Brazilian Portuguese and the Null Subject Parameter 9783964561497

Table of contents :
TABLE OF CONTENTS
PREFACE
THE LOSS OF THE AVOID PRONOUN' PRINCIPLE IN BRAZILIAN PORTUGUESE
THE PRONOMINAL SUBJECT IN ITALIAN AND BRAZILIAN PORTUGUESE
VISIBLE SUBJECTS AND INVISIBLE CLITICS IN BRAZILIAN PORTUGUESE
NULL SUBJECTS IN BRAZILIAN PORTUGUESE: DEVELOPMENTAL DATA FROM A CASE STUDY
BRAZILIAN PORTUGUESE AS A DISCOURSE-ORIENTED LANGUAGE
MAIN AND EMBEDDED NULL SUBJECTS IN BRAZILIAN PORTUGUESE
NULL SUBJECTS WITHOUT 'RICH' AGREEMENT
BRAZILIAN PORTUGUESE VS ORDER: A DIACHRONIC ANALYSIS
SYNTACTIC CODIFICATION OF CATEGORICAL AND THETIC JUDGMENTS IN BRAZILIAN PORTUGUESE
THE PARTIAL PRO-DROP NATURE AND THE RESTRICTED VS ORDER IN BRAZILIAN PORTUGUESE
INDEX
LIST OF CONTRIBUTORS

Citation preview

Mary Aizawa Kato / Esmeralda Vailati Negrao (Eds.) Brazilian Portuguese and the Null Subject Parameter

Editionen der Iberoamericana Ediciones de Iberoamericana Serie A: Literaturgeschichte und -kritik / Historia y Crítica de la Literatura Serie B: Sprachwissenschaft / Lingüistica Serie C: Geschichte und Gesellschaft / Historia y Sociedad Serie D: Bibliographien/Bibliografías

Herausgegeben von / Editado por: Walther L. Bernecker, Frauke Gewecke, Jürgen M. Meisel, Klaus Meyer-Minnemann B: Sprachwissenschaft / Lingüística, 4

Mary Aizawa Kato Esmeralda Vailati Negrao (Eds.)

Brazilian Portuguese and the Null Subject Parameter

Vervuert- Iberoamericana 2000

Die Deutsche Bibliothek - CIP-Einheitsaufnahme [Iberoamericana / Editionen / B] Editionen der Iberoamericana=Ediciones de Iberoamericana. Serie B, Sprachwissenschaft = Lingüística. Frankfurt am Main : Vervuert Reihe Editionen, Serie B zu: Iberoamericana. Hervorgegangen aus: Iberoamericana / Editionen / 03 4. Brazilian Portuguese and the Null Subject Parameter. - 2000 Brazilian Portuguese and the null subject parameter / Mary Aizawa Kato ; Esmeralda Vailati Negräo (ed.). Madrid : Iberoamericana; Frankfurt am Main : Vervuert, 2000 (Editionen der Iberoamericana : Serie B, Sprachwissenschaft ; 4) ISBN 84-95107-69-4 (Iberoamericana) ISBN 3-89354-883-1 (Vervuert) © Iberoamericana, Madrid 2000 © Vervuert Verlag, Frankfurt am Main 2000 All rights reserved Cover Design: Michael Ackermann Illustration: Sketch for „Angelus" by Ernest Mange (1999) Printed on acid free paper Printed in Germany

5

Table of Contents

TABLE OF CONTENTS PREFACE MARY A IZAWA KA TO

7

T H E LOSS OF THE 'AVOID PRONOUN' PRINCIPLE IN BRAZILIAN PORTUGUESE MARIA EUGENIA LAMOGLIA DUARTE THE

PRONOMINAL

SUBJECT

IN

ITALIAN

AND

BRAZILIAN

PORTUGUESE

MARILZA DE

37

OLIVEIRA

VISIBLE SUBJECTS AND INVISIBLE CLITICS IN BRAZILIAN PORTUGUESE SONIAM. L. CYRINO, M. EUGENIA L. DUARTE AND MARY A. KATO N U L L SUBJECTS IN BRAZILIAN PORTUGUESE: DEVELOPMENTAL DATA FROM A CASE STUDY LUCIENE J.

17

55

75

SIMOES

BRAZILIAN PORTUGUESE AS A DISCOURSE-ORIENTED LANGUAGE ESMERALDA VAILATI NEGRAO AND EVANI VIOTTI MAIN AND EMBEDDED PORTUGUESE MARIA CRISTINA FIGUEIREDO

NULL

SUBJECTS

IN

105

BRAZILIAN 127

SIL VA

N U L L SUBJECTS WITHOUT ' R I C H ' AGREEMENT

147

Marcello MODESTO BRAZILIAN PORTUGUESE V S ORDER: A DIACHRONIC ANALYSIS BERLINCK

175

Rosane de ANDRADE SYNTACTIC

CODIFICATION

OF

CATEGORICAL

AND

THETIC

JUDGMENTS IN BRAZILIAN PORTUGUESE HELENA

195

BRITTO

T H E PARTIAL P R O - D R O P NATURE AND THE RESTRICTED O R D E R IN BRAZILIAN PORTUGUESE

VS 223

MARY A IZAWA KA TO INDEX

259

LIST OF CONTRIBUTORS

269

Preface

7

PREFACE Mary A izawa KA TO The generative conception of language in the Principles and Parameters (PP) program has opened two important fields in comparative grammar: a) crosslinguistic synchronic analysis and b) intralinguistic diachronic studies. In both enterprises, the concept of parameters and their positive or negative setting are fundamental in handling the contrasts found crosslinguistically in natural languages or diachronically within the same language. Synchronic language internal variation is not explicitly planned as an enterprise in the theory, but the answer to an interview question, following The Managua Lectures, shows that Chomsky considers language internal variation to be a case of the co-existence of different languages: Question: A child can learn two languages simultaneously, one in the house and the other in the street. Does that mean that the child relates the position of the switches to the environment? Answer: Well, this is a very important question which I have been pretending all along does not arise. The question is a very mysterious one, because the child learns different languages, say Spanish at home and English in the streets. But, in fact, the problem is really more general, because every human being speaks a variety of languages. We sometimes call them different styles or different dialects, but they are really different languages, and somehow we know when to use them, one in one place and another in another place. N o w each of these different languages involves a different switch setting. In the case of Spanish/English it is a rather dramatically different switch setting, more so than in the case of the different styles of Spanish that each of you has mastered. (Chomsky 1988: 188)

It becomes clear, then, that the linguist working synchronically with one language can face differences in linguistic judgement and also differences in language behavior, since in the same object there may be different grammars at play. The result is that whatever the aim of description — synchronic or diachronic — the linguist working in the principles and parameters framework will be doing comparative grammar while focusing on some proposed parameter. One of the most discussed parameters, not only in its theoretical formulation, but also in the ways it functions in language acquisition and language change, is the null subject (NS) parameter, also referred to as the 'pro-drop parameter'. Underlying the latter terminology is the assumption that the empty subject is always a pronoun, while in the former there is no such assumption. The idea of 'parameter' was first conceptualized as a possibility of violation of some principle. Thus, just as Principle A of the Binding Theory could be violated by languages that have long distance anaphors, some languages could

Mary Aizawa KATO

8

violate the Extended Projection characterized in this way.

Principle (EPP).

NS

languages could

be

Parametric variation is later reconceptualized as a function o f the lexicon, more specifically o f the functional lexicon. In this perspective, a parameter is not understood as a particular syntactic property o f the surface, like the obligatoriness or the optionality o f lexical subjects, but as some abstract morphological property o f functional heads. In the case o f the N S parameter the functional category in question was first considered to be I N F L and its agreement features. However, the existence o f empty subjects has also been correlated with languages that are topic prominent, or discourse oriented, like Chinese and Japanese, in which agreement inflection is considered to be nonexistent. This led the N S parameter, when positively-marked, to be further parametrized in terms o f the way the empty category is identified: by inflection or by an antecedent. As it is, the N S parameter actually defines at least three types o f languages, as seen in the figure below: Fig. 1: NS

Identification by I N F L

a Italian Spanish

c English French

b Chinese Japanese

Notice that languages o f type a. and b. can be distinguished in terms o f morphology, but there is still some common property that puts a. and b. together, as opposed to c. It seems, therefore, that non-violation o f the E P P is still needed to account for c. A parameter has also been defined as a set o f surface phenomena derivable from the underlying abstract property, the N S parameter being a good example o f this. According to Rizzi ( 1 9 8 2 : 117) and Chomsky ( 1 9 8 1 : 2 4 0 ) , N S languages tend to manifest the following clustering o f properties:

9

Preface

(i)

missing subjects [1] Ho trovato il libro. ('I found the book')

(ii)

free inversion in simple clauses [2] Ha telefonato Gianni. ('Gianni called')

(iii)

'long wh-movement' of subjects [3] L'uomo [che mi domando [chi abbia visto]] (with the interpretation: 'the man x such that I wonder who x saw')

(iv)

empty resumptive pronouns in embedded clauses [4] Ecco la ragazza [che mi domando [chi crede [che possa VP]]] ('this is the girl who I wonder who thinks that she may VP')

(v)

apparent violation of the *[that-t] filter [5] Chi credi [che partira]? ('who do you think [(that) will leave]')

Though free inversion is considered by many linguists to be a property of the NS parameter, such a correlation is not found in discourse oriented languages like Japanese and Chinese, and has even been contested in the study of some dialects of Italian, where the INFL system is extremely rich in affixes and clitics (Safir 1982; Franchi & Ilari 1986; Nicolau 1996). As for the other properties, the diagnostics are difficult because there are other intervening factors, such as the fact that these languages have wh-in-situ and that Japanese has an empty complementizer. Therefore, the nature of the abstract property and the predictable cluster of surface properties that define a NS language constitute a major problem in the PP theory. In this collection of articles, the N S parameter and the questions that it raises are studied in Brazilian Portuguese (BP), both from a diachronic and a synchronic perspective and also covering questions regarding language acquisition. The challenge of studying BP resides in two aspects: a)

it is argued that, since the end of the 19th Century, BP has been undergoing changes that will ultimately lead to a reestructuring of its grammar (see Roberts & Kato 1993; and the articles therein);

b)

in the most extensively described variety, which is the dialect spoken in the Southeast of the country (S2o Paulo, Rio and Minas Gerais), BP does not fit any of the sub-types in Fig. 1, in its synchronic stage.

Two interpretations can be given to the facts in b: i) BP data is a mixture of the old and new grammars in competition; and ii) BP data is all licensed within a single stable grammar. The two positions can be found in this collection. Obviously, those who work in a diachronic perspective tend to embrace the

Mary Aizawa KATO

10

former view and those who work synchronically tend to assume the latter position. The empirical diachronic facts attested to in previous literature on BP and pertinent to the topic under investigation are the following: a) b)

a substantive loss of referential null subjects in the first and second persons, and a more gradual loss in the third (Duarte 1993, 1995); an increase of referential null objects (Tarallo 1983; Cyrino 1994);

c)

an expressive increase of subject doubling (Duarte 1995; Britto 1998);

d)

the increase of the SVO order, where VSO or VOS would be expected in N S languages (Torres Morais 1995; Lopes Rossi 1996; and Andrade Berlinck 1995).

Most of these studies also provide a theoretical analysis of one of these changes, but there are also studies that try to interpret a set of phenomena attributing them to some abstract property of grammar (cf. Galves 1993; and Roberts 1993). Previous synchronic studies on BP have focused on a)

theoretical issues such as: a . l ) the asymmetry between the null subject and the null object (Moreira da Silva 1983); a.2) the restrictive nature of free inversion (Nascimento 1984; Kato & Tarallo 1988); a.3) the structure of unaccusatives (Nascimento & Kato 1995); a.4) the licensing conditions for the N S and inversion (Figueiredo Silva 1994); a.5) the nature of the AGReement head (Galves 1993, 1998; Nicolau 1996); a.6) the functions of left dislocation and the order SV (Britto 1998); a.7) the inexistence of pro (Kato 1999); and a.8) the nature of the third person empty category in BP (NegrSo & MUller 1996; and NegrSo 1997);

b)

the comparison of BP with other N S languages as in: b. 1) Galves (1987), who compares the null subject in BP and in EP; and b.2) De Oliveira (1996), who compares the use of the NS in Italian and in BP (see also this volume);

c)

the description of spoken BP as in: c . l ) realization and position of subjects (Kato et alii 1998); and c.2) existential structures (Franchi et alii 1998).

The relevant results are summarized in the studies included in this volume. The present volume contains ten articles, five on the possibility and nature of empty subjects (Duarte, De Oliveira, Figueiredo Silva, NegrSo & Viotti and Modesto), two on the (im)possibility and nature of VS order (Andrade Berlinck and Britto), one on the correlation between the null subject and free inversion (Kato), one on

Preface

11

the correlation between null subjects and empty objects (Cyrino, Duarte & Kato), and one on the null subject in the acquisition of BP (SimSes). The articles that deal with the empty subject can be divided into two types: a) those that discuss the three grammatical persons to examine what sort of agreement paradigm licences the null subject (Duarte and De Oliveira); and b) those that take the perspective of Binding (Figueiredo Silva, Modesto and NegrSo & Viotti). The conclusions in these papers suggest that while with the first and second persons, marked as [+person] in Benveniste's terms (cf. Nicolau 1996; and Galves 1998), the NS phenomenon may be exclusively a function of morphology, the NS of the third person, or [-person], can involve more than one type of licensing and identification, or even a distinct type of category. In the first group, the discussion centers around the definition of inflection as 'rich' or 'poor'. The strongest claim is that of Duarte, for whom the neutralization of the second and third persons is making BP change into a nonNS language. Not only does she find massive use of overt referential subject pronouns, but also the beginning of the loss of null subjects in non-referential contexts. De Oliveira compares BP with contemporary Italian. Using Italian facts, her claim is that a neutralization that affects first and second person does not affect identification, while the neutralization that involves the third person affects licensing of pro. For this author, the presence of lexical pronouns is also subject to morphological properties. Thus, lui, an accusative form in Italian, can only be co-indexed with an object, requiring pro for the co-indexed subject, while ele, a nominative pronoun, can be co-indexed with a subject. The second group works basically with the binding properties of the null third person. Figueiredo Silva claims that BP has really lost referential pro, but her definition of a referential pronoun is different from that of the first group. In her analysis, the first person is possible in root contexts, when CP is not involved: Comprei o carro ( (I) bought the car) vs *Onde comprei o carrol (where (I) bought the car?). According to her analysis, the empty subject is a variable of a [+person] DP in spec of CP. Since in wh-questions Spec is occupied, the subject cannot move to that position, hence no null subject is possible in such contexts. Concerning embedded null subjects, she considers the possibility of splitting them into two types: 'anaphoric' and 'variable' null subjects. She presents some problems for this analysis and suggests that Modesto's analysis, in this volume, may be the correct one to solve the problems raised. This means that the distinction between null subjects that are A-bound and null subjects that are A'bound may only be an apparent one, in BP. Modesto claims that pro is possible in BP only in contexts where it can be bound by another DP in an A'-position, AGRP being an A'position above TP. Confirmation of the analysis is obtained by examining the behavior of null subjects in islands and discussing null objects in Chinese, which can also be analyzed as A'-bound null pronominals.

12

Mary A izawa KA TO

NegrSo & Viotti's analysis falls within the second group, but they start with the claim that there is no correlation between the loss of verbal inflectional morphology and the decrease in the use of null subjects. Following Galves (1993) and Figueiredo Silva (1994), who have argued that the loss or the weakening of the verbal inflectional paradigm in BP has triggered a restructuring of its sentence pattern that has approximated BP to other discourse-oriented languages, the article claims that Brazilian Portuguese is a discourse-oriented language. Rather than the substitution of a null subject for a lexical subject, they propose a specialization in the use of overt pronouns as opposed to empty pronouns. In the contexts analyzed, overt pronouns are interpreted as E-type pronouns, and such an interpretation is dependent upon a certain structural configuration that corroborates the claim that BP is a discourse-oriented language. The work of Cyrino, Duarte & Kato tries to make sense of the correlation found by Tarallo (1983) between decrease of null subjects and increase of null objects. The authors claim that the triggers for the changes were independent, but that both changes were governed by the same 'referential hierarchy', which optimally maps referential (((-feature items with lexical DPs and non-referential (|>-feature items with empty categories. Still within the NS topic, there is a chapter by SimSes on the null subject in language acquisition. Her work examines the properties of early null subjects in Brazilian Portuguese, and finds support for the view that, from early stages, children acquiring BP exhibit a highly constrained type of null category in subject position and that their speech is grammar-based concerning the use of null arguments. She shows that, when sentences involve CP elements and embedded contexts, the child's use of null subjects is not asystematic or driven by other kinds of restrictions, similar to the ones found for other empty categories examined both in child and adult languages. No evidence in her sample of data favored the expectation of a future parameter resetting in the children's development, and the observed properties of sentences lead to the interpretation that the empty category present in null referential contexts is similar to the one described in Figueiredo Silva (1994) for adult I-language. The volume includes two papers that deal mainly with the problem of subject/verb inversion and one that works with the correlation of null subjects with VS syntax. Andrade Berlinck's previous diachronic work (1989, 1995) had already attested to the loss of VS syntax in BP. In the chapter she develops for this volume, she undertakes a careful quantitative analysis comparing the loss of VSO and VOS, on one hand, and inergative VS vs unaccusative/ergative VS, on the other. She concludes that both VSO and VOS have been almost categorically lost and that the remaining VS structures are with unaccusative verbs, which can

13

Preface

be derived with the postverbal DP in-situ. Her results show that, when compared to Duarte's (1993) data, there is an effective correlation between free inversion and null subjects. Britto's work tries to correlate the loss of inversion and null subjects with the finding that BP uses subject doubling and Left Dislocation (LD) profusely, without the marked reading of LD in other languages. LD constructions in BP and SV constructions with lexical subjects in EP are compared and found to be equivalent from the interpretative point of view: both represent the syntactic codification of categorical judgements in the sense of Kuroda (1972). On the other hand, the comparison of SV constructions in BP with VS structures in EP reveals yet another similarity: both are interpreted as the syntactic codification of thetic judgement. She then discusses the syntactic derivation of this parallelism. The empirical correlation found between subjects and free inversion is proposed by Kato to be derivable from the structure of a NS language. NS languages are claimed to have no spec of TP projected: agreement affixes are pronominal, and, like clitics, are merged as arguments, moving to T to check Case and «(¡-features. The definite lexical subject of a NS language is claimed to be external to TP, having a 'default' nominative case. Indefinite NPs are internal to VP, headed by the pronominal agreement, which moves as a head to T, therefore not pied-piping the nominal element; in the case of the definite DP, it is the TP that moves over the external lexical subject. BP has lost both types of movements since agreement ceased to be pronominal and, as a consequence, Spec of TP is projected. The author argues that the pronominal agreement affixes were replaced by weak pronouns, except for the neuter -0, which can still function as the //-type expletive in existentials and unaccusative VS constructions. These are attested to be the only currently productive VS constructions according to Andrade Berlinck's diachronic analysis in this volume. This collection of papers still raises other interesting questions: a)

Are there languages that have referential null subjects, but no null subject bound pronouns? If not, how can these two types of null subjects be related?

b)

Figueiredo Silva claims that BP has no null resumptive pronouns, which means that BP does not comply with the third and fourth properties of NS languages. On the other hand, it complies with the 'that-t' filter, though it lost post verbal subjects, allegedly the position from which subjects are extracted. How can Fig. 1 be reformulated in order to accommodate a language such as BP? Can its properties be attributed to change underway alone?

14

Mary Aizawa KATO

c)

There is a claim that BP is a discourse oriented language (cf. Pontes 1987). Negnio & Viotti explain the NS in BP under this hypothesis. How do the other facts in BP fit this typology?

d)

For Adams (1983), Old French lost its VSO word order and the null subject was lost along with it, both the postverbal S and the null subject being assumed to have been licensed by government. However, Italian does not have VSO order and is still a NS language. How may we account for the fact that in one language VSO and the null subject seem to have been correlated and in another the null subject is possible without this order?

The final question has to do with the explanatory level: "how does a child get to know all that we have learned about BP as seems to have been the case of the children in SimSes' study? Of course the question is unanswerable without UG, but even with a theory that includes a parameter such as the NS, the question that remains is: what triggering experience does the child use to learn BP grammar as opposed to Italian grammar? It seems that BP grammar is actually a sub-set of Italian grammar. In this sense, the child should start using third person null subjects, like BP, before using the null subject for the other persons (see this proposal in Kato 1999). If no evidence to extend the null subject to other persons appears, the child will continue speaking BP. On the other hand, if the child hears the null subject with the first person in the primary data, learning that the language is a strong NS language like Italian is also instantaneous, considering the referential hierarchy proposed by Cyrino, Duarte & Kato (this volume). Despite some different viewpoints and interpretations of the data, on the whole this book presents an important consensual picture of the nature of BP: a) it is different from a prototypical N S language like Italian; b) the parameter does not apply equally to all persons; and c) it has peculiarities, but they are systematic, and therefore leamable. From the range of perspectives — comparative grammar, diachronic analysis and language acquisition — the book is a contribution to the PP theory and its research agenda. What is most important in the book, however, is the explicit questions that are raised for future research. The development of many of the works contained in this book was due to research grant by Brazilian funding agencies: Esmeralda Vailati NegrSo (CNPq / FAPESP), Evani Viotti (CNPq), Helena Britto (FAPESP), Marcello Modesto (CNPq), Maria Eugénia Lamoglia Duarte (CNPq), Rosane de Andrade Berlinck (CNPq), Sonia M. L. Cyrino (CNPq), and Mary A. Kato (CNPq / FAPESP).

Preface

15

References ADAMS, Marianne (1987) Old French, Null Subjects and Verb-second Phenomena. UCLA, Ph.D. Dissertation. ANDRADE BERLINCK, Rosane de (1989) "A construyo VSN no Portugués do Brasil: urna visào diacrònica do fenòmeno da ordern". In: F. Tarallo (ed.) Fotografías Sociolingüísticas. Campinas, SP, Editora da UNICAMP/Pontes: 95-112. ANDRADE BERLINCK, Rosane de (1995) La Position du Sujet en Portugals: étude diachronique des variétés brésilienne et européenne. Leuven, Katholieke Universiteit, Ph.D. Dissertation. BRITTO, Helena de Souza (1998) Deslocamento à Esquerda, Resumptivo-sujeito, Ordern SV e a Codificando Sintática de Juizos Categórico e Tético no Portugués do Brasil. Campinas, SP, UNICAMP, Ph.D. Dissertation. CHOMSKY, Noam (1981) Lectures on Government and Binding. Dordrecht, Foris. CHOMSKY, Noam (1988) Language and Problems of Knowledge. The Managua Lectures. Cambridge, Mass., The MIT Press. CYRINO, Sonia (1994) O objeto nulo no Portugués do Brasil: um estudo sintáticodiacrönico. Campinas, SP, UNICAMP, Ph.D. Dissertation. CYRINO, Sonia (1997) O objeto nulo no Portugués do Brasil: um estudo sintáticodiacrónico. Londrina, Editora VEL. DE OLIVEIRA, Marilza (1996) Respostas Assertivas e sua Variando na Linguas Románicas: seu papel na aquisifäo. Campinas, SP, UNICAMP, Ph.D. Dissertation. DUARTE, Maria Eugenia L. (1993) "Do pronome nulo ao pronome pleno". In: I. Roberts & M. A. Kato (eds.) Portugués Brasileiro: Urna viagem diacrònica (Homenagem a Fernando Tarallo). Campinas, SP, Editora da UNICAMP: 10728.

DUARTE, Maria Eugenia L. (1995) A Perda do Principio 'Evite Pronome' no Portugués Brasileiro. Campinas, SP, UNICAMP, Ph.D. Dissertation. FIGUEIREDO SILVA, Maria Cristina (1994) La Position Sujet en Portugals Brésilien. Università de Genève: Doctoral Dissertation, Published in Portuguese (Editora da UNICAMP), 1996. FRANCHI, Carlos & Rodolfo ILARI (1986) "Clíticos nominativos e inversäo do sujeito em bielès". In: D.E.L.T.A. 2 (1): 77-103. FRANCHI, Carlos, Esmeralda V. NEGRÄO & Evani VIOTTI (1998) "Sobre a gramática das o r a l e s impessoais com Ter/Haver". D.E.L.T.A. 14. N.° Especial: 105-32. GAL VES, Charlotte (1987) "A sintaxe do portugués brasileiro". Ensaios Lingüísticos, 13: 31-50. GAL VES, Charlotte (1993) "O enfraquecimento da concordancia no Portugués Brasileiro". In: I. Roberts & M. A. Kato (eds.): 387-408. GALVES, Charlotte (1998) "Tópicos e sujeitos, pronomes e concordancia no portugués do Brasil". Cadernos de Estudos Lingüísticos 34: 19-32. KATO, Mary A. (1999) "Strong pronouns, weak pronominals and the null subject parameter". PROBUS 11 (1): 1-37.

16

Mary Aizawa

KATO

KATO, Mary A. & Fernando TARALLO (1988) "Restrictive VS syntax in Brazilian Portuguese: its correlations with invisible clitics and visible subjects". 38th Georgetown Roundtable in Language and Linguistics. Campinas, SP, UNICAMP, ms. KATO, Mary A., Eunice NICOLAU, Rosane de ANDRADE BERLINCK & Helena BRITTO (1998) "Padröes de P r e d i c a l o na Gramática do Portugués". In: M. A. Kato (ed.) Gramática do Portugués Falado, Vol V. Campinas, SP, Editora da UNICAMP/FAPESP: 201-74. KURODA, S.-Yuki (1972) "The Categorical and The Thetic Judgment". Foundations of Language 9: 153-85. LOPES ROSSI, Maria Aparecida (1996) A Sintaxe Diacrònica das Interrogativas-Q do Portugués. Campinas, SP, UNICAMP, Ph.D. Dissertation. MOREIRA DA SILVA, Samuel (1983) Études sur la Symétrie et l'Asymétrie Sujet/Objet dans le Portugals du Brésil. Università de Paris Vili, Ph.D. Dissertation. NASCIMENTO, Milton do (1984) Sur la Posposition du Sujet dans le Portugals du Brésil. Université de Paris VIII, Ph.D. Dissertation. NASCIMENTO, Milton do & Mary A. KATO (1995) "O estatuto dos nomináis pósverbais dos verbos inacusativos". Revista de Estudos da Linguagem, ANO IV, n.° 3: 31-74. NEGRÄO, Esmeralda V. (1997) "Asymmetries in the Distribution of Overt and Empty Categories in Brazilian Portuguese". In: J. R. Black & V. Motapanyane (eds.) Clitics, Pronouns and Movement. Current Issues in Linguistic Theory 140. Amsterdam & Philadelphia: John Benjamins Ltd. NEGRÄO, Esmeralda V. & Ana Lucia MÜLLER (1996) As mudan?as no sistema pronominal brasileiro: substituido ou especializado de formas. D.E.L.T.A. 12 (1): 125-52. NICOLAU, Eunice (1996) As Propriedades de Sujeito Nulo e de Inversäo no Portugués Brasileiro. Campinas, SP, UNICAMP, Ph.D. Dissertation. PONTES, Eunice (1987) O Tópico no Portugués do Brasil. Campinas, SP, Pontes. RIZZI, Luigi (1982) Issues in Italian Syntax. Dordrecht, Foris. ROBERTS, Ian (1993) Verbs and Diachronic Syntax. Dordrecht, Kluwer. ROBERTS, Ian & Mary A. KATO (eds.) (1993) Portugués Brasileiro: urna viagem diacrònica (Homenagem a Fernando Tarallo). Campinas, SP, Editora da UNICAMP. SAFIR, Ken (1982) Syntactic Chains and the Defìniteness Effect. Cambridge, Mass.: The MIT Press, Ph.D. Dissertation. TARALLO, Fernando (1983) Relativization Strategies in Brazilian Portuguese. University of Pennsylvania, Ph.D. Dissertation. TORRES MORAIS, Maria Aparecida (1995) Do Portugués Clàssico ao Portugués Moderno: um estudo da cliticizafäo e do movimento do verbo. Campinas, SP, UNICAMP, Ph.D. Dissertation. ZUBIZARRETA, Maria Luiza (1998) Word Order, Prosody and Focus. Cambridge, Mass.: The MIT Press.

The Loss of the 'Avoid Pronoun' Principle in Brazilian Portuguese

17

THE LOSS OF THE AVOID PRONOUN' PRINCIPLE IN BRAZILIAN PORTUGUESE Maria Eugenia Lamoglia DUARTE1 (UFRJ) 0.

Introduction

This chapter brings together the results of research attempting to trace the course of a parametric change in progress in Brazilian Portuguese (BP), which is evolving from a null to a non-null subject language. In a number of diachronic studies (Tarallo 1983; Berlinck 1989, 1996; Duarte 1992, 1993; Lopes Rossi 1993; Torres Morais 1994; Ribeiro 1994; Cyrino, Duarte & Kato (this volume), among others), BP has been shown to have lost the obligatory subject omission in certain syntactic environments (see Calabrese 1986, for Italian; and Fernandes Soriano 1989, for Spanish) and the VS order, a behavior that clearly sets it apart from other Romance null subject languages, such as Italian, Spanish and European Portuguese, and, at the same time, makes it more similar to French, a non-null subject language. The purpose here is to use the Principles and Parameters framework (Chomsky 1981), which presents the set of properties characterizing null subject languages, in association with certain theoretical assumptions of the variationist approach to language change (see Weinreich, Labov & Herzog 1968), to show, on the basis of quantitative and qualitative analyses, the progress already made in this direction and to predict further developments. Section 1 presents diachronic results which clearly suggest that the decrease in null referential subjects can be related to the reduction of the inflectional paradigm of verbs in BP. The section also presents the results of the analysis of a synchronic sample, which corroborates the results of the diachronic study and shows a further consequence of the change: the tendency to fill arbitrary reference subjects. In Section 2, BP and European Portuguese (EP) are compared in relation to the expression of definite and arbitrary referential subjects. The section closes with the claim that the defective system of null subjects exhibited by BP today is a stage in the parametric change towards obligatorily filled pronominal subjects. Section 3 presents evidence of the 'embedding' or the 'side effects' of this change into the system: the appearance of left-dislocated subjects, a structure which is not compatible with a pro-drop language but which is typical of non-pro-drop systems, such as French. A comparison is then made of these results in BP and some examples from Middle French offered by Vance (1989). It will be shown that, while exhibiting a

1

This paper had the support of CNPq (Proc. n.° 35 07 31/99-3).

Maria Eugènia Lamoglia

18

DUARTE

defective system of null subjects, Middle French initiated the licensing of leftdislocated constructions, a consequence of the cliticization process underway in a situation very similar to what one finds in BP today. Section 4 presents some data which makes it possible to suggest hypotheses regarding the progress of the change in the direction of the filling of nonreferential (expletive) subjects. Final remarks appear in Section 5.

1.

From null to full referential subjects

1.1.

Definite referential evidence of change

subjects:

diachronic

and

synchronic

The loss of the property that leads to the obligatory omission of pronominal subjects in certain syntactic environments (see Calabrese 1986) has been related by some authors (see, among others, Roberts 1993a, for French, and Galves 1993, for BP) to a drastic reduction in the inflectional paradigm of the verb. In BP, such a reduction has been motivated by a change in the set of pronominal subjects rather than by phonological processes. As a consequence, a paradigm which once exhibited six different inflectional endings has been reduced to four and, in the speech of the members of the younger generation, to three 2 . The reduction began with the replacement of the second person pronominal forms tu ('you'), 2nd singular, and vói ('you'), 2nd plural, which combined with verb forms bearing exclusive inflections, with the former address form vocé(s) ('you'), which combines with third person verb forms. More recently, the first person plural pronoun nós ('we') is being replaced by the pronominal expression a gente3 ('one'), which also requires the third person singular verb form. Table 1 below shows the three paradigms for the regular verb: amar (to love):

2

In some tenses, such as the imperfect of the indicative and the subjunctive mood, only two distinctive forms are used. 3 The pronoun a gente is a grammaticalized form originated from the noun gente ('people') (see Lopes 1999). A s a pronoun, it can bear either definite reference, 'we' (I and other people), or arbitrary reference, 'one'.

19

The Loss of the 'Avoid Pronoun' Principle in Brazilian Portuguese

Table 1: Pronominal and inflectional paradigms in BP Pers./Nb. 1" sing. 2 nd sing. 3rd sing. 1* plur. 2 nd plur. 3rd plur.

Pronouns Eu Tu Voce Ele/Ela Nós A gente Vós Voces Eles/Elas

Paradigm 1 Paradigm 2 am o am o am a s am a am a am a am a am a mos am a mos am a am a is am a m am a m am a m am a m

Paradigm 3 am o -

am a am a -

am a -

am a m am a m

In a diachronic study based on popular plays written in the 19th and 20th centuries, Duarte (1993) shows that, in fact, the tendency to fill pronominal subject positions can be related to the impoverishment of the inflectional paradigm of BP verbs. The results are presented in Figure 1 below. Notice that coordinated structures with coreferential subjects are not included in the analysis because null subjects in such structures are not exclusive to null subject languages: Fig. 1: Full pronominal subjects during seven periods (%)

The first three periods examined reveal Paradigm 1, compatible with the 'functional richness' to which Roberts (1993a) refers - one zero ending and two synchretisms - perfectly able to license and identify null subjects in all grammatical persons. Expressed subjects in these plays, particularly those in the

20

Maria Eugènia Lamoglia DUARTE

third person, seem to be used for emphasis or focus, as is usual in pro-drop systems (see Calabrese 1986, for Italian). The next two periods represent Paradigm 2, revealing that, with four distinctive endings, the functional richness has become weaker and the expression of the pronominal subject involves a variable rule. Finally, in the last two periods, in which Paradigms 2 and 3 coexist, the rate of expressed subjects rises significantly. An examination of these data reveals that the effects of the erosion of the inflectional paradigm are felt gradually and do not act uniformly on all grammatical persons. In the case of BP, the data shows that the second person was the first to change, increasing from an average of 18% of full subjects in the first three periods to 75% in the fourth, a rate that is maintained throughout the final period, where 78% of full subjects are observed. Full first person pronouns also increase dramatically, albeit more gradually, from 31% in the 1845 text to 82% in the 1992 one. It is the third person pronouns that offer the strongest resistance to an expressed subject. An average of 28% of full third-person subjects prior to 1937 4 increases from 41% to 48% in the more recent texts analysed. The problem is how to explain such resistance when it is specifically in the third person form, especially the singular, that the ambiguity is found, and there are fewer chances for referential identification from the inflection. NegrSo and Muller (1996) and NegrSo (1997) suggest that there has been a certain stabilization in the process and that the alternation filled/empty subject in BP could be explained by postulating a bound empty category. Another possible answer would lie in the hypothesis that, once full identification from the inflection is lost, the third person pronoun would depend on the N P in the previous context, that is, there would be an external referent to reinforce the identification of an empty pronoun; hence, the stronger resistance to the implementation of the change in the third person. This is the hypothesis defended by Duarte (1995) and sustained here. As we will show in the remaining of this section and on the basis of evidence from Middle French, the change seems to be in progress and the difference between null subjects in BP and other Romance languages is not qualitative. Rather, it is quantitative. In other words, the contexts in which one still finds null subjects in BP are those described as obligatory in a null subject language (see Calabrese 1986, for Italian, and Fernandes Soriano 1989, for Spanish). The difference is that now such obligatory null subject contexts have become optional 5 , a phenomenon which is part of a process of change.

4

This period seems to have been decisive for the loss of obligatory V S order in questions, as this structure decreases from then on (Duarte 1992). On the basis of the analyses, Kato & Duarte (1998) show that it was the variable system of null subjects triggered the reanalysis of V S into SV in such structures. 5 According to Kroch (1989) such optionality would reflect the co-existence of grammars in competition.

whtwo that two

21

The Loss of the 'Avoid Pronoun' Principle in Brazilian Portuguese

The second step in this study consisted of the examination of the oral language, in order to test the hierarchy found in the written language and verify the behavior of the third person pronouns. This study involved research based on the contemporary speech of 12 college-educated speakers born in Rio de Janeiro belonging to three age groups 6 (Duarte 1995). Figure 2 below suggests a change in progress. Fig. 2: Full subjects in spontaneous speech according to grammatical person and age group (%

100% 90% 80% (V 70% 60% 80% 40% 30% 20%

- - 4 r - -

10% 0%

Group 1 - - < > • - - 1st. Person

Group 2 - 2nd. Person

Group3 fi — 3rd. Person

This figure shows that individuals in Group 1 (over 46 years of age) exhibit the lowest rates of expressed subjects for all grammatical persons, followed by those in Group 2 (35-46 years) and Group 3 (25-35 years). At the same time, the figure shows that the highest rates of expressed subjects are found for the second person in the speech of every group (80%, 94% and 92%, respectively) confirming the diachronic results. The first person follows, with 67% for Group 1 and 79% for Groups 2 and 3. In last place come the third person forms, with the lowest rates: 50%, 65%, and 71%, respectively. The hierarchy, maintained throughout for the behavior of the three groups, reveals a greater difference 6

The sample, part of the NURC-RJ, was recorded in 1992 and is known as Recontact.

Maria Eugènia Lamoglia

22

DUARTE

between Group 1, on the one hand, and Groups 2 and 3, on the other. Notice, moreover, that the graph does not suggest a constant variation: while the rate of expressed third person subjects for Group 1 approaches the rate found in the 1992 7 play, the rates exhibited for Groups 2 and 3 clearly suggest a change in progress. These results were confirmed by the analysis of four hours of radio interviews with people from different regions of the country; this allows the extension of the findings from the speech of Rio de Janeiro to standard spoken BP 8 . It was possible to show that in BP today there does not seem to be a single context in which referential pronominal subjects are obligatorily null; in other words, a once obligatory rule has become optional (Duarte 1998a). This can be seen with the explicit coreferential subjects expressed in [1] and [2], and the [-human] antecedents in [2] and [3]: [1]

De repente ela¡ sabe que ela¡ quando crian9a ficava meio triste por isso. 'It may happen that she knows that she as a child would be sad for that'

[2]

A casa¡ virou um filme quando ela¡ teve de ir abaixo. 'The house became a movie when it had to be demolished'

[3]

Nova Trento¡ é do tamanho da rúa S3o Clemente de Botafogo. Ela¡ é desse tamanho. Ela¡ náo tem paralelas. 'Nova Trento is the size of S3o Clemente Street in Botafogo. It is very small. It has no parallels'

1.2.

Arbitrary subjects

The analysis of arbitrary referential subjects presented in this section is based on the assumption that such subjects should be influenced in a manner similar to that of definite referential ones, i.e, arbitrary pronominal subjects would also be preferably filled. The overall rates obtained in the analysis of the same contemporary sample described above show that expressed pronominal forms (66%) are much more prevalent than null ones (34%). And contrary to the dictates of traditional grammars, the use of the clitic se is the least frequent form, confirming the findings of Kato and Tarallo (1986). Figure 3 shows the distribution of pronominal strategies for indeterm¡nation in BP:

7 This may suggest that texts of popular plays succeed in approximating the vernacular of their times, thus being a good source for diachronic research. Texts from informal letters, for example, exhibit higher rates of null subjects, particularly for the first person (see Paredes Silva 1988). 8 There are some regions of the country, however, where Paradigm 1 at least partially survives. As far as I know, there are no detailed accounts regarding the expression of pronominal subjects.

The Loss of the 'Avoid Pronoun' Principle in Brazilian

Portuguese

23

Fig. 3: Strategies for indétermination in BP (%)

The preferred pronominal form for indétermination today is the use of second person pronoun você ('you'), followed by third person eles ('they') 9 and first person plural a gente ('one'), which together account for 73% of the occurences. Next we see the use of the clitic se + 3rd. person verb form (8%) in variation with zero + 3rd. person (17%) 10 ; the former strategy is restricted to the speech of Group 1 and the latter shows a regular distribution throughout the three age groups. Finally, we have the use of nos ('we') in only 2% of the examples of arbitrary subjects, suggesting a tendency similar to that observed for definite reference subjects (Table 1), i.e, it seems to be disappearing from the pronominal system of BP. Examples [4-9] illustrate these strategies. Notice that, even in embedded environments with coreferential subjects (examples [4-6]), the pronominal forms tend to be expressed rather than null: [4]

9

Você quando você viaja, você passa a ser turista. Entâo voce passa a fazer coisas que voce nunca faria no Brasil. 'You when you travel, you become a tourist. So, you start doing things that you would never do in Brazil'

We are not taking into account possible nuances of meaning in the use of third person as opposed to the other strategies. According to Galves (1987), it is exactly the filling of definite reference subjects that makes the arbitrary interpretation of an empty category possible in such contexts. Notice, however, that it is not as frequent as the full pronominal options.

24

Maria Eugènia Lamoglia

DUARTE

[5]

Quando eles querem eles fazem. Quando eles querem eles acham dinheiro. 'When they want they do (it). When they want, they find the money'

[6]

Hoje em dia, quando a gente levanta as coisas, e que a gente ve tudo o que aconteceu. Mas na 2 ganhou na loto. b.O Pedroi disse que prom ganhou na loto. P. said that (he) won the lottery

(BP) (EP)

In context-neutral situations, i.e, 'out of the blue' contexts, sentence [2]b. can be interpreted as saying that 'Pedro thinks that Paulo won the lottery', if Paulo is salient from the situational context (as when the speaker is looking at a photo of Paulo) or if the speaker is intentionally pointing to Paulo. Sentence [2]a., however, cannot have this interpretation in either of those situations. The difference seems clear: in EP, the grammatical features of the empty category, 7

On the impossibility of having an overt pronoun bound by a quantifier in BP, see Negrao & Muller (1996); NegrSo (1997). This fact is also discussed briefly later in this paper. 8 In fact, Negrao (1990) concludes that impoverished agreement does not lead to a higher percentage of overt subjects, which points in the same direction argued for here, i.e, that the pro-drop phenomenon in BP is not related to agreement.

Marcello MODESTO

150

i.e, 3 rd person singular, are given by the verb inflection and the empty category is able to take any reference that fits such a description (either the higher subject Pedro, or somebody else available from the situational context who is a singular 3 rd person element). In BP, on the other hand, verbal inflection is inert and the empty category cannot refer in the same way that it does in EP; it necessarily gets its interpretation from a preceding subject. This fact makes referential null subjects impossible in matrix clauses in BP 9 : [3]

*pro comeu pizza, (s/he) ate pizza

In order for the empty category in [2]a. to have an antecedent other than the matrix subject Pedro, the sentence has to be in a context such as an answer to a question like 'what about Paulo?'. In these contexts, matrix null subjects are also acceptable; however, the empty category in those cases has a very different nature. As observed by Huang (1984), BP is a language that allows for NP-topic deletion. As an answer to the question above, both [2]a. and [3] become possible with the empty category referring to Paulo, but only because that empty category is the trace of a topicalized phrase that has been deleted. This raises the question whether all null subjects in BP can be analyzed this way. The answer is no. Take [2]a. again, with the interpretation given by the indices above (the 'out of the blue' interpretation). The empty category in that sentence must be pronominal in nature. Movement of a 'zero' topic from that position would lead to a Crossover configuration, since both embedded and matrix subjects are coreferential. It must be concluded, then, that, at least in these cases, the empty subject in BP is pronominal. In view of these facts, I take it as clear that a) the empty category in subject position in BP can be the trace of a moved 'zero' topic but it can also be the category pro\ b) when pro is the subject, it is not identified by agreement; it needs a preceding antecedent and therefore null (referential) matrix subjects are impossible in an 'out of the blue' context. From now on, I will concentrate on the cases where the null subject is pro (so all the examples should be considered as being in an 'out of the blue' context). I will detail the differences between null subjects in BP and EP (which I take to be representative of pro-drop Romance languages in general) on one hand, and between null and overt pronouns in BP, on the other. The prediction is that overt pronominal subjects in BP should pattern with null pronominal subjects in EP since, in both cases, the subject contains or acquires from the agreement inflection grammatical features such as person and number. In other words, overt pronouns come with person

9

Expletives and arbitrary subjects can still be null since they do not need to be identified.

Null Subjects without 'Rich ' Agreement

151

and number specifications from the lexicon; pro, by hypothesis, is not lexically specified for person and number, but acquire them from the rich verbal inflection and therefore can behave like an overt pronoun. Null pronominal subjects in BP, on the other hand, remain without these features since agreement is weak and need a preceding DP to be their antecedent. The fact that pro in BP does not acquire person and number features from agreement, then, explains its unexpected behavior (to which I immediately turn)10. As noted by Negrlo (1997), a pro subject in BP is able to co-refer only with a ccommanding DP, unlike overt pronouns, or empty pronouns in more typical prodrop languages: [4]

a . [ 0 amigo do Pedro2]i disse que/?roi/.2/«3 ganhou na loto.

(BP)

b . [ 0 amigo do Pedro2]i disse que ele 1/2/3 ganhou na loto.

(BP)

c . [ 0 amigo do Pedro2]i disse que pro\an ganhou na loto. the friend of Pedro said that (he) won the lottery

(EP)

As noted by Figueiredo Silva (1994), pro in BP takes only the closest higher subject as its antecedent (where closeness is defined in terms of c-command) (in [5]) and cannot refer to a higher object (in [6])", again unlike overt pronouns and null subjects in other pro-drop languages: 10 Although most of my examples from now on use 3rd person singular and a past tense, the same facts would obtain with all other persons and (indicative) tenses. " I am glossing over some complications caused by subjunctive tenses since they are absent from children's speech and so are a product of schooling. Mary Kato has also called my attention to the fact that reference to an object by a null subject becomes possible with verbs in the indicative mood if there is a modal in the embedded clause. It is possible that, as suggested to me by Hagit Borer (p.c.), the modal brings a hidden subjunctive. Anyhow, it is clear that in the core cases (shown in the text) a null subject can only take a higher subject as its antecedent and this is the syntactic phenomenon I will try to explain, leaving the exceptions for further research. Note that the subject orientation of empty subjects in BP is a problem for Kato (1999) and Cyrino, Duarte & Kato (this volume) who analyze cases of 'controlled' embedded null subjects as a pro which is doubled by PRO. This is problematic because PRO subjects may be controlled by objects; null subjects of finite sentences, however, cannot (usually) have objects as their antecedents. This problem cannot be reduced either to properties of particular verbs or to pragmatic factors. In [i], we see that under verbs like convencer (to convince), a PRO subject has to be controlled by the matrix object. So convencer has to be considered an object control verb. Yet, null subjects of finite sentences embedded under convencer (as seen in [6] above) cannot take objects as their antecedent. In [ii], we see that the same applies to null subjects embedded under verbs like dizer (to tell). However, there is no pragmatic reason why the subject could not refer to Maria in [ii], as its version with an overt pronoun shows:

152

[5]

Marcello

a. O Paulo) disse que o Pedro 2 acredita que p r o l a n ganhou

MODESTO

(BP)

b. O Pauloi disse que o Pedro 2 acredita que elei/2/3 ganhou

[6]

c. O Pauloi disse que o Pedro 2 acredita queproxys ganhou Paulo said that Pedro believes that (he) won

(EP)

a. O Paulo! convenceu o Pedro 2 q u e p r o m n tinha que ir embora.

(BP)

b. O Pauloi convenceu 0 Pedro 2 que elei/2/3 tinha que ir embora. c. O Pauloi convenceu o Pedro 2 q u e p r o m n tinha que ir embora. Paulo convinced Pedro that (he) had to go away From the paradigms in [4] - [6] it can be attested that the prediction that pro in pro-drop languages patterns with overt, but not with null, pronouns in BP is borne out. I have loosely explained this fact by assuming that pro in EP acquires grammatical features such as person and number from (i.e, it is identified by) the verbal agreement while pro in BP does not, which forces it to have an antecedent in its sentence. Overt pronouns obviously have these features inherently, although their reference may not be set, i.e, it may depend on contextual variants. I take it that these grammatical features (person, number, and, in some languages, gender) are a sine qua non condition for an expression to refer. Although the reference of pronouns, unlike nouns, is sometimes contextually dependent, the presence of the features [person] and [number] defines a set of possible antecedents without which contextual reference would be impossible. In this sense, pro in EP becomes a 'normal' pronoun after it acquires person and number features from the verbal inflection. I will describe this difference between overt and null pronouns by assuming that pro, unlike overt pronouns, is always generated without a referential index (which, in this sense, signals the absence of person and number features). Identification of pro in EP (and other pro-drop languages) is then the process of acquiring a referential index from the 'rich' agreement (which is possible by being in a Spec-Head relation with it). Note that pro's reference is not set after it gets its index, just like the reference of overt pronouns is not set even though they inherently possess referential indices. The need for an antecedent in the same sentence for pro in BP is then explained by the fact that pro cannot refer to anything, since the 'weak' inflection of BP is unable to provide it with the necessary grammatical features. A lot of questions can be raised at this point: a) how can pro 'corefer' with its antecedent if it cannot 'refer' at all (if it doesn't have a referential index)?; b) why must its antecedent be a 'sentence-mate' (but not someone salient from [i]

O Pauloi convenceu a Maria2 a PRO'U2 sair. P. convinced M. to go out

[ii]

O Paulo! disse pra Maria2 q u e p r o \ n / ela2 ganhou na loto. P. told M. that (he) / she won the lotery

Null Subjects without 'Rich ' Agreement

153

discourse)?; c) why may only subjects (and, in fact, only the closest subject) qualify as pro's antecedent? In order to answer these questions, I will now turn to cases not noticed in the literature on BP where an empty subject takes something other than a subject as its antecedent. These cases are those in which an object (for instance) is overtly moved to an A'-position. (Note that this fact, i.e. the fact that objects are not usually possible antecedents for pro but become possible antecedents once they are A'-moved, provides strong evidence that the identification of null subjects in BP is not related to agreement). Take the examples below: [7]

a. Quemi que o Pedro 2 convenceu ti que proxmi tinha que ir embora? who (that) Pedro convinced that (he) had to go away b. A Mariai, o Pedro 2 convenceu t, q u e p r o m tinha que ir embora. Maria Pedro convinced that (s/he) had to go away

In [7]a., the wh-word is moved from the object position, a position that usually cannot serve as the antecedent for pro (compare [7] with [6] above). Being moved, not only does the wh-phrase become a possible antecedent for pro but it becomes the only possible antecedent: pro cannot be interpreted as coreferent with the matrix subject. Similarly, in [7]b., an object becomes a possible antecedent for the empty subject through topicalization. In this case, we have an ambiguous sentence. In trying to explain the facts presented so far as to what can serve as the antecedent for null subjects, I see two logical possibilities. On one hand, the relevant notion for being a possible antecedent for pro is that of being a subject, and moved elements somehow acquire some kind of 'subjecthood'. In other words, objects would somehow become subject-like when moved, explaining the contrast between [6] and [7], On the other hand, the relevant notion might be the fact that an argument can only serve as an antecedent for pro if it occupies an A'-position. The sentences in [7] seem to point in that direction, since objects become possible antecedents when topicalized or wh-moved to A'-positions. This line of thought would imply that the subjects (but not the objects) in [6] occupy A'-positions. I will pursue this second option, since it sounds more plausible 12 . Let's start by assuming that the data in [4] through [6], which shows that the null subject can take only the closest c-commanding subject as its antecedent, shows the same basic process seen in [7], i.e, that the null subject must be A'-bound. It must then be 12

The first possibility is not as implausible as it may seem at first. It may be assumed that (at least one of the landing site position(s) of moved arguments is interpreted as a 'subject of predication' position. In this sense, arguments become 'subjects' when moved. As we will see below, my analysis does involve something like this and so the two possibilities are in fact equivalent in every sense. However, I decided it would be clearer to speak of independently needed A'-positions then to try to define what this 'subject of predication' position would be.

154

Marcello MODESTO

concluded that the subject in BP normally occupies an A'-position and that this is what distinguishes subjects from objects with respect to the possibility of serving as an antecedent for pro. The assumption that subjects in BP occupy an A'-position also helps us to answer some of the questions left pending above: the 'antecedent' of pro, now understood as its A'-binder, must be local (a 'sentence-mate') since A'-binding is local. Only subjects, but not (unmoved) objects, can be the 'antecedent' of pro because only subjects occupy an A'-position. In sentences like [5] above, both overt subjects occupy an A'-position, therefore the closest c-commanding subject is the only possible binder (which may be explained as a minimality effect, as discussed below). Pro, in fact, does not corefer with an antecedent. Since it has no referential index, pro cannot corefer with anything and it is un interpretable at LF because it has no reference. The binder of pro rescues the derivation, since it provides an interpretation for pro (the interpretation of a bound variable). The logic is as follows: since pro does not have a referential index, and agreement is too meager to provide it with one, a derivation containing pro in subject position in BP will only be interpreted at LF if pro can be interpreted as a variable. In other words, pro is only possible in this language when bound by some DP in an A'-position. Note that the c-command requirement shown in [4] is also explained. Since pro needs to be A'-bound (and does not refer like a normal pronoun) only a c-commanding DP will be a possible antecedent/binder. Coreference with a non c-commanding DP is impossible, since pro does not refer. Accepting that subjects in BP occupy an A'-position, the question now is why, and what that position is. The first possibility to consider would be that subjects in BP occupy a topic position. I will discard such a possibility in view of the fact that quantifiers such as ninguém 'nobody', which are barred from topic position, obviously make good subjects. Besides, assuming that subjects are topics would buy us more problems than advantages. Consider the ambiguous sentence [7]b. again. If the ambiguity was caused by the fact that there are, in fact, two topics in the sentence, the lack of ambiguity in [8] would be mysterious: [8]

a. A Maria^ o Cadu 2 convenceu elai que pro*m tinha que ir embora. Maria Cadu convinced her that (he) had to go away b. A Maria h O Cadu 2 quase chorou depois que pro,ln foi embora.13 Maria Cadu almost cried after that (he) went away

13

Although presented as grammatical, the fact that the topic in [8]b. is not related to the comment clause makes the sentence very odd. This sentence is acceptable with the interpretation 'as for Maria, Cadu almost cried after he went away'.

Null Subjects without 'Rich ' Agreement

155

In both sentences in [8], the topic a Maria was base generated in the dislocated position. In [8]a. this is signaled by the presence of an overt pronoun in the position related to the topic. In [8]b., base generation is the only option, since the position with which the topic could be related is inside an adjunct island. Both sentences are unambiguous in that only the matrix subject can bind the embedded pro. So far, then, we have seen that subjects in BP are possible 'antecedents' for pro because they occupy an A'-position. However, if a wh-phrase is moved, it becomes the only possible 'antecedent'. If a topic is present, the sentence is ambiguous if and only if the topic could have been moved to that position. The generalization that seems to emerge from those facts is that when an argument is moved to an A'-position, the subject loses its role as binder of pro. In other words, [7]b. would be structurally ambiguous: in the derivation where the topic has been moved, the subject cannot bind pro, just like in [7]a. On the other hand, in the derivation where the topic has been base generated, the subject is the only possible binder for pro, as in [8], There are then two different (and unambiguous) structures associated with [7]b. The importance of the movement operation is also clear from the fact that if the wh-phrase remains in situ (which is always a possibility in BP), the subject becomes the only possible binder again (compare [9] with [7]a.): [9]

O Pedro2 convenceu quem, que pro.,/2 tinha que ir embora? Pedro convinced who that (he) had to go away

We then conclude that subjects in BP occupy an A'-position which is not the topic position. When either a wh-phrase or a topic is moved to the left periphery of the sentence, those phrases will use that very same position as a landing site (but maybe not their final landing site), preventing the subject from moving there (and thus from being the binder/antecedent of an embedded pro subject). Topics not derived by movement will not occupy that position, leaving it free to be occupied by the subject, which then becomes the only possible antecedent for a null subject. The two questions above, however, still remain: what is the position occupied by subjects when no other argument is A'-moved and why does the subject move there? In the next section, I will conclude that such a position is the specifier of the highest INFL projection and that movement of some argument to that position is necessary in order to license that projection.

1.2.

The position of subjects in BP

In the preceding section, I concluded that null pronominal subjects are identified by being in a Spec-Head relation with 'rich' agreement in EP, but by being A'bound in BP. The position that A'-binds null embedded subjects in BP is

Marcello MODESTO

156

occupied by any A'-moved phrase or, in case there is no overt A'-movement, by the matrix subject. According to Diesing (1990) and Koopman & Sportiche (1991), among others, the specifier of the highest INFL projection is a possible A'-position in some languages. More specifically, according to the latter authors, this specifier position is an A-position when case is assigned to (or checked in) it but A ' when it is not a case position. Let's assume, following Belletti (1990), that the highest INFL projection is AgrP. If [Spec AgrP] is the A'-position occupied by subjects in sentences like [9], for instance, nominative case must have been checked in [Spec TP], Is this at all plausible? My answer would be yes. Consider EP: Costa (to appear) has argued that pre-verbal subjects in EP are used to convey old information, whereas post-verbal subjects present new information. Assuming that the verb moves to Agr° in EP (uncontroversially); and that nominative case is assigned to [Spec TP], movement of subjects to a pre-verbal position, [Spec AgrP], an A'-position, would force a topic interpretation for subjects, explaining the old information interpretation. 14 As argued in Modesto (in progress), BP has lost long verb movement, i.e, verb movement to Agr°, which has made post-verbal subjects impossible (in transitive sentences). However, it might be true that case checking remains exactly as it is in EP: nominative case is checked in [Spec TP] and [Spec AgrP] is a possible A'position. The structure of a sentence like [2]a. would then be as in [10] and the structure of [7]a. as in [11] (I will ignore a possible AgrOP projection) 15 : [2] a. O Pedro! disse q u e p r o x n ganhou na loto. Pedro said that (he) won the lottery [10]

AgrP

t) tv que pro ganhou na loto 14

Alexiadou & Anagnostopoulou (1998) argue convincingly that pre-verbal subjects

occupy an A'-position in Romance in general. 15 In [10] and [11], I assume that the verb moves overtly to the head of TP. It is possible, however, that the verb stays even lower in the structure. This will not bear on the point being made here.

157

Null Subjects without 'Rich ' Agreement

[7] a. Quemi que o Pedro 2 convenceu t, que pro\mi tinha que ir embora? who (that) Pedro convinced that (he) had to go away [11]

CP

Agr

TP

ti t v t 2 que pro tinha que ... In both structures, the subject o Pedro checks nominative case in [Spec TP]. In [10], it is moved to [Spec AgrP], which counts as an A'-position and so the subject A'-binds the embedded pro, being interpreted as its antecedent. In [11], on the other hand, the subject remains in [Spec TP] and the wh-phrase is moved to [Spec AgrP] on its way to [Spec CP]. In this case, the subject does not A ' bind pro and it cannot be interpreted as its antecedent. Recall that a basegenerated topic (in [8]), and a wh in situ (in [9]) cannot bind pro. This is because of the fact that, since movement is absent in both cases, the matrix subject has to move to [Spec AgrP] and thus it will be the only possible binder for an embedded pro16 Movement to [Spec AgrP] is then obligatory: if no (wh or topicalized) constituent is moved to that position, the subject has to climb there. Of course, this is not caused by the need of having an antecedent for pro, since a derivation with an overt pronoun is always possible. On the contrary, I am arguing that the derivation with pro is possible, exactly because there will always be an A ' element to bind it. Assuming Rohrbacher's (1994) hypothesis that agreement morphemes may have their own lexical entry, Speas (1994) develops a theory in which pro-drop is

16

I am assuming that unmoved topics are base-generated in a position higher than [Spec AgrP],

158

Marcello

MODESTO

related to verb movement. In that theory, a functional head X can only project a phrase XP if XP has content, according to [12]: [12] A node X has content if and only if X dominates a distinct phonological matrix or a distinct semantic matrix. In her words, "if XP [...] dominates no phonological material except that which is in the complement YP, then XP dominates no distinct phonological matrix. Similarly, if XP dominates no semantic material except that which is in the complement YP, then XP dominates no distinct semantic matrix." In this way, either the head X or the specifier of XP must dominate some material with content in order for the projection XP to be licensed. Radically empty projections, which function solely as landing sites for movement, are disallowed. In the case of AgrP, two instances must be considered. In weak agreement languages, agreement morphemes do not have a lexical entry, so the verb is base generated already inflected for agreement. Verb movement by itself is not enough to license the projection of AgrP (which, as Speas suggests, may be because the Agr projection cannot be licensed by a head of a category other than Agr). So the only way to license the AgrP projection in these languages is to fill its specifier position (with an overt expletive, or by moving a DP into that position overtly). Those languages will then always have an overt subject. On the other hand, in rich agreement languages, Agr morphemes have their own lexical entry and therefore are base generated as the head of the Agr projection. The head of AgrP has content so there is no necessity for the specifier of AgrP to be filled. The null subject (pro), which Speas assumes to be base generated in a VP internal position, remains in that position and the specifier of AgrP is truly empty (until LF, when pro moves in order to be identified by Spec-Head agreement). The agreement morpheme, however, is dependent on the verb since it cannot be spelled out by itself. This theory, then, implies that, in all null subject languages, the verb must be overtly moved to AgrP. In this way, null subjects are dependent on verb movement, although the reverse is not true (differing here from Rohrbacher 1994). Although it is not my intent here to discuss every detail (and problems) of this theory (referring the reader to Modesto (in progress)), I will assume portions of Speas's theory. Specifically, I will assume that the Agr projection must be licensed by a contentful category. According to Rohrbacher, the property of having inflectional morphemes generated under the Agr head is what makes agreement rich. In that case, BP will not have lexicalized morphemes and so the verb will be base generated already inflected in the VP. The need to move the subject to [Spec AgrP] is thus explained by the need to license that projection. If another constituent is moved, it can be used to license AgrP and the subject will

Null Subjects without

'Rich ' Agreement

159

remain in [Spec TP], It is then the need to license AgrP that will provide an A'element to license pro since some DP will always have to be moved to [Spec AgrP] 17 . Summarizing, I have argued that pro is a null pronominal element which lacks a referential index. In rich agreement languages, those in which inflectional agreement morphemes are base generated under Agr°,pro will be able to acquire the features it lacks by being in a Spec-Head relation with the Agr head at LF. It will then refer in the same way an overt pronoun would. In weak agreement languages, a pro subject may license the Agr projection. However, it will be uninterpretable at LF because, even if the verb moves to the Agr head, it is unable to furnish pro with person and number features (actualized as a referential index). In some of these weak agreement languages, as in BP, a derivation with a pro subject may be saved if the null pronominal is interpreted as a variable bound by a DP in an A'-position. The fact that nominative case is checked in [Spec TP] in BP and that the Agr projection must be licensed will conspire in a way such that there will always be an A'-binder for pro subjects in embedded clauses. In other words, if no DP has been moved to an A'-position, the subject has to move to [Spec AgrP] in order to license that projection. Since [Spec AgrP] is a possible A'-position, these subjects will always be able to A'-bind an embedded pro. The theory outlined so far (since it concerns only referential null arguments) indicates that a licensing requirement for pro (as in Rizzi's 1986 theory) is unnecessary. A null pronominal is possible in subject position (and maybe in

17

Note that the embedded pro in sentences like [2]a. will also be moved to [Spec AgrP] to license that projection. Differing here from Speas, I assume that pro does license the Agr projection since it contains at least one distinct feature: its categorial feature. A question arises as to how pro can be interpreted as a variable at LF if it occupies an A'-position. It might be the case that [Spec AgrP] is a dual position, as proposed by Diesing (1990): it is a possible A'-position (since it is not a theta nor a case position) but it may be taken to be an A-position. Alternatively, it could be maintained that [Spec AgrP] is always an A ' position if the pro in that position is taken to be parallel to traces of wh-phrases moved by successive cyclic movement. A matrix subject would bind pro in [Spec AgrP], an A ' position, and pro in turn would bind its trace in [Spec TP] which would then be interpreted as a variable. This is parallel to a derivation where a wh-phrase in the matrix CP antecedent governs an intermediate trace as in fVho, does Mary think tl that John saw ti? The sentence in [i] seems to indicate that this is the right approach since, in that sentence, an embedded null subject binds another embedded pro, showing that the intermediate empty category must itself occupy an A'-position: [i] O Pedro, disse que pro, achava que pro, ia perder o jogo. Pedro said that (he) thought that (he) was going to loose the game

160

Marcello

MODESTO

object position as well, as we will see below) if it can be identified 18 . The way in which pro may be identified seems to vary from language to language and the parameter involved in that choice will be discussed in section 3.

2.

Confirmation of the analysis

Confirmation of the analysis presented in section 1 can be provided in two ways: a) by presenting further data on BP that would not be explained otherwise; and b) by presenting cross linguistic evidence that some languages resort to A'-binding in order to identify empty categories. In this section, I show that the interpretation of null subjects and the behavior of BP subjects in islands is predicted by the analysis presented, supporting it. Firstly, it will be shown that null subjects in BP are obligatorily interpreted as variables, which follows naturally from the analysis presented here. Secondly, I will show that null subjects are impossible in relative clauses but not in other complex NP or adjunct islands, which also follows naturally from the fact that pro in BP needs to be A'bound. I will also show that the same process of rescuing a derivation by A'binding can be found in Chinese.

2.1.

The interpretation of null subjects

In the last section, I concluded that subjects are possible antecedents for pro because they normally occupy A'-positions. This fact enables a derivation containing a pro subject to be interpreted at LF because, being A'-bound by a higher subject, pro is interpreted as a variable. It thus follows that a null subject in BP should always be interpreted as a variable. This prediction is borne out. In actuality, the core of the analysis presented here was designed exactly to explain that very same fact. Negrao (1997) was the first to note that there is a specialization of forms deriving the use of null versus overt pronouns in BP: the null variant is used whenever a bound variable reading is intended. Overt pronouns, on the other hand, can never be interpreted as variables. Negrao observes that null subjects in BP give rise only to sloppy identity readings in contexts of VP ellipsis19:

18

Maybe the licensing requirement needs to be stated only for expletive pro because those empty pronominals do not need to be identified. However, it is unclear at this point what property of INFL could license null expletives, since both weak and rich agreement languages (BP and EP, for instance) license them. 19 Examples [12] and [13] are taken from Negrào (1997).

Null Subjects without 'Rich ' Agreement

[13]

161

a. O Pedroi acha que pro\ é inteligente e o Paulo também. (BP - sloppy only) b. O Pedro, acha que pro\ é inteligente e o Paulo também. (EP strict/sloppy) Pedro thinks that (he) is intelligent and Paulo does too

As she notes, assuming an analysis of VP ellipsis where the elided VP is reconstructed as a copy of the first VP and where variables are bound by a lambda operator, the absence of the strict identity interpretation in [13]a. is expected, according to her analysis, because the null subject is always interpreted as a bound variable. A further argument given by NegrSo as evidence that null pronouns are always interpreted as bound variables (and overt pronouns are never interpreted in that way) is the fact that overt and null pronouns give rise to different interpretations and, consequently, truth values, in sentences where their 'antecedent' is the phrase só DP (only DP): [14]

a. Só Malufi acha q u e p r o \ vai ganhar as elei?8es. b. Só Maluf) acha que ele, vai ganhar as ele¡95es. only Maluf thinks that (he) will win the elections

In [14]a., we have the interpretation that Maluf is the only candidate who thinks of himself as a winner so the sentence is true in a situation where the other candidates in the election think that they will lose. Sentence [14]b., on the other hand, says that Maluf is the only person who thinks that Maluf will win, so it can be true if Covas, another candidate, thinks that he will be the winner. NegrSo notes that the different truth conditions between the two sentences come from the fact that "in the appropriate level of representation the empty category [in [14]a. MM] is translated as a bound variable and the sentence can be paraphrased as: the only x such that x thinks that x is going to win the election is M a l u f ' . Sentence [14]b., however, would have to be translated as "the only x such that x thinks that Maluf will win is M a l u f ' because overt pronouns are never interpreted as bound variables. The only obstacle to accepting Negrao's claim that pronominals forms in BP have a specialized use, i.e, that null pronominals can only appear in bound contexts and overt pronoun can never be bound, is that (for me and other speakers I have consulted) the sloppy interpretation is obtainable when an overt pronoun is used in a VP ellipsis construction. Compare [13]a. with [15]:

162

[15]

Marcello MODESTO

O Pedroi acha que elei é inteligente e o Paulo também. strict/sloppy) Pedro thinks that he is intelligent and Paulo does too

(BP -

Assuming Negráo's analysis of [13]a. where variables are bound by a lambda operator, it is predicted that [15] should not have a sloppy interpretation since, by hypothesis, overt pronouns cannot be bound. In fact, the author rules out a sloppy interpretation for [15]. The fact that most speakers find a sloppy identity interpretation possible for [15] turns out not to be a problem if the VP ellipsis theory of Fiengo and May (1994) is assumed. Without going into many details, it is sufficient to note that, in that theory, when the antecedent VP is reconstructed into the elided VP (at LF), referential indices need not be preserved (i.e, they can be changed). In this way, the sloppy reading of [15] does not present any counter argument for NegrSo's claim that overt pronouns cannot be bound because the sloppy reading can be construed without resorting to taking the pronoun in the antecedent VP to be a variable. However, assuming Fiengo and May's theory, to preserve the fact that only the sloppy interpretation is possible in [13]a., a theory like the one presented here is necessary. When the antecedent VP is reconstructed, changing or preserving the index of the null subject pronoun is indifferent since pro does not have any index. The reconstructed pro inside the elided VP at LF will conform to the same restriction as the empty category in the antecedent VP, i.e, it will have to be A'-bound by the closest subject, therefore, only the sloppy identity interpretation will be possible. Consider the LF structure of [13]a. with the reconstructed VP below: [13]

a'. O Pedroi acha que pro é inteligente e o Paulo também (acha que pro é inteligente)

In the first conjunct, the only way to interpret the empty category in subject position is to take it as a variable bound by the subject o Pedro. Similarly, in the second conjunct, pro must be interpreted as bound by the subject o Paulo. Binding the second instance of pro by the highest subject is impossible due to the lack of c-command; coreference, as always, is not an option with null pronouns (since those pronouns lack a referential index). From the data in [13] and [14], it can then be concluded that Negr3o was right in claiming that overt and null pronouns are specialized forms: the latter can only be used in contexts where it can be interpreted as a bound variable; the former can never be bound. The first generalization is explained by my analysis. Null pronouns must be interpreted as bound variables in BP in every context because they are not identified by agreement and therefore remain without a referential index at LF. The second generalization still lacks a clear explanation. However, the option of having a bound overt pronoun is restricted in many languages (cf.

Null Subjects without 'Rich ' Agreement

163

Montalbetti 1984; Picallo 1994) so the fact observed by Negrito may not be related to any characteristic specific to BP.

2.2.

Minimality effects

In [5]a. above, it can be seen that only the closest higher subject can be the antecedent for a null subject in BP. Since the 'antecedent' is in fact an A'binder, in the analysis presented here, the impossibility of having the subject Paulo binding pro over the possible binder Pedro, in [5]a., repeated below, could be explained as an interception effect of the kind referred to as Minimality (cf. Rizzi 1990; Chomsky 1995; and Aoun & Benmamoun 1998). [5]

a. O Paulo i disse que o Pedro 2 acredita que pro^yy*! ganhou. Paulo said that Pedro believes that (he) won

Such Minimality effects can be explained by the fact that A'-binding of pro is a strategy to save the derivation: since pro does not have a referential index, it is uninterpretable at LF unless it can be interpreted as a variable. Because pro must find a binder in order to be interpreted, it cannot choose any binder; it must take the closest c-commanding potential binder. We then predict that any ccommanding potential A'-binder will intervene between pro and a higher subject. This prediction is borne out: [16]

a.*A Maria! achou um carro [Op quepro t tem grana pra comprar t].20 Maria found a car that (she) has (enough) money to buy b.*A Mariai n5o gostou do livro [Op quepro\ revisou t], Maria did not like the book that (she) revised

In [16], the relative operator is an A'-specifier that c-commands the null subject. Therefore, it qualifies as the closest possible binder for pro and it intervenes between the pronominal and the subject in the higher clause. The interpretation where pro is bound by the higher subject is then impossible. However, the interpretation in which the relative operator binds pro is also ill-formed due to principle C of the binding theory: pro c-commands (and so it would bind) the trace of the relative operator in object position. An empty category in subject position inside a relative clause is then only possible if the relative operator originates in that position (i.e, a pro subject is never allowed in relative clauses). Note that, besides the operator in relative clauses, nothing leads us to expect that empty subjects will be barred in strong islands, since this empty category is not generated by movement and nothing prevents pro from being bound by the 20

Sentence [16]a. is from Figueiredo Silva (1994) who uses the anaphoric Agr theory of Borer (1989) to explain some of the properties of null subjects in embedded contexts in BP. For a discussion of that analysis (and its problems) see Modesto (in progress).

Marcello

164

MODESTO

higher subject when in an adjunct clause or a complex NP (i.e, no possible A'binder intervenes). Accordingly, the sentences in [17] are grammatical: [17]

a.A Maria se esquece do fato de que pro vai ganhar menos no novo emprego. Maria is not considering the fact that (she) will earn less in the new job b.O presidente negou os rumores de que pro recebeu dinheiro de emprescirios. The president denied the rumors that (he) got money from businessmen c.A Maria quase chorou depois que pro viu o estrago. Maria almost cried after that (she) saw the damage d.O Pedro estava bebado quando pro bateu no Daniel. Pedro was drunk when (he) hit Daniel

2.3.

Null arguments in Chinese

The analysis presented here for null subjects in BP resembles very closely Huang's (1982, 1984) analysis of null objects in Chinese. Huang argues that null pronominals are subject to a rule of control (in [18]) which, coupled with principle B of the binding theory, would give the following result: pro is banned from object position in every language lacking object agreement; pro is possible as a subject when there is a rich agreement element, or when there is no agreement at all: [18]

Generalized Control Rule (GCR) Coindex an empty pronominal with the closest nominal element (NP or Agr).

For Huang, 'closest' is defined in terms of c-command but also by the notion of clause boundaries, so 'minimal distance' will not distinguish between a ccommanding subject and a c-commanding object within the same clause. This is because, in Chinese, a null subject in a tensed clause, just like PRO (the subject of infinitival clauses), is able to take either a higher subject or object as its antecedent, the choice between them being determined by pragmatic factors. Although a pro object is impossible, according to Huang, in languages like Chinese (since there is no object agreement), empty objects are still possible due to the fact that the language is 'topic-oriented', i.e, it allows for topic NPdeletion. Null objects are then possible when they are variables bound by a (overt or zero) topic. To better understand Huang's theory, consider the sentences in [19]:

Null Subjects without 'Rich ' Agreement

[19]

165

a. ec came. b. John saw ec. c. ec saw ec. d. John said that ec saw Bill. e. John said that Bill saw ec.

In a language like English, all of the sentences in [19] will be ungrammatical. Since English does not allow for topic deletion, all empty categories in [19] should be null pronouns. In [19]a., according to the GCR, the subject must be coindexed with Agr, the closest nominal element. But Agr in English is too meager to determine the content of a subject pro and the sentence is ill-formed, for it violates the principle of recoverability. In [19]b., by the GCR, a pro in object position should be coindexed with a nominal element in its own clause: either Agr or the subject John. In the absence of object agreement, the empty category must be coindexed with the subject John. However, coindexing of pro and the subject in [19]b. would lead to a principle B violation (the pronoun would be bound in its governing category) and so the sentence is ungrammatical. The ungrammatically of [19]c. is then obvious. The ill formedness of [19]d. is explained in much the same way as [19]a.: the embedded Agr is unable to identify pro, and coindexing pro with any other (more distant) nominal element would violate the GCR. In [19]e., the empty category cannot be identified by any element in the embedded clause, as in [ 19]b. It also cannot be identified by any of the matrix nominal elements by force of the GCR, as in [19]d., hence it is also ill-formed. In rich agreement languages, such as Italian, the empty categories in [19] are, once again, pronominal, since these languages are not topic-oriented. In [19]a. and [19]d., both well-formed sentences, the subject is coindexed with Agr and Agr is able to identify the empty category. (Note that coindexation with Agr does not violate principle B since Agr is not an argument). In [19]b., which is ungrammatical, the empty category cannot be coindexed with John (or with Agr, which is itself coindexed with John) because of the binding theory. Sentences [19]c. and [19]e. also will not be able to satisfy the GCR without violating the binding theory and so are both ill-formed. Next, consider Chinese. Since it is a topic-oriented language, the empty categories in [19] can be null pronouns but can also be variables bound by a zero-topic. In [19]a., the empty subject cannot be pro since there is nothing to identify it, but it can be a variable bound by a null topic (as discussed above for BP). In [19]b., similarly, it cannot be pro because of the interaction between the GCR and the binding theory but it can be a variable. The same applies for [19]e. In [19]c., both empty categories can only be variables (since the language allows for multiple topics). In [19]d., the empty category can be either a null

166

Marcello MODESTO

pronominal or a variable. If it is pro, it is bound by the matrix subject John, which is the closest nominal element, since Huang argues that there is no Agr in Chinese. It can also be a variable, since nothing prevents this. All the sentences in [19] are, then, well-formed in Chinese. From the discussion above, it is clear that the data in Chinese is similar to the BP data presented here. On one hand, null pronominal subjects in Chinese are also only possible in embedded contexts. Matrix null subjects are impossible in 'out of the blue' contexts. On the other hand, null objects in Chinese are always A'bound (which is what we are arguing for regarding null subjects in BP). It is then plausible to raise the question whether Huang's theory, or some variant of it, could explain the data in BP. As for null subjects, it is easy to see that the GCR is not the principle underlying the behavior of that language. BP is a weak agreement language, so it should pattern with English. Even if the presence of Agr could be circumvented, the impossibility of taking an object as the antecedent of an embedded empty subject differs from the facts in Chinese and it is unexpected according to Huang's theory. Moreover, the fact that objects become possible 'antecedents' when A'-moved would be unexplained. Consider the pair below (where [20]a. = [7]a.): [20]

a.Quemi que o Pedro 2 convenceu ti q u e p r o { / r , 2 tinha que ir embora? who (that) Pedro convinced that (he) had to go away b.O Pedro 2 convenceu a Mariai que pro*u2 tinha que ir embora. Pedro convinced Maria that (he) had to go away

According to the GCR, pro in [20]a., b. should be controlled by the embedded Agr, which is too meager to identify it, so the sentences should be ill-formed. Suppose now that the presence of agreement in BP can be circumvented; [20]b. is expected to be ambiguous and also [20]a., if it is assumed that the variable in object position can control pro (if not, only Pedro would be a possible antecedent since there is a clause boundary between the subject and the whphrase). The fact that the sentences in [20] are not ambiguous and, further, that the only interpretation in [20]a. is that where pro 'corefers' with the wh-phrase indicates that the GCR is not at work in BP. Therefore, the GCR must not be a universal principle. However, since it is theoretically awkward that such a principle would hold only in Chinese (or in some languages but not others), we have an indication that Control is not responsible for the interpretation of null pronominals. Instead, as I have been arguing, I will take the right notion to be Binding. We have seen already that pro subjects in BP must be A'-bound. PRO subjects, on the other hand, can take objects as their antecedents, which indicates that A'binding does not play any role in this case. If we were to subsume the Chinese data under the analysis presented here, a similar problem would arise: the empty

Null Subjects without 'Rich ' Agreement

167

category in embedded subject position in Chinese (either PRO or pro) is able to take either a matrix subject or object as its antecedent. If the right generalization is that null pronominals in general have to be bound, it could be argued that pro in Chinese, unlike in BP, must be A-bound. Moreover, what Huang claims for null objects in Chinese is very close to what I argue here for null subjects in BP, i.e, that those empty categories must always be A'-bound. Incorporating the Chinese data under the analysis presented here seems, then, promising. In what follows, I will account for null subjects in infinitival clauses in BP and all null subjects in Chinese using the theory of Borer (1989) and extend my analysis to null objects in Chinese.

2.3.1. Null subjects in Chinese As seen above, null subjects in Chinese pattern with null subjects of infinitival clauses in BP in that they can take either subjects or objects as their antecedent (compare [21] with [6] above): [21 ]

a.O Pedro convenceu a Maria¡ a [PRO¡ vir á festa], Pedro convinced Maria [PRO to come to the party] b.O Pedro¡ prometeu pra Maria [PRO¡ vir á festa], Pedro promised to Maria [PRO to come to the party]

The choice of the antecedent pro in Chinese, as argued by Huang, is determined by pragmatic factors 21 . Similar facts exist in Hebrew, where a pro subject can take either a c-commanding subject or object as the 'antecedent' of the empty category: [22]

a.Talilai ?amra le-Itamar 2 she proi/2 yavo/tavo. Talila said to-Itamar that (s/he) will come b.Talilai ?amra le-Itamar 2 she hem 1+2 /3/pro*i +2 /. 3 yavo?u. Talila said to-Itamar that *(they) will come

Borer (1989) uses the Hebrew data as empirical confirmation for a theory that subsumes controlled PRO to pro. She argues that the head Agr in infinitival clauses is [+anaphoric] and so must be bound by a referential DP in order to receive a person feature, giving rise to control structures. In Hebrew, 3rd person agreement in past and future tenses is also anaphoric. In matrix and embedded contexts, an overt pronoun in subject position can bind the anaphoric Agr since it 21

Huang (1984, 1989) does not give any example in which a null subject takes an object as its antecedent, however, Huang (1989) states in a footnote that 'the reference of pro in Chinese is not subject-oriented'.

168

Marcello

MODESTO

stands in a Spec-Head relation and is coindexed with the Agr head. On the other hand, if the subject is null (pro), after binding takes place, the anaphoric Agr is still too impoverished to identify the subject, that very same pro, since no person feature was transferred.22 Null 3rd person subjects in matrix clauses then are expected to be ungrammatical. However, in embedded contexts, the anaphoric Agr head can move to the head of COMP, being bound by a DP in the matrix clause. After binding takes place, Agr possesses a person feature, which is transmitted to the embedded pro subject by transitivity. Borer underlines the correlation predicted by her theory between control and an empty COMP. Therefore, one crucial assumption is that the complementizer she in Hebrew cliticizes to INFL, leaving the head C empty thus allowing the raising of Agr. The fact that an embedded null subject cannot take split antecedents, as shown in [22]b., provides a strong argument in favor of the anaphoric character of Agr. Borer proposes that Agr in non-finite clauses is universally [+ anaphoric], Hebrew 3rd person agreement would also be anaphoric in past and future tenses. She also notes that the Chinese data can be nicely incorporated under her theory if it is assumed that Chinese also contains a head Agr and that Agr is [+anaphoric] both in finite and infinitival clauses in that language. Since Agr is A-bound, the fact that PRO in BP and pro in Chinese and Hebrew may take subjects or objects as their 'antecedent' follows from this.

2.3.2. Null objects in Chinese According to Huang, as seen above, null pronominal objects are banned in all languages lacking object agreement. All the empty categories in object position in Chinese are then analyzed as variables. It is clear, however, that null objects in Chinese are not always created by movement operations since they are possible inside relative clauses and other islands for movement (see Xu 1986; Huang 1987). Huang's point is solely that, at LF, all empty categories in object position must be variables, regardless of their derivational history, so the conclusion that null objects are not pronominal depends crucially on the functional definition of empty categories of Chomsky (1982) assumed by Huang. As he himself notes (1987: 331), "it has been argued that the functional definition of ECs proposed by Chomsky may not be desirable (see Epstein 1983, 1984; Brody 1984; Lasnik 1985; and Saito 1985, among others) and that an EC may not change its identity in the course of a derivation." If this is correct, then variables are always created by movement, and A'-bound empty resumptive pronouns are not variables. In fact, as I argue here, a functional definition of empty categories is not necessary. Null subjects in BP and null objects in Chinese may be treated as A'-binding identified null pronominals. Obviously, 22

Translating this into my analysis, pro cannot bind Agr since it does not have a referential index so Agr has to be bound by some other DP.

Null Subjects

without

'Rich ' Agreement

169

some empty categories may be variables left by movement of a zero topic, both in Chinese and in BP and I am referring here to those cases in which a topicalization analysis is not plausible (contexts that are 'out of the blue' or in which the antecedent-gap relation does not obey Subjacency). That the two phenomena (i.e, null subjects in BP and null objects in Chinese) should be collapsed into one analysis is clear from the fact that those empty categories always need to be A'-bound. Consider for instance [23] from Huang 1984 (and compare it with [16] above): [23]

Li Xiaojie hai zhao-bu-dao [yige [ ec xinzhong xihuan ec de] nanren]. Li miss still can't-find one in-heart like DE man 'Miss Li sitll cannot find a man who she loves in her heart' not: 'Miss Li still cannot find a man who loves her in his heart'

In [16], I have shown that a null subject occurring inside a relative clause in BP must necessarily be related to the relative operator. It cannot be pro because the empty operator would intervene between pro and its binder, rendering the sentence ungrammatical. The same state of affairs is found in Chinese. [23] shows that a null object is only possible in a relative clause if the object empty category is related to the head of the relative. Null subjects in Chinese, however, do not conform to this restriction since their identification is done by A-binding (of the Agr head), as discussed above. Chinese will then be a language where only the identification of subjects is agreement-related (assuming the anaphoric Agr analysis discussed above). The identification of null objects is done by A'binding, in the same way as argued for BP subjects, which provides support for the claim made here that pro arguments may be identified by A'-binding in some languages 23 .

3.

Parametrization

In the preceding sections, I have argued that, as far as null subjects go, there are two types of languages: a) those in which identification of a pro subject is only possible through Spec-Head relation with Agr; and b) those in which pro is given an interpretation by being A'-bound by another DP. The first group is subdivided with respect to the ability of Agr to identify a pro subject. Rich agreement languages, those in which pro is identified, can be said to contain [+pronominal] agreement. In weak agreement languages, then, Agr will be

23

I will not discuss how identification of null objects in BP is obtained, referring the reader to Cyrino (1996).

Marcello MODESTO

170

[-pronominal]. Those [-pronominal] Agr languages may still identify empty subjects if Agr is [+anaphoric] 24 : [24]

y

Agr is [+pronominal]

Italian

identification by agreement ^ Agr is [-pronominal]

identification by A'-binding

Agr is [-anaphoric]

English

Agr is [+anaphoric]

Chinese

/

BP

Agreement related identification seems to be the unmarked parametric choice. Weak agreement languages such as Chinese and (future and past 3 rd person inflection in) Hebrew use the anaphoric Agr strategy (available in UG as the 'normal' agreement in non-finite clauses) as a way to identify null subjects through agreement. This seems to indicate that Identification through A'binding implies weak and anaphoric agreement (or no agreement, as in Chinese null objects). It is suitable to ask, then, why BP did not resort to the anaphoric Agr strategy and why English does not use any alternative strategy and does not allow null subjects in any contexts. As mentioned above, Borer stresses the correlation between anaphoric Agr and an empty COMP (clear from the facts in Saramaccan discussed by the author). The lack of an overt complementizer in Chinese and the clitic character of the complementizer she in Hebrew (according to Borer) allow the anaphoric Agr to move into COMP and get bound by a matrix argument. In BP and English, however, the presence of the complementizer in embedded finite sentences prevents an anaphoric Agr from moving and getting its interpretation. As for why English does not resort to A'binding, the answer is tentative. I believe that A'-binding identification is only possible in discourse-oriented languages, i.e, those exhibiting a cluster of properties which include topic-NP deletion. The fact that BP (for subjects) and Chinese (for objects) are the two languages in which null pronominals are identified by A'-binding seems to indicate that such a relation (between A'binding identification and discourse-orientation) does in fact hold. However, it is not yet very clear why this should be so. In BP, at least, it might be the case that null subjects started being identified by A'-binding as an analogy to other empty categories bound by zero topics. Consider this hypothetical scenario. At a certain point in time, probably around the 1930's and 1940's (cf. Duarte 1995), 24

It can be supposed that a [+pronominal, +anaphoric] AGR would violate some principle of UG and so it is absent from the inventory of languages.

Null Subjects without 'Rich ' Agreement

171

2 nd person inflection had already been lost in BP. However, inflectional morphemes were still part of the lexicon, as they are in EP to this day, and so null subjects are still plentiful. The young generation acquiring the language at that time would then be faced with contradictory input. On one hand, the child acquires a weak verbal paradigm (as defined in Rohrbacher 1994) with unique inflections only for the 1 st person. Consequently, the child interprets that in this language, inflectional morphemes are not lexicalized and that the verb does not move overtly to the Agr projection (a parametric change). Importantly, the child also automatically interprets that agreement is not capable of identifying null pronominals in subject position (since agreement is weak) and starts to fill that position in most contexts. On the other hand, children still encounter a great number of empty subjects in the input (since their parents' grammar is different than the one they have acquired) and are forced to accommodate that input. One way of accommodating empty subjects is to interpret them as the result of topicalization and topic NP deletion. Assuming Raposo's (1986) analysis for null objects in EP, we see that topic NP deletion is possible in EP, and thus, it is very likely to have been possible in the BP of the beginning of the century. The strategy of considering subject gaps as topic deletion, however, is not possible when an empty subject is embedded and is coreferent with a higher argument, since topicalization would give rise to a Crossover configuration. My claim is that, by analogy with the other instances of empty subjects which are treated as variables left by topicalization, embedded subjects which are coreferent with a matrix argument also start being interpreted as variables. These variables, however, are not created by movement. Since Universal Grammar gives the possibility of interpreting pronouns as bound variables, the child uses that alternative. Since pre-verbal subjects occupy an A'-position (in EP and in the BP of that time), the accommodation of empty embedded subjects coreferent with a higher subject is straightforward. Coreference with objects, however, ceases to be possible, since objects do not normally occupy A'-positions. There is another parametric choice that is crucial for the analysis presented here: the possibility of checking nominative case in [Spec TP], I have assumed that both European and BP allow nominative case checking in [Spec TP], leaving [Spec AgrsP] as a (possible) A'-position while other languages would necessarily check nominative case in the highest INFL projection (AgrP). The difference between these two group of languages could be viewed as different settings of a parameter involving the TP projection. A version of such a parameter is defended by Bobaljik & Jonas (1996); Jonas (1996); and ThrAinsson (1996). B&J interpret that the relevant parametrized choice is whether [Spec TP] is licensed as a syntactic position or not. In the affirmative case, [Spec TP] would be the nominative checking position. They propose that: "In a given language, if tense morphology blocks agreement morphology, then that language does not license [Spec TP]". English would then be the case of a language that does not license [Spec TP] since past tense morphology never cooccurs with

Marcello

172

MODESTO

agreement morphology. Jonas affirms that [Spec TP] is only available as a case checking position in those languages where the verb raises overtly to T. Finally, Thrdinsson argues that only languages in which tense and agreement morphology are 'independent' will have separate Agr and Tense functional projections. 'Independent' morphology is understood in the sense of B&J: if a verb form can bear both agreement marker and a tense marker simultaneously, then agreement and tense morphology are independent. Thriinsson understands that in languages like English, where agreement and tense morphology are dependent, AgrP and TP are fused into one INFL projection (differing from B&J), deriving the unavailability of two specifier positions in English opposed to the availability of two Spec positions observed in Icelandic. All three theories lead one to expect that EP would check nominative case in [Spec TP], As for BP, Jonas' theory would lead to the same expectation if it is assumed that the verb in BP also reaches the head T°, while the other would lead to that expectation as straightforwardly as in EP. We can then conclude that the assumption that BP and EP check nominative case in [Spec TP] is well-founded.

4.

Conclusion

I have presented evidence that verbal agreement in BP is unable to identify an empty category in subject position, however, empty pronominals are still allowed as subjects in this language if an alternative way of identifying them is provided. I have argued that such an alternative way is to interpret the empty pronominal as a variable at LF, bound by the closest possible A'-binder. A wide range of facts is explained by this analysis, such as the impossibility of having objects as antecedents for empty subjects; minimality effects by intervening subjects; the impossibility of empty subjects in relative clauses etc. The analysis is also applicable to null objects in Chinese and challenges one of the most accepted views on null arguments: that there is a correlation between rich agreement and empty arguments in every language. Although such a correlation does exist in a number of languages, I have argued that it is parametrized and therefore not universal.

References ALEXIADOU, Artemis & Elena A N A G N O S T O P O U L O U (1998) "Parametrizing AGR: Word Order, V-Movement and EPP-Checking". Natural Language and Linguistics Theory: 491-539. A O U N , Joseph & Elabbas B E N M A M O U N (1998) "Minimality, Reconstruction, and PF-Movement". Linguistic Inquiry 29: 569-97. BELLETTI, Adriana (1990) Generalized Verb Movement. Aspects of Verb Syntax. Torino, Rosenberg & Sellier. BOBALJIK, Jonathan D. & Dianne JONAS (1996) "Subject Positions and the Roles of TP". Linguistic Inquiry 27: 195-236. BORER, Hagit (1989) "Anaphoric AGR". In: O. Jaeggli and K. Safir (eds.): 69-109.

Null Subjects without 'Rich ' Agreement

173

BRODY, Michael (1984) "On Contextual Definitions and the Role of Chains". Linguistic Inquiry 15: 355-80. CHOMSKY, Noam (1981) Lectures on Government and Binding. Dordrecht, Foris. CHOMSKY, Noam (1982) Some Concepts and Consequences of the Theory of Government and Binding. Cambridge, MIT Press. CHOMSKY, Noam (1995) The Minimalist Program. Cambridge, MIT Press. CINQUE, Guglielmo (1990) Types of A' Dependencies. Cambridge, MIT Press. COSTA, Joäo (to appear) "Word Order and Constraint Interaction". Seminários de Lingüística, Universidade do Algarve. CYRINO, Sonia Maria (1996) O Objecto Nulo no Portugués do Brasil. UNICAMP, Ph.D. Dissertation. DIESING, Molly (1990) "Verb Movement and the Subject Position in Yiddish". Natural Language and Linguistic Theory 8: 41-79. DU ARTE, Maria Eugènia L. (1995) A Perda do Principio "Evite Pronome" no Portugués Brasileiro. Campinas, SP, UNICAMP, Ph.D. Dissertation. EPSTEIN, Samuel (1983/1984) "A note on Functional Determination and Strong Crossover". The Linguistic Review 3: 299-305. FIENGO, Robert & Robert MAY (1994) Indices and Identity. Cambridge, MA., MIT Press. FIGUEIREDO SILVA, Maria Cristina (1994) La Position Sujet en Portugais Brésilien. Université de Genève, Ph.D. Dissertation. HUANG, C.-T. James (1982) Logical Relations in Chinese and the Theory of Grammar. MIT, Ph.D. Dissertation. HUANG, C.-T. James (1984) "On the distribution and reference of empty pronouns". Linguistic Inquiry 15: 531-74. HUANG, C.-T. James (1987) "Remarks on Empty Categories in Chinese". Linguistic Inquiry 18: 321-37. HUANG, C.-T. James (1989) "Pro-drop in Chinese: a generalized control theory". In: Osvaldo Jaeggli and Ken Safir (eds.): 185-214. JAEGGLI, Osvaldo & Ken SAFIR (1989) "The Null Subject Parameter and Parametric Theory". In: Osvaldo Jaeggli and Ken Safir (eds.) The Null Subject Parameter. Dordrecht, Kluwer: 1-44. JONAS, Dianne (1996) "Clause Structure, Expletives and Verb Movement". In: W. A b r a h a m e t a l . (ed.) Minimal Ideas. John Benjamins, Philadelphia: 167-88. KATO, Mary A. (1999) "Strong Pronouns, Weak Pronominals and the Null Subject Parameter. PROBUS 11 (1): 1-37. KOOPMAN, Hilda & Dominique SPORTICHE (1991) "The position of Subjects". Lingua 85: 211-58. LASNIK, Howard (1985) "Illicit NP Movement: Locality Conditions on Chains?". Linguistic Inquiry 16: 481-90. MODESTO, Marcello (In progress) "On the Identification of Null Arguments". University of Southern California, Doctoral Dissertation. MONTALBETTI, Mario (1984) "After Binding". MIT, Doctoral Dissertation. NEGRAO, Esmeralda V. (1990) A distribuiçâo e a interpretaçào de pronomes na fala de crianças da escola pública. Unpublished paper. NEGRÄO, Esmeralda V. (1997) "Asymmetries in the distribution of overt and empty categories in Brazilian Portuguese". In: J. R. Black & V. Motapanyase (eds.) Clitics, Pronouns and Movement. Current Issues in Linguistic Theory 140. Amsterdam & Philadelphia, John Benjamins: 217-35.

174

Marcello

MODESTO

NEGRÄO, Esmeralda V. & Ana L. MÜLLER (1996) "As Mudar^as no Sistema Pronominai do Portuguès do Brasil: Substituifäo ou Especializa^äo de formas?" D.E.L.T.A. 12: 125-52. PICALLO, Carme (1994) "Catalan Possessive Pronouns: The Avoid Pronoun Principle Revisited". Natural Language and Linguistic Theory 12: 259-99. RAPOSO, Eduardo (1986) "The Null Object in European Portuguese". In: Osvaldo Jaeggli and Carmen Silva-Corvalan (eds.) Studies in Romance Linguistics, Dordrecht, Foris: 373-90. RIZZI, Luigi (1986) "Null Objects in Italian and the Theory of pro". Linguistic Inquiry 17: 501-57. RIZZI, Luigi (1990) Relativized Minimality. Cambridge, MIT Press. ROBERTS, Ian & Mary A. KATO (1993) Portuguès Brasileiro: urna viagem diacrònica. Campinas, SP, Editora da UNICAMP. ROHRBACHER, Bernhard (1994) The Germanic VO Languages and the Full Paradigm: A Theory of V to I Raising. University of Massachusetts, Ph.D. Dissertation. SAITO, Mamoru (1985) Some Asymmetries in Japanese Syntax and Their Theoretical Implications. MIT, Ph.D. Dissertation. SPEAS, Margareth (1994) "Null Arguments in a Theory of Economy of Projection". In: E. Benedicto and J. Runner (eds.) Functional Projections. UMOP 17: 179-208. TARALDSEN, Knut Tarald (1978) "On the NIC, Vacuous Application and the that-trace Filter", MIT, ms. THRÀINSSON, Höskuldur (1996) "On the (Non-) Universality of Functional Categories". In: W. Abraham et al. (ed.) Minimal Ideas. John Benjamins, Philadelphia: 253-81. XU, Liejiong (1986) "Free Empty Category". Linguistic Inquiry 17: 75-93.

Brazilian

Portuguese

VS Order:

A Diachronic

Analysis

175

BRAZILIAN PORTUGUESE VS ORDER: A DIACHRONIC ANALYSIS 1 Rosane de ANDRADE BERLINCK (UNESP / Araraquara) 0.

Introduction

As pointed out by Jaeggli & Safir (1989), the explanation of the already classical distinction between null subject languages and non-null subject languages has been a major problem for linguistic theory since the early work of Perlmutter (1971). With the development of the Generative Theory, within the framework of Government and Binding, the null subject property soon emerged as a strong candidate for a parameter associated with linguistic variation. This idea emerged as several studies revealed that such a property seems to co-vary with, at least, two other grammatical properties: subject-inversion, and the possibility of long distance extraction of the subject over a lexically filled complementizer (thaitrace condition violations). Besides the importance that the parametric approach has for the studies of language typology, it has obvious implications for the study of language internal variation and change. Thus, once a language property changes over a certain period of time, the other linguistic properties associated with the same parameter are expected to change as well. As far as the null subject parameter is concerned, the attested non-chance correlation between the possibility of a null subject and that of subject-inversion implies that any modification of the former will correspond to a parallel modification of the latter. The changes undergone by Brazilian Portuguese (BP) constitute a good opportunity to test the validity of this association. In this variety, null pronominal subjects used to be preferred to lexical pronouns in tensed sentences until the second half of the 19' Century. Since then, we observe a noticeable decrease in the frequency of this pattern (Tarallo 1983; Duarte 1993, 1995; Cyrino, Duarte and Kato, this volume). Similarly, the subject-inversion in declaratives, still productive in the beginning of 18th Century, became progressively more restricted from the end of that century (Andrade Berlinck 1989, 1995; Torres Morais 1993)2. 1

The ideas discussed in this paper and the results presented here are based mainly on a revised version of chapters 6 and 7 of my Ph.D. Dissertation (Andrade Berlinck 1995). I am grateful to CNPq (Conselho Nacional de Desenvolvimento Científico e Tecnológico Brasil) for a Doctoral Research Fellowship (Proc. n.° 20.0875/90.6) which made it possible for me to carry out this research. A strong decrease in V S is also observed in interrogative sentences, although here the process began later. According to Duarte (1992), by the end of 19th century, SV appears

176

Rosane de ANDRADE

BERLINCK

Considering, however, that both phenomena are still possible and found in current BP, we can indeed ask what kind of null subject and what kind of subject-inversion BP has lost. The aim of this paper is to contribute to an answer to the second part of this question. Section 2, below, discuss the heterogeneous character of VS in Portuguese, presenting a descriptive picture of VS constructions in Modern BP, compared to that of other varieties of Portuguese. The third section presents the results of an analysis of VS constructions in BP declarative sentences based on a large corpus of plays, personal letters and travel narratives from the 18th to the 20th Centuries 2 . Finally, a revision of some of the ideas presented in section 2 is proposed on the basis of the analysis discussed in section 3.

1.

The heterogeneous character of VS in Portuguese

The affirmation of a non-chance correlation between null subjects and free subject inversion implies, incorrectly, the idea that subject inversion is a simple phenomenon. On the contrary, this term comprises constructions heterogeneous in their nature and behaviour. As far as Portuguese is concerned, the analysis of this heterogeneity must consider, besides the distinctions internal to the system, contrasts observed both along a horizontal axis (geographical varieties) and a vertical one (distinct language periods).

1.1.

Modern Brazilian Portuguese

The heterogeneous character of VS in Modern Brazilian Portuguese has largely discussed by Tarallo and Kato (1989) and Kato & Tarallo (1993). authors propose that what is understood as subject inversion corresponds to different structures in this variety of Portuguese: Right-dislocation or inversion [1], Verb-fronting [2] and Unaccusative construction [3]3. [1]

been The three False

a.(ele) Ta pronto o vestido azul. b.[[ 1P proy [p (es)tai [Vp tk [y t( pronto]]]] o vestido azul k ] it is ready the blue dress

in association with a new construction - the use of the expletive e que. This correlation is maintained until the second half of 20th Century, when SV became the preferred construction, independent of the presence of the expletive. For further discussion on V S in interrogatives, see also Lopes Rossi (1993, 1996) and Kato (1993). 2 See the others works in this volume for further discussion of the null subject property in BP. 3 The translations proposed are not intended as ideal English versions for the Portuguese examples. Their purpose is only to help the reader in the task of interpretation.

Brazilian Portuguese VS Order: A Diachronic Analysis

[2]

177

a.Ali moravam os meninos. b.[ IP Ali, [p moravanij [Vp os meninos [y tj t a ]]]] there lived the boys

[3]

a.Apareceu um aluno/ o Papa. b.[i P 0 [p apareceui [ V pum aluno/o Papa k [y tj t k ]]]] (there) appeared a student/the Pope

The structural position of the subject opposes [1] to [2] and [3]: while it is outside the IP in [1], it is inside the VP both in [2] and [3]. These last two, however, differ as to the subject base-generated position. The argument is projected as external constituent in [2], thus occupying [Spec VP]; but in [3] it appears as sister to the verb. As the unaccusative verb is unable to assign Case to its internal argument, when it raises to I to receive its inflection, the N P argument must follow it, at least to [Spec VP], in order to receive Case. Another important difference between [2] and [3] concerns the presence of an initial trigger in [2] which clearly favours a VS construction, in a pattern similar to Germanic inversion (V2), although in Portuguese, V3 and V4 are also possible. According to the authors, in Modern BP this initial constituent has preferably a non-nominal nature: it may be a wh-word, a deitic locative (as in [2]), a measure phrase like price or weight, but not an object. Such an element is not necessary in [3]. Besides these formal differences, the three constructions contrast as to their subject discursive value. For Kato & Tarallo, in Right-dislocation structures, the dislocated NP is a secondary focus, "which operates a predication upon the pronominal subject (pro)". In V-fronting, the subject is placed under the scope of the verb, being interpreted as part of the sentence focus. This is also the reading obtained by the Unaccusative construction, despite the different syntactic status of its argument. Finally, a fourth type of subject inversion in Modern BP is mentioned by Kato & Raposo (1994): Free inversion (the Kayne & Pollock's Romance inversion), exemplified in [4], Contrary to the other three structures, in Romance inversion the NP subject is neither inside the VP nor outside the IP. It is adjoined to the VP as illustrated in [4]b., the corresponding structure of [4]a. [4]

a.Telefonou ontem um amigo meu que eu nao vejo ha 10 anos/o Prefeito. b.[ip pro,.xpi [r telefonoui [Vp [vptk [v ti] ontem] um amigo meu que eu n3o vejo ha 10 anos/o Prefeito k ]]] has called yesterday a friend of mine who I haven't seen the last ten years/the Mayor

178

Rosane de ANDRADE

BERLINCK

Besides significative structural differences, the constructions diverge also as to their productivity. While the Unaccusative construction is quite frequent in Modern BP, V-fronting, Romance inversion and Right dislocation may be considered marginal in this variety4. Despite these contrasts, all four structures share a common feature: they seem to obey the same general constraint, namely a close association with monoargumental verbs. If this characteristic is inherent in the case of the Unaccusative construction, it is by no means a necessary condition for the other ones. This will become evident as patterns present in Modern European Portuguese (EP) as well as Old Portuguese are discussed in the following section.

1.2.

Modern European Portuguese and Old Portuguese

Recent studies about Modern EP5 show that, while bearing some similarities to BP, this variety presents quite a different picture from that of BP as VS constructions are concerned. The main shared characteristic is based on the Unaccusative construction, where VS is the non-marked order for both. As for the other constructions, EP imposes less constraints on VS than BP. For certain contexts, this word order is preferred to non-inverted subjects and even considered obligatory. This is the case of sentences with semantically weak verbs and a preposed constituent in [5] (cf. Ambar 1992). [5]

a.Em Lisboa mora o Pedro. in Lisbon lives Peter b.A verdade disse o Paulo, the truth said Paul

According to Kato & Raposo (1994), the trigger element in V-fronting may also be an object in EP, as indicated by [6]6, contrary to what is observed for BP: [6]

O bolo comeu a Maria the cake-DO ate Mary-SUBJ

The authors state that there is also a difference in the discursive value of the initial element in this construction for each variety of Portuguese: in EP it has always a topical reading, while in BP it can also express the sentence focus.

4

See Kato et al. (1996) for a data-based study on VS in Modern spoken BP. Cf. Brito & Duarte (1982), Ämbar (1992), Kato & Raposo (1994), Andrade Berlinck (1995). 6 Kato & Raposo (1994). 5

Brazilian Portuguese VS Order: A Diachronic Analysis

179

As to the position occupied by the postposed subject, both V-fronting [7]a. and Romance inversion [7]b. are possible 7 : [7]

a.Acontecem essas coisas frequentemente. (there) happen these things frequently b.Acontecem frequentemente essas coisas. (there) happen frequently theses things

However, Ambar (1992) asserts that the first case constitutes the preferred order. The construction becomes relatively less acceptable as the subject NP is placed more to the right, as in [8] 8 : [8]

a.Nesse dia, tinha a Maria/ela emprestado os seus melhores discos ao Pedro. b.Nesse dia, tinha emprestado a Maria/ela os seus melhores discos ao Pedro. c.*Nesse dia, tinha emprestado os seus melhores discos a Maria/ela ao Pedro. d.Nesse dia, tinha emprestado os seus melhores discos ao Pedro ?? a Maria/A MARIA/*ela/ ?*ELA 9 . 'That day, Mary/she had lent her best records to Peter'

Besides all these contrasts, EP differs from BP in a very important way: as examples [5] to [8] indicate, VS is not limited to mono-argumental verbs in EP. N o information about the productivity of these constructions is available for EP, but some are reported as fairly productive in Old Portuguese. Padua (1960) points out XVS as general possibility in OP, more frequent than Romance inversion. As it is still observed in Modern EP, the preposed constituent of XVS could be a direct object, as well as an adjunct. This characteristic led Ribeiro (1992, 1995) to propose that OP presented a V2 syntax; i.e. the verb used to occupy the second clause position, being preceded by any phrasal constituent. This structure has been analyzed as a case of V-fronting. The example in [9] illustrates the pattern 10 : [9]

7

E todo o contràrio faz a Escritura (DSG 3.12.12) and all the contrary does the Scripture And the Scripture does all the contrary

The examples in [7] are taken from Brito & Duarte (1982: 241). The examples in [8] are taken from Ambar (1992). 9 The capital letters indicate emphasis on the constituent. 10 The examples [9] and [10] come from Ribeiro (1992).

8

180

Rosane de ANDRADE BERLINCK

Ribeiro's analysis also shows too that a VI pattern with an immediatley postverbal subject (VS(X)) constitutes a possible, although less frequent construction in OP [10]: [10]

Diremos n6s ora, padre, que... (DSG 1.4.16) say-will-lpp we now, father, that... 'We will say now, father, that...'

Torres Morais (1993) concludes that XVS and VSX are still productive in Portuguese by the 16th and the 17th Centuries. This phenomenon is reduced in Modem BP to residual occurrences, as Ribeiro suggests and as our analysis confirms. Nevertheless, its productivity in earlier periods assures its inclusion in any diachronic analysis of BP.

2.

What is kept, what is lost: a diachronic analysis of VS in BP

Having established the heterogeneous character of VS in Portuguese, we go back to the main question with which this work is concerned: the characterization of VS syntax in BP diachrony. Before discussing the results of my analysis of this phenomenon, some words about methodology are required.

2.1.

On methodology

The study of diachronic phenomena imposes the analysis of empirical data, necessarily taken from written sources. The composition of our corpus of analysis followed the assumptions of the labovian model for the study of linguistic variation and change. According to Labov (1966: 99), in order to study a variable linguistic phenomenon, the researcher must gather the data from the most natural linguistic manifestation - the vernacular. Thus, I looked for written sources which correspond to instances of more informal linguistic expression: personal letters, dramatic literature (mainly comedies), and travel narratives (cf. Rissanen 1986). The corpus comprises three historical periods - the 18th, the 19th and the 20th Centuries. A sub-corpus for each half century was compiled, as an attempt to maintain a difference of fifty years between them. In the following discussion, I will be refering to Times I to VI, which correspond basically to 1730, 1780, 1830-50, 1880, 1930, and 1970-90, respectively. My analysis is restrict to the declarative finite sentences present in the corpus. As indicated by the discussion in section 2, the model of clause structure adopted in the description was the one proposed in the Government and Binding theoretical framework (Chomsky 1981, 1982; Koopman & Sportiche 1991). The sentences described were submitted to a quantificational treatment, following the

Brazilian Portuguese

VS Order: A Diachronic

Analysis

181

methodology of quantitative Sociolinguistics". In section 2.2. the object of analysis is defined and in section 2.3. comments are made on the results of the quantitative analysis.

2.2.

The object of analysis

One of the objectives of this analysis was to determine which VS constructions were present in BP throughout different periods of its history. The following structural possibilities were available after the corpus was investigated12: [11]

A.[IP Xx [p Vj [Vp S [y tj tx]]]]13 or B . [ , P _ [ r Vi [ V P S [ y ti X]]]] a.Logo na barra tem este rio urna perigosa cachoeira; (...) soon at its mouth has this river a dangerous fall 'At its mouth this river has a dangerous fall' (A. Taunay, "Relatos mon5oeiros", 1730) b.Tinha eu meus quinze anos quando là apareceu, vindo do MaranhSo, 0 Sr. Ambròsio. had I my fifteen years when there appeared, coming from Maranhào, Mr. Ambròsio 'I was about fifteen when Mr. Ambròsio appeared there, coming from Maranhào' (M. Pena, "ONovi?o", 1845)

[12]

A.[ 1P pr 0cxpl [,. Vi [VP tk [y ti X ]] Sk]]] B.[, P /vo expl [,. Vi [VP [VP _ [ v ti tk] X ] NPk]]]14 a.Adoro Portinari. Alias sei que eie também me quer um imenso bem e derivou multo disto o retrato que eie fez de mim. 1 love Portinari. Otherwise (I) know that he too wants me a great well and derived much of this the portrait that he made of me

" Cf. Labov (1971, 1972, 1980), Cedergren & Sankoff (1974). 12 Some other VS constructions where the NP subject neither immediately follows the verb nor is in sentence final position appeared in the corpus, as in: (a) Tive noticia, que noanno desepte centos trinta e dous, representara a V. Magde Ignacio Alvares daSilva osprogressos destas minhas deligencias. (1734) 'I have notice that at the year of seven hundred thirty two related to Your Majesty Ignacio Alvares da Silva-SUBJ the progresses of these my diligences-DO' Their marginality, however, has justified their exclusion from the main analysis. 13 X stands for a lexical complement (be it a verbal argument or an adjunct). 14 X is analyzed here, tentatively, as an adjunct to VP.

182

Rosane de ANDRADE

BERLINCK

'I love Portinari. And I know that he likes me a lot too and the portrait that he made of me derived mainly from this' (M. de Andrade, Querida Henriqueta, 1930) b.(...) e que seu IrmSo n3o viera ja ali a falar-lhe porque chegara a sua casa um homem de Sao Joao de El-Rei, que o embarapou; (...) and that his brother did not come yet there to speak to him because arrived to his home a man from S3o JoSo de El-Rei, who embarassed him '...and that his brother did not come yet to speak to him because a man from SSo Jo3o de El-Rei arrived at his home, and embarrassed him' (A utos da Inconfidenc ia Mineira, 1789) [13]

[[\?proy [p Vj [Vp tk [y t| (X)]]]] Sk] Outra vez desaparecido! Esta ficando incorrigivel esse cachorro... again disappeared! (It) is becaming incorrigible, this dog... (R. MagaMes Junior, "O homem que fica", 1939)

[14]

[,p0[rVi[vpNPk[vtitk]]]] Se aparecerem os foguetes, mando-lhe dizer. 'If (there) appear the fireworks, (I) send someone to tell you (this)' (G.Ramos, Cartas, 1930)

For sake of simplification, I will be refering to these possibilities in the following discussion as VSX, VXS, VX#S and VS, respectively. They all are opposed to a construction with a preposed subject, which will be indicated generically as SV(X).

2.3.

The analysis

The results below refer only to declarative finite clauses with nominal subjects. Constructions with prepositional subjects are excluded of the study as they present a categorical VS syntax. The cases of Right-dislocation (VX#S) are computated separately, due to the low frequency of this construction throughout the corpus. They represent less than 0,5% of the total number of sentences analyzed. The majority of these cases, as those in [13] and [15] below, comes from the plays used in the analysis. This suggests that the construction may be found mainly in colloquial speech, a modality that dramatic literature usually aims to reproduce.

Brazilian Portuguese

[15]

VS Order: A Diachronic

183

Analysis

E bem bonita a Quinota! (she) is very pretty, Quinota (A. Azevedo, "Otribofe", 1891)

The discussion, then, will be centered on the variation among SV(X), VSX and VXS. Table 1 presents the general frequencies for these three constructions 15 . These results do not include the data refering to unaccusative verbs, which will be discussed later. Table 1: Word order frequencies in BP across six periods of time Period of time Word order SV VSX VXS

I

II

III

IV

V

VI

42% 34% 24%

76% 14% 10%

90% 6% 4%

89% 5% 6%

94% 2,5% 3,5%

96% 2% 2%

The values show a clear contrast between the beginning of the 18th Century and the second half of 20th Century. At Time I the distribution between pre- and postverbal subjects indicates a small predominance of VS constructions. Time VI shows, on the contrary, an almost categorical presence of SV constructions. In between, we observe a progressive decrease in the frequency of VS constructions, whose point of departure can be established at Time II (by the end of 18th century). Considering these frequencies, we conclude that the process is not uniformly gradual. It is possible to talk about two prevailing moments of change. First, the passage from Time I to II, when the most intense decrease is observed. Second, the passage from Time II to III (first half of 19th Century). After this moment, the constructions present slight differences in frequency between each period. It seems that by the second quarter of the last century the present pattern was already well-established.

l5 The frequencies in Table 1 correspond to the following numbers of tokens. amounts given in parentheses are the totals of tokens analyzed for each period:

Period of time Word order SV VSX VXS

The

I

II

III

IV

V

VI

87(205) 69 49

313(410) 58 39

1204(1335) 80 51

1171(1309) 68 70

2002(2136) 60 74

1065(1106) 22 19

184

Rosane de ANDRADE

BERLINCK

As we take into consideration general frequencies, the process described seems to affect both two types of subject inversion equally. But important differences emerge as these constructions are analyzed according to the type of verb involved. Tables 2 and 3 below present the results of this analysis for the VSX and VXS construction, respectively. Table 2: Frequency of VSX in BP according to type of verb, across six periods of time Period of time Type of verb inerg. intransitive copula direct transitive indirect transitive oblique bi-transitive

I

II

50% 14% 34%

6% 17,5%

-

-

-

45% 55%

16% 18%

III

IV

V

VI

4% 7% 5% 8% 4,5% 8%

0% 3,5% 4% 0% 6% 11%

0% 2% 1% 3% 1,5% 9%

0% 3,5% 1% 8% 0,8% 2,4%

Table 3: Frequency of VXS in BP according to type of verb, across six periods of time Period of time Type of verb inerg. intransitive copula direct transitive indirect transitive obliques bi-transitive

I

II

36% 34% 11,5%

16% 5%

-

-

-

30% 15%

10% 7,5%

III

IV

V

VI

6% 7% 2,5% 7,5% 3% 2,5%

0% 11% 3% 0% 5% 3%

8% 6% 2% 14% 5% 0,5%

4% 3% 1% 7% 0% 1,5%

The comparison of the results for the two constructions at Time I, when both were still productive, reveals that they were possible with any of the types of verbs considered. However, they did not occur with all types of verb in the same way. The higher frequencies of VSX appear with transitive verbs (direct transitive, oblique and bi-transitive). The sentences in [16] examplify this: [ 16]

a.Logo na barra tem este rio uma perigosa cachoeira; soon at its mouth has this river a dangerous fall (A. Taunay, "Relatos monsoeiros", 1730) b.Soube Jo3o Leite desta procura93o... has known Jo3o Leite of this proxy (A. Taunay, "Relatos sertanistas", 1738)

Brazilian Portuguese KS Order: A Diachronic Analysis

185

c. ...vacas, cameiros e alguns cavalos, entre os quais estava um, que disse o Cacique fora seu, ... cows, sheeps and some horses, among which was one, that said the Chief had been his (A. Taunay, "Relatos monpoeiros', 1730) On the other hand, VXS occurs more often with non-transitive verbs (inergative intransitive and copula), as in [17]: [17]

a. ...ferido, acudiu logo todo o mais gentio, que andava ao Corredio, wounded, came in haste soon all the rest of the gentiles, that were on the run b.mas suposto lanfou duas, nem por isso foi muito o milho de que as proveu\ but despite he had sent two [expeditions], not because of this was too much the corn with which he supplied them (A. Taunay, "Relatos sertanistas", 1738)

These clear cut distinction between mono-argumental and pluri-argumental verbs is still perceptible at Time II. Despite the great general decrease in the frequencies, VSX continues to appear preferably with transitive verbs (16% to 18%), while VXS occurs relatively more often with the copula (16%). After this moment, the contrast became much less marked as the frequencies drop to very low values. The results for the sentences with direct transitive verbs, highlighted in Figure 1 below, can serve as an illustration of this process. Fig. 1:

Frequency of SV, VSX and VXS with direct transitive verbs in BP, across six periods of time

• SV • VSX EVXS

186

Rosane de ANDRADE

BERLINCK

The frequencies obtained for VSX with bi-transitive verbs deserve comment, as they present a different pattern of reduction. Contrary to other transitive verbs, this construction keeps a frequency of about 10% of the word orders found in the corpus until the first half of 20th Century (Time V)16. A closer analysis reveals that these sentences are construed mainly with dicendi verbs, as in [18]. [18] a ...e me recolhendo a minha fazenda da Borda do Campo no mez de Dezembro dito acima me perguntou o Padre Manoel Rodrigues se V. Exca. ja saberia destas cousas. and going back to my farm at 'Borda do Campo' in the month of December above mentioned asked me Father Manoel Rodrigues if Your Excellency would already know about this things ('Autos da Devassa da Inconfidencia Mineira', 1789) b.Disse-me a tal senhora que o novi90 Carlos estava naquele quarto, said me the woman that the novice Carlos was in that room (M. Pena, "O Noviijo", 1845) c.Entre dois sanduiches sabiamente mastigados contou-me voce que a tradutora em causa 'foi mal recebida' na Editora. between two sandwiches wisely chewed told me you that the translator in cause 'has been received badly' at the publisher (M. Lobato, "Cartas escolhidas", 1940) According to Ribeiro (1992), these cases can be analyzed as the type of construction Hirschbtthler & Junker (1988) call discursive. In current BP, such constructions are felt to be a marked, quite unusual word order. The preferable association between VSX and transitive verbs correlates, in an obvious way, to the differences in discursive values presented by the subjects of each construction. As discussed above (Section 2), the subject occurring in VSX has a non-focal interpretation. It has already been given in discourse and, moreover, its antecedent is in the neighbouring preceding sentences. Transitive verbs are usually associated with given entities to which they attribute predications in the ongoing flow of discourse. Thus, it seems natural that transitive verbs are more frequent in VSX than in VXS, whose subject presents a focal reading and often refers to a new entity.

l6 The frequencies obtained for indirect transitives also do not follow the tendencies observed. But in this case, both VSX and VXS present unusual results. As this category of verbs is not very productive, further considerations on its behaviour necessitate an enlargement of the corpus that is beyond the limits of this study.

Brazilian Portuguese

VS Order:

A Diachronic

187

Analysis

The constructions with unaccusative verbs have taken a rather different path, as it can be observed in Figure 2, below 17 : Fig. 2: Frequency of SV, VS, VSX and VXS with unaccusative verbs in BP, across six periods of time

I

II

III

IV

V

VI

Here the 'inverted constructions' present higher frequencies than SV until Time II (the end of the 18th Century). But although SV becomes the preferred construction in Time III, it never reaches the same values obtained with other kinds of verbs. In fact, from Times III to VI, we observe a very stable distribution of word orders: SV and VS maintain almost the same frequencies (around 63% and 28%, respectively) and VSX and VXS present very few occurrences. Moreover, it becomes evident that BP has not lost VS, as the variation of the results along two centuries is not very pronounced 18 .

''Although the nominal argument of ergative verbs is not usually considered their subject, the labels SV, VS, VSX and VXS will be kept for ease of comparison. The figures correspond to the following frequencies:

Period of time Word order

1

II

III

IV

V

VI

SV 25% 40% 60% 62,5% 62,5% 66% VS 33% 40% 30% 26,5% 30% 29% VSX 21% 6,5% 6% 2% 1,5% 2% VXS i 21% 4% 13,5% 9% 6% 3% Word order frequencies with ergative verbs in BP across six periods of time 18 For a profound and enlightening discussion on why this kind of construction is maintained in BP, see Kato (this volume). For a data-based analysis of V S constructions in spoken BP, see Kato et al. (1996).

Rosane de ANDRADE BERLINCK

188

What BP has indeed lost in this case is VSX and VXS. Thus, despite the differences due to verbal nature, unaccusative constructions share a similar process with the other types of verbs. Both constructions have become significantly less frequent since Time II, no matter which type of verb it is. In the characterization of VS constructions, the presence of a preposed complement is usually mentioned as an important feature. This has been indicated, for instance, as a major characteristic of the Germanic inversion. As expected, many cases were found among the data analyzed, the sentences in [19] serving as illustration. [19]

a. ...com a Còpia da Conta que a S. Mg. deo a Camera desta Vila. with the copy of the bill that to His Majesty gave the Chamber of this town (I. J. Mongiardino, "Carta ao Governador da Bahia", 1790) b.À mulher cabe a parte do consolo, ou da ternura; to the woman fits the part of consolation, or of tenderness (J. Alencar, "O demònio familiar", 1857) c.Com o Portinari também me sucedeu isso vàrias vezes. with Portinari also succeeded me this several times (M. de Andrade, Querida Henriqueta, 1930)

As a general loss of subject inversion is observed in BP, it would be expected that such a correlation become weaker. This is, indeed, what the present analysis reveals, as showed in Figure 3, below. Fig. 3: Frequencies of XSV and XVS in BP across six periods of time 1 _

XSV(X) -•-XVS(X)

II

III

IV

V

VI

The figure presents a complete inversion of situation from Times I to VI. This change becomes more dramatic if we consider that the cases of XVS in Modern BP are predominantly ergative constructions, as in [19]c. above. Besides this, it

Brazilian Portuguese VS Order: A Diachronic Analysis

189

is significant that we find in modern BP instances of SV even in contexts which seemed to present 'crystallized' VS syntax, as those of dicendi verbs in reported speech. The sentences [20]a. e b. are quite illustrative of this variation, as they were produced by the same speaker. [20]

a.Ah! compreendo, disse-me ele. ah! (I) understand, said me he b.Pouco, muito pouco, dona Lili me disse. little, very little', Mrs. Lili said me (G.Ramos, Cartas, 1930)

2.4.

Some conclusions

The aim of this analysis was to characterize the loss of the VS order in Brazilian Portuguese, determining which of the VS constructions BP has lost and how this process has taken place. The study showed that both V-fronting (VSX) and Romance inversion (VXS), productive constructions in the beginning of the 18th Century, became marginal structural alternatives in Modern BP. A significant decrease in the frequencies of these two constructions was already registered by the end of the 18th Century. This suggests that the starting point of the process could be placed in between these two moments. The characterization of these two VS constructions revealed that each of them co-occur with different types of verbs: VSX are especially found with transitive verbs, while VXS is more usual with non-transitive ones. This led to differences in the pattern of decreasing use of each one, as they have kept relatively higher frequencies with one kind of verb than with another. As for what the language retained of VS syntax, besides the residual cases still observed in today's BP, only the VNP construction with unaccusative verbs continues to be as productive as it was in earlier periods. The motivations of such maintainance are discussed thoroughly by Kato (this volume).

3.

The heterogeneous character of VS revisited

The analysis of empirical data permits and implies the consideration of the context in which each element/sentence analyzed is inserted. In diving into BP's diachrony for an investigation into the loss of VS syntax, new facts about the heterogeneous character of VS have emerged, which deserve a comment. The association between the postposed subject and focus is a common point in the literature on VS constructions. The NP arguments in Verb-fronting (VSX),

190

Rosarte de ANDRADE

BERLINCK

in the Unaccusative construction (VNP) and in Romance inversion (VXS) are usually described as bearing this discursive value. However, this similarity must be reviewed as we compare [21] to [22], for example. [21]

O médico aconselhou-me uma série de banhos de mar; e, morando aqui em Santa Tereza, seria isso impossivel... the doctor suggested me a series of sea baths; and living here at Santa Tereza, would be this impossible 'The doctor suggested to me a series of sea baths; but living here at Santa Tereza, this would be impossible...' (A. Gonzaga, "O hospede do quarto n.° 2", 1937)

[22]

Preparei um trabalho que será de grande transcendencia moral! Que terá resultado estupendíssimo e que muito lucrará com ele a sociedade. (I) prepared a work that will be of great moral transcendency! which will have a stupendous result and that a lot will gain with it the society 'I've prepared a work that will be of great moral transcendency! Which will give a stupendous result and with which the society will gain a lot' (M. Pena, "O Diletante", 1845)

The sentences illustrate VSX and VXS, respectively. This structural difference correlates to distinct discursive values. Only the subject in the VXS construction is interpreted as the sentence focus. In VSX, on the contrary, it makes part of the comment portion of the sentence and does not play, isolately, any marked discursive function 19 . In regard to the informational status, the subject normally refers to a given entity. The subject in VXS, on the other hand, often presents a low degree of accessibility 20 : its referent is less accessible than the complement, what means that it is relatively more 'charged' with information 21 . The subject's informational status does not allow us to analyze these cases as instances of Right-dislocation, as indicated by the impossibility of a variant structure with a correferential pronoun in initial position [23]. [23] 19

* [[IP ela k [ r lucraráj [ Vp tk t¡ com ele]]] a sociedade k]

By 'marked discursive function', I mean the cases in which the phrasal constituent is the main focus or topic, for instance. 20 The concept of 'accessibility' comes from a tradition of studies that include Chafe (1974, 1976, 1987); Givon (1983, 1988, 1990); Payne (1987 a, b), among others. 21 While the complement referent (com ele) was given explicitly in the preceding text, the subject (a sociedade) refers to an entity known to the speaker and the hearer but not yet mentioned in the discourse. For a more detailed discussion about the role of informational distinctions in Portuguese VS structures see Andrade Berlinck (1997).

Brazilian Portuguese

VS Order: A Diachronic

Analysis

191

Thus, there is complementary distribution between VSX and VXS, concerning the discursive function of the verb's external argument. Moreover, the analysis of cases such as those illustrated by [21] makes difficult to sustain that [Spec VP] constitutes a focal position. This affects equally the analysis of Unaccusative constructions, as the NP argument is also argued here to raise to [Spec VP] to satisfy structural requirements. In doing so, it cannot be interpreted, then, as

focus [24 a-b], [24]

a.Vinha eie entre as fumaça da locomotiva. came he among the smokes of the engine 'He came among the smokes of the engine' (G. Guamieri. "Urn grito parado no ar", 1973) b.(...) o dito seu Irmâo tinha ido para Carijós, logo no outro dia apareceu

o mesmo na villa de Sào José; his brother has gone to Carijós, soon in the next day appeared the same one at the town of Sào José 'His brother has gone to Carijós, and in the next day the same one appeared at the town of Sào José' Comparing now [24] to [25], we conclude that the focal interpretation is associated, indeed, with 'real' postposition: [25]

Ai està corno acontece a um naturalista urna coisa que nada tern de natural! 'Here it is how (it) happens to a naturalist something that is not at all natural' (A.Azevedo. "Otribofe", 1891)

If we must conceive the base generated position of the phrase uma coisa que nada tem de natural as the sister branch of the verb in V', then we must equally suppose that this constituent does not occupy this position in [25], but it was 'moved' to the very end of the sentence.

References AMBAR, Manuela (1992) Para urna sintaxe da Inversao Sujeito-verbo em português. Lisboa, Colibri. ANDRADE BERLINCK, Rosane de (1989) "A construçâo V SN no português do Brasil: uma visào diacrònica do fenòmeno da ordem". In: F. Tarallo (ed.) Fotografias Sociolinguisticas. Campinas, SP, Pontes: 95-112. ANDRADE BERLINCK, Rosane de (1995) La position du sujet en portugais: étude diachronique des variétés brésilienne et européenne. Leuven, Katholieke Universiteit, Doctoral Dissertation.

192

Rosane de ANDRADE

BERLINCK

ANDRADE BERLINCK, Rosane de (1997) "Nem tudo que é posposto é novo: estatuto informational do SN e posiçào do sujeito em portugués". Alfa 41: 57-78. BRITO, Ana Maria & Inés S. DUARTE (1982) "Condiçôes sobre posposiçào do sujeito em portugués". Boletim de Filologia. Tomo XXVII. Fascículos 1-4. Lisboa, Instituto Nacional de Investigaçâo Científica: 191-253. CEDERGREN, Henrietta J. & David SANKOFF (1974) "Variable rules: performance as a statistical reflection of competence". Language 50: 333-55. CHAFE, Wallace (1974) "Language and consciousness". Language 50: 111-33. CHAFE, Wallace (1976) "Givenness, contrastiveness, definiteness, subjects, topics and point of view". In: C . L i (ed.) Subject and Topic. New York, Academic Press. CHAFE, Wallace (1987) "Cognitive constraints on information flow". In: R. Tomlin (ed.) Coherence and Grounding in Discourse. TSL 11. Amsterdam, John Benjamins. CHOMSKY, Noam (1981) Lectures on Government and Binding. Dordrecht, Foris. CHOMSKY, Noam (1982) "Some Concepts and Consequences of the Theory of Government and Binding." Linguistic Inquiry monograph 6. Cambridge, Mass., MIT Press. DUARTE, Maria Eugènia L. (1992) "A perda da ordem V(erbo) S(ujeito) em interrogativas q u - n o portugués do Brasil." D.E.L.T.A. 8, n.° especial: 37-52. DUARTE, Maria Eugènia L. (1993) "Do pronome nulo ao pronome pleno: a trajetória do sujeito no portugués do Brasil". In: I. Roberts & M. A. Kato (eds.) Portugués Brasileiro - urna viagem diacrònica. Campinas, SP, Editora da UNICAMP: 10728. DUARTE, Maria Eugènia L. (1995) A perda do principio "Evite Pronome" no portugués brasileiro. Ph.D. Dissertation, Campinas, SP, UNICAMP. GIVON, Talmy (ed.) (1983) Topic Continuity in Discourse: Quantitative Crosslanguage Studies. TSL 3. Amsterdam, John Benjamins. GIVÓN, Talmy (1988) "The pragmatics of word-order: predictability, importance and attention". In: M. Hammond et al. (eds.) Studies in Syntactic Typology. Amsterdam/ Philadelphia, John Benjamins: 243-84. GIVON, Talmy (1990) Syntax. A Functional-typological Introduction. Vol. II. Amsterdam/ Philadelphia, John Benjamins. HAEGEMAN, Liliane (1991) Introduction to Government and Binding Theory. Cambridge: Basil Blackwell. HIRSCHBÜHLER, Paul & Marie-Odile JUNKER (1988) "Remarques sur les Sujets Nuls en Subordonnées en Ancien et en Moyen Français". In: Paul Hirschbiihler & Anne Rochette (eds.) Revue Québécoise de Linguistique Théorique et Appliquée 7 (3). JAEGGLI, Osvaldo & Kenneth SAFIR (1989) "The Null Subject Parameter and Parametric Theory". In: Osvaldo Jaeggli & Ken Safir (eds.) The Null Subject Parameter. Dordrecht, Kluwer. KATO, Mary A. (1992) "Variaçâo sintètica e estilo". Cadernos de Estudos Lingüísticos 22: Homenagem a Carlos Franchi: 127-37. KATO, Mary A. (1993) "Word order change: the case of Brazilian Portuguese whquestions". Paper presented at The XI International Congress of Historical Linguistics, Los Angeles. KATO, Mary A. & Fernando TARALLO (1988) "Restrictive VS syntax in Brazilian Portuguese: its correlations with invisible clitics and visible subjects". Georgetown Roundtable in Language and Linguistics 38.

Brazilian Portuguese

VS Order: A Diachronic

Analysis

193

KATO, Mary A. & Fernando TARALLO (1993) "The loss of VS syntax in Brazilian Portuguese". In: Ingedore V. Koch and Brigitte Schlieben-Lange (eds.) Linguistik in Brasilien. Tübingen, Niemeyer. KATO, Mary A. & Eduardo RAPOSO (1994) "European and Brazilian word order; questions, focus and topic constructions". In: C. Parodi, A. C. Quicoli & M. L. Zubizarreta (eds.) Romance Linguistics in Los Angeles. Georgetown University Press. KATO, Mary A. et al. (1996) "Padröes de predicado no portugués falado no Brasil". In: M.A. Kato (org.) Gramática do Portugués Falado. Vol. V: Convergencias. Campinas, SP, Editora da UNICAMP; Säo Paulo, SP, FAPESP. KAYNE, Richard (1972) "Subject Inversion in French Interrogatives". In: J. Casagrande and B. Saciuk (eds.) Generative Studies in Romance Languages. Rowley, Mass.: Newbury House: 70-126. KAYNE, Richard & Jean-Yves POLLOCK (1978) "Stylistic Inversion, Successive Cyclicity, and Move NP in French". Linguistic Inquiry 9 (4): 595-621. KOOPMAN, Hilda & Dominique SPORTICHE (1991) "The position of subjects". Lingua 85: 211-58. LABOV, William (1966) The Social Stratification of English in New York City. Washington, D.C.: Center for Applied Linguistics. LABOV, William (1971) "Methodology". In: W. Dingwall (ed.) A Survey of Linguistic Science. College Park, University of Maryland: 412-97. LABOV, William (1972) Sociolinguistic Patterns. Philadelphia, University of Pennsylvania Press. LABOV, William (1980) (ed.) Locating language in time, space and society. Quantitative analyses of linguistic structure. Vol.1. New York, Academic Press. LOPES ROSSI, Maria Aparecida (1993) "Estudo diacrònico sobre as interrogativas do portugués do Brasil". In: I. Roberts and M. A. Kato (eds.) Portugués Brasileiro urna viagem diacronica. Campinas, SP, Editora da UNICAMP: 307-42. LOPES ROSSI, Maria Aparecida (1996) A Sintaxe diacrònica das interrogativas-Q do portugués. UNICAMP, Ph.D. Dissertation. PADUA, Maria da Piedade C. e M. de (1960) A Ordern das palavras no Portugués Arcaico. Frases de verbo transitivo. Coimbra, Instituto de Estudos Románicos. PAYNE, Doris L. (1987a) "Information structuring in Papago narrative discourse". Language 62: 783-804. PAYNE, Doris L. (1987b) "Meaning and pragmatics of order in selected South American Indian Languages". Paper presented at the Conference on The Role of Theory in Language Description. Ocho Rios, Jamaica. PERLMUTTER, David M. (1971) Deep and Surface Constraints in Syntax. New York: Holt, Rinehart and Winston. PERLMUTTER, David M. (1978) "Impersonal passive and the unaccusative hypothesis" In: J. Jaeger et al. (eds.) Proceedings of the 4th Annual Meeting of the Berkeley Linguistics Society: 111-43. RIBEIRO, Ilza M. de O. (1992) "Evidence for a V2 phase in Old Portuguese". In: A. Battye & I. Roberts (eds.) Clause structure and language change. N. York, Oxford. RIBEIRO, liza M. de O. (1995) A sintaxe da ordem no portugués arcaico: o efeito V2. UNICAMP, Ph.D. Dissertation.

194

Rosane de ANDRADE

BERLINCK

RISSANEN, Matti (1986) "Variation and the study of English Historical Syntax". In: D. Sankoff (ed.) Diversity and Diachrony. Amsterdam/Philadelphia: John Benjamins Publishing Company. TARALLO, Fernando (1983) Relativization Strategies in Brazilian Portuguese. Doctoral dissertation. University of Pennsylvania, University Microfilms International. TARALLO, Fernando & Mary A. KATO (1989) "Harmonía trans-sistêmica: variaçâo intra- e inter-lingüística". Preediçâo 5. Campinas, SP. TORRES MORAIS, Maria Aparecida (1993) "Aspectos diacrônicos do movimento do verbo, estrutura da frase e caso nominativo no portugués do Brasil". In: I. Roberts & M. A. Kato (eds.) Portugués Brasileiro - uma viagem diacrônica. Campinas, SP: Editora da UNICAMP: 263-306.

Syntactic Codification of Categorical and Thetic Judgments in Brazilian Portuguese

195

SYNTACTIC CODIFICATION OF CATEGORICAL AND THETIC JUDGMENTS IN BRAZILIAN PORTUGUESE Helena BRITTO 1 ( U N I C A M P ' )

0.

Introduction

In contrast to European Portuguese (hereafter EP) and Classic Brazilian Portuguese, present-day Brazilian Portuguese (hereafter BP) fails to license preferential] pro (cf. Figueiredo Silva 1996) and allows VS order only with unaccusative verbs (cf. Andrade Berlinck 1996). Moreover, it has a significantly high frequency of left-dislocated constructions with a lexicalized resumptive subject (hereafter LD), similar to what is found in French (cf. Auger 1995). Also like French (cf. Auger op. cit.; Zribi-Hertz 1994; Jakubowics & Rigaut 1997), the LD constructions of BP are [-marked] from the interpretative point of view, and are thus not equivalent to the LD of null subject languages such as EP (cf. Duarte 1987). This peculiarity of BP has been the object of various investigations by scholars of BP (among them, Pontes 1987; Duarte 1987; Kato 1988, 1993; Galves 1993; Braga 1987; Kato et al. 1996; Duarte 1995; and Figueiredo Silva 1996). In the lines proposed by Britto (1998), the present paper returns to this topic, from the point of view of the thetic/categorical distinction and ga-NPs/wa-NPs of Japanese (cf. Kuroda 1972). LD constructions in BP and SV constructions with lexical subjects in EP, a true null subject language, are compared and found to be equivalent from the interpretative point of view: as with LD constructions in French (Kuroda 1972, 1992), LD constructions in BP are considered to represent the syntactic codification of categorical judgments - equivalent to the SV constructions with a lexicalized subject in EP. The dislocated DP and the lexicalized preverbal subject are the topic of these structures (i.e., the constituent to be interpreted in the appropriate component as the subject of a categorical judgment (Kuroda 1992)). On the other hand, the comparison of SV constructions in BP with VS structures in EP (Martins 1994) reveals yet another similarity: both are interpreted as the syntactic codification of a thetic judgment. The syntactic derivation of this parallelism will be discussed here. This paper is organized as follows. Following the introductory section, the second section shows a descriptive comparison of data from BP with data from Japanese presented by Kuroda (1972, 1976), data concerning unaccusatives in BP presented by Kato (1988), and data from EP presented by Martins (1994). The third section involves a study of strong and weak nominative pronouns in BP, in order to demonstrate that LD constructions in this language are 1

This paper had the support of FAPESP / Grant n.° 98/12075-3.

Helena BRITTO

196

interpretatively equivalent to the SV constructions of null subject languages since both constructions project the same syntactic structure, the difference being the null or lexicalized nature of the weak pronouns. From the distribution of strong and (lexicalized) weak pronouns in BP and the corresponding distribution in EP, defended at the end of the section, the following parallelism is proposed: [1]

a.

[Esses rapazes these guys

[jp eles sâo muito bonitos.]] they are very handsome

(BP)

b.

[Esses rapazes

[lPpro sSo muito bonitos.]]

(EP)

[¡Ppro s3o muito bonitos.]]

(EP)

c. [Esses rapazes [eles

Based on these basic configurations, the fourth section extends the analysis proposed to LD constructions with resumptive subjects and SV structures in BP, as well as to SV and VS constructions in EP, within the framework of Chomsky (1995). Finally, a brief comparison of the syntactic representation of LD structures in BP, French and EP is made in an attempt to derive both the [-marked] interpretative nature of such structures in BP and French, and the [+marked] one in EP.

1.

The syntactic expression of thetic and categorical judgment in Japanese, EP and BP

1.1.

Kuroda's (1972; 1976) proposal for Japanese

Classical Logic claims that each and every judgment (i.e., the affirmation or negation of a property for an entity) should be expressed in the form of a structure composed of a subject (logical, i.e., the entity to which some property(ies) is(are) attributed) and a predicate (i.e., the property(ies) asserted or negated). In contrast to that, Kuroda (1972, 1976, 1992) defends the position that there are two distinct types of judgments: the categorical judgment, with the subject-predicate structure, and the thetic judgment, with no such structure. According to the author, a thetic judgment describes an event or state, and consists of a single cognitive act, namely, the presentation of an object, either an entity or an event. A categorical judgment, on the other hand, expresses a predicate, and, thus, consists of two separate acts, one of them being the act of recognition of that which is to be made the subject, and the other, the act of affirming or denying what is expressed by the predicate about that subject (Kuroda 1972: 154). The latter is the only type of judgment which can be framed within the traditional logical paradigm. Based on the observation that English sentences such as [2] can be translated into Japanese either as [3]a. or [3]b., Kuroda argues that the two different forms in Japanese correspond to the two different types of judgments (Kuroda 1976: 5): thetic judgments, expressed by independent sentences with a definite or indefinite NP marked as nominative

Syntactic Codification of Categorical and Thetic Judgments in Brazilian Portuguese

197

(i.e., -ga), and categorical judgments, expressed by independent sentences with a definite NP marked as topic (i.e., -wo): [2]

the cat is sleeping there.

[3]

a.neko ga asoko de nemutte iru a/the cat there sleeping is b.neko wa asoko de nemutte iru the cat there sleeping is

To prove that wo-NPs will eventually be interpreted as the subject of a categorical judgment in the appropriate interpretative component of the grammar, the author observes that they cannot be semantically indefinite, but must rather be specific, i.e., they are not used by the speaker when he merely refers to an entity that is not especially individualized, or one that has not had its identity previously established in his mind. Rather, wa-NPs are used only when the speaker is referring to a specific individual entity which is differentiated from all other objects. This interpretative condition on wa-NPs observed by Kuroda is the main argument used to link wa-NPs to logical subjects, since this characteristic of wa-NPs seems to reflect the main property of subjects of categorical judgments (as is also pointed out by Ladusaw, 1994): their presuppositional nature. Kuroda considers the following contexts. Absolute generic sentences, with a neutral interpretation, have wa-NPs in the initial position, since constituents of generic sentences are interpreted either generically or definitely, but never indefinitely. Absolute universal sentences, considered by Kuroda as a special case of generic sentences, also represent categorical judgments. Subordinate conditional and alternative coordinate sentences with wa-NPs are ungrammatical, even under a generic interpretation, while completives and causals with vra-NPs are grammatical, the presence of an NP-ga in the latter being unconditionally blocked. To justify the fact that wa-NPs are obligatorily present in absolute generic sentences and generic causals, but categorically blocked in conditional and alternative generic sentences, Kuroda observes that, given a sentence such as Given the fact that dogs bark, John prefers cats, it is possible to extract the judgment that 'dogs bark'. On the other hand, the judgment that 'men are animals' cannot be extracted from a dependent construction such as If men are animals,(...). Thus, for Kuroda, independent clauses always represent judgments, whereas some dependent clauses only provide material for judgment, i.e., they are interpreted as a constituent of the judgment represented by the whole sentence. Similarly, in sentences such as Since Fido is following John, John is running the judgment 'Fido is following John' can be extracted, but this is not possible with the sentence If Fido is following John, John must be running. In other words (adapting the terms of Kuroda (1972) to those of Rivero (1971)), one can say that dependent clauses must be interpreted as the expression of

Helena BRITTO

198

categorical judgments rather than as material for a judgment whenever the speaker has a positive attitude and assumes that the action to which this sentence is referring has occurred, i.e., the speaker presupposes the truth of the sentence (Rivero 1971: 307).

1.2.

Unaccusative constructions in BP (Kato 1988)

Unaccusatives are the only verbs in BP which accept the VS order. Kato (1988) tried to establish a comparison between the two types of subject (preposed and postposed) and wa-NPs and ga-NPs in Japanese, respectively: [4]

[5]

a.As cartas chegaram. the letters arrived

(=wa-NP)

b.Chegaram as cartas. arrived the letters

(=ga-NP)

a.O pneu furou. the tired punctured b.Furou o pneu. punctured the tire

(=wa-NP) (=ga-NP) (in Kato 1988: 116-24)

Considering the observations in Li & Thompson (1979) and Kuno (1973) about wa-NPs and ga-NPs, Kato (1988) points out that the characteristics of ga-NPs (i.e., interpretation in terms of indefiniteness, lists, and contrastive effect) are equivalent to those already identified by Nascimento (1984) for post-verbal subjects in BP. On the other hand, the characteristics presented for wa-NPs (they establish a referential frame in which a predicate is truly and necessarily definite) seem to be more similar to a preverbal subject, rather than to what was defined by Pontes (1981) as a topic. With these observations, Kato makes an analysis similar to that proposed by Kuroda (1976). She also points out that, due to the influence of Fregean logic in classic syntactic studies (cf. Fillmore 1968 and Bach 1968), as well as in more recent approaches (cf. Fukui & Speas 1986), the category subject becomes a category of surface structure rather than of deep structure. In the same line, Kuroda suggests that ga-subjects are internal to the VP, while wa-subjects are external.

1.3.

EP (Martins 1994)

Considering the arguments by Kuroda (op. cit.) and Kato (op. cit.), Martins (1994) assumes that the logical distinction between categorical and thetic judgments is also formally identified in EP: non-dependent (independent) sentences with SV order are interpreted as the expression of categorical judgments, whereas those with VS order are considered to represent thetic

Syntactic Codification of Categorical and Thetic Judgments in Brazilian Portuguese

] 99

judgments. To justify this assumption, Martins compares the Japanese sentences in [6] with their translations into EP [7]: [6]

a.neko wa asoko de nemutte iru. the/a cat TOP there in sleep is b.neko ga asoko de nemutte iru. the/a cat NOM there in sleep is

[7]

a.O gato estd a dormir. the cat is there to sleep a'.Um gato esti Id a dormir. a cat is there to sleep b.Esti o gato a dormir. is the cat there to sleep b'.Est£ um gato la a dormir. is a cat there to sleep

Assuming the discursive context in which speaker A says Vou por o cao la fora (.I'm going to put the dog outside), Martins says that speaker B's answering with a sentence like [7]b.' would be perfectly adequate, with the DP um gato interpreted only as a participant of the event which the speaker B describes, and worthy of the attention of the speaker only in such way. Answer [7]b. would also be appropriate in the discursive context mentioned, with the difference lying in the fact that, in [7]b., the identity of the cat is known (not the case in [7]b.'), although this identity is not relevant (cf. Martins 1994: 394). On the other hand, replying to speaker A with sentences such as [7]a. and [7]a.' would be inadequate, since they involve a topicalized phrase, which means that the identity of the individual named by o/um gato - the topic of the phrase - is essential not only for the speaker, who is saying something about it, but also for the listener, who will evaluate the assertion o/um gato esta la a dormir as a function of what he knows about the topic (cf. Martins 1994: 393). Martins uses the functional projection XP (adapted from Laka 1990) in a configuration in which I P dominates AgrP, while the latter, in turn, dominates TP, and defends the hypothesis according to which the final location of the subject in visible syntax ([spec, EP], [spec, AgrsP] or [spec, TP]) reveals interpretative distinctions about the judgment: (i) for independent sentences with SV order, the subject is found in [spec, XP] with the topic Case, and the sentence expresses a categorical judgment; ii) for independent sentences with VS order, the subject is found in [spec, TP], with the nominative Case, and the sentence expresses a thetic judgment; (iii) for dependent clauses with SV order, the subject is found in [spec, AgrsP], with nominative Case, and the sentence is considered material for judgment; (iv) for dependent sentences with VS order, which can be analyzed as presumed focus

200

Helena BRITTO

(as in Chomsky 1971), the subject is found in [spec, TP], with nominative Case, and the sentence expresses a response judgment.

1.4.

LD + resumptive subject: expression of categorical judgment in BP

The productivity and diversity o f syntactic contexts in which L D constructions are encountered in B P have been pointed out by various authors, including Duarte (1995), whose analysis predicts that they should face no restrictions in use, since they appear freely in the following situations: - in root and embedded sentences: [8] a.[A populafSo neotrentina]j ela; 6 meio flutuante porque os homens (...) the neotrentian population it is relatively unstable because the men... b.EntSo voce acredita que [a pris2o do PC]; ela, so vai acontecer por acaso? then you believe that the imprisonment of PC it only will happen by chance - with definite and indefinite reference [9] a.EntSo [o Instituto de F.]j ele; manda os piores professores, (...) then the Institute o f F. it sends the worst professors... b.[Um homem comum]; ele; tem um conforto compativel com (...) a common man he has a comfort compatible with... - with dislocated free relatives [10] [Quem vem fazer compras no Serra e Mar]; ele; n3o faz compras, ele; passa momentos de alegria. (speaker in grocery store) whoever comes to buy things in the Serra e Mar he not buys things, he spends moments o f happiness - with quantified DPs: [11]

[Toda pessoa que assiste uma pe?a]; ela; tem uma opiniao (...) every person who attends a play he has an opinion (...)

The data below seem to indicate, however, that there are restrictions in relation to such constructions. Adapting the examples from Martins (1994) to BP, it can be seen that for the context in [12], from which the author was able to extract sentences expressing a thetic judgment, only the (a) sentences in [13] and [14] are adequate responses in BP:

Syntactic Codification of Categorical and Thetic Judgments in Brazilian Portuguese

[12]

A person gets home and, before he asks anything, someone says:

[13]

a.[Joana] ligou. Joana called

201

b.*[A Joana]| ela; ligou Joana she called [14]

a.(Vo)ce viu o jornal? [Um judeu] matou o Primeiro Ministro de Israel. you saw the newspaper? A Jew killed the Prime Minister of Israel b.*(Vo)ce viu o jornal? [Um judeu]| ele, matou o Primeiro Ministro de Israel. you saw the newspaper? A Jew he killed the Prime Minister of Israel

On the other hand, for a context such as [15], similar to the situations considered by Martins for obtaining sentences expressing categorical judgments, a sentence like [13]b. (repeated below as [16]) is now perfectly adequate: [15]

For a week, no one in the family talks about anything else except for a certain trip and the travel agent, called Joana, who was responsible for all the arrangements. The husband gets home, asks 'what's new?' and his wife replies:

[16]

[A Joana]j elaj ligou e disse que esta tudo certo. Joana she called and said everything is OK

Moreover, non-specific NPs in BP, which are permitted in existential sentences, probably because they have no presupposed existence (cf. En9 1991), do not participate in dislocation ([17]a. vs. [17]b.), in the same way that non-specific subjects are blocked in EP SV sentences (cf. Martins op. cit.) [17]

a.*[Dois homens]| elesj querem falar contigo. two men they want to talk with you b.[Aqueles/os dois homens]j elesj querem falar contigo. those/the two men they want to talk with you

In two other contexts - generic and universal sentences - which are syntactically coded as categorical judgments in EP and Japanese, dislocation with resumptive subjects also occurs in BP: - generic sentence: [18] Gatosj elesj comem peixe. cats they eat fish

Helena BRITTO

202

- universal sentence: [19] a.[Toda criaba]; ela¡ aprende rápido a gostar de Coca-Cola every child he/she learns quickly to like Coca Cola b.[Todo homem]¡ ele¡ odeia se sentir fraco. every man he hates feeling himself weak Also in dependent sentences, the examples below seem to show that, as in Japanese, categorical judgments can be extracted from the subordinate sentences in BP, as long as the complements involve a positive presupposition in relation to truth value (as in Rivero 1971: 307). [20]

a. 70s corredores n3o acreditam que [aquele belga] ¡ ele¡ ganhe a corrida. the racers not believe that that Belgian he win-subjunctive the race b.Os corredores nSo acreditam que [aquele belga]¡ ele¡ ganhou a corrida, the racers not believe that that Belgian he win-indicative the race

Finally, canonical predicative copular sentences versus inverted predicative copular sentences, in the sense of Moro (1993), behave as predicted by the hypothesis defended here: when an NP cannot be interpreted as the subject of a categorical judgment nor the subject of predication, in the sense of Calabrese (1986), it cannot be dislocated in (present-day) BP. The kind of asymmetry found in [21] can be derived indirectly from the fact that only the arguments of a predicate, never the predicate itself, can be interpreted as the subject of a categorical judgment or the subject of predication: [21]

a.[0 roubo do material]¡ ele¡ foi a causa do problema. (canonical copular sentence) the theft of the material it was the cause of the problem b.*[A causa do problema]¡ ela¡ foi o roubo do material. (inverted copular sentence) the cause of the problem it was the theft of the material

In summary, the examples presented here seem to indicate that BP, prevented from free use of the SV/VS distinction, which exists in EP, and not presenting Case morphology as Japanese does, has adopted the use of LD constructions to express categorical judgments, thus liberating SV order for the expression of thetic judgments.

2.

SV sentences in EP and LD constructions in BP

Following the hypothesis to be defended here, SV sentences in null subject languages are interpretively equivalent to LD in BP because both project

Syntactic Codification of Categorical and Thetic Judgments in Brazilian Portuguese

203

identical syntactic structures, with the difference residing in the null versus lexical nature of the weak preferential] nominative pronoun. To confirm this hypothesis, the following topics will be discussed below: (i)

the partial pro-drop nature of BP, based on the results of Negrao (1997) and Figueiredo Silva (1996), to show that in (present-day) BP only [-referential] pro exists;

(ii)

the behavior of strong and weak pronouns in Italian and French, based on the work by Cardinaletti (1997). This will be done to compare the behavior of weak pronouns in a pro-drop language with those in a clearly non-pro-drop one, thus providing a point of reference for the behavior of nominative pronouns in BP;

(iii)

the behavior of strong and weak pronouns in BP, to show that resumptive subject pronouns accompanying LD constructions in BP should be classified as weak pronouns. The structures in [22] and [23] for SV sentences in a pro-drop language and LDs in BP and French, respectively, are then presented:

[22]

[xp DP [ip pro

[p V (...)]]]

(EP)

[23]

a.[XpDP[IPele

[ r V (...)]]]

(BP)

b.[XpDP[,Pil

[pV (...)]]]

(French)

2.1.

Partial pro-drop nature of BP

Negrao (1997) compared the syntactic distribution of the lexicalized nominative pronominal forms ele/ela and the null form pro with the distribution of the corresponding possessive forms seu/sua and the compound forms of prepositions + strong pronoun dele/dela in BP. She found that both the nominative pro and the possessive forms seu/sua occur in contexts where they are linked to either variables ([24]a. and [25]a.) or other pronouns ([26]a., [27]a., [28]a. and [29]a.). By contrast, the lexical nominative pronominal forms ele/ela and the prepositional forms dele/dela are definitely blocked in such contexts ([24]b to [29]b). However, these forms are regularly present in referential contexts - in which pro and seu/sua can also occur, since, as can be seen in [30] and [31], there is a relation of c-command between these categories and the DP to be taken as the antecedent: [24]

a . [ 0 Guilherme]j disse que [ninguim]^ acredita que 0*i/k/*w vai ganhar. Guilherme said that no one believes that will win b . [ 0 Guilherme]i disse que [ninguem]^ acredita que elej/*^ w vai ganhar. Guilherme said that no one believes that he will win

Helena BRITTO

204

[25]

a.[0 Guilhermejj disse que [ninguém]^ telefonou para [sua¡/k/*w m3e]. Guilherme said that no one telephoned to his mother b.[0 Guilhermejj disse que [ninguém]^ telefonou para a [müe dele;/*^]. Guilherme said that no one telephoned to his mother

[26]

a.[Quem]¡ [t]¡ acha que 0 ¡ disse que 0 ¡ é inteligente? who think that said that he is intelligent b.[Quem]¡ [t]¡ acha que 0 ¡ disse que ele*¡ é inteligente? who thinks that said that he is intelligent

[27]

a.[Ninguém]¡ [t]¡ acha que 0 ¡ disse que 0 ¡ é inteligente? no one thinks that said that is intelligent b.[Ninguém]¡ [t]¡ acha que 0 ¡ disse que ele*¡ é inteligente? no one thinks that said that he is intelligent

[28]

a. [Quem]¡ [t]¡ disse que 0 ¡ acha que [seu¡ filho] é inteligente? who said that thinks that his son is intelligent b.[Quem]¡ [t]¡ disse que 0 ¡ acha que [o filho dele*¡] é inteligente? who said that thinks that his son is intelligent

[29]

a.[Ninguém]¡ [t]¡ disse que 0 ¡ acha que [seu¡ filho] é inteligente? no one said that thinks that his son is intelligent b.[Ninguém]¡ [t]¡ disse que 0 ¡ acha que [o filho dele*¡] é inteligente? no one said that thinks that his son is intelligent

[30]

a.[0 Guilherme]; disse que [o Pedro]|< acredita que 0*¡/k/*w vai ganhar. Guilherme said that Pedro believes that will win b . [ 0 Guilherme]j disse que [o Pedro] ^ acredita que ele¡/k/w vai ganhar. Guilherme said that Pedro believes that he will win

[31]

a.[0 Guilherme], disse que [o Pedro]^ sempre telefona para a [ s u a ^ ^ mSe], Guilherme said that Pedro always telephones to his mother b . [ 0 Guilherme]j disse que [o Pedro]|< sempre telefona para a [m3e dele i/k / w ]. Guilherme said that Pedro always telephones to his mother

Figueiredo Silva (1996: 120-24) also defends the non-existence of preferential] pro in BP, based on constructions like those is [32] to [35]: [32]

a . * 0 que (que)pro comprei ontem? what that bought yesterday

Syntactic Codification of Categorical and Thetic Judgments in Brazilian Portuguese

205

b.O que (que) eu comprei ontem? what that I bought yesterday [33]

a.O Joao vai trazer a salada? Jo2o will bring the salad b.*N§o, O VINHO pro vai trazer. no, THE WINE will bring c.NSo, O VINHO ele vai trazer. no, THE WINE he will bring

[34]

a .pro parece que o JoSo passou por aqui. (it) seems that Joao passed by here b.*Isso/*ele parece que o JoSo passou por aqui. this/it seems that Jo3o passed by here

[35]

a .pro choveu a noite inteira. (it) rained the [whole] night long b.*Isso/*ele choveu a noite inteira. this/he rained the [whole] night long

If BP and EP are compared in the contexts discussed by Figueiredo Silva ([32] to [35]), it becomes clear that the main difference between the two languages lies in what happens in referential contexts, since this is the kind of context in which EP systematically presents pro and blocks the presence of the lexicalized pronoun. The comparison between the respective nominative pronominal paradigms and their corresponding verbal paradigms found in EP and in BP reveals that number inflection is preserved in the morphology of BP, while person features have been drastically reduced. This reduction in the inflectional paradigm, however, does not affect the formal licensing of pro in contexts like [34] or [35], since in both cases the relation of agreement is maintained between the specifier of the maximal projection which harbors pro (i.e., [spec, AgrP]) and its head. The identification of a referential pro in [32] and [33], however, is not possible: as Figueiredo Silva (op. cit.) and Rizzi (1986) affirm, although the features of number in Agr are sufficient for the identification of quasi-argumental and nonargumental pro (as in Chomsky 1981), the identification of referential pro requires the presence of person features in Agr as well. Since these features are absent in Agr° in BP, referential pro is blocked.

2.2.

Strong and weak pronouns in Italian and French

Based on the pronominal typology defended by Cardinaletti & Starke (1993), Cardinaletti (1997) observed that, from the point of view of syntactic rules, the empty pro category of truly pro-drop languages is a weak pronoun. She

Helena BRITTO

206

compares its distribution with that o f both strong and weak pronouns in French, a clearly non-pro-drop language. Both pro and the weak pronoun il in French share distributional equivalence, but contrast with the syntactic behavior o f the corresponding strong pronoun o f il (i.e., lui). Both pro and the weak pronoun il can be used (i) as an expletive; (ii) as a quasi-argument (as in Chomsky 1981); (iii) as an impersonal form, and (iv) as a pronoun with a referential interpretation, either [+human] or [-human]: [36]

a.Il/*lui est arrivé trois filles. it is arrived three girls b.Il/*lui pleut. it rains c.Ilsi m p/*euxj m p m'ont vendu un livre endommagé, dans ce magasin. they to-me have sold a book damaged, in that shop d.Ils/*eux sont très beaux. they are very nice

[37]

a.prof*lui/*loro

(=les gra?ons; =les livres)

sono arrivate tre ragazze,

b .pro/* lui piove. c.proj m p /*loroj m p mi hanno venduto un libro rovinato, in quel negozio. d.pro/*loro sono molto belli. (=i ragazzi; =i libri) Weak lexical pronoun forms in Italian are used as an example for the application o f the remaining tests presented in Cardinaletti & Starke for the classification o f pronouns as strong, weak or clitic (Table 1). The following results are obtained: just as for il, the series o f egli/esso cannot (i) be modified; (ii) participate in a coordinate relation; (iii) remain at S-structure in its base position; nor (iv) occupy a peripheral syntactic position ([38] to [41]): [38]

a.*[Anche/Solo egli] ha dichiarato la propria disponibilità. [also/only he] has declared the his availability b. [Anche/Solo lui] ha dichiarato la propria disponibilità.

[39]

a.*[Egli e suo fratello] hanno dichiarato la propria disponibilità. [he and his brother] have declared the they availability b.[Lui e suo fratello] hanno dichiarato la propria disponibilità.

[40]

a.*Ha aderito Egli. has adhered he b.Ha aderito Lui.

Syntactic Codification of Categorical and Thetic Judgments in Brazilian Portuguese 207

[41]

a.*Egli me hanno detto che non si presentará. he to-me have told that [he] not himself will-present b.Lui me hanno detto che non si presentará.

Table 1: Criteria for pronominal classification (adapted from Cardinaletti & Starke 1993) Pronominal

Classification

Criteria Interpretation

Syntactic Distribution Theta position in SS

LORD position

Clitic

_

Weak

-

-

Strong

+

+

2.3.

COOK? MBDHFC.

-

+

D antee.

Expl

+

+

+

+

-

-

Imperson

Projection

Prosody Word Stress

XP

•Human

Reduction wlea

+

+

+

+

+

+

+

+

-

-

+

+

-

Strong and weak nominative pronouns in BP

In BP (Galves 1997: 20-2), as in German and Hebrew (cf. Laenzlinger & Shlonsky 1997), the application of these tests makes it possible to identify two homophonous series of accusative pronouns, one of them being formed by strong pronouns and the other, by weak ones. [42]

a.Eu deixei ele em casa (= o menino; o Iivro). I left him/it at home (= the boy; the book) a.'Eu vi ELE (e n3o ela). I saw him (and not she) a."*Eu li ELE (e nSoela). I read IT (and not it) b.Eu encontrei ele e ela (= o rapaz e a mofa). I met him and her (= the boy and the girl) b.'*Eu li ele e ela (= o livro e a revista). I read it-masc.sing and it-fem.sing (= the book and the magazine)

Differently from Hebrew and homophonous nominative series, examples, the (a) and (b) versions (c) and (d) versions contrast weak

German, however, BP also presents two as indicated in the examples below (in these should be interpreted as paraphrases, while the nominative and strong pronoun):

Helena BRITTO

208

[43]

a.O JOAO bebe demais (e niSo a Maria). JOAO drinks too much (and not Maria) b.ELE bebe demais (e nSo ela). HE drinks too much (and not she) c.O JoSo bebe demais. Jo3o drinks too much d.Ele bebe demais. he drinks too much

[44]

a.O CARRO bebe demais (e n3o a moto). THE CAR consumes too much (and not the motorcycle) b.*ELE bebe demais (e n3o ela, a moto). IT consumes too much (and not it, the motorcycle) c.O carro bebe demais. the car consumes too much d.Ele bebe demais. it consumes too much

[45]

a.O MEU CACHORRO € grande (e n3o a minha gatinha). MY DOG is big (and not my kitten) b.ELE 6 grande (e nSo ela). HE is big (and not she) c.O meu cachorro 6 grande. my dog is big d.Ele £ grande. he is big

[46]

a.O SOFA e grande (e nSo a mesa de jantar). THE SOFA is big (and not the dining table) b.*ELE e grande (e n2o ela). IT is big (and not it) c.O soft € grande. the sofa is big d.Ele € grande. it is big

[47]

a.O JoSo e a Maria est3o bebendo demais. JoSo and Maria are drinking too much b.Ele e ela est3o bebendo demais. he and she are drinking too much

[48]

a.O carro e a moto estao bebendo demais. the car and motorcycle are consuming too much

Syntactic Codification of Categorical and Thetic Judgments in Brazilian Portuguese

209

b.*Ele e ela estao bebendo demais. he and she are consuming too much [49]

a.Aqui em casa, so meu irm3o bebe demais (o resto da familia quase n2o bebe). here at home, only my brother drinks too much (the rest of the family almost doesn't drink) b.Aqui em casa, so ele bebe demais (o resto quase n3o bebe). here at home, only he drinks too much (the rest almost doesn't drink)

[50]

a.Aqui em casa, so o meu carro bebe muito (os outros s3o bem economicos). here at home, only my car consumes too much (the others are very economical) b.*Aqui em casa, s6 ele bebe muito (os outros s3o bem economicos). here at home only it consumes a lot (the others are very economical)

On the one h a n d , s t r o n g p r o n o u n s , w h i c h are n e v e r i n t e r p r e t e d as h a v i n g a [-animate] referent, are the only ones which are subject to contrastive stress. Thus, according to the criteria for typological classification of pronouns presented in table 1, once a pronoun has been stressed, it can only refer to a [+animate] referent, a prediction confirmed by the unacceptability of [44]b. and [46]b.. Also, according to the criteria in table 1, since only strong pronouns can undergo coordination, coordinated pronouns will never be interpreted as having [-animate] referents, because having [-animate] referents is a property reserved for weak pronouns. This correlation is also confirmed by the contrast between [47]b. and [48]b.. Furthermore, when pronouns are accompanied by modifiers, only a [+animated] interpretation of the pronoun is possible; hence, only strong pronouns occur in this syntactic distribution ([49]b. versus [50]b.). Still other restrictions are encountered in BP: as expected, only the strong pronoun ele can occupy a peripheral syntactic position ([51]b. and [52]b.). [51]

a.O Jo3o, o cabelo dele estd horrivel. JoSo, his hair is horrible b.Ele, o cabelo dele esta horrivel. he, his hair is horrible

[52]

a.O carro, o motor dele esta horrivel. the car, its motor is horrible b.*Ele, o motor dele esta horrivel. it, its motor is horrible

On the other hand, with respect to weak forms, the weak pronouns in BP occur as lexicalized pronouns in contexts in which they refer to [±animate]

Helena BRITTO

210

antecedents, taking on the preferential] interpretation; while the null form, as already suggested by Figueiredo Silva (op. cit.), takes on the function of expletives and quasi-arguments: [53]

BP a.ElaI*pro estâ muito curta.(=a saia) it is very short (the skirt) b.ElaJ*pro estâ muito comprida.(=a garota) She is very tall (the girl) c.pro/*e\e chegou um pacote ontem. (it) arrived a package yesterday d.pro/*ele choveu bastante hoje. (it) rained a lot today

[54]

EP &.pro/*e\a estâ muito curta. (=a saia) it is very short (the skirt) b.pro/*e\a estâ muito comprida. (=a garota) she is very tall (the girl) c.pro/*e\s chegou um pacote agorinha. (it) arrived a package right now d.pro/*e\e choveu demais hoje. (it) rained too much today

[55]

French a.Elle/*pro est très belle. (= la jupe) b.Elle/*pro est très belle. (= la jeune fille) c.Il !*pro est arrivé. d.Il I*pro pleut.

Also only the null weak interpretation in BP. [56]

pronoun pro

can take on the

[+impersonal]

a .pro telefonaram para câ hoje cedo. Eu acho que foi o teu irmâo. (they) called here today early. I think that it was your brother b.*eles telefonaram para câ hoje cedo. Eu acho que foi o teu irmâo. they called here

Finally, in agreement with the prosodie criterion affecting deficient pronouns, BP lexicalized weak pronouns lead to a clear syllabic reduction, with the disyllables becoming monosyllables, as already pointed out by Kato (1996) and Nunes (1994).

Syntactic Codification of Categorical and Thetic Judgments in Brazilian Portuguese

[57]

a.second person singular:

[vo'se]

->

[elis]

[se] [e1]

b.third person singular masculine: [eli] c.third person plural masculine:

211

->

[ez]

3.

Projections of SV sentences and LD structures

3.1.

Basic syntactic configuration for LD constructions

Once the weak lexicalized nominative pronoun ele in BP behaves like nominative pro in EP and the other truly pro-drop languages, the syntactic and interpretative parallelism found between BP LDs and EP SV sentences can be appropriately derived, if one considers the proposals according to which pro is always found in SV structures in null subject languages, and the preverbal lexical subjects are adjoined to IP (cf. Soriano 1989; Barbosa 1997): [58]

[59]

[60]

a.[Ella [¡ppro [ha comprado (...) ]]]. (Spanish) b.[Ela [ j p p r o [comprou (...) ]]].

(EP)

c.[Ela [jp ela [comprou (...) ]]].

(BP)

a.[Maria

[ i p p r o [ha comprado (...)]]].

(Spanish)

b.[A Maria [ j p p r o [comprou (...) ]]].

(EP)

c.[A Maria [ip ela [comprou (...) ]]].

(BP)

a.[Maria

[Ella

[¡ppro [ha comprado (...) ]]]].

(Spanish)

b.[A Maria

[Ela

[ i p p r o [comprou (...) ]]]].

(EP)

c.

[A Maria

[]p Ela [comprou (...) ]]].

(BP)

Notice that the structures in [58] and [59] capture the fact that both of them have identical interpretations as expressions of categorical judgments. Those in [60], on the other hand, yield a derivation for the LDs interpretation as [+marked] contrastive topic, as proposed in Creider (1979) or marked topic, as postulated in Duarte (1987) - only in pro-drop languages, but not in BP.

3.2.

Adjunct position in DP dislocation

Although proposals such as that of Soriano (1989) and Barbosa (1997), among others, suggest the pertinence of the hypothesis defended here, they raise certain questions in face of data from BP. First, if the weak pronouns occupy a specific preverbal position [spec, AgrP], as Cardinaletti & Starke (1993) and Cardinaletti (1997) suggest, would the adjunction of the left dislocated subject DP to IP

Helena

212

BRITTO

guarantee its interpretation as the subject o f a categorical judgment? Certainly, in proposals which defend the adjunction o f left-dislocated subject DPs to IP, a certain relationship between the dislocated DP and the weak pronoun in [spec, AgrP] is guaranteed, and this relation, as already pointed out by Cardinaletti & Starke (1993), can be formally derived. In the framework proposed by these authors, strong pronouns project CP, but weak pronouns do not, and CP is where the referentiality index is located. Therefore, weak pronouns have no index: Lacking index, nothing forces them to be referential: they may occur as expletives, impersonals, etc. But since they lack a referential index, they can be interpreted as referential only if they are associated to a (non-deficient) antecedent, through coreference. As a consequence, deficient pronouns can only be referential if they are 'old information' or 'specific'. They are uninterpretable in and by themselves (Cardinaletti & Starke, op. cit.: 28).

Cardinaletti & Starke (1993) further observe that, in dealing with coreference, they are dealing with a configuration in which coindexation is considered to be a function. In this function, only one of the members o f the pair has a reference: the non-referential member will be associated with the referential one, in a proposal similar to that o f Fiengo & May (1994). However, in spite of being able to guarantee such a relationship between the weak pronoun and its antecedent, analyses which propose the adjunction o f the L D DP subject to IP do not explicitly explain why preverbal DP subjects in Spanish (e.g. Soriano 1989) or EP (e.g. Barbosa 1997) have to be interpreted as Thema or subject o f predication (as in Calabrese 1986), or as topic (as in Kuroda 1972; 1992). In addition, nothing is mentioned about why SV sentences in these languages are always interpreted as predicative sentences or as the syntactic expression o f categorical judgments. A s with null weak pronouns in pro-drop languages ([62] from Calabrese op. cit.), in BP lexicalized nominative weak pronouns always have a preverbal lexical subject as antecedent, and this is the only possible subject o f predication: [61]

a.Quando Carlosj bateu em Antonioj, elej/*j estava bebado. when Carlos hit Antonio he was drunk b.Quando Carlosj bateu em Antonioj, ELEj/*j estava bebado. when Carlos hit Antonio H E was drunk

[62]

a.Quando Carloj ha picchiato Antonioj, proy*^ era ubriaco. b.Quando Carloj ha picchiato Antonioj, luij/j* era ubriaco.

In his analysis o f the obligatory coindexation o f Italian null pronouns with the Thema o f the sentence, Calabrese does not consider hypotheses such as the

Syntactic Codification of Categorical and Thetic Judgments in Brazilian Portuguese

213

adjunction of DP subjects to IP. If, however, this type of proposal were applied to examples similar to [62], the configurations obtained would be the following: [63]

a.[[Quando [ipC.j [ippro [ha picchiato A.j, ]]]] [ipproj/*j t e r a ubriaco.]]] b.[[Quando [ipC.j [ I P pro [ha picchiato A.j, ]]]] [ipluij/j* [era ubriaco.]]]

Given these configurations, the question arises again: how can one derive the obligatory interpretation of a phrase in an adjunct position as Thema, or the subject of predication? And also, as a corollary, how can one explain the fact that such sentences with preverbal DP subjects adjunct to IP can only be interpreted as predicative sentences, since interpretation as presentative sentences (as in Calabrese) or the syntactic codification of thetic judgments (as in Kuroda) is completely blocked?

3.3.

Specifier positions for the dislocated DP and the resumptive pronoun

SV sentences in pro-drop languages with S in a specifier position located above [spec, IP] are found in the syntactic generative literature. Cardinaletti (1997) suggests specific syntactic positions for hosting various types of preverbal elements commonly called subjects, including preverbal subjects in both null subject languages, like Italian, and non-null subject languages, like English and French. She suggests the following syntactic distribution: [64] a.[ Agr ip {Gianni/lui} [ A g r 2 P {/>/-o/tuweak}

Vfm [

b.[ Ag riP {John/He} [ A g r 2 P {it} Agr2°

v

fm [

c.UgrlP {Jean/lui} [ A g r 2 p {il}

v

fin [

She defends what she calls the hypothesis of specialization, in which the two subject positions present in [64] are specialized for subjects of different types, specifically strong and weak, with the position [spec, AgrlP] specially reserved for subjects of predication, as in Calabrese {op. cit.). Along the same lines, Martins (1994) suggests that subjects can have more than one position for their landing site: not only can they surface in [spec, TP] or [spec, IP] (i.e., [spec. AgrP]) above the VP, both of which check nominative Case, but they can also surface in [spec, SP], which specializes in checking topic Case2. Any constituent in [spec, Z] is interpreted as the subject of a categorical judgment in the appropriate component. Notice, however, that in both analyses, the filling of the DP subject of the specifier position located immediately above AgrP necessarily

2

See Kato (this volume), for whom the Case of the nominal in this position is the 'default' Case.

Helena BRITTO

214

implies the movement of this DP from the [spec, AgrP] position, where a trace remains. The analysis proposed here departs from what is suggested by Cardinaletti (op. cit.) and Martins (op. cit.), but is close to what is defended by Figueiredo Silva (1996) with respect to double subject constructions in BP, since it considers that both the dislocated DPs and the resumptive subject are taken to occupy specifier positions, the one occupied by the resumptive being the lowest. Thus, [65]a. and [65]b. are the syntactic configurations for the sentences [59]b. and [59]c.: the specifiers of the TopP and IP are specialized in checking topic and nominative Cases, respectively: [65]

a.[T 0p p Mary b.[j o p p Mary

A Maria [ipproj []• telefonouj [yp t¡ tj ] pro telephoned A Maria [ip ela¡ [[• telefonouj [yp t¡ tj ] she telephoned

(PE) (PB)

Since both constituents of the doubled subject in [65] occupy specifier positions, despite the fact that only one thematic role can be attributed by the verb to an external argument, it becomes necessary to explain how formal problems, related to checking formal features and attributing thematic roles in [65], are avoided. Thus, the following questions which were not considered by Figueiredo Silva will be discussed below: (i) Case assignment to phrases generated in a specifier position in the base structure; and (ii) the separation of the assignment of Case and thematic roles.

3.4.

Review of Case Theory and Theta Criterion

The minimalist framework of Chomsky (1995) brought a new approach to Case issues. By defining the lexicon as a list of exceptions (or idiosyncratic properties) coded in an optimal manner, Chomsky (1995: 235-37) suggests that, for any lexical item, one of the idiosyncratic properties coded in the lexical entry would be the sound-meaning relation. This implies a phonological matrix of the familiar kind expressing exactly what is not predictable, and a comparable representation of semantic properties. It is also observed that each item in the lexicon should be accompanied by its formal categorial features, which consist of one of the formal intrinsic features. As for the optional formal features which include the Case feature, Chomsky observes that they should not be included in lexical entries, since nothing predicted by any of the principles of the grammar should be redundantly present in the lexicon. Thus, once it is accepted that the lexicon should only contain what is not the result of general grammatical principles, it is also clear that the lexical entry of an item should not include Case or ((»-features,

Syntactic Codification of Categorical and Thetic Judgments in Brazilian Portuguese

215

since these features would presumably be the natural result of belonging to the category N, and thus part of the principles of Universal Grammar. It is not the case that Chomsky thinks that optional features are added to a lexical item, since Case and «(»-features are added arbitrarily as a noun is selected from the numeration. Once DPs have Case features added, these continue to be submitted to the computational system, where feature checking occurs. Based on this model, the data below can be derived as follows: [66]

a.[x o p p A Maria

[ipelaj

[p telefonouj [yp tj tj ]

b.

[ j p A M a r i a j [p telefonouj [yp tj tj ]

Since features such as Case and «(»-features always should be added to nouns, the proposal to be defended here is that, for a convergent derivation of a structure such as [66]a., the topic Case feature is added to the DP A Maria, whereas the nominative Case feature is applied to the pronoun ela (lexically marked as [+weak]). These features are checked overtly in Top 0 and 1°, respectively. The DP located in [spec.TopP] is marked with topic Case, so that it will be interpreted as a topic in the interpretative component of the grammar, in the sense of Kuroda (1972-1992), with respect to NPs-wa from Japanese. On the other hand, for a convergent derivation of a structure such as in [66]b., the nominative Case feature is added to the DP A Maria to be checked, again in the visible syntax, in 1°. In this case, the structure is interpreted as representing a thetic judgment. If, from the minimalist perspective, the configurations in [66] do not seem to present problems in relation to Case, from the perspective of the Government and Binding Theory, such structures may be theoretically unfeasible if the usual definition of the Visibility Condition is considered. This condition says that DPs receive Case in order to make their thematic roles visible in LF. In [66]a., the weak pronoun is an external argument, which has a thematic role in VP and Case in IP. The DP in [spec, TopP] cannot, therefore, originate in argument position. Thus, if it is generated in [spec, TopP] in the base structure, the assignment of (topic) Case to it would be blocked under this view. However, as defended by Chomsky (1993), if thematic roles are inherently associated with the heads which attribute them (and not with the expressions which receive them), from the conceptual point of view it seems more appropriate to define the Theta-Criterion as a condition on attribution, rather than on reception of these roles (cf. Nunes 1995: 168). As Nunes (1995) observes, under a version of the Thematic Criterion which considers that chains of arguments should receive a theta-role, a sentence such as [67]a. would be able to converge and receive an interpretation equivalent to a passive structure such as [68]:

Helena

216

[67]

a.*It kissed Mary. b.[xp It T [ v p [yp kissed Mary]]]

[68]

Mary was kissed.

BRITTO

On the other hand, in a version of the Thematic Criterion which considers that theta role assigners should saturate their attributions for the convergence of the derivation, the ungrammatically of a sentence such as [67]a. would be justified by the fact that, in the derivation in [67]b., the assignment of the external thematic role by the light verb doesn't occur. Using the same argument, the contrast in [69] can also be explained. In [69]a., the derivation is canceled because it violates the Last Resort Condition. When the phrase John moves from its initial position of insertion (where it satisfies the EPP) to the position of [spec, vP] (where it is assigned a thematic role by v), it bypasses checking operations, since theta role assignment is always in complementary distribution with feature checking: [69]

a.*John; [ v P tj [yp expected ft; to be someone in the room]]] b.Someonej was expected [tj to be tj in the room]

In [69]b., both movements of the phrase someone are licensed due to feature checking: it will move from the base position to [spec, TP[.fini,e]] in order to check the N-feature of T, and from this specifier position to the position [spec,TP[+finite]] for checking of the N-feature of T and the ((»-features and Case features of the DP. Also based on the concept of the Thematic Criterion in Chomsky (1993), Nunes (1995: 91) proposes that the ungrammaticality of [70]b. should be derived on the basis of the lack of thematic role assignment by the verb rather than the lack of a role on the argument, as defended by Chomsky (1995) 3 : [70]

a.John) [ v p tj [yp expected [ someonej to be tj in the room ]]] b.*Johnj [ v p [yp expected [tj to be someonej in the room ]]]

At this point, considering the Thematic Criterion as defended by Nunes (1995), the derivations in [66] should be reconsidered: [71 ]

a-hopP A Maria [jp ela; b.

3

[j- telefonouj [yp tj tj ]

[]p A Mariaj [p telefonou; [yp tj t; ]

According to Chomsky (1995), the ungrammaticality of structures such as [75]b. is explained based on the idea that arguments which do not receive a thematic role violate Full Interpretation conditions in LF.

Syntactic Codification of Categorical and Thetic Judgments in Brazilian Portuguese

217

In [71], the assigner of theta-roles (V in such cases) saturates its assignments completely for the external argument ela in [71]a. and a Maria in [71]b.. The complementary distribution of thematic assignment and feature checking is thus respected, and the DP generated in [spec, TopP] only checks formal features without being linked to any thematic role assignment by V. Since this violates neither the Thematic Criterion nor the restrictions imposed by the Theory of Case, the derivations in [71] are licit.

4.

Final remarks: from BP to French

As Auger (1995) points out, two main approaches are predominant in studies of clitic subject in French. On the one hand, there are those who, strongly influenced by the different syntactic behavior of the clitics in French and those in Trentino, defend the idea that the nominative pronoun il in French is a syntactically independent element which occupies the [spec, IP] position, with the lexical subject being a structure such as [72] in a position of left-dislocation (see Rizzi 1986, and Dufresne 1993): [72]

Pierre il aime la musique

On the other hand, motivated by the recognition of the [-marked] nature of constructions such as [72] in French, others suggest that the phrase Pierre in this structure occupies the position of specifier of IP, whereas il is just a agreement maker attached to V (see as Auger (1993), Roberge (1990), and Zribi-Hertz {op. cit.). If it is assumed that French LDs syntactically codify categorical judgments (see Kuroda 1972), such constructions should be considered equivalent to the LDs in BP. Moreover, BP and French are similar, not only in relation to the syntactic codification of categorical judgments, but also in relation to thetic judgments, since both languages seem to share a similar codification: SV structures stricto sensu. Based on the similarities found in the BP and French data, the following observations about the approaches mentioned by Auger should be made. In contrast to the approach by Auger and Roberge about the position of clitic subjects in French, it is worthwhile to consider the approach of Svolacchia et al. (1995: 92) concerning the presence or absence of resumptive subjects in Somali. They state that the presence vs absence of the resumptive pronoun has semantic consequences: when it is present, the DP subject is topicalized, while when it is absent, the subject is basically new. In French, the interpretative difference between constructions with left-dislocation + resumptive subject and SV constructions is a counter-argument for the thesis that, in this language, the clitic subject is only an element of agreement; it does not seem coherent to assume that the absence versus presence of agreement elements should be able to provoke the interpretative distinction between presentative and predicative sentences. However, the attempt of these authors to mirror the derivation of LD sentences and their [-marked] nature in French is

Helena BRITTO

218

quite adequate from the interpretative point of view for a model of analysis in which different syntactic structures reflect semantic distinctions. On the one hand, the hypothesis in Rizzi {op. cit.), Dufresne {op. cit.) and Barbosa {op. cit.), according to which LD phrases in French are derivationally adjunct to IP, does not, from an cross-linguistic point of view, capture the [-marked] nature of such constructions, since structures with phrases in adjunction are generally interpreted as [+marked]. On the other hand, the hypothesis that subject clitics in French occupy a [spec, IP] is adequate to explain the contrast which exists between clitic subjects in Trentino and those of French (cf. Rizzi 1986). The pertinent aspects of these two approaches can be reconciled if the configurations in [73] to [75] are considered. After all, the explanation contrasting BP to EP would also be valid for French: the [-marked] interpretative nature of a representation such as [73] can be derived from the location of the phrase Pierre in [spec, TopP], and it should be interpreted as a topic, in the sense of Kuroda {op. cit.): [73]

[jopP Pierre

[jp ilj [p aimej la musique [yp tj tj]]] (French)

[74]

[jopP O Pedro[ip elej [p adoraj musicafyp tj tj]]] (BP)

[75]

[jopp O Pedro[ip pro-, [p adoraj musica

[yp tj tj]]] (EP)

The syntactic tests by Kayne (1975) applied exhaustively to the French pronoun il and the other members of the French paradigm by Cardinaletti & Starke {op. cit.) and Cardinaletti {op. cit.) indicate that, syntactically, such pronouns are not clitics, but weak pronouns. The interpretative distinction between LD constructions with lexical resumption and SV sentences in French is codified by the derivations in [76]a. and [76]b.: just like in BP, the absence of the TopP projection in [76]b. bars the interpretation of a categorical judgment, and the sentence is interpreted as a thetic judgment: [76]

a.[x o p p Pierre [jp ilj b.

[p aimej la musique [yp tj tj]]]

[]p Pierrej

aimej la musique [yp tj tj]]]

Turning to EP, whether in SV sentences, which are [-marked] from the interpretative point of view, or in LD constructions, which are usually classified as marked topic constructions, constituents above the IP in the LD configurations in [77] below would always be adjuncts of the projection, as is proposed by Barbosa {op. cit.): [77]

a.[[pJo3o [\ppro

[j- disse-me

]]]

b.[ip Jo3o [ip ele

[ippro disse-me

]]]

Syntactic Codification of Categorical and Thetic Judgments in Brazilian Portuguese

219

However, if it is true that S is adjoined to IP in EP [-marked] SV sentences ([77]a.) as well as in [+marked] LD together with the dislocated non-subject element ([77]b.), it is necessary to assume that the properties present in the [-marked]/[+marked] opposition are not qualitative, but rather quantitative. On the other hand, if, instead of [77], one assumes the representations in [78], expressed in terms of the argumentation presented above, the [+marked] nature of the sentences is qualitatively represented in the syntax by the absence versus presence of adjunction. [78]

5.

a.[x 0 pp Joâo

[IPpro

[p disse-me

]]]

MïoppJoâo

[Toppele

[i P pro [p disse-me

]]]

Conclusions

Assuming that natural languages syntactically codify judgments, it was shown here that, as in French (see Kuroda 1972), BP LD constructions are the syntactic codification of categorical judgments, while BP SV clauses codify thetic judgments. As for the syntactic configuration of thetic judgments, it was observed that, differently from EP, VS order is not available for BP in general, this language making use of SV order for the codification of such judgments. On the other hand, as far as the syntactic configuration of categorical judgments is concerned, the interpretative parallelism found between BP LD constructions and EP SV sentences was syntactically derived from arguments according to which such structures project the same syntactic structure, the difference between them resting in the null nature of the weak nominative pronoun in EP, on one hand, and its lexical nature in BP, on the other. It was also defended that the dislocated DPs, as well as the resumptive nominative pronouns in EP, French, and BP occupy a specifier position at the end of the syntactic derivation. The [spec, TopP] position is reserved for the dislocated DPs, while [spec, IP] is occupied by a null nominative pronoun, in the case of EP, or by a lexicalized one, in the case of French and BP. Finally, based on the minimalist framework of Chomsky (1993) and (1995) and on questions about Case checking and thematic assignment, some arguments in favor of such configurations were presented.

References ANDRADE BERLINCK, Rosane de (1996) La Position du sujet en Portugais. Leuven, Katholieke Univ. / UNICAMP, Ph.D. Dissertation. AUGER, Julie (1993) "More Evidence for Verbal Agreement-Marking in Colloquial French". In: W. Ashby et al. (ed.) Santa Barbara Romance Papers. Amsterdam, Benjamins. AUGER, Julie (1995) "Les Clitiques Pronominaux en Français Parlé Informel: Une Approche Morphologique". Revue Québécoise de Linguistique 24 (1): 22-60.

220

Helena

BRITTO

BACH, Emmon (1968) "Nouns and nouns phrases". In: E. Bach and R. Hams (ed.) Universals in Linguistic Theory. New York, Holt, Reinhart and Winston. BARBOSA, Pilar (1996) "Clitic placement in European Portuguese and the position of subjects". In: Halpern, A. and A. Zwicky (ed.) Approaching Second: Second Position Clitics and Related Phenomena. CSLI Publications. BARBOSA, Pilar (1997) "Subject Positions in the null subject languages". Seminários de Lingüística 1. Universidade do Algarve - UCEH, Faro: 39-63. BRAGA, Maria Luíza (1987) "Essa dupla manifestaçâo do sujeito, ela é condicionada lingüísticamente". A na is do 34° Seminário do GEL. Campinas, SP.: 106-15. BRITTO, Helena (1998) Deslocamento à Esquerda, Resumptivo-sujeito, Ordern SV e a Codificaçâo Sintática de Juizos Categórico e Tético no Portugués do Brasil. Ph.D. Dissertation. CALABRESE, Andrea (1986) "Pronomina". In: N. Fukui, T. Rapoport and E. Sagey (ed.) MIT Working Papers in Linguistics 8: 1-46. CARDINALETTI, Anna (1997) "Subjects and clause structure". In: L. Haegeman (ed.) The New Comparative Syntax. Londres, Longman. CARDINALETTI, Anna & Michel STARKE (1993) "The Typology of Structural Deficiency". University of Venice/University of Geneva, ms. CHOMSKY, Noam (1971) "Deep structure, surface structure and semantic representations". In: D. Steinberg and L. Jakobovits (ed.) Semantics: An Interdisciplinary Reader in Philosophy, Linguistics, and Psychology. Cambridge, MA., Cambridge University Press. CHOMSKY, Noam ( 1981 ) Lectures on Government and Binding. Dordrecht, Foris. CHOMSKY, Noam (1993) "A minimalist Program for Linguistic Theory". In: K. Hale and S. Keyser (ed.) The View from Building 20: Essays in Linguistics in Honor of Sylvian Bromberger. Cambridge, MA., MIT Press. CHOMSKY, Noam (1995) The Minimalist Program. Cambridge, MA., MIT Press. CREIDER, Chet (1979) "On The Explanation of Transformations". Syntax and Semantics 12:3-21. DU ARTE, Inés (1987) A Construçào de Topicalizaçâo na Gramática do Portugués: Regência, Ligaçâo e Condiçôes sobre Movimento. Universidade de Lisboa, Ph.D. Dissertation. DU ARTE, Maria Eugenia L. (1995) A Perda do Principio 'Evite Pronome ' no Portugués do Brasil. Campinas, SP, UNICAMP, Ph.D. Dissertation. DUFRESNE, Monique (1993) L'Articulation Syntaxique et Phonologique de la Cliticisation: Les Cas des Pronoms Sujets en Moyen Français. Université du Québec à Montréal, Ph.D. Dissertation. ENÇ, Miirvet (1991) "The Semantics of Specificity". Linguistic Inquiry 22 (1): 1-25. FIENGO, Robert & Robert MAY (1994) Indices and Identity. Cambridge, MA., MIT Press. FIGUEIREDO SILVA, Maria Cristina (1996) A Posiçào do Sujeito no Portugués do Brasil: Frases Finitas e Infinitivas. Campinas, SP, Ed. da UNICAMP. FILLMORE, Charles (1968) "The Case for Case". In: E. Bach and R. Harms (ed.) Universals in Linguistic Theory. New York. Holt, Reinhart and Winston. FUKUI, Naomi & Margaret SPEAS (1986) "Specifiers and Projections". MIT Working Papers 8: 128-72. GALVES, Charlotte (1993) "O enfraquecimento da concordancia no PB". In: I. Roberts and M. A. Kato (orgs.) Portugués Brasileiro: uma viagem diacrônica: homenagem a Fernando Tarallo. Campinas, SP, Editora da UNICAMP.

Syntactic Codification of Categorical and Thetic Judgments in Brazilian Portuguese

221

GALVES, Charlotte (1997) "La Syntaxe Pronominale du Portugais Brésilien et la Typologie des Pronoms". In: A. Zribi-Hertz (ed.) Les Pronoms: morphologie, syntaxe et typologie. Saint-Denis, Presses Universitaires de Vincennes (PUV): 1134. JAKUBOWICS, C. & C. RIGAUT (1997) "L'acquisition des clitiques nominatifs en français". In: A. Zribi-Hertz (ed.) Les Pronoms: morphologie, syntaxe et typologie. Saint-Denis, Presses Universitaires de Vincennes (PUV): 57-99. KATO, Mary A. (1988) "Sujeito e Tòpico: Duas categorías na Sintaxe?" Cadernos de Estudos Lingüísticos 17: 109-32. KATO, Mary A. (1993) "Recontando a História das Relativas em urna Perspectiva Paramétrica" In: I. Roberts and M. A. Kato (orgs.) Portugués Brasileiro: Uma Viagem Diacrònica. Campinas, SP, Ed. da UNICAMP. KATO, Mary A. (1996) "Phonological, and Semantics Feature Correlation in Language Change and Grammar Selection". Campinas, SP, UNICAMP, ms. KATO, Mary A., Milton do NASCIMENTO, Eunice NICOLAU, Rosane BERLINCK and Helena BRITTO (1996) "Padrôes de Predicaçâo no Portugués Falado no Brasil". In: M. A. Kato (org.) Gramática do Portugués Falado. Vol V: Convergências. Säo Paulo-Campinas, SP, FAPESP-UNICAMP. KAYNE, Richard (1975) French Syntax: The Transformational Cycle. Cambridge, MIT Press., M.A. Thesis. KUNO, Susumu (1973) The Structure of the Japanese Language. Cambridge, MA. MIT Press. KURODA, S-Yuki (1972) "The Categorical and The Thetic Judgment". Foundations of Language 9: 153-85. KURODA, S-Yuki (1976) "The concept of subject in grammar". In: M. Shibatani (ed.) Syntax and Semantics: Japanese Generative Grammar. New York, Academic Press. KURODA, S-Yuki (1988) "Whether We Agree or Not: Rough Ideas about the Comparative Syntax of English and Japanese". Linguisticae Investigationes XII 1: 1-47. KURODA, S-Yuki (1992) Japanese Syntax and Semantics. Dordrecht-Boston-London, Kluwer Academic Publishers. LADUSAW, William (1994) "Thetic and Categorical, Stage and Individual, Weak and Strong". In: M. Harvey and L. Santelmann (ed.) SALT ¡V: 220-29. LAENZLINGER, Christopher & Ur SHLONSKY (1997) "Weak Pronouns as LF Clitics: Clustering and Adjacency Effects in the Pronominal Systems of German and Hebrew". Studia Linguistica 5 (2): 154-85. LAKA, Itziar (1990) Negation in Syntax. MIT, Ph.D. Dissertation. LI, Charles & Sandra THOMPSON (1979) "Third-Person Pronouns and Zero-Anaphora in Chinese Discourse". Syntax and Semantics 12: 311-35. MARTINS, Ana Maria (1994) Clíticos na História do Portugués. Universidade de Lisboa, Ph.D. Dissertation, MORO, Andrea (1993) I Predicati Nomimali e la Struttura della Frase. Padova, UniPress. NASCIMENTO, Milton do (1984) Sur la posposition du sujet dans le Portugais du Brésil. Université de Paris VIII, Ph.D. Dissertation. NEGRÂO, Esmeralda V. (1997) "Asymmetries in the distribution of overt and empty categories in Brazilian Portuguese". In: J. R. BLACK and V. MOTAPANYANE (eds.) Clitics, Pronouns and Movement. Current Issues in Linguistic Theory 140. Amsterdam and Philadelphia, John Benjamins.

222

Helena

BRITTO

NUNES, Jairo (1994) "Nominative Pronoun Reduction in Brazilian Portuguese". University of Maryland, ms. NUNES, Jairo (1995) The Copy Theory of Movement and Linearization of Chains in the Minimalist Program. University of Maryland, Tese de Doutorado. PONTES, Eunice (1981) "Da Importância do Tópico em Portugués". A nais do V Encontro Nacional de Lingüistica, Rio de Janeiro, RJ, PUC-Rio. PONTES, Eunice (1987) O Tópico no Portugués do Brasil. Campinas, SP, Pontes. RI VERO, Maria-Luisa (1971) "Mood and Pressuposition in Spanish". Foundations of language 7: 305-36. RIZZI, Luigi (1986) "On the Status of Subject Clitics in Romance." In: O. Jaeggli (ed.) Studies in Romance Linguistics. Foris, Dordrecht. ROBERGE, Y. (1990) The Syntactic Recoverability of Null Arguments. Kingston, McGill-Queen's University Press. SORIANO, Olga F. (1989) "Strong Pronouns in Null-Subject Languages and the Avoid Pronoun Principle". MIT Working Papers in Linguistics 1 1 : 228-39. SVOLACCHIA, Marco, Lunella MEREU and Annarita PUGLIELLI (1995) "Aspects of Discourse Configuration in Somali". In: K. Kiss (ed.) Discourse Configurational Languages. New York-Oxford, Oxford University Press. ZRIBI-HERTZ, Anne (1994) "La syntaxe des clitiques nominatifs en français standard et en français avancé". In: G. Kleiber and G. Roques (ed.) Travaux de Linguistique et de Philologie. Strasbourg-Nancy, Klincksiek: 131-47.

The Partial Pro-Drop Nature and the Restricted

VS Order in Brazilian Portuguese

223

THE PARTIAL PRO-DROP NATURE AND THE RESTRICTED VS ORDER IN BRAZILIAN PORTUGUESE Mary Aizawa KATO1 ( U N I C A M P )

0.

The aims of this paper

The null subject (NS) parameter has been proposed to be a cluster of properties 2 , including: a) b)

the possibility of null subjects free

inversion / postposed subjects (cf. Chomsky 1981; Rizzi 1982).

Adams (1987) and Roberts (1993a) showed that the loss of null subjects and of VSX (or long verb movement) are related in a diachronic study of Old French. Kato & Duarte (1998) showed, however, that BP maintained the null subject after the loss of long verb movement in interrogatives and that the loss of the null subject correlates with the loss of VOS, or free inversion. However, the correlation between NS and free inversion has been challenged by Safir (1982), who found out that Northern Italian dialects, while disallowing the former, license the latter. Brazilian Portuguese (BP), on the other hand, has been exhibiting a change in progress in both properties, a fact that can be used in favor of a single parameter (cf. Tarallo and Kato 1989; Duarte 1993; Roberts 1993b). Curiously, though Duarte's (1993, 1995) studies show a significant decrease in the use of null subjects 3 , they show that the change affected mainly the first and second persons, leaving a still reasonably stable use of null subjects for the third person. Andrade Berlinck's (1989, 1995) studies show, on the other hand, that free inversion is being lost, but that VS is still productive with the copula and with unaccusative verbs.

' This paper had the support of CNPq (Proc. n.° 300814/88-89R) and FAPESP (Grant 98/14557-5 e 98/13726-8). It was partly developed during my stay at USC as a visiting scholar, under the sponsorship of FAPESP (Proc. n.° 98/14557-5). I thank Joseph Aoun, Hagit Borer, Hajime Hoji, Yuki Kuroda, and Maria Luiza Zubizarreta for the opportunity to present and discuss the paper at USC and for the valuable comments they provided. The same holds for Eduardo Raposo, who invited me to present this paper at U. California, Santa Barbara. My thanks to Jazon Santos for his support and role as devil's advocate. I also thank Evani Viotti for her careful reading of the first version of this paper and, last but not least, I am greatly indebted to Jairo Nunes for his extensive comments and discussions during the conception and writing of this paper. All the remaining faults are of my own responsability. I also thank Angel Mori for the Spanish data and Marilza de Oliveira for the Italian data. 2 See the complete list in the Preface of this volume. 3 See also Tarallo (1983).

224

Mary A izawa KA TO

The still licensed null subjects and VS forms could be viewed as residual products of the former setting of the parameter. But, if one uses Borer's (1984) view that parameters are related to morphology, apparent residual phenomena can be accommodated as part of a stable grammar, given the fact that morphology is not necessarily regular. The aim of this paper is to first present a theory of the N S parameter, based on Kato (1999), which shows that both the possibility of null subjects and the possibility of free inversion can be derived from the same morphological property of the agreement system. I will then show the changes that occurred in BP, regarding the parameter in question, and the nature of the apparent 'residual' phenomena. The paper is organized as follows. Section 1 will present Kato's (1999) theory of the null subject, which is based on Rizzi's (1982) and Burzio's (1986) original insight that INFL in NS-languages is a sort of pronominal or clitic, and on Everett's (1996) claim that agreement affixes, clitics and pronouns are 'allomorphic' realizations of «{»-features. The proposal extends this view to crosslinguistic variation: for the same function, languages optimally choose one of these forms for the nominative pronominal - free weak 4 pronouns, subject clitics or pronominal Agr. The analysis eliminates referential pro as a descriptive category. Section 2 will show how subject inversion (I-subjects) can be derived in languages that choose subject clitics and pronominal agreement (Agr), but not in languages that choose free nominative pronouns. Indefinite and definite VOS constructions are claimed to have differences in their derivations. The former involves only movements for feature checking; the latter derive from the order SVO, with a prosodically-motivated movement of VO, exhibiting weight constraints. Section 3 analyzes existential constructions with and without agreement and it shows that agreement patterns are a function of the Case of the associate and of the way in which the D-feature of T is checked. For languages like Spanish and Portuguese, it is proposed that the postverbal nominal of existential constructions checks accusative case, and a - 0 n e u t e r affix checks the D- and the Case feature of T. The locative elements {there, ci and y) are claimed to distribute like the personal pronouns into: weak pronouns, clitics and affixes and to appear in

4

Pronominals are assumed to be strong or weak as in Cardinaletti & Starke (1994), but the split in Kato (1999) is different: weak pronominals are further divided into: free, clitic or affixal. Strong pronouns are the ones that appear in predicative and left-dislocated position.

The Partial Pro-Drop Nature and the Restricted FS Order in Brazilian Portuguese

225

doubling constructions with PPs. Such elements are seen as meaningful and as requiring raising to check some abstract feature of V+T. Section 4 analyzes the nature of the changes that occurred in BP. The appearance of a free nominative pronoun paradigm led to the reanalysis of agreement affixes as non-pronominal. The appearance of the weak free nominative paradigm forces the projection of Spec of TP. The loss of inversion properties are shown to be merely a consequence of this. The chapter then analyzes the unaccusative and existential constructions, both of which still exhibit VS constructions. It is shown that BP conforms to the requirements of a language like French, with a null agreement expletive of the il-type.

1.

The nature of the agreement system and the null subject parameter

1.1.

Agreement as D-arguments

The decrease of null subjects in BP was shown by Duarte (1993) to have been triggered by the replacement of the second person 'tu' by the addressee form 'Vossa Merce' (=Your Grace) in its reduced form 'voce', which triggers third person agreement (cf. Duarte's table I, repeated in [1]). [1]

(Adapted from Duarte 1993: 109)

Number

Person

Paradigm 1

Paradigm 2

Sg Sg Sg Sg PI PI PI PI

1st 2nd 2nd 3rd 1st 2nd 2nd 3rd

cant-o canta-s canta-0 canta-0 canta-mos canta-is canta-m canta-m

cant-o

direct (tu) indirect (você)

direct (vós) indirect (vocês)

canta-0 canta-0 canta-mos canta-m canta-m

The development of a poorer agreement paradigm thus provides support for the hypothesis of 'rich' morphology as the determiner of the pro-drop parameter (cf. Taraldsen 1980; Borer 1989; and others). But the intriguing fact is that the first person, which is the only form still marked with a distinct morphology, was the

Mary Aizawa

226

KATO

first subject to become more frequently expressed 5 . Moreover, though the indirect second person 'voce' and the third person ('e/e/ie/a') have the same 0inflection, the former spells-out the pronoun more often than the third person (cf. Duarte, this volume). Galves (1993) interprets Duarte's empirical facts as a change in the agreement system. In the same line as Rohrbacher (1992) and Speas (1994), she uses AGR as a clause head and a general Principle of Economy, claiming that AGR 6 is projected only when a language has a strong (rich) agreement morphology: a) languages like Japanese would have no AGRP; b) languages with a strong (rich) agreement paradigm would have AGR filled with the agreement morpheme; c) languages with a weak agreement system have no AGRP, the agreement morpheme being generated as a synchretic category with Tense. Thus, in Galves'analysis, BP lost the independent AGR projection and T is now a synchretic category (T+AGR). For Galves (1993), neutralization of the second and first persons provoked a change in the strength of the A G R head, making it [-person], AGR becomes empty, but its Spec can be filled by a Topic. Though Chomsky (1995) eliminates AGR as a functional category 7 , Galves (1998) proposes that a projection between CP and TP is still necessary to account for cross-linguistic differences and proposes that her former AGR head is a projection of the category Person, a deictic category like Tense. Kato's (1999) analysis retains Galves'idea that agreement in NS-languages does not form a synchretic category with Tense. But instead of proposing that it is the head of a clause projection, it proposes that agreement inflection is the head of a DP, which merges with the verb as its argument. Any determiner -feature carrier, be it a free pronoun, a clitic or a pronominal affix, appears as an item in the numeration and starts the derivation in a similar manner, by being merged in 5

The loss of morphological uniformity (Jaeggli and Safir's (1989) hypothesis) could be claimed to have triggered the loss of null subjects, but it does not explain why in the third person, and especially with expletives, the subject is still null. Moreover, the BP expletive, which is obligatorily null, cannot be accounted for in terms of 'formal licensing' (Rizzi 1986), like the expletive in German, which may be null in certain contexts. 6 We are using: [a] AGR for clause head and [b] Agr for pronominal agreement affixes. When agreement is not independent from the verb and tense entry, it will not be represented. 7 Functional categories in the minimalist work are reduced to only those that have interpretation. Agreement for Chomsky (1995) is a relation and has no independent meaning and, therefore, should not constitute a clausal head.

The Partial Pro-Drop Nature and the Restricted VS Order in Brazilian Portuguese

227

D position. Pronominal Agr is thus syntactically defined as the agreement inflection that appears in the numeration as an independent item from the verb, which appears inflected only for tense. Compare the Agr(eement) like English Trentino and [2]

first stage of the derivation of [2]a., a language that has pronominal like Spanish, with [2]b., languages that have free subject pronouns and German, and [2]c., languages that have subject clitics like Fiorentino:

a. Spec of VP in languages with pronominal Agr like Spanish

b. Spec of VP in languages with free pronouns like English and German

VP

VP

D

V

-o

habl(a)-

XP

V he Ich

c. Spec of VP in languages Fiorentino

XP

speaks spreche

clitic subjects like Trentino and

VP

D teyou

V

XP

parlispeak+pres

Like free weak pronouns and clitics, these affixes have Case and (^-features. If T has strong V-features, it attracts the tensed V or the Auxiliary in order to have its features checked. If T has strong D-features, it will attract the free pronoun, the clitic or the pronominal affix. Pronouns, clitics and pronominal affixes are minimal and maximal categories, but pronouns land in Spec of T and clitics and affixes move as heads. Spec of T is projected in the former case, but not in the latter case.

228

[3]

Mary Aizawa KA TO

Languages with and without projection of Spec of TP: Type a. Spanish

Type c. Fiorentino TP

T

A

Agr

I

VP

/\

T

DP

V'

I

I

I

-Oj habla

tj

Cl

I

V

tej

T

DP

I

V

I

parli

tj U

tv

te parli => te-parli

o habla-=> hablo

Type b2. German

T y p e b l . English

TP

TP

r

DPi

I He

T

VP

DP

V'

I /,

/ \ V speaks

XP

V+T spreche

DP | ti

V V

DP t..

Thus, in both languages of Type b, the free pronoun raises to [Spec,TP]. The difference between them is overt V-to-T movement, as in German, and covert movement (or movement of only the Formal Features), as in English. In languages of Type a., like Spanish, and Type c., like Fiorentino, the subject clitics and pronominal affixes are attracted and adjoined to T. T has its nominative case eliminated after checking, but the «(»-features of the pronominal Agr are retained 8 , as with free pronouns 9 , and Spec of T is not projected 10 . The 8

Recall that, in N S languages, the agreement affix is an independent item in the numeration as in Galves', Rohrbacher's & Speas' proposals. So the verb appears inflected only for tense.

The Partial Pro-Drop Nature and the Restricted VS Order in Brazilian Portuguese

229

difference between clitics and affixes lies only in the order they appear morphologically realized. The Agr chain and the Clitic chain, with their heads in INFL and tails in SPEC of VP, are interpreted at LF as the syntactic subject of the clause. Agr and the subject clitic can thus be interpreted at LF as the external argument of the verb. For Nunes (1998a), checking through adjunction to the head is the optimal configuration and specifiers are only projected if the morphology does not tolerate adjunction to the head. NS languages would be, in his conception, the optimal case, though he does not deal specifically with this problem.

1.2.

The 'default' nominative case of lexical 'Subjects'

Following Kato (1999), it will be assumed that in NS languages, both lexical pronouns and ordinary DP subjects are external to TP 11 . Partially following Martins (1996), it will be assumed that the external position is Spec of EP, interpreted as the 'subject' of a categorical 12 sentence 13 . The difference between Martins and the present analysis is that the former assumes the Specifier of E to co-exist with pro, like Soriano (1989) for Spanish and Barbosa (1997) for European Portuguese, while here pro is excluded as a descriptive category as in Kato's (1999) analysis. Also, Martins postulates AGRP as an intermediate projection between TP and £P, the latter a root category. In the present analysis, both root and embedded sentences can be either LP or TP, depending on the presence or absence of a full DP doubling the Agreement affix. Kato's (1999) analysis proposes that strong pronouns can double any weak feature form: weak pronouns, clitics or even Agreement affixes when these are

9

According to Chomsky (1995: 279), "some features remain visible at LF even after they are checked: for example phi-features of nouns, which are interpreted". 10 Under Chomsky (1993), the EPP is reduced to the conception that the N feature of T be checked in overt syntax. In our analysis, what moves is visible morphology and not only Formal Features and this is taken to satisfy the EPP. " The same claim is made by Soriano (1989) for Spanish, and by Raposo (1994) and Barbosa (1996) for European Portuguese. Others claim that subjects are internal to TP (Zubizarreta 1998, for Spanish; Costa 1998, for European Portuguese). 12 The thetic vs. categorical judgement of classical philosophy was used in Kuroda (1972) to distinguish sentences with -wa and -ga in Japanese, and by Kato (1989) and Martins (1996) to distinguish SV and VS in Romance. The opposition is similar to that given by terms like predication vs. presentational sentences (cf. Gueron 1980; Nascimento 1984; Franchieia/. 1998). 13 See Britto (this volume) for a proposal on the categorical/thetic sentences in BP, coherent with the proposal presented here.

Mary A izawa KA TO

230

[+pronominal] 14 . Thus, subject doubling in NS languages is not a phenomenon that involves a silent pro, as proposed by the above mentioned authors, but the Agreement affix itself. Thus, while French doubles the subject clitic and English the weak pronoun, Spanish doubles the agreement itself. [4]

a.Me¡, I¡.... b.Moij, je¡.... c.Yoj, V+ Agr¡

Comparing the three, Kato (1999) proposes that the strong pronoun has a 'default' case, nominative being the 'default' in Romance N S languages. It could be argued that [4]c. is actually the configuration of nominative case by agreement, in terms of a Spec-head relation. However, the same contrast found in [4] can also be found in other configurations, like [5] and [6], which are not environments of nominative checking. [5]

a. It is me. b. C'est moi. c. Soy yo.

[6]

a. There is me. b. II y a moi. c. Estoy yo.

In Nunes (1998b), the concept of 'default' case has been used to deal with the accusative case of an English sentence similar to [6], He proposes that nominal elements in English (including pronouns) are underspecified with respect to the type of Case they bear, and that a default morphological rule realizes pronouns with unchecked Case-features as accusative" (p. 36). For Kato, NS languages have two types of nominative: the 'default' nominative of lexical DPs, which is not checked, and the nominative feature of pronominal Agr, which has to be checked against T. Checking eliminates all noninterpretable features, which means that only in [4]c. are the D and «(»-features of the verb agreement retained 15 because the agreement is a pronominal with interpretable features.

14 As in Speas (1994), agreement was considered [+pronominal] when the verb incorporated the pronoun, though such incorporation is not always transparent. 15 We may suggest that the 'default' case is not a feature since it is not 'selected' in the numeration. It is visible at PF, but not at LF.

The Partial Pro-Drop Nature and the Restricted VS Order in Brazilian Portuguese

231

The analysis there extends the 'default' case for the doubled DP subjects, as they can appear in coordination with strong pronouns {Juan y yo). The representation of sentences containing a strong pronoun or a DP in pre-verbal position, as in [7]a. and [b.], are shown in [7]' below: [7]

a.Yo lo comi. I it ate+lpSg ('I ate it') b.Juan lo comio. Juan it ate+3pSg ('Juan ate it')

[7]'

a', fep Yoj [TP lo com-ii [yp .••]]] b'. [EP Juan; [ j p lo comi-6j [yp ...]]]

Since the pronominal agreement performs all the necessary checking operations within TP, the lexical elements (yo, Juan) are merged with a head above TP like the normal cases of Left-Dislocation (LD), as in [8]16: [8]

John, he ate the cake.

In Spanish, a NS language, the weak pronominal is the agreement itself, and in [7]' it appears as the resumptive pronominal of the lexical element in ZP. This lexical DP has the properties of a dislocated nominal: it is definite or specific, has the 'default' nominative case, which needs no checking, and is interpreted as the 'subject' of the assertion contained in TP.

1.3.

Evidence of the complementary pronominal forms in Old French

distribution

of

weak

To provide empirical support for the theory proposed, Kato reports to Dufresne and Dupuis's (D&D 1996) analysis of pronouns in Old French. As is well known, since the pioneering work of Adams (1983), Old French has been analyzed as a null subject language which lost this property along with the VS order. For D&D, pronouns in OF were generated as full NPs/DPs, while in Modern French (MF), they are pure Ds17. Their study is relevant for the present analysis, 16 A doubling structure like [8] was analyzed by Kato (1998) as the left-dislocation (LD) of a secondary predicate, of which the subject was a weak pronominal: [DP he [DP John]]. However, this analysis cannot prevent resumptive pronouns from appearing inside islands, the reason why it had to be abandoned. I thank Joseph Aoun for pointing this problem out to me. 17 The authors also assume that after the pronouns become a pure functional D, they also become candidates for phonological clitics.

Mary A izawa KA TO

232

since the form of the ((»-feature carriers present during and after the NS-period is crucial for the hypothesis that is entertained here. D&D show that personal pronouns in OF appeared separated from the verb by intervening elements and could also be conjoined with full NPs, both situations impossible in Modern French. [9]

a.Elle colpes non avet. she sinned never had b.Je meisme cil Yvain sui. I myself this Yvain am c.Mais si tu é li tuns lignages se tresturned de mei. but if you and the your family turn away from me

From the end of the 15th century on, the pronoun has always been adjacent to the verb. Doubling was sometimes possible, as can be seen in the examples by Foulet (apud Roberts 1993a: 112): [10]

a.Renars respond: 'Jou, je n'irai' R answers 'I I won't go' (Coronnement Renart, A. Foulet (ed.) 1929: 598) b.Et jou je cuit... and 11 believe... (ibid, 1616)

The strong nominative pronoun paradigm in OF was replaced by the objective case paradigm, which seems to have already been in competition with the strong nominative forms in OF, as in the examples of Moignet's (apud Dufresne & Dupuis). [11]

a.Et lors en irons moi et vos après le chevalier. and then shall go me and you after the horseman b.Dist li rois à Boort qu'il venist a cort, lui et sa compaignie. aid the king to Boort that he should come to court, him and his courtship

In OF, the strong dative pronouns seem to occur in dislocations to the right, while the strong nominative paradigm seems to occur to the left. In MF, the dative form has become generalized as the strong form on both sides. The initial hypothesis in Kato (1996a) was that when the strong form was nominative, the language was a NSL. But languages like German and the facts of BP presented here have shown that the hypothesis was too strong. In both languages, the strong pronouns are nominative and yet German is a non-NSL and BP is on its way to becoming a non-NSL language. What can be maintained is

The Partial Pro-Drop Nature and the Restricted VS Order in Brazilian Portuguese

233

that if the strong form is not nominative, then the language is [-NS]; the inverse not necessarily being true. The interpretation now given to the facts in French is the following: when strong nominative pronouns co-existed with pronominal Agr, the null subject was possible. When a weak quasi-homophonous nominative paradigm appeared, Agr ceased to be pronominal, the null subject disappeared and the dative form took over the function of the strong pronouns 18 . This confirms the present hypothesis that pronominal Agr and weak subject pronouns are morphological doublets in the sense of Kroch (1994) and that both can co-occur with strong pronouns. [12]

a.Jou, Agr+pronominal]

{OF}

b.Jou, j e agr [-pronominal],..{between OF and MF} c.Moi, j e agr [-pronominal]...{MF} Below, we will show how the present theory of weak pronominals explains the possibility of free inversion in NS languages in a natural manner.

2.

The nature of agreement and the possibility of free inversion

2.1.

Defining free inversion

Postverbal subjects can appear in two positions in Spanish: a) immediately after the verb and before the object as in [13]b.; or after the whole predicate, as in [13]c. [13]

a.Juan comió la torta. Juan ate+3pSg the cake ('John ate the cake') b.Comió Juan la torta. ate+3pSg Juan the cake ('John ate the cake') c.Comió la torta Juan. ate+3pSg the cake Juan ('JOHN ate the cake')

Though Adams (1987) relates the loss of NS in Old French to the loss of its VSO order, the pattern that appears more generalized in NS languages is VOS. The former is found in Romance languages of the Germanic type, like Spanish (SP) and European Portuguese (EP) (cf. Zubizarreta 1998), but not in other Romance languages, like Italian.

18 We may also entertain the hypothesis that the opposite was the case: when Agr ceased to be pronominal, a weak pronoun paradigm was created.

Mary A izawa KA TO

234

In Zubizarreta's (1998) analysis of Spanish and Italian, one of the most extensive studies on word order, it is claimed that: a)

the basic order in SP is [yp VSO];

b)

surface VSO results from raising of V-to-T and the Formal Features (FF) of S to T;

c)

SVO is derived by raising S to Spec of TP, where other constituents (adverbs, complements) can appear;

d)

VOS is derived from VSO by movement of VP2, which after Vmovement to T contains only O: [TP V+T [ V P 2 0 0

[ V P i S [ V P 2 t V t 0 ]]] ];

e)

the basic order in Italian is [ypSVO];

f)

Italian VOS results from SVO, by movement of VP2 over S, which occupies a TP external position: [ F P S [ T P V + T [...0]]

[ [ T P V + T [...O]] 0 [ppS [ T P t t ]]]];

g)

in both languages the VOS forms are derived by a prosodicallymotivated movement (P-movement) 19 , which places the subject in a position to receive nuclear stress and, thus, be interpreted as the focus of the sentence 20 ;

h)

Spanish has no weight constraint like Italian because VOS is obtained by moving a VP that contains only O, while Italian moves the TP, which contains V+T+O.

Contrary to Zubizarreta's positions, this study will assume that: a)

Romance NS languages (RNSL) have uniformly an [yp SVO] order in the base (Kayne 1994);

b)

the surface SVO order results from the visible movement of AGR to T to check nominative Case and by merging a DP (a strong pronoun or definite/specific DP) with TP (see section 2.4), (Kato 1999).

19

Inspired by earlier work by Chomsky (1971) and Jackendoff (1972), Zubizarreta postulates the Focus Prosody Correspondence Principle (FPCP), which states that "the focused constituent (or F-marked constituent) of a phrase must contain the intonation nucleus of that phrase" (p. 38). The effect of P-movement is to have the focalized subject in a position where it can get nuclear stress. According to Zubizarreta, P-movement is not constrained by economy principles like greed. 20 Formally, Zubizarreta defines the place of nuclear stress in terms of Kayne's (1994) LCA (Linear Correspondence Axiom), which correlates linearization with asymmetric ccommand. Her definition of the Nuclear Stress Rule, which applies to the syntactic tree that is the input to Spell-out, is: "Given two nodes Cj and Cj that are metrical sisters, the one lower in the syntactic asymmetric c-command ordering is more prominent" (p. 40).

The Partial Pro-Drop Nature and the Restricted VS Order in Brazilian Portuguese

235

Also contrary to Zubizarreta's analysis, it will be proposed that: c)

the derivation of VOS and of VSO are independent of each other;

d)

VOS with definite subjects is uniformly derived in NS languages from [EpS[ipV[ V p...O]], by moving TP, which contains VO, an analysis proposed by Zubizarreta only for Italian; a different derivation is proposed for indefinite postverbal subjects;

e)

Iberian NSLs (Spanish and European Portuguese) will be proposed to have VSO because they have an extra V- movement to a higher projection, a movement not available for Italian.

2.2. The constrained nature of free inversion with definite postverbal nomináis Sentences with overt 'subjects' in NS languages were shown above to place such 'subjects' outside TP, doubling Agr, and having 'default' nominative case. Let us compare the representation of sentences [7], repeated here as [14], with those of [15]: [14]

a.Yolocomi. I it ate+lpS ('I ate it') b.Juan lo comió. Juan it ate+3psg ('Juan ate it')

[14]'

a'.[ £ P Yo¡

[ T P lo com-íitvp ...]]]

b ' . k p J u a n j [jp lo comi-ó¡ [yp •]]] [15]

a. L o c o m i y o . it ate+lpS I ('I ate it') b.Lo comió Juan, it ate+3pSg Juan ('JUAN ate it')

Free inversion with strong pronouns and definite DPs 21 adjoins the maximal projection TP to EP, a prosodically-motivated movement in the sense of Zubizarreta (1998), so that the nuclear stress falls on Spec of Z, as it becomes the rightmost element, or the deepmost, in Cinque's (1993) terms. The DP in Spec of £ remains in-situ and encodes the 'default' nominative case. P-movement is applied when the subject is focalized in NS languages. Thus the V(0)S pattern

" We will see in the next section that free inversion with indefinite nominals has this nominal in-situ.

Mary Aizawa

236

KATO

has the subject marked [+F], In order to get the V (O) S order in [15], NSLs can move TP upwards, adjoining it to EP 22 : [15]'

a " . [ I P [ T P locom-i

[ V P ...] hp yo [ + F ] [ t T P ]

b" [ip [TP lo comi-6 [ V p ...] [IP Juan ( + F ] [ t T P ] English, and also other NS languages, have no VOS form because, in order to leave the subject in the canonic nuclear stress position at the end of the sentence, T', a non-maximal projection, has to move as can be seen in [16]': [16]

a.John can speak Tagalog. b.* Can speak Tagalog John.

[16]'

a.[TPJohn [T' can [VP, speak Tagalog]]] b.* [ T' [can speak Tagalog [ TP John [T' tj.]]

English has a different strategy to focalize the subject, which is internal to TP: focalization in-situ. This is possible because, according to Zubizarreta (1998: 46): "defocalized constituents are metrically invisible for the NSR (Nuclear stress rule) in German and English". [ 17]

JOHN can speak Tagalog.

This section showed that VOS can be naturally derived in NS languages, but not in non-NS languages. It is also the order uniformly distributed in NS languages, contrary to the VSO order, which is a property of the Iberian Romance, but not of NS languages in general. Therefore, the claim that the structure of VOS in Spanish is derived from VSO structure seems difficult to accept. Zubizarreta's strongest argument to claim that VOS in Spanish is not derived in the same fashion as in Italian has to do with the fact that the latter, but not the former, has weight constraints on the VOS order. In favor of the uniform derivation for the VOS order in Italian and Spanish, we will present below some facts attested to in empirical studies of Spanish, which show that Spanish is also subject to weight constraints in its VOS order. Speaking of VS orders in Spanish, Bentivoglio and Introno (1978: 10) say that they occur mainly with intransitive verbs or with transitive verbs when the object is a clitic. Terker (1984: 276) finds the same sort of facts, saying that: "in fact, any transitive verb

22

We assume with Kayne (1994) that movement is always to the left. But in the GB framework, when there was no such assumption, many interesting proposals were made for the position and case of postposed subjects in Romance. For Portuguese, in particular, see Raposo (1988).

The Partial Pro-Drop Nature and the Restricted VS Order in Brazilian Portuguese

231

with two arguments will freely occur in sentences with VS order as long as the object is a clitic, as in 3) and 4)" ([18] a. and b., in the text): [18]

a.Lo instaló Esteban. it installed Estaban ('Esteban installed it') b.Quería hacerlo Juan, wanted do-it Juan

('Juan wanted to do it')

For Terker, even verbs with three arguments readily permit VS order if all the objects are clitics, as in [19]: [19]

Se lo iba a decir el estudiante. him it went to say the student ('The student was going to tell it to him')

As the effect of the use of clitics is to reduce lexical arguments, it becomes clear that Spanish is also sensitive to weight constraints. Studying word order in Italian, Benincá & Salvi (1988: 125) also show that the presence of clitics favors inversion (see also Rizzi 1991). [20]

a.L'ha mangiata la mamma. it has eaten the mother (The mother has eaten it) b.?Ha mangiato la torta la mamma, has eaten the pie the mother (The mother has eaten the cake)

[21]

a.L'ha letto mio fratello. it has read my brother (My brother has read it) b.?Ha letto il libro mio fratello. has read the book my brother (My brother has read it).

As in both languages the presence of clitics, which makes the clause lighter, facilitates P-movement, we claim that VOS can have a uniform derivation in N S languages.

2.3

The unconstrained nature of free inversion with indefinite nomináis

It has often been pointed out that unaccusative constructions with post-verbal arguments are constrained by a definiteness effect. But a definiteness effect does not seem to be a privileged property of unaccusatives. Benincá & Salvi (1988) show that in Italian, if the subject is indefinite in transitive constructions, the VOS sentences are unmarked. In other words, for Benincá & Salvi, inversion is unmarked if the sentence has a presentational function, or a thetic nature in Kuroda's (1972) sense.

Mary A izawa KA TO

238

[22]

a.Ha mangiato la torta un gatto. has eaten the pie a cat (a cat has eaten a cake) b.Ha letto il libro un bidello. has read the book a janitor (a janitor has read a book)

This shows that either the weight constraint is not operative when the subject is indefinite or that there is no P-movement in these cases. The first hypothesis would weaken Zubizarreta's proposal considerably. Instead, the latter will be proposed to be the case. Thus, we will be distinguishing a)

free inversion derived by P-movement

b)

free inversion without P-movement.

We will be also claiming that unaccusative verbs can have both kinds of inversion in NS-languages. The only difference between unaccusative and unergative verbs is VP internal: the unaccusative has its Agr morpheme merged as its internal argument and the unergative has it as its external argument. In both cases, this affixal argument undergoes head movement to T, where it checks the non-interpretable Case feature of T, which is then erased. [23]

Llegaron. arrived+3pPl ('They arrived')

[23]'

a.

VP / V

\

I llega+past

b. DP

T

I -onl+lmm3pPII

-on.,

VP

V+T

V

DP

[+3pPl]

t

> llegaron

[24]

Llamaron. calIed+3pPl ('They called')

[24]'

a.

Yp DP I on

[+„om,+3pPi

b

-

TP

V

T

I

/

l^ma+past

D on i,3„p//

\ T Hama+past

VP

=> llamaron

DP |

V |

'

"

The Partial Pro-Drop Nature and the Restricted VS Order in Brazilian Portuguese

239

In the previous example Agr was the DP acting as the internal or the external argument of the verb. We may assume that the internal structure of this DP can but need not be more complex. Using Postal's (1969) old idea that pronouns are articles, some linguists assume that Romance accusative clitics take an empty Noun as their complement (Corver and Delfito 1993; Uriagereka 1995; Raposo 1996, 1997). The analysis for subject Agreement adopted here goes in the same direction: the pronominal Agr is the head of a DP with a QP as its complement. While the pronominal affix has a nominative feature that has to be checked, the QP has the 'default' nominative case, which does not require checking and is only relevant at PF. Both have interpretable -features. The representation of this DP can be seen in [25]: [25]

DP D

QP

on

tres hombres

l+nom,+3pPl¡ ¡+3pPI, "default am"/

The DPs in non-null subject languages can only have ordinary determiners in D. But N S languages, for example, can have this structure as the argument of the verbs llegar ('arrive'): [26]

Llegaron

tres mujeres.

arrived+3pPl three women

('There arrived three w o m e n ' )

The derivation of [26] is illustrated in [26]': [26]'

a.

VP

I llega+past

b.

TP

/ \ D on

QP [tres mujeres]

-on¡ V+T l+3pPI/ llega+past

f+nom,+3pPIJ /+3pPI. "de/atih am"/

V

DP

|

llegaron tv

D tj

QP [tres mujeres] [+3pPI, "default" nom]

The indefinite QP is assumed to start with 'default' nominative case and interpretable ((»-features as in [25], The whole DP [DP -on [QP tres mujeres]] merges with the verb as its internal argument (fig. [26]b.). Attracted by the

Mary A izawa KA TO

240

strong V-feature of T, the verb raises to T, yielding [jpllega+past [DP -on /ires clientes]]. The strong D-feature of T attracts the pronominal affix -on and, in the checking process, the case features of both are eliminated. The result is in b. Morphology does the rest, affixing -on to its right position. The QP stays insitu, inside DP, since the "default" case does not require checking. Indefinite NPs are, thus, natural candidates for focalization, since they receive the nuclear stress without undergoing P-movement. The derivation with inergatives is quite similar, except that the DP merges with the verb as its external argument. [27]

Llamaron tres clientes. telephoned+3pPl three clients ('Three clients called')

[27]'

a.[Vp[DP -on [+n0 m, + 3pPI] [QP t r e s clientes]] llama+ past] d.[ T P - o n j [+3PPI] llama+past [ V p [DP [QP tres clientes] ['default'nom] M ]

With transitive verbs, we may propose a similar derivation. The pronominal affix raises to check the features of T and the indefinite nominal stays in-situ, but in Spec of vP. The object moves to the "outer Spec" of vP in a multiple-Spec configuration to check the strong D-features of v, since the subject that originated there, with the 'default' case, does not. The surface order will, therefore, be VOS. [28]

Comio la torta un gato. Ate+3pSg the cake a cat ('A cat ate the cake')

[28]'

[ T P comi-6j [ v P la tortaj [ v P [ D P tj un gato] [ V p tj [ t v [ tj ]]]]]]

In free inversion with definite I-subjects, it was seen that the structure derives from a sentence with a lexical pre-verbal 'subject' placed in the Spec of an extra projection. Movement of TP over this 'subject' was considered P-movement, sensitive to weight constraints. This is not the case of free inversion with indefinite subjects, which were shown to have only pronominal Agreement moving to T, with the lexical subject staying in-situ.

2.4.

Indefinite pre-verbal subjects and the order VSO

Two questions remain to be answered: a) how to derive sentences with preverbal indefinite subjects; and b) how to account for the VSO order.

The Partial Pro-Drop

Nature and the Restricted

VS Order in Brazilian

Portuguese

241

The claim was that Spec of Z was a position for definite and specific nomináis. Phrases like muchas de las mujeres are a possible DP in Spec of I 2 3 , but muchas mujeres is not allowed in such a position. I will also make the strong claim that definite/specific subjects cannot appear as complements of Agr, like the indefinites, as they already contain D. They can only merge as Spec of Z as a full DP. Following Raposo (1994) and Uriagereka (1995), FP (Focus Phrase) is proposed above ZP as a position to host quantified/indefinite nomináis in its Spec or the verb in its head. Let us compare [29]a. and b.: [29]

a.Muchas mujeres; me llamaron;. many women me called ('Many women called me') b.Muchas mujeres amó¡ Juan¡. many women loved +3pSg Juan ('Many women John loved')

[29]'

a.[pp Muchas mujeresj [me llamaro-n¡ [jp ty+j-tvpt li l j 1

]

b.[pp Muchas mujeres [amó¡ [^p Juan¡ [jp .... When a quantified object is preposed, the verb is attracted to the head of F and the subject stays below. If the subject is definite, it is in Spec of Z. This shows that FP is above ZP. For the quantified subject in [29] a, the same position is proposed. It starts as the complement of Agr. Agr raises to check the nominative feature of V+T, and V+T also raises to F. Such movements are of the same nature as those in WH-movement24. [30]

A quién amó Juan? To whom loved Juan ('Who did John love')

[30]'

[cp a quién [amó [ £P Juan [yp

In Raposo (1994), the order VSO results from movement of V+T to the F head, and this is the derivation that will be assumed here: [31 ]

Comió Juan la torta. Ate+3pSg Juan the cake ('John ate the cake')

[31]'

[ F P comió [ip Juan [ T P T a g r + v + t [ v p'a torta [ V p t A g r

23

These are the cases that can take the particle -wa in Japanese. Following Rizzi's (1991) wh-criterion, the verb can be assumed to have the [+F] feature, which is checked against the head F. The quantified nominal raises to its Spec, complying with the F-criterion. 24

Mary A izawa KA TO

242

Like languages that have WH-movement but no movement of the inflected verb, Italian has weak V features in F but has strong D-features, yielding the following structures involving FP: [32]

a.Molte ragazze Gianni ha amato. Many women Gianni has loved b.Molte raggazze hanno amato Gianni. Many women have loved Gianni c.*Molte raggazze ha amato Gianni. d.*Ha amato Gianni molte ragazze.

The structures for [32]a. and b. are: [32]'

a.[pp Molte raggazzej [Sp Gianni [ j p h a + 0 j amato [ v p t| [ tj... b.[pp Molte raggazej [xp hannoj amato [ v p Gianni [ V p tj...

Summing up, the following differences were proposed in the derivation of definite and indefinite VOS sentences: a)

b)

In NS languages, full definite/specific DPs merge as Spec of S, doubling pronominal Agr; indefinite nominals merge with Agr as its complement, forming the subject-argument of the verb; In both indefinite and definite VOS, Agr raises to check nominative in T;

c)

P-movement of TP over Spec of SP, in order to make the nuclear stress fall on the subject, is subject to weight constraints; VOS, which results from checking operations only, namely movement of Agr to T, is an unmarked construction;

d)

Both the DP in Spec of E and the QP as complement of Agr have 'default' nominative case.

The difference in the derivation of definite and indefinite free inversion proposed here can explain why free inversion with indefinite nominals is unconstrained with regard to weight in Italian and Spanish and also why VOS in the latter is also an unmarked construction. It also explains why free inversion with definite I-subjects is often analyzed as being marked. This is due to the fact that markedness has to do with P-movement and not with checking movement. We have shown in this section that unaccusative constructions in Romance NS languages result from the same sort of derivation that is proposed for free inversion with inergative and transitive verbs. The section also analyzed indefinite free inversion as derivationally distinct from free inversion with

The Partial Pro-Drop Nature and the Restricted VS Order in Brazilian Portuguese

243

definite DPs. In the next section we will analyze existential sentences, which present a different pattern from free inversion in Spanish and Portuguese.

3.

Existential constructions in NS languages

3.1.

Existentials and unaccusative constructions: differences and similarities

Unaccusatives and existentials pattern together in many languages. Thus, in English, a non NS-language, unaccusative constructions with presentative verbs {arrive, appear etc.) and existential constructions with be exhibit similar behavior with regard to agreement: the inflected verb agrees with the postverbal nominal, referred to in the generative literature as the expletive associate. [33]

a.There are cats under the table, b.There have arrived many letters.

Italian, a NS language, behaves like English: the inflected verb agrees with the associate, both in existential and in unaccusative constructions. [34]

a.Ci sono dei gatti sotto il tavolo. there are of cats under the table ('There are cats under the table') b.Sono arrivati alcuni uomini. are arrived some men ('There arrived some men')

French unaccusatives and existentials also exhibit similar behavior where agreement is concerned, but, contrary to English and Italian, there is no agreement relation between the verb and the postverbal nominal. What the verb agrees with is the expletive //. [35]

a.Il y a des chats sous la table. expl loc-cl has of cats under the table ('There are cats under the table') b.Il est arrive plusieurs des lettres. expl is arrived many of letters ('There arrived many letters')

Iberian NS languages, on the other hand, present an asymmetry in agreement behavior in these constructions: the unaccusatives are more like English, with agreement holding between the verb and the postverbal nominal, and the existentials are more like French, without agreement between the verb and the postverbal NP/DP. [36]

a.Hay gatos debajo de la mesa. has cats under the table ('There are cats under the table')

Mary Aizawa KATO

244

b.Llegaron muchas cartas. Arrived+3pPl many letters ('There arrived many letters') [37]

a. H i gatos embaixo da mesa. European Portuguese (EP) has cats under the table ('There are cats under the table') b.Chegaram muitas cartas. arrived+3pPl many letters ('There arrived many letters')

3.2.

The case of the expletive and of the associate

In Chomsky (1995), we find a theory that accounts for the variation found between English and Italian, on the one hand, and French, on the other. According to his theory: a)

the verb agrees with the associate if the expletive lacks Case and features 25 {there in English, pro in Italian), but not if the expletive has its own case and «(»-features (// in French);

b)

in the case of English and Italian, since the expletive has no case and features, the features of the verbal complex adjoined to T are checked by the features of the associate;

c)

Spec of TP is projected in the three languages to satisfy the EPP, being occupied by the expletive there in English, by pro in Italian, and by il in French;

d)

the FF of the associate adjoins to there in English and to the expletive pro in Italian, checking their case and «(»-features against V+T 26 ;

e)

in French, il (+nominative, +3S) checks all the relevant features of the complex V+T.

As for the case of the postverbal nominal of existentials, many linguists claim that the have-type auxiliary attributes case (see Roberts 1987; Nunes 1995; Kayne 1993), and Nunes (1995) proposes that it is the accusative case for the expletive of the il type 27 . Cardinaletti (1997) shows, however, that, when the same overt neuter form can be used for both the nominative and the accusative, as in Galician, the verb exhibits agreement with the associate, and, in this case, its case cannot be claimed to be accusative.

25

See Lasnik (1995a & b), for whom there has case, but no ^-features. This is also the analysis proposed by Zubizarreta ( 1998). 27 It should be observed, however, that, if the auxiliary is of the be-type, there is no guarantee that the associate will be nominative, as French has no agreement with unaccusatives and its auxiliary with these verbs is the verb être. 26

The Partial Pro-Drop

Nature and the Restricted

VS Order in Brazilian Portuguese

245

T h e solution for the unaccusative cases is not so clear either. First, it is paradoxical to say that the complement o f the unaccusative verb has accusative case. Secondly, the auxiliary is o f the be-type in French. One way out would be the partitive case proposed by Belletti ( 1 9 8 8 ) and adopted by Lasnik ( 1 9 9 5 ) . T h e other possibility is to say that the case for both the existential and the unaccusative constructions is the 'default' dative case, as both constructions admit a strong dative pronoun when the associate is definite. [38]

a.Il y a lui. Expl loc-cl has him+dat ( ' T h e r e is h i m ' ) b.Il est arrivé

lui

E x p l is arrived him+dat ( ' T h e r e arrived h i m ' ) W e would, in this case, abandon the theory that for have-type constructions the case is accusative. T h e 'default' alternative will be assumed here 2 8 . In the analysis presented here for N S languages, instead o f raising the abstract F F s , what is raised is the pronominal Agr itself. Thus, what happens in an invisible manner in English happens visibly, before spell-out, in N S languages. Checking is obtained, therefore, without the projection o f S p e c o f T P . T h e existential constructions o f N S languages like Spanish and Portuguese are more similar to French existentials and unaccusatives in that the verb lacks agreement with the associate and the auxiliary is o f the have-type in the former. However, as Spanish and Portuguese may have the accusative clitic with the existential verb, the Case o f the associate in these languages will be assumed to be accusative. [39]

a . Y o no creo en brujas, pero que los hay hay. I not believe in witches but that them have have ( ' I do not believe in witches, but they do e x i s t ' ) b.Eu nâo acredito em bruxas, mas que as hâ, hâ.

Instead o f postulating a pro, it is the third person 0-afT\\ itself that is claimed to play the role o f //. B u t though part o f the numeration, this affix has no argument

Of the languages discussed, the only one that cannot be treated uniformly is English, which resorts to the 'default' case only when the associate is definite, in which case there is no agreement. I thank Hagit Borer for having pointed out this to me. 28

(i) (ii) (iii)

There is him. There are them. There is/are a man/some men

(i') *There has arrived him. (ii) *There have arrived them. (iii) There has/have arrived a man/some men.

Maty A izawa KA TO

246

role. The third person 0-afñx Case and ^-features.

adjoins directly to hay/tem (V+T) to check its

Though the agreement pattern is different, existential constructions in Spanish and European Portuguese resemble indefinite free inversion with a singular associate in that the nominal stays in-situ. What distinguishes them is that the element that checks the features of V+T is, for free inversion, a 0 affixal-D that raises from an argument position, and for the existential construction, it is a 0 affixal-D directly adjoined to T. Both are third person singular j0-suffixes bearing nominative Case, but the argumental affix -0\ also has a [± human] component (like in he and she), while the expletive -02 is neuter (like it). If agreement affixes are pronominal in NS languages, such a distinction is conceivable. As the difference between 0 1 and 0 2 can be the manifestation of gender features, the corresponding FF can be [± masc.] for the former and [0masc.] for the latter29. Thus the PF is the same, but the ((»-features are not. Notice that when the gender feature is involved, its contrast becomes visible if there is a predicative adjective. [40]

a.Maria llegó cansada. Maria arrived tired+fem. ('Maria arrived tired') b.Llegó cansada. arrived+3pSg tired+fem. ('She arrived tired') c.Mario llegó cansado. Mario arrived tired+masc. ('Mario arrived tired') d.Llegó cansado. arrived+3pSg tired+masc. ('He arrived tired')

The contrast at the checking point between an existential sentence and an unaccusative construction with a singular associate of the type seen in [36] above, and repeated here as [41], can be seen in [41]': [41]

a.Hay un gato debajo de la mesa. has a cat under the table ('There is a cat under the table') b.Llegó una carta, arrived +3pSg a letter ('There arrived a letter')

[41]'

a.[xp 0 [ + n o m , +3pSg]-hay[+ n 0 m,] [ t v [DP " " gato [+ a c cus]] [•••]] b.TP 0 i [+nom,+3pSg,[-mascl llegó+past[+ n0IT1) ] [t v [DP ['default' nom]]

29

I thank Jairo Nunes for this suggestion.

una carta

The Partial Pro-Drop

3.3.

Nature and the Restricted

VS Order in Brazilian Portuguese

247

The role of the locative

We are now left with the question about the locative there-, if all features are checked by the FF of the associate, why do we need there? In French, the expletive // is inserted and, therefore, Spec of T is projected, but its insertion has to do with feature checking. According to Chomsky (1995), there-insertion would be necessary only to satisfy the EPP in English, which seems in this view to be independently necessary for feature checking. The claim in this paper is that the locative is a meaningful expression30 and that it requires checking. Existential constructions exhibit, in general, a weak locative element even when a language has no strong D-feature in T, as is the case of Italian. Thus, Italian has the clitic ci, and Spanish has the locative incorporated as an affix in the verb haber (ha+y)3'. Old Portuguese used to have a locative pronoun hi/i, but today there is no overt clitic or affix with that function. The proposal in this paper is that weak locatives distribute across languages like the weak personal pronominals. It was seen that weak personal pronouns can be a) free (/ in English), b) a clitic (te in Fiorentino); or c) an affix (-o in Spanish). Likewise, weak locatives are claimed to be: a) a free form (there in English); b) a clitic (y in French); or an affix (y in Spanish). These weak forms can be doubled by PPs. The weak locative forms either raise to Spec of TP (there in English) or adjoin to T as clitics or affixes (y and y in French and Spanish, respectively). Similarly to the doubling structures involving personal pronominals, we can claim that locative pronominals form a DP with the PP complement. What moves to TP is the weak form, leaving the lexical PP in-situ. [42]

a. [op there [pp under the table]] MDP Y

[pp s o u s ' a table]]

c.fop -y

[pp debajo de la mesa]]

The locative element is also proposed to check some abstract feature of the existential verb in V+T. In order to check the formal features of V+T, some feature-carrying element has to be raised. Portuguese has no weak locative, and,

30

This is the core of Nascimento & Kato's (1995) analysis, but this paper modifies their analysis of Case and function of the associate, claimed to be the predicate of the locative in a small clause relation. The associate is proposed to have no case. 31 I thank Evani Viotti for bringing this fact to my attention. However, Zubizarreta pointed out to me that the j> appears only with this form and not with other persons and tense. Spanish could, thus, be proposed to have a null locative as in Portuguese.

Mary A izawa KA TO

248

therefore, no raising of any type occurs. The result of the derivations can be seen in [43]: [43]

a.[xpFFj [ + n o m , + 3 p P i ] + t h e r e j [T table]]]]] b.[TP

are+T [ D P catsj [ D P tj [ P P under the

H[+nom,+3pSgl h" yj + a+T [dp

des

chats [ D P tj [ P P sous la table]]]]]

c.[tp 0 [ + n o m + ,3 p S g | + hayj +T [ D P gatos [ D P tj [ P P debajo de la mesa]]]] d.[jp 0 [ + n o m + ,3 p S g | + hd +T [ D P gatos [ P P embaixo da mesa]]] In the next section the changes that ocurred in BP will be seen against the theory of the NS and inversion proposes so far.

4.

The change(s) in Brazilian Portuguese

According to Kato's (1999) analysis, BP lost the referential null subject 32 , and a weak nominative pronoun appeared in place of the pronominal Agr system, which used to be identified as the grammatical subject. [44] fala- = present tense stem of the verb 'speak' EP and BP before the change

after the change

IpS 2pS 3pSg lpPl 2pPl 3pPl

eu falò cê fala ele/ela fala nós falamos ces falam ele/elas falam

fal-o fala-s fala-0 fala-mos fala-is fala-m

The third person is shown by Duarte (1993, 1995) as still productively null, but Negr3o & Miiller (1996), Figueiredo Silva (1994) and Modesto (this volume) show that many instances of the third person null subject should be analyzed as bound pronouns or variables. What actually remains, then, is the null expletive, analyzed above as the neuter 0 2 . The non-argumental 0 2 is the only affix that can still appear as an independent item in the numeration and be merged with V+T to check its D-features. All the other agreement affixes are now part of the V entry and, in their place, the weak pronouns appear. We will now see how this relates to the loss of inversion. 32

I am assuming that what remains of these referential null subjects is the result of schooling or is data connected to older generations (cf. Duarte 1995).

The Partial Pro-Drop

Nature and the Restricted

KS Order in Brazilian Portuguese

249

As agreement is no longer analyzed as [+pronominal], it cannot appear as an independent item from the verb in the numeration. The categories that can check the case and «(»-features of V+T are the newly formed weak nominative pronouns or a full DP. The weak subject pronouns, similar to the nominative pronouns in English, require not only the projection of Spec of TP, but a strong D-feature in T, yielding the order SVO. [45]

a. Spec of VP before the change

b. Spec of VP after the change

VP

-o

VP

fal(a)-

a \ Output of the derivation Before the change

b'. Output of the derivation After the change TP DP

V+T

ti

falo -o falo

=> falo

Spoken Brazilian Portuguese (BP) retains a productive VS order in unaccusative and existential constructions (examples in [46]), despite its increasing loss of free inversion (examples in [47])33. [46]

33

a.Tem um gato embaixo da mesa. has a cat under the table ('There is a cat under the table')

We exclude here the conservative forms with haver found in written language, where old forms are used for stylistic reasons: (i) Ha gatos embaixo da mesa. has cats under the table ('There are cats under the table') (ii) Haviam chegado muitas cartas. have+3pPl arrived many letters ('There have arrived many letters') Kato (1996b) assigns such language fossils the status of stylistic morphology.

Mary A izawa KA TO

250

b.Tinha chegado muitas cartas. had+3pSg arrived many letters ('There arrived many letters') [47]

a.Respondeu-a/a pergunta a Maria. *BP Ok EP answered+3pSg clitic/the question the Maria ('Mary answered the question') b.Respondeu a pergunta uma aluna. *BP Ok EP answered+3pSg the question a student ('A student answered the question') c.Cantam os passaros/ muitos passaros. *BP Ok EP sing the birds / many birds ('The birds/ many birds sing')

In present-day BP, unaccusative constructions exhibit unified behavior with existentials where agreement is concerned, as can be seen in [46]b. 34 , in which the associate does not agree with the verb. Free inversion with definite subjects was shown to be impossible in English because Spec o f TP is projected. Movement o f the predicate, or the TP, would constitute an illegal operation, since it would move an intermediate projection. The same can be shown to be the case in BP. [48]

Os p£ssaros cantam. Ok EP Ok BP the birds sing+3pPl ('The birds sing')

[49]

Cantam os passaros. sing+3pPl the birds

[48]'

European Portuguese

Ok EP *BP ('THE BIRDS sing') [49]'

European Portuguese

SP DP

TP

os passaros t-FI

34

IP

T

/ \ Agr V + T cantam

VP I tv

Agr V+T cantam

1

t

os passaros [+F]

Franchi et al. (1998) show that even in data taken from interviews with educated Brazilians, agreement may be absent. In colloquial popular BP, lack of agreement seems to be the norm. They also show that ter, compared to haver and existir, is by far the most frequent (50% of the total).

The Partial Pro-Drop Nature and the Restricted KS Order in Brazilian Portuguese

[49]" Brazilian Portuguese

[48]" Brazilian Portuguese TP

•TP

DP os passaros

251

T'

T'

T

VP

I

I

cantam