Study in the Logic of Self-Consciousness. Ratio 7: 130- 57. Castillo, J. 2001. Thematic Relations between Nouns. Doctoral dissertation, University of Maryland. 364 Castillo, J., J. Drury and K. Grohmann. 1999. The Status of Merge over Move Preference. University of Maryland Working Papers in Linguistics 8. University of Maryland, College Park. Cavalcante, S. 2003. On the Arbitrary/Generic Null Subject in Brazilian Portuguese. Ms, University of Campinas. Cerqueira, V. C. 1996. A Forma Genitiva ?Dele? e a Categoria de Concord?ncia (AGR) no Portugu?s do Brasil. In M. Kato and I. Roberts (eds.). Cheng, L.-S. and E. Ritter. 1987. A Small Clause Analysis of Inalienable Possession in Mandarin and French. Proceedings of NELS 18: 65-78. Chierchia, G. 1984. Anaphoric Properties of Infinitives and Gerunds. Proceedings of the 3rd West Coast Conference on Formal Linguistics 3: 28-39. Chierchia, Gennaro: 1990. Anaphora and attitudes de se. In R. Bartsch, J. van Benthem & P. van Emde Boas (eds.) Semantics and Contextual Expression Foris: Dordrecht. Chomsky, N. 1957. Syntactic Structure. The Hague: Mouton. Chomsky, N. 1965. Aspects of the Theory of Syntax. Cambridge, Mass.: MIT Press. Chomsky, N. 1970. Remarks on Nominalization. In R. A. Jacobs and P. S. Rosenbaum (eds.) Readings on English Transformational Grammar. Waltham, Mass.: Ginn. Chomsky, N. 1973. Conditions on Transformations. In S. R. Anderson and P. Kiparsky (eds.) A Festschrift for Morris Halle. New York: Holt, Rinehart & Winston. Chomsky, N. 1977. On Wh-Movement. In P. Culicover, T. Wason and A. Akmajian. (eds.) Formal Syntax. New York: Academic Press. Chomsky, N. 1981. Lectures on Government and Binding. Dordrecht: Foris. 365 Chomsky, N. 1982. Some Concepts and Consequences of the Theory of Government and Binding. Cambridge, Mass.: MIT Press. Chomsky, N. 1986a. Knowledge of Language: Its Nature, Origin and Use. New York: Praeger. Chomsky, N. 1986b. Barriers. Cambridge, Mass.: MIT Press. Chomsky, N. 1993. A Minimalist Program for Linguistic Theory. In K. Hale and S. J. Keyser (eds.). The View from Building Twenty. Cambridge, Mass.: MIT Press. Chomsky, N. 1995a. Bare Phrase Structure. In G. Webelhuth. (ed.) Government and Binding Theory and the Minimalist Program. Oxford: Blackwell. Chomsky, N. 1995b. Categories and Transformations. In The Minimalist Program. Cambridge, Mass.: MIT Press. Chomsky, N. 2000. Minimalist Inquiries: the Framework. In R. Martin, D. Michaels and J. Uriagereka (eds.) Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik. Cambridge, Mass.: MIT Press. Chomsky, N. 2001a. Derivation by Phase. In M. Kenstowicz (ed.). Ken Hale: A Life in Language, 1-52. Cambridge, Mass.: MIT Press. Chomsky, N. 2001b. Beyond Explanatory Adequacy. MIT Occasional Papers in Linguistics 20. MIT, Cambridge, Mass. Chomsky, N. and H. Lasnik. 1977. Filters and Control. Linguistic Inquiry 8: 425-504. Chomsky, N. and H. Lasnik. 1993. The Theory of Principles and Parameters. In J. Jacobs, A. von Watechow, W. Sternefeld and T. Vennemann (eds). Syntax: An International Handbook of Contemporary Research. Berlin: Walter de Gruyter. 366 Cinque, G. 1980. On Extraction from NP in Italian. Journal of Italian Linguistics 5: 47-99. Cinque, G. 1994. Evidence for Partial N Movement in the Romance DP. In G. Cinque (ed.) Paths toward Universal Grammar, Studies in honor of Richard Kayne. Georgetown University Press. Citko, B. 2002. ATB Wh-Movement and the Nature of Merge. Proceedings of the North Eastern Linguistics Society 34. Clark, R. 1992. The Selection of Syntactic Knowledge. Language Acquisition 2: 83-149. Cole, P., G. Hermon and C.-T. J. Huang. 2001. Introduction. In Cole, P., G. Hermon and C.-T. J. Huang (eds.) Long Distance Reflexives. New York: Academic Press. Crisma, P. 1990. Functional Categories inside the Noun Phrase: A Study of the Distribution of Nominal Modifiers. Doctoral dissertation, University of Venice. Crisma, P. 1996. On the Configurational Nature of Adjectival Modification. In K. Zagona (ed.), Grammatical Theory and Romance Languages. Selected Papers from the 25th LSRL. Cyrino, S. M. 1996. O Objeto Nulo no Portugu?s do Brasil. Doctoral dissertation, University of Campinas. Culicover, Peter W. and P. M. Postal. 2001. Parasitic Gaps. Cambridge, Mass.: MIT Press. Davidson, D. 1969. On Saying That. Synthese 19:130-46. den Dikken, M. 1998. Predicate Inversion in DP. In A. Alexiadou, A. and C. Wilder (eds.) Possessors, Predicates, and Movement in the Determiner Phrase. Amsterdam: John Benjamins Publishing. 367 De Wit, P. 1997. Genitive Case and Genitive Constructions. Utrecht: UiL OTS, Utrecht University. Diesing, M. Indefinites. Cambridge, Mass.: MIT Press. Dowty, David R. 1985, On Recent Analyses of The Semantics of Control. Linguistics and philosophy 8: 291-331. Duarte, M. E. 1995. A Perda do Princ?pio Evite Pronome no Portugu?s Brasileiro. Doctoral dissertation, University of Campinas. Duarte, M. E. 1996. Do Pronome Nulo ao Sujeito Pleno. In M. Kato and I. Roberts (eds.) Portugu?s Brasileiro: Uma Viagem Diacr?nica. Campinas, University of Campinas. Engdahl, E. 1983. Parasitic gaps. Linguistics and Philosophy 6: 5-34. En?. M. 1991. The Semantics of Specificity, Linguistic Inquiry 22: 1-25. Epstein, S. and D. Seely. 1999. SPEC-ifying the GF "Subject"; Eliminating A-Chains and the EPP within a Derivational Model. Paper presented at the LSA Summer Institute Workshop on Grammatical Relations. Fabb, N. 1990. The Difference between English Restrictive and Nonrestrictive Relative Clauses. Journal of Linguistics 26: 57-77. Farrell, P. 1995. Backward Control in Brazilian Portuguese. Proceedings of the Eastern States conference on linguistics 11: 116-127. Ferreira, M. 2000. Argumentos Nulos em Portugu?s Brasileiro. MA thesis, Campinas: University of Campinas. Fiengo, R and J. Higginbotham. 1981. Opacity in NP. Linguistic Analysis 7: 395-421 368 Figueiredo Silva, M. C. 1996. A Posi??o do Sujeito no Portugu?s Brasileiro. Campinas: Editora da Unicamp. Freeze, R. 1992. Existentials and Other Locatives. Language 68: 553-595. Freindin, R. and J-R. Verganaud. 2001. Exquisite Connections: Some Remarks on the Evolution of Linguistic Theory. Lingua 9: 639-666. Fodor, A. J. 1970. Three Reasons for Not Deriving ?Kill? from ?Cause to Die?. Linguistics Inquiry 1: 429-438. Fodor, A. I. Sag. 1982. Referential and Quantificational Indefinites. Linguistic and Philosophy 5: 355-98. Galves, C. 1996. O Enfraquecimento da Concord?ncia no Portugu?s Brasileiro. I n I. Roberts e M. Kato (eds.). Galves, C. 1998. T?picos, Sujeitos, Pronomes e Concord?ncia no Portugu?s Brasileiro. Cadernos de Estudos Ling??sticos IEL ? Unicamp 34: 7-21. Galves, C. 2001. Ensaios sobre as Gram?ticas do Portugu?s. Campinas: Editora da Unicamp. Gavruseva, E. 2000. On the Syntax of Possessor Extraction. Lingua 110: 743-772. Geuder, W. 2000. Oriented Adverbs: Issues in the Lexical Semantics of Event Adverbs. Doctoral dissertation, Universit?t T?bingen, Konstanz. Giorgi, A. and G. Longobardi. 1991. The Syntax of Noun Phrases. Cambridge: Cambridge University Press. 369 Gonz?lez, N. 1988. Object Raising in Spanish. New York: Garland. Goodall, Grant. 1987. Parallel Structures in Syntax. Cambridge: Cambridge University Press. Grimshaw, J. 1990. Argument Structure. Cambridge, Mass.: MIT Press. Groat, E. & J. O'Neil. 1996. Spell-Out at the Interface: Achieving a Unified Syntactic Computational System in the Minimalist Framework. In W. Abraham, S. D. Epstein, H. Thr?insson & C. J.-W. Zwart (eds.) Minimal Ideas: Syntactic Studies in the Minimalist Framework. Amsterdam: John Benjamins. Grohmann, K. 2003. Prolific Domains. On the Anti-Locality of Movement Dependencies. Amsterdam: John Benjamins. Grohmann, K. and L. Haegeman. 2002. Resuming Reflexives. Proceedings of the 19th Scandinavian Conference in Linguistics 19, University of Troms?. Gueron, J. 1995. Inalienable Possession, PRO-Inclusion and lexical chains. In J. Gueron, J. Pollock and H. Obenauer, (eds.) Grammatical Representation. Foris: Dordrecht. Guimar?es, M. 2000. In Defense of Vacuous Projections in Bare Phrase Structure. University of Maryland Working Papers in Linguistics 9, University of Maryland, College Park. Guimar?es, M. and C. Rodrigues. 2002. Locative Determiner and Argumental Pseudo- Adverbs. Ms., University of Maryland, College Park. Gutman, 1999. Null-Subjects: A Theory of Syntactic and Discourse Identification. Doctoral dissertation, University of Delaware. Ha?k, I. 1985. The Syntax of Operators. Doctoral dissertation, MIT. 370 Hakulinen, A. 1979. Suomen kielen rakenne ja kehitys [The Structure and Development of the Finnish Language]. Nelj?s, korjattu ja lis?tty painos. Helsinki: Otava. Hakulinen, A. and L. Karttunen. 1973. Missing Persons: On Generic Sentences in Finnish. Chicago Linguistics Society 9: 157-71. Hale, K. and S. J. Keyser. 1993. On Argument Structure and the Lexical Expression of Syntactic Relations. In: The View from Building Twenty, 53-109. Cambridge, Mass.: MIT Press. Hale, K. and S. Keyser. 1998. Bound Feature, Merge and Transitivity Alternations. In: L. Pylkk?nen, A. van Hout and H. Harley (eds.) Papers from the Uppen/MIT Roundtable on the Lexicon. MIT Working Papers in Linguistics, 35: 49-72. MIT, Cambridge, Mass. Hale, K. and S. Keyser. 2002. Prolegomenon of a Theory of Argument Structure. Cambridge, Mass.: MIT Press. Helasvuo, M.-L. 1994. Are relative clauses either restrictive or non-restrictive? A study of relative clauses in the Finnish Pear Stories. In A. Holmberg and U. Nikanne (eds.). Case and other functional categories in Finnish syntax. Berlin: Mouton de Gruyter. Higginbotham, J. 1988. Contexts, Models, and Meanings: a Note on the Data of Semantics. In R. Kempson (ed.) Mental Representation: the Interface between Language and Reality. Cambridge: Cambridge University Press. Higginbotham, J. 1992a. Reference and Control. In: R. Larson, S. Iatridou, U. Lihari and J. Higginbotham (eds.) Control and Grammar. Dordrecht: Kluwer. Higginbotham, J. 1992b. Anaphoric Reference and Common Reference. Ms, MIT, Cambridge, Massachusetts. Hintikka, J. 1986. The Semantics of ?a Certain?. Linguistic Inquiry, 17: 331-36. 371 Hoinenen, T. 1995. Null Subjects in Finnish: From Either-Or to More-Or-less. Ms, Research Institute for the Languages of Finland. Holmberg, A. 2003. Null Subjects and Uninterpretable Features: Evidence from Finnish. Ms., University of Oslo. Homlberg, A. The Syntax of Yes and No in Finnish. Studia Linguistica, 55: 141-75. Holmberg, A. and U. Nikanne. 2002. Expletives, Subjects, and Topics in Finnish. In: P. Svenonius, (ed.) Subjects, Expletives, and the EPP. Oxford: University Press. Holmberg, A. and U. Nikanne. 1994. Expletives and Subject Positions in Finnish. Proceedings of NELS, 24: 173-187. Holmberg, A., U. Nikanne. 1993. Introduction. In A. Holmberg & U. Nikanne (eds.) Case and Other Functional Categories in Finnish Syntax. Mouton de Gruyter, Berlin. Hornstein, N. 1995. Logical Form: From GB to Minimalism. Oxford: Blackwell. Hornstein, N. 1998. Movement and Chains. Syntax, 1:99-127. Hornstein, N. 1999. Movement and Control. Linguistic Inquiry, 30: 69-96. Hornstein, N. 2001. Move! A Minimalist Theory for Construal. Oxford: Blackwell. Hornstein, N. 2002. A Grammatical Argument for Neo-Davidsonian Semantics. In: G. Preyer and G. Peter (eds.) Logical Form and Language, 345-64. Oxford: Clarendon Press. Hornstein, N. 2003. On Control. In: R. Hendriks (ed.), Contemporary Grammatical Theory. Oxford: Blackwell. 372 Hornstein, N. and A. Weinberg. 1981. Case Theory and Preposition Stranding. Linguistic Inquiry, 12: 55-92. Hornstein, N and D. Lightfoot. 1981. Introduction. In N. Hornstein and D. Lightfoot (eds.) Explanation in Linguistics. Longman: London. Hornstein, N., S. Rosen and J. Uriagereka. 1995. Integral Predication. Proceedings of WCCFL 13. Hornstein, N. and A. Weinberg. 1999. On the Necessity of LF. The Linguistic Review, 7: 129-67. Hornstein, N. and I. San Martin. 2000. Obviation as Anti-Control. Paper presented at GLOW 2000, Universidad de Deusto & Universidad del Pais Vasco. Hornstein, N. and J. Nunes. 2002. On Asymmetries between Parasitic Gap and Across-the-Board Constructions. Syntax, 5:26-54. Hornstein, N. and M. Motomura. 2002. Psych Verbs, Theta Roles and Reconstruction. In: Proceedings of 2002 LSK International Summer Conference, Vol II, 39-58. Linguistics Society of Korea. Huang, C.-T. J. 1982. Logical Relations in Chinese and the Theory of Grammar. Doctoral dissertation, MIT. Huang, C.-T. J. 1984. On the Distribution and Reference of Empty Pronouns. Linguistic Inquiry, 15:531-74. Jackendoff, R. 1977. X-bar Syntax. Cambridge, Mass.: MIT Press. Jackendoff, R. 1983. Semantics and Cognition. Cambridge, Mass.: MIT Press. Jackendoff, R. 1990. Semantic Structure. Cambridge, Mass.: MIT Press. 373 Jackendoff, R. 1997. The Architecture of the Language Faculty. Cambridge, MA.: MIT Press. Jaeggli, O. 1982. Topics on Romance Syntax. Foris: Dordrecht. Jaeggli, O. and Safir, K. 1989. The Null Subject Parameter. Dordrecht: Kluwer. Jenkins, L. 2000. Biolinguistics. Cambridge, England: Cambridge University Press. Johnson, K. 1994. Bridging the Gap. Ms., University of Amherst. Johnson, K. 1991. Object Positions. Natural Language and Linguistic Theory. 9: 577- 636. Kaiser, E. 1998. 2001. Locality in Anaphoric Relations: A Look at Finnish. Paper presented at the 75th LSA Annual Meeting. Washington D.C. Kanerva, J. M. 1987. Morphological Integrity and Syntax: The Evidence from Finnish Possessive Suffixes. Language, 63: 498-521. Kepetangianni, K. and D. Seely. 2004. On Control in Modern Greek: Explanation by Deduction;Agreement by Degrees. Ms, Eastern Michigan University. Kato, M. 1999. Weak and Strong Pronominals. Probus, II: 1-37. Kato, M. 2000. The Partial Pro-Drop Nature and the Restricted Word Order. In: M. Kato and E. Negr?o. Brazilian Portuguese and the Null Subject Parameter. Frankfurt: Vervuert-Iberoamericana. Kato, M. and F. Tarallo. 1988. Restrictive VS Syntax in Brazilian Portuguese. Paper Presented at The Georgetown Round Table on Languages and Linguistics. 374 Kayne, R. 1975. French Syntax: the Transformational Cycle. Cambridge, Mass: The MIT Press. Kayne, R. 1989. Facets of Romance Past Participle Agreement. In: P. Beninc? (ed.), Dialect Variation on the Theory of Grammar. Dordrecht: Foris. Kayne, R. 1993. Toward a Modular Theory of Auxiliary Selection. Studia Linguistica, 47: 3-31. Kayne, R. S. 1994. The Antisymmetry of Syntax. Cambridge, MA.: MIT Press. Kiguchi, H. 2002. Syntax Unchained. Doctoral dissertation, University of Maryland, College Park. Kitahara, H. 1995 Target Alpha: Deducing Strict Cyclicity from Derivational Economy. Linguistic Inquiry, 26: 47-77. Koizumi, M. 1993. Object Agreement Phrase and the Split VP Hypothesis. In: J. D. Bobaljik and C. Philips (eds.) Papers on Case and Agreement I. MIT Working Papers in Linguistics, 18: 99-148. MIT, Cambridge, Mass. Kornfilt, J. 1984. Case Marking, Agreement, and Empty Categories in Turkish. Doctoral dissertation, Harvard University. Koizumi, M. 1994. Secondary Predicates. Journal of East Asian Linguistics, 3: 25-79. Kroch, A. 1989. Reflexes of Grammar in Patterns of Language Change. Journal of Language Variation and Change, 1: 199-244. Kuroda, S.-Y. 1972. The Categorical and the Thetic Judgment. Foundations of Language, 9:153-185. Kuroda, S.-Y. 1999. Notes on So-Called Head-Internal Relative Clauses in Japanese. In: M. Muraki and E. Iwamoto (eds.). Linguistics: In Search of the Human Mind: A Festschrift for Kazuko Inoue, 414-29. Tokyo: Kaitakusha. 375 Koster, J. 1984. On Binding and Control. Linguistic Inquiry, 15:417-59. Landau, I. 1999a. Elements of Control. Doctoral dissertation, MIT. Landau, I. 1999b. Possessor Raising and the Structure of VP. Lingua, 107: 1-37. Landau, I. 2000. Elements of Control: Structure and Meaning in Infinitival Constructions. Dordrecht: Kluwer. Landau, I. 2003. The Scale of Finiteness and the Calculus of Control. Ms. Ben Gurion University. Larson, R. K. 1988. On Double Object Constructions. Linguistic Inquiry, 19: 335-91. Larson, R. K. 1990. Double objects revisited: Reply to Jackendoff. Linguistic Inquiry, 21: 589-632. Larson, R. K. 1991. "Promise" and the Theory of Control. Linguistic Inquiry, 22:103-39. Larson, R. K. 2004. The Projection of DP. Paper presented at the University of Stuttgart, Stuttgart, Germany. Lasnik, H. 1976. Remarks on Coreference. Linguistic Analysis, 2:1-22. Lasnik, H. 1989. On the Necessity of the Binding Conditions. In: H. Lasnik. Essays on Anaphora. Dordrecht: Kluwer. Lasnik, H. 1995. Last Resort. In: S Haraguchi and M. Funaki (eds.). Minimalism and Linguistic Theory. Tokyo: Hitizi Syobo, 1-32. Lasnik, H. 1999a. A Note on Pseudogapping. In: Minimalist Analysis, 151-74. Oxford: Blackwell. 376 Lasnik, H. 1999b. Chains of Arguments. In: S. D. Epstein and N. Hornstein (eds.) Working Minimalism, 189-215. Cambridge, MA.: MIT Press. Lasnik, Howard. 1999c. Minimalist Analysis. Oxford: Blackwell. Lasnik, H. 2000. Syntactic Structures Revisited: Contemporary Lectures on Classic Transformational Theory. (With M. Depiante and A. Stepanov) Cambridge, Mass.: MIT Press. Lasnik, H. 2002. The Minimalist Program in syntax. Trends in Cognitive Sciences, 6:432- 437. Lasnik, H. 2003. Minimalist Investigations in Linguistic Theory. London: Routledge. Lasnik, H. and J. Uriagereka. 1988. A Course in GB Syntax. Cambridge, Mass.: MIT Press. Lasnik, H. and M. Saito, 1992. Move a: Conditions on Its Application and Output. Mass.: Cambridge, MIT Press. Lasnik, H. and M. Saito. 1991. On the Subject of Infinitives. In: L. Dobrin, L. Nichols, and R. Rodriguez (eds.) Papers from the 27th Regional Meeting of the Chicago Linguistic Society, 324-343. Lasnik, H. and M. Saito. 1984. On the Nature of Proper Government. Linguistic Inquiry, 15: 235-89. Lees, R. B. 1960. The grammar of English nominalizations. Doctoral dissertation, MIT, Cambridge: Mass. Lightfoot, D. 1989. The Child Trigger Experience: Degree-0 Learnability. Behavioral and Brain Sciences, 12: 321-75. 377 Lightfoot, D. 1991. How to Set Parameters: Arguments from Language Change. Mass.: Cambridge, MIT Press. Lightfoot, D. 1994. Degree-0 Learnability. In B. Lust, G. Harmon and J. Kornfilt (ed.) Theory and First Language Acquisition: Cross-Linguistic Perspective. Vol. II: Binding Dependencies and Learnability. Hillsdale, NJ: Lawrence Erlbaum Associates. Lightfoot, D. 1999. The development of language: Acquisition, change and evolution. Oxford: Blackwell. Lightfoot, D. and C. Rodrigues. 2003. Subject Inversion and Clitic Climbing. In: I. Cagri, L. Meroni and G. Tesan. University of Maryland Working Papers in Linguistics, 12. University of Maryland, College Park. Lobato, L. 2000. Causes and Consequences of Linguistic Change: the Case of Portuguese Subject Position. Paper presented at DIGS VI Diachronic Generative Syntax. University of Maryland, College Park. Longobardi, G. 2001. The Structure of DPs: Some Principles, Parameters and Problems. In: M. Baltin and C. Collins (eds.). L?pez, L. 2001. On the (Non)complementarity of q-theory and Checking Theory. Linguistics Inquiry, 32: 694-716. Lunguinho, M. 2003. Parti??o de Constituintes no Portugu?s do Brasil: Caracter?sticas Sint?ticas e Sem?nticas. Ms. University of Brasilia. Manzini, Maria Rita: 1983. On control and control theory. Linguistic inquiry, 14: 421- 446. Manzini, M. R. and A. Roussou. 2000. A Minimalist Theory of A-Movement and Control. Lingua, 110: 409:447. 378 Manzini, M. R and L. Savoia. 2002. Parameters of Subject Inflection in Italian Dialects. In P. Svenonius (ed.). Marantz, A. 1997. No Escape from Syntax: Don't Try Morphological Analysis in the Privacy of Your Own Lexicon. In: A. Dimitriadis, L. Siegel, et al. (eds.) University of Pennsylvania Working Papers in Linguistics, Vol. 4.2: 201-225. Marantz, A. 1995. The Minimalist Program. In: G. Webelhuth. (ed.), 349-381. Martin, J. 1995. On the syntactic structure of Spanish DPs. Doctoral dissertation, University of Southern California. Martin, R. 1996. A Minimalist View of PRO and Control. Doctoral Dissertation, University of Connecticut. Martin, R. and J. Uriagereka. 2000. Some possible foundations of the minimalist program. In: R. Martin, and J. Uriagereka (eds.) Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik, 1-30. Cambridge, Mass.: MIT Press Martins, A. M. 1994. Cl?ticos na Hist?ria do Portugu?s. Doctoral dissertation, University of Lisbon. Matsuya, A. 2000. Theta Movement in Japanese Syntax. Doctoral dissertation, University of Maryland, College Park. Matushansky, O. 2003. Going through a Phase. Ms. University of Paris-8. Menuzzi, S. 1999. Binding Theory and Pronominal Anaphora in Brazilian Portuguese. Ph.D. dissertation, Leiden University. Modesto, M. 2000. Null Subjects without ?Rich? Agreement. Doctoral dissertation, University of South California. Moltmann, F. 2003. Propositional Attitudes without Propositions. Synthese, 135: 77-118. 379 Montalbetti, M. 1984. After Binding. Doctoral dissertation, MIT. Moreira da Silvia, 1983. ?tudes sur la Sym?trie et l?Asym?trie Sujet/Objet dans le Portugais du Br?sil. Doctoral dissertation, University of Paris VIIII. M?ller, G. 1998. Incomplete Category Fronting: A Derivational Approach to Remnant Movement in German. Kluwer. Munn, A. 1993. Topics in the Syntax and Semantics of Coordinate Structures. Doctoral dissertation, University of Maryland, College Park. Muromatsu, K. 1994. Integrals in Japanese. In J. Nunes, E. Thompson and S. Varlokostova (eds.) University of Maryland Working Papers in Linguistics, 2: 136- 158. Negr?o, E. V. 1999. O Portugu?s Brasileiro:Uma L?ngua Voltada para o Discurso. Ms., University of S?o Paulo. Negr?o, E, and A. L. M?ller. 1996. As Mudan?as no Sistema Pronominal do Portugu?s Brasileiro: Substitui??o ou Especializa??o de Formas? Delta, 12: 125-152. Nevis, J. 1984. Five Morphemes in Finnish: Possessive Suffixes or Anaphoric Clitics. Ohio State Working Papers in Linguistics, 29: 174-207. Nevis, J. 1985. Finnish Particle Clitics and General Clitic Theory. Doctoral dissertation, Ohio State University. Nikanne, U. 1994. On Movement of Spec(IP) Position on Finnish. In: N. Cover and H. van Riemsdijk (eds.) Studies on Scrambling. Berlin: Mounton the Gruyter. Nunes, J. 1995. The Copy Theory of Movement and Linerarization of Chains in the Minimalist Program. Doctoral dissertation, University of Maryland at College Park. 380 Nunes, J. 2004. Linearization of Chains and Sideward Movement. Cambridge, Mass.: MIT Press. Nunes, J. and J. Uriagereka. 2000. Cyclicity and extraction domains. Syntax, 3: 20-43. Oehrle, R and Nishio, H. 1981. Adversity. Coyote Papers 2, University of Arizona Linguistic Circle: Tuscon Arizona, 163-93. Ogawa, Y. 2001. A Unified Theory of Verbal and Nominal Projections. Oxford: Oxford University Press. O?Neil, J. 1995. Out of Control. In: J. N. Beckman (ed.) Proceedings of NELS, 25:361- 71. GSA, University of Massachusetts, Amherst. Ord?nez, F. 1997. Word Order and Clausal Structure in Spanish and other Romance Languages. Doctoral dissertation, University of New York. Ormazabal, J. 1991. Asymmetries on Wh-movement and Some Theoretical Consequences. Ms. University of Connecticut, Storrs. Pagotto. E.1992. A Posi??o dos Cl?ticos em Portugu?s: um Estudo Diacr?nico. Masters thesis, University of Campinas. Parsons, T. 1990. Events in the Semantics of English: A Study in Subatomic Semantics. Cambridge: MIT Press. Partee, B. 1987. Noun Phrase Interpretation and Type-Shifting Principles. In J. Groenendijk, D. de Jongh, and M. Stokhof (eds.) Studies in Discourse Representation Theory and the Theory of Generalized Quantifiers, GRASS 8. Foris: Dordrecht. Partee, B. H. and V. Borschev. 2000. Possessives, Favorite, and coercion. In A. Riehl and R. Daly (eds.) Proceedings of ESCOL 99, 173-190. Ithaca, NY: CLC Publication. 381 Pesetsky, D. 1992. Zero Syntax, Vol 2. Ms., MIT, Cambridge, Mass. Picallo, M. C. 1991. Nominals and Nominalizations in Catalan. Probus, 3: 279-316. Picallo, M. C. 1994. Catalan Possessive Pronouns: the Avoid Pronoun Principle Revisited. Natural Language and Linguistic Theory, 12: 259?99. Pierrehumbert, J. 1980. The Finnish Possessive Suffixes, Language, 56, 603-621 Pires, A. 2001. The Syntax of Gerunds and Infinitives: Subject, Case and Control. Doctoral dissertation, University of Maryland, College Park. Pires, A. and C. Rodrigues. 2002. Left-dislocated Non-finite Adjuncts. Paper presented at 2002 Meeting of LSA- Linguistic Society of America. San Francisco, California. Platzack, C. 1995. Null Subjects, Weak Agr and Syntactic Differences in Scandinavian. Working Papers in Scandinavian Syntax, 53: 85-110. Polinsky, M. and Potsdam, E. 2002. Backward Control: Evidence from Malagasy. In: A. Rakowski and N. Richards (eds.) Proceedings of the eighth meeting of the Austronesian Formal Linguistics Association. MIT Working Papers in Linguistics 44: 257-272 Postal, P. M. 2002. Skeptical Linguistic Essays. Ms., New York University. Postal, P. M. 1974. On Raising: One Rule of English Grammar and its Theoretical Implications. Cambridge, Mass.: MIT Press. Pylkk?nen, L. 2002. Applicative and Depictive Secondary Predication. Paper presented at LingLunch, MIT. 382 Raposo, E. 1999. Towards a Minimalist Account of Nominal Anaphora in Spanish and English. Ms, University of California, Santa Barbara. Raposo, E. and J. Uriagereka. 1995. Two Types of Small Clauses: Toward a Syntax of Theme/Rheme Relations. In: A. Cardinaletti and M.T. Guasti (eds.) Small Clauses, 179-206. New York: Academic Press. Raposo, E. and J. Uriagereka, 2002. Two Types of Small Clauses: Toward a Syntax of Theme/Rheme Relations. In: Derivations: Exploring the Dynamics of Syntax. London: Routlegde. Reinhart, T. and E. Reuland. 1993. Reflexivity. Linguistic Inquiry, 24: 657-720. Reinhart, T. and E. Reuland. 1991. Anaphors and Logophors: An argument Structure Perspective. In J. Koster and E. Reuland (eds.) Long Distance Anaphora. Cambridge: Cambridge University Press. Richards, N. 1997. Competition and Disjoint Reference. Linguistic Inquiry, 28:178-87 Ritter, E. 1991. Two Functional Categories in Noun Phrases: Evidence from Modern Hebrew. Syntax and Semantics, 25: 37-62. Rivas, A. A Theory of Clitics. Doctoral dissertation, MIT. Rizzi, L. 1982. Issues in Italian Syntax. Dordrecht: Foris. Rizzi, L. 1986. Null Objects in Italian and the Theory of Pro. Linguistic Inquiry, 17: 501- 57. Rizzi, L. 1990. Relativized Minimality. Mass.: Cambridge: MIT Press. Roberts, I. 1986. The Representation of Dethematized Subjects. Dordrecht: Foris. 383 Roberts, I. 1997. Restructuring, Head Movement, and Locality. Linguistic Inquiry, 28. Roberts, I and A. Roussou. 1999. A Formal Approach to Grammaticalisation. Linguistics, 37: 1011-1041. Rodrigues, C. 2000. Deriving Referential Null Subjects From Movement. Generals paper, University of Maryland, College Park. Rodrigues, C. 2002. Loss of verbal Morphology and Referential Subjects in Brazilian Portuguese. In: D. Lightfoot (ed.) Syntactic Effects of Morphological Change, 160- 178. Oxford: Oxford University Press. Rodrigues, C. 2003. What?s at the Left Periphery of Hebrew Copular Constructions. University of Maryland Working Papers in Linguistics 12. Department of Linguistics, University Maryland, College Park. Romero, J. 1997. Construcciones de Doble Objeto y Gram?tica Universal. Doctoral dissertation, Universidad Aut?noma de Madrid, Madrid. Ross, J. R. 1967. Constraints on Variables in Syntax. Doctoral dissertation, MIT. Ross, J. R. 1982. Pronoun Deletion in German. Paper presented at the annual meeting of The Linguistic Society of America. San Diego, California. Ross, J. R. 1984 Inner islands. Proceedings of BLS 10. Berkeley Linguistic Society, University of California, Berkeley Rouveret, A. 1989. Cliticisation et temps en Portugais Europ?en. Revue des Langues Romanes, 2: 337-71. Safir, K. 1986. Relative Clauses in a Theory of Binding and Levels. Linguistic Inquiry, 17: 663-89. 384 Silva, G. M de O. 1984. Varia??o no Sistema Possessivo de Terceira Pessoa. Tempo Brasileiro, 78/79: 54-72. Szabolcsi, A. 1981 The Possessive Construction in Hungarian: a Configurational Category in a Non-Configurational Language." Acta Linguistica Academiae Scientiarum Hungaricae, 31: 161-189. Szabolcsi, A. 1983. The Possessor That Ran Away from Home. The Linguistic Review, 3: 89-102. Szabolcsi. A. 1987. Functional Categories in the Noun Phrase. In I. Kenesei (ed.) Approaches to Hungarian, vol. 2. Jate Szeged. Szabolcsi, A. 1994. The Noun Phrase. In: F. Kiefer and K. ?. Kiss (eds.) The Syntactic Structure of Hungarian. New York: Academic Press. Saito, M and H. Hoshi. 2000. Japanese Light Verb Construction. In: D. Michaels, R. Martin , and J. Uriagereka (eds.), 261-295. Saito, M. and K. Murasugi. 1993. Subject Predication within IP and DP. Ms., University of Connecticut and Kinjo Gakuin University. Salmon, N. 1986. Reflexivity. Notre Dame Journal of Formal Logic, 27: 401-29. San Martin, I. 2002. On Subordination and the distribution of PRO. Doctoral dissertation, University of Maryland, College Park. Schachter, P. 1973. Focus and Relativization. Language: 19?46 Schein, B. 1995. Small Clause and Predication. In: A. Cardinaletti and M. T. Guasti (eds.) Syntax and Semantics: Small Clauses, 49-76. San Diego, CA: Academic Press. 385 Schoorlemmer, M. 1998. Possessors, Articles and Definiteness. In: A. Alexiadou and C. Wilder (eds.) Possessors, Predicates and Movement in the Determiner Phrase. Amsterdam: Benjamins. Sells, P. 1987. Aspects of Logophoricity. Linguistic Inquiry, 18: 445-479. Shlonsky, U. 1997.Clause Structure and Word Order in Hebrew and Arabic: An essay in Comparative Semitic Syntax. New York: Oxford University Press. Soltan, U. 2003. Locative Inversion: A Phase-Based Analysis. Generals Papers, University of Maryland, College Park. Speas, M. 1994. Economy, Agreement and the Representation of Null Arguments. Ms., University of Massachusetts, Amherst. Sportiche, D. 1983. Structural Invariance and Symmetry in Syntax. Doctoral dissertation, MIT. Sportiche, D. 1988. A Theory of Floating Quantifiers and Its Corrolaries for Constituent Structure. Linguistic Inquiry, 19: 425-449. Sportiche, D. 1996. Clitic Constructions. In: J. Rooryck and L. Zaring (eds.) Phrase Structure and the Lexicon. Dordrecht: Kluwer. Sportiche, D. 1998. Partitions and Atoms of Clause Structure. London: Routledge. Stowell, T. 1981. Origins of Phrase Structure. Doctoral dissertation, MIT. Cambridge, Mass. Stowell, T. 1982. The Tense of Infinitives. Linguistic Inquiry, 13: 561-70. Sulkala, H. and Karjalainen, M.1992. Finnish. London: Routledge. 386 Stowell, T. 1983. Subjects Across Categories. The Linguistic Review, 2: 285-312. Taraldsen, K.T. 1980. On the Nominative Island Condition, Vacuous Application, and The That-Trace Filter. Indiana University Linguistics Club. Tarallo, F. 1996. Diagonosticando uma Gram?tica Brasileira: O Portugu?s D?Aqu?m e D?Al?m-Mar ao Final S?culo XIX. In M. Kato and I. Roberts (eds.). Tellier, C. 1990. Underived Nominals and the Projection Principle: Inherent Possessors. Proceedings of NELS, 20: 472-486. Tellier, C. 1991. Licensing Theory and French Parasitic Gaps. Dordrecht: Kluwer. Tellier, C. and D. Valois. 1995. Agreement and Extraction out of DPs. Proceedings of WCCFL, 14: 525-540. Ticio, E. 2003. On the Structure of DPs. Doctoral dissertation, University of Connecticut, Storrs. Torrego, E. and J. Uriagereka. 1995. Parataxis. Ms, University of Maryland and Boston University. Watanabe, A. 1999. Multiple Theta Marking: A Preliminary Study. In: E. Iwamoto and M. Muraki (eds.), Linguistics: In Search of the Mind: A Festschrift for Kazuko Inoue, 708-36. Tokyo: Kaitakusha. Williams, E. 1980. Predication. Linguistic Inquiry, 11: 203-38. Williams E. 1982. The NP-Cycle. Linguistic Inquiry, 13.2 Williams, E. 1994. Thematic Structure in Syntax. Cambridge, Mass: MIT Press. 387 Ura, H. 1994. Varieties of Raising and the Feature-Based Bare Phrase Structure Theory. MIT occasional papers in linguistics, 23: 297-316. Cambridge, Mass. MITWPL. Uriagereka, J. 1988. On Government. Doctoral dissertation, University of Connecticut. Uriagereka, J. 1993a. Lecture Notes. Ms, University of Maryland, College Park. Uriagereka, J. 1993b. Specificity and the Name Constraint. In C. Mason, S. Powers and C. Schmitt (eds.) University of Maryland Working Papers in Linguistics. Uriagereka, J. 1995. Aspects of Syntax Clitic Placement in Western Romance. Linguistic Inquiry, 26: 79-132. Uriagereka, J. 1996. From Being to Having. In J. C. Castillo, V. Miglio and J. Musolino, (eds.) University of Maryland Working Papers in Linguistics, 4: 152-172. Uriagereka, J. 1997. An F position in Western Romance. In K. Kiss (ed.) Discourse Configurational Languages. Oxford: Oxford University Press. Uriagereka, J. 1998. Rhyme and Reason. Cambridge, Mass.: MIT Press. Uriagereka, J. 1999. Multiple Spell-out. In S. D. Epstein, and N. Hornstein (eds.) Working Minimalism. Cambridge: MIT Press. Valois, D. 1991. The Internal Syntax of DP. Doctoral dissertation. UCLA, Los Angeles, California. Vainikka, A. 1989. Deriving Syntactic Representations in Finnish. Mass.: GLSA, University of Massachusetts, Amherst. Vanikka, A. and Y. Levi. 1999. Empty Subjects in Finnish and Hebrew. Natural Language and Linguistic Theory, 17: 613-671. 388 Vilkuna, M. 1989. Free Word Order in Finnish: Its Syntax and Discourse Functions. Helsink: Suomalaisen Kirjallisuuden Seura. -- Viotti, E. 1999. A Sintaxe das Senten?as Existenciais do Portugu?s do Brasil. Doctoral dissertation, Universidade de S?o Paulo. Vainikka, A. and I. Levy. 1999. Empty Subjects in Finnish and Hebrew. Natural Languages and Linguistic Theory, 17: 613-71. Vergnaud, J. R. 1974. French Relative Clauses. Doctoral Dissertation, MIT. Vergnaud, J. R. and Zubizarreta. M. L. 1992. The Definite Determiner and Inalienable Construction in French and English. Linguistic Inquiry, 23: 595-652. Zamparelli, R. 1995. Layers in the Determiner Phrase. Doctoral dissertation, University of Rochester. Zoerner, E. and B. Agbayani. 2001. Moving Away from Deletion. Paper presented at the Annual Meeting of the Linguistic Society of America. Washington, DC. Zwart, J-W. 1988. The First Case. The Nominative as a Default Case and Consequences for the Control Theory. MA Thesis, University of Groningen. Zubizarreta, M. L. 1982. On the Relationship of the Lexicon to Syntax. Doctoral dissertation, MIT.