Word Order in Brazilian Portuguese [Reprint 2012 ed.] 9783110869552, 9783110171938

Citation preview

Word Order in Brazilian Portuguese

W G DE

Studies in Generative Grammar 57

Editors

Harry van der Hulst Jan Köster Henk van Riemsdijk

Mouton de Gruyter Berlin · New York

Word Order in Brazilian Portuguese

by

Gláucia V. Silva

Mouton de Gruyter Berlin • New York

2001

Mouton de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter G m b H & Co. KG, Berlin.

The series Studies in Generative Grammar was formerly published by Foris Publications Holland.

© Printed on acid-free paper which falls within the guidelines of the ANSI to ensure permanence and durability.

Library of Congress Cataloging-in-Publication

Data

Silva, Gláucia V. (Gláucia Valeria), 1963Word order in Brazilian Portuguese / by Gláucia V. Silva. p. cm. - (Studies in generative grammar ; 57) Includes bibliographical references and index. ISBN 3 11 017193 7 (alk. paper) I. Portuguese language - Brazil - Word order. I. Title. II. Series. PC5443 .S55 2001 469.7'98-dc21

Die Deutsche Bibliothek - Cataloging-in-Publication

2001026719

Data

Silva, Gláucia V.: Word order in Brazilian Portuguese / by Gláucia V. Silva. - Berlin ; New York : Mouton de Gruyter, 2001 (Studies in generative grammar ; 57) ISBN 3-11-017193-7

© Copyright 2001 by Walter de Gruyter G m b H & Co. KG, D-10785 Berlin. All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Printing & Binding: Hubert & Co, Göttingen. Cover design: Christopher Schneider, Berlin. Printed in Germany.

Contents

1

Introduction 1 Word order in Brazilian Portuguese 2 Overview of the Minimalist Program 3 Analyses before Minimalism 3.1 Declaratives 3.2 PFA-questions 3.2.1 Analyses assuming V-to-C movement 3.2.2 Analyses assuming no V-to-C 3.2.3 Implications for Brazilian Portuguese 4 Minimalist accounts 5 Structure of the study

1 1 5 7 7 9 10 11 15 19 26

2

Basic order in declaratives 1 Basic word order 2 Adverb position 2.1 Overt verb movement in French 2.2 Adverb position in a structure without AgrP 3 Sentential adverbs 4 Verb position and other adverbs in Brazilian Portuguese 4.1 Adverbs and compound verbs 4.2 The position ofjá 5 Conclusion

29 29 31 36 47 51 60 64 68 72

3

Postverbal subjects in Brazilian Portuguese 1 Introduction 2 The Unaccusative Hypothesis 2.1 Unaccusatives in Spanish Summary 3 Unaccusatives in Brazilian Portuguese Summary 4 Verb-subject order in Brazilian Portuguese 4.1 Locative inversion

77 77 78 80 83 84 91 92 96

vi

Contents

Summary 4.1.1 The locative PP and pro in BP 4.1.2 Overt Case checking by the subject Summary 4.2 VS and unaccusatives 4.2.1 Checking of accusative Case 4.2.2 Postverbal indefinite subjects: feature checking 4.2.3 Lack of VOS and VSO 4.2.4 Postverbal definite subjects 4.3 VS with other verbs Conclusion

117 127 130 136 139

Verb movement in interrogatives 1 Introduction 2 Interrogatives in Romance: paradigms 2.1 Diagnostics for w/z-movement 3 The PF/z-Criterion 4 Accounts for Spanish: no V-to-C Summary 5 Questions raised by V-to-C Verb position in interrogatives 5.1 Summary 5.2 Subject position in interrogatives Summary 6 Embedded interrogatives 7 Previous studies of Brazilian Portuguese interrogatives 8 Conclusion

143 143 144 148 150 154 159 160 165 172 174 181 182 185 187

5 4

102 103 105 108 108 115

Notes

191

References

199

Index

207

Chapter 1 Introduction

This work investigates the restrictions on clausal word order in the Brazilian variant of the Portuguese language. The goal of this study is to provide an analysis of word order in Brazilian Portuguese under the Minimalist Program framework. This chapter introduces the Minimalist Program (Chomsky 1993, 1995b), the framework that guides this study, and reviews a few analyses of word order in Spanish and in Portuguese. The chapter begins with the parameters for word order found in Brazilian Portuguese, showing the motivation for a study such as the one presented here. Then, an overview of the Minimalist Program is provided, in order to lay the foundations for the study. Following this overview, some analyses of word order are reviewed, starting with those put forth before Minimalism, and including analyses of declaratives and of w/z-questions. The implications of those studies for Brazilian Portuguese are discussed before introducing a few analyses provided within the Minimalist framework. We will then see that those accounts also do not explain satisfactorily the phenomena found in Brazilian Portuguese, a fact that motivates the study in this book.

1. Word order in Brazilian Portuguese Brazilian Portuguese (BP) exhibits some unexpected facts regarding word order. It does not allow for a postverbal1 subject in interrogatives, contrary to what one might expect, given the facts in other closely related languages. This is interesting from a theoretical point of view, since most analyses of other Romance languages assume an obligatory 'subject-verb inversion' in questions (cf. Rizzi 1991, Zubizarreta 1992 among others). In Brazilian Portuguese, such inversion yields ungrammaticality, and we find SVO to be the order used in interrogatives as well as in declaratives. In other related languages (Spanish, Italian, and even European Portuguese) there are other or-

2

Introduction

ders possible in affirmative sentences - i.e., in these languages a postverbal subject may also be found in declaratives. Within the framework of the Minimalist Program (Chomsky 1993), it is very important to account for word order facts: since movement before Spellout is reflected in the output, this "unexpected" order in Brazilian Portuguese happens because there is more or less overt movement required in this language (compared to the other Romance languages). For the most part, declaratives in Brazilian Portuguese exhibit SVO order. This is the default order for sentences involving transitive and unergative2 verbs: (1)

a. A Ana comprou muita coisa nesta loja. 'Ana bought much stuff at this store' b. * Comprou a Ana muita coisa nesta loja. Bought Ana much stuff at this store c. * Comprou muita coisa a Ana. Bought much stuff Ana

(2)

a.

O Ivo trabalha todo dia. 'Ivo works every day' b. * Trabalha todo dia o Ivo. Works every day Ivo c. * Todo dia trabalha o Ivo. Every day works Ivo

As we can see in (1) and (2), declaratives involving transitives or unergatives show the order subject-verb. As a rule, a postverbal subject yields ungrammaticality. This is not true, however, of sentences that contain unaccusative verbs, where the order verb-subject is also possible: (3)

a. A Maria chegou. 'Maria arrived' b. Chegou a Maria. Arrived Maria

Word order in Brazilian Portuguese

3

Sentences with ser/estar 'to be' may also contain VS in declaratives: (4)

a.

Os meninos säo impossíveis. 'The boys are impossible' b. Sao impossíveis os meninos.3 Are impossible the boys

In interrogatives, the order found with transitives and unergatives is again SV: (5)

a.

O que o Paulo comprou? What Paulo bought? 'What did Paulo buy?' b. *0 que comprou o Paulo? What bought Paulo?

(6)

a.

Quando a Bia trabalha? When Bia works? 'When does Bia work?' b. * Quando trabalha a Bia? When works Bia

The examples in (5) and (6) contrast with their counterpart in most dialects of Spanish, one closely related language, which allows only postverbal subjects in most w/z-questions: (7)

a. * ¿Qué Pablo compró? What Pablo bought? b. ¿Qué compró Pablo? What bought Pablo? 'What did Pablo buy?' c. * ¿Cuándo Ana trabaja? When Ana works? d. ¿ Cuándo trabaja Ana? When works Ana? 'When does Ana work?'

4

Introduction

In Brazilian Portuguese, a different order with a transitive verb may yield a different meaning, whereas in other languages we may find ambiguity: the Spanish example (8a) and the Italian example (8b) are ambiguous, while Brazilian Portuguese (8c) and (d) are not: (8)

a.

¿Qué contiene la caja? (Spanish) 'What does the box contain?'/'What contains the box?' b. Chi ama Maria? (Italian) 'Who loves Maria?'/'Who does Maria love?' c. Quem ama Maria? (Brazilian Portuguese) 'Who loves Maria?'/*Who does M. love? d. Quem Maria ama? (Brazilian Portuguese) 'Who does Maria love?'

Although SV is the obligatory order with transitives and unergatives (with the exception noted in endnote 2), VS as well as SY is found in w/j-questions with unaccusatives4, both with inanimate and with animate subjects: (9)

a.

a'. b.

b'. c. c'.

Por onde passa o ônibus? By where passes the bus 'Where does the bus go by?' Por onde o ônibus passa? By where the bus passes Quando sai o jornal? When comes out the newspaper 'When does the newspaper come out?' Quando o jornal sai? When the newspaper comes out A que horas sai a Bia? At what time leaves Bia A que horas a Bia sai? At what time Bia leaves 'What time does Bia leave?'

Overview of the Minimalist Program

5

As we can see from the examples in (9), with unaccusative verbs VS yields grammatical results, just as SV does. That does not happen with transitives or unergatives, as was shown before. These preliminary data already show that word order in Brazilian Portuguese is an intriguing topic: for the most part, BP does not accept subject-verb inversion, not even in interrogatives, where inversion is found in so many languages (including European Portuguese, though not obligatorily). Nevertheless, inversion is perfectly acceptable with unaccusatives and the copular verbs ser/estar 'to be', both in interrogatives and in affirmatives. Therefore, we ought to investigate why it is that clauses - both declarative and interrogative - in Brazilian Portuguese with transitives and unergatives cannot exhibit inversion, and also why unaccusatives may yield inversion in both kinds of clauses. An account of the facts in Brazilian Portuguese raises important questions for recent theoretical accounts of 'inversion' and word order in general. The next section sketches an overview of the framework under which the analyses here are carried out.

2. Overview of the Minimalist Program According to the Principles and Parameters approach to linguistic theory, the variations found among languages can be derived from a highly constrained set of parameters which interact with universal principles of language. The Minimalist Program is a particular variant of the Principles and Parameters framework, based on principles of economy. Within this framework, developed in Chomsky (1991, 1993, 1995b), invariant principles determine what counts as a possible derivation; a derivation converges at one of the interfaces (PF and LF) if it yields a representation satisfying FI (Full Interpretation) at this level; otherwise it crashes. Chomsky argues that multiple levels of representation can raise empirical problems; thus it is best to dispense with them in a representation of the language system. There are, for example, instances of expressions interpretable at LF that are not interpretable in their D-structure positions, which make special assumptions for postulating a D-structure lose

6

Introduction

credibility. Chomsky (1993) dispenses with this level of representation and relies on Generalized Transformations for lexical access. Without D-S, the computational system selects an item from the lexicon and projects it to an X' structure (but cf. Chomsky 1995a, 1995b (Chapter 4), where X' Theory is eliminated as a template on phrase structures). S-structure is also dispensed with, since it interferes with optimality and increases complications in the derivation (a linguistic expression is an optimal realization of interface conditions). In Minimalism, a lexical item enters the derivation already equipped with all the features necessary. Taking a verb as an example, it is assumed to have inflectional features in the lexicon as an intrinsic property. These features are checked against Inflection). If the features of I and the verb match, the strong features of I disappear and the verb enters the PF component under Spell-out: the derivation converges. If those features do not match, the derivation crashes at PF, because the strong features cannot be eliminated and are uninterpretable at PF. This checking procedure may a priori take place at any point in the derivation, before or after Spell-out, depending on feature strength (so, feature checking is important for reasons other than preventing crashing at PF). Following Chomsky (1993), the functional elements Tense and Agr both have nominal and verb features (D/N- and V-features). (Chomsky 1995b revises this system by eliminating Agr as a functional head. I will treat this point in more detail in the next chapter.) These features may be parameterized with either a strong or a weak value. Strong features are required to be checked in the derivation by Spell-out (i.e., in the overt syntax), whereas weak features are checked after Spell-out. The morphological features of Tense and Agr have two functions: first, to check properties of the DP that raises to their specifier position; and second, to assure that DP and V are properly paired. The interaction of these features with principles such as Procrastinate (which states that if movement is not required to be overt, then it will be covert) will determine whether certain steps of the derivation occur before Spell-out (overtly) or at LF (covertly). The D-features correlate with the Spec positions, ruling DP movement, and the V-features with the functional heads, ruling

Analyses before Minimalism

7

head movement. The Case Filter is recast as an interface condition, the condition that all morphological features must be checked somewhere, for convergence. French and English, for example, have strong D- features of Tense, which means that in these languages a DP argument must raise overtly to check its features. French also has strong V-features of Tense. Thus, Chomsky (1991) argues that V in French moves overtly to T, and the V+T complex head then raises to AgrS. In English, however, the V-features of Agr are weak and are not checked overtly. Following Pollock (1989) and Chomsky (1991, 1993) this difference between strong and weak features accounts for observed contrasts in word order between the two languages. The only relevant difference between them, therefore, is the specification for the V-features of Agr. Thus, within Minimalism, word order has to be accounted for in terms of the distinction between strong/weak features, which correlates with whether movement happens overtly or covertly. In the next section we will discuss a few accounts of word order undertaken within the Government and Binding theory framework, prior to Minimalism. We will also briefly introduce analyses based on the Minimalism framework. Here, our goal is to verify what insights can be derived from these analyses independently of the particular theoretical assumptions on which they are based.

3. Analyses before Minimalism 3.1. Declaratives

As we have seen, word order is predominantly SV in Brazilian Portuguese. In a work that has been taken as departure for many others, Koopman and Sportiche (1991) propose that subjects do not originate in [Spec, IP], but rather move to this position for Case reasons. At DStructure, subjects are found adjoined to Vmax (alternatively, in [Spec, VP], as in Contreras 1991). This idea is quite well motivated in their work, and here it is assumed to be correct. Thus, a simple declarative such as (10) would have (11) as its D-Structure (in pre-Minimalism terms):

8

(10)

Introduction

Joäo viu o presidente. 'Joäo saw the president' IP

(11) NPA

I' ymax

I NP* Joäo

VP V

NP

ver

o presidente

NPA receives Structural Case from I. This is an instance of Case assignment by agreement, where a head assigns Case to an NP in its specifier position (or, in Minimalism terms, Case is checked in the Spec position). Given the Brazilian Portuguese example above, it would appear that movement to NPA is necessary in Brazilian Portuguese. Unlike in Spanish (as [13]), postverbal subjects are not found in matrix declaratives: (12)

a. Paulo deu urn presente ao amigo dele. Paulo gave a gift to his friend b. *Deu Paulo um presente ao amigo dele. Gave Paulo a gift to his friend

(13)

a. Pablo le regalò el violin a su amigo. Pablo him-gave the violin to his friend b. Le regalò Pablo el violin a su amigo. To-him Pablo gave the violin to his friend

Consider how Koopman and Sportiche (1991) would deal with Spanish. NPA in this language is not necessary, because Infi seems to be able to govern NP*5. This is explicitly assumed by Contreras

Analyses before Minimalism

9

(1991) to account for VSO order in Spanish. He further argues that a preverbal subject in Spanish is adjoined to IP, a defective category in Spanish that lacks an intermediate projection. Thus, SVO order arises from adjunction rather than movement to [Spec, IP]. Unlike Spanish, preverbal subjects in Brazilian Portuguese are not only an option; they are not adjoined, but rather move to [Spec, IP] for Case checking reasons. Thus, Contreras's account would analyze Brazilian Portuguese as English, in which IP is not a defective category. Furthermore, his account relies on the notion of government, which is not available in the Minimalism framework. For Italian VOS, under the Principles and Parameter framework it was generally assumed that there the subject moved from a preverbal position to a right-adjoined position (as in Rizzi 1982). Lack of VSO in Italian was accounted for with the impossibility for the subject to be governed and Case-marked in its base position. Again, the Principles and Parameters accounts deal with ideas that are not part of the Minimalism framework, and thus would need to be reformulated. For example, the Minimalist Program does not include the notion of government, although it does keep Case, even if now it is no longer "assigned", but rather "checked". However, we can derive the insight that Case needs drive movement to NP A . Later it will be seen that these Case needs have different realizations in Brazilian Portuguese vs. Spanish.

3.2. Wh-questions In questions, as observed above, preverbal subjects are also the rule for transitive and intransitive verbs in Brazilian Portuguese. Some languages, however, do not accept this type of order in interrogatives. The accounts of word order put forth before Minimalism, as we see, rested crucially on where subjects can be governed and Casemarked. Furthermore, those analyses generally did not take into account movement of the verb (although the general assumption is that the verb moves to Infi by S-S in Romance, as opposed to English). Here I outline a few of those analyses.

10

Introduction

3.2.1. Analyses assuming V-to-C movement Rizzi (1991) accounts for inversion in w/z-questions in English and Italian.by proposing the ^-Criterion: (14)

The ^-Criterion a. A ffTz-operator must be in a Spec-head configuration with o 311 X

[+WH]·

o b.

An X [+WH] must be in a Spec-head configuration with a FF/z-operator.

According to Rizzi, English and Italian exhibit I to C movement: the verb, either the main verb (in Italian) or an auxiliary (in English), first moves to Infi, and then to C in questions, in order to satisfy the WhCriterion. He claims that this criterion is responsible for the SStructure distribution and the LF interpretation of w/z-operators. The PF/z-Criterion, according to him, expresses the fact that interrogative operators must be in the spec of CPs which are interpreted as questions, and, conversely, CPs interpreted as questions must have interrogative operators as specifiers. Rizzi argues that the criterion is met in English and Italian w/z-questions because Infi carrying [+Wh] moves to C, and the wA-operator moves to its Spec: (15)

[ c p[ c .what[ c0 has[+WH] [Mary t said t]]]]

This same account is used to explain the data in Italian: Wh is licensed in main clauses under I, the inflected verb moves to C and the PF/z-Criterion is met. This is supported by hypothetical clauses in Italian, where I to C movement is also found (parallel to English): the hypothetical complementizer se 'if can be dropped only with a postverbal or null subject, as in (16) from Rizzi (1991): (16)

a. *(Se) Gianni fosse arrivato, tutti sarebbero stati contenti. 'If Gianni had arrived, everybody would have been happy' b. (Se) fosse arrivato Gianni,...

Analyses before Minimalism

11

c. (Se) fosse arrivato in tempo, Gianni sarebbe stato contento. (If) had arrived in time, G. would have been happy. In Rizzi's analysis, se can be replaced with the inflected verb under I to C movement. The non-application of this movement requires se to be in the sentence, its absence yielding ungrammaticality as in (16a). Rizzi's approach is explicitly adopted for Spanish wA-structures by Zubizarreta (1992) with little further argument. Mallén (1994) also proposes that V moves to C in Spanish wh-structures, triggered by a mechanism similar to that in V2 languages, i.e. the necessity for a constituent to comply with a Case requirement, which in Spanish is expressed as a requirement that a variable or the head of the movement chain containing a variable must be properly governed by a Case assigner6. Nevertheless, evidence for V-to-C is not provided in his work.

3.2.2. Analyses assuming no V-to-C Contra the analyses arguing for V to C movement, other analyses take as their starting point the idea that 'inversion' is due to the subject not moving to [Spec, IP]. Sufier (1994) argues that V in Spanish does not move to C, and the empirical validity of her argument needs to be emphasized. In her analysis, she uses the position of phrasal adverbs as tests: if V moved to C in questions, these adverbials would necessarily follow V when a wh-phrase is in [Spec, CP]. However, this is not the case in either Spanish (17a, b) or Brazilian Portuguese (17c, d) (in BP, the very position of the subject would indicate lack of Vto-C): (17)

a. ¿A quién jamás ofenderías tú con tus acciones? Whom never would-you offend with your actions b. ¿Qué idioma todavía estudia Pepita en su tiempo libre? What language still studies Pepita in her free time (Suñer 1994: [21a] and [21c])

12

Introduction

c.

Que idioma o Ivo ainda estuda no seu tempo livre? What language Ivo still studies in his free time d. */?Que idioma o Ivo estuda ainda no seu tempo livre? What language Ivo studies still in his free time As Sufier points out, in English either Aux or do moves to C, leaving the adverb behind: (18)

a. Which student do you still tutor? b. * Which student still do you tutor?

Another piece of evidence for the claim that V does not move to C in Spanish is given by sentential negation when combined with adverbials like those in (17). When negation occurs together with VPexternal adverbs, the adverbials and the negation word no (or näo in Portuguese, [19c]) should follow the verb if it were in C. Again, (19) shows the opposite: (19)

a ¿Qué aún no le dio Mafalda a su mamá? 'What has Mafalda not given her mother yet?' b. ¿A quién ya casi no le escribes tú carta? 'To whom don't you write letters hardly anymore?' (Sufier 1994: [25a] and [25b]) c. A quern voce já quase näo escreve? 'To whom don't you hardly write anymore?'

The sentential negator and the adverbials are placed before the verb, showing that V is not in C, but rather within IP. Therefore, like Spanish (assuming Sufier to be correct), Brazilian Portuguese does not have movement of V into C. In the case of verb movement, then, Brazilian Portuguese parallels Spanish, a parallelism which is not found with the position of the subject. As we have seen, SV is the preferred order in Brazilian Portuguese. For Spanish, there are several analyses that propose an account for the obligatory postverbal subject in w/z-questions. Ideally, an account of postverbal subjects in Spanish should also be able to account for the lack thereof in Brazilian Portuguese. Sufier (1994) proposes a

Analyses before Minimalism

13

condition to account for obligatory inversion, the so-called Argumentai Agreement Licensing (AAL). Suñer postulates that whphrases have the feature [+Argumentai], which means that by SpecHead agreement, C° also has the same feature. Then, the head and its Spec have to match in being [+wh, +Arg]. However, Suñer continues, this selected feature seems to need licensing by the predicate in I (or T). She considers selection to involve feature matching or checking between the verb and its argumentai elements. Thus, her Argumentai Agreement Licensing requires that [Spec, CP] and C agree with respect to the feature [+ Argumentai]. As a consequence, a subject cannot raise to [Spec, IP] when an argumentai w/z-phrase is in [Spec, CP], and subject-verb 'inversion' obtains. Since the subject in Brazilian Portuguese does move to [Spec, IP] even in wA-questions, a condition such as the AAL would be particular to Spanish. Nevertheless, there are varieties of Spanish which this condition does not account for, as is the case with Puerto Rican Spanish, which allows for preverbal subjects in w/z-questions. Suñer suggests that the AAL is simply not part of the grammar of Puerto Rican Spanish. A different analysis is given by Goodall (1993), who argues that in Wi-clauses in Spanish, the wA-phrase must move into [Spec, IP] before moving into [Spec, CP], preventing the subject NP from landing in [Spec, IP]. This would mean that [Spec, IP] is a potential A'-position in Spanish. Koopman and Sportiche (1991) suggest that for English the status of NPA depends on how to account for the distribution of do. In Spanish, where the subject may remain in situ, NP A can plausibly be an A'-position. Finally, Contreras's (1991) explanation of Spanish word order in questions rests crucially on his idea that a preverbal subject in Spanish is adjoined to IP. This is seen in (20), the structure for Maria compró un libro:

14

Introduction

(20)

IP NPI Maria\

IP I

VP VP

V comprar

NP 2

NP ti un libro

(20) could be generated by movement or by base-adjunction7. If generated by movement, NP 2 dominates a variable, and NP, is licensed at S-structure by binding the variable. This derivation would then allow a preverbal subject in a w/z-question. Nevertheless, Contreras proposes that a closed domain cannot contain any unlicensed elements (Closed Domain Condition, CDC). (21)

*¿ Quéi Maríaj compró ti tj ?

The A'-chain consisting of qué and its trace constitutes a closed domain, since it contains all that is necessary for its interpretation. Violation of the CDC thus yields the ungrammatically in (21). A baseadjoined preverbal subject is equally ungrammatical in a whmovement structure, since preverbal subjects, by his analysis, are not licensed until LF. Because w/z-movement creates a chain at Sstructure, the CDC must be met at that level, hence the ungrammaticality. The analyses put forth by Sufier, Goodall and Contreras may give us a few insights regarding the impossibility of SV in Spanish wÄ-questions. The essential idea in Suñer's account is related to agreement between [Spec, CP] and C. The fact that feature mismatch yields ungrammaticality is in tune with the basic idea in the Minimalist Program. This approach (feature mismatch) is taken up again in Ordóñez (1997) to account for the same phenomenon, but now under a more current framework (I will return to Ordóñez's analysis in

Analyses before Minimalism

15

Chapter 4). In Contreras's work we find the idea that preverbal subjects in Spanish occupy a more peripheral position, a concept that has been argued for in the work of several other researchers, and for languages other than Spanish as well (for Spanish, this idea is present in Zubizarreta 1994; Ordóñez 1997). If updated to the Minimalism framework, Goodall's account may be related to (possibly EPP) feature checking by the w/i-word in [Spec, TP], which would prevent the subject from moving to that position overtly. The insight that feature checking prevents further overt movement of the subject is explored in more detail in Chapter 3 of this study, to account for postverbal subjects in the locative inversion construction and for inversion with unaccusatives. Therefore, we see that all three analyses contain ideas that can be helpful if refined and updated.

3.2.3. Implications for Brazilian Portuguese These previous analyses of word order present shortcomings regarding Brazilian Portuguese, as we shall see. In justifying the WhCriterion, Rizzi (1991) takes some examples from French: he observes that along with having Subject-Aux inversion, French also has SV order, which, as we have seen, is the usual order for Brazilian Portuguese (examples from Rizzi 1991): (22)

a. Elle a rencontré qui? She has met who? b. Quii elle a rencontré t¡ ? Who she has met? c. Quii üj-t-elle tj rencontré t¡ ? Who has she met?

(22c) looks like its English counterpart, and is analyzed in the same way, i.e., the object and the inflected verb move in order to meet the #7i-Criterion. According to Rizzi, the FfTi-Criterion is normally met by means of static agreement: I contains the Wh feature, and the Criterion is satisfied at S-structure and at LF, given that I moves to C. This leaves (22a) and (b) to be explained. Rizzi's account for these

16

Introduction

sentences rests on the notion of dynamic agreement: a w/z-operator is able to confer the Wh feature to a clausal head under agreement. He formalizes this as in (23): (23)

Wh-Op y => Wh-Op y [ + W H ]

In (23), the specifier (i.e., [Spec, CP]) can pass the relevant feature specification to the head C. At D-structure in (22b), no clausal head contains the feature [+WH], JfTz-movement applies in the syntax, and C can get the feature through dynamic agreement. Thus, besides static agreement, French is also endowed with dynamic agreement, which applies freely in the syntax or at LF. Now, the Brazilian Portuguese counterparts to (22a)8 and (b) are also well-formed, whereas the sentence corresponding to (22c) would not be found: (24)

a.

Eia tinha encontrado quem? She had met who? b. Quem eia tinha encontrado? Who she had met? c. * Quern tinha eia encontrado? Who had she met?

Brazilian Portuguese patterns with French in having in situ whelements and preverbal subjects in questions. Furthermore, as in French and in English, in subcategorized questions only the structure resulting from the simple movement of the w/z-element to [Spec, CP] is well-formed. This would lead us to reject a simpler account of (22) and (24) that might state that the ^-Criterion does not apply at S-S (or before Spell-out) - in other words, it does not drive overt movement: (25)

a. * Eu näo sei [(que) [eia encontrou quern]] I don't know she met who b. Eu näo sei [quem¡ [eia encontrou t j ] I don't know who she met c. *Eu näo sei [quem¡ tinhaj [eia tj encontrado tj] I don't know who had she met

Analyses before Minimalism

17

Given (25), we need to assume that the ^-Criterion is met at Sstructure (in pre-Minimalism terms). Since the arguments for dynamic agreement in French can be mapped onto Brazilian Portuguese, the lack of inversion in w/2-questions could be accounted for by means of this kind of agreement proposed by Rizzi. Unlike French (and English, Italian, Spanish...), however, Brazilian Portuguese does not allow postverbal subjects in (most) questions. This suggests that static agreement is not an option in this language, and that the proposed ^-Criterion is met only by means of dynamic agreement. The claim that the FP7z-Criterion may not be met through its default form is strong. Rizzi regards dynamic agreement as a special extra option that a language such as French has. Thus, we would not expect a language to prohibit the use of static agreement, as would seem to be the case in Brazilian Portuguese, where dynamic agreement could not be only an option, but in fact the only mechanism possible. We saw above that Rizzi argues that Italian satisfies the WhCriterion via Y to I to C movement, as was assumed for Spanish by Zubizarreta (1992). Nevertheless, we have already seen that, following Suñer, there may in fact not be V to I to C movement in Spanish (I will return to this issue in Chapter 4 and re-examine the facts in Italian). Since it is clear that lack of inversion is the default construction in Brazilian Portuguese, and since a given criterion would most probably not apply only in its marked form, it appears that the Whcriterion should not account for the lack of inversion in BP the same way it happens in French. In section 3.2.2, we see that Suñer argues (convincingly) for the lack of V to C movement in Spanish, and that her analysis carries over to Brazilian Portuguese. Taking that into consideration, we reach the conclusion that the analyses of obligatory inversion in Zubizarreta (1992) and in Mallén (1994) are not empirically supported in Spanish, and since Brazilian Portuguese does not show V-to-C either, such accounts would not be able to handle the facts in this language. We have seen that Suñer (1994) proposes the AAL to account for inversion9. Puerto Rican Spanish, however, exhibits preverbal subjects in questions. In Puerto Rican Spanish, w/i-phrases pattern

18

Introduction

alike regardless of their argumentai or non-argumental status. A lexical subject can always appear between selected wA-phrases and the verb, suggesting that the A A L is not respected. If this dialect has dispensed with the A A L , sentential adverbials and left-dislocated constituents would not yield ungrammaticality if placed between C and V. Suñer's Puerto Rican Spanish data confirm this prediction, and one would expect BP to do the same, since the subject NP moves to [Spec, IP] as in Puerto Rican Spanish, appearing between the whphrase and the verb. However, the data do not translate straightforwardly to Brazilian Portuguese, and it becomes very difficult to show that this language, like Puerto Rican Spanish, does not comply with the AAL 1 0 . Given that inversion yields ungrammaticality for most cases in Brazilian Portuguese, saying that a condition such as the A A L is dispensed with would not be enough, since we would not have accounted for why it is that inversion is not only not obligatory (since it is at least optional in Puerto Rican Spanish with noncomplex subjects), but in fact is prohibited. Recall that for Goodall, [Spec, IP] is a potential A'-position in Spanish. We have seen that in Brazilian Portuguese this position appears to be an A-position to which the subject NP moves for Case reasons. Following this account, the word order difference between Brazilian Portuguese and Spanish would be a parametric one regarding Case assignment (checking) to NP*: while [Spec, IP] is only an A-position in Brazilian Portuguese, it is a potential A'-position in Spanish. Contreras's analysis assumes different structures for Spanish and English. In Spanish, according to Contreras, preverbal subjects are 'adjuncts' in the non-canonical direction, and hence unlicensed until LF. So, Contreras's CDC, as formulated, rules on 'inversion' in Spanish because of the nature of preverbal subjects in that language. For Brazilian Portuguese, the grammaticality of preverbal subjects would actually be expected. However, English root questions would pose a problem for his analysis, since his CDC would disallow VS. As mentioned before, the accounts of word order in Spanish reviewed here may each contain interesting insights, especially if we can update them to the current framework. However, an account of word order in Brazilian Portuguese will need more than an update,

Minimalist accounts

19

since those analyses reviewed cannot, as has been shown, apply to BP. Having looked at analyses of word order put forth before Minimalism, we can now move to those accounts made available in the present framework.

4. Minimalist accounts Recall that within Minimalism, word order has to be accounted for in terms of strong/weak features, corresponding to overt vs. covert movement. In order to account for the various surface word orders in French, Drijkoningen (1994) uses this notion of morphological strength. French does confirm Kayne's (1994) hypothesis that S(pecifier)-H(ead)-C(omplement) is the universal basic order, but CH and H-S are also encountered. H-S occurs with Stylistic Inversion: (26)

Quand a dormi Jean? 'When did Jean sleep?'

The order in Stylistic Inversion is restricted to plain DPs, however. Use of a subject clitic yields ungrammaticality: (27)

* Quand a dormi-t-il? When has slept he?

One of the major differences between plain DPs and subject clitics is a difference in morphologically visible Case. Subject clitics are lexically marked for nominative Case. If morphologically visible Case is the trigger for obligatory overt movement, [Spec, AgrSP] will have to be reached overtly. As the H-S order with Stylistic Inversion only obtains in w/z-questions, subjects do not need to move to AgrSP when there is a w/z-element in CP. Drijkoningen then proposes that dynamic agreement be interpreted as a principle that allows strength of one functional head to induce weakness of a governed functional head (recall that in Rizzi's original proposal, dynamic agreement would be

20

Introduction

the ability of a w/z-operator to confer the w/z-feature to a clausal head under agreement). H-S in French also occurs with Clitic Inversion, which excludes plain DPs: (28)

a.

Quand a-t-il dormi? When has he slept? 'When did he sleep?' b. * Quand a Jean dormi? When has Jean slept?

Drijkoningen's interpretation of dynamic agreement implies that (28b) is an instance of the principle Procrastinate, since the plain DP does not have to move overtly because the strength of AgrS is suspended. The subject clitic, on the other hand, has to move because of its morphologically realized nominative nature. He concludes that the subject in French declaratives must move because AgrS is strong, yielding the order S-H-C. If the subject of a w/z-question is a plain DP, dynamic agreement induces weak AgrS, and the subject, when not a clitic pronoun, remains in its base position. If the subject is a clitic, it moves to AgrS in order to morphologically check lexical Case, yielding Clitic Inversion as in (28a). Drijkoningen's interpretation of dynamic agreement does not apply to the examples in Brazilian Portuguese, since in this language we do not generally find inversion in interrogatives, except for the instances mentioned in section 1 (note that the facts do not change when the subject is a pronoun). The basic SVO order in Brazilian Portuguese indicates that the Ν or D feature of the Spec position to which the subject moves is strong11, and it is not made weak in any example of interrogatives, for we find subject-verb inversion in BP to be related to the class of verbs used, rather than to whether the sentence is an affirmative or an interrogative. Ordófiez (1997) puts forth a different account of the various word orders in Spanish, Catalan and Italian. For VOS in declaratives, he states that the subject is focused and occupies its base position, while the object moves to the Spec of a higher inflectional projection. Ordóñez departs from previous analyses of VOS which argued that

Minimalist accounts

21

the subject was right-adjoined to the VP (also the usual assumption for Italian VOS under the Principles and Parameters framework). The difference between the two hypotheses is in the hierarchical representation of VOS: in the right-adjunction hypothesis, the subject adjoined to the VP c-commands the object in its base position, while the proposal in Ordóñez's light predicate raising hypothesis is that the object, having moved to a projection above VP (but below TP), ccommands the subject in its base position. Ordófiez points out that his account of VOS encompasses a hierarchical difference between VSO (where the subject c-commands the object) and VOS in Spanish which was not available in the right-adjunction hypothesis, and which conforms with Kayne's (1994) antisymmetry proposal. Some examples from Ordófiez (1997) are given below: (29)

a.

Demà comprará un llibre la Magda. Tomorrow will-buy a book Magda b. Ha recensito il libro un professore. Has summarized the book a professor

(Catalan) (Italian)

For the VOS examples of (29) (and for Italian and Catalan generally) Ordófiez claims that subjects do not end up in a postverbal position by simple head movement of the verb to a higher position, as they do in Spanish VSO; rather, he argues that postverbal subjects are focused (with verbs that are not unaccusative), and thus must move overtly to the Spec of a FocP, which is above TP. This move is followed by light predicate raising of the TP to a projection X o above the FocP, and VOS obtains. For Spanish VSO his proposal involves movement of the subject to a neutral phrase (NeutP): Ordófiez claims that in Spanish VSO subjects are neutral, whereas in VOS they are focused. The lack of VSO in Italian and Catalan is accounted for by assuming 19 that these languages lack NeutP . We will examine Ordóñez's proposals in further detail in Chapter 4, when discussing languages other than Brazilian Portuguese. We have already seen that BP does not allow VS in most interrogatives. Naturally, SV is also the order most commonly found in affirmatives. 'Inversion' will cause ungrammatically where it is grammatical in other Romance languages, sometimes depending on

22

Introduction

discourse context and notions such as topic and focus (cf. Zubizarreta 1998): (30)

a. Ha mangiato un dolce il ragazzo. (Italian) Has eaten a dessert the young man b. Sabe la lección María. (Spanish) Knows the lesson Maria. c. Leu a Maria esse livro. (European Portuguese) Read Maria this book.

In (30), we find examples of postverbal subjects in declaratives, which are all ungrammatical in BP, regardless of considerations such as topic and focus: (31)

a. *Comeu um doce o rapaz. Ate a dessert the young man b. *Sabe a liçâo Maria. Knows the lesson Maria c. *Leu a Maria esse livro Read Maria this book

The examples in (31) involve transitive verbs, and above it is shown that SV is the obligatory order with transitive verbs even in questions. The same is true of sentences with unergative verbs (but see note. 2). We have also seen that unaccusative verbs (and ser/estar) may exhibit a postverbal subject in interrogatives and in affirmatives. There is some work within the Minimalist framework on word order in European Portuguese. This dialect differs from Brazilian Portuguese in many aspects, one of which is word order. As mentioned above, European Portuguese allows for inversion in interrogatives as well as in declaratives (if only as an option): (32)

a.

Os alunos lêem muitos livros? The students read many books? b. Lêem os alunos muitos livros? Read the students many books? 'Do the students read many books?'

Minimalist accounts

(33)

23

a. Muita cerveja bebeu o Pedro ontem à noite. Much beer drank Pedro last night b. So esse livro leu a Maria. Only this book read Maria

Raposo (1994) proposes that the sentences in (33) have an element in FP (Functional Phrase)—which does not mean, however, that inversion must necessarily take place. He draws a comparison between EP and European Spanish with regard to affective operators. He claims that VS order in EP is TP-internal, occurring both in root and subordinate clauses. A basic assumption for this claim is that we find a Functional Phrase which hosts affected elements between CP and TP. This is shown in the structure in (34): (34)

FP Spec

F' TP 13

F [+F]

Spec

T' Τ

AgrSP

Δ So este CD

me

recomendou Maria

As in (34), the head of FP can be an abstract [+F], or it can host morphological material. He takes F to be an operator-related category, endowed with one feature taken from a set of possible operator features such as wh, focus, emphasis. Raposo argues that operator-like elements (which would have at least one of the features just mentioned) can raise to [Spec, FP], where they are checked by [+F], Nevertheless, the resulting word order may show either a preverbal or a postverbal subject (in both cases, the clitic, if present, appears before the verb; postverbal clitics are unmarked in EP). When no affected

24

Introduction

element raises to [Spec, FP], the finite verb raises to F; the result is a presentative sentence: (35)

Telefonou-me um cliente ontem. Telephoned me a client yesterday Ά client called me yesterday'

(EP/unmarked)

If V moves to F when there is no affected element involved, we need to ask what the landing site is for the subject when it is preverbal, as in (36a); Raposo gives it the structure in (36b): (36)

a.

Pedro comeu o pastel. 'Pedro ate the pastry' b. Pedro [p' [F comeu [ρ ν o pastel}].

In (36b), the landing site for the subject can only be [Spec, FP], for that would be the only way of obtaining the word order (assuming the same structure as in Raposo 1994). If the subject has its Case checked in [Spec, AgrSP] when it is postverbal (which obtains because the structure assumed by Raposo has TP as the highest node, above AgrSP), there would be no Case-related reason for the subject to move beyond [Spec, AgrSP]. It could be that if there is no other focused element, the subject is the default focus. For Brazilian Portuguese, Kato and Raposo (1994) conclude that there is no V-movement to F, given generalized proclisis in this language. On the other hand, fronted highlighted Focus is, according to them, evidence for movement of affected elements: (37)

So os CD's a Maria me recomendou. Only the CDs Maria me recommended

They note that there is "residual TP-internal VS order". There is no mention, however, of a possible relation between VS and unaccusative verbs (which can be seen in examples given above). They do give examples of VS with (what seem to be) unaccusatìve verbs, as in (38):

Minimalist accounts

(38)

25

a.

Furou o pneu do carro Went flat the tire of the car b. O pneu do carro fur ou. The tire of the car went flat

It is possible that Kato and Raposo did not associate unaccusatives with VS because they also give examples of VS order with a transitive verb which they list as grammatical in Brazilian Portuguese: (39)

So os CD's me recomendou a Maria. Only the CDs me recommended Maria

For me and for several other native speakers of Brazilian Portuguese, however, (39) is not grammatical, despite the focus intonation on the object. Even with highlighted Focus, there is a strong preference for preverbal subjects14. Thus, there seems to be a link between VS and verb classes, since only with unaccusatives and in the locative inversion construction (see Chapter 3) is VS fully acceptable. As a matter of fact, Tarallo and Kato (1993), who note the loss of pro-drop properties in Brazilian Portuguese, state in a footnote that "unaccusative VS order is resistant to change because the system still retains the zero expletive" (1993: 111). I will address the question of the nature of pro in BP in Chapter 3. One account of the predominant SV order in declaratives and interrogatives in Brazilian Portuguese is attempted by Galves (1996), who correlates loss of inversion and weakening of Agr in BP. She argues that, with a weaker Agr, there is dissociation between the Agr morpheme and the Agr node, causing the Comp position to become "of difficult access" to the verb, since an intermediate node is between the normal position of the verb (T) and Comp. Galves claims that there is evidence for an AgrP that is independent of morphological realization, and that this node would be located above TP. Note, however, that Galves's AgrP exists independent of morphological realization. Galves's work will be discussed in more detail in Chapter 4. We will see that, at first sight, her analysis might appear attractive. However, one drawback of her analysis can be pointed out here: there is evidence, as we have seen, that the verb in

26

Introduction

Spanish does not move as high as C in interrogatives (or in inverted declaratives). But in Spanish, VS is a widespread order, and is obligatory in w/z-questions. Therefore, it appears that lack of V to C movement should not necessarily be associated with lack of inversion in a given language. Here we can conclude the review of some analyses of word order. Having already introduced the Minimalism framework, we can now conclude this chapter with the structure of this study and a brief summary of the content of each chapter.

5. Structure of the study In Chapter 2 we will discuss word order in declaratives in BP, especially the obligatory SVO order with transitives. That discussion will lead us to an analysis of the basic structure in BP sentences. In the process of putting forth that account, we will consider examples from Spanish, Italian and French, looking especially at adverb position in all four languages, a fundamental tool for the analysis. Chapter 3 contains the analyses of different occurrences of VS in Brazilian Portuguese. That chapter will start with a discussion of the Unaccusative Hypothesis and the tests used for diagnosis of unaccusative verbs in different languages (Italian and Spanish). Those tests will be applied to Brazilian Portuguese, in an attempt to determine which verbs are unaccusative in that language, and whether inversion is found with all the verbs which pass the tests for unaccusativity. We will proceed to an account of the locative inversion construction (which is found with both unaccusatives and unergatives), followed by the analysis of VS with unaccusatives (without a locative preposition in sentence-initial position). To reach those analyses, we will also examine pro in BP and how it may differ from pro in Spanish and Italian. Chapter 4 discusses interrogatives in Brazilian Portuguese and other Romance languages. As we have seen, in other languages, we find obligatory VS in w/z-questions, whereas in BP that order yields ungrammaticality with transitives and unergatives, and is optionally available with unaccusatives, just as much as in declaratives. We will provide an account of that ungrammaticality, showing why the order

Structure of the study

27

subject-verb is grammatical in Brazilian Portuguese interrogatives. An account will also be given of obligatory VS order in languages such as Spanish and Italian, which are compared and contrasted with Brazilian Portuguese. For those accounts, we will study the position of the verb and of the subject in these languages, arguing that it is the subject which occupies different positions in Spanish/Italian and Brazilian Portuguese. English will also be discussed, and a comparison will be drawn between Romance and English. We will also discuss previous accounts of interrogatives in Brazilian Portuguese, showing that a synchronic account of the facts was in order.

Chapter 2 Basic order in declaratives

This chapter discusses the position of adverbs in order to determine the position of the subject and the verb in the sentence. The study of Brazilian Portuguese (BP) is done comparatively, showing also structures for declaratives in Spanish, Italian and French. Adverb position is often used as a diagnostics for determining the position of subjects and verbs in a clause. After examining the facts involving adverbs in Brazilian Portuguese, Spanish, Italian and French, we will confirm overt movement of the subject in Italian and French SVO. Movement is also overt in Brazilian Portuguese and in Spanish SVO sentences, though the evidence is not the same as in Italian and French. In this chapter we also find a discussion of AgrP, concluding that this node may not be present universally, though it is posited to exist in French (above TP). The Asp(ect) node is also discussed, and it is argued that the verb in Brazilian Portuguese and in Spanish SVO declaratives may land in either Asp or T, while the (tensed) verb in Italian raises overtly to T. The chapter is divided as follows: first, the basic paradigms for word order in Brazilian Portuguese are reintroduced; then, a discussion of adverb position follows, which helps us specify the landing sites for the subject and the verb in declaratives. Sentential adverbs are also analyzed, leading to the positing of AgrP in French. The discussion of the position of certain adverbs is expanded to other examples, including sentences with compound verbs, and the Portuguese adverb já ('already') is looked at in more detail, before concluding the chapter.

1. Basic word order The basic word order in declaratives in Brazilian Portuguese is SVO. This is the order found in constructions involving transitive and uner-

30

Basic order in declaratives

gative verbs. Recall the basic paradigms with respect to 'inversion' or lack thereof which we saw in Chapter 1 : (1)

(2)

a. A Bia comprou urn livro aquí. 'Bia bought a book here' b. * Comprou a Bia um livro aqui. Bought Bia a book here c. * Comprou urn livro a Bia. Bought a book Bia

(*VSO)

a.

(SVAdv)

O Ivo trabalha todo dia. 'Ivo works every day' b. Todo dia o Ivo trabalha. Every day works Ivo c. * Trabalha todo dia o Ivo. Works every day Ivo d. * Todo dia trabalha o Ivo. Every day works Ivo

(SVO)

(*VOS)

(AdvSV) (VAdvS) (AdvVS)

The same rule does not apply, however, to sentences that contain unaccusative verbs or copular verbs: (3)

(4)

(5)

a. A Ana chegou às 10 horas. 'Ana arrived at 10 o'clock' b. Chegou às 10 horas a Ana. Arrived 10 o'clock Ana c. ΊChegou a Ana às 10 horas. Arrived Ana at 10 o'clock a.

Os meninos säo impossíveis. 'The boys are impossible' b. Sao impossíveis os meninos 15 Are impossible the boys a. A comida está na mesa. 'The food is on the table'

(SVXP) (VXPS) (VSXP) (SVXP) (VXPS)

(SVXP)

Adverb position

b.

Está na mesa a comida. Is on the table the food c. *Está a comida na mesa. Is the food on the table

31

(VXPS) (VSXP)

As can be seen in examples (3), (4) and (5), the order verb-subject is acceptable with certain classes of verbs (noting that the order, in these cases, is generally V-XP-S). I will return to these classes and the orders they yield. At this point, we should notice that (3 c) is only possible with contrastive focus on Ana·, (3c) is interpreted as 'it was Ana, not Joäo, that arrived'. Furthermore, when we find a VP that contains more material than the verb and the subject, it becomes more difficult for the subject to immediately follow the verb. This difficulty is what we see in (5c), where we find the (apriori) ungrammatical order V-SPP. If the sentence presents a final subject, it is acceptable, as seen in (5b) and (3b). The order V-S-PP may be acceptable, but it depends on where focus is placed. A discussion of focus in relation to the position of the subject is provided in Chapter 3. Postverbal subjects in Brazilian Portuguese will be discussed in more detail in section 2. For now, I shall concentrate on the order for sentences with transitives and intransitives, in order to account for the rigidity found in Brazilian Portuguese, as opposed to European Portuguese or Spanish. I will start by looking at the position of adverbs in Portuguese, Spanish, Italian and French.

2. Adverb position In this section, we will discuss the position of adverbs and how they can help determine where the subject and the verb are located. As hinted above, the study of adverbs will confirm overt movement of the subject in SVO in Brazilian Portuguese, Spanish, Italian and French. Let us look at the assumptions regarding adverbs and the insights they may provide. In analyzing the position of adverbs in the four languages mentioned above, I will assume that adverbs adjoin to a projection of the head that they modify (Travis 1988, Bowers 1993). I will also as-

32

Basic order in declaratives

sume that an adverb can adjoin to an X', as postulated in Chomsky (1995a). In her work, Travis argues that adverbs cannot project to a phrasal category; they remain as heads. She suggests that heads that do not project must be licensed differently from maximal projections, and assumes that it is a feature of the licensing head (noun or verb) that licenses the adverb. Some of the head features she suggests are Manner, a verb feature, and Event and Agreement, listed as features of Infi. Travis also notes that some adverbs may change meaning depending on the position in which they appear in the sentence (since many adverbs can surface in more than one position) - in other words, interpretation of the adverb is affected by the feature (of the head) that licenses it. The following examples are from Travis (1988): (6)

a. b. c. d.

The police carelessly will arrest Fred. Fred carelessly will be arrested by the police. The police arrested Fred carelessly. Fred was arrested carelessly by the police.

In (6a, b), the subject-sensitive adverb is licensed by Infi, whereas in (6c, d) the agent-sensitive adverb is licensed by the verb. She assumes that the relevant feature in Infi is Agr(eement), and in the verb it is Manner. Thus, the semantic differences found among the examples in (6) can be accounted for: in (6a) the adverb seems to be predicated of the police, whereas in (6b) it seems to be predicated of Fred. In discussing possibilities for structures that involve adjunction, Chomsky (1995a) rules out movement of an adverbial that is adjoined: given the principle of Greed, adverbs have no morphological properties that require movement. If adjunction is not the result of movement, it must follow from Merge. Nevertheless, base adjunction is barred if XP is "semantically active". Chomsky then argues that adverbials can "be adjoined to such phrases as AgrP or IP, or to any X'". As a matter of fact, not only can adverbs adjoin to an X', but in general they must do so if they are prohibited from adjoining to an XP which is either an argument or a predicate, following Chomsky (1995a). According to him, adjunction results in a two-segment category, but not a term of the structure formed. Adjunction cannot form

Adverb position

33

a new term in the derivation, because this term would have no interpretation at the LF interface. These assumptions do not rule our adjunction to the root, since the root is not a predicate or an argument of anything. So, adjunction cannot form a new term in the derivation. Here, I will adopt aspects of both Travis's (1988) and Chomsky's (1995a) works. I assume with Travis that adverbs adjoin to a category that they modify. Given the assumptions in Chomsky (1995a), adverbs can adjoin to X' (or to the root of a sentence). As a result of these postulations, I will assume that adverbs adjoin to an X' that they modify. Given these assumptions regarding adverbs, we can look at the position they occupy. The adverb 'already' has a somewhat fixed place in the four languages mentioned. The most common sites are the ones in (7): (7)

a.

Eu jä sei portugués. Ί already know Portuguese' b. Yoya sé español. Ί already know Spanish' c. Io so già italiano. I know already Italian d. Je sais déjà français. I know already French

In Portuguese and Spanish, the adverb 'already' appears before the inflected verb, whereas in Italian and French its usual position is after the inflected verb and before the direct object. Though these are the generally preferred positions for 'already', other possibilities exist in Portuguese, Spanish and Italian: (8)

a.

Eu sei portugués jä. I know Portuguese already b. *Eu seijá portugués,16 I know already Portuguese c. Yo sé español ya. I know Spanish already

34

Basic order in declaratives

d. ΊΥο sé ya español}1 I know already Spanish e. Io già so italiano. Ί already know Italian' As we can see, in (8a) and (c) we find S-V-DO-Adv, whereas (8d) shows S-Adv-V-DO (which is the order in (7a) and (b), the "default" order for Brazilian Portuguese and Spanish). (8b) and (d) show the order S-V-Adv-DO, which do not seem, at least with this adverb, to be (fully) acceptable. In Portuguese there seems to be no real difference in focus between (7a) and (8a), although (8a) is far less common. In Italian, the use of a preverbal già does imply a difference in focus: while postverbal già is perceived as 'neutral', in preverbal position this adverb acquires a focused reading. In Spanish, (8c) has a more focused reading than (7b). Of these four languages, French is the only one that does not allow any variation regarding the position of 'already'. This lack of flexibility in relation to position of adverbs is much more noticeable in French than in any of the other three languages, as we will see. I will analyze the more common occurrences of 'already', those in (7). Verb movement in French has been much discussed (Pollock 1989; Chomsky 1991, 1993; Koizumi 1993; Drijkonigen 1994). Pollock has argued that differences between French and English can be accounted for by different extensions of verb movement. He thus accounts for the contrasts in negative sentences such as those of (9): (9)

a. Jean n'aime pas Marie. 'Jean doesn't love Marie' b. *Jean ne pas aime Marie. c. *John likes not Mary.

The contrast in (9) shows that the verb moves to the inflectional head in French, but not in English. That the French verb moves to the inflectional head can also be shown by differences regarding adverb position in the two languages:

Adverb position

(10)

35

a. *John kisses ¡ often t¡ Mary. b. Jean embrasse,· souvent t¡ Marie. c. John often kisses Mary. d. *Jean souvent embrasse Marie.

If adverbs in French and English cannot move to the right, often in (10a) can only end up between the verb and the object if the verb moves to the inflectional head. Since that movement does not happen in English, (10a) is ungrammatical, and only (10c) is acceptable (without verb movement). The French verb does move to the inflectional head, yielding (10b). Pollock argues that the ungrammaticality of (lOd) is due to the obligatoriness of such movement in French, assuming, as a first pass, that French and English share the D-structure in (11), where (Adv) is an optional adverbial position that can be occupied by VP adverbs like often/souvent (Pollock assumes that the adverb originates in VP): (11)

[ipNP I ([Neg not/pas]) [VP(Adv) V ...]]

Within the Minimalist Program framework (Chomsky 1993), as seen in Chapter 1, a lexical item does not need to pick up inflection as it moves. Movement does occur, but for the item to have its features checked. In the case of strong features, movement takes place prior to Spell-out, the point at which the derivation branches to PF and LF. If an item has weak features, then movement only happens after Spell-out. Note that Chomsky (1995b) argues that the operation Move is driven by the requirement that some feature F must be checked, which means that the minimal operation moves simply the feature F. However, in the case of overt movement the entire lexical item is pied-piped, for PF reasons. In terms of Minimalism, the difference between French and English is that the inflected verb in French checks strong (verb) features before Spell-out, while verbs in English do not show overt movement (i.e., movement before Spellout), because the relevant (verb) features are weak. The inflected verb in French moves overtly to a functional head higher than VP. Here, we need to specify where the inflected verb in French moves to.

36

Basic order in declaratives

There is more than one option, and these are investigated in the next subsection.

2.1. Overt verb movement in French The landing site for the inflected verb in French will depend on the structure being assumed. In this section I will first review various proposals that have been made concerning clausal structure in UG. I will then return to the specific question of the landing site of verb movement in French as indicated by adverb position. Pollock (1989) argues that the structure of IP is more highly articulated than had been assumed until then. He does not view Inflection as one constituent with two different sets of features (Tense and Agr). Rather, he argues that each of these sets of features is the syntactic head of a maximal projection, AgrP and TP. The need for AgrP is derived from the need for short verb movement in French (and English have/be), which Pollock assumes is in fact verb movement to Agr. He also argues for the existence of another maximal projection in French and in English, NegP, proposing the following structure: (12)

As mentioned above, Pollock accounts for certain differences between French and English on the basis of verb movement to Infi (or

Adverb position

37

Τ), which he argues to be obligatory in French and blocked in English, where movement to Agr (the first step) is barred. Chomsky (1991) adopts the split Infi hypothesis put forth by Pollock. However, he does not assume that Agr is dominated by Tense. According to him, Agr should be expected to dominate T, to yield standard subject-verb agreement. Furthermore, Chomsky argues for Agr dominating Τ on the basis of morphological evidence that suggests the same conclusion: in several languages where it is possible to obtain relevant evidence, the agreement element is "outside" the tense element in the verbal morphology. The assumption that Agr dominates Τ had in fact been postulated by other researchers. Baker's (1985) "Mirror Principle" states that in a given word the respective order of affixes which may be present reflects the order in which the affixation has taken place through syntactic movement. Thus the affix which is closer to the root must be the one that has moved first, and so on. Given the Mirror Principle, Belletti (1990) considers an inflected verb in Italian where the respective order of the affixes is overtly visible and shows that the order of tense and agreement features in the verbal morphology in Italian is T...Agr. This order of features leads her to the conclusion that Τ is lower than Agr in Italian. In a study of functional categories, Ouhalla (1991) suggests an Agr/Tns parameter related to c-selection that should indicate whether Tense is above Agr or vice-versa. In VSO languages, Agr is inside Tense, and the subject appears inside the verb, whereas in SVO languages Agr is outside Tense, and the subject is found outside the verb. So, if Τ c-selects Agr, TP is above AgrP; if Agr c-selects T, AgrP is above TP. The Romance languages are primarily SVO, which means, following Ouhalla's argument, that Agr appears outside Tense. According to Ouhalla, then, AgrP should appear above TP. Given the arguments in Chomsky (1991), Belletti (1990) and Ouhalla (1991), it can be assumed that, if Agr is indeed present, it dominates T. In the tree diagram shown in (13), we can see that, in this case, overt verb movement is to AgrS (note that the following structure omits AgrOP, and that it includes the NegP proposed by Pollock [1989]):

38

Basic order in declaratives

(13) Spec

AgrSP AgrS'

NP V+T+AgrS

NegP

Neg

TP Spec

T' Τ

VP Spec

V V

Given (13), we have the first possible landing site for the inflected French verb: AgrS. Nevertheless, the status of Agr has been questioned. In more recent work, Chomsky (1995b) subjects functional categories to a minimalist critique. Functional categories are fundamental because of their role in feature-checking, which is what drives movement. According to Chomsky, certain features enter into the interpretation at LF, while others are uninterpretable and must be eliminated for convergence. This difference is expressed in terms of [± Interpretable] features. Among the [+Interpretable] features are categorial features and the phi-features of nomináis. Interpretable features cannot be deleted and so remain accessible to the computation and visible at LF. Chomsky argues that the categories T, C and D have Interpretable features: they "provide" instructions at either or both interface levels. Agr, however, consists solely of [-Interpretable] features. There is evidence from interface relations about T, C and D, but Agr is present for theory-internal reasons only. Looking more closely at the role of Agr, Chomsky argues that Agr lacks an independent Case-assigning feature. This feature is provided by the V or Τ that adjoins to it. When Agr is strong, the position is phonetically indicated by the overt categories that raise to it,

Adverb position

39

which is exemplified by "multiple-subject constructions" (MSCs) with object raising, such as the Icelandic "transitive expletive construction" (TEC) that follows (English words): (14)

there painted [τρ a student] t j [vpt.Wj]]] [AgrP

[AgrP

the house2

According to the analysis provided by Jonas and Bobaljik (1993), the expletive in (14) is in [Spec, AgrSP]; painted is the verb raised to T, which then raises to AgrS; the subject a student is raised to [Spec, T] and the object the house to [Spec, AgrO]. VP, as indicated, contains only traces. In a TEC, AgrS and Τ each have overt specifiers, thus each has a strong feature. In a TEC, three pre-VP positions are required within IP for nominal expressions: expletive, subject and object. One position is provided by T. There is evidence then for two non-Interpretable functional categories, AgrS and AgrO. As mentioned, AgrS is strong in MSCs, providing a specifier and a position for V-raising above the domain of strong T. When it has strong features, Agr provides a position for raising of Τ, V or DP. If it has no strong features, then PF considerations give no reason for it to be present. Therefore, Chomsky concludes that Agr exists only when it has strong features, being simply the indication of a position that must be occupied by overt movement. The function of Agr is to provide a structural configuration in which features can be checked, but the Case-assigning feature is intrinsic to the heads (V, T) that raise to Agr for checking of DP in [Spec, AgrS], so there is no reason to also assign it to Agr. As for phi-features, if Agr has them, they are [-Interpretable]. The conclusion is that Agr consists only of the strong features that force raising. With this conclusion, Chomsky notes that there is no need to deal with optionally strong Agr, or with the difference of strength between AgrS and AgrO (the best case, he had noted earlier, is for AgrO to have the same constitution as AgrS). Since Agr consists only of strong features, it cannot attract covert raising, which must then target Τ and V directly. There is thus no reason to postulate AgrO unless it induces overt raising of DP to [Spec, AgrO]. AgrS appears

40

Basic order in declaratives

in MSCs, but, still according to Chomsky, it lacks independent motivation elsewhere, which means that for French and English, for example, Agr is not in the lexicon18. Ultimately, Chomsky (1995b) rejects Agr (even when it is strong) in favor of a multiple Spec approach, which might affect our conclusion regarding the presence of Agr in French. However, here I am not following the multiple Spec approach, but rather one that argues for the presence of Agr when morphologically/syntactically indicated (cf. Bobalijk and Thráinsson 1998). Since French has the morphology that indicates the presence of Agr (namely, subject clitics), an ultimate conclusion regarding the existence of AgrP in French is indeed possible here. If Agr is not in the lexicon in French, the structure in (13) needs to be reformulated, and with that, one possible landing site for the inflected verb in French will be lost; then, the only likely candidate for Case checking is [Spec, TP], with the V in T. Chomsky argues, as mentioned above, that Case can be checked by T. Let us assume that this is indeed what happens. In this instance, the subject in French (and English) raises to [Spec, TP] and has its nominative Case checked there (movement of the subject is actually driven by the EPP feature, which needs to be checked. A more detailed discussion of movement driven by the EPP feature will be found in Chapter 3.). Since Agr is not present, the subject will not move any further. The inflected verb, then, should raise to check its strong features in T. If Agr does not exist in our system, and the French verb moves overtly to T, we still need a functional projection below TP, in order to be able to account for differences between French and English (as in Pollock (1989), where it is assumed that short Vmovement, present in the grammars of English (for have/be) and French, is movement to Agr) and to be able to account for differences found regarding adverb position in the Romance languages under investigation here. Tenny (1987) investigates aspect and its interaction with syntax. She discusses, among other topics, the place of aspectual markers in X-bar theory. As is well known, Aspect differs from Tense. The latter refers to a moment in time determined by the context in which the expression is used, while Aspect refers to the internal time of the event. Tense must be determined partly through extragrammatical

Adverb position

41

information, whereas Aspect is determined through information contained within the linguistic expression. The Aspect that Tenny discusses as the one determined through information in the linguistic expression is lexical Aspect or Aktionsart. Tenny argues that Tense is syntactically independent from Aspect. The English progressive marker be-ing has a different syntactic distribution from Tense: it may occur where overt Tense may not: (15)

a. * Steve wants PRO goes/went go to Vermont. b. Steve wants PRO to be going to Vermont. (Tenny 1987: 203)

In (15) we see that a syntactic constraint on the occurrence of overt Tense markers is irrelevant for Aspect. English overt Tense must cooccur with a lexical subject or trace, and it cannot appear with PRO. Aspect, however, is not subject to that constraint. In Tenny's analysis, Tense takes syntactic scope over Aspect. She discusses several instantiations of Aspect, and concludes that the possible structures are three, as below: (16)

a.

IP Γ Infi tns/mod

AspP Asp' Asp

VP V' V

NP

42

Basic order in declaratives

IP NP

Γ

Infi tns mod

AspP Asp' Asp

NP

In (16b) Aspect is a specifier of the verb phrase. Given present assumptions that the subject is base generated in the [Spec, VP] position, (16b) can no longer be considered as a plausible option for a phrase structure that includes Aspect. (16c) basically renames VP as AspP, but it is a way in which Aspect might head a syntactic phrase. Nevertheless, we will not consider it, since a verb phrase independent from Aspect is needed in the structure, at least for the reason already mentioned (base-generated position of the subject NP). Also, it does not follow the idea that the aspect morpheme is an independent morpheme from the verb phrase. This idea is captured in (16a). The structure in (16a), with an AspP intervening between TP and VP, is largely identical with structures proposed by other researchers who focus on the projection of a functional category for encoding sentential Aspect (as in Laka 2000). Here I am adopting Tenny's structure, but with AspP referring to sentential Aspect rather than lexical Aspect; I leave open the question of whether there is another projection either below sentential AspP or between the two VPs (cf. Travis 1991, Slabakova 1997) for encoding lexical Aspect. The idea that the aspect morpheme is independent from the verb phrase, captured in (16a), needs to be expressed in the syntax, given that an independent aspect morpheme may appear in some lan-

Adverb position

43

guages, as in Louisiana Creole. Rottet (1993) analyzes that language and concludes that tense and aspect are morphemes separate from the verb (he also teases out mood marker and negation phrase). (17) is an example from Rottet (1993) and shows a sentence where the morphemes for tense and aspect appear separately; (18) is the structure for a sentence where the tense and aspect morphemes appear on the surface, independent from V: (17)

(18)

a.

Pendant yé té apé dinin,... while they ANT PROG dine 'While they were eating dinner...'(ANT= anterior tense marker; PROG=progressive aspect) TP Spec mo

T' Τ

AspP

te

Asp VP ape mözhe I PAST PROG eat Ί was eating' Given the facts in languages such as Louisiana Creole, we should favor a structure where AspP is separate both from the TP and the VP. From the structures in (16), the only one that gives us the possibility of teasing apart tense, aspect and verb is (16a). Tenny (1987) does note that that structure would be proven by a language with aspectual morphology (such as Louisiana Creole, which she does not discuss), and that it is an option in UG. Since (16a) is the only structure that captures the fact that verb and aspect can be separate, it is the one that should be adopted in an analysis that involves AspP. Nevertheless, that structure needs to be updated, given the split Infi hypothesis. Though Tenny has tense and modality under Infi, here I will not be concerned with modality, and Infi will be substituted by Tense:

44

(19)

Basic order in declaratives

TP Spec

T' AspP Spec Asp' Asp

VP Spec

V'

(19) gives us a structure where Agr is not present, which is consonant with the previous discussion about the absence of Agr in phrase structures. (19) also conforms with the fact that Tense takes scope over Aspect, as argued by Tenny (1987). Assuming then (19) as a possible working structure, (13) is reformulated as follows, with a NegP node projected below TP and above AspP: (20)

Adverb position

45

A French declarative such as (21a) might then correspond to the tree diagram (21b) at the moment of Spell-out, in a structure that omits Agr and includes an Asp node (only the relevant projections are represented): (21)

a. Jean parle français. b. TP Spec Jean

T' Τ

AspP

parle

Spec

Asp' Asp

VP Spec

V' V

NP I français

In (21b), the subject has moved to [Spec, TP] from its original position ([Spec, VP], presumably), and the verb has moved to T, in order to check its strong features. Assuming (20), the adverb déjà may adjoin to Asp', following Chomsky's (1995a) assumption that adverbs can adjoin to any X', yielding the observed order of W-déjà-object DP. Here we need to clarify some points regarding movement of the subject. According to Chomsky (1995a), for a Spec position to be projected, it should be necessary that an element moves to or through it. If nothing moves through it, a Spec position is not projected at all, and the head will be a daughter of XP, not of X'. In (21), the subject's landing site is [Spec, TP]. For the argument that the adverb adjoins to Asp' to be valid, it is obviously necessary that Asp' actually be present, which would only happen if a Spec position for AspP were occu-

46

Basic order in declaratives

pied, or if an element moved through it, so that it would be projected. The only element that can move through [Spec, AspP] is the subject, on its way to its final landing site. Taking into account the Minimal Link Condition (MLC), the subject would indeed move through [Spec, AspP], since that is the shortest move in the derivation: movement straight to [Spec, TP] would be longer, violating the MLC. Thus, [Spec, AspP] is projected, for the subject moves through that position on its way to [Spec, TP]. If the Spec position is actually present in AspP, then Asp' is also present, and the adverb can adjoin to The suggestion made here that déjà may adjoin to Asp' can also be based on the semantic relationship between the adverb and the projection it modifies. An adverb such as 'already' could modify Tense or Aspect. Given the structure argued for here, this adverb does not modify Tense in French: adjunction of déjà to T' would yield an ungrammatical sentence: (22)

a. *Jean déjà parle français. TP

ts

V I

tv

NP Δ

français

Adverb position

47

Given the ungrammatically of (22a), Aspect is left as the possible projection to be modified by déjà. This agrees with Cinque's (1999) analysis of 'already' as an aspectual adverb (however, I am going to partially modify his analysis of 'already' below). I return now to the discussion of the position of adverbs in Brazilian Portuguese, Spanish, Italian and French.

2.2. Adverb position in a structure without AgrΡ Looking at Brazilian Portuguese, we see in (7a), repeated here as (23), that the adverb já appears before the verb: (23)

Eu já sei portugués. I already know Portuguese

If we assume that the subject moves out of VP to have its nominative Case checked in [Spec, TP], it is clear that the adverb is below that position. But it is necessary to show that the subject does overtly move out of VP, as Koopman and Sportiche (1991) propose. They assume that the subject NP moves from [Spec, VP] to [Spec, IP] for Case-assigning reasons; updated to the Minimalism framework, this assumption means that the subject NP has its Case checked in [Spec, TP]. One of the arguments that Koopman and Sportiche use to strengthen their proposal is Q-floating. They assume that raising of the subject optionally pied-pipes Q. If it doesn't, there is a 'floated' Q next to the trace of the subject NP. According to them, this explains why Q appears between Infi and VP (or between Τ and VP): because the trace of the subject does. Since the quantifier is adjacent to this trace, it is anaphorically dependent on the subject in [Spec, TP], Q-floating is found in Brazilian Portuguese, as well as many other languages. (24) gives us some examples: (24)

a.

Todos os meninos viajaram/chegaram. 'All the boys traveled/arrived' b. Os meninos viajaram/chegaram todos. The boys traveled/arrived all

48

Basic order in declaratives

The subject Todos os meninos moves from its base position in [Spec, VP] to [Spec, TP] in (24a). There, however, the movement is not apparent. It becomes clear in (24b), where we see that the quantifier was left stranded in its base position, and only os meninos and the verb moved. Thus we see that the subject NP does move out of its base position inside VP. The structures for (24a) and (b) are given in (25a) and (b), respectively: (25)

a.

TP

Spec T' [Todos os meninos]¡ Τ AspP viajaram-j Spec Asp' Asp^^VP Spec ti b.

V' Δ t¡

TP

Spec Τ' [Os meninos\\ Τ AspP viajarams Spec Asp' Asp Spec todos t¡

VP V' L 20

Adverb position

49

Returning to adverbs, it is possible that in Portuguese and Spanish (and Italian, cf. [7c]) an adverb is adjoined to T' if adverbs can adjoin to X'. The structure in this instance would be as follows: (26)

TP

Spec Subject ya/jä

T' Τ' T+V

AspP

Spec

Asp' Asp

Note that in (26), an important assumption is that ya/ja can modify tense in Spanish and Brazilian Portuguese, thus adjoining to T'. Based on sentences that contain this adverb, it is also possible for the adverb to adjoin to Asp'; in that case, the verb would only move as high as Asp. This option with regard to the landing site of the tensed verb in Spanish and Brazilian Portuguese is discussed further later in this chapter (section 2.4.). For now, suffice it to say that Spanish and Brazilian Portuguese do exhibit that optionality regarding the position the verb raises to in declaratives. Let us take a moment to examine the possibilities for positioning of adverbs. In Spanish, for example, adverbs such as ya and siempre can appear before or after the verb in declaratives. Below are the possible orders (excluding sentence-final position): (27)

a.

Ya habla muchas cosas en español. (ya-V-DO) Already speaks many things in Spanish b. Habla ya muchas cosas en español. (V-^a-DO) Speaks already many things in Spanish c. Siempre estudia sus lecciones. (siempre-Y-DO) Always studies his/her lessons

50

Basic order in declaratives

d.

Estudia siempre sus lecciones. Studies always his/her lessons

(V-siempre-DO)

(27) shows that ya and siempre might be able to adjoin to Asp' or to T', and that the verb also enjoys some flexibility in terms of its landing site, which could be either Asp or T. So, if the verb always raised to T, then in (27a) and (c) the adverb would be adjoined to T \ and in (27b) and (d) it would adjoin to Asp'. However, it is possible that the verb optionally raise only as high as Asp, which could be the case in (27a) and (c); in those instances, then, the adverb could be in either Asp' or in T'. Thus, there are two alternatives that might account for (27): optional position of ya, and optional raising of the verb to T. Given the facts with ya and siempre in Spanish, it is possible that Cinque's (1999) universal order of adverbs might not be accurate. If indeed those adverbs can adjoin to either T' or Asp', the only way to keep Cinque's universal order would be to claim that there are two lexical entries for those adverbs. Otherwise, his universal hierarchy of adverbial specifiers may not be precise (in spite of the examples from different languages that Cinque shows). The flexibility in the landing site of the tensed verb that may be possible in Spanish and BP is not present in Italian and French. For Italian (as in [7c]) and French the structure would then be (28): (28)

TP T'

Spec Subj T+V

AspP Spec Asp' ts già/déjà Asp' VP

Asp tv

tv

DO

Sentential adverbs

51

At this point, it might seem possible for French to exhibit a structure such as (26), with déjà adjoined to T', and the subject in [Spec, TP], Nevertheless, that would not yield a grammatical result in French: (29)

*Jean déjà parle jrançais.

So far, nothing rules out (29). Therefore, we will have to postulate that the subject and the verb in French are actually at a position above TP, an assumption that will predict the correct order in French. But before reaching that point, let us discuss some differences between French and Italian with regard to the. position of adverbs. Although the position of già/déjà may lead us to believe that Italian and French adverbs behave similarly, these two languages exhibit great differences regarding the position of sentential adverbs. As a matter of fact, the two languages do not show exactly the same behavior regarding 'already' either: whereas the only possible place for French déjà is immediately following the inflected verb, Italian già may precede the verb, just as Spanish ya and Portuguese jä. The preverbal già seems to be "focused", as noticed earlier. Nevertheless, Italian allows for that possibility, whereas French does not. The differences found between these two languages may be an indication of a difference regarding the position of the subject or the verb. The position of sentential adverbs may shed some light on subject/verb position. Therefore, I turn to sentential adverbs next.

3. Sentential adverbs French and Italian behave differently in relation to sentential adverbs. That difference can be seen in the examples below: (30)

a.

Probabilmente Gianni ha sbagliato. Probably Gianni has mistaken b. Gianni probabilmente ha sbagliato. Gianni probably has mistaken c. Lui ha probabilmente sbagliato. He has probably mistaken

52

Basic order in declaratives

In Italian, we see that the sentential adverb probabilmente can appear in sentence initial position, after the subject and between the auxiliary and the main verb. In French, sentential adverbs do not enjoy so much freedom: (31)

a.

Jean a probablement fait des erreurs. 'Jean has probably made mistakes' b. *Jean probablement a fait des erreurs. Jean probably has made mistakes c. * Probablement Jean a fait des erreurs. Probably Jean has made mistakes

As we see in (31), the only acceptable position for probablement is between the auxiliary verb and the main verb. The adverb cannot appear in sentence initial position, nor can it be located between the subject and the inflected verb. In case a simple verb appears, the adverb can only follow the verb, not precede it nor be the first element in the sentence: (32)

a.

Jean aime probablement la linguistique. Jean loves probably linguistics b. *Jean probablement aime la linguistique. Jean probably loves linguistics c. * Probablement Jean aime la linguistique. Probably Jean loves linguistics

Analyzing the examples in (30), Belletti (1990) proposes the possibility of free recursion of AgrP. According to her, this allows a sentential adverb such as probabilmente to freely modify any AgrP. It could modify the highest AgrP (as in [30a]), which is located above AgrSP. Or it could modify an AgrP lower than the highest, yielding the order in (30c). She summarizes her proposal with the following structure:

Sentential adverbs

(33)

53

AgrP (probabilmente) (NP)

AgrP Agr'

(probabilmente)

AgrP (NP)

^Agr^ Agr

AuxP Aux

In (30c), the adverb does not modify the highest AgrP, but one below it. The auxiliary verb fills the highest and empty Agr, hence the order S-Aux-Adv-Pst Prt. To explain (30b), Belletti proposes that the adverb occupies the same position as in (30a), but the subject is found in a peripheral position, left dislocated or topicalized. Evidence in favor of this analysis, she argues, comes from sentences whose subject is an indefinite quantifier such as nessuno 'nobody'. Since an indefinite quantifier cannot be left dislocated, her analysis (outlined above) predicts that, if nessuno is the subject, only the topicalization structure should be available. Given sentences such as (34), where a clear contrastive stress on nessuno is necessary (since the order nessuno probabilmente is not allowed without stress), Belletti concludes that these structures require left dislocation or topicalization (only topicalization in the case of an indefinite quantifier): (34)

a. NESSUNO probabilmente telefonerà alle 5. Nobody probably will call at 5 b. Dicono che NESSUNO probabilmente telefonerà alle 5. They say that nobody probably will call at 5.

54

Basic order in declaratives

The sentences in (34) thus provide evidence, in Belletti's analysis, that the sentence-initial subject in (30b) is in a topicalized position: nessuno in the examples above is topicalized, and so is the subject in (30b). Belletti's analysis for (30b) may also hold for Brazilian Portuguese. First, let us exemplify the Brazilian Portuguese correspondents of (30)21: (35)

a. Provavelmente Joäo tinha errado. Probably Joäo had made a mistake b. Joäo provavelmente tinha errado. Joäo probably had made a mistake c. Joäo tinha provavelmente errado. Joäo had probably made a mistake

Assuming Belletti's analysis of (30b) (i.e., that the subject in [30b] is topicalized), Joäo in (35b) could also be topicalized. If this analysis is correct, in the case of an indefinite quantifier subject only topicalization should be possible, as in Italian. In fact, the data in (34) find parallels in Portuguese, where the order Ninguém provavelmente V is only possible if Ninguém is stressed. Otherwise, the result is ungrammatical: (36)

a. NINGUÉM provavelmente telefonará às 5. Nobody probably will call at 5 b. *Ninguém provavelmente telefonará às 5. c. Provavelmente ninguém telefonará às 5. Probably nobody will call at 5

Thus we see that a subject that precedes a sentential adverb will also be in a topicalized position in Brazilian Portuguese. Now, following our assumptions, in a structure without Agr, TP is the absolute root. A topicalized element is thus found in a position above TP. Since we no longer assume a structure with Agr, the argument that accounts for the position of sentential adverbs using recursion of Agr does not hold: the sentential adverb has to be adjoined to some other projection. Still assuming that the subjects in (30b) and (35b) are indeed topicalized, the adverb may be adjoined to T', given the assumption

Sentential adverbs

55

in Chomsky (1995a) that adverbs may adjoin to X'. However, being a sentential adverb, provavelmente/probabilmente may also be adjoined to TP, since that type of adverbs should adjoin to the root. Adjunction to TP, in this case, does not go against Chomsky's assumptions, since TP is neither an argument nor a predicate. According to Chomsky (1995a), as mentioned before, adjunction results in a two-segment category, but "not a term of the structure formed." Adjunction to the root is not barred under these assumptions: the root is not a predicate or an argument of anything, as just mentioned for TP; although adjunction to it would create a new term, this should not be problematic at the LF interface. I will assume that this is what happens with sentential adverbs, i.e., they modify the sentence and adjoin to the root. Below we have an example with a simple verb and one possible structure for it: (37)

a. Gianni probabilmente sbaglia b. TopP

probabilmente

TP

Asp VP When the adverb is located in sentence initial position, as in (30a), it is simply adjoined to TP, with the subject in the [Spec, TP] position. As for a sentential adverb placed after the inflected verb, as in (30c), we could assume that it is adjoined to Asp'. But that assumption

56

Basic order in declaratives

would go against the idea in the paragraph above, i.e., that sentential adverbs adjoin to the root because they modify the sentence. If we are to keep this idea, we need to analyze (30c) as another instance of topicalization of the subject. But now the inflected verb must also have moved to a position above TP, to yield the order S-VaUx-AdvVpart. However, a structure such as (37b), with an adverb adjoined to a maximal projection, might seem contrary to the view in Chomsky (1995a), where adverbs adjoin to X'. But in that work adverbs are assumed to also be able to adjoin to AgrP. In a structure without topicalization, adjunction to TP should then be possible. Nevertheless, (37a) involves topicalization, which means that TP is no longer the root of the sentence. The relationship between Top and TP is normally understood to be one of predication, but it is not clear that the sentence is an argument of Top. Since adjunction should not be to a predicate or an argument, in the light of Chomsky (1995a, b) adjunction to TP would not be possible if Top were present, if in fact there is a predication relation between Top and TP. Thus, adjunction to TP in this instance seems to pose theoretical problems. Therefore, here too the sentential adverb may be adjoined to T', instead of TP, thus yielding the following structure: (38)

TopP Spec Subj

Top' TP

Top Vaux + Τ

Spec

T'

ts

sent. adv.

T'

tv+T

AspP

Spec ts

Asp'

Δ

Sentential adverbs

57

As we see in (38), the sentential adverb may be adjoined to T'. We can then keep the idea that sentential adverbs only adjoin to TP or T', since they modify the sentence. But adjunction to TP would bring several theoretical problems, as discussed. Therefore, adjunction to T' seems to be the best alternative, since the adverb would still be in the sentential phrase, and there is no violation of the ban on adjunction to arguments. Some problems are also posed by the examples in French, such as (32). We need to explain why probablement can only be found in postverbal position. Our explanation shall relate to the position of the subject and the verb in French. Recall that it was suggested earlier that the differences in adverb position between French and Italian may indicate different subject positions in these two languages. Suppose that the subject in French actually raises to a higher position (above TP). Certain peculiarities about French may lead us to this supposition. For example, French is the only Romance language under discussion (excluding, importantly, northern Italian dialects) that has subject clitics: those are absent in Spanish, Italian, and Portuguese. This may signify that the subject in French does not occupy the [Spec, TP] position, as it seems to be the case for Italian and Brazilian Portuguese22. If this is the case, what position does the subject occupy in French? One candidate for that projection is Agr (see endnote 4). In general, French does not allow topicalization. Belletti (1990: 45) mentions that "it is well known (although the reason is not understood) that no topicalization process is available in French."23 So, a projection above TP in French would not be a topicalized position, as it is in Italian and Brazilian Portuguese, which leads us to postulate Agr as a possible projection above TP in French. With a projection above TP we are able to account for the ungrammaticality of (29): Jean is in [Spec, AgrP], the verb is in Agr, but déjà (or any other adverb) cannot adjoin to Agr, since adjunction is only possible if an element adjoins to a category it modifies (and Agr is not "modifiable"). Thus, (29) is ungrammatical, and the adverb is only found in postverbal position.

58

Basic order in declaratives

If AgrP is in fact projected in French, it has strong features, which causes the subject to raise to [Spec, AgrP], and the verb to Agr before Spell-out: (39)

[Agrp Subj [ A gr· V

[TP t s [Τ· tv . . . ] ] ] ]

In (39), the subject moves to [Spec, AgrP] leaving a trace in the [Spec, TP] position. The verb presumably also raises to Agr in order to check the strong V-feature in that position, leaving its trace in T, as can also be seen in the structure in (39). At this point, some theoretical considerations must be taken into account. In a structure where AgrP dominates TP, is the latter an argument of the former? Is it a predicate? Structurally, if AgrP dominates TP, TP is a complement of AgrP. Semantically, however, there is no notion that TP is an argument of AgrP: it does not get a theta-role from AgrP. That being the case, even though TP is a complement of AgrP structurally, it is not one semantically. Thus, in principle adjunction to TP in French could be possible. The assumption here is that the only maximal projection to which anything can adjoin is TP, which is (although not always) an absolute root. As mentioned earlier, Chomsky (1995a) argues against adjunction to a predicate or an argument, which would create an extra term that would not be interpretable at the interface. Our suggestion that French projects Agr allows us to account for the ungrammaticality of sentence initial adverbs in this language: if adjunction should capture a modification relation between the element that adjoins and the position it adjoins to, a sentential adverb should only be able to adjoin to TP or T'. The adverb cannot adjoin to AgrP because modifying AgrP would not be semantically interpretable (so, part of Belletti's proposal has to be rejected). Thus we have one more reason why French should not allow adjunction of an adverb to AgrP, and one more reason why probablement in (32a) can only be in postverbal position: (40)

a. * Probablement Jean aime la linguistique. b. *Jean probablement aime la linguistique. c. Jean aime probablement la linguistique.

Sentential adverbs

59

Probablement cannot be in sentence initial position, as in (40a), for the reasons discussed above. It cannot appear between the subject and the verb either ([40b]): if the subject is in [Spec, AgrP] and the verb is in Agr, the adverb would need to adjoin to Agr' to appear in that position. But adjunction to Agr has been argued against, be it AgrP or Agr'. Thus the only position left for the adverb in French is after the verb, (40c), as we can see in the following structure: (41)

[AgrP Jean [Agr· aime [χρ ts [τ· probablement [τ- tv ...]]]]]

In (41) we have the sentential adverb adjoined to T', not to TP. As in the Italian example, the sentential adverb could be adjoined to TP, but that adjunction raises theoretical problems, as noted above: since AgrP is located above TP in French, in order for adjunction to TP to be possible, we would need to argue that TP is not an argument of AgrP. However, adjunction to T' seems to yield the same results as adjunction to TP would. Therefore, here I opt for adjoining probablement to T', rather than to TP. The account for sentential adverbs in French can extend to déjà, which also has only one possible surface position. However, we need not assume that déjà (not a sentential adverb) adjoins to T \ It can be adjoined to Asp' too, as are the correspondent adverbs in certain orders in Spanish, Portuguese and Italian. The question of where déjà should adjoin bears on what it should modify, since it has been assumed here that adjunction of an adverb occurs to the projection it modifies. The possible structures for a sentence such as (7d), repeated here as (42a), should be (42b) and (42c): (42)

a. Jean parle déjà français. b. [AgrP Jean [Agr' parle [Tp t s [ r déjà [ r t v [asPp t s [as P · t v [vp ts [v* tv français] c. [AgrP Jean [Agr' parle [TP t s [τ t v [asPp h Usp· déjà [Asp· t v [vp t s [v· ty français]

As (42b) and (c) show, adjunction of déjà is possible both to T' and to Asp'. In either case, the order obtained is the grammatical one in (42a).

60

Basic order in declaratives

To summarize, the differences found regarding the position of sentential adverbs in Italian (and Brazilian Portuguese) and French have led us to argue that adjunction to TP occurs in Italian (and Brazilian Portuguese), even when topicalization occurs and TP is not the absolute root of the sentence. The facts in French are accounted for by postulating an AgrP above TP, to where the subject and the verb move before Spell-out. Since adjoining to Agr would have no semantic import, adverbs in French do not adjoin to that projection. This assumption predicts that no adverb is found between the subject and the verb, which is indeed the case in French, where adverbs are found only in postverbal position.

4. Verb position and other adverbs in Brazilian Portuguese Our discussion of the position of certain adverbs can be expanded to other examples. We have already seen that sentential adverbs can appear in different surface positions. Let us now look at an adverb such as sempre 'always', which can appear in different positions in the sentence: (43)

a. A Bia sempre faz o dever de casa. b. A Bia faz sempre o dever de casa. c. A Bia faz o dever de casa sempre. 'Bia always does the homework'

In (43a), the adverb appears between the subject and the verb, like the adverb já. Following the analysis for the position of ja, we may assume that the adverb in (43a) is also adjoined to T', which means that the verb would be located in T. In (43b), where it follows the verb, the adverb may be adjoined to Asp'. Given the previous discussion about adverb position, the adverb should adjoin to the projection it is modifying. Thus, if we assume that the verb is in T, the adverb in (43a) would modify T, whereas in (43b) it would modify Asp. When it is found in VP-final position, as in (43c), the adverb is right adjoined to V', and it has a focalized reading. Given these assumptions, the structures for the sentences in (43) should be the following:

Verb position and other adverbs in BP

(44)

a. b. c.

61

[TP A Bia [r sempre [R faz [Aspp t s [Asp' t v [VP t s [V t v o dever de casa]]]]]]] [TP A Bia [ r faz [AspP t s [Asp' sempre [Asp· t v [VP t s [ν· t v o dever de caja]]]]]]] [TP A Bia [r faz [Aspp t s [ASP' t v [VP t s [V [ Ν tv o dever de casa] sempre]]]]]]24

The structures in (43) show the adverb adjoined to different positions, whereas the inflected verb is always in T. An alternative analysis, as discussed earlier, would have the adverb in (43a) and in (43b) adjoined to the same position; the verb would be the element whose position would vary. It may be the case that in declaratives the verb optionally lands in Asp, not raising all the way up to T. Thus, it is possible that in (43 a) the verb is in Asp, and in (43 b) it is located in T. In that case, the adverb in both examples would be adjoined to the same position: Asp'. This alternative analysis would entail the following structures for (43 a) and (b): (45)

a. b.

[TP A Bia [ τ Τ [AspP t s [Asp· sempre [Asp, faz [ Vp t s [v tv o dever de casa]]]]]]] [τρ A Bia [τ faz [AspP t s [Asp' sempre [Asp, ts [ Vp t s [v ty o dever de casa]]]]]]]

The structures in (45) show that the verb in Brazilian Portuguese declaratives might optionally raise only up to Asp (for another possible analysis of [43c], see note. 11). This alternative analysis maintains the idea that the adverb adjoins to a projection it modifies, in this instance Asp'. This optionality is corroborated by facts in Spanish: (46)

a. b.

Pablo a menudo juega con su hermana. Pablo often plays with his sister Pablo juega a menudo con su hermana. Pablo plays often with his sister

In (46) we see that α menudo 'often' may surface either before or after the verb in Spanish. This is an adverb that seems to be clearly as-

62

Basic order in declaratives

pectual in nature; if this is so, to conform with the assumption that an adverb adjoins to a projection it modifies, a menudo would be adjoined to Asp', which in turn means that the verb is the element whose landing site varies. (46a) indicates that the verb would be located in Asp, whereas in (46b) it would be in T. This discussion regarding the landing site of the verb raises the question of whether the verb must move to a certain position, as opposed to where it can (optionally) move to. Taking the position that a given adverb can only adjoin to one position means that the verb is the element that moves to different projections. If Spanish a menudo, for instance, can only modify aspect, then what the data show is that minimally in declaratives the verb has to move at least as high as Asp, with the option of moving as high as T. In this case, the assumption would be that the strength of the feature in Τ is optional, which accords with the idea that parametrization is in functional categories, not lexical categories. A priori, both analyses (the verb always moves to T, or the adverb always adjoins to Asp') appear to yield the same result. However, as will be discussed in greater detail in Chapter 4, the verb in Spanish and Brazilian Portuguese interrogatives does raise to T, given the data below: (47)

a.

¿Con quién juega a menudo Pablo? With who plays often Pablo b. * ¿Con quién a menudo juega Pablo? With who often plays Pablo

(Spanish)

(48)

a.

(BP)

Que medidas o governo toma amiúde? What measures the government takes often b. *Que medidas o governo amiúde toma? What measures the government often takes

The w/z-questions in (47a) and (48a) show that the verb is located above the adverb: the adverb being adjoined to Asp', the verb raises to T. What the ungrammatically of (47b) and (48b) shows is that this adverb cannot freely adjoin to either T' or Asp': if that were the case, (47b) and (48b) would be grammatical, since the verb would be in Τ

Verb position and other adverbs in BP

63

and the adverb would adjoin to T \ But, since this adverb apparently can only adjoin to Asp' in both Spanish and Brazilian Portuguese, placing it before the verb results in ungrammaticality. (Chapter 4 contains a detailed discussion of mandatory V-to-T in interrogatives.) Therefore, it appears that at least certain adverbs must adjoin to only one projection, which means that in declaratives it is the verb that optionally raises to Asp or to T. Earlier in this chapter, we had already seen that the verb in Spanish declaratives, for example, might optionally raise to Asp or to T, based on facts with the adverbs siempre and ya. (cf. [27] and preceding and following discussion.) Here we can examine briefly some data containing those adverbs in w/j-questions. In that environment, both adverbs may surface in preverbal position: (49)

a.

¿Quéya aprendiste? What already you-learned b. ¿Qué siempre estudias? What always you-study

Given our previous conclusion that in Spanish interrogatives the verb must raise to T, in (49) the adverbs must then be adjoined to T', unlike the adverb a menudo, for example, which must adjoin to Asp' in interrogatives. The discussion above indicates that the verb in Spanish and Brazilian Portuguese declaratives may optionally raise to Τ or stop at Asp in declaratives, based on the position of adverbs such as a menudo in Spanish and amiúde in Brazilian Portuguese, whose nature is aspectual, meaning that it would have to adjoin to Asp'. However, not all adverbs display such clear semantic import. It is possible that certain adverbs indeed adjoin to several different projections, as sentential adverbs do. This assumption - that there are adverbs that may adjoin to more than one projection - is taken below, where we discuss compound verbs and the position of adverbs.

64

Basic order in declaratives

4.1. Adverbs and compound verbs When we have a verb tense that involves an auxiliary, certain adverbs may also appear in more than one position. We have seen that this happens with sentential adverbs, such as provavelmente, which can appear in sentence-initial position, before the auxiliary verb, and between the auxiliary and the main verb. Adverbs of manner can also appear in several positions in the sentence. Let us take cuidadosamente 'carefully', for instance: (50)

a. Cuidadosamente (,) o Ivo tinha feito o trabalho. b. O Ivo tinha cuidadosamente feito o trabalho. c. O Ivo tinha feito cuidadosamente o trabalho. d. O Ivo tinha feito o trabalho cuidadosamente. e. IO Ivo cuidadosamente tinha feito o trabalho. 'Ivo had carefully done the work'

Before discussing the position of the adverb in (48), it is necessary to consider the position of the participle feito. In Chomsky (1991), a past participle is viewed as heading AgrP, following Kayne (1989). Belletti (1990) assumes that the past participial Agr takes a further functional projection as a complement (which she calls AspP25), whose head contains the past participial inflection proper (-t in Italian), while the Agr head contains the typical agreement features of gender and number, which can be overtly expressed in Italian, or realized with the unmarked masculine singular ending, depending on different syntactic structures. In this work, I have argued against a "default" presence of AgrP.26 Thus the participle could not move to an Agr projection, simply because it is not present in the structure assumed here. It still needs to move, however, as we see from (50c), where an adverb is found between the participle and the direct object. In Belletti (1990), the participle raises to a phrase immediately above VP. That is also the position of AgrOP in Chomsky (1991). Thus, we can assume that the participle moves to a phrase above VP. Let us call this phrase PartP, for Participle Phrase. Since this phrase is above VP, where the subject originates, it needs a Spec position for the subject

Verb position and other adverbs in BP

65

to go through, in order not to violate the MLC. Therefore, the participle will move to the head of that phrase. We also need to address the auxiliary verb, and the position where it starts. I will follow Belletti (1990) and assume that the auxiliary verb starts off in an AuxP, found above the PartP. AuxP should project a Spec position, for the reasons given above (namely, movement of the subject and the MLC). A sentence such as (51a) then has the structure in (51b): (51)

a. b.

O Ivo tinha feito o trabalho. [τρ O Ivo [ R tinha [ A P t s [ASP' taux [AUXP t s [PartP feito [ v p t s [ y t v o trabalho]]]]]]]}] SP

[AUX' taux

The participle in (51a) moves from VP to PartP, and the auxiliary verb moves from Aux to T, where it checks the V-features of Tense. The subject moves from its base generated position in [Spec, VP] to [Spec, TP] via the other Spec positions in the structure, leaving its trace. Returning now to the positions of the adverb in (50), we can say that in (50a) the adverb adjoins to TP, thus modifying the whole sentence. In (50b), where it appears after the tensed verb, it is adjoined to Asp'. Following our assumptions about adverbs - that they adjoin to projections they modify, as in Travis (1988) - , the adverb should not adjoin to Aux' or Part', because, given the assumptions in Travis (1988) and Chomsky (1995a), the result of adjunction to an auxiliary or a participle would be semantically uninterpretable27. Thus, in (50b) there is only one projection that the adverb can semantically modify: Asp. In (50c), the adverb modifies the verb, and is thus adjoined to V'. (5Od) shows an adverb with a focalized reading, as does (43c): in both instances the adverb is right adjoined to V', or alternatively, as in Cinque (1999), the adverb is located in the spec position of a VP 'shell', followed by successive movement to the left of the lower VP to a higher specifier (see note. 11). With the inflected verb in T', the adverb in (50e) would be adjoined to that position, but as the sentence shows, adjunction of cuidadosamente to T' does not seem possible, which is accounted for by our assumption that an adverb will adjoin to a projection it modifies, and this adverb would not

66

Basic order in declaratives

modify Tense. (52) below shows the structures for the sentences in (50); (53) is the tree diagram corresponding to (52a), included here for ease of exposition: (52)

a.

[TP Cuidadosamente [χρ O Ivo [ r tinha [ A s p p t s Usp' taux [AuxP taux [PartPfeito [ V p t s [ y t v o trabalho]]]}]]]]}

b.

[χρ O Ivo [χ' tinha [ a s p p t s [Asp' cuidadosamente[Asp'taux

[AuxP taux [PartP feito[yP t s [ y t v o trabalho]]]]]]}]] [χρ O IVO [ χ ' tinha [ a s p P t s [ a s P ' taux [AuxP taux [ P a r t P / e # o [ V P ts [ v cuidadosamente [ y t v o r r a ¿ a / / z o ] ] ] ] ] ] ] ] ] ] [χρ Ο Ivo [τ> tinha [ A s p P t s [ a s P ' taux [AuxP taux [PartPßito [vp ts [ y t v o trabalho] y cuidadosamente ]]]]]]]]]

DP o Ivo T+V+Aux tinha

AspP ts

Asp' Asp

AuxP

taux

taux

PartP

t v o trabalho As we see in the structures in (52) above, certain adverbs can adjoin to several projections in BP. In this case, it looks as though the tensed

Verb position and other adverbs in BP

67

verb does check features in Τ before Spell-out: in (52b) the only projection that the adverb can adjoin to is Asp', and the auxiliary verb will only be found before the adverb if the auxiliary has moved to Τ prior to Spell-out. As we can see, the auxiliary verb tinha does not stop at Asp, but rather moves as high as T. This behavior is not like that of main verbs, which do have the option of stopping at Asp in declaratives, as we have seen here. However, the behavior of the auxiliary verb in Brazilian Portuguese is in accordance with the facts in English, where auxiliary verbs must move as high as they can (i.e., T), although main verbs do not seem to move out of VP. These facts are expected if raising of auxiliary verbs, which do not have any semantic content, is a purely syntactically-driven movement. Movement of main verbs, on the other hand, happens both for syntactic reasons (such as feature checking) and LF interpretation. To summarize the discussion on adverbs in Brazilian Portuguese, we have seen that jâ can adjoin to T', but that is not the only projection to which adverbs can adjoin. An adverb adjoins to the projection it modifies: when it modifies the sentence, the adverb adjoins to TP, which is the only maximal projection that allows for adjunction, following Chomsky (1995a); when modifying Tense or Aspect, the adverb will adjoin to T' or Asp', respectively. And adjunction to V' happens when the adverb modifies the verb. Some adverbs, as seen in (52), can adjoin to different positions. The discussion of adverbs and compound verbs shows that, although a main tensed verb in Brazilian Portuguese does not need to move beyond Asp, an auxiliary such as tinha is found necessarily in the Τ node. Thus, the discussion of position of adverbs can shed light on the landing site for main verbs and auxiliaries as well. Returning to the position of the adverb já 'already', earlier in this chapter, we find a discussion that suggests that this adverb, as well as its Spanish counterpart ya, could adjoin either to T' or to Asp'. The position of that adverb is the topic of the next subsection.

68

Basic order in declaratives

4.2. The position of ja In this chapter we have discussed the position of subjects and the possible lack of AgrP in the structure, which would lead to the subject's Case being checked in [Spec, TP]. With the subject checking its Case in [Spec, TP], a preverbal adverb could indicate adjunction to T', if the verb moves to Τ overtly. I assume that the verb in Italian does move to Τ before Spell-out. In Spanish (and Brazilian Portuguese), we have seen that the verb in declaratives has the option of raising only as far as Asp. In declarative sentences with ya/ja, the verb can, a priori, be located in either Τ or Asp. For ease of exposition, let us show only one of those possibilities here, namely adjunction to T':

ya

AspP I habla Spec

Asp'

I t s Asp

VP

I tv

V'

NP

V

NP

I

Δ

t

español

In (54), the adverb adjoins to T' and the verb moves to T, leaving a trace in Asp. The subject NP raises to [Spec, TP] from its base position within VP, since it checks Case overtly in this construction. The object has its Case checked covertly by the verb, according to as-

Verb position and other adverbs in BP

69

sumptions in Chomsky (1995a). In Chapter 3,1 will analyze other orders in Spanish, and we will see that in fact it is possible for the subject to remain in its base position. In Chapter 4, the question of Case checking and checking of the EPP feature in Spanish (and Italian) is revisited in light of the syntax of w/z-constructions in the language. Italian would also exhibit a structure such as (54) for instances of preverbal già. When già is postverbal, however, it could be adjoined to Asp'. A priori, it could also be adjoined to V', which is a possible candidate for adjunction. But this adverb does not modify the verb, and thus cannot be adjoined to V'. (55) is an example of a structure for 'neutral' (i.e., postverbal) già: TP

(55)

T'

Spec Gino

Τ

AspP

parla Spec ts

Asp' Asp'

Adv I già

VP

Asp tv

NP

V' V g t

NP 4 italiano

In (55) we see that the adverb is adjoined to Asp', and the verb and the subject raise to Τ and [Spec, TP] as they do in (54). Although Italian già can optionally (and emphatically) be positioned before the tensed verb, where it is normally found in Portuguese and Spanish, the possibility for jà to immediately follow the

70

Basic order in declaratives

tensed verb is not found in Brazilian Portuguese ([56]). A postverbal ya may appear in (at least several dialects of) Spanish ([57]): (56)

*A Ana fez já o dever de casa. Ana did already the homework

(57)

%Ana hizo ya la tarea. Ana did already the homework

Postverbal ya in Spanish declaratives has the same reading as preverbal già in Italian, i.e., it is focalized. In Brazilian Portuguese, however, this adverb cannot appear immediately after the verb, an impossibility that needs to be accounted for. One possible explanation is related to the position of the verb at the moment of Spell-out. If the verb does not move all the way up to T, but rather stops at Asp (a possibility that does exist in the language) and only checks its features in Τ after Spell-out, then the adverb would not have another projection to adjoin to, since this adverb does not modify the verb and thus cannot adjoin to V'. Although it accounts for the facts with já, we have seen that the verb in Brazilian Portuguese may indeed raise to T, and the examples with já do not show clearly that the verb raises only to Asp. Recall that our discussion of optionality of the landing site of the verb was based on sentences with adverbs that surfaced in more than one position, such as sempre 'always', which can indeed appear after the tensed verb: (58)

O Ivo faz sempre o trabalho. Ivo does always the work 'Ivo always does the work'

In the sentence in (58), the adverb appears between the verb faz 'does' and the direct object o trabalho 'the work'. If the verb always moved to thte Asp node, no temporal adverb would be able to adjoin to a projection between that node and the participle: adjunction is only to X', and adverbs adjoin to elements they modify. Adverbs of time can naturally modify T, and also possibly Asp. But, unlike manner adverbs, they do not adjoin to V', for they do not modify the verb.

Verb position and other adverbs in BP

71

Therefore, sempre 'always' in example (58) has to be adjoined to Asp'. The adverb being adjoined to Asp' means that the verb, being above it, has in fact raised to T. This is clearly seen in the structure for (58), given below: (59)

tv

Spec ts

V'

V

I

tv

NP

Δ

o trabalho

Having confirmed that the verb may indeed raise to Τ before Spell-out, we can entertain another hypothesis about the only possible position of jä in Brazilian Portuguese. This adverb may in fact be cliticizing to the verb in BP. Besides being monosyllabic, it patterns with other clitics with regard to disallowing other words (besides other clitics) to intervene between it and the verb: (60)

a. *Bia já provavelmente fez o dever de casa. Bia already probably did the homework b. A Bia aínda provavelmente está fazendo o dever de casa. Bia still probably is doing the homework

Example (60a) shows that other adverbs cannot appear between já and the tensed verb, whereas we see from the example in (60b) that

72

Basic order in declaratives

other adverbs may allow for that possibility. However, clitics are found in exactly that position (between the adverb and the verb) with já: (61)

O Ivo já te

chamou.

Ivo already you called 'Ivo already called you' Since only clitics can intervene between já and the verb, it is reasonable to assume that this adverb cliticizes to the verb in Brazilian Portuguese, where clitics are found preverbally. Therefore, it would not be possible for já to appear postverbally. That explains the impossibility, in Brazilian Portuguese, for já to appear after the verb, even with a focalized reading.

5. Conclusion Having discussed the position of adverbs in Brazilian Portuguese, Spanish, Italian and French we have a clearer idea of what a structure for those languages might look like. Brazilian Portuguese, Spanish and Italian all admit preverbal adverbs, such as 'already' (although preverbal già in Italian denotes a heavier focus, as noted). The three languages also exhibit similar behavior regarding sentential adverbs. In French, on the other hand, adverbs can only occupy the postverbal slot in a sentence with a tensed verb (for the position of adverbs with infinitivals in French, see Pollock 1989). All four languages seem to prefer the SVO order, although Spanish and Italian allow for VOS, and Spanish also has VSO. Some speakers of Spanish may perceive the subject as topicalized in an SVO sentence. Regarding the structure of declaratives in these languages, the discussion on position of adverbs confirms overt movement of the subject and the verb in Italian and French. In these two languages, 'already' is found in postverbal position. Given the arguments in Chomsky (1995a) with regard to adjunction, we assume that adverbs adjoin to X' or TP. In the case of a postverbal adverb, it is obviously

Conclusion

73

not adjoined to TP. If the subject and the inflected verb had not moved overtly, the adverb would not appear in postverbal position, for there would be no projection below V' for it to adjoin to. In French and Italian, then, overt movement of the subject and the verb is confirmed. As for Portuguese, it is clear that movement is also overt, given negation and the floating quantifiers discussed in subsection 2.2. Spanish and Portuguese behave alike regarding negation, and it seems that both the subject and the verb also check features before Spell-out in this language's SVO sentences. We have seen that there is reason for AgrP not to be universally present (Chomsky 1995a). In a structure without AgrP, the subject moves to [Spec, TP] to have its nominative Case checked. Nevertheless, the preverbal subject may be topicalized, as shown by Belletti (1990) for sentences in which probabilmente appears between the subject and the verb. Her analysis may be extended to Portuguese and presumably to Spanish too. We have seen that the structures for declaratives in Spanish, Italian and Brazilian Portuguese show similarities such as the position of the subject and the verb. Preverbal ya/jâ adjoins to T' or Asp' (which goes against Cinque's [1999] universal hierarchy of adverbs, as noted before), postverbal già adjoins to Asp'. Furthermore, a preverbal subject might be topicalized if a sentential adverb appears between the subject and the verb, as argued by Belletti (1990). French, however, behaves differently from the other three languages with regard to position of adverbs: for the most part, it only accepts adverbs in postverbal position. This is true for all kinds of adverbs, including sentential adverbs, which cannot appear in the beginning of a sentence, as in (32) above, where we see that probablement can only surface after the verb. Having to rule out the possibility of adjunction to a projection above the verb, and to one before the subject, I have argued for the existence of an AgrP above TP in French. The subject then moves overtly to [Spec, AgrP] and the verb moves to Agr. Adjunction to AgrP (or Agr') is barred under our assumptions, since the adverb should adjoin to a projection it modifies, and it would not modify agreement. Therefore, the postverbal adverb in French may actually be adjoined to T', or even Asp' as in Spanish or Italian. As for sen-

74

Basic order in declaratives

tential adverbs, they can adjoin to T' in French. Since TP is below AgrP, and the verb raises overtly to Agr, the adverb will still surface in postverbal position. The structure for a sentence such as (32a) should then be: (62)

^AgrP^ Spec Jean

Agr' Agr TP I aime Spec T' I ts probablement Τ' T tv

AspP Spec ts

Asp'

Asp tv

VP Spec ts

V' V I tv

NP A la ling.

As we can see in (62), the subject moves to [Spec, AgrP] through [Spec, TP], where it checks its nominative Case. The verb raises to Agr, leaving traces in Asp and in T. The adverb adjoins to T' (not to Asp'), being a sentential adverb. In the case of an adverb like déjà, it could adjoin to T' or to Asp', depending on which projection is believed to be modified by the adverb. Summarizing the conclusions on basic word order, I have argued that the subject raises to [Spec, TP] before Spell-out in Brazilian Portuguese, Spanish and Italian. In the latter, the (tensed) verb raises

Conclusion

75

to Τ overtly, whereas in Spanish and in BP the verb may land in Asp or in T. In either case (verb moving to Τ or to Asp), SVO is the order that yields. The existence of an AgrP above TP has been posited for French: in this language, the subject moves overtly to [Spec, AgrP] and the verb to Agr. Having looked at the SVO order in all four languages, we can now proceed to investigate those instances when postverbal subjects are possible in declaratives. As mentioned in the introductory chapter, subjects can appear after the verb in BP in constructions that involve unaccusative verbs. Unaccusativity in Brazilian Portuguese is discussed in the next chapter.

Chapter 3 Postverbal subjects in Brazilian Portuguese

1. Introduction In the two previous chapters, we saw that the default word order in Brazilian Portuguese is SVO. However, we do find instances of sentences in which the subject appears after the verb. As mentioned in Chapter 1, unaccusative verbs do allow for both SV and VS orders. An account of that fact is provided in this chapter. Moreover, we will also look at instances of postverbal subjects with unergative verbs, and analyze them as well. We will begin by looking at the Unaccusative Hypothesis and the diagnostics used for unaccusativity, showing examples from Italian and Spanish. We will see that some of the same diagnostics used for Spanish will also apply to Brazilian Portuguese. We will also find that there is only a weak correlation between belonging to a certain semantic class and having the syntactic property of unaccusativity in Brazilian Portuguese. Examples of postverbal subjects in this language are found with unaccusative verbs and in locative inversion constructions - the latter are analyzed in section 4.1, where we see that the prepositional phrase acts like a subject not only in terms of position in the sentence, but also in terms of feature checking: it is argued that locative inversion is possible in Brazilian Portuguese because the PP can check the EPP feature in [Spec, TP]. On the other hand, pro in Brazilian Portuguese is argued to check Case as well as the EPP feature, which will account for the ungrammatically of plain inversion with verbs other than unaccusatives: the subject of this class of verbs does not have to check nominative Case covertly, given the nature of the Case of unaccusatives. The discussion of postverbal subjects with unaccusatives will lead to the conclusion that inherent Case is checked in its merge position. Furthermore, we will see that, though the Case of unaccusatives need not be partitive, it is always inherent. As we will see in section 4.2.4., the gram-

78

Postverbal subjects in Brazilian

Portuguese

maticality of postverbal definite subjects may correlate to the difference between "new" and "old" information in the discourse. After this brief summary of the findings in this chapter, let us proceed and start by discussing the Unaccusative Hypothesis. This discussion is followed by the analysis (divided in various subsections) of the occurrence of postverbal subjects in Brazilian Portuguese.

2. The Unaccusative Hypothesis It has been maintained that the argument of some intransitive verbs is generated in object position and the subject position is nonthematic. This claim is known as the Unaccusative Hypothesis, which was put forth by Perlmutter (1978). Perlmutter's original work was stated in terms of Relational Grammar. Nevertheless, the idea contained in the unaccusative hypothesis has been embraced in the Principles and Paramaters framework as well, given that unaccusative verbs have been observed to display a syntactic behavior that is unlike that of other intransitives: the subject of unaccusatives seems to act more like the object of transitive verbs. A preliminary structure for a VP headed by an unaccusative could be as follows: (i)

A

V

NP

Later, I will assume a structure that incorporates the notion of VP shell, following Larson (1988), in which unaccusatives would differ from other verbs (since they lack that shell). For now, suffice it to say that unaccusatives differ from other verbs because they lack the external argument, as is clear from (1). Burzio's (1986) generalization states that a verb which lacks an external argument fails to assign accusative Case (hence the term unaccusative), and if a verb fails to assign accusative Case, it has no external theta-role. Unaccusative verbs may appear in syntactic constructions where a transitive or an unergative verb would not. In Italian, cliticization of the particle ne and the selection of essere 'to be' as an aux-

The Unaccusative Hypothesis

79

iliary are instances of the unique syntactic behavior of unaccusatives. With transitive verbs, «e-cliticization is only possible with the direct object, whereas unaccusatives allow for cliticization of ne with the postverbal subject. This is shown in the examples below: (2)

a.

Giovanni ne inviterà molti. Giovanni of-them will invite many 'Giovanni will invite many of them' b. *Ne esamineranno il caso molti. Of-them will examine the case many c. Ne arriveranno molti. Of-them will arrive many. 'Many of them will arrive'

Another example of the syntactic behavior of unaccusatives in Italian is selection of essere 'to be' as an auxiliary, which occurs in passive sentences, where the subject is an underlying object. Similarly, unaccusatives also select essere as auxiliary, thus indicating that their subject originates in object position. The following examples illustrate selection of essere as an auxiliary: (3)

a. b.

Giovanni è stato invitato. Giovanni is been invited Giovanni è arrivato. Giovanni is arrived

(3) shows that essere is the auxiliary used for passives, as in (3a), and unaccusatives, as in (3b). We then see that both tests presented above can be used as differentiation between unaccusative and unergative verbs. jVe-cliticization and the use of essere (instead of avere) as an auxiliary with an unergative verb yields an ungrammatical result, as shown in (4) with telefonare: (4)

a. *Ne telefoneranno molti. Of-them will telephone many b. * Giovanni è telefonato. Giovanni is telephoned

80

Postverbal subjects in Brazilian

Portuguese

Given these observations, verbs that behave like arrivare 'arrive' have been classified as unaccusatives. They differ from unergatives in several aspects, and their behavior may vary from one language to another. For example, in Italian (as we saw) and in French, use of a certain auxiliary verb may be an indication that a verb is an unaccusative. In languages such as Spanish, which has no auxiliary selection in perfective tenses, unaccusatives must be characterized by means of other diagnostics. Spanish unaccusatives are the sbuject of the next subsection.

2.1. Unaccusatives in Spanish For languages such as Spanish, which uses only one auxiliary, other tests have been used to establish whether a verb is an unaccusative. One of these tests involves the use of the past participle of the verb in question, in a construction referred to in González (1988) as the Past Participle as Adjective (PPA) construction. González gives some examples of this construction, such as (5): (5)

a.

La bruja hechizó a la princesa. The witch enchanted the princess b. La princesa hechizada durmió profundamente. The enchanted princess slept soundly c. *La bruja hechizada odia a la princesa. The enchanted witch hates the princess (González 1988: 27)

In (5), we see that the past participle of a transitive verb such as hechizar 'to enchant' can be used as an adjective, as in (5b). However, the noun that this adjective modifies is the object of the verb, as in (5a). (5c) shows that the subject of the verb cannot be modified by the past participle as adjective28. Since the subject of an unaccusative is argued to be generated in the position of an object, the PPA construction should be possible with unaccusative verbs. And it is, as exemplified below:

The Unaccusative Hypothesis

(6)

81

a.

La mujer desmayada fue llevada al hospital. The fainted woman was taken to the hospital. b. Las joyas desaparecidas son de un gran valor. The disappeared jewels have a great value. La bomba explotada en la embajada hizo noticia. The exploded bomb at the embassy was big news. (González 1988: 30)

The Spanish verbs in (6) (>desmayar 'to faint', desaparecer 'to disappear', and explotar 'to explode') are verbs that only have one argument and can occur with the PPA construction. The same is not true of other verbs that do not take objects: (7)

a. *El hombre trabajado está muy cansado. The worked man is very tired b. *El payaso reído apareció en el circo. The laughed clown appeared in the circus c. *La niña jugada se fue a dormir. The played girl went to sleep (González 1988:30)

As can be seen in (7), unergative verbs do not exhibit the same behavior as unaccusatives in the PPA construction: if used as an adjective, the past participle of an unergative verb will yield an ungrammatical result. Another test for Spanish that distinguishes unaccusatives from unergatives is the participial absolute (PA) construction. Based on a transitive clause such as (8a), we can have the PA construction in (8b): (8)

a.

El poeta leyó la carta. 'The poet read the letter' b. Leída la carta, el poeta procedió a suicidarse. Read the letter, the poet went on to commit suicide c. * Leído el poeta, procedió a suicidarse. Read the poet, (he) went on to commit suicide

82

Postverbal subjects in Brazilian

Portuguese

For a transitive verb such as leer 'to read', the PA construction renders grammatical results if the object of the verb follows it, as in (8b). If the subject accompanies the verb, the result is ungrammatical, as seen in (8c). The PA is relevant for the Unnacusative Hyposthesis because unaccusatives can occur in that construction, whereas unergatives cannot: (9)

a

Desaparecidas las joyas, llamaron a la policía. Disappeared the jewels, (they) called the police b. * Hablado el profesor, los alumnos se retiraron. Spoken the teacher, the students left

The PA is grammatical in (9a) because this example contains an unaccusative verb. Since the verb in (9b) is unergative, use of the participial absolute results in ungrammaticality. A third possible test for unaccusatives in Spanish is the use of the impersonal se: this class of verbs does not occur with impersonal se, as stated in Demonte (1989): (10)

a. * Se llega tarde alas fiestas. 'People arrive late at parties' b. Se trabaja mucho aquí. 'People work a lot here' c. Se compra libros usados en esta librería. 'People buy used books in this bookstore'

The examples in (10) illustrate the use of the impersonal se with an unergative verb, as in (10b), and with a transitive verb, seen in (10c). However, the use of impersonal se with an unaccusative yields an ungrammatical result, as attested in (10a). In Spanish, the expression quedar por 'to remain, to be left' + infinitive differentiates between transitive objects and unaccusative subjects, on the one hand, from transitive subjects and unergative subjects, on the other. With the former (the objects), the expression is grammatical, whereas with the latter (the transitive and unergative subjects) it is not. In (11) we find examples of the four constructions referred to above:

The Unaccusative Hypothesis

(11)

83

a.

Quedan tres cartas por escribir. 'There are three letters left to write' b. Quedan tres profesores por llegar. 'There are three professors left to arrive' c. * Quedan tres profesores por escribir las cartas. There are three professors left to write the letters d. * Quedan tres profesores por trabajar. There are three professors left to work

As is clear from (11), quedar por does not occur either with transitive subjects (as in [11c]), nor unergative subjects (exemplified in [lid]). However, the expression is grammatical in (1 la) and (b), where it appears with a transitive object and an unaccusative subject, respectively. The patterning here is therefore the same as what is seen with the PA and the PPA tests. Thus we have several tests that allow us to identify an unaccusative verb in Spanish.

Summary In Italian and in Spanish, as we have seen, there seem to be clear diagnostics for unaccusativity, yielding quite accurate results. In Italian, these tests show certain syntactic constructions that only occur with unaccusatives, namely cliticization of the particle ne and the use of the verb essere as an auxiliary (as opposed to avere, which is the auxiliary used for transitives and unergatives). In Spanish, we have listed four tests for unaccusativity. The first one was the Past Participle as Adjective (PPA) construction, exemplified in (6), which shows that the past participle is only used as an adjective for objects or for subjects of unaccusative verbs. Then, we looked at the participial absolute (PA) construction, also used only with objects of transitive verbs or subjects of unaccusatives, as seen in (8) and (9). The use of impersonal se, which occurs only with transitives and unergatives, can also be used as a diagnostics for unaccusativity: if a verb appears in that construction, it is not unaccusative. Finally, we saw that the expression quedar por + infinitive, like the Past Participial as Adjec-

84

Postverbal subjects in Brazilian Portuguese

tive (PPA) and the Participial Absolute (PA) constructions, is also possible only with transitive objects and unaccusative subjects. The same accuracy in diagnosing whether a verb is unaccusative may not be found in all languages. In German, for example, the unaccusativity tests commonly used may not be as reliable as in Italian and Spanish, as discussed in Seibert (1993), who notes some problems with the diagnostics for unaccusativity. However, although the syntactic tests for German unaccusativity may be murky, that does not mean that the Unnacusative Hypothesis does not hold. There may be tests for unaccusativity that are purely syntactic, such as auxiliary selection or «e-cliticization; others, however, may indicate the existence of some semantic constraints, which means that unaccusativity might be a necessary but not sufficient condition. At this point, it is necessary to investigate the behavior of Brazilian Portuguese with regard to the hypothesis. This is the topic of the next section.

3. Unaccusatives in Brazilian Portuguese As we have seen in chapter 1, postverbal subjects in Brazilian Portuguese declaratives are possible if the sentence contains an unaccusative verb29. This is again exemplified below: (12)

a. A água congelou. The water froze b. Congelou a água. Froze the water

In (12) we have an example of possible postverbal subject with an unaccusative verb, congelar 'freeze'. As seen before, postverbal subjects in Brazilian Portuguese are not generally allowed with transitives and unergatives, as shown below: (13)

a. A Bia leu a revista. Bia read the magazine b. *Leu a revista a Bia. Read the magazine Bia

Unaccusatives in Brazilian Portuguese

85

c. * Leu a Bia a revista. Read Bia the magazine (14)

a

A criança brinca em casa. The child plays at home b. * Brinca em casa a criança. Plays at home the child c. * Brinca a criança em casa. Plays the child at home

Examples (13) and (14) indicate that a postverbal subject in Brazilian Portuguese does not occur with transitive verbs (as shown by the ungrammaticality of [13b, c]), nor with unergatives (cf. [14b] and [c]) when the verb is in sentence-initial position30. However, we must make sure that the verbs that allow for inversion are indeed unaccusatives, for there might be inversion with other classes of verbs. The two most common tests for unaccusativity, as mentioned before, are selection of auxiliary and attributive use of the past participle. Portuguese, like Spanish (and unlike Italian and French), uses only one auxiliary, and thus needs other tests for diagnosing unaccusativity. Since Spanish is a closely related language, it seems reasonable to see whether (some of) the tests that are used for Spanish also apply in Portuguese. Although Portuguese has only one auxiliary, another common test for unaccusativity, the adjectival use of the past participle (PPA constructions), can still be used. Taking examples used by González (1988), we have the following results for Brazilian Portuguese: (15)

a.

A bruxa enfeitiçou a princesa. The witch enchanted the princess b. A princesa enfeitiçada dormiu profundamente. The enchanted princess slept profoundly c. *A bruxa enfeitiçada odeia a princesa.31 The enchanted witch hates the princess

The sentences in (15) show the same results as those found for Spanish (cf. [5] above). If a verb that is not transitive allows for its subject

86

Postverbal subjects in Brazilian Portuguese

to be found in a construction such as that in (15b), we have evidence that that subject originated in object position (as do the objects of transitive verbs). Those verbs would be unaccusatives, based on that test. In other words, unaccusative verbs, but not unergative verbs, allow for the PPA construction in Portuguese. Examples of unaccusatives in Spanish given by González have similar results when translated to Portuguese: (16)

a. As jóias desaparecidas säo de grande valor. The disappeared jewels have great value b. O crime ocorrido em janeiro ainda näo foi solucionado. The crime (that) occurred in January has not yet been solved c. Os muros rachados cairam com o terremoto. The cracked walls fell down with the earthquake d. O metal expandido funciona como termostato. The expanded metal works as a thermostat

As we can see in (16), desaparecer 'disappear', ocorrer 'occur', rachar 'crack', expandir 'expand' are verbs that pass the PPA test for unaccusativity, and could thus be regarded as unaccusative verbs in Portuguese. Note that here, rachar 'crack' and expandir 'expand' are naturally being used in their intransitive sense. These verbs have a transitive counterpart, exemplified below: (17)

Eu rachei o prato. Ί cracked the dish'

(18)

A presidente da companhia expandiu o mercado desde que assumiu. 'The president of the company has expanded the market since she took office'

In (17) and (18) we have the transitive counterparts of rachar and expandir. Note also that when these verbs are used intransitively they do not have se, as they would in Spanish:

Unaccusatives in Brazilian Portuguese

(19)

O prato rachou. The dish cracked

(20)

O mercado expandiu desde que a presidente assumiu. The market has expanded since the president took office

87

Since (19) and (20) do not exhibit se, there is little motivation for a syntactic derivation, as opposed to a lexical derivation, from transitive to intransitive rachar - otherwise, o prato in (19), for example, would be an underlying object of a transitive verb, with a derivation similar to that of a passive sentence, in particular, a sentence with se. As in Spanish, the PPA construction in Portuguese cannot be used with unergative verbs. Some examples are given in (21): (21)

a. * O home m trabalhado está muito cansado. The worked man is very tired b. *0palhaço rido aparecen no circo. The laughed clown appeared at the circus c. *A menina brincada foi dormir. The played girl went to sleep d. * O rapaz dançado foi descansar. The danced young man went to rest

The ungrammaticality of the examples in (21) should indicate that those verbs are not unaccusatives, for it is not possible to use their past participle as an adjective. Another test that might work for Portuguese is the participial absolute (PA) construction, which is grammatical with unaccusatives but ungrammatical with unergatives. As explained above, the PA construction is grammatical with transitive verbs when they are followed by their objects, as exemplified for Spanish in (8) above. The same results are found in Portuguese: (22)

a.

O poeta leu a carta. 'The poet read the letter' b. Lida a carta, o poeta se suicidou. 'Having read the letter, the poet committed suicide'

88

Postverbal subjects in Brazilian

Portuguese

c. *Lido o poeta, ele se suicidou. Read the poet, he committed suicide An example of the PA with an unaccusative verb in Portuguese is given below: (23)

Morios os soldados inimigos, a batalha foi considerada vencida. Dead the enemy soldiers, the battle was considered won

Morrer 'die', as an unaccusative, admits the use of the PA construction, as shown in (23). The same result does not obtain when the verb used is a 'regular' intranstive: (24)

*Trabalhados os professor es, a escola permanecen aberta. Worked the teachers, the school remained open

The contrast between (23) and (24) should constitute evidence that only certain intransitive verbs fall into the unaccusative class, whereas other intransitives do not. The latter are classified as unergatives. Several verbs in Portuguese admit the use of the PA. These are, roughly, the same verbs that admit the PPA construction as well. Here I list several verbs that take both the PPA and the PA construction, and could therefore be regarded as unaccusatives in Portuguese. The semantic classification below is based on Levin and Rappaport Hovav's (1995) verb classes. (25)

Verbs that take the PPA and the PA constructions (unaccusatives): (a) Verbs of change of state: aborrecer-se 'get angry', acordar 'wake up', afundar 'sink', aumentar 'increase', brotar 'sprout' , começar 'begin', congelar 'freeze', crescer 'grow', cristalizar 'cristalize', derreter 'melt', desenvolver 'develop', desmaiar 'faint', deteriorar 'deteriorate', diminuir 'decrease', enferrujar 'rust', enfurecer-se 'get mad', enrugar 'wrinkle', entristecerse 'get sad', envelhecer

Unaccusatives in Brazilian Portuguese

(b) (c)

(d)

(e)

(f)

(g)

89

'grow old', enxugar-se 'dry', escorrer 'drain', escurecer 'become dark', evaporar 'evaporate', evoluir 'evolve', expandir 'expand', explodir ' e x p l o d f e r v e r 'boil', florescer 'flourish', melhorar 'improve', molhar-se 'get wet', mudar 'change', nascer 'be born', piorar 'worsen', quebrar 'break', rachar 'crack', secar 'dry', terminar 'end', virar 'become'. Verbs of emission: feder 'stink',pingar 'drip', suar 'sweat', vazar 'leak'. Verbs of inherently directed motion: avançar 'advance', cair 'fall', chegar 'arrive', descer 'go down', entrar 'enter', escapar 'escape', fugir 'escape', ir 'go', parar 'stop', passar 'pass, go by', retornar 'return', sair 'leave', subir 'go up', vir 'come', virar 'turn around', voltar 'come/go back'. Verbs of manner of motion: derrapar 'slide', desfilar 'parade', girar 'spin', rolar 'roll', quicar 'bounce', viajar 'travel', virar 'turn' Verbs of existence, appearance and occurrence: acontecer 'happen', aparecer 'appear', continuar 'continue', estrear 'open' (a play, etc), materializarse 'materialize', ocorrer 'occur', reunir-se 'meet', suceder 'happen', surgir 'appear', viver 'live'. Verbs of disappearance: desaparecer 'disappear', expirar 'expire', morrer 'die'. Verbs of spatial configuration: ajoelhar-se 'kneel', deitar-se 'lie down', descansar 'rest', encostar 'lean', equilibrarse 'balance', levantarse 'get up', sentarse 'sit down'

The list in (25) (which is by no means exhaustive) includes verbs that pass two tests for unaccusativity. However, there are verbs that belong to some of those classes but do not have the syntactic characteristics of unaccusative verbs. A few examples are: voar 'fly', galopar 'gallop', pular 'jump', which belong to the verbs of manner of mo-

90

Postverbal subjects in Brazilian Portuguese

tion class. These and several other verbs that fall into the semantic categories listed in (25) cannot be classified as unaccusatives (some of these are included in the unergative list below). Thus we see that, in Brazilian Portuguese, there seems to be only a weak correlation between belonging to a certain semantic class and having the syntactic property of unaccusativity. Unergative verbs should not occur with the PPA and the PA constructions, and this is indeed the case, as shown in examples (21) and (24) above. Below are several intransitive verbs (unergatives) that do not pass the tests for unaccusativity: (26)

Unergative verbs: andar 'walk', bocejar 'yawn', brilhar 'shine, twinkle', brigar 'fight', brincar 'play', caminhar 'walk', cantar 'sing', chorar 'cry', conversar 'talk', correr 'run', discutir 'argue', dormir 'sleep', engatinhar 'crawl (children)', escorregar 'slide', espirrar 'sneeze', esquiar 'ski', flutuar 'float', gargalhar 'laugh loudly', galopar 'gallop', lutar 'fight', meditar 'meditate', mentir 'lie', mergulhar 'dive', nadar 'swim', passear 'walk leisurely',patinar 'skate',pular 'jump', rebolar 'sway', rir 'laugh', sorrir 'smile', tossir 'cough', trabalhar 'work', voar 'fly', vomitar 'vomit'.

The list in (26) (also not exhaustive) comprises verbs that do not allow for the adjectival use of the past participle, nor for the participial absolute construction. Thus, these are verbs that, according to the UH, do not fall into the unaccusative category; they do not pass the diagnostics discussed above. Note that some of the verbs in the unergative list have cognate or very close semantic objects, as in cantar urna cançao 'sing a song', rir um riso gostoso 'laugh a nice laugh', or chorar rios de lágrimas 'cry rivers of tears'. This fact is not unexpected: Hale and Keyser (1993) develop the idea that unergatives are actually covert transitives. They assume that unergative verbs have an initial lexical structure of the simple transitive type, and that they are formed by means of incorporation (as in Baker 1988). Hale and Keyser's argument accounts for the fact that we find (semantically close) objects for some of the unergative verbs in Brazilian Portuguese.

Unaccusatives in Brazilian Portuguese

91

Here I have discussed only two diagnostics for unaccusativity in Portuguese. A third possible test also parallels one used for Spanish: the use of the impersonal se. As in Spanish, the impersonal se only seems to occur with certain verbs, not with others. It does appear with transitives and unergatives, as illustrated below: (27)

a.

Vendem-se artigos importados . 'We sell imported goods' b. Se trabalha muito nesta fábrica. 'People work a lot at this factory"

In (27a) we have an example of the impersonal se used with a transitive verb, vender 'sell'. The sentence in (27b) illustrates the use of the impersonal se with an unergative verb, trabalhar 'work'. In Spanish, the impersonal se does not occur with unaccusatives, according to Demonte (1989). This seems to be the case for Portuguese too, since use of this particle with verbs from (25) renders ungrammatical results: (28)

a. *Sempre se vai às festas do presidente '(One) Always goes to the president's parties' b. *Se chega tarde aqui. '(One) Arrives late here'

As seems clear from (28), verbs included in the list in (25) cannot occur with the impersonal se. Thus, the use of this particle may be considered another test for unaccusativity.

Summary In section 3 we have seen that some of the tests that can be used as diagnostics for unaccusativity in Spanish can also be used in Brazilian Portuguese. Examples (15) and (16) show that the use of the PPA construction yields the same results as in Spanish, i.e., the construction only occurs with objects of transitive verbs or subjects of unaccusatives. The second test used for Spanish, the PA construction, also

92

Postverbal subjects in Brazilian Portuguese

works for Portuguese, as attested in examples (22) and (23). Based on the results of those two tests, two lists of examples of verbs in Brazilian Portuguese (neither one exhaustive) were presented: the one in (25), with verbs that take the PPA and the PA construction, and can be considered unaccusative; and another in (26), with unergative verbs. The last test used for unaccusativity in Brazilian Portuguese also parallels one used for Spanish: the impersonal se, a construction that does not appear with unaccusative verbs. There is, however, one more aspect which appears to be related to unaccusativity, as mentioned above: subject-verb inversion. Having looked at verbs that can be considered unaccusatives in Brazilian Portuguese, I will turn to inversion next.

4. Verb-subject order in Brazilian Portuguese In Brazilian Portuguese, the presence of postverbal subjects is restricted to certain environments. It does not happen necessarily in interrogatives; as a matter of fact, a postverbal subject in a question may yield an ungrammatical result, depending on what verb is used. With transitive verbs, for example, the subject must appear before the verb, be it in declaratives or in interrogatives, as we have seen in Chapter 1 : (29)

a.

O que a Ana viu? What Ana saw? 'What did Ana see?' b. A Ana viu um disco voador. 'Ana saw a flying saucer'

(30)

a. * O que viu a Ana?33 What saw Ana? b. * Viu um disco voador a Ana. Saw a flying saucer Ana c. * Viu urn disco voador a Ana? Saw a flying saucer Ana? 'Did Ana see a flying saucer?'

Verb-subject order in BP

93

In (29), both examples contain the order SVO: (29a) is a w/z-question, which is fully grammatical, and (29b) could be the answer to (29a). The order VS will result in ungrammatically, as illustrated in (30). (30a) shows that a transitive verb cannot occur before the subject even in a w/z-question; in declaratives, VS is equally disallowed, as seen in (30b); (30c) illustrates the ungrammaticality of VS in yes-no questions. In sum, a transitive verb cannot precede the subject in Brazilian Portuguese, in any circumstances. It also seems to not occur with unergative verbs, judging from the examples below (but see endnote 3 and section 4.1 later): (31)

a. * Quando trabalha a Isa1? When works Isa? b. * Trabalha a Isa todo dia. Works Isa every day c. * Trabalha todo dia a Isa. Works every day Isa

(31) contains examples with the unergative verb trabalhar 'to work'. As with the transitive verb illustrated in (30), the examples with an unergative verb are also ungrammatical if a postverbal subject is used, be it in a w/z-question, as in (31a), or in declaratives, as in (31b) and (c). Notice that (31b) contains an adverbial in the end of the sentence. The position of the adverbial is different in (31c), where it appears between the verb and the subject. In either instance, the result is ungrammatical. The order VS would not be rescued even if an adverbial (or any other element) were not present, as seen below: (32)

a.

Quem nada aquil 'Who swims here?' b. Os atletas (nadam aqui). 'The athletes do' c. * Nadam os atletas. Swim the athletes

The examples in (32) seem to indicate that no VS is acceptable with a verb such as nadar, for the question in (32a), the only possible an-

94

Postverbal subjects in Brazilian Portuguese

swer contains a preverbal subject, if the verb even appears in the answer (the parentheses indicate the optionality of the verb and the adverbial in the answer). Thus we see that with certain verbs inversion is impossible, as with transitive verbs, independent of the type of clause (declaratives, Wz-questions, yes-no questions). Nevertheless, VS is indeed possible with unergatives, but in restricted environments: the locative inversion construction allows for postverbal subjects: (33)

Nesta piscina nadam os melhores atletas. In this pool swim the best athletes

Although VS does not occur with an unergative verb that occupies a sentence-initial slot, it is possible if there is a (many times locative) PP before the verb. This construction is known as locative inversion, and is also present in English (the locative inversion construction also occurs with unaccusatives, as we will see). VS then is very restricted with unergatives, and not possible with transitives. However, postverbal subjects may indeed appear in Brazilian Portuguese with certain verbs, some of which are usually considered core examples of unaccusative verbs. They can be found with verbs such as chegar 'to arrive', nascer 'to be born', or acontecer 'to happen': (34)

a. a'. b. b'. c. c '.

Chegou o Ivo. Arrived Ivo O Ivo chegou. Ivo arrived Nasceu o filho da Cida. Was born the son of Cida O filho da Cida nasceu. The son of Cida was born Aconteceu urna tragèdia! Happened a tragedy Urna tragèdia aconteceu ! A tragedy happened

Verb-subject order in BP

95

In (34), examples (a'), (b') and (e') contain a preverbal subject. Examples (a), (b) and (c) show that VS is just as acceptable in those instances. The order VS is found with certain verbs, as the ones included in the list in (25), which contains unaccusative verbs. So, besides passing two tests for unaccusativity, those verbs have one more aspect in common: they allow for postverbal subjects. Could the possibility for VS order then be one more test for unaccusativity in Brazilian Portuguese? This question is not easily answered. While all the verbs in (25) allow for postverbal subjects, not all verbs that allow postverbal subjects are included in that list. For example, the verb ficar 'to stay' does not pass the tests for unaccusativity, and yet it does admit the order VS: (35)

a. *Os alunos ficados aprenderäo inglés. The students stayed will learn English b. * Ficados os alunos, continuamos a aula. Stayed the students, we continued the class

(36)

a.

Quem ficou na sala? Who stayed in the living room? b. Ficaram as meninas. Stayed the girls

In (35) we see that ficar does not yield grammatical results when used with the PPA construction (35a), as well as with the PA construction (35b). However, as is shown in (36b), ficar can indeed appear before the subject. Other such verbs are ligar 'to telephone' and atender 'to answer the phone'. These examples will be discussed later in this chapter. As mentioned above, unergatives may appear with postverbal subjects in Brazilian Portuguese. In the subsections that follow, we will look at postverbal subjects in locative inversion, with unaccusative verbs, and with a few verbs that are not, a priori, classified as unaccusatives.

96

Postverbal subjects in Brazilian Portuguese

4.1. Locative inversion The locative inversion construction is found with unergative and unaccusative verbs. In this construction, the subject may appear postverbally if a PP appears before the verb. According to Levin and Rappaport Hovav (1995), in English locative inversion is possible with intransitive verbs, though not all intransitive verbs are found in this construction. Some examples from Hoekstra and Mulder are given below: (37)

a. Into the room entered a man. b. Down the street rolled the baby carriage. c. Round and round spins the fateful wheel. (Hoekstra and Mulder 1990:28 [61])

The examples above illustrate the locative inversion construction with unaccusative verbs. This construction is characterized, as mentioned above, by the word order PP-V-NP. Furthermore, the PP is typically a locative or directional PP (hence the name of the construction). Other examples of locative inversion are given in (38): (38)

a. b. c.

Out of the barn ran a horse. Into the room walked a man. Ouf of the house strolled my mother's best friend. (Hoekstra and Mulder 1990:31 [70])

In (38) we have examples of locative inversion with unergative verbs. Thus we see that both unaccusative and unergative verbs can appear in a locative inversion, though (at least in English) not all such verbs can. (For a list of English verbs found in this type of construction, see Levin and Rappaport Hovav 1995). The locative inversion construction is not found in English only: we can find instances of sentences with the word order PP-VNP in Brazilian Portuguese as well: (39)

a.

Do Nordeste vêm os peöes de obra. From the Northeast come the construction workers

Verb-subject order in BP

b.

97

Nesse edifìcio trabalham cerca de 150 pessoas. In this building work some 150 people.

The sentences in (39) illustrate the locative inversion construction in Brazilian Portuguese. Example (39a) contains an unaccusative verb, vir 'come', whereas in (39b) we find an unergative verb, trabalhar 'work'. A few more examples of this construction in Brazilian Portuguese are given in (40), in order to better illustrate its occurrence: (40)

a.

Nessa casa mora o presidente. In this house lives the president b. Naquele quarto dormem as crianças. In that room sleep the children c. De muitos países vieram os imigrantes. From many countries came the immigrants d. Para o Rio fogem os criminosos dos filmes de Hollywood. To Rio escape the criminals in Hollywood movies.

(40a) and (b) demonstrate again that unergative verbs can be found in the locative inversion construction, as well as unaccusatives, shown in (40c) and (d). Levin and Rappaport Hovav (1995) point in the direction of a possible account of this construction: they claim, with Bresnan (1993), that at some point of the derivation the preverbal PP has the status of a subject. According to Levin and Rappaport Hovav, the PP moves from a VP-internal position to subject position, which can be interpreted as [Spec, TP] in Minimalist terms. Their account is related to discourse function, and they do not include an analysis of Case in their proposal; as a matter of fact, they "note that Case assignment is also a problem facing any analysis of the locative inversion construction" (Levin and Rappaport Hovav 1995: 268). Collins (1997) provides another discussion of the locative inversion construction, also arguing that the PP moves to [Spec, TP] from within the VP, satisfying Last Resort and Minimality. However, Collins assumes that locative inversion is restricted to unaccusative verbs, which "will simplify the account a great deal" (Collins 1997: 27). As is clear from Levin and Rappaport Hovav's (1995) work, that construction does occur

98

Postverbal subjects in Brazilian Portuguese

with several unergative verbs in English, and also in Brazilian Portuguese. An account of the locative construction is provided by Hoekstra and Mulder (1990), whose solution involves assigning (or checking, in Minimalism) nominative Case to the PP in [Spec, IP] (or [Spec, TP] in the structure assumed here). They also propose that the "PP originates in the predicative part of a small clause-complement, and provides nominative Case to its subject under head-spec agreement from its base position" (Hoekstra and Mulder 1990: 29). For this proposal to work, first they must show that the PP forms a small clause with the postverbal NP. However, as pointed out by Levin and Rappaport Hovav (1995), the small clause analysis would also require that the verb in this construction not have an external argument. Thus, only unaccusatives would be found in locative inversion examples. To account for the presence of unergatives in this construction, Hoekstra and Mulder (1990) argue that an unergative verb may either take an individual as external argument, or a propositional internal one. In this case, the two instantiations of the verb would result from projection of distinct sets of lexical requirements. According to them, certain predicates vary such that they may take arguments of different types. If this is so, then an unergative verb might pose as an unaccusative, which would be the consequence of this shift in the meaning of the verb. In view of previous accounts of this construction, a different explanation could be proposed, one that could account for feature checking without having to resort to verb meaning shift. That does not constitute an impossible task, given the assumptions within the current framework. Here, as Levin and Rappaport Hovav (1995) and Hoekstra and Mulder (1990) do, I also assume that the PP in a locative inversion sentence is in "subject position", i.e., it is located in [Spec, TP]. According to Bresnan (1993), as has just been mentioned, the preverbal PP is a subject at some level of linguistic representation: she argues that the PP is a subject at LFG's f-structure, which in Minimalism may be translated as prior to Spell-out. Evidence can be found in Brazilian Portuguese that allows us to claim, as Bresnan does for English,

Verb-subject order in BP

99

that the preverbal PP raises from a VP-internal position to subject position (or [Spec, TP]): (41)

a.

Neste quarto parecem ter dormido as crianças. In this room appear to have slept the children b. As crianças ¡parecem [t¡ ter dormido t¡ neste quarto]. c. * Neste quarto ¡parecem [as crianças/. ter dormido tj t j .

In (41) we can see the interaction of locative inversion with raising. When the locative inversion construction is embedded under a subject-raising verb, the PP moves up and takes the place of the surface subject of the matrix verb. This is shown in (41a). (41b) shows that the NP subject of the non-inverted counterpart of (41a) can also raise. However, elements that are not subjects normally do not move up: as shown in (41c), raising of the PP yields an ungrammatical result when the embedded verb has a subject. Thus it looks as if the PP is indeed acting like a subject in the locative inversion construction. Further evidence of such behavior comes from sentences with factive verbs, which do not allow for extraction of the subject. This is illustrated in (42): (42)

a. *Quem voce lamenta que ganhou o premio? Who do you regret that won the prize? b. *Em que quarto voce lamenta que dormiram os meninosi In which room do you regret that slept the boys? c. Em que quarto voce lamenta que os meninos dormiram? In which room do you regret that the boys slept d. Em que quarto voce acha que dormiram os meninos? In which room do you think that slept the boys? e. Quem voce acha que ganhou o premio? Who do you think that won the prize?

The sentences found in (42a) demonstrate that the subject of a factive verb {lamentar 'regret') cannot be extracted. The same result obtains with a PP in place of the surface subject, as in (42b), with (42c) showing that it is not extraction of the PP per se which causes ungrammaticality. However, with a verb that is not factive, as in (42d),

100

Postverbal subjects in Brazilian

Portuguese

the sentence-initial PP yields a grammatical sentence, as does the preverbal wA-subject in (42e). Therefore, the examples in (42) also suggest that the PP acts like a subject, and occupies the [Spec, TP] position at Spell-out. Since it is in [Spec, TP] the PP will check the EPP feature34, which needs to be checked prior to Spell-out. This feature is divorced from Case, according to Chomsky (1995b). If the PP can check the EPP feature in the locative inversion construction, that means that the subject cannot do it. Then, the subject can check Case covertly: with the EPP feature checked overtly, Case can wait until LF to be checked (and by Procrastinate it would have to). Since the subject agrees with the verb in the locative inversion construction, the Case it checks (covertly) must be nominative. The locative inversion construction, as pointed out by Levin and Rappaport Hovav (1995) is possible with certain unaccusative and certain unergative verbs, though not all of them. But these verbs have in common the fact that they do not have an object in the usual sense (since unergatives might have an incorporated object, according to Hale and Keyser 1993). According to Chomsky (1995b), at least one argument has to move in the overt component of the syntax: covert Case checking of both arguments is barred. Chomsky explains that both arguments cannot raise covertly because of the c-command relationship at the LF interface. He starts his discussion with a supposition that a language has covert raising of both arguments, which then means that both FF(Subj) and FF(Obj) raise covertly to TP. After the necessary movements apply, Chomsky arrives at the structure below: (43)

Τ FF(Obj)

Τ

FF(Subj) Vb

Τ Τ

Verb-subject order in BP

101

Having arrived at (43), Chomsky continues the discussion, supposing that output conditions at the LF interface rule out (43) "under the rather natural requirement that FF(Subj) must c-command FF(Obj) in T o m a x i f b o t h gj.g p r e s e n t » (Chomsky 1995b: 3 6 2 ) Then, the object would have to adjoin to T 0max before the subject, which, he argues, is impossible because the subject is closer to Τ than the object if both remain in situ. Thus, at least one of the arguments must raise overtly if the expression is to converge35. The discussion on one argument having to move overtly does not affect verbs that have only one argument, as is the case of unaccusatives and unergatives. Since it is the only argument of the verb, the subject of unaccusatives and unergatives should in principle be able to move covertly. To show this formally, we follow Chomsky (1995b) and assume a Larsonian shell, which is applicable to unergatives 36 , where ν is a light verb to which V raises overtly (Chomsky 1995b, Larson 1988):

In a locative inversion construction, the PP checks the EPP feature, while the Case feature of the subject is checked at LF. Thus, both unaccusatives and unergatives can be present in a locative inversion construction. The restriction that exists regarding transitives (one argument has to move overtly) does not apply to the other two classes of verbs (i.e., these verbs have only one argument, which can in fact move after Spell-out).

102

Postverbal subjects in Brazilian Portuguese

Summary The occurrence of postverbal subjects in Brazilian Portuguese is limited to unaccusative verbs and the locative inversion construction, as shown in examples (33) (locative inversion) and (34) (unaccusatives). It is true that postverbal subjects can appear with a few verbs that do not pass the tests for unaccusativity, as exemplified in (36) with the verb ficar; it will be argued later that this verb might be considered unaccusative, and that other verbs that allow for postverbal subjects may be going through some structural change (see discussion in 4.3.). The discussion of the locative inversion construction leads to the conclusion that the PP in the construction occupies the [Spec, TP] position, and therefore must check the EPP feature. Hoekstra and Mulder (1990) argue that the PP also assigns (checks) nominative Case. However, this proposal faces problems, since the PP needs to form a small clause with the postverbal NP, which in turn means that the verb can only have one argument, limiting the occurrence of locative inversion to unaccusatives. They account for the presence of unergatives in the construction by proposing that unergatives may either take an individual as external argument, or a propositional internal one. This would mean that there would be two projections of sets of lexical requirements for one verb. Given these complications, we need a different account of the locative inversion construction. This account involves the checking of nominative Case: while the PP checks the EPP feature in [Spec, TP], it does not check nominative Case, which is left for the subject NP to check after Spell-out. We have seen that the ban on covert movement of both arguments of the verb prevents transitives from appearing in a locative inversion construction, since one of the arguments (either the subject or the object) would have to raise before Spell-out. This discussion does not apply to unergatives and unaccusatives, whose subject, being the only argument of the verb, can in principle move covertly. In the next subsection, we compare the locative PP and pro in Brazilian Portuguese in terms of syntactic features, establishing a fundamental difference between them.

Verb-subject order in BP

103

4.1.1. The locative PP vs. pro in Brazilian Portuguese Throughout this work, we have seen that unaccusatives can indeed occur with postverbal subjects, even in the absence of a preposed PP. We will see later that their occurrence with pro can be compared to the unaccusative constructions in English where there appears in sentence initial position: (45)

a. b.

Vieram muitos alunos à aula. Came many students to class There came many students to class.

In (45) we see that the unaccusative construction in Brazilian Portuguese and in English is very similar, and that both may exhibit a postverbal subject, with the presence of pro in Brazilian Portuguese and the expletive there in English. Our account of Brazilian Portuguese (45a) will involve a comparison between Brazilian Portuguese pro and the English existentials, and the difference between structural and inherent Case. For now, however, suffice it to say that whatever allows for (45 a) is different from the mechanism underlying the locative inversion construction: with this construction, both unaccusatives and unergatives are possible, but in (45) only an unaccusative can render the examples grammatical. The presence of an unergative, as seen before, would yield an ungrammatical result. Inversion with unaccusatives (without a preposed PP) is thus not the same phenomenon as the locative inversion construction. Still, the subject will not check the EPP feature in inversion with unaccusatives (nor does it have to check nominative Case - as we will see later, unaccusatives may check inherent Case). Thus, in those instances of inversion where there is no preposed PP, pro must suffice to check the EPP feature. With instances of locative inversion, however, the account is the same for examples with unaccusatives and with unergatives, since it is the PP that checks the EPP feature. In principle, the subject of unergative verbs should be able to check Case covertly too, since pro can check the EPP feature as just mentioned (and the EPP is the only feature that actually drives movement). However, as we have seen and the examples below dem-

104

Postverbal subjects in Brazilian

Portuguese

onstrate again, inversion with unergatives only occurs if there is a preposed PP: (46)

a. *Esquiaram os turistas ontem. Skied the tourists yesterday b. No Chile esquiam muitos turistas. In Chile ski many tourists

Given that postverbal subjects are ungrammatical with unergative verbs in the absence of a preposed PP, as (46) shows, pro in Brazilian Portuguese must check not only the EPP feature, but also (obligatorily) nominative Case. If pro checks nominative Case, there is no Case feature for the subject to check covertly. Thus, the crucial difference between a PP and pro in Brazilian Portuguese is that pro checks Case, whereas the PP can check only the EPP feature. Our conclusion that pro in Brazilian Portuguese must check Case is reached because of the grammaticality of the locative inversion construction and the ungrammaticality of plain inversion (without a preposed PP) with unergative verbs. This conclusion also explains why transitives do not appear in the locative inversion construction (or in any other inversion construction for that matter). This impossibility is exemplified below: (47)

a. Clinton's security staff brought heavy weapons to Rio. b. * To Rio brought Clinton's security staff heavy weapons. c. *To Rio brought heavy weapons Clinton's security staff.

(48)

a. A segurança de Clinton trouxe armas pesadas para o Rio. b. *Para o Rio trouxe a segurança de Clinton armas pesadas c. *Para o Rio trouxe armas pesadas a segurança de Clinton

In (47) we have examples of the ungrammaticality that results when a locative inversion is attempted with a transitive verb. (48) is the Brazilian Portuguese translation of English (47), showing that in this language too transitives cannot appear in the locative inversion construction: the PP would check the EPP feature, and both nominative and accusative Case would have to be checked covertly, which does

Verb-subject order in BP

105

not happen due to the fact that at least one of the arguments has to raise before Spell-out. Given the ban on covert raising of both arguments, the subject cannot appear postverbally with transitives in languages such as English and Brazilian Portuguese, where the object always seems to move covertly: with the impossibility of the two arguments checking Case after Spell-out, the subject will necessarily move overtly. But then we need to explain why the subject must move overtly in transitive constructions, or, in other words, why it is the subject and not the object that checks Case overtly.

4.1.2. Overt Case checking by subject In null subject languages, VOS is a possible order (in languages such as Spanish and Rumanian, VSO is also possible). Given the ban on covert raising for both arguments, in VOS the subject may remain VP-internal if the language allows for object shift, for example37. Appearing only in null subject languages, VS with transitives is associated with the occurrence of pro. In other words, if the language has pro, then it allows for postverbal subjects with all kinds of verbs. Portuguese has been commonly regarded as a null subject language; however, Brazilian Portuguese seems to be losing its null subject character. Although Brazilian Portuguese does exhibit some of the properties associated with the null subject parameter, such as empty expletive elements, others are optional, such as empty resumptive subject pronoun in an embedded clause, or non-existent, as free inversion of subjects38. Some examples are given below: (49)

a. pro Parece que Juca vai 1er o livro. (It) Seems that Juca will read the book b. Juca i disse que [ej/ele^ comprou o livro. Juca said that he bought the book c. * Comprou o livro o Juca. Bought the book Juca.

106

Postverbal subjects in Brazilian

Portuguese

With the examples above, we can attest that Brazilian Portuguese does exhibit some properties of a NS language, but not all of them: (49a) shows that expletives are empty, and that the presence of a pronoun would render the sentence ungrammatical; in (49b), however, we see that the resumptive subject pronoun in the embedded clause can be empty, but it does not have to be: a full pronoun could be referring to the subject of the matrix clause, just as it could be referring to another subject. (49c) shows that Brazilian Portuguese does not admit free inversion of the subject, as Spanish and Italian do. Thus, we can conclude that Brazilian Portuguese is in the process of losing its null subject properties and becoming a non-null subject language (a process that has happened in French, as Adams 1987 shows). Although Brazilian Portuguese is presently going through the process of becoming a NNS language, it still does have pro, as seen in the examples above. Another occurrence of pro in Brazilian Portuguese is given below: (50)

Chove muito no veräo. Rains a lot in the summer

In (50) we have an example of a weather sentence, which also involves the appearance ofpro in Brazilian Portuguese. Rizzi (1986) argues that pro is allowed in some grammatical systems with a reduced range of uses. He adopts a typological classification that distributes pro among different systems. Rizzi's classification goes from no occurrence of pro, as in English, to pro as a nonargument, a quasi argument and a referential argument, as in Italian (and Spanish). Following Rizzi, pro in Brazilian Portuguese would fall somewhere between the two ends of the spectrum, even if closer to the Italian/Spanish end: in Brazilian Portuguese pro occurs as a nonargument (pro seems that TP) and as a quasi argument (pro is raining)', it can also appear as referential, as in (49b), though not as "fully" as it occurs in Spanish and Italian, languages that allow for VS order ("fully" referential pro, as we might call it, meaning that pro can appear referentially in the beginning of sentences, thus allowing for inversion to happen). But why does pro in Brazilian Portuguese not occur with transitive verbs?

Verb-subject order in BP

107

This question is related to the one above: why does the subject, and not the object, move overtly? They can be answered if we assume that pro in Brazilian Portuguese is somehow different from pro in Spanish and Italian. This difference may be related to feature checking: it is possible that in one language (Brazilian Portuguese), pro has to check features that it doesn't always in others (Spanish and Italian). We have seen that in the locative inversion construction, the PP checks the EPP feature. Ifpro is possible, then it satisfies the EPP (Chomsky and Lasnik 1993). If a language allows for pro, then pro checks the EPP feature in T, as has already been mentioned. That is true of any occurrence of pro, as it is true of English expletives. Thus pro in Brazilian Portuguese will check the EPP feature just as it does in NS languages. Therefore, the difference between pro in Brazilian Portuguese and in other languages must be related to some other feature. One feature that may solve this problem is Case. As discussed in Chapter 2, the subject in Brazilian Portuguese transitive constructions raises in order to check its (structural) Case feature (and also, naturally, the EPP feature). In a null-subject language, however, the subject does not necessarily check Case overtly. So, there must be a reason for overt raising of the subject in Brazilian Portuguese. That reason is associated with the properties of pro in Brazilian Portuguese, which does not appear with transitive verbs. We can then think of the difference between the two kinds of pro in terms of Case checking: pro in Brazilian Portuguese would always check Case, whereas in Spanish and Italian it does not have to. If pro in Brazilian Portuguese always checks the Case feature of T, we have an explanation for why it doesn't occur with transitives: the subject would not have a Case left to check covertly if that has been done by pro, and the derivation would crash. When we find the order VOS in languages such as Spanish, pro is present but it cannot have checked the Case feature in T, which is then left for the subject to check covertly. Thus the difference between pro in Brazilian Portuguese and in nullsubject languages is related to Case checking: in Brazilian Portuguese it has to check Case, whereas in null-subject languages that obligatoriness doesn't exist. The Case checking property of pro in Brazilian Portuguese is thus once more argued for: the ungrammaticality of

108

Postverbal subjects in Brazilian

Portuguese

unergatives in plain inversion is ultimately linked to the necessity of pro to check Case, as is the lack of VOS in Brazilian Portuguese. In null-subject languages, on the other hand, the subject can check its nominative Case covertly, since pro does not have to check Case.

Summary In section 4.1 we have investigated the locative inversion construction and accounted for the occurrence of unaccusatives and unergatives in such construction. The findings in this section may be summarized as follows: 1. The locative inversion construction: PP checks EPP feature, but nominative Case is checked covertly by the subject NP. 2. pro in Brazilian Portuguese checks Case, besides the EPP feature. 3. Locative inversion does not occur with transitive verbs due to ban on covert Case checking for both arguments of the verb. 4. The difference between pro in Brazilian Portuguese and pro in Spanish and Italian is related to checking of Case: in Brazilian Portuguese it can check Case, whereas in Spanish and Italian it does not, thus allowing for postverbal subjects with transitives in those languages. After this brief summary of the discussion on the locative inversion construction, feature-checking and pro in Brazilian Portuguese, we can turn to examples of VS order with unaccusative verbs in sentence-initial position, i.e., not in a locative inversion construction. In the next section we will look at the Definiteness Effect proposed by Belletti (1988) for Italian and discuss whether it applies to Brazilian Portuguese.

4.2. VS and unaccusatives The occurrence of postverbal subjects with unaccusatives is not a phenomenon particular to Brazilian Portuguese. As we have seen in

Verb-subject order in BP

109

the subsection above, English also allows it, as long as an expletive or a locative is found preverbally: (51)

a. b. c. d. e.

There arrived a woman at the office. *Arrived a woman at the office. * There worked a woman at the office. * There wrote a woman a paper. * There arrived the woman at the office.

In (51a), we can see that an unaccusative verb may appear before the subject, if an expletive is in preverbal position. (51b) shows that the verb can't be the first element in the sentence. For the most part, the subject can occupy a postverbal slot if the verb is unaccusative; this possibility is not available if the verb is an unergative39 or a transitive, as seen in (51c) and (d). (5le) shows that the inverted subject (generally) cannot be definite when the expletive there is in preverbal position. Italian also allows for postverbal subjects in sentences with unaccusative verbs (though not only with unaccusatives). Belletti (1988) claims that partitive Case is inherently assigned by unaccusatives, and argues that this is the reason why the Definiteness Effect exists. The Definiteness Effect is seen in (51a) and (e) in English. It also seems to be present in French: (52)

a.

Il est arrivé trois filles. There arrived three girls b. *// est arrivé la fille. There arrived the girl

Given the Unaccusative Hypothesis, the postverbal subject of these sentences is in fact the object of the verb. The Definiteness Effect (DE) thus concerns the nature of the DP argument of unaccusative verbs: it must be indefinite if it does not raise to [Spec, TP], In Belletti's (1988) analysis, the DP arguments of unaccusatives must be indefinite because these verbs (inherently) assign partitive Case to

110

Postverbal subjects in Brazilian Portuguese

their thematic object, and partitive Case always selects an indefinite meaning for the NP. The DE can also be found in Italian. However, this language allows for VS in sentences with verbs other than unaccusatives, and with definite subjects, as will be exemplified and discussed later. For now, the discussion will regard the instances where the DE does appear. Following Belletti (1988), relevant examples are of the following abstract form: (53)

NPJvpVNPjPP] for V=unaccusative (or passive), NP 2 its direct argument and PP a thematic complement of V.

With this configuration, one can be sure that the relevant postverbal NP is in the NP 2 object position of the verb. NPi is the position where NP2 would move to if it appeared preverbally. Given Belletti's approach, the DE is expected to affect the NP2 object of (53). Consider examples such as the following in Italian: (54)

a. All 'improvviso è entrato un uomo dalla finestra. Suddenly has entered a man through the window. b. *All'improvviso è entrato l'uomo dalla finestra. Suddenly has entered the man through the window c. All 'improvviso l'uomo è entrato dalla finestra. Suddenly the man has entered through the window (Belletti 1988: 9)

The examples in (54) seem to show that the DE is indeed present in Italian, given the contrast between (a) and (b). (54c) shows that a definite NP yields a grammatical sentence if it is preverbal. The constraints found in English, French and Italian regarding postverbal subjects with unaccusative verbs may not appear, at first sight, to be present in Brazilian Portuguese: while postverbal subjects in English, French and Italian appear to be possible only if the DP is indefinite, in Brazilian Portuguese that does not happen. In sentences that contain an unaccusative verb, a postverbal subject is possible, independently of its definiteness:

Verb-subject order in BP

(55)

a. b.

111

Morreu urna princesa. Died a princess, Morreu a princesa de Gales. Died the princess of Wales.

In (55), we see that unaccusatives may appear before the subject in Brazilian Portuguese regardless of the definiteness of the latter. Thus, it might seem that the DE does not exist in Brazilian Portuguese. However, the picture changes when some material is added to the VP: (56)

a.

Morreu urna princesa em Paris. Died a princess in Paris. b. ΊΊMorreu a princesa em Paris. Died the princess in Paris. c. Morreu em Paris a princesa. Died in Paris the princess.

As can be seen in (56), a postverbal subject with an unaccusative is indeed possible no matter whether the subject is definite or not. Nevertheless, a postverbal definite subject is only fully accepted when it appears in VP-final position, as in (56c). This indicates that we are faced with two types of postverbal subject in Brazilian Portuguese, which do depend on the definiteness of the subject. The same type of judgment is observed in (57), which is a translation of Italian (54): (57)

a.

De repente, entrou um homem pela janela. Suddenly, entered a man through the window b. V.De repente, entrou o homem pela janela. Suddenly, entered the man through the window

Although (57b) is not fully accepted, notice that (58) is. In (58), the subject is definite but appears after the PP:

112

(58)

Postverbal subjects in Brazilian

Portuguese

De repente, entrou pela janela o homem. suddenly, entered through the window the man

The differences in grammatically found in (56), (57) and (58) suggest that Brazilian Portuguese may also exhibit the DE. Therefore, in examples where the postverbal subject is an indefinite NP, Brazilian Portuguese may actually be patterning with English, French and Italian. Following Belletti's (1988) analysis, the postverbal indefinite subject receives partitive Case. She rules out the possibility for (nominative) Case transmission to the immediately postverbal position via chain. If indefinite postverbal subjects of unaccusative verbs receive partitive Case, and preverbal subjects receive nominative Case, what Case do indefinite preverbal subjects of unaccusative verbs receive? The question refers to subjects of sentences such as the following: (59)

Urn homem entrou pela janela. A man entered through the window.

Belletti discusses that question and concludes that inherent Case can combine with structural Case. She finds in Finnish overt manifestation of the combination of inherent partitive Case with structural nominative Case. Therefore, an indefinite preverbal subject of an unaccusative verb would actually be able to receive both (inherent) partitive and (structural) nominative Case, according to her analysis. But unaccusatives are not the only verbs that admit VS in Italian. Postverbal subjects are possible in that language in sentences that do not contain an unaccusative verb, such as (60): (60)

Ha parlato Gianni. spoke Gianni

Belletti (1988) proposes that such sentences are possible with a representation where the postverbal subject is adjoined to VP: (61)

[¡ρ NP¡ [j I [VP [VP V] NP J]]]

Verb-subject order in BP

113

In (61), which is the representation in Belletti (1988), for (60), the Case-assigning Infi governs not only the preverbal subject position, but also the adjoined postverbal NP. The higher VP in (61) does not constitute a barrier for government of an adjoined category, which makes it possible for nominative Case to be assigned to the postverbal subject. The structure in (61) is also assumed to be the one for sentences with an unaccusative verb and a definite subject, such as (62): (62)

È arrivato Gianni. arrived Gianni

In the light of examples such as (62), Belletti assumes that the definite postverbal subject is adjoined to VP just as the one in (62), receiving then nominative Case. As for assignment of partitive Case to indefinite subjects, it becomes the only option if the subject is not VP-adjoined:

The verb in (63) is an unaccusative verb. Recall that nominative Case, according to Belletti, cannot be transmitted via chain to the thematic object of an unaccusative verb in its base position, which makes partitive Case the only one available for that position. Examples (55) and (56) seem to show that the subject of unaccusative verbs do receive (or check) partitive Case in Brazilian Portuguese, following Belletti. At this point, we need to make clear that Belletti worked within the Government and Binding theoretical framework, whereas the analyses provided here follow Chomsky's Minimalist Program (1995b). In Belletti (1988), for example, Case is received by an NP, while here we refer to checking of Case by an NP

114

Postverbal subjects in Brazilian Portuguese

(or a DP). Keeping these differences in mind, we can proceed to an analysis of the Brazilian Portuguese data. As mentioned above, postverbal indefinite subjects of unaccusative verbs in Brazilian Portuguese do seem to check partitive Case. In that instance, the structure for a sentence such as (55 a), repeated below as (64a), could be (64b): (64)

a. b.

Morreu urna princesa. TP

T+Asp+V morreu

urna princesa In (64b)40 we have a structure where the subject NP remains in postverbal position after Spell-out, where I assume that it checks partitive Case from the unaccusative verb (specific assumptions about checking of inherent vs. structural Case will be discussed below). This structure applies for sentences with an unaccusative verb and an indefinite subject, where the latter checks partitive Case. The checking of partitive Case by indefinite subjects of an unaccusative verb appears to happen in Brazilian Portuguese, given the data above (cf. [57]). Nevertheless, the order verb-subject in Brazilian Portuguese is possible with definite subjects also, as long as the verb is an unaccusative (cf.[55J). We could follow Belletti (1988) and assume that nominative Case may be checked by the NP in object position. But then the obvious question would be: why don't other classes of verbs allow for postverbal subjects? If nominative Case could be checked by the subject in its base position, and it did not

Verb-subject order in BP

115

have to move to do so, then verb-subject would always be possible in Brazilian Portuguese. Since it is not, the checking of nominative Case in the merged position of the subject is ruled out. However, in the discussion of the locative inversion construction we have seen that, rather than a strong nominative Case feature, it is actually the EPP feature that determines whether or not inversion is possible: if this feature can be checked overtly by an element other than the subject, then inversion can occur (with one-argument verbs, as we have seen). For now, however, suffice it to say that an indefinite subject that appears postverbally in Brazilian Portuguese (with an unaccusative verb) should check partitive Case, as is argued by Belletti for Italian, English and French. Nevertheless, we now need to account for definite subjects of unaccusatives which may also appear after the verb (although in a different order relative to PP adverbials): definite subjects pose a problem regarding Case, since here it has been argued that pro in Brazilian Portuguese checks nominative Case. Before we proceed to investigate postverbal definite subjects, however, we should depart a little from unaccusatives and determine how accusative Case is checked in Brazilian Portuguese. The discussion of the checking of accusative Case may prove relevant to an analysis of the facts with unaccusative verbs and postverbal subjects.

4.2.1. Checking of accusative Case In discussing the status of Agr in Minimalism, Chomsky (1995b) argues, at first, that Agr exists only when it has strong features. He ends up dispensing with Agr entirely, opting instead for the idea of multiple Specs. According to him, the Case-assigning feature is intrinsic to the heads (V, T). Chomsky assumes transitive constructions to be of the form in (65), ignoring [Spec, V]:

116

Postverbal subjects in Brazilian

Subj

Portuguese

v'

V

Obj

According to Chomsky, overt object raising can be forced by adding to ν itself a strong D-feature that requires overt substitution in the "outer Spec" of a multiple-Spec configuration. Object-raising takes place when the light verb ν that heads the transitive construction in (64) is assigned the strong feature as it is drawn from the lexicon and placed in the numeration. Chomsky notes that the choice is arbitrary, forced or unavailable, as the language has optional, obligatory or no overt object-raising, respectively. Given the principle of Procrastinate, the most economical derivation would be one in which neither the subject nor the object moved overtly. Nevertheless, following Chomsky (1995b) and Alexiadou and Anagnostopoulou (1997), at least one argument has to move in the overt component: if both move covertly, we would not obtain the right structure at the LF interface, as discussed previously. In Chapter 2, where I discussed the basic word order in Brazilian Portuguese, I argued that preverbal subjects move before Spell-out. Overt raising of the subject allows for the object to move covertly, which is what I also assumed in that chapter, and which would be in accordance with Procrastinate. Taking an approach that would conform with what is argued in Chomsky (1995b), if accusative Case is checked covertly, the formal features of the object are attracted directly to the verb in T. This would be what happens in Brazilian Portuguese, since I have assumed that objects do not raise before Spellout. Overt movement of the subject and covert raising of the object (and consequent covert checking of accusative Case) explains the SVO order found in Brazilian Portuguese. However, SVO would also obtain if the object raised overtly, as long as the verb moved higher than the object. It has been argued that the object in English moves

Verb-subject order in BP

117

before Spell-out (Johnson 1991). For Brazilian Portuguese, however, that does not seem to be the case: participial agreement (as in Italian) would indicate overt object movement, and Brazilian Portuguese does not display that agreement. In the lack of uncontroversial data that this is so, I will assume that there is no overt raising of the object in Brazilian Portuguese (perhaps for economy reasons). If the object were to raise overtly, then the subject could move after Spell-out, which would yield VOS order (as in Spanish). But that possibility is not available in Brazilian Portuguese, as seen before, because pro would check the Case feature and the subject would have no Case left to check when it moved covertly. We can have a subject not move to [Spec, TP] overtly if there is something else available to check the strong D feature of Τ without checking Case, as in locative inversion. Note, however, that the need for the subject to move overtly in Brazilian Portuguese transitive constructions does not necessarily preclude the possibility of the object also moving overtly. Overt movement of both arguments seems to be a marked option, perhaps because it violates Procrastinate. But the option does exist, as evidenced by Icelandic (see discussion of Transitive Expletive Constructions in Chapter 2). Brazilian Portuguese does not display this marked option (at least there is no conclusive evidence that it does), and if the subject moves overtly, the object will move covertly. Thus, checking of accusative Case in Brazilian Portuguese happens after Spell-out. The discussion of checking of accusative Case leads us to the discussion of feature checking in sentences with postverbal subjects. Since we have already seen the data with indefinite subjects that appear after the verb in Brazilian Portuguese, I will turn to an analysis of those data next.

4.2.2. Postverbal indefinite subjects: feature checking We have seen that indefinite subjects of unaccusative verbs that appear postverbally in Brazilian Portuguese may check partitive (inherent) Case. Now we can take a closer look at how this feature checking might take place.

118

Postverbal subjects in Brazilian

Portuguese

Unaccusative verbs differ from transitives and unergatives in that they lack a higher VP shell. This shell is exemplified in (66), where we have the structure for a ditransitive verb, following Larson (1988): (66) VP DP a book

V' V

PP

give

to Mary

As we can see in (66), the direct complement to the verb is in [Spec, VP], while the indirect complement is a sister to V. As for unaccusative verbs, we can assume that their subjects have the theta status of a complement to the verb. Since they lack the higher v', their subject would occupy the position that the direct complement of a ditransitive verb does, i.e., [Spec, VP]. We can also assume that a directional PP, on the other hand, is the indirect complement to an unaccusative verb of motion, such as enter, and as such it would occupy the position that the PP in (66) does. The analysis of Case checking by a postverbal indefinite subject requires the assumption that there is a difference in checking of structural and inherent Case. Chomsky (1995b) argues that structural Case can only be checked by a nontrivial chain. In other words, it cannot be checked in the position to which the DP merged. We propose that inherent Case, on the other hand, is checked in its immediate merge position. This is a theta-position, and inherent Case is linked to theta considerations. Thus, for a sentence such as (67a), we can have the structure in (67b), after merging of the V, the DP and the PP: (67)

a.

Entrou urn homem pela janela. entered a man through the window

Verb-subject order in BP

b.

119

VP um homem

V'

V

PP

entrar

pela janela

In (67b) we see that the subject occupies the [Spec, VP] position, as does the direct object of 'give' in (67). The PP pela janela 'through the window' is a sister to V, as is the indirect object in (67). In (67b) the DP urn homem 'a man' would have its inherent (partitive) Case checked immediately upon merging into [Spec, VP], After the verb raises to Τ to have its features checked, we get the order V-S-PP in (67a), which Belletti (1988) argues is the unmarked order for indefinite subjects with unaccusative verbs (at least for Italian). An idea that is crucial to the assumptions above and the analysis that results is that there is a pro to check the strong D (or N) feature of T. However, pro does not check the phi-features, since those match the subject DP (here, um homem). This means that the phifeatures of the DP are checked covertly, thus allowing for the verb to agree with the postverbal subject. In an analysis such as the one outlined here, pro behaves like existential there in English with regard to checking of the DP's phi-features, since the verb also agrees with the subject DP, as in (68):41 (68)

There are three women at the office.

In (68) it is clear that the verb agrees with the subject DP three women, as does the unaccusative verb in (67a). In this respect, thus, pro in the Brazilian Portuguese unaccusative construction is not like il in French, since the French verb agrees with singular il and not with the postverbal DP: (69)

Il est arrivé trois hommes. Arrived-sing. three men

120

Postverbal subjects in Brazilian

Portuguese

The position of the subject at LF can be confirmed with binding facts. Following Chomsky (1995b), the associate in the English existential construction should have the binding and control properties of the overt subject: covert raising of the subject's formal features to the verb in Τ establishes the relevant c-command configuration for binding and control. In a null subject language, he continues, the counterparts to the examples in (70) should be admissible, as opposed to those in (71): (70)

a. There arrived three men without identifying themselves. b. There arrived with their own books three men from Eng land.

(71)

a. * I met three men without identifying themselves. b. *Ifound with their own books three men from England.

The contrast between Italian (72) and French in (73) shows that the expectations regarding binding and control seem to be correct: (72)

a.

Sono entrati tre uomini senza identificarsi. Three men entered without identifying themselves b. Ne sono entrati tre t senza dire una parola. Of-them three entered without saying a word

(73)

a. *Il est entré trois hommes sans s'annoncer. Three men entered without identifying themselves b. *// en est entré trois t sans s'annoncer. Of-them three entered without identifying themselves

The null subject expletive in Italian has the relevant properties of English there. LF-raising of the formal features of the subject, then, assigns A-position properties to the subject for binding and control, which makes clear that the subject is overtly in object position. In French, where the expletive il is analogous to English it, the LF operation does not take place: the features of the matrix TP (INFLphrase in Chomsky's structure) have already been checked by the expletive. There is no covert raising, and no binding or control.

Verb-subject order in BP

121

The control facts in Brazilian Portuguese confirm that the postverbal subject in Brazilian Portuguese patterns with that in English. In Brazilian Portuguese, raising of the subject at LF also occurs: (74)

a.

Entraram [très alunosj depois de PROJalar com a professera. Entered three students after speaking with the professor b. Entrou [cada homemj com sua [pròpria J esposa. Entered each man with his own wife c. Vi [très alunosj depois de PROjfalar com aprofessora. I saw three students after speaking with the professor.

The subject très alunos 'three students' (or its formal features) in (74a) has moved at LF, showing control properties. The binding properties are demonstrated in (74b): pròpria 'one's own' is an anaphor that needs a c-commanding antecedent in its binding domain. (74c) demonstrates that objects do not exhibit the same control properties as subjects: in that example, PRO is not controlled by the object très alunos. The properties shown in (60b) then parallel the English example in (70a). Thus, pro in the Brazilian Portuguese unaccusative construction, where we find a postverbal subject, does indeed share relevant properties of the English existential. Like there, pro in Brazilian Portuguese checks the EPP feature. But is the EPP the only feature checked by there? According to Chomsky (1995b), there must lack Case or phifeatures, or both, or all features of TP would be checked and the associate would not raise. However, Lasnik (1995) builds up an argument to weaken the condition of Greed, put forth in Chomsky's Minimalist Program (1993), and this argument includes the proposal that the associate of there has its Case checked independently of there. The EPP follows from the fact that the D feature of Τ is strong, which in turn means that Τ must check that feature before Spell-out. When Τ checks a feature, the strong D feature of Τ disappears, because if that did not happen, there would be no PF convergence, whether or not the EPP was satisfied. Thus, Lasnik (1995) argues that Τ must check the Case of there in the overt component, or the derivation will crash at PF. But note that correlation does not necessarily follow, since the

122

Postverbal subjects in Brazilian Portuguese

strong D feature (the EPP) and the Case feature of Τ are two different features. The possibility must exist for (at least some of the time) some languages to check the strong D feature without checking Case (as, for example, in locative inversion, or as in VOS/VSO in Spanish). If the Case feature of Τ is strong and is not checked, then indeed the derivation will crash. But Chomsky (1995b) limits strong features to categorial features. Therefore, Lasnik's theoretical argument may not be without flaws. However, he does present empirical evidence for overt checking of Case by the expletive there. His argument for Case checking by there has several consequences, but the first one seems to be the fact that an element that is comparable to there also checks Case in the overt component of the syntax. Therefore, if pro in Brazilian Portuguese can be compared to the expletive there in English, then the former should also check Case as the latter does. Furthermore, as we will see later in this section, pro in Brazilian Portuguese can also be compared to expletive it in English, in that both check Case (as will be discussed later, pro is also present in sentences with the verb parecer 'to seem'). This correlation between the expletives in English and Brazilian Portuguese pro confirms the argument in 4.1.1, where pro is said to check Case in Brazilian Portuguese, thus accounting for the lack of VOS in this language. One difference between the expletives it and there in English is that it could be said to also check phi-features, unlike there. However, one can also argue that the third person singular is default when the subject (i.e., a postverbal clausal subject) has no phi-features, an argument which is more compatible with facts analyzed later in this section: we will see that pro in Brazilian Portuguese can appear in environments similar to where it appears in English (with postverbal sentential subjects). But then, if it is believed to check phi-features, then pro in Brazilian Portuguese would appear to also be able to do so - but only sometimes. This 'optionality', so to speak, regarding checking of features can be problematic: what would stop pro from checking those features if there were a postverbal DP? The approach that the third person singular is default in those instances is thus compatible with all occurrences ofpro, as we will see again later. Returning to checking of the Case feature, if English there checks Case, the relevant feature of Τ is no longer available to check

Verb-subject order in BP

123

the Case feature of the associate NP or DP. Thus, the associate NP or DP must have its Case checked in some other way. As has been assumed here, unaccusatives (as well as the verb to be in English, and ser/estar in Portuguese) check partitive Case. Since to be is assumed to check partitive Case, some discussion of that point may be in order. Lasnik (1995) raises the question of how partitive Case can be licensed by be in simple existential constructions such as (75), given the standard small clause analysis: (75)

There will be [a man available].

The question appears given the ungrammatically of Italian (76), discussed in Belletti (1988): (76)

*Sono considerati [alcuni studenti intelligenti]. Are considereed some students intelligent

When the Italian verb considerare is passivized, it can't take a small clause complement with an overt subject, as seen above (the possibility does exist for active sentences). The ungrammaticality of (76) is expected, since there is no theta-relation between the passive verb (which, according to Belletti [1988] is the partitive Case licenser) and the subject of the small clause. But then how can (75) be explained? As discussed in Lasnik (1995), be does not theta-mark a man (although it may check the Case feature), but the predicate of the small clause (available) does. Lasnik discusses this problem, and explores the theory of inherent Case in the search for a solution. Chomsky (1986) distinguishes inherent from structural Case in three aspects. The first difference was related to D-structure, where inherent Case would be assigned, and S-structure, where assignment of structural Case would take place. That difference has disappeared, given the developments in the theory, namely Chomsky's Minimalist Program, where he proposes that those levels do not exist. But there are still two differences between the two types of Case: the standard configuration of inherent Case is the head-complement relation between licenser and licensee, and the Case licenser must also license a thetarole to the licensee. As pointed out by Lasnik, recent minimalist re-

124

Postverbal subjects in Brazilian

Portuguese

search has looked for symmetry. He then adopts the null hypothesis and extends the symmetry proposed for (structural) accusative Case and (structural) nominative Case to inherent Case, which would mean that the three are licensed in a spec-head relation to a functional head. If this is so, the LF configuration of an unaccusative verb and its complement would parallel that of a normal transitive verb and its complement. This means that the theta-marking requirement remains as the only special property of inherent Case. But the spec-head configuration just mentioned can be considered as one of theta-licensing for unaccusative verbs. Then we are left with (75). For that sentence, Lasnik proposes the structure in (77), which is a standard ECM configuration not in conformity with the theta-marking requirement: (77)

AgrOP NP AgrO' a man AgrO V be

VP AgrO (Lasnik 1995: 34)

Lasnik notes that now be is a light verb, having no theta-roles of its own to assign. He follows Saito and Hoshi (1994) and argues that the resulting configuration is indeed one of theta-marking: the merged predicate combines the Case feature of the higher verb be with the marking property of the lower predicate available. With the merged predicate in AgrO (for the sake of discussion, I simply quote Lasnik's structure), all requirements for the licensing of inherent partitive Case are satisfied. Lasnik notes that other unaccusative verbs will not license partitive Case exceptionally if this raising process is restricted to instances where the higher predicate is "light". With this analysis, Lasnik discusses the apparent fact that the passive verb (in a sentence such as [75]) does not license partitive Case: this licensing would be done by the complex of be and the passive verb. He concludes that be is responsible for partitive Case in

Verb-subject order in BP

125

English and in Italian passive constructions. Following his argument, I will extend his analysis to ser/estar in Portuguese, assuming that those verbs, just as be in English, check partitive Case. Having discussed Case checking by be, we can now return to our considerations on pro. Since we have drawn a parallel between pro and English existential there, it seems natural to suppose that pro in Brazilian Portuguese may occur as a locative-like pleonastic. The idea of pro in Brazilian Portuguese not being "fully" referential is supported by the fact that Brazilian Portuguese does not exhibit all the properties of a full-fledged null subject language such as Spanish, where pro does occur as a referential argument. In 4.1.1, where we discussed pro in Brazilian Portuguese, we saw the properties of the null subject parameter that Brazilian Portuguese still holds, and those that it does not display anymore. We have also seen that Rizzi (1986) adopts a typological classification that distributes pro among different systems, in a classification that goes from no occurrence of pro to pro as a nonargument, a quasi argument and a referential argument. Thus pro in Brazilian Portuguese is found between the two ends, since it occurs as a nonargument (pro seems that TP) and as a quasi argument (pro is raining), as has been mentioned before. Here I repeat examples of both occurrences ofpro in Brazilian Portuguese: (78)

a. pro Parece que o aluno näo entendeu o assunto. (It) Seems that the student didn't understand the subject b. pro Nevou muito cedo este ano nos EUA. (It) Snowed very early this year in the USA

These two uses of pro correspond to the English expletive it, as opposed to the unaccusative construction, where, as we have seen, pro seems to pattern with there in English. Here it has been argued that in the Brazilian Portuguese unaccusative construction (where a postverbal subject occurs), pro checks the Case feature of T, following Lasnik's (1995) argument for there checking Case overtly. However, as hinted at before, not all instances of pro in Brazilian Portuguese parallel English there: as in the examples above, pro in Brazilian Portuguese may correspond to English it. The fact that pro in Brazilian

126

Postverbal subjects in Brazilian Portuguese

Portuguese finds correspondence in both English expletives could be problematic for the analysis being developed here. Nevertheless, there seems to be agreement among researchers that the English expletive it does check Case. Chomsky states that "English it and French il (...) have a full complement of features." (Chomsky 1995b: 288) Therefore, whether pro in Brazilian Portuguese is compared to it or to there, it will check Case overtly, as do the English expletives, since pro only occurs in those environments where the expletives do. One difference between the English expletives it and there appears to be checking of phi-features, which it seems to do, since the verb agrees with it. However, as briefly discussed earlier, we may consider the third person singular that occurs with the expletive it the default form in cases where there is an "extraposed" sentential subject: that kind of subject cannot check phi-features. So, instead of claiming that pro in Brazilian Portuguese only checks phi-features sometimes, we can maintain that the third person singular in (78) is a default form when the subuject does not have phi-features. Summarizing our analysis for indefinite postverbal subjects in Brazilian Portuguese, we assume that the subject of an unaccusative verb behaves as a complement to the verb with regard to theta considerations, and that the direct complement of a ditransitive verb is in [Spec, VP]; a directional PP is the indirect complement to an unaccusative verb of motion. Following Chomsky (1995b), the difference between structural and inherent Case is established, which allows for the inherent (partitive) Case of the indefinite subject to be checked in the merge position. After the verb raises to T, we get the surface order V-S-PP. For this analysis to be possible, it is assumed that there is a pro in Brazilian Portuguese that checks the DP feature of T. This pro, however, does not behave as a referential argument (as it does in Spanish), although it does check Case (as do there [following Lasnik 1995] and it in English). Having outlined an analysis for indefinite postverbal subjects with unaccusative verbs we can outline some considerations on pro with respect to the lack of VOS and VSO in Brazilian Portuguese, seeing how the lack of those orders can be related to the characteristics of pro in this language. We will return to definite postverbal subjects with unaccusatives in subsection 4.2.4.

Verb-subject order in BP

127

4.2.3. Lack of VOS and VSO The fact that pro in Brazilian Portuguese checks Case overtly may help us understand why this language does not allow for VOS or VSO orders. Looking at one analysis of those orders, Alexiadou and Anagnostopoulou (1997) propose an account for VOS that states that this order occurs in Italian and Spanish for the same reason why Icelandic allows TECs. Since at least one argument has to move in the overt component, a suggestion by Chomsky that is looked at in more detail in the discussion of locative inversion (4.1), they argue that if a language permits subject movement to Τ and/or object shift, postverbal subjects will be possible in the presence of an object. They assume that the D feature of Τ is optionally strong in languages such as Italian. If strength on Τ is not selected, the subject remains in its VP internal position, and thus overt object movement must occur. The reason why at least one of the arguments has to move in the overt component is related to the need to obtain the correct c-command relationship at the LF interface (Chomsky 1995b). The subject in Brazilian Portuguese transitive constructions would be able to move covertly if there were a pro that could check the EPP feature. Since pro in Brazilian Portuguese can do that, then the lack of VOS must be related to the object not being able to move overtly. If the object doesn't, the subject has to. An object would only move overtly if it had to check some strong D or Ν feature of a functional head. Here I have been working with the assumption that objective Case is checked in Asp. This being so, Asp in Brazilian Portuguese cannot have a strong D or Ν feature, not even optionally. Assuming thus that the D or Ν feature in Asp is always weak, the possibility for overt movement of the object is barred, which in turn accounts for the lack of VOS. Stipulating that Asp in Brazilian Portuguese must have a weak D/N feature serves the purpose of accounting for the lack of VOS, but there appears to be no independent motivation for that 'weakness'. A stronger explanation for the lack of VOS would involve an argument that could be motivated independently. This argument is at hand: if pro in Brazilian Portuguese (like the expletives there and it in English) checks not only the EPP but also the Case feature of T, then the

128

Postverbal subjects in Brazilian

Portuguese

subject of the transitive verb would have no Case left to check when it moved covertly, as discussed in 4.1.1. The subject must check its nominative Case, which would not be possible if pro were occupying the [Spec, TP] position, for that would mean that the Case feature of Τ would be checked by pro and would not be available when the subject moved in the covert syntax. With this account, we need not discuss whether or not the object moves overtly in Brazilian Portuguese (although we assume it does not, perhaps for economy reasons): the subject has to move to [Spec, TP] to check Case, and that position cannot be occupied by pro as it is in Italian and in Spanish. Here the difference between pro in Brazilian Portuguese and pro in Spanish and Italian is fundamental. In the locative inversion section we argued that pro in Brazilian Portuguese must always check Case, whereas in Spanish and Italian Case checking by pro may be optional (in instances of VOS, pro does not check Case, for the reasons mentioned above). If pro does not always have to check Case in null-subject languages, then VOS can be allowed at least when Asp has a strong D/N feature. The fact that pro in Brazilian Portuguese checks Case may also account for the lack of VSO, the reasoning being the same as for the explanation for the lack of VOS. But this account will depend on the structure assumed for VSO. If we assume, with Ordófiez (1997), for example, that in VSO the subject moves out of the VP but not as high as [Spec, TP]42, the verb moves to Τ and the object remains in situ, the subject would still need to check its Case covertly. Since pro in Spanish does not have to check Case, in such a derivation it will not, and the Case feature will be available for the subject to check after Spell-out. In other words, both in VOS and in VSO (assuming Ordófiez's derivation for VSO, for example) the subject would somehow have to check its nominative Case covertly. But that possibility becomes unavailable if pro checks that Case in the overt component, which is what happens in Brazilian Portuguese. Alexiadou and Anagnostopoulou (1997) suggest that VSO is an order related to the existence of clitic doubling in a given language: only languages that allow clitic doubling also have VSO, such as Spanish, Rumanian and Greek. They analyze the relation between V and S in VSO in these languages as an instantiation of doubling.

Verb-subject order in BP

129

Their argument seems plausible, and it would also apply to Brazilian Portuguese: if there is no clitic doubling, there is no VSO. And in Brazilian Portuguese there are barely any clitics: along with null subject characteristics, clitics are also being less and less used in Brazilian Portuguese. Since clitics are not used, clitic doubling is obviously not possible. Thus, whether we follow Alexiadou and Anagnostopoulou (1997) or rely on Brazilian Portuguese pro checking Case overtly, we reach the same result: VSO is not allowed in this language. The conclusion just reached relies heavily on a derivation of VSO where the subject does not move to [Spec, TP]. That could be obtained either with a derivation such as Ordóñez's (1997), or if both the subject and the object, as well as the verb, remained in their base position. The latter option, however, is made impossible if we follow (as we have here) Chomsky's reasoning that at least one of the arguments has to move in the overt component. Nevertheless, VSO could still be the result of a third derivation: it is possible that the subject might move to [Spec, TP], and the verb would move past it to a position above TP. This possibility, among others, is raised by Sufier (1994). However, she argues that V does not raise past I (or T) in Spanish: assuming the projection of NegP for negation, Neg must be higher than Τ in Spanish, where the particle no always appears before the inflected verb. Furthermore, other adverbials precede no, suggesting that the verb does not move past T: if it did, the adverbials would be expected to follow the verb, not precede it. An example from Sufier (1994) is given below: (79)

Drea siempre sabe el resultado. Drea always knows the result (Sufier 1994 [23a])

In (79) we have a preverbal subject, but similar results obtain with postverbal subjects: (80)

Siempre sabe Drea el resultado. Always knows Drea the result

130

Postverbal subjects in Brazilian Portuguese

If the verb in Spanish VSO were to move past Τ the adverbials, as mentioned above, should have to follow it, especially the negation adverb, assuming the existence of NegP above TP. But even in VSO that does not happen, as seen in (80). Therefore, it seems that VSO cannot in fact be derived with V moving past T. So, the subject in this order must not be in [Spec, TP]. And if it is not in that position overtly, then it must check its Case after Spell-out, which is made impossible in Brazilian Portuguese because pro would check Case before the subject did. Having ruled out VOS and VSO in Brazilian Portuguese, we can now turn to definite subjects that appear after the verb.

4.2.4. Postverbal definite subjects As we have seen, postverbal subjects in Brazilian Portuguese appear (mostly) with unaccusative verbs. Nevertheless, definite and indefinite subject DPs/NPs yield different results when the VP has more material than simply the verb and the subject. The results obtained with indefinite DPs/NPs may be accounted for by an analysis such as Belletti's (1988). But what about a definite NP that occurs postverbally? It is clear that a different account is needed for examples such as (56c) or (58): partitive Case cannot be assigned to definite NPs. An example of a definite postverbal subject is the unaccusative construction (59b), repeated below, where pro must be present in [Spec, TP]. At Spell-out, the structure for (81a) must be (81b): (81)

a

Morreu a princesa de Gales. Died the princess of Wales b. [χρ proj [χ· morreu [AspP ti [Asp' tv [VP tv « princesa de Gales]}]]]

As discussed in section 4.2.2, in (81), pro is a quasi-locative element that checks the DP/NP features of TP, making it possible for the DP/NP subject to remain in its base position. In (81b) pro is merged in [Spec, AspP], and raises to [Spec, TP], Evidence for the [Spec, AspP] as the base position of pro comes from sentences such as (82):

Verb-subject order in BP

(82)

131

Vêm sempre os alunos. come always the students

In the discussion of position of adverbs in Chapter 2, we reached the conclusion that the tensed verb in Brazilian Portuguese raises to Τ (or, optionally, to Asp) overtly in declaratives. Adverbs can only adjoin to X' (or TP), and adverbs of time do not adjoin to the verb, given that adverbs adjoin to the categories they modify (and there would be no significance in adjoining an adverb of time to a verb). That being so, in (86) the verb has raised to Τ and sempre 'always' must be adjoined to Asp', which in turn means that AspP must be present. Since we find an adverb adjoined to Asp', [Spec, AspP] must be present in the structure. But, as discussed in Chapter 2, a Spec position is projected only when an element is present there (or moves through it). Therefore, an element must have been present in [Spec, AspP] at some point in the derivation, otherwise it would not be projected, which would bar the projection of Asp' and the consequent possibility for an adverb to adjoin to it (to Asp'). The only candidate that could have occupied [Spec, AspP] in (87) is pro, which must then have originated in that position. The structure for (82) is given below: (83)

[ T P proi [ r vêm [ A s p p tj Usp' sempre

[ A s p ' t v [VP t v os alu-

nos]]]]]] In (83), pro is occupying the [Spec, TP] position, having moved there from its merge position, [Spec, AspP].4 This description of the facts, however, is not an explanation for why it is that Brazilian Portuguese admits definite NPs in postverbal position with unaccusative verbs. Looking at some English examples, we can see that this language also allows for postverbal definite subjects: (84)

There comes the ice cream vendor.

The sentence above shows that, in actual fact, definite subjects can also occur postverbally in English (with an unaccusative verb). The

132

Postverbal subjects in Brazilian

Portuguese

same is true for the existential construction, as discussed in Birner and Ward (1993): (89)

Q : " What is there on your deskT A: "There is the usual telephone and the red pen."

As we can see in (85), given the appropriate context, definite NPs can also appear with there+to be. Thus, it is not only Brazilian Portuguese that allows for inverted definite subjects with unaccusative verbs (or existentials). The possibility of getting postverbal subjects with unaccusatives is then broader than simply indefinite versus definite NP. Thus, it may be the case that postverbal subjects with unaccusatives are related rather to "new" versus "old" information, to the extent that a postverbal subject is syntactically possible. For Spanish, Zubizarreta (1994) notes that new information appears postverbally, whereas old information is found in preverbal position. In her work, the notion of Universe of Discourse (UD) plays an important role: it includes the set of referents and properties that are shared by the speaker and hearer in the instant of the utterance. Zubizarreta then defines old and new information with respect to the UD: in the instant of the utterance U within a discourse D, "old" information is constituted by referents and properties that have been introduced prior to U and "new" information is constitued by referents and properties that had not been introduced prior to U. She illustrates that contrast with the examples below: (86)

a.

Vino el cartero esta mañana. Para mi gran alivio, llegó la carta (que estábamos esperando). Came the postman this morning. To my great relief, arrived the letter (that we were watiting for) b. Esta mañana llegó una carta de Paris. This morning arrived a letter from Paris. c. Mario me ha escrito una carta. %Llegó la carta ayer. Mario to me-has written a letter. Arrived the letter yesterday.

Verb-subject order in BP

d.

133

Mario me ha escrito una carta. La carta llegó ayer. Mario to me-has written a letter. The letter arrived yesterday. (Zubizarreta 1994: 182)

As we can see, in (86a) a definite NP is found in postverbal position, constituting "new" information within the discourse. The same happens in (86b), where we find an indefinite subject. When the subject is old information as defined above, however, it must appear preverbally, as shown with the contrast between (86c) and (d). The Spanish examples above translate to Brazilian Portuguese, suggesting that the facts correlating "new" information and VS order also apply to this language, at least for unaccusative verbs.44 The grammatical notions of "topic" and "focus" are based on the discourse notions of "old" and "new" information. Given the facts in (86), it seems that postverbal subjects are focused. Now the question is how to articulate the discourse notions with the grammatical ones. Zubizarreta adapts a proposal by Herburger (1992), which states that all NPs/DPs take an e(vent) argument and that if an NP/DP is in the scope of the EV(ent) operator (which is assumed to be contained in the Aspect node), its e argument is bound by the EV operator. In such a case the NP/DP constitutes "new" information. If the NP/DP is outside the scope of the EV operator (the Asp node), its e argument is free, and it is interpreted as linked to an event previously introduced in the discourse, thus constituting "old" information. In the light of this correlation, Zubizarreta states that at LF, if a DP is focus, it is within the scope of the Aspect node; if it is topic, it is outside the scope of the Aspect node45. A DP that is found in postverbal position is within the Asp node, and thus should be focused. Given (86), which finds correspondence in Brazilian Portuguese, we see that this is indeed the case: in a VS construction, the subject is interpreted as new information in the discourse, and is found within the Asp node, thus being focused. The association made above between a focused postverbal NP and new information in the discourse permits us to state discourse notions in grammatical terms. However, we have not accounted for

134

Postverbal subjects in Brazilian Portuguese

Case checking of definite postverbal subjects. How does it take place? Indefinite subjects of unaccusative verbs supposedly check partitive Case (Belletti 1988), and definite subjects are supposed to check nominative Case. Belletti identifies the inherent Case of unaccusative verbs with partitive, but this correlation is not necessary if the Definiteness Effect is pragmatic, and not syntactic. The status of the DE is, according to Chomsky (1995b: 384), "unclear for many reasons, including the fact that it (...) interacts with other properties, such as focus." If the DE does not hold necessarily, we need not assume that the Case of unaccusatives is only partitive, although it is indeed inherent (and not structural). Recall that our assumption regarding inherent Case is that it is checked in situ. If unaccusative verbs check inherent Case even when the subject is definite, we can then explain the fact that postverbal definite subjects are possible with that class of verbs. We have seen that there in English and pro in Brazilian Portuguese check Case. We have also seen that English there may occur with an unaccusative verb and a postverbal definite NP. Under our assumptions, the postverbal definite subject checks inherent Case, which poses no problem for the previous assumption that there checks the Case feature of T. The same applies to postverbal definite subjects in Brazilian Portuguese, which occur in unaccusative constructions where pro checks Case in T. Having explained the occurrence of postverbal definite subjects with unaccusative verbs in Brazilian Portuguese, we now need to discuss the difference in judgment found when there is more material in the VP. This is exemplified above in (56b) and (c). Another example is found below: (87)

a.

Veio ao Rio o papa. Came to Rio the pope b. IWeio o papa ao Rio.

As we can see, the presence of a PP is better accepted if it appears before the DP subject, as in (87a), than in sentence-final position (as in [87b]). In our discussion of postverbal definite DPs above, we fol-

Verb-subject order in BP

13 5

low Zubizarreta (1994) and conclude that those DPs are interpreted as new information, and that they are focused, since they are within the Asp node. In order to account for the difference in judgment between (87a) and (b), we assume, with Cinque (1993), that the accentual prominence of a lexical item is a function of its depth of embedding. Therefore, to the extent that the postverbal definite DP is focused, it is also the last element in the clause. With the structure that I assume here, the order V-PP-definite subject obtains when the PP moves past the subject, which must happen for it to get out of the scope of the focus feature (since it does not have to move for Case checking reasons). The prediction of this analysis is that, given a certain pragmatic context, it must be possible to find the order V-definite argument-PP, since in V-PP-definite subject the PP is "old" information, while the subject is "new". This prediction is in fact borne out: (88)

a.

Os músicos de jazz geralmente väo a Säo Paulo, mas este ano veio Neneh Cherry ao Rio. The jazz musicians usually go to SP, but this year came Neneh Cherry to Rio.

AspP veio Asp' VP Asp tv DP N.C.

V' PP

Δ ao Rio

136

Postverbal subjects in Brazilian Portuguese

As (88) shows (with the structure in [88b] conforming to previous discussion on the Larsonian shell - see example [67]), the order Vdefinite S-PP is indeed possible, if the PP does not have to 'escape' the focus feature, being "new" information in the discourse. So, in the right context, we can also get the order V-definite S-PP, although it may not be the 'preferred' order in Brazilian Portuguese. In this subsection we discussed the occurrence of definite postverbal subjects with unaccusatives, which happens in Brazilian Portuguese as well as in English. We saw that new information in the discourse appears postverbally, and, following Zubizarreta (1994), an association was established between new information and focus, leading to the conclusion that postverbal subjects are focused. In order to account for Case checking of definite subjects that appear after the verb, our proposal is that unaccusative verbs do not need to check partitive Case only (which is a Case that can only be checked by indefinite NPs), although they do check inherent Case for postverbal definite NPs too. The difference in judgement found between examples such as (87a) and (87b) was explained in the light of an algorithm proposed by Cinque (1993): the most prominent accent in a sentence falls on the most embedded constituent, which means that a focused NP should be the last element in the clause. However, the PP may appear in clause-final position if it constitutes "new" information in the discourse, as shown in (88). With an account of postverbal definite and indefinite subjects with unaccusative verbs, we can now turn to examples of VS order in Brazilian Portuguese with verbs that cannot, a priori, be classified as unaccusatives.

4.3. VS with other verbs In the beginning of section 4, we saw that not only verbs that are classified as unaccusatives allow for a postverbal subject. It was shown that a few verbs that do not pass the tests for unaccusativity still allow for the order VS, like the verb ficar 'stay, remain'. Here I give a few more examples of VS with verbs that are not diagnosed as unaccusatives:

Verb-subject order in BP

(89)

137

a.

Ficou um aluno depois que a professera saiu. Stayed one student after the teacher left b. Onde ficou o aluno? Where stayed the student? c. Ligou o Carlos. Telephoned Carlos d. Quando vai ligar a Ana? When will telephone Ana? e. Eia ligou ontem, e atendeu o Paulo. She called yesterday, and answered (the phone) Paulo

In (89) we have examples of VS with three verbs, ficar 'to stay, to remain', ligar 'to telephone' and atender 'to answer the phone'. (89a), (c) and (e) show instances of postverbal subjects in declaratives, whereas (89b) and (d) demonstrate the occurrence of VS in interrogatives. These examples resemble those with unaccusative verbs, with which we can find postverbal subjects without a PP in the sentence-initial position. However, these verbs do not pass the tests for unaccusativity: (90)

a. *Ficado o aluno, a professora saiu. Stayed the student, the teacher left b. *0 aluno ficado vai esperar por sua mäe. The student stayed will wait for his mother. c. * Ligada a Ana, nos pudemos sair. Called Ana, we could go out d. *A menina ligada vaifalar com o namorado. The girl called will talk to the boyfriend e. *Atendido o Paulo, nos soubemos da novidade. Answered Paulo, we heard the news f. *0 rapaz atendido falou com todo mundo.46 The young man answered talked to everyone

As (90) shows, the verbs ficar, ligar and atender do not pass two common tests for unaccusativity. Examples (a), (c) and (e) illustrate the use of the pasticipial absolute (PA) construction, which is one possible test for unaccusativity in Brazilian Portuguese and Spanish,

138

Postverbal subjects in Brazilian Portuguese

as seen earlier in this chapter. Sentences (b), (d) and (f) show instances of the PPA (past participle as adjective) construction, which is another test used for unaccusative diagnostics. Use of these two constructions results in ungrammatically, signaling that the verbs chosen should not be unaccusatives. Here we are faced with a problem for the account of postverbal subjects proposed earlier: if only unaccusatives occurred in VS constructions (other than the locative inversion), how can we explain VS with ficar, ligar and atender? First of all, although the verb ficar does not pass the tests for unaccusativity, it "feels" like an unaccusative verb. It is present in expressions that indicate change of state, one characteristic commonly found among unaccusatives. As a matter of fact, these expressions also find unaccusative correspondents in Brazilian Portuguese. A few examples are listed below: (91)

aborrecer-se = ficar aborrecido/a 'get mad' aumentar = ficar maior 'increase, get bigger' diminuir = ficar menor 'decrease, get smaller' enfurecer-se =ficarfurioso 'get furious' entristecerse = ficar triste 'get sad' envelhecer = ficar velho/a 'get old' melhorar = ficar melhor 'improve, get better' molharse = ficar molhado 'get wet' piorar = ficarpior 'worsen, get worse'

Coupled with this characteristic of the verb ficar, the fact that it admits postverbal subjects leads us to believe that this verb might be an unaccusative, even though it doesn't pass the standard tests for unaccusativity. The verbs ligar and atender do not seem to have the "unaccusative feeling" that ficar does, but they may be going through some process that may affect their structure. Another explanation that may be offered for inversion with these verbs is related to the difference between grammatical and acceptable data. It is possible that these examples are simply acceptable, even if they are not fully grammatical. They may be undergoing the process of becoming grammatical, but

Conclusion

139

may not have reached that status yet. It is also possible that these verbs (especially ligar and atender) are mere exceptions to the rule that states that only unaccusatives in Brazilian Portuguese admit postverbal subjects (in the absence of a preposed locative preposition).

5. Conclusion Verb-subject order in Brazilian Portuguese is found with unaccusative verbs and in the locative inversion construction. In this chapter (section 2) we have discussed the Unaccusative Hypothesis (UH) and the tests used for diagnosis of unaccusativity in different languages, such as Italian and Spanish. For Italian, we find tests such as necliticization and the use of the verb essere in the participle, which seem to be clear indications of unaccucsativity in that language. For Spanish, which (like Portuguese) only has one auxiliary for the participle, the tests are different from those in Italian, and involve the use of the participial absolute and of the past participle as an adjective. We have also discussed unaccusativity in Brazilian Portuguese, applying tests to that language (basically, the same tests as used for Spanish) in order to determine which verbs can be classified as unaccusatives. We have seen that the verbs that pass the tests for unaccusativity also admit postverbal subjects (section 3). The different instances of postverbal subjects in Brazilian Portuguese are discussed in section 4. The locative inversion construction is accounted for: the PP in [Spec, TP] checks the EPP feature, which means that the subject does not need to move overtly (since it is the EPP feature that drives overt movement), and can check Case after Spell-out. We have established the difference between the locative PP and pro in Brazilian Portuguese in terms of Case: while the locative PP only checks the EPP feature, pro in Brazilian Portuguese also checks Case, thus explaining the ungrammaticality of plain inversion with unergatives (the subject of unaccusatives need not check nominative Case covertly, given the nature of the Case of unaccusatives, but the same is not true for unergatives).

140

Postverbal subjects in Brazilian

Portuguese

Postverbal subjects with unaccusatives are thoroughly discussed throughout section 4.2, which includes considerations on checking of accusative Case (4.2.1). The discussion of postverbal indefinite subjects leads to the conclusion that inherent (partitive) Case is checked in its immediate merge position, which is a theta-position (inherent Case is linked to theta-considerations). A comparison is established between pro in Brazilian Portuguese and the English expletives: in some instances, pro can be parallel to there, in others, to it. Given Lasnik's (1995) argument for Case checking by there in English, we conclude that pro must also check Case in all instances, since there seems to be agreement on the Case checking capability of the expletive it in English. The fact that pro in Brazilian Portuguese must check Case can account for the lack of VOS in this language, and also possibly for the lack of VSO, as pointed out in 4.2.3. VSO can be explained in different manners, depending on the structure assumed, but all possibilities were rejected for Brazilian Portuguese. Postverbal definite subjects are discussed in 4.2.4. The grammaticality of definite subjects in postverbal position may be related to the difference between "new" and "old" information, as proposed by Zubizarreta (1994). New information in the discourse may be interpreted as focus in syntactic terms, which would mean that the postverbal subject is focused. As for Case checking, we conclude that the Case of unaccusatives need not be partitive (since inversion is not exclusively found with indefinite subjects), though it is always inherent. We also discuss the difference in judgment found between the orders V-definite S-PP (which may not be fully accepted) and V-PP-definite S (which is accepted). Again, we reach the conclusion that this difference is related to "new" vs. "old" information. Section 4.3 shows a few verbs that admit inversion but do not pass the tests for unaccusativity. The next chapter deliberates on word order in interrogatives. As we have seen throughout this work, verb-subject order is only grammatical in Brazilian Portuguese in the locative inversion construction and in constructions with unaccusatives. In interrogatives with transitives and unergatives the order verb-subject yields ungrammaticality. This fact contrasts sharply with other languages, where a postverbal subject is mandatory in w/z-questions. In Chapter

Conclusion

141

4 we will find an explanation for those facts. We will also discuss obligatory VS in interrogatives in Italian and in (non-Caribbean) Spanish, besides French and English.

Chapter 4 Verb movement in interrogatives

1. Introduction Throughout this work we have seen that the default word order in Brazilian Portuguese is SVO. In Chapter 1 we saw that a postverbal subject yields ungrammatical results in sentences that contain a transitive or an unergative verb. (But note the exception of the locative inversion construction, which does allow for postverbal subjects in the presence of intransitives, as discussed in Chapter 3.) We also saw that interrogatives do not escape this trend in Brazilian Portuguese: a postverbal subject in a question with a transitive or an unergative verb makes it ungrammatical. In this chapter we will discuss these facts, and provide an account for them. Section 2 outlines the paradigms for interrogatives in Romance, with subsection 2.1 showing that whmovement in Romance is indeed A'-movement. Section 3 discusses Rizzi's WTi-Criterion and whether this criterion accounts for the data in Italian and the other Romance languages. In Section 4, we (re-)visit accounts for Spanish interrogatives that do not involve V-to-C movement. Section 5 ponders on questions raised by V-to-C in Romance, especially whether it actually takes place overtly. The conclusion that V-to-C is not overt in Romance is reached in subsection 5.1, which discusses the position of the verb in interrogatives. Subsection 5.2 considers the position of the subject in w/z-questions, concluding that the subject in Brazilian Portuguese and in French interrogatives raises to [Spec, TP], whereas in Spanish and Italian it remains in its base position. Section 6 briefly discusses embedded questions. Finally, Section 7 examines previous accounts of Brazilian Portuguese interrogatives, showing why those analyses were insufficient to explain the lack of inversion in interrogatives in this language.

144

Verb movement in interrogatives

2. Interrogatives in Romance: paradigms We have seen that in Romance languages there is a difference in word order between interrogatives and declaratives, except for Brazilian Portuguese, where the word order found is the same for both interrogatives and declaratives. Let us recall the paradigms for whquestions in Spanish (1), Italian (2), French (3) and Brazilian Portuguese (4)47: (1)

a. ¿Dónde trabaja Juaríi Where works Juan b. * ¿Dónde Juan trabaja? Where Juan works c. Juan trabaja en la escuela. Juan works at the school

(Spanish)

(2)

a. ...Comeparla Gianni? How speaks Gianni b. *Come Gianni parlai How Gianni speaks c. Gianni parla bene. Gianni speaks well

(Italian)

(3)

a.

Quand étudie Jean! When studies Jean b. * Quand Jean étudie! When Jean studies c. Jean étudie pour le matin. Jean studies in the morning

(French)

(4)

a. * O que leciona a Maria! What teaches Maria b. O que a Maria leciona? What Maria teaches c. Maria leciona portugués. Maria teaches Portuguese

(Brazilian Portuguese)

Interrogatives in Romance: paradigms

145

The sentences above show us that in Spanish and Italian w/z-questions the order VS is obligatory. The paradigm for French in (3), though accurate, does not show the whole picture: regarding VS, French has Complex Inversion and Subject Clitic Inversion; I will return to those phenomena later. In Brazilian Portuguese Wi-questions, however, the order found is the same as in declaratives: SV. A preverbal subject is not found in Spanish or in Italian wÄ-questions, even if pronouns are used in place of plain NPs; however, SV is still the only order possible in Brazilian Portuguese also when pronouns are used: a. * ¿Dónde él trabaja?48 Where he works b. *Come lui parlai How he speaks c. Onde eia leciona? Where she teaches *Onde leciona eia? d. Where teaches she

(Spanish) (Italian) (Brazilian Portuguese) (Brazilian Portuguese)

Thus we see that there is a difference with regard to word order in Spanish and Italian declaratives and interrogatives, but that difference is not present in Brazilian Portuguese. Note that VS may be encountered in Italian (VS, VOS, but *VSO) and in Spanish (VS, VOS, VSO) declaratives; however, w/z-questions differ from declaratives in that SV is ungrammatical in interrogatives, but not in declaratives, as shown above. Therefore, Brazilian Portuguese does not follow the general paradigm in Romance w/z-questions, since it does not allow for VS, even as an option. This is also the pattern found in yes-no questions: (6)

a. A Ana fez uma mousse? Ana made a mousse b. *Fez a Ana uma mousse? Made Ana a mousse

Again, in Spanish and Italian VS is a possible order:

146

(7)

Verb movement in interrogatives

a. b.

¿Hizo Juan la tarea? Did Juan the homework Parla bene Gianni? Speaks well Gianni

(Spanish) (Italian)

We see then that VS is not even optional in Brazilian Portuguese. The order found in declaratives in Brazilian Portuguese holds for interrogatives (but we must note the exception constituted by sentences with unaccusative verbs, which do admit VS, as discussed at length in Chapter 3). Having shown the paradigms for Spanish and Italian interrogatives, we can now turn to French. As seen in (3) above, French may exhibit VS in wA-questions; furthermore, the order wh-S-V is not allowed if the subject is a plain NP, as illustrated by (3b). However, French has peculiarities regarding interrogatives that are not present in Spanish and Italian. For instance, French does allow for a whelement in situ, as seen below, as well as for the order wh-S-V, if a subject clitic is used. The more expected wh-V-S is also found with a subject clitic (here I repeat a set of data first presented in Chapter 2, which appeared originally in Rizzi 1991): (8)

Elle a rencontré qui?49 She has met who b. Qui elle a rencontré? Who she has met c. Qui a-t-elle rencontré? 'Who has she met' a.

(8) indicates that French has several options for word order in whquestions. What we see in (8c) is an example of Subject Clitic Inversion (SCI), which can also occur in the presence of a preverbal nonpronominal DP: (9)

Depuis quand Jean est-il malade? Since when Jean is he sick

Interrogatives in Romance: paradigms

147

Another construction found in French is Stylistic Inversion (SI), which occurs when a subject non-pronominal DP is found postverbally, as in (3a). SI is also found in contexts other than root whquestions. It occurs in relative clauses, and in certain subjunctive clauses: (10)

a.

La maison où habite cet homme est très jolie. The house where lives this man is very beautiful b. Je veux que parte Paul. I want that leave Paul

Note, however, that SI cannot occur with yes-no questions, while SCI can: (11)

a. * Partira ton ami? Will leave your friend b. Partira-t-il? Will leave he

Although there are several possible orders for French interrogatives, we do not encounter wh-S-V-O, wh-V-O-S, or w/j-V-S-0 when the subject is a non-pronominal DP: (12)

a. * Quand Marie écrit des lettres? When Marie writes letters b. * Quand écrit des lettres Marie? When writes letters Marie c. * Quand écrit Marie des lettres? When writes Marie letters

Even in French, where we find a less straightforward picture with regard to word order in interrogatives, the syntax of questions and the syntax of declaratives are quite different, illustrating once again that in Romance languages there appears to be some kind of special interrogative syntax. But Brazilian Portuguese is an exception to that tendency: word order in declaratives and interrogatives is indeed the same.

148

Verb movement in interrogatives

With the paradigms for interrogatives laid out, we can try to analyze the data and to find an answer for why Brazilian Portuguese differs from other Romance languages. Before doing so, however, we will take a closer look at w/z-movement, in order to establish the kind of movement that takes place in Romance w/z-questions.

2.1. Diagnostics for Wh-movement In the various Romance languages under study here, the w/z-phrase moves to [Spec, CP] in interrogatives, showing the characteristics of A'-movement. We can attest w/z-movement with a few diagnostics. Let us start with island constraints, with Spanish examples in (13): (13)

a.

Juan me dijo [cp cuandoi [TP vio a Maria tjJJ. 'Juan told me when he saw Maria.' b. Juan me dijo [epa quién¡ [τρ vio t¡ la semana pasada]]. 'Juan told me who he saw last week.' c. [cp A quién¡ [TP te dijo Juan [cp que [TP vio t¡]]]? 'Who did Juan tell you that he saw?' d. *[CP A quién¡ [TP te dijo Juan [CP cuándoj [TP vio t¡ tj]]J? Who did Juan tell you when he saw?

In (13d) A quién 'whom' is extracted from an embedded question, which is introduced by a moved element cuando 'when'. Comparing (13d) with the grammatical (13c), we can see that movement is blocked by the presence of the w/z-element cuando, leading to the conclusion that w/z-questions are islands for movement. Another constraint on w/z-movement is the superiority condition, which restricts extraction of w/z-elements: XP is superior to YP if XP and YP are in the same IP and XP c-commands YP (definition in Haegeman 1991). This effect can be seen in the English and Brazilian Portuguese examples below (identical meanings): (14)

a. * What did who seel b. *0 que quern viu?

Interrogatives in Romance: paradigms

149

As (14) shows, subjects are superior to objects, and extraction of a w/z-object over a w/z-subject yields ungrammaticality. For Spanish, it has been argued that superiority effects are voided by the subject being able to be in postverbal position (Jaeggli 1982), but even this claim is not uncontroversial. The Spanish translation of (14) seems to be only marginally accepted: (15)

? ¿Qué vio quién? What saw who 'What did who see'

Even if (15) can be accepted to a certain degree, its status (as a question mark) still shows that superiority effects exist in Spanish too. It has also been argued that wA-elements license parasitic gaps, which are null elements whose existence depends on the presence of another null element. The following Brazilian Portuguese example can illustrate the occurrence of parasitic gaps: (16)

Paulo é um homem [cp em quern [TP voce confia EI [cpquando [cp voce conhece e2]]]. 'Paulo is a man whom you trust when you meet'

The null elements ei and e2 in (16) occur in a position in which they check accusative Case. The trace ej would naturally be coindexed with the w/z-element em quern 'whom', and ez, a parasitic gap, is licensed by ei. Licensing of parasitic gaps is acknowledged as one of the diagnostics for w/z-movement, and as seen in (16), w/z-phrases do indeed coindex with traces that license parasitic gaps in Romance (with Brazilian Portuguese as an example). Another characteristic of w/z-movement is long movement: the w/z-element can move cyclically and be extracted from several clauses. Consider the following Brazilian Portuguese example: (17)

Quern¡ [τρ voce acha [cp t¡ que [χρ o Ivo disse [cp t¡ que [τρ pro lele ia convidar t¡]]]]]? Who you think that Ivo said that he would invite

150

Verb movement in interrogatives

In (17), we can see that the wA-element quem 'who' has cyclically moved up from the lowest clause via [Spec, CP], thus characterizing A'-movement. With the Spanish and Brazilian Portuguese examples in this section, we see that it is well established that movement of the whphrase itself is A'-movement to [Spec, CP]. With the knowledge of what kind of movement happens in Romance w/z-questions, we can now turn to the study of subject-verb inversion and lack thereof in Romance interrogatives. At this point, we know that VS may obtain due to one of two possibilities: either the subject does not raise to [Spec, TP] (which, presumably, is its landing site in SVO declaratives), or the verb moves higher than Τ - although this higher position might not be C. Either way, Romance languages in general seem to exhibit some type of interrogative syntax which is not found in Brazilian Portuguese. In the next section, we will start by recapitulating Rizzi's ^ - C r i t e r i o n and the assumptions that it entails.

3. The ^-Criterion As discussed in Chapter 1, Rizzi (1991) offers an account for inversion in w/i-questions in English and Italian by proposing the WhCriterion50, which I repeat here: (18)

The PFA-Criterion a. A PF/z-operator must be in a Spec-head configuration with anX°[+wH]· b. An X°[+WH] must be in a Spec-head configuration with a WA-operator.

First of all, it must be emphasized that the part of the Criterion above that is of relevance for this chapter is the (b) clause; the first part of the Criterion, the (a) clause, is what forces (for Rizzi) movement of the w/z-phrase itself. For the Criterion to be satisfied, there must be Ito-C movement in English and in Italian. This movement would be necessary because I is assumed to carry the [+wh] feature. Thus, the verb first moves to I (or T, in a more articulated structure) and then to

The Wh-Criterion

151

C, satisfying the Criterion. As we saw in Chapter 1, Rizzi claims that this criterion is responsible for the S-Structure distribution and for the LF interpretation of Wh-operators. We also saw that Rizzi allows for a 'weaker' version of the ^-Criterion, which he calls dynamic agreement, as opposed to the static agreement that would be the default form through which the Criterion would be met. Rizzi proposes that dynamic agreement accounts for the optionality found in French, where Wi-questions may exhibit the same order as in Italian (Wz-VS), or Brazilian Portuguese (w/z-S-V) in the presence of a subject clitic. The problem then, as discussed before, is that in Brazilian Portuguese the ^-Criterion would never be met through its default form, but only by means of dynamic agreement. If we are to adopt Rizzi's criterion, we will also be left with the question why it cannot be met through its default form in Brazilian Portuguese. There are other problems for Rizzi's ^-Criterion, which is based on the assumption that V moves to C in wft-questions. Overt Vto-C is assumed not to occur in Spanish, as argued by Suñer (1994), and discussed in Chapter 1. Recapitulating what is said in that chapter, Suñer uses the position of phrasal adverbs to show that V-to-C does not take place in Spanish: if V moved to C in questions, these adverbials would follow V when a wA-phrase is in [Spec, CP]. As the example below shows, the adverbials do not appear after the verb in the presence of a w/z-phrase in Spanish (examples from Suñer 1994): (19)

a

¿A quién jamás ofenderías tú con tus acciones? (Span) Whom never would-offend you with your actions 'Who would you never offend with your actions?' b. ¿Qué idioma todavía estudia Pepita? What language still studies Pepita

Although the adverbial data seem to indicate that V-to-C does not occur in Spanish, in Italian the picture appears to be different. Adverbs exhibit a different distribution in Italian than they do in Spanish: (20)

a. Maria spesso parlava di lui. Maria often spoke of him

(Italian)

152

Verb movement in interrogatives

b.

Di chi parlava spesso Mariai Of whom spoke often Maria c. *Di chi spesso parlava Maria? Of whom often spoke Maria

In (20a) we find the adverb in preverbal position in an affirmative sentence. In (20b), a w/z-question, the adverb follows the verb, and it cannot appear before the verb, as is clear from the ungrammatical (20c). Since the adverb must surface postverbally in a w/z-question, it would seem that V-to-C in Italian does take place, although it does not in Spanish (or Brazilian Portuguese). Let us discuss for a moment what may trigger V-to-C movement, if it actually exists. According to Rizzi's claim, V-to-C would be motivated by a criterion: the movement would happen so that the w/z-element and Τ (or I, in Rizzi's terms) could be in a Spec-head configuration, since he assumes that Τ has a [+wh] feature. This way, the verb would move to C and the ^-Criterion would be met. However, movement of the verb to C could be driven by a factor other than a criterion: it might be the case that this kind of movement happens because of feature checking requirements, which would conform to a minimalist account of the facts. In Minimalism, movement happens before Spell-out if a given feature is strong and needs to be checked prior to LF. Thus, it could be the case, if V-to-C indeed occurs in Italian, that the V-feature in a wh C° in this language is strong, which would require that it be checked overtly. If it is checked by the verb moving to C, then that movement would take place before Spellout, giving us the surface result we see in a sentence such as (20b). At this point, it should be noted that there are different assumptions with regard to the locus of the w/z-feature: Rizzi and Chomsky, for example, assume distinct sites for the feature. Rizzi believes that the w/z-feature originates in I: he proposes that among other specifications, the main inflection can also be specified as [+w/z]. For Chomsky (1995b), on the other hand, it is the w/z-feature in C which forces movement of the wft-phrase itself to [Spec, CP], an effect that is achieved by Rizzi via movement of the inflected verb (V+I) to C. Chomsky argues that the feature Q, one of the formal features of the functional category C, is plainly Interpretable, and

The Wh-Criterion 153 needs to be checked overtly only if it is strong, as is the case in English (the strong Q feature is satisfied by a feature FQ). For English questions, then, there may be raising of I-to-Q, and if this is raising of a verbal feature, then FQ in this case is V. However, if the ννΛ-clause is embedded, I-to-Q raising is impossible, but the w/z-phrase may raise overtly to the embedded [Spec, Q], yielding the following example (from Chomsky 1995b): (21)

they remember [which book Q [John gave t to Mary]]

For Chomsky, even in a main clause V+I moving to C will not check FQ (since this movement does not eliminate the need for the whelement itself to move), which leads him to argue that I-to-C in interrogatives happens due to phonological reasons. Thus, we notice that the assumptions by Rizzi and by Chomsky regarding the site of the wÄ-feature are indeed different, having distinct consequences in terms of verb movement in interrogatives. Rizzi argues that movement of V+I to C is responsible for w/z-movement to [Spec, CP], whereas for Chomsky (1995b) w/z-movement happens due to the w/z-feature in C. Returning to Rizzi's WTi-Criterion, another problem is pointed out by Ordóñez (1997): Rizzi proposes that V-to-C in interrogatives destroys the Spec-head configuration necessary for Case assignment (Case checking) to the subject in [Spec, IP]. Thus Rizzi's claim entails that nominative Case is alternatively assigned (or checked, in Minimalism terms) by Tense to the right. Ordóñez notes that this idea does not accord with a Minimalist perspective: according to Chomsky (1995b), once a configuration for Case is obtained at a certain stage of the derivation, it cannot be undone later. Nevertheless, Rizzi's account could be made compatible with this idea, which means that this particular criticism does not really weaken the essence of the WhCriterion approach: I/T can check nominative Case on the subject in [Spec, TP], and Τ can then move to C (conversely, the DP in [Spec, TP] checks the [-Interpretable] Case feature on T). As formulated, the PPTz-Criterion presents several drawbacks, as we can see from the discussion above, and one major problem is the obligatory V-to-C that the Criterion entails. In section 4.5 we will discuss questions that are raised by the hypothesis that there is

154

Verb movement in interrogatives

movement of the verb to C. Interrogatives in Spanish have been analyzed without the assumption that V-to-C occurs, as discussed in Chapter 1. Once again, let us recapitulate that discussion.

4. Accounts for Spanish: no V-to-C In Chapter 1, several analyses for obligatory VS in Spanish interrogatives were discussed. Sufier (1994), for example, proposes the AAL (Argumentai Agreement Licensing), a condition that requires that [Spec, CP] agree with C with regard to the feature [+ argumentai]. The consequence of that condition is that no argumentai element, such as a subject, could intervene between C and V if there were an argumentai wA-phrase in [Spec, CP]. Sufier's analysis encounters a few problems. Besides the ones already discussed in Chapter 1 (such as not being able to account for varieties of Spanish, such as Puerto Rican Spanish, where SV is indeed found in wA-questions), Ordóñez (1997) reinforces the fact that Sufier's approach is problematic crosslinguistically, since it would not be clear why subjects in [Spec, TP] in English or German would not have the same blocking effects, given the grammaticality of basic English sentences such as (22): (22)

What did you do?

Problematic as it may be, Sufier's analysis does bring to light relevant facts regarding (lack of) V-to-C in Spanish. Her view that VS in Spanish is not due to V-to-C is shared by other authors as well. According to Contreras (1991), there would be different structures for Spanish and English SV sentences. In Spanish, a preverbal subject would be adjoined to TP. Recall that Contreras proposes that a closed domain cannot contain any unlicensed elements. A preverbal subject that were adjoined to TP, having moved to that position, would violate Contreras's CDC (Closed Domain Condition) in a wA-question: the A'-chain consisting of the wh-element and its trace constitutes a closed domain that would not tolerate the presence of the preverbal (adjoined) subject.

Accounts for Spanish: no V-to-C

155

Goodall (1993) argues that in w/i-clauses in Spanish the whelement must move into [Spec, IP] (or [Spec, TP]) before rising to [Spec, CP]. This movement would prevent the subject from rising to [Spec, TP]. This means that [Spec, TP] is a potential A'-position in Spanish. Besides explaining obligatory VS in Spanish wA-questions, Goodall's account makes a few predictions. First, with this account we would expect that in clauses that are [-wh] but in which there is movement (such as relative clauses), the wA-phrase would move directly to [Spec, CP]. The prediction is borne out, as Goodall shows with the lack of contrast between (23a) and (23b): (23)

a. Esta es la muchacha que todos conocemos. This is the girl who all we-know b. Esta es la muchacha que conocemos todos. This is the girl who we-know all

Since the w/j-phrase does not move through [Spec, TP], the subject {todos) is free to raise to that position, as in (23 a). Another prediction made by Goodall's analysis is that if the clause is [+wh\ but there is no wA-movement, the subject can move to [Spec, TP]. Goodall tests this prediction in two cases: with wh-words such as why, and with complementizers of the whether class, which we show here. These complementizers appear to be [+wh], as in (24): (24)

a. No sé si Juana se fue a Albuquerque. No I-know whether Juana went to Albuquerque b. No sé si se fue Juana a Albuquerque. No I-know whether went Juana to Albuquerque Ί don't know whether Juana went to Albuquerque.'

Goodall argues that the prediction made by his analysis is correct, in that the subject, by his analysis, has moved into [Spec, TP] in (24a). We have seen that the preverbal subject in Brazilian Portuguese raises to [Spec, TP] (the equivalent to Goodall's [Spec, IP]) for EPP and Case checking, indicating that in Brazilian Portuguese the [Spec, TP] position is an A-position, not an A'-position. But here what we need to highlight is that the three accounts (Sufier 1994,

156

Verb movement in interrogatives

Contreras 1991 and Goodall 1993) assume no V-to-C in Spanish; rather, VS would be due to the subject not moving to the position that it would possibly occupy in SV declaratives. Ordóñez (1997) provides a more recent account for the obligatory VS in Spanish w/z-questions. He follows Kayne's (1994) antisimmetry proposal and proposes an analysis based on the idea that Spanish (and Catalan) preverbal subjects are in [Spec, TopP]. According to Ordófiez, the head Top conflicts in features with the whrequirements of the w/z-element. He argues that preverbal lexical subjects and pro occupy different positions in Spanish and Catalan, assuming that lexical DPs are in the Spec of a topic projection. However, Ordóñez does not take an A' minimality approach, which might at first sight account for the ungrammaticality of preverbal subjects in w/z-questions in Spanish and Catalan: if subjects in these two languages were in an A'-position, he states, it might be possible to account for such ungrammaticality in terms of A'-minimality. Preverbal subjects in Spanish and Catalan occupy a more peripheral position, Ordóftez argues, than subjects in English, and from this perspective, topics in A'-position would block the movement of the wh-word to [Spec, CP], thus explaining the ungrammaticality of SV in Spanish and Catalan w/z-questions. Therefore, the subject can not occupy a preverbal position, and will necessarily be found postverbally. This explanation would also account for why lexical subjects, but not pro, have a blocking effect in interrogatives. But he points out that such an approach alone would be too simple: this idea would contrast with Torrego's (1984) facts. Torrego shows that, in Spanish, the extraction of a wA-word over a preverbal subject is in fact available in embedded declaratives where long distance w/z-movement has taken place. If a subject in topic position were blocking extraction of the w/z-word, there would be no contrast between (25a) and (25b): (25)

a.

Qué pensabas que [χ0ρρ la revista [ip había publicado]]? What you thought that the magazine had published b. *Qué [χορρ la revista [ip había publicado]]? What the magazine had published

Accounts for Spanish: no V-to-C

157

Ordóñez argues that the contrast shown in (25) indicates that the blocking of the lexical DP subject only happens in the clause where the wh-element lands. He claims that there is a local principle (an agreement relationship) that explains obligatory VS in Spanish whquestions, in the spirit of Rizzi (1991). Nevertheless, his approach differs crucially from Rizzi's: Ordóñez proposes that this agreement is with the complementizer, not with the verb. This agreement relationship, according to him, involves an empty complementizer in Spanish and the auxiliary do in English: (26)

[cp 0%O[+wh] did[+wh] [ip he I [Vp visit t]]]?

(27)

[cp A quién[+wh¡ [co [+wh] [IP pro [p visitó t]]]]?

Examples (26) and (27) illustrate Ordóñez's agreement relationship. According to him, topics would interfere with this relationship, and if Spanish preverbal lexical subjects occupy a topic position, as he claims, then they cannot appear in wA-questions. He argues that pro, in contrast, does not occupy a topic position in Spanish and Catalan, which would account for why lexical subjects, but not pro, have a blocking effect in interrogatives. A lexical subject in TopP creates a mismatch between the complementizer and the wA-element in CP. For that argument to work, Ordóñez adopts a CP recursion approach. Furthermore, to explain the grammatical presence of preverbal subjects in embedded declaratives, he argues that the empty complementizer of a matrix interrogative and the complementizer in an embedded sentence start off in different places: the latter would start off in the lower CP shell and enter in a Spec-head relation with a topic XP, while the former, not being part of a topic, would acquire its feature by direct agreement with the w/z-feature of the XP moved to [Spec, CP] (traces left by w/2-elements are assumed to be [-wh]). This difference is illustrated below: (28)

a. Quéjpensabas [tj[.wh] gwe,[-wh] [TopP Ia revista ti [ip había publicado tj]]

158

Verb movement in interrogatives

CP Quéj

C

0[+wh] I P

pro

I'

pensabas

CP tj[-wh]

C'

que ¡[.Wh]

TopP

la revista

Top' Top

CP t«[-wh]

IP

había publicado tj (29)

a. b.

[Qué [+wh]

0[+wh]

[iP había publicado la revista]]?

CP Qué[+Wh]

C'

0[+wh]

IP

había publicado la revista As we can see in (28) and (29), Ordóñez accounts for the presence of a pre verbal subject in Spanish embedded sentences and the obligatoriness of postverbal subjects in w/z-questions by means of feature matching: the embedded que 'that' carries a [-wh] feature, whereas

Accounts for Spanish: no V-to-C

159

the emptv complementizer in a matrix question is [+wA]. He assumes that Top is always specified as [-wh], which does not bring problems to his analysis of the embedded declarative in (28a) because he follows Lasnik and Saito (1992), who argue that traces left by whelements are [-wh]. He concludes that the matrix/embedded asymmetry reduces to a difference between wA-traces being [-wh] and elements/phrases being [+wA]. His account also assumes, like the others mentioned above, that VS in Spanish interrogatives obtains because the subject does not move to its usual preverbal position in SVO, not because the verb moves higher than the subject: the verb is assumed, in all these analyses, to raise just as high as, but no higher than, it does in declaratives. In an earlier work, Ordófiez (1996) takes one more step regarding the fFA-Criterion. He still assumes that Spanish does not have V-to-C, and that lexical subjects in SV occupy a more external position in Spanish than in English. He also maintains that there is local wA-agreement, as in Rizzi (1991). But he adopts the idea that this agreement can be satisfied by movement of the verb to C at LF, as proposed by Sportiche (1998) for complex inversion in French. Therefore, Ordóñez (1996) argues that the agreement relation is obtained overtly in English (as in example [26] above), and covertly in Spanish (as in [27]). If the supposition in Ordóftez (1996) regarding V-to-C is correct, we need to address a few questions raised by the very assumption that V-to-C does indeed take place.

Summary To summarize what has been discussed so far in this chapter, we have seen the paradigms for interrogatives in Romance. For wA-questions, we see that the paradigms vary among Spanish, Italian, French and Brazilian Portuguese. In Spanish and Italian, as exemplified in (1) and (2), we find obligatory VS order. A priori, French patterns with Spanish and Italian, at least when a non-pronominal DP subject is present, as in (3). However, SV is also a possibility in the presence of a subject clitic, as attested in example (8) (which also shows that French allows for the wA-element to remain in situ). We have also

160

Verb movement in interrogatives

seen examples of Stylistic Inversion and of Subject Clitic Inversion in that language. Brazilian Portuguese, unlike the other Romance languages studied here, does not allow for VS in interrogatives (unless the verb belongs to the unaccusative class, see Chapter 3). After looking at the paradigms for interrogatives in Romance, we looked at w/i-movement, in an effort to determine the type of movement that happens in Romance interrogatives. We have seen that the four languages exhibit w/z-movement to [Spec, CP] (even though the w/z-element may optionally remain in situ in French and Brazilian Portuguese). This movement was attested to by means of diagnostic examples, such as constructions exhibiting w/i-island constraints (as in Spanish [13]), the superiority condition (shown in [14]), or parasitic gaps (which appear in examples [16] and [17]). Next, we revisited Rizzi's (1991) Wh-Criterion, according to which there is I-to-C movement in English and in Italian interrogatives. However, some problems for the Criterion were pointed out, such as the well-accepted view that there is no V-to-C in Spanish (cf. Suñer 1994). In light of the discussion of the JfTz-Criterion and of Vto-C movement, we also revisited the accounts for Spanish VS order that do not assume that type of movement. Having recapitulated those accounts, we will proceed to the discussion of some questions raised by the possibility of the occurrence of V-to-C.

5. Questions raised by V-to-C As we have seen, the adverbial data in Italian seem to indicate that the verb moves past Τ in interrogatives (perhaps V-to-C). If in fact this movement happens (overtly) in Italian (and, presumably, in French when VS is found in interrogatives), we need to discuss a few questions. The first issue regarding possible overt movement of the verb to C has already been mentioned here: is this movement driven by feature checking needs, or is it driven by a criterion? If V-to-C is a feature-driven movement, it would be in accordance with the assumptions in the Minimalism framework: feature checking is the operation that drives movement (Chomsky 1995b). In that case, the V-feature in C° (the feature that presumably forces V-

Questions raised by V-to-C

161

to-C) might be strong in Italian, if we find evidence for overt V-to-C, and weak in Brazilian Portuguese, where V-to-C does not appear on the surface. Presumably, this feature would also be weak in Spanish, given the arguments for lack of V-to-C in that language (Suñer 1994, Ordóñez 1997). If one assumes that this feature is weak in one language (Spanish and, presumably, Brazilian Portuguese) and strong in another (Italian and, perhaps, French), the accounts for VS in interrogatives in those languages would be different. However, one may assume that movement of the verb to C is driven by a criterion; in that case, languages that do not exhibit this movement on the surface might either satisfy the criterion covertly (as in Ordóñez 1996), or follow Rizzi's (1991) original assumption for French: that dynamic agreement takes place, and the criterion is met even without movement of the verb to C. The second question that we need to raise is one that challenges the assumption in the first question: does V-to-C actually take place in Italian and French? The data containing adverbs in Italian do appear to indicate that the verb moves higher in interrogatives than it does in declaratives. The relevant paradigm is repeated here for convenience: (30)

a.

Maria spesso parlava di lui. Maria often spoke of him b. Di chi parlava spesso Mariai Of whom spoke often Maria? c. * Di chi spesso parlava Mariai Of whom often spoke Maria

(Italian)

(30) shows adverb placement in declaratives and in w/2-questions. We can notice that, while the frequency adverb precedes the verb in an affirmative sentence, it appears postverbally in an interrogative, which would suggest that the verb has moved to a higher position than it occupies in the declarative (30a). Now the question to be asked is: is that higher position necessarily C? It would not be implausible to think that the verb moves higher, but to a projection other than C. One of Ordóñez's (1996, 1997) criticisms of the ^-Criterion is that V-to-C does not conform with Kayne's (1994) antisimmetry

162

Verb movement in interrogatives

proposal, particularly with regard to the position of clitics. Ordófiez argues that, according to Kayne's LCA (Linear Correspondence Axiom), the pro-clitic in a sentence such as (31) cannot be adjoined to the same inflectional projection as the verb: (31)

Juan le escribió. Juan to-him wrote

(Spanish)

Ordófiez explains that the LCA bars multiple adjunction to the same head; therefore, pro-clitics must be adjoined to an inflectional projection above the landing position of the verb. So, if the verb moves past the clitic to C the order predicted in interrogatives is V-cl. But that is not the order that we find: (32)

a.

¿Quién le escribió? Who to-him wrote b. * ¿Quién escribióle? Who wrote to-him

(Spanish)

Since V-cl does not obtain in Spanish interrogatives, as seen in (32) where the clitic le must appear preverbally, the verb would not have moved to C, if one is to follow Kayne's (1994) antisimmetry proposal. If Ordóñez's argument regarding Kayne's LCA and lack of Vto-C holds for Spanish, it should also be valid for Italian, since the order cl-V is also found in Italian interrogatives, as observed by Zanuttini (1997): (33)

a.

Cosa gli diceva Gianni? What to-him said Gianni? b. *Cosa dicevagli Gianni? What said to-him Gianni?

(Italian)

Thus we see one more reason why V-to-C may not take place, or, at least, needs to be questioned. Although this particular argument is valid only if one follows Kayne's proposal, it is worth noting that there exists one more argument against V-to-C, and that it seems to hold for Italian as well as Spanish.

Questions raised by V-to-C

163

The view that the verb does not move to C overtly in interrogatives is corroborated by Sportiche (1998), as briefly mentioned above. In discussing complex inversion and subject clitic inversion in French, Sportiche concludes, against previous analyses of the same phenomenon (such as Kayne 1984, or Rizzi and Roberts 1989), that overt V-to-C does not occur in French. Rather, he argues, V-to-C takes place covertly. The evidence he uses against overt V-to-C involves predictions concerning what he calls "IP initial material". If there is V-to-C, that kind of material would surface to the right of the verb+subject clitic (V+SC1) complex, whereas lack of overt V-to-C predicts that it appears to the left of the V+SC1 complex. Here I give Sportiche's example to support lack of overt V-to-C in French. The position of the temporal clausal adjunct in (34) and in (35) is crucial to his argument: (34)

a.

Quand le vote a eu lieu les électeurs sont allés a la pêche. When the election took place the voters went fishing b. ULes électeurs sont quand le vote a eu lieu allés a la pêche. The voters did when the election took place go fishing

(35)

a. Dans quelle ville quand le vote a eu lieu (les électeurs) sont-ils allés a la pêche? In which town when the election took place (the voters) did-they go fishing b. 17Dans quelle ville les électeurs sont-ils quand le vote a eu lieu allés a la pêche? In which town the voters did-they when the election took place go fishing

In (34) we see that the IP initial material (the phrase in italics) can appear before the subject in a declarative sentence, but only very marginally can it surface after the inflected verb. (35) is a whquestion, and it indicates that the same results obtain in interrogatives: the material can indeed appear to the left of the V+SC1 complex, as in (35a), but not to the right, as in (35b). Thus, Sportiche ar-

164

Verb movement in interrogatives

gues, the prediction of an analysis with no overt V-to-C is borne out, whereas the prediction of the analysis involving overt V-to-C is not. Therefore, he concludes that V-to-C does not take place overtly. However, we have data involving adverbs in Italian and Spanish that seem to indicate that the verb in Italian interrogatives does move higher than it does in Spanish questions, as seen in Spanish (19) and Italian (20) (repeated later as [30]). The Spanish data seem to indicate that the verb does not raise to C in interrogatives, since the adverbs appear in preverbal position. A priori, the data for Italian appear to indicate the opposite, i.e., that the verb does raise to C, for the adverb surfaces after the verb in a w/z-question, a result that is expected if there is V-to-C. However, there does not seem to be enough evidence to determine that the position to which the verb moves is indeed C. In studying Romance negation facts, Zanuttini (1997) follows Kayne's (1994) assumptions and states that it is not possible to argue that the w/z-operator is in the specifier of the head in which the finite verb occurs (based on negation and clitic facts). She goes on to argue that in w/z-questions (examining Italian dialects) the verb moves to a head other than the one that has the w/z-operator in its specifier. If this is true, she proceeds, then the verb does not move to C, but to a lower head. Let us suppose, with Zanuttini (1997), that the verb in Romance interrogatives does not go up to C. This assumption would suggest that the relation between verb movement and the presence of a wh-element may not be so direct as it is usually believed: the verb would move to a lower head, not to C. Thus, if there were indeed some type of checking relation between the verb and the w/z-operator, it would not be one that needed to be checked in a Spec-head configuration, as in a Criterion. As will be argued here, inversion in interrogatives is not accounted for by a criterion, but by feature-checking relations, which involves head adjunction checking rather than Spechead agreement. Summarizing our questions, we see that the hypothesis that Vto-C takes place is disputed: supposing the movement occurs, what drives it, a requirement based on a criterion, or feature-checking relations? And if it does, can it happen before or after Spell-out? But does it occur indeed? So far we seem to have evidence that the verb in

Questions raised by V-to-C

165

Italian does move to a position higher than it does in declaratives, and that that movement does not take place in Spanish. Let us now turn to a more detailed examination of the positions of the verb in interrogatives in Romance.

5.1. Verb position in interrogatives In Italian and Spanish, the order wh-S-V is disallowed51. Whether the verb moves to C or not, what seems clear is that there is some type of relation between the verb in interrogatives and the w/z-operator, and the subject cannot interfere with that relationship. There are two options that may account for this fact: either 1) the subject cannot move up to the position where it does in SVO declaratives (presumably, [Spec, TP]); or 2) the verb moves higher than it does in declaratives (but not to C). Several analyses of Spanish word order argue for the first option (Contreras 1991, Suñer 1994, Goodall 1993, Ordófiez 1996, 1997). Since there seems to be evidence that the verb in Spanish does not move any higher than it normally does (as argued for in Suñer 1994 and corroborated in Ordóñez 1997), we are led to the conclusion that the Spanish subject in interrogatives does not move as high as it potentially does in SVO sentences, thus refraining from intervening with the relationship between the verb and the w/i-operator. As we saw in Chapter 3, it is possible for the Spanish subject in declaratives to remain in its base position and have Case checked covertly, since pro in that language will check the EPP feature (but not nominative Case). Thus it is possible that in w/2-questions, as I will argue below, the wÄ-element may check the EPP feature, still making it possible for the subject not to raise to [Spec, TP]. In Italian, the postverbal subject in questions could be a result of the same process that possibly takes place in Spanish: the subject could stay in its base position as well. The data that suggest that there is V-to-C in Italian are not conclusive. Rizzi (1991) argues for V-to-C based on the obligatoriness of VS and the impossibility of SV in whquestions, as in (2a) and (2b) above. That fact alone, however, would not necessarily mean that the verb has moved to C, since in Spanish

166

Verb movement in interrogatives

we find similar facts, even though there is evidence that the verb does not move to C. Note that Rizzi does not include any data with adverbs to conclusively establish V-to-C movement. Rizzi also argues for movement of I (or T, in our terms) to C based on hypothetical clauses in Italian, which, as seen in Chapter 1, parallel English hypotheticals: the complementizer se 'if can only be dropped if there is a postverbal or a null subject. I repeat here the data from Chapter 1 (which appeared originally in Rizzi 1991): (36)

a. * (Se) Gianni fosse arrivato, tutti sarebbero stati contenti. (If) Gianni had arrived, everybody would have been happy b. (Se) fosse arrivato Gianni, tutti sarebbero stati contenti. (If) had arrived Gianni, everybody would have been happy c. (Se) fosse arrivato in tempo, Gianni sarebbe stato contento. (If) had arrived in time, Gianni would have been happy

Rizzi argues that se can be replaced with the inflected verb moved to C: if that movement does not take place, se must be present in the sentence, otherwise ungrammaticality will result, as in (36a). However, the fact that V-to-C might occur in the Italian construction above is not necessarily evidence that it also occurs in interrogatives. Furthermore, (36b) has the subject appearing after the past participle, not immediately after the tensed verb, as would be the case in English: (37)

Had John arrived on time, everybody would have been happy.

For English, one may argue for the existence of V-to-C in (37): the tensed verb is in C, and the subject (supposedly in [Spec, TP]) appears immediately after the inflected verb. In Italian, however, the subject surfaces after the past participle. The past participle may be assumed to be in AspP, as in Belletti (1990), or it may be assumed to be able to raise to a head above certain Adverb Phrases, according to Cinque (1999). Either way, the projection hosting the past participle is below TP, which is the projection to which the verb moves in (nonhypothetical) declaratives. Therefore, if the past participle is located

Questions raised by V-to-C

167

below TP, the sentences in (36) could be an indication that the subject in (36b) does not raise to [Spec, TP], rather than constitute evidence that the tensed verb moves to C. Furthermore, the auxiliary verb and the past participle in Italian are most likely separated by several functional projections, given the grammaticality of Aux-Adv-Past Participle in that language. This fact would constitute one more piece of evidence against V-to-C in Italian. Thus we see that the arguments for V-to-C in Italian (obligatory VS in interrogatives and hypothetical sentences) are not at all conclusive: it is possible that in both cases the subject stays in a lower position. The Italian data do not offer much clarity regarding the position of the verb in interrogatives. For example, the data involving frequency adverbs in Italian in (20)/(30) do seem to indicate that the tensed verb moves higher than it does in declaratives. However, there are a few factors to be taken into consideration here. First, as mentioned before, and following Zanuttini (1997), it may be the case that the verb does indeed move higher than T, but not to C. Another issue would be related to interrogatives: it might be possible that nothing can come between the wA-element and the verb in Italian, along the lines of Ordóñez's (1996) account for VS in Spanish interrogatives: he argues that the (topicalized) subject cannot interfere with an agreement relation between the w/z-phrase in Spanish. In Italian this agreement relation could be such that nothing could intervene between the verb and the wA-element. This possibility would require further investigation that is beyond the scope of this work. Here I am simply trying to demonstrate that the data in (20)/(30), which show the impossibility of a preverbal adverb in Italian interrogatives, do not necessarily prove that there is V-to-C in that language. A third possible explanation for the data in (30) explores the position of the adverb. If we assume that the frequency adverb spesso 'often' is adjoined to Asp' (thus below TP), the verb in a declarative sentence (such as [30a]) does not need to move past Asp (although it is obviously possible that the verb does move to T). Therefore, (30c), repeated here, minimally indicates that the verb has moved higher than Asp, but does not necessarily show that it has moved higher than T:

168

(38)

Verb movement in interrogatives

*Di chi spesso parlava Maria?

One more fact that casts doubt on the existence of overt V-toC in Italian interrogatives is the ungrammaticality of VSO in that language. If the verb indeed moved to C in w/z-questions, we would expect to find the order wh-Y-S-O, since we would assume that the subject occupies the same position in interrogatives as it does in declaratives. However, this is not what we encounter in Italian; rather, the order w/j-V-O-S is what is found: (39)

a. *Come ha risolto Gianni il problema? How has solved Gianni the problem b. Come ha risolto il problema Gianni? How has solved the problem Gianni

Once again, there is evidence that the verb does not move to C overtly in interrogatives in Italian. Returning to the possibilities mentioned in the beginning of this section regarding obligatory VS in interrogatives (i.e., either the subject stays in its base position, or the verb moves to C), we cannot be sure that the verb moves to C in Italian (in fact, there is evidence that it does not), and we know that it does not in Spanish. Even if the verb in Italian interrogatives does move higher than T, there is nothing that proves that it moves to C. Therefore, we can work with the assumption that the verb in Italian does not move as high as C, at least not overtly. As for Spanish, it is clear that the subject remains in a lower position in interrogatives than it does in SVO declaratives. Exploring the possibility that the verb in Italian actually moves beyond T, but not as far as C, we would need to ask what position it moves to. Raposo (1994) postulates one possible position. He argues for the existence of an F (for 'Functional') node above IP but below CP. In a more articulated structure, Raposo's FP would be located above TP. Thus F appears, at first sight, to be a plausible position for the verb to land. As a matter of fact, one of Raposo's arguments for the existence of morphological F in European Portuguese (EP) is the fact that F would provide an account of enclisis (postverbal clitics) in EP. He adopts the idea that clitics left-adjoin to Infi or

Questions raised by V-to-C

169

to the highest category resulting from the split of Infi (in this work, that category is T). According to him, when the verb raises to F, it leaves the clitic behind (in Infi), resulting in enclisis. Take the following example: (40)

(European Portuguese)

a.

O Pedro deu-lhe o livro.

b.

Pedro gave to-him the book FP O Pedro

F'

Ihe

Τ

ts tv

o livro

tv

In (40), Raposo's analysis places the verb in F, while the clitic is leftadjoined to Τ (as shown in [40b]), resulting in enclisis. He argues also that preverbal clitics (proclisis) would follow from the verb not raising to F. We must note here that Raposo's analysis, adjoining the clitic to Τ which itself contains the trace of the raised V+T, goes against Kayne (1991), who argued that clitics cannot adjoin to heads that are traces. Enclisis does not occur in simple sentences in the other Romance languages though. Therefore, we need to look again at the argument based on clitics. The order clitic-verb, which obtains in Spanish, Italian, European Portuguese, and Brazilian Portuguese whquestions, would yield if the clitic adjoined to a functional head above the one that hosts the verb. If the verb moved past Τ to F in whquestions, we could expect the order verb-clitic. However, enclisis does not obtain in w/z-questions, even in European Portuguese (where, as seen above, verb-clitic can be found in declaratives):

170

Verb movement in interrogatives

(41)

a. *Quem comprou-οΊ Who bought it? b. * ¿Quién comprólo? c. *Chi halo compratto?

(42)

(EP, BP) (Spanish) (Italian)

CP Quern

C' FP F' comprou

TP T' vP

Δ tv

tstv

What (41) shows is that enclisis is not possible in w/z-questions in the Romance languages analyzed here. (42) is the tree structure for (41a), an ungrammatical sentence, demonstrating that, if the verb were to move past Τ in interrogatives, the clitic would appear in postverbal position, which does not happen. According to Raposo, (42) is ruled out because the verb only raises to F when no other element moves there. When a Wz-element (an operator) moves to [Spec, CP] via [Spec, FP], it M-licenses52 F, which means that the verb does not need to move for F to be licensed. By Economy, then, the verb will not raise, thus staying in Τ and yielding proclisis in w/i-questions. Tying this discussion to w/z-questions in Italian, then, it seems that the verb would not move past T, which in turn means that it is the subject that stays in a lower position when VS obtains. This conclusion brings the advantage of unifying the analyses for Spanish and for Italian obligatory postverbal subjects in interrogatives.5

Questions raised by V-to-C

171

If the verb in Italian and Spanish interrogatives must move to T, we have another question to answer: why is this movement obligatory in wA-questions, while in declaratives it seems to be optional? In Chapter 2 we saw that in declaratives the verb in Spanish, for example, may but need not move as high as T. This is shown by data such as those in (43 a, b), which contrasts with the interrogative examples of (43 c, d): (43)

a. Juan a menudo habla con sus amigos. Juan often talks with his friends b. Juan habla a menudo con sus amigos. Juan talks often with his friends c. ¿Con quién habla a menudo Juan? With who speaks often Juan d. *¿Con quién a menudo habla Juan? With who often speaks Juan

A menudo 'often' is an adverb that may modify aspect, based on its semantic import. If that is the case, in (43a) a menudo would be adjoined to Asp', which would mean that the verb habla would be located in Asp. In contrast, in (43b) the verb will have raised to Τ before Spell-out. In the interrogative sentence in (43 c) we see that the verb is located above the adverb, which shows that the verb has moved to a higher position than in the declarative sentence (43 a). (43 d) demonstrates that the verb cannot remain as low as it can in declaratives, having to raise to the higher slot. Thus we can conclude that the verb in interrogatives must raise to T: the adverb does not adjoin to T, remaining in a lower position (Asp')54. In declaratives, however, the verb has the option of staying in Asp and not moving any higher, as the data above show. So we need to account for the fact that the verb in interrogatives in Spanish must move to T, whereas in declaratives the verb appears to be able to stop at Asp. Rizzi (1991), as we have seen throughout this work, argues for V-to-C movement in interrogatives. He maintains that this movement happens in w/z-questions because there is a feature in C that forces raising of the verb. We have seen here that there are several reasons to believe that the verb in Romance actually does not

172

Verb movement in interrogatives

raise to C, but rather moves as high as Τ only. Nevertheless, raising of the verb to Τ in Romance interrogatives seems to be obligatory, whereas in declaratives there is at least an option for the verb to remain in Asp in Spanish and Brazilian Portuguese. Having to explain this obligatoriness, we can follow the spirit of Rizzi's idea: there is indeed some [+wh\ force that forces the verb to move to T. But, contrary to Rizzi, there is no reason to believe that that [+wh\ feature is in C: the evidence for the Romance verb moving to Τ indicates that such feature is located in T, driving the verb in a Romance interrogative to move to Τ rather than staying in Asp. Arguing that in Romance interrogatives V-to-T is obligatory due to the presence of a [+wh] force in T, we can establish the difference between interrogatives in Romance and in English: Τ in English does not carry this [+wh] feature. Since C is the only position that has the [+w/z] force in English, in yes-no questions inversion is obligatory: an auxiliary verb must be in C. In Romance, on the other hand, VS in yes-no questions is not obligatory because Τ does carry [+wh\ force (as well as C), which means that the verb does not need to move to C, as it does in English. In English, as is clear now, the verb has no other position to raise to in matrix interrogatives55, be it yes-no questions or wÄ-questions, thus having to land in C.

Summary Summarizing the issues raised by V-to-C, the first question addressed here was whether this movement, if it exists, is driven by a criterion or by feature checking needs. If it is related to checking of features, it would conform with the assumptions in Minimalism. Another issue, one that should actually come first, is: does that movement in fact happen in Italian and in French? (We have evidence that it does not take place in Spanish and in Portuguese.) We have seen that the data in Italian are not conclusive and do not prove that V-to-C exists in that language. We have also seen that V-to-C has been questioned by Ordófiez (1996, 1997) because it is not in accordance with Kayne's (1994) antisimmetry proposal. Both Spanish and Italian behave in a

Questions raised by V-to-C

173

similar fashion regarding the clitic facts used by Ordóñez to explain why V-to-C appears problematic. Sportiche (1998) advocates that overt V-to-C does not take place in French, using data that contain "IP initial material" (as it is called by Sportiche), as seen in examples (34) and (35). Zanuttini (1997), examining Italian dialects, maintains that the verb in whquestions moves to a head other than the one whose specifier has the w/z-element, which means that V-to-C does not take place; rather, the verb would move to a head lower than C. Looking at the questions raised by V-to-C, we have proceeded to investigate the position of the verb in interrogatives in Romance. We have seen that the obligatory wh-V-S order in Italian and Spanish might be explained by one of two possibilities: either the subject does not raise as high as it does in declaratives, or the verb moves higher than it does in declaratives (though not to C). For Spanish, there are several works that argue for the first possibility (a position which, incidentally, is also taken in this study). The order in Italian interrogatives may result from the same process that is believed to happen in Spanish, i.e., the subject might stay in its base position. Having looked at some of the data used by Rizzi (1991) to argue for V-to-C, we have seen that they are not conclusive: as a matter of fact, the data involving hypothetical clauses in Italian (cf. [36]) show an order different than that found in English (example [37]), a language where we may expect V-to-C to occur. Besides not showing V-to-C, the sentence in (36b) seems to indicate that the subject in fact remains below [Spec, TP]. Therefore, we see that the order in Italian interrogatives and in hypothetical clauses do not argue conclusively for Vto-C. Furthermore, we have seen that the data involving adverbs in w/j-questions also appear to make the case against verb movement to C in Italian. We have investigated the possibility that the verb in Italian may move higher than T, even if it doesn't move to C, as suggested by Zanuttini (1997). One possible position for the verb to raise to in Italian interrogatives, if it indeed moves past T, might be the Functional node proposed by Raposo (1994). Nevertheless, the discussion of the position of clitics (on which Raposo bases his argument for the FP node) shows that enclisis does not occur in w/z-questions, not even

174

Verb movement in interrogatives

in European Portuguese, where verb-clitic is quite commonly found in declaratives. If the verb did move past Τ in interrogatives, the clitic would be expected to appear after the verb, as in the (ungrammatical) structure in (42). Raposo accounts for that ungrammaticality, and in his account the verb must remain in T. Thus, the verb in w/z-questions does not appear to move past T. The next issue raised was the mandatory movement of the verb to Τ in interrogatives in Spanish and Italian. In Spanish (and in Brazilian Portuguese) we had seen that the verb in declaratives does not have to raise to Τ (though it may). In w/z-questions, however, the option is not available: the sentences in (43) show that the verb in interrogatives must indeed move to T. The explanation for that obligatoriness presented here follows an idea put forth by Rizzi (1991): there is a \+wh\ feature that makes the verb raise. However, Rizzi argues that this feature is present in C in Romance as well as in English. Here, the account offered argues that the [+wh] feature in Romance is present (also) in T, thus forcing movement of the verb to that position. The difference between English and the Romance languages, then, is related to the locus of the [+wh] feature: while in English that feature is only present in C, in Romance it is found in T. Since Τ does not carry the [+wh\ feature in English, the verb has to raise to C in matrix interrogatives, unlike the verb in Romance languages. Having investigated where the verb in Romance interrogatives moves, and why it does so, the question that remains is: what rules out preverbal subjects in Spanish and Italian w/z-questions? This is discussed in the next subsection.

5.2. Subject position in interrogatives As has been discussed throughout this study, interrogatives in several languages do not allow SV order. Here we need to provide an account of Spanish and Italian only. In English questions, the subject is clearly in its usual place; in French, the subject in interrogatives may also be in its normal position, as argued in Sportiche (1998). Thus, we are left with Spanish and Italian, languages that still need an account for their obligatory postverbal subjects in interrogatives.

Questions raised by V-to-C

175

This explanation is tied to the reason why the verb in interrogatives raises to T. Since T, along with C, is endowed with [+wh], the w/i-element in these languages needs to move first to [Spec, TP], and then, presumably, to [Spec, CP]. The idea that the ννΛ-operator moves to [Spec, CP] via [Spec, TP] is argued for in Goodall (1993). He maintains that it is that movement which prevents the subject DP from raising to [Spec, TP] in interrogatives. As Goodall argues, [Spec, TP] then is an A'-position. This movement of the w/z-element to [Spec, TP] prevents movement of the DP subject to that position. Chomsky (1995b) speaks of A- and A'-positions only informally, since, in the Minimalist Program, they are no longer so well defined. Zubizarreta (1998) discusses the idea of Τ being a type of syncretic category, such that [Spec, TP] can either be a landing site for the subject (where it would check the EPP feature), or an A'-type landing site for topicalization and focus. The fact that [Spec, TP] might be an A'-position is thus not problematic. Therefore, in Spanish and Italian, languages that have [+wh] in Τ in interrogatives, the subject is not able to raise to [Spec, TP], yielding the obligatory VS order in whquestions. As we have seen in Chapter 3, the EPP feature needs to be checked overtly. In w/î-questions it will be checked with the whelement itself, indicating that the EPP can be satisfied if [Spec, TP] is at least some of the time an A'-position. This idea is present in Goodall, as we have seen, and also in Koopman and Sportiche (1991), who claim that [Spec, TP] may be an A'-position in English, but only when wA-phrases occur there. Therefore, if the vvA-element checks the EPP feature on its way to [Spec, CP], then the subject need not, and thus cannot, raise to [Spec, TP] overtly: there being no EPP feature for the subject to check, there is no reason for it to raise (since raising of the subject is driven by the need to check the EPP feature). As for Case checking, the wA-element will either already have its Case checked (if it is an object, for example), or will not have a Case to check (if it is an adverbial). Thus, the Case feature on Τ will be available for the subject to check its formal features covertly. If the whelement is also the subject, it checks nominative Case on its way to [Spec, CP].

176

Verb movement in interrogatives

Summarizing the paragraph above, in the analysis of Spanish and Italian w/z-questions outlined here, the subject does not check the EPP feature, which is checked by the w/z-element, and the Case feature on Τ remains available for the subject to check covertly. The subject in Italian and Spanish interrogatives, then, remains in its base position, not being able to raise to Τ because the w/z-element has moved to C via T, which in those languages carries a [+wh\ feature. Brazilian Portuguese does not behave the same way as Spanish and Italian, as is well known: in Brazilian Portuguese, we do not find subject-verb inversion in interrogatives (except, possibly, for those cases where inversion is also found in declaratives, as seen in Chapter 3). This difference between Brazilian Portuguese, on one hand, and Spanish and Italian, on the other hand, can be explained if the [+wh\ feature in Τ is not the same in all three languages: unlike Spanish and Italian, the [+wH\ feature in Τ in Brazilian Portuguese is not an N/D feature, but only a V-feature. Thus, the w/z-element does not go through [Spec, TP] on its way to [Spec, CP]. That being so, the EPP feature in Τ has to be checked by the subject, just as in declaratives. This would also be the mechanism in French interrogatives, but recall that, as discussed in Chapter 2, the lexical subject in French is found in [Spec, AgrSP], a position to where the subject will raise after checking the EPP feature in T. The mechanisms are shown below, where (44a) illustrates what happens in Spanish and Italian, and (44b) in Brazilian Portuguese:

Questions raised by V-to-C

(44)

177

a. CP wh

C TP twh

T'

V+T

vP Subj

v'

tv

VP tv

Uh

b. CP wh

C TP Subj V+T

T' vP tsubj

ν

tv

VP tv

twh

The structures above represent the mechanisms in Spanish/Italian and Brazilian Portuguese interrogatives, respectively, with the whelement as an object. We can see in (44a) that the subject remains in [Spec, VP], because the ννΛ-element raises to [Spec, CP] via [Spec, TP], thus checking the EPP feature in that position and forcing the

178

Verb movement in interrogatives

subject to remain in its base position. Because of this mechanism, we find the obligatory VS order in Spanish and Italian interrogatives. In (44b), on the other hand, we see that the w/z-element raises directly to [Spec, CP], leaving the EPP feature to be checked by the subject, which must then raise to [Spec, TP]. That way, subject-verb inversion does not yield, as confirmed by the data in Brazilian Portuguese. Here, one question that needs to be asked is why the [+wh] feature in Τ in Brazilian Portuguese is a V-feature, not an N/D feature. Is there any morphological or syntactic feature to which it could be linked? The conclusion that the [+wh] feature in Brazilian Portuguese is a V-feature has been reached based on an analysis of the word order and on a comparison to other languages. There does not seem to be any obvious feature in Brazilian Portuguese syntax or morphology that would point to a definite answer to the question why the [+wh\ feature in Brazilian Portuguese is a V-feature. Thus, the issue must be addressed from a different angle: we can now ask whether it is plausible that this is a V-feature in Brazilian Portuguese, or whether Brazilian Portuguese resembles English in that aspect, having the [+wh\ feature only in C. Recalling the discussion on obligatory inversion (and lack thereof) in yes-no questions in the section above, a correlation was established between [+wh] in C only and mandatory inversion, as in English: since Τ carries no [+Wz] force in English, in interrogatives the verb is left with C to land in (not having the option of landing in T), which explains obligatory VS. If Brazilian Portuguese were to resemble English, i.e., if in Brazilian Portuguese there were no [+wh\ feature in Τ but only in C, then VS would also be obligatory in Brazilian Portuguese yes-no questions. However, in Brazilian Portuguese we find a different scenario: not only is VS not obligatory, it is not possible in yes-no questions. Therefore, the [+wh] feature must not be present in C only (as it is in English): it has to be present in Τ as well. If Brazilian Portuguese is not like English in what concerns the [+wh] force, then Τ must carry a [+wh\ feature, and this must be a V-feature, not an N/D feature (which is also the case in French56·57). In terms of plausibility, the fact that Brazilian Portuguese carries a V[+wh~\ feature in Τ must be favored, given that the other possibility (Brazilian Portuguese being like English) has been ruled out in face

Questions raised by V-to-C

179

of the facts in Brazilian Portuguese. Thus, we can account for the lack of inversion in [+wA] questions, as seen above: unlike Spanish and Italian interrogatives, the [+wh] element does not move through [Spec, TP], thus not disabling this position for the landing of the subject. If [Spec, TP] is still available for the subject to move to in Brazilian Portuguese interrogatives, then this movement has to take place due to the obligatory checking of the EPP feature. We then see that the [+wA] feature in Τ must be only a V-feature in Brazilian Portuguese. To sum up, the subject in Spanish and Italian interrogatives must remain in its base position because in those languages Τ carries a D feature with [+wA] force: the wA-element raises to [Spec, CP] via [Spec, TP], checking the EPP feature in that position, and thus blocking the subject from moving overtly. Case checking of the subject is done covertly. In Brazilian Portuguese and in French, the [+wA] feature is only a V-feature, not an N/D feature: the w/i-phrase does not go through [Spec, TP]. Therefore, the subject must still move to that position in order to check the EPP feature. These various options can be summarized in a chart: (45)

Summary of options: wA-feature in Τ V N/D Eng BP/ F S/I

-

-

+

-

+

+

movement V-to-C; wh-mov't to C V-to-T; wh-mov't to C iS V-to-T; wh-mov't via [Spec,TP]

The chart in (45) shows the options regarding the wA-feature in Τ and the syntactic reflexes that correspond to those options. As we can see, the wA-feature in Τ can be negative for both the V and the N/D features, as in English; it can be positive for V and negative for N/D, as in Brazilian Portuguese and French; and it can be positive for both V and N/D, as in Spanish and Italian. The only option not attested here is [+wh] in Τ negative for V but positive for N/D. Presumably, that option would be expected in a language where we could find V-to-C

180

Verb movement in interrogatives

(not V-to-T), and w/z-movement through [Spec, TP]. In such a language, the subject in interrogatives would remain in [Spec, VP], not being able to raise to [Spec, TP] for the reasons outlined above. However, the verb would indeed move to C in interrogatives, as it does in English. In such a language, there would inevitably be inversion in interrogatives, because the verb moves to C and because the subject remains in its base position. Given this "double" reason for inversion, it may be difficult to attest that option {[+wh] in Τ as N/D feature only), since we might resort to the English-like option, or to the Spanish-like one. Naturally, the account for inversion depends on the facts of each language, and it may be that one given language would indeed exhibit facts (such as position of adverbials) that would lead to the option that is unattested in our chart. Nevertheless, that has not been the case among the languages analyzed here. After this summary, one question remains: if the verb does not move to C overtly in Romance, does it do so covertly? In English, Vto-C is overt, but in Romance there may not be the need for it at all: if the [+wh] feature is present in Τ as well as C, presumably covert Vmovement to C would not be necessary. Lack of even covert V-to-C is a possibility mentioned in Zanuttini (1997), as we have seen before. Sportiche (1998) argues for lack of overt V-to-C in French, based on Complex Inversion and clitic doubling facts. He states that I-to-C is a triggered rule, which applies only if some specific property (such as question) triggers it. In that case, C contains [+Q], a property which must be realized. If a complementizer is available, then raising of I to C is not required, but, Sportiche continues, if the [+Q] property appears in main clauses, the only way to license this property is to raise I-to-C. Sportiche concludes that there is no variation in the configurations allowing and thus requiring I-to-C. Whether this raising is done overtly or covertly depends on the language and on the particular construction, he argues, but that would be the only variation possible: I-to-C in interrogatives would always take place. Sportiche does not entertain the hypothesis argued for in this chapter, i.e., that Τ in Romance is also endowed with [+wA], a hypothesis that might allow for the verb not to raise to C in interrogatives. This is the hypothesis that might mean that V does not need to raise to C, and by Economy, cannot do so. This question is a purely

Questions raised by V-to-C

181

theoretical one, since there seems to be no way of proving (or disproving) covert V-to-C in Romance. Perhaps V-to-C is indeed, as Sportiche argues, universal and without variation (other than covert versus overt). In fact, universal (and thus covert in Romance) V-to-C is an attractive solution to this discussion, for universality is ultimately sought. Nevertheless, the possibility does exist for this movement not to take place at all in certain languages, and this possibility needs at least to be brought to light. Whether or not V-to-C ultimately takes place in Romance is a question for which I will not attempt to provide a final answer.

Summary In section 5, we have discussed the position of subjects and verbs in Romance interrogatives, contrasting these positions with English. The conclusion reached is that in Spanish and Italian interrogatives, VS obtains not because the verb raises to C, but rather because the subject stays in its base position: if the EPP feature is checked by the w/z-element, the subject has no reason to raise overtly; by Procrastinate, then, it does not move before Spell-out. The Case feature of the subject can be checked covertly (and if the wA-element is the subject it also checks nominative Case when it moves through TP). Movement of the w/i-element via [Spec, TP] occurs because Τ is also endowed with a [+wh\ feature, as argued previously for the account of obligatory V-to-T in interrogatives in Spanish and Italian. Therefore, the subject is prevented from moving to [Spec, TP] because the w/j-element in those languages moves to C via T, thus checking the EPP feature in that position and eliminating the need for the subject to raise. In Brazilian Portuguese and in French the subject in interrogatives moves to its regular spot, since the w/i-element does not check the EPP feature on its way to C. We have seen that lack of VS order in Brazilian Portuguese interrogatives (as opposed to mandatory VS in Spanish and Italian) can be explained if the [+Wi] feature in Τ is only a V-feature, which does not allow the w/z-element to move through [Spec, TP]. If no other item checks the EPP feature (which

182

Verb movement in interrogatives

needs to be checked before Spell-out), then the subject has to move overtly, thus yielding the wh-S-V order found in Brazilian Portuguese. One final question that has been addressed regards the nature of the [+w/z] feature in Brazilian Portuguese. Since there seems to be no feature in the syntax or the morphology of that language that determines the nature of the [+wh] feature (i.e., that it is indeed a Vfeature), the conclusion that it is not an N/D feature was reached based on the word order facts and on comparison to other languages. That being the case, the options regarding the nature of the [+wh] feature were the following: is it plausible that in Brazilian Portuguese this is a V-feature, or does Brazilian Portuguese only has the [+wh] feature in C, as in English? Since the facts in English and in Brazilian Portuguese interrogatives are crucially different (actually, they are opposite, since English requires inversion and Brazilian Portuguese does not allow it), we are led to conclude that Brazilian Portuguese cannot pattern with English, and thus cannot have [+wh\ force in C only. That [+wh\ force has to be present in Τ as well, driving the verb to that node in interrogatives. Therefore, we have favored the plausibility that the [+wh] feature is a V-feature in Brazilian Portuguese: that being so, the w/z-element does not move through [Spec, TP] (unlike what happens in Spanish and Italian), and the subject must still move to that position overtly, in order to check the EPP feature, just as in declaratives. Finally, I have also pointed out that, on theoretical grounds, it is possible that V-to-C may not take place at all in Romance. However, I have left that question open for further discussion. One aspect of interrogatives that has not been discussed yet is embedded questions: what happens in interrogatives when they are not the main clause? Even though matrix questions constitute the focus of this work, we should discuss embedded contexts briefly.

6. Embedded interrogatives Throughout this chapter, we have discussed and analyzed the mechanisms in interrogatives in main clauses. However, the syntax of em-

Embedded interrogatives

183

bedded interrogatives has not been discussed yet. In English, embedded questions do not show the subject-auxiliary inversion present in main questions; in Romance, word order in embedded interrogatives does not seem to change very much. Let us start with the data. In Spanish, for example, VS is obligatory in embedded whquestions as well as matrix wA-questions: (46)

a.

¿Qué dijo Ana? What said Ana b. Yo pregunté qué dijo Ana. I asked what said Ana c. * Yo pregunté qué Ana dijo. I asked what Ana said

In (46), we can attest that the word order that appears in main clauses such as (46a) also surfaces in embedded questions, as (46b). Lack of inversion yields ungrammaticality, shown in (46c). At this point, we can presume that the account for subject-verb inversion in matrix whquestions also explains the obligatoriness of inversion in sentences such as (46b): the embedded subject must still remain in its base position, for it seems that the embedded w/i-element also raises to [Spec, CP] via [Spec, TP], Therefore, it appears that the w/j-phrase must always move first to [Spec, TP] in the clause in which it ends up in [Spec, CP]. In English, however, we do not find inversion in embedded contexts: (47)

a. When are you leaving? b. *I asked when were you leaving. c. I asked when you were leaving.

What (47) shows is the opposite of what is seen in Spanish: in English, the subject-auxiliary inversion that is mandatory in main questions does not appear in embedded questions. Rizzi (1991) suggests that the specification WH may fill the embedded C, making it unavailable as a landing site for the verb (he does not elaborate on this suggestion). Chomsky (1995b) proposes that it is the feature Q for interrogative clauses which determines inversion in English matrix

184

Verb movement in interrogatives

questions, and that if Q is embedded, the verb will not be able to raise to Q. Both proposals could be adapted to conform with the analysis proposed in this chapter for inversion in English matrix questions; however, I will not go into detail here. Suffice it to show that English differs from Romance in that the mechanism in embedded whquestions is not exactly the same as in matrix questions. For Brazilian Portuguese, as one would expect, the same lack of inversion that is seen in main questions is also found in embedded contexts. Therefore, not much has to be said about Brazilian Portuguese, where the mechanisms seem to be very much the same for both matrix and embedded interrogatives. Here we still need to discuss embedded yes-no questions. In Spanish, for example, we have seen that inversion in matrix yes-no questions is optional. That is not a problem if the w/i-feature is in T. In embedded yes-no questions, that optionality also exists: (48)

a.

¿Se va a casar Juan con Ana? Will marry Juan with Ana? b. ¿Juan se va a casar con Ana? Juan will marry with Ana? c. Me preguntaron si se iba a casar Juan con Ana. Me they-asked if would marry Juan with Ana 'They asked me if Juan was going to marry Ana'. d. Me preguntaron si Juan se iba a casar con Ana. Me they-asked if Juan would marry with Ana

In (48) we attest that inversion is optional in matrix yes-no questions, as shown in (48a) and (b), and also in embedded yes-no questions, seen in (48c) and (d). Thus, it appears as though the same mechanism that triggers obligatory VS in w/î-questions in Spanish (and optional postverbal subjects in yes-no questions) is also at play in embedded contexts. VS in Spanish and Italian matrix questions obtains, as argued here, because the wA-element moves to [Spec, CP] via [Spec, TP], not because of V-to-C movement. In English, on the other hand, postverbal subjects in interrogatives occur due to V to C, which explains why VS is not possible in embedded questions in this language.

Previous studies of BP interrogatives

185

Before closing this chapter, we need to look at previous accounts of questions in Brazilian Portuguese. The next section discusses analyses of Brazilian Portuguese interrogatives that were available before this work.

7. Previous studies of Brazilian Portuguese interrogatives Several researchers have noted the order SV in Brazilian Portuguese interrogatives. Duarte (1992) shows the diachronic change from VS to SV in w/j-questions in Brazilian Portuguese, analyzing data that confirm the change. She states that the appearance of é que 'is that' in interrogatives in Brazilian Portuguese was a determining factor for the loss of VS in interrogatives (or, she believes, for the loss of V-toC). Examples of interrogative with è que are given below: (49)

a. b.

Onde é que a Maria comprou o presente? Where is that Maria bought the present? Quando é que o Paulo chegou? When is that Paulo arrived?

The occurrence of é que in Brazilian Portuguese interrogatives is very common in spoken language. However, it does not appear to be related to the loss of VS in w/i-questions: as Lopes Rossi (1996) points out, modern European Portuguese (EP) also contains é que, but EP has not lost VS in interrogatives, which remains as a possible order. Therefore, the correlation between the appearance of é que and the loss of VS in interrogatives does not seem to hold. Lopes Rossi (1996) hypothesizes about the process of reanalysis that interrogatives have gone through. She believes that Brazilian Portuguese indeed had V-to-C before the 19th century, and that the movement was lost because of the weakened verbal agreement in Brazilian Portuguese. In her data, she also observes the appearance of the non-clitic interrogative pronoun O que 'what', which has substituted the clitic que since the 19th century. According to Lopes Rossi, que differs from o que: the former seems to be a clitic (like French que) which occurs in interrogatives with V-to-C because

186

Verb movement in interrogatives

it needs to cliticize to a verb moved to CP. The interrogative pronoun o que, she continues, can occur without V-movement to C and also in situ, since it is not a clitic. However, she notes that the appearance of O que was not the only reason why Brazilian Portuguese has lost VS in questions: the language, according to her, has gone through a change in a functional category (Agr, she claims) that has been responsible for the change in compliance with Rizzi's Wh-Criterion (from static to dynamic agreement). In other words, the weakening of agreement between the verb and its subject has also been a factor in the disappearance of VS in Brazilian Portuguese interrogatives. She reasons that EP also has o que, but has not lost VS in questions, because in EP verb-subject agreement has remained strong. While some of the reasons for the loss of VS in Brazilian Portuguese presented in Lopes Rossi may be sound, at least one of them should be questioned: why there should be any relationship between strength of verbal agreement and V-t-C. English is a language where V-to-C is present, but we find weak agreement on the verb (here, 'weak' does not mean that it does not force overt movement. We speak of 'weak agreement' here in the pre-minimalist sense of not licensing null subjects). Therefore, this kind of relationship does not seem to explain why there is (or not) inversion in questions. Furthermore, the reasons presented by Lopes Rossi do not account for the synchronic obligatoriness of SV in questions. In other words, there is still a need for an explanation to clarify why it is that movement occurs (or does not). That clarification is attempted by Galves (1996). She attributes the historical shift from VS to SV to a parametrical change, from strong Agr to weak Agr (as defined by Galves). That assumption makes sense, naturally, since language change occurs when children reset a certain parameter given the evidence they are exposed to when acquiring the language. Galves links the resetting of the parameter to the nature of the agreement morpheme in Brazilian Portuguese, which is also claimed by Lopes Rossi (1996), as seen above. Galves defines Agr as weak when it does not contain features for person (lst/2nd/3rd). Indeed, Brazilian Portuguese does not show different verb morphemes for all the persons; it does not distinguish between 2nd and 3rd person endings, which leads to the

Conclusion

187

assumption that Agr is weak. The lack of distinction between 2nd and 3rd person can be seen below: (50)

eu escrevo voce escreve ele escreve nos escrevemos voces escrevem eles escrevem

I write you write he writes we write you write they write

(50) shows that the first person is the only one that is distinguished from the others, which leads to the claim that Agr is weak in Brazilian Portuguese. Since it is weak, it has no independent projection, according to Galves, and the subject moves to [Spec, TP], while the verb moves to T. The fact that Agr is not projected in Brazilian Portuguese has also been argued here, in Chapter 2 (though for different reasons). However, the fact that the subject does not need to move to Agr to check its features, as argued by Galves, does not necessarily explain why it must move to TP to do so. She does not make it clear why there is movement at all, which is what would need to be explained under a minimalist view. Therefore, Galves's account lacks an explanation that is fundamental to a proper analysis under Minimalism: the reason why movement takes place. This reason is provided in this study, where we find a minimalist explanation for the lack of inversion in Brazilian Portuguese interrogatives, as well as for the obligatory inversion in Spanish and Italian w/i-questions, contrary to the previous accounts of SV in Brazilian Portuguese interrogatives, which could not provide a solid explanation of the facts analyzed in this work. 8. Conclusion This chapter has discussed interrogatives in Romance, in search for an account of obligatory subject-verb inversion in such w/i-questions in Spanish and Italian, as opposed to the ungrammaticality that inversion yields in Brazilian Portuguese. First, we have seen the paradigms for different Romance languages (Spanish, Italian, French, Brazilian

188

Verb movement in interrogatives

Portuguese) regarding interrogatives, having also shown that those languages exhibit w/i-movement to [Spec, CP]. Then, we have looked at Rizzi's ^-Criterion, showing that this criterion does not provide a satisfactory account for the data in Italian and in other Romance languages. The basic assumption in Rizzi's ^-Criterion is that there is movement of the verb to C. We have discussed analyses for Spanish interrogatives that did not presuppose V-to-C, as well as the questions raised if that movement is assumed. In order to find out whether such movement is indeed present in Romance, we have discussed the position of the verb in interrogatives, looking at data used by Rizzi to argue for V-to-C, as well as examples involving the position of adverbs and clitic pronouns in w/z-questions. Our conclusion is that in Romance the verb does not raise to C in interrogatives, but rather to Τ only. A difference between English and Romance languages was established: in English interrogatives, the verb does indeed move to C. This difference is due to the presence of a [+wh\ force in Τ (besides C) in Romance, which is not present in English, where the [+wh\ force only exists in C. So, as desired, parametric variation is 'located' in functional categories and their specifications. After analyzing the position of the verb in interrogatives in Romance, we have turned to the position of subjects, showing why VS yields in Spanish and Italian, but not in Brazilian Portuguese. In the first two languages, the subject is argued to remain in its base position, whereas in Brazilian Portuguese questions the lexical subject raises to [Spec, TP]. It has been argued, along the lines of Goodall (1993), that subject-verb inversion yields in Spanish and Italian because the w/*-element moves to [Spec, CP] via [Spec, TP], thus checking the EPP feature in [Spec, TP] and making it impossible for the subject to raise to that position overtly. In Brazilian Portuguese, w/z-movement occurs directly to [Spec, CP], because the [+wh\ feature in Τ is not an N/D feature, but simply a V-feature (as opposed to Spanish/Italian). Since the w/i-element does not raise via [Spec, TP], the EPP feature has to be checked before Spell-out, which is done by the subject. This chapter also includes a brief discussion of embedded questions. We have seen that obligatory VS in embedded interroga-

Conclusion

189

tives in Spanish, for example, is triggered by the same mechanism that occurs in main wA-questions: the w/z-element moves to [Spec, CP] via [Spec, TP]. The optionality of postverbal subjects in Spanish embedded yes-no questions is also explained the same way as optionality of VS in main yes-no questions. As we have also seen, postverbal subjects in English interrogatives occur due to V to C, an account that explains the lack of VS in embedded questions in this language. Thus, differences in embedded word order in English and Romance interrogatives reduce to the presence or the absence of V to C in main questions: since Romance does not exhibit V to C, the order found in embedded interrogatives is the same as in matrix questions, whereas in English, which does contain V to C, the order is different, for that movement will obviously not occur in embedded environments. Finally, we have looked at previous accounts of word order in vc/j-questions in Brazilian Portuguese, showing that none of the those analyses contained a satisfactory explanation for the facts, since none of them explains why it is that VS is disallowed in Brazilian Portuguese w/z-questions (with transitive verbs).

Notes

191

Notes

Notes to Chapter 1 1. Here postverbal means after the verb, not the VP. 2. In Chapter 3 we will study a construction that in fact allows for postverbal sub jects with unergative verbs, the so-called locative inversion. The default order with unergatives, however, is indeed SV. 3. Notice that (i) is grammatical, but it carries a different meaning: (i) Sao os meninos impossiveis. 'It is the impossible boys' 4. The verbs ser/estar 'to be' also allow for VS in questions: (i) Quem è voce? 'Who are you?' (ii) Onde está o dicionário? 'Where is the dictionary?' Though not obligatory, VS seems to be the preferred order in wA-questions involving ser/estar. 5. In Minimalism, the subject NP in Spanish would not necessarily have to move to [Spec, AgrSP] before Spell-out. 6. It is not universally agreed that V2 is driven by Case requirements. 7. As we will see in Chapter 2, adjunction is restricted in Minimalism. 8. Rizzi notes that the equivalent to (22b) is found in Brazilian Portuguese, but he does not mention the equivalent to (22a). However, it sounds perfect to my ear, without echo intonation. 9. Inversion also occurs with some predicates that select constituents which they do not theta-mark, as the nonthematic measure phrase in (i): (i) ¿Cuántopesará esta beba? 'How much would this baby girl weigh?' Thus, Suffer uses the term 'argumentai' to encompass real (i.e., referential) arguments and those such as in (i). 10. In Brazilian Portuguese, left-dislocated constituents seem not to appear in the position Sufier claims possible for Puerto Rican Spanish: (i) * O que ao Rafaelfizeram? What to Rafael did-they-do This could serve as evidence that her predictions about the non-compliance with the AAL are not borne out in BP, and that the condition might need adjustments. Nevertheless, (i) can be an instance of topicalization. It is not clear whether topicalized elements move to [Spec, CP] or adjoin to IP (Lasnik and Saito 1992). It may very well be the case that topicalization is movement to [Spec, CP] (Authier 1992), given that the facts in BP do not prove otherwise. If this is

192

Notes

so, (i) would not prove anything in relation to the AAL, because a topicalized element would not adjoin to IP anyway. 11. In Chapter 2 I argue that BP does not have AgrSP, and that preverbal subjects move to [Spec, TP]. 12. The lack of NeutP is not sufficient to account for lack of VSO in Italian and Catalan, since this order could obtain if only the verb moved out of VP. Possible explanations for the lack of VSO in languages other than Spanish (and a few others) are found in Chapters 3 and 4. 13. Although there seems to be no reason for the presence of a [Spec, TP] position in (34) (since no element moves there), I maintain the structure in Raposo and Kato (1994). They assume that clitics are adjoined to Τ in finite clauses in EP. 14. For marked Topic, Kato and Raposo have the following examples: (i) Estes CD's recomendou-me a Maria (EP/*BP) (ii) Os CD's. a Maria me recomendou. (*EP/BP) I agree with their judgments on these sentences.

Notes to Chapter 2 15. The third order possible for (4) would be: (i) Sao os meninos impossíveis. But in this case, as already noted in Chapter 1, there would be a change in meaning: (i) means 'Here/These are the impossible boys'. 16. Later in this Chapter we find a dicussion of the position ofjà in the structure of the sentence. 17. Note that (8d) improves with a longer direct object: (i) Yo sé y a las reglas de órdenes de palabras. I know already the rules of order of words 18. Later on, we will argue that Agr may indeed be present in French, a language which shows different behavior from other Romance languages regarding, among other things, adverb position. Here I want to disregard Agr as universally present in phrase structure. 19. Note that movement of the subject through [Spec, AspP] would violate Last Resort (Chomsky 1995b), since it would check no feature in that position, unless it is assumed that Asp, like T, has a strong D-feature to be checked. An alternative analysis to adverb adjunction might be drawn along the lines of Cinque (1999). He suggests that adverbs might be located in the Spec positions of distinct maximal projections, which would mean that each adverb projects a different functional projection. By this analysis, déjà itself would occupy [Spec, AspP], 20. Note that here the verb is assumed to move to T. As is shown later, the verb in Brazilian Portuguese declaratives may optionally raise only to Asp.

Notes

193

21. The sentences in (35) do not have exactly the same meaning as Italian (30), given that Portuguese expresses preterite with a simple verb, while Italian does so with a compound verb. A compound verb is used in (35) to obtain a close syntactic correspondence to the sentences in (30). 22. Spanish is a little more complex, given that it may, depending on the speaker, allow VSO. For the speakers who do accept VSO, the order SVO seems to give a topicalization "flavor" to the subject. I will return to the order in Spanish declaratives in Chapter 3. 23. Belletti does note that left dislocation is in fact available in French: (i) Jean, probablement il aime la linguistique. Jean, probably he loves linguistics However, she argues that the structure is not relevant, "because it would involve a subject clitic in the sentence following the dislocated object." (Belletti 1990: 130) 24. Again, an alternative analysis of adverb placement is found in Cinque (1999). The alternative for right adjunction in (43c) would be to have a base structure such as (i), where the adverb is in the spec of a VP 'shell', followed by obligatory successive leftward movement of the lower VP to a higher specifier: (i)

XP

I sempre VPj I Bia faz o dever de casa 25. While Tenny (1987) argues for a syntactic category of Aspect based on semantic facts, Belletti (1990) assumes the presence of an AspP without further argument. 26. Although AgrP is assumed here to not be always present, sometimes it may indeed be found in the derivation, as has been argued for French. Since French may have AgrS, it is possible that a language such as Italian, that may exhibit agreement between the participle and the direct object, in fact has AgrO. 27. Note that the subject does not move through AuxP or PartP, for only AspP and TP are assumed to have strong D features.

Notes to Chapter 3 28. Note that (5c) is indeed possible in both Spanish and English, but not with a meaning that would correspond to (5a).

194

Notes

29. Here I am referring to verbs that are clearly considered unaccusatives in other languages, such as aparecer 'appear', nascer 'be born', morrer 'die', chegar 'arrive', sair 'leave', and several others. 30. We will see later that unergatives may allow for a construction know as 'locative inversion', where a locative PP is the first element in the clause, followed by an unergative verb and its subject. 31. As noted in endnote 1 for the example in Spanish, (15c) is also grammatical, but not with a meaning that could correspond to (15a). 32. Example (27a) is an instance of written language, as would be found at the entrance of a shop. In spoken Brazilian Portuguese, the impersonal se (as other clitics) tends to appear in preverbal position, as in (27b). 33. (30a) is grammatical with the meaning 'What saw Ana?'. 34. Chomsky (1995b) assumes that it is the D feature of a DP that enters into a checking relation with the EPP feature of T, and a PP would not normally check the D (or N) feature. However, Collins (1997) argues that, in the locative inversion construction, the PP may in fact check the EPP feature. He considers two possibilities, one of which I quote here: " (...) to allow the set of features that may enter into a checking relation with the EPP feature of Τ to be widened from simply the D feature of a DP to any categorial feature". (Collins 1997: 28) Thus, checking of the EPP feature by a PP seems to be possible. 35. Hale and Keyser (1993) argue that the initial lexical structure of unergatives verbs can be compared to that of a simple transitive verb. In other words, (at least certain) unergatives can be thought of as covert transitives. This view is not in disagreement with what is claimed here for the locative inversion construction. There can be no covert raising of two arguments, but some unergatives are considered covert transitives because the object has incorporated to the verb, and does not check Case. Therefore, covert Case checking of the subject is still possible in those instances. 36. Unaccusatives will not have the light verb. 37. This is a generalization of what must happen if in VOS the subject is indeed still within VP. We must note the possibility that both the object and the subject move out of the VP. But minimally any account of VOS would require that the object move overtly. 38. For more on the (lack of) null subject properties in Brazilian Portuguese, see Duarte (1996), Galves (1996), Kempchinsky (1985). 39. With unergatives, VS is found in the locative inversion construction, as seen in the previous section, but not with an expletive in sentence-initial position. 40. Recall that in the analysis of verb movement in Brazilian Portuguese in Chapter 2, we reach the conclusion that in declaratives it is possible that the verb raise to Asp or to T. For ease of exposition, however, in this chapter I only show structures with the latter possibility. 41. Chomsky (1995b) argues that existential there in English does not check Case, unlike our proposal for pro here. Nevertheless, as will be seen later, Lasnik (1995) does argue for Case-checking ability of there.

Notes

195

42. The account of Spanish VSO in Ordófiez (1997) involves the projection of a Neutral Phrase (NeutP) to where the subject would move. This NeutP is below TP and above VP. 43. Another argument for pro merging in [Spec, AspP] is the fact that ifpro merged directly in [Spec, TP], nominative Case would be checked by a trivial chain. But Chomsky (1995b) assumes that structural Case can be checked only by nontrivial chains, which means that, ifpro checks nominative, it cannot merge in [Spec, TP], 44. In Spanish transitive constructions, VOS and SVO also correlate with subject as new or old information (respectively). In Brazilian Portuguese, where the only order possible with transitives is SVO, the subject will be preverbal regardless of whether it is old or new information, as is also the case in English. 45. For a different view on placement of Event in the grammatical representation, see Stowell (1996). 46. Note that examples (90e) and (f) are grammatical with a different meaning. Atender can also mean 'to see (a patient)'. With that transitive meaning, those examples are naturally acceptable, with the DP interpreted as object, not subject. They become ungrammatical if we use the meaning referred to in the main text, i.e., 'answer (the phone).'

Notes to Chapter 4 47. Here I will not use unaccusative verbs, since these show different behavior than transitives and unergatives, as seen in Chapter 3. 48. Note that this construction is grammatical in Puerto Rican Spanish (PRS), a dialect that admits SVO in interrogatives. For more on interrogatives in PRS, see Ortiz-López (1993), Suffer (1994). 49. Under the Government and Binding theory, it was assumed that the wA-element in (8a) would move at LF. Chomsky (1995b) states that "if the operator feature [WJ-] is unchecked, it raises to an appropriate position, covertly if possible (by Procrastinate)". (Chomsky 1995b: 272) 50. A logical question at this point would be how the WA-Criterion might account for verb movement in yes/no questions. I will discuss the relevance of yes/no questions toward the general issue of the syntax for wA-questions later in the chapter. 51. But note the exception of por qué 'why' in Spanish, which does allow for por qué-S-V (along with por qué-V-S). For accounts of that exceptional case, see Goodall (1993), Ordóñez (1997). The same is true for Italian, as discussed by Rizzi (1991). 52. Here, "M" is for "morphological". 53. But note that Italian does not admit the order VSO - here I am referring to mandatory VS in interrogatives, be it VOS or VSO. The lack of VSO in Italian

196

Notes

might be explained following Ordófiez (1997): this language does not project NeutP (the projection to which the subject would raise in VSO in Spanish). However, as noted in Chapter 1, even if there is not a NeutP for the subject to move to and it stays within VP, the object still moves past the subject to yield VOS. Recalling Chapter 3, Chomsky (1995b) suggests that both arguments cannot check Case covertly; thus, the object would need to move overtly if the subject did not (because the wA-element moved to [Spec, TP]). Therefore, VSO occurs in Spanish because of some other mechanism for checking Case on the subject, as proposed by Alexiadou and Anagnostopoulou (1997), who correlated the existence of clitic doubling in a given language to the occurrence of VSO: the relation between the verb and the subject would be an instantiation of doubling. 54. This entails that the adverbs discussed by Suñer (1994) must be adjoined to T', or, at least, have the possibility of adjoining to T' (given the discussion in Chapter 2 of the variable placement of ya). 55. Naturally, the picture in English embedded questions is different, as discussed below. 56. As shown in the sentences in (12), SCI seems to be the "normal" syntax for French wA-questions. When there is SI, it is possible that the verb may in fact move as high as C, with the subject in its usual position. Thus, there would not be any reason to argue that in French the wA-feature on Τ is anything other than a V-feature, corroborating the conclusion that French and Brazilian Portuguese pattern together in this respect. 57. Here I have not discussed the syntax of French SCI structures. A possible account for SCI might argue for the presence of an XP in French that would be found below CP and above AgrP, as in (ii): (i) Qui Jean a-t-il vu? (ii) CP

Qui

c*

T° VP Clitic vu In (ii), XP might be a functional phrase, such as the FP proposed in Raposo (1994).

Notes 58. Both French and Brazilian Portuguese exhibit wh in situ as an option, though present only in main questions: (i) a. Voce viu que filme onteml (BP) You saw what movie yesterday b.*Eu perguntei você viu que filme ontem. I asked you saw what movie yesterday c. Eu perguntei que filme você viu ontem. I asked what movie you saw yesterday

197

References

Adams, Marianne 1987 "From Old French to the theory of pro-drop", Natural Language and Linguistic Theory 5, 1-32. Alexiadou, Artemis and Elena Anagnostopoulou 1997 "Postverbal subjects", paper presented at the 7th Coloquio de Gramática Generativa, Oviedo, Spain. Authier, Jean-Marc 1992 "Is French a null subject language in the DP?", Probus 4, 1 -16. Baker, Mark 1985

Incorporation, Chicago: University of Chicago Press.

Belletti, Adriana 1988

"The Case of unaccusatives", Linguistic Inquiry 19, 1-34.

Belletti, 1990 Adriana Generalized verb movement, Turin: Rosenberg and Sellier. Birner, Betty J. and Gregory Ward 1993 "There-sentences and inversion as distinct constructions: a functional account", in Proceedings of the 19th Annual Meeting of the Berkeley Linguistics Society, 27-39, Berkeley, CA: Berkeley Linguistics Society. Bobaljik, Jonathan and Hoskuldur Thráinsson 1998 "Two heads aren't always better than one", Syntax 1,37-71. Bowers, John 1993

"The syntax of predication", Linguistic Inquiry 24, 591 -656.

Bresnan, Joan 1993 "Locative inversion and the architecture of UG", ms., Stanford University. Burzio, Luigi 1986 Italian Syntax, Dordrecht: Reidel.

200

References

Chomsky, Noam 1986 Knowledge of language: Its nature, origin, and use, New York: Praeger. Chomsky, Noam 1991 "Some notes on economy of derivation and representation", in R. Freidin, ed., Principles and parameters in comparative grammar, 417-454, Cambridge, MA: The MIT Press. Chomsky, Noam 1993 "A Minimalist Program for Linguistic Theory", in K. Hale and S. J. Keyser, eds., The View from Building 20: Essays in Linguistics in honor of Sylvain Bromberger, 1-52, Cambridge, MA: The MIT Press. Chomsky, Noam 1995a "Bare phrase structure", in Héctor Campos and Paula Kempchinsky, eds., Evolution and revolution in linguistic theory: Essays in honor of Carlos Otero, 51-109, Washington, D.C.: Georgetown University Press. Chomsky, Noam 1995b "Categories and Transformations", in The minimalist program, 219394, Cambridge, MA: The MIT Press. Chomsky, Noam and Howard Lasnik 1993 "The Theory of Principles and Parameters", in Joachim Jacobs, Arnim von Stechow, Wolfgang Sternefeld and Theo Vennemann, eds., An International handbook of contemporary research, Berlin/New York: Walter de Gruyter. Cinque, Guglielmo 1993 "A null theory of phrase and compound stress", Linguistic Inquiry 24, 239-297. Cinque, Guglielmo 1999 Adverbs andfunctional heads: a cross-linguistic perspective, New York: Oxford University Press. Collins, Chris 1997 Local Economy, Cambridge, MA: The MIT Press. Contreras, Heles 1991 "On the position of subjects", in Susan Rothstein, ed., Perspectives on phrase structure: Heads and Licensing (Syntax and Semantics 25), 63-80, New York: Academic Press.

References

201

Demonte, Violeta 1989 Teoría sintáctica de las estructuras a la reacción, Madrid: Síntesis. Drijkoningen, Frank 1994 "Morphological strength and word order variation in French", paper presented at the IV Colloquium on Generative Grammar, Tarragona, Spain. Duarte, Maria Eugenia Lamoglia 1992 "A perda da ordem V(erbo) S(ujeito) em interrogativas Qu- no portugués do Brasil" [" The loss of order V(erb) S(ubject) in whquestions in Brazilian Portuguese"], DELTA 8 (Special issue): 37-52. Duarte, Maria Eugenia Lamoglia 1996 "Do pronome nulo ao pronome pleno: a trajetória do sujeito no portugués do Brasil " ["From null pronouns to full pronouns: the trajectory of the subject in Brazilian Portuguese"], in Ian Roberts and Mary Kato, eds, Portugués brasileiro: urna viagem diacrònica [Brazilian Portuguese: a diachronic voyage], 107-128. Campinas, SP: Editora da UNICAMP. Galves, Charlotte 1996 "O enfraquecimento da concordância no portugués brasileiro " ["The weakening of agreement in Brazilian Portuguese"], in Ian Roberts and Mary Kato, eds, Portugués brasileiro: urna viagem diacrònica [.Brazilian Portuguese: a diachronic voyage], 387-408. Campinas, SP: Editora da UNICAMP. González, Nora 1988 Object and raising in Spanish, New York: Garland Publishing, Inc. Goodall, Grant 1993 "Spec of IP and Spec of CP in Spanish ^//-questions", in William Ashby, Marianne Mithun and Giorgio Perissinotto, eds., Linguistic Perspectives on the Romance Languages, 199-209, Amsterdam/Philadelphia: John Benjamins Publishing Co. Haegeman, Liliane 1991 Introduction to Government and Binding Theory, Cambridge, MA: Blackwell Publishers. Hale, Kenneth and Samuel Jay Keyser 1994 "On argument structure and the lexical expression of syntactic relations", in Kenneth Hale and Samuel Jay Keyser, eds., The view from Building 20, 53-109, Cambridge, MA: The MIT Press.

202

References

Herburger, Elena 1992 "Focus and the LF on the NP Quantification", ms., University of Southern California, LA. Hoekstra, Teun and Rene Mulder 1990 "Unergatives as copular verbs; locational and existential predication", The Linguistic Review 7, 1-79. Jaeggli, Osvaldo 1982 Topics in Romance syntax, Dordrecht: Foris. Jonas, Dianne and Jonathan Bobaljik 1993 "Specs for subjects: the role of TP in Icelandic", Papers on Case and Agreement 1, MIT Working Papers in Linguistics 18, 59-98. Johnson, Kyle 1991 "Object positions", Natural Language and Linguistic Theory 9, 577636. Kato, Mary and Eduardo Raposo 1994 "European and Brazilian Portuguese word order: Questions, Focus and Topic constructions", paper presented at LSRL XXIV, Los Angeles, CA. Kayne, Richard 1984 Connectedness and Binary Branching, Dordrecht: Foris. Kayne, Richard 1989 "Null subjects and clitic climbing", in Osvaldo Jaeggli and Kenneth Saflr, eds., The null subject parameter, 239-261, Dordrecht: Kluwer. Kayne, Richard 1991 "Romance clitics, verb movement and PRO", Linguistic Inquiry 22, 647-68. Kayne, Richard 1994 The Antisymmetry of Syntax, Cambridge, MA: The MIT Press. Kempchinsky, Paula 1985 "Brazilian Portuguese and the null subject parameter", MESTER, 316.

Koopman, Hilda and Dominique Sportiche 1991 "The position of subjects", Lingua 85, 211-258.

References Laka, Itziar 2000

203

"Thetablind Case: Burzio's Generalization and its image in the mirror", in Eric Reuland, ed., Arguments and Case: explaining Burzio's generalization (Linguistics Today 34), 103-129, Amsterdam/Philadelphia: John Benjamins.

Larson, Richard 1988 "On the double object construction", Linguistic Inquiry 19, 335-92. Lasnik, Howard 1995 "Case and Expletives revisited: on Greed and other human failings", Linguistic Inquiry 26,615-633. Lasnik, Howard and Mamoru Saito 1992 Move a, Cambridge, MA: The MIT Press. Levin, Beth and Malka Rappaport Hovav 1995 Unaccucsativity: at the syntax-lexical semantics interface, Cambridge, MA: The MIT Press. Lopes Rossi, Maria Aparecida Garcia 1996 "Estudo diacrònico sobre as interrogativas do portugués do Brasil" ["A diachronic study of interrogatives in Brazilian Portuguese"], in Ian Roberts and Mary Kato, eds., Portugués brasileiro: urna viagem diacrònica [Brazilian Portuguese: a diachronic voyage], 307-342. Campinas, SP: Editora da Unicamp. Mallén, Enrique 1994 "Verb movement in Spanish and the Licensing Condition on variables", Linguistische Berichte 148,445-476. Ordóffez, Francisco 1996 "The inversion construction in interrogatives in Spanish and Catalan", ms., The City University of New York. Ordófiez, Francisco 1997 Word order and clause structure in Spanish and other Romance languages, Doctoral dissertation, The City University of New York, NY. Ortiz-López, Luis 1993 "La regla de inversión de sujeto-verbo en la interrogativa del español de Puerto Rico: estudio de los diferentes contextos", ms., University of Iowa.

204

References

Ouhalla, Jamal 1991 Functional categories and parametric variation, New York: Routledge Pollock, Jean-Yves 1989 "Verb movement, universal grammar, and the structure of IP", Linguistic Inquiry 20, 365-424. Raposo, Eduardo 1994 "Affective operators and clausal structure in European Portuguese and European Spanish", paper presented at LSRL XXIV, Los Angeles, CA. Rizzi, Luigi 1982 Issues in Italian Syntax, Dordrecht: Foris. Rizzi, Luigi 1986 "Null objects in Italian and the theory of pro", Linguistic Inquiry 17, 501-557. Rizzi, Luigi 1991 "Residual verb second and the W7z-Criterion", Technical Reports in Formal and Computational Linguistics, Université de Génève. Rizzi, Luigi and Ian Roberts 1989 "Complex Inversion in French", Probus 1, 1-30. Rottet, Kevin 1993 "Variation and verb raising in Louisiana Creole: a GB perspective", paper presented at the 23rd Linguistic Symposium, Northern Illinois University. Saito, Mamoru and Hiroto Hoshi 1994 "Japanese light verb construction and the Minimalist Program", ms., University of Connecticut, Storrs. Seibert, Anna 1993 Intransitive Constructions in German and the Ergative Hypothesis, Working Papers in Linguistics, University of Trondheim. Slabakova, Roumyana 1997 "Bulgarian pre-verbs: Aspect in phrase structure", Linguistics 35, 673-704.

References

205

Sportiche, Dominique 1998 "Subject Clitics in French and Romance: Complex Inversion and Clitic Doubling", in Ian Roberts and Kyle Johnson, eds., Beyond Principles and Parameters {Studies in Natural Language and Linguistic Theory 45), 189-221, Dordrecht: Kluwer. Stowell, Tim 1996 "The syntax of Tense", in Johan Rooryck and Laurie Zaring, eds., Phrase Structure and the Lexicon, Dordrecht: Kluwer. Suffer, Margarita 1994 "V-movement and the licensing of argumentai wh-phrases in Spanish", Natural Language and Linguistic Theory 12, 335-372. Tarallo, Fernando and Mary Kato 1993 "Filling syntactic boundaries in spoken Brazilian Portuguese", Language Variation and Change 5, 91-112. Tenny, Carol 1987 Grammaticalizing aspect and affectedness, Cambridge, MA: MIT Working Papers in Linguistics. Torrego, Ester 1984 "On inversion in Spanish and some of its effects", Linguistic Inquiry 15, 103-129. Travis, Lisa 1988 "The syntax of adverbs", McGill working papers in Linguistics: Proceedings of the IVth workshop on Comparative Germanic Syntax, 280-310. Travis, Lisa 1991 "Derived objects, inner aspect and the structure of VP", NELS, U. of Delaware. Uriagereka, Juan 1995 "Aspects of the Syntax of Clitic Placement in Western Romance", Linguistic Inquiry 26, 79-123. Zanuttini, Raffaella 1997 Negation and Clausal Structure, New York: Oxford University Press. Zubizarreta, Maria Luisa 1992 "Word order in Spanish and the nature of Nominative Case", ms., University of Southern California, Los Angeles, CA.

206

References

Zubizarreta, María Luisa. 1994 "Grammatical representation of topic and focus: implications for the structure of the clause", Cuadernos de Lingüística del I. U. Ortegay Gasset 2, 181-208. Zubizarreta, María Luisa 1998 "Prosody, focus and word order", MIT monograph.

Index

A'-position, 155-156,175 adverb position, 31-36,49-50,6061,64 sentential adverbs, 51 -59 adjunction, 32-33, 55,64-71 agreement static, 15-16, 151 dynamic, 16-17, 20, 151, 161 antisimmetry, 156,161,172 Argumentai Agreement Licensing, 13, 154 Aspect, 40-43, 61-63, 67 binding, 120 Case accusative, 115-116 Case Filter, 7 checking, covert, 100,116 checking, overt, 105 inherent, 117, 123-124, 134 nominative, 98,113,114115,153 partitive, 109-110, 112-115, 117,123-125, 130, 134 structural, 8, 97,112,118, 123 Catalan, 21 c-command, 100-101 clitic doubling, 128-129 Clitic Inversion, 20 Closed Domain Condition, 14, 18,154

control, 120, 121 default order, 2,29-30 Definiteness Effect (DE), 109110,112, 134 discourse, 97 é que, 185 ECM, 124 Economy, 170, 180 embedded questions, 183-184 emphasis, 23 enclisis, 169, 170, 173 English, 7, 40, 119, 120, 183 EPP, 40, 100, 101, 103-104, 107, 121, 139, 165, 175,176-179 European Portuguese, 22-24, 168-169, 174, 185 expletives, 107,108,122 factive verb, 99 feature interpretable features, 38 phi-features, 119, 122 strong and weak features, 6 D/N feature, 116, 119, 127,179 V-feature, 176-179, 181, 182 wA-feature, 152-153, 157, 172, 176 Finnish, 112 focus, 23,132-136, 140 French, 7,19,40,45-46, 51, 52, 57-59, 74,109,119,

208

Index

120,144,146-147,163 Functional Phrase, 168,173 German, 84 Government and Binding, 7 Icelandic, 39, 117 interrogatives, 3, 92 islands, 148 it, 122,125-126,140 Italian, 4, 9,21,22,51,53,55, 79, 110,112-113,120, 144,145, 151-152,162, 166 Last Resort, 97 LCA (Linear Correspondence Axiom), 162 LFG, 98 light verb, 101,116,124 locative inversion, 94, 96-101, 107,117,139 Merge, 32 Minimal Link Condition (MLC), 46,65 Minimalist Program, 2, 5, 113, 115,123 Minimality, 97 A'-minimality, 156 Mirror Principle, 37 movement A'-movement, 148,150 w/i-movement, 148-149, 153,155, 156, 160 ne-cliticization, 79 null subject, 105-106 parametrical change, 186 parasitic gaps, 149 participial absolute (PA), 81, 88, 95,137

Past Participle as Adjective (PPA), 80-81, 87, 88, 95, 138 Principles and Parameters, 5, 9 pro, 103, 105-107, 117, 119, 121-122,125-126, 128, 131, 134, 139-140, 157, 165 proclisis, 169 Procrastinate, 6, 100, 116-117 Q-floating, 47 raising, 99,116, 120, 121 ser ¡estar, 3, 123 SCI (Subject Clitic Inversion), 147 shell, 65, 118, 136 small clause, 98,123 Spanish, 3, 8,11-12, 22, 33, 49-50, 61-63, 80-83, 132-133, 144, 151, 157-158,162,171, 183 Puerto Rican Spanish, 13, 17-18, 145 Stylistic Inversion, 19, 147 superiority effects, 149 Tense, 6, 7 there, 103, 108, 120, 121-122, 125-126, 132, 140 topic, 132-134 topicalization, 54, 56 Transitive Expletive Construction (TEC), 39 Universe of Discourse (UD), 132 Unnacusative Hypothesis, 78, 139 verb morphemes, 186

Index V O S , 20-21,105,108,127-128, 140,168 V S O , 21,105,127-130,140, 168

209

^ - C r i t e r i o n , 10,15-17,150153,159,186