Peripheral subjects in two varieties of Portuguese: evidence for a non-unified analysis

João Costa (Universidade Nova de Lisboa) & Charlotte Galves (UNICAMP)

1. Introduction.

Both in European Portuguese (EP) and in Brazilian Portuguese (BP), subjects appear to be external. In spite of exhibiting data that confirm the existence of V-to-I movement, the two languages contrast with French (Emonds 1975, Pollock 1989) in allowing the adjacency between subject and verb to be broken by adverbs and floating quantifiers:

(1) a. O Pedro provavelmente viu a Maria.

Pedro probably saw Maria

b. O Pedro viu provavelmente a Maria.

(2) a. Os meninos todos viram a Maria.

the children all saw Maria

b. Os meninos viram todos a Maria.

It is possible to show that this type of behavior does not necessarily derive from an analysis in which preverbal subjects are in a peripheral position with respect to IP, i.e are adjoined to IP or CP, as in Belletti (1990), Barbosa (1995)’s left-dislocation analysis. From such a point of view, the only A-position available for subjects is the VP internal position.

Belletti (1990) argues in favor of such an analysis for sentences like (3a) in Italian, observing that N-words without focal stress cannot precede a sentence adverb. This behavior is not found in Portuguese:

(3) a. *Nessuno probabilmente ha sbagliato. (Italian)

noone probably has failed

b. Ninguém provavelmente falhou. (EP/BP)

noone probably failed

This contrast between Portuguese and Italian argues in favor of an analysis attempting to derive the externality of subjects in Portuguese without necessarily resorting to Left-Dislocation. The goal of this paper is to show that differences between the two varieties of Portuguese provide evidence in favor of accepting the idea that left-dislocation is not the only strategy for deriving external subjects. In particular, it will be argued that, in both varieties of Portuguese, the verb undergoes short-verb-movement, not targeting the highest inflectional nucleus, AgrS. This type of V-movement will be able to derive the non-adjacency between subject and verb, independently of the position of the subject (Spec,IP or left-dislocated).

It will be further defended that the position occupied by preverbal subjects in each variety is different.

EP as well as BP differ from other Romance languages in that they allow for more than one position for pre-verbal subjects. In EP, according to the syntactic context, preverbal subjects can occupy either Spec,IP (Spec,AgrSP), or Spec,TP. BP has the particularity that two pre-verbal subject positions can be projected at the same time: an “internal” one in Spec,TP, and an “external” one, adjoined to IP (AgrSP). This difference in behavior will derive several differences between the two varieties.

2. Evidence for short-V-movement in Portuguese.

Since Emonds (1975) and Pollock (1989), Adverb-placement and the distribution of floating quantifiers are taken as diagnosis to detect whether there is V-to-I movement in a given language. English only has the order Subj-{FQ,Adv}-V-O, which seems to indicate that there is no V-to-I movement. French exhibits subject-verb adjacency, which is taken to indicate that the subject and the inflected verb are in a spec-head configuration, which is the result of movement:

(4) a. John often kisses Mary.

b. *John kisses often Mary.

c. The children all kissed Mary.

d. *The children kissed all Mary.

(5) a. Jean embrasse souvent Marie.

Jean kisses often Marie

b. *Jean souvent embrasse Marie.

Jean often kisses Marie

c. Les enfants embrassent tous Marie.

the children kiss all Marie

d. *Les enfants tous embrassent Marie.

the children all kiss Marie

As shown by Costa (1996) for EP, and by Galves (1994) for BP, in Portuguese, adverbs and floating quantifiers may appear both preverbally and postverbally. Hence, the contrasts that exist in English and French are not visible in this language:

(6) a. O João beija frequentemente a Maria.

João kisses often Maria

b. O João frequentemente beija a Maria.

João often kisses Maria

c. As crianças beijam todas a Maria.

the children kiss all Maria

d. As crianças todas beijam a Maria.

the children all kiss Maria

As mentioned in the introduction, the lack of adjacency between the subject and the verb has been noted for Italian in Belletti (1990). According to this author, it is possible to defend that, when the subject is peripheral, there has been left-dislocation of the subject, since N-words cannot be left-dislocated, and, as expected, cannot surface in the order S-Adv-V, unless they bear a focal stress:

(7) *Nessuno probabilmente ha sbagliato. (Italian)

noone probably has failed

This type of ungrammaticality does not arise in Portuguese, as illustrated in (8):

(8) Ninguém provavelmente falhou.

noone probably failed

This type of behavior seems to indicate that the type of non-adjacency between subject and verb existent in Portuguese is different from the one described by Belletti (1990). Costa (1996) defended that this difference between Portuguese and Italian can be explained if one assumes that there is short-V-movement in Portuguese. According to his analysis, the subject is in Spec,AgrSP, but the verb has moved only up to T, which explains the lack of adjacency between the non-focused N-word and the verb. This analysis also explains the fact that floating quantifiers and adverbs may intervene in between the subject and the verb.

One might think that the type of data presented so far could show that V-to-I movement is optional in Portuguese, rather than indicating that there is short-V-movement. Such an analysis would predict ADV-V or FQ-V orders if the verb does not move, and V-ADV or V-FQ orders if there is V-movement. However, there are reasons to discard such a hypothesis.

First, if one takes a low adverb, such as bem (well), or atentamente (carefully), it is possible to observe that this adverb only surfaces postverbally. If optional V-movement were at stake, one would expect it to optionally appear preverbally, which is not true (cf. Galves 1994 e Figueiredo Silva 1996, for BP):

(9) a. O Pedro leu bem/atentamente o livro.

Pedro read well the book

b. *O Pedro bem/atentamente leu o livro.

Pedro well read the book

Second, it is possible to show that the verb can surface in between two adverbs or in between a floating quantifier and an adverb, as in (10):[1]

(10) ??a. Os meninos todos beijam frequentemente a Maria.

the children all kiss often Maria ??

b. Os meninos frequentemente beijam todos a Maria (EP).

the children often kiss all Maria

c. Os meninos ontem leram bem o livro (EP/BP).

the children yesterday read well the book

Finally, let us consider the reading of adverbs that are ambiguous between a manner and subject-oriented reading, such as inteligentemente (cleverly). As shown in Costa (1997, 1998), the preferred reading in preverbal position is the subject-oriented one, while the postverbal position favors the manner reading:

(11) a. O Pedro inteligentemente leu o livro. (Subject-oriented/*Manner)

Pedro cleverly read the book

b. O Pedro leu inteligentemente o livro. (*Subject-oriented/Manner)

Pedro read cleverly the book

The difference in readings argues against an analysis in terms of optional V-movement, since the reading of the adverb should be the same independently of what happens to the verb. Actually, the distribution of this type of adverb provides an additional argument in favor of the short-V-movement approach, if one compares Portuguese with French (Costa 1998). In this language, the postverbal position is ambiguous between the two readings:

(12) Pierre lit intelligemment le livre. (Subject-oriented/Manner)

Pierre reads cleverly the book

This is expected if the verb moves to a higher position in French than in Portuguese, as illustrated in the partial representations in (13):

(13) a. Portuguese:

[_AgrSP S [_TP Adv_SO [_TP V [_VP Adv_Manner

b. French:

[_AgrSP S V [_TP Adv_SO [_TP t_V [_VP Adv_Manner

Independent evidence for the fact that the two positions available for adverbs are identical in the two languages comes from Williams’ (1994) observation that negation disambiguates the reading in French playing a role similar to the one played by the verb:

(14) a. Pierre ne lit intelligemment pas le livre. (Subject-oriented/*Manner)

Pierre ne reads cleverly not the book

b. Pierre ne lit pas intelligemment le livre. (*Subject-oriented/Manner)

Pierre ne reads not cleverly the book

The ambiguity of the postverbal position in French, that is not found in Portuguese, may be taken as an additional argument for the idea put forward here that the landing site of the verb in Portuguese is lower than in other Romance languages.

Throughout this paper, we will refer to these subjects that may surface in non-adjacency to the inflected verb as external subjects, and will assume the short-V-movement analysis, since it appears to make correct predictions.

3. Differences between EP and BP.

In the previous section, no differentiation was made between the European and the Brazilian varieties of Portuguese, since the arguments for short-V-movement hold in both varieties. In spite of the fact that the two languages have external subjects, EP and BP display several differences that we will be dealing with throughout this paper. The differences to be described in this section have to do with the behavior of preverbal and postverbal subjects.

A. Pronominal doubling.

In BP, but not in EP, pronominal doubling of the subject by a pronoun in preverbal position is quite frequent (Galves 1983, 1987, 1997, 1998, Britto 1998). Although doubling is not impossible in EP, it seems to be available in a very limited set of contexts (cf. Costa 2000). In BP, on the contrary, pronominal doubling is very frequent, and according to some authors, a preferred option in many contexts (see e.g. Kato 1999).

(15) O Pedro, ele telefonou. (BP/??EP)

Pedro, he rang

For (15) above, the contrast is not very obvious, since, as mentioned, there are some contexts in which doubling is possible. Duarte (1995) notes that, in embedded contexts, doubling is also possible in BP. Galves (1998, 2000) points out that this contrasts very strongly with EP, in which a sentence like (16) is ungrammatical.

(16) Eu acho que o povo brasileiro_i ele_i tem uma grave doença. (BP/*EP)

I think that the people brazilian he has troubles

We will thus try to explain why there is a preference for pronominal doubling in BP, and why this is a marked option in EP.

B. Topic-orientation.

BP is a topic-oriented language, in the sense that it allows for a topic to agree with the verb (Galves 1998, 2000). This type of construction is illustrated in (17):

(17) a. O relógio estragou os ponteiros. (BP/*EP)

the clock damaged-3ps/sg the hands

b. Aquele carro furou os pneus.

that car emptied-3ps/sg the tyres

c. A mesa quebrou as pernas.

the table broke-3ps/sg the legs

As already pointed out by Duarte (1987), this type of construction in which the verb agrees with a topic is impossible in European Portuguese.

C. VSO word order.

Like in other null subject Romance languages, VSO word orders are possible in EP, as in (18):

(18) Comeu o Pedro o bolo.

ate Pedro the cake

As mentioned above, pronominal doubling is much more frequent in BP than in EP. Above, we mentioned what happens when a subject is doubled by a pronoun in preverbal position. If the pronoun is postverbal, the results are different. Pronominal doubling in postverbal position is possible in EP, but not in BP, as illustrated in example (19):

(19) O Pedro, leu ele o livro. (*BP/EP)

Pedro, read he the book

This type of word order is impossible in BP. The ungrammaticality of (19) in BP is also a case of VSO order.

In the remainder of the paper, we will try to derive these differences from the behavior of preverbal subjects.

4. Proposal.

As mentioned in the introduction and in section 2, preverbal subjects appear to be external in both varieties of Portuguese. It was also referred that subjects may be external if they are left-dislocated, as proposed in Belletti (1990) for Italian and by Barbosa (1995) for all null subject languages.

We propose the following partition between languages in order to derive the externality of subjects:

(20) A subject is external if:

a. It is left-dislocated.

b. It is in Spec,AgrS, and the verb undergoes short verb movement to T.

c. It is left-dislocated and the verb undergoes short verb movement to T.

Option (20a) is attested in Trentino (Barbosa 1995), as illustrated in (21), and many other languages. Trentino is a good example of this type of option, since the doubling of the external subject by a subject clitic is obligatory.[2]

(21) Maria *(la) parla. (Trentino)

Maria she-cl speaks

We claim that EP is an instantiation of option (20b), while BP is an instantiation of option (20c).

(20c) means that AgrS in BP is a very defective category, which has neither V nor D-features attracting the verb or the subject. This entails that the Extended Projection Principle can only be satisfied in Spec/TP, which, in accordance to what has been argued above, is also a V-related position. This defectiveness is likely to be correlated with the poverty of the verbal paradigm in BP, which ultimately explains the peculiar null subject properties of the language (cf. Galves 1987, Duarte 1995). BP has null subjects, but their behavior indicates that they are no longer identified by the verbal agreement (see Figueiredo Silva 1996, Barbosa, Duarte and Kato 2000, Modesto 2000). As a consequence, Spec,TP tends to be filled by a lexical pronoun.[3]

We have thus the following representations for sentences with pre-verbal subjects in each of the languages:

(22) Trentino, Catalan (Barbosa 1995):

[_IP Subj [_IP (cl_i)V [_VP pro_i t_V ]]]

(23) European Portuguese:

[_AgrSP Subj [_TP t_Subj V [_VP t_Subj t_V ]]]

(24) Brazilian Portuguese:

[_AgrSP Subj [_AgrSP [_TP pro/ele V [_VP t_pro t_V]]]]

It is important to emphasize that the there is a crucial difference between the behavior of subjects in the three options. In the case of Trentino, the A-position of the subject is Spec,VP, preverbal subjects are left-dislocated. In EP, there are several A-positions for the subject (Spec,AgrSP, Spec,VP and Spec,TP).[4] In BP, Spec,TP is obligatorily occupied by a subject-related pronominal category [5], and Spec,VP is not an available surface position for subjects.

5. Analysis of the differences between BP and EP.

The properties described above bring BP very close to languages in which the subject is left-dislocated. In particular, the construction involving pronominal doubling is expected if the preverbal subject is in an adjunction position. If the subject were occupying a clause-internal position, there should be no place for the pronoun, since pronominal doubling of a clause-internal constituent is impossible in Portuguese, as the following examples attest:

(25) a. *Eu não o vi o Pedro.

I not him saw the Pedro

b. *Eu não lhe dei o livro à Maria.

I not to-her gave the book to Maria

The fact that the subject may be doubled by a pronoun seems to be a stronger argument for its external status than the observation that there is no subject-verb adjacency, since, as pointed out above, the lack of adjacency may be independently derived in terms of short-V-movement.

For EP, it is possible to argue that the subject occupies an A-position (for arguments in favor of this claim, see Ambar 1992, Duarte 1987 and Costa 1998, 2000). In particular, the following arguments constitute evidence for this claim:

i) No A-bar minimality effects in embedded questions:

It is possible for a wh-phrase to move to Spec,CP crossing the subject position without yielding A-bar minimality effects in the dialects in which left-dislocation is ruled out in this context:

(26) a. Perguntei quem o Pedro tinha visto.

I asked whom Pedro had seen

b. *Perguntei quem ao Pedro tinha dado o livro.

I asked who to Pedro had given the book

ii) SVO in unmarked context:

In contexts of sentence-focus, the emerging word order is SVO. Since left-dislocation is ungrammatical in this context, it must be the case that the subject occupies an A-position in this context, alledgedly, Spec,IP:

(27) A: What happened?

B: a. O João partiu um copo.

João broke a glass

b. *Partiu o João um copo.

broke João a glass

If the subject were left-dislocated in EP, VSO could be a legitimate answer to the question in (29), since this is a possible word order in this language, and the postverbal position would correspond to the only A-position for the subject.

Returning to the differences between BP and EP, and focusing on pronominal doubling, it is legitimate to assume that BP is closer to languages in which the subject is left-dislocated than EP. There is however one crucial property of BP that is not shared by other languages with left-dislocated subjects, namely topic-orientation. The fact that EP behaves differently from BP, and that BP is not exactly like the other languages in which preverbal subjects are left-dislocated led us to the proposal made above in which there are three possible representations for external subjects.

The externality of the subject in the three groups is derived by the fact that in no language are the subject and the verb in a Spec-Head relation that would induce adjacency. The fact that the subject occupies an A-position in EP derives the contrast with Italian, regarding the placement of N-words, and the evidence for the A-status of preverbal subjects in this language.

Bearing these different representations in mind, let us then turn to the difference between the two varieties of Portuguese, and see how this analysis derives them.

A. Preverbal pronominal doubling:

As we described above, doubling a subject by a nominative pronoun is very common in BP, and may happen both in root and in embedded contexts, as shown in (28), (29) and (30) (cf. Kato 1993, and Galves 2000). Doubling in root context is marginal in EP. In embedded contexts, it is ungrammatical:

(28) O Pedro, ele telefonou. (BP/??EP)

Pedro, he rang

(29) Eu acho que o povo brasileiro_i ele_i tem uma grave doença. (BP/*EP)

I think that the people brazilian he has troubles

(30) Um país em que o presidente, ele não tem mais a confiança do povo, não pode ir para frente

a country in which the president he no longer has the trust of the people cannot go ahead

Although it is commonly associated with categorical judgements (Britto 2000), subject doubling by a pronoun may be found in some out of the blue contexts, as in (31):

(31) O Edmilson, ele está?

Edmilson he is?

‘Is Edmilson there?’

Doubling in an out-of-the-blue context is ungrammatical in EP.

This difference between the two varieties of Portuguese follows straightforwardly from the analysis proposed above. In BP, lexical subjects are adjoined to IP, and there is an internal pre-verbal position If this is the case, it is expected that pronominal doubling is found more often in this variety. We shall argue below that pronominal doubling in BP is a case of hidden clitic-left dislocation.

In EP, on the contrary, the behavior of subjects is not different from the behavior of other categories. Therefore, EP allows pronominal doubling of the subject in the set of contexts in which clitic-left-dislocation of non-subjects is also allowed. For instance, in answers to wh-questions in which there is no assertion regarding the object, this category may be left-dislocated, as in (32) (cf. Ambar 1999):

(32) A: O que é que o Pedro comeu?

what did Pedro eat

B: A sopa…ele comeu-a, o resto…não sei.

the soup…he ate it, the rest…I don’t know

The same happens with the subject if there is no assertion:

(33) A: Quem foi à praia?

who went to the beach

B: O Pedro…ele foi, os outros não sei

Pedro…he went, as for the others…I don’t know.

Note that the possibility for subjects to be left-dislocated in EP is not ruled out by our hypothesis. They may be clitic-left-dislocated, as any other category may be, and in the same contexts where other categories may be. The proposal that subjects are in Spec,AgrSP predicts that the range of contexts in which they are dislocated is more limited than in BP.

B. Topic-orientation:

It was shown above that BP has a kind of construction that does not exist in EP nor in Catalan or Trentino. In this type of construction the verb agrees with a topic, as in the examples repeated from above (Galves 1998, 2000):

(34) a. O relógio estragou os ponteiros. (BP/*EP)

the clock damaged-3ps/sg the hands

b. Aquele carro furou os pneus.

that car emptied-3ps/sg the tyres

c. A mesa quebrou as pernas.

the table broke-3ps/sg the legs

d. A revista xerocou

the journal xeroxed ("the journal was xeroxed")

In our proposal, we predict that there are three types of external subjects: either they are left-dislocated, or there is short-V-movement with subject in Spec,AgrSP, or there is short-V-movement with left-dislocation of the subject. These three options have consequences for the way AgrS is licensed. There is an obvious difference between the case in which the subject-is left-dislocated and the case in which the subject is in Spec,AgrSP and there is short-V-movement, on the one hand, and the case in which there is short-V-movement and the subject is left-dislocated, on the other hand: only in the latter case, which we claim to be the analysis for BP, is the highest A-position (AgrS) radically empty. This can be seen in the structure below for BP:

(35) Brazilian Portuguese:

[_AgrSP Subj [_AgrSP [_TP pro V [_VP t_pro t_V]]]]

These different structures predict that there will be different strategies for identifying AgrS. In languages like Trentino, AgrS will be identified by the verb, which is moved up to this functional nucleus, since there is no evidence for short-V-movement. In EP, there is no movement of the verb until AgrS. Hence, V-movement is not the strategy for identifying this category. However, since the subject is in its specifier, AgrS may be identified by the subject via Spec-head agreement. In BP, AgrS has an empty nucleus and an empty specifier. Thus, neither V-movement nor Spec-head agreement can be employed as strategies for identifying this functional nucleus. Galves (2000) argues that AgrS is identified by pronominal features which move from the argument position co-indexed with the left-dislocated phrase, like hidden clitics.[6]

One important characteristic of the construction exemplified in (34) is the complete loss of the agent argument. This claim is supported by the contrast between (35a) and (35b):

(35) a. A revista foi xerocada para ganhar tempo.

the journal was xeroxed to gain time

b. * A revista xerocou para ganhar tempo.

the journal xeroxed to gain time

In the passive construction (35a), there is an implicit agent which controls the null subject of the purpose clause. No such implicit agent is available in (35b). We conclude from this that there is no Spec/TP projected in these constructions. Their structure will be the one in (36):

(36) [_AgrSP DP_i [_AgrSP AgrS_i [_TP V-f_i [_VP t_V pro_i]]]]

In this structure, since the agreement features of V (f_i) do not enter in a Spec/head agreement relation with Spec/TP , they are free to identify the AgrS features, and agree with the pre-verbal DP. But crucially, this DP cannot be in Spec/AgrP because, if it would, the feature of AgrS would be identified by the DP in its specifier, and the verbal features would remain without identification. Moreover, the pre-verbal DP could not move from the internal domain of a verb in its active form to the external domain of this verb. Finally, no violation of Principle B of Binding occurs in (36) since no subject A-position is instantiated.

The claim that in (34), preverbal DPs are adjoined to AgrSP and the corresponding argumental positions are filled by null pronouns makes these sentences minimally differ from the sentences in which the subject is doubled by a pronoun. The difference between the two constructions reduces to the absence of a specifier for TP in the former.[7] Both of them can only be generated in a language in which AgrS has neither V-features nor D-features. Our claim is that this is what characterizes topic-oriented languages. The fact that such constructions are not legitimate in EP in spite of the fact that EP subjects are external, supports the claim that they are not adjoined to AgrSP but raise to Spec/AgrP.[8]

C. VSO and postverbal pronominal doubling:

The last difference between EP and BP may be shown to follow from the proposed analysis combined with the difference between the two varieties regarding the null subject parameter. As mentioned above, VSO word orders are possible in EP, but not in BP. Likewise, when a subject is doubled by a pronoun in EP, the pronoun may be either preverbal or postverbal:

(37) O Pedro, leu ele o livro. (*BP/EP)

Pedro, read he the book

(38) Comeu o Pedro o bolo. (*BP/EP)

ate Pedro the cake

We bring these two differences together, since both involve cases of postverbal subjects. We relate this difference between the two languages with the null subject parameter. In EP, null subjects are identified by a rich agreement, like in Italian. A well-known fact about BP is that it is less permissive with respect to null subjects than EP. The sentence in (39), from Duarte (2000), is an example of the loss of the ‘avoid pronoun’ principle in BP, typical from non-null subject languages:

(39) Você quando você viaja, você passa a ser turista. Então você passa a fazer coisas que você nunca faria no Brasil.

you when you travel, you become a tourist. So, you start doing things that you would never do in Brazil.

It is claimed above that null subjects in BP are not identified by rich agreement, which was lost (cf. Galves 1994, 2000), but via some other process (cf. references above). Another property of languages in which null subjects are identified by rich agreement on the verb is that subject-verb inversion is possible. This is easily derivable from the fact that the rich verbal agreement satisfies the D-features of the functional category hosting the verb, allowing for the subject to stay in-situ. In languages in which agreement is not rich, some other process must happen in order to check those D-features. In our analysis, the loss of the null subject property is therefore associated with the obligatory presence of a nominal expression in Spec,TP .[9] This explains that no postverbal position for the subject may be used. This will be true irrespectively of the status of the subject as a full DP or as a pronoun. Since EP is a well-behaved null subject language, the subject may be stranded in postverbal positions (Martins 1994, Costa 1996). Hence, inversion with pronouns and full DPs will be possible in this language.

6. Conclusions.

.

In conclusion, although subjects can be characterized as "external" in EP as well as in BP, we have shown that this externality cannot be characterized in an unitary way . Moreover, both BP and EP differ from languages like Italian in which only referential subjects can be external. We have claimed that this is due to the fact that both in EP and BP , the verb does not raise up to AgrS but stays in Tense. In addition, in BP, AgrS have no D-features, and pre-verbal subjects are adjoined to AgrSP. This accounts for the topic-oriented characteristics of this language, which are not present in EP.

References.

Ambar, Manuela. 1992. Para uma sintaxe da inversão sujeito-verbo em português. Colibri,

Lisbon

Ambar, Manuela. 1999. Aspects of the syntax of focus in Portuguese. in G. Rebuschi e L.

Tuller (eds) The Grammar of Focus. John Benjamins

Barbosa, Pilar. 1995. Null Subjects. Doctoral Dissertation, MIT

Barbosa, Pilar, E. Duarte and Mary Kato. 2000. A distribuição do sujeito nulo no português

europeu e no português brasileiro. paper presented at PE/PB Colloquium, Coimbra

Belletti, Adriana. 1990. Generalized Verb Movement. Turin: Rosenber and Telier

Britto, Helena. 2000. Syntactic Codification of Categorical and Thetic Judgements in Brazilian

Portuguese. in M. Kato and E. Negrão (eds) Brazilian Portuguese and the Null

Subject Parameter. Vervuert

Coelho Izete, João Costa, Cristina Figueiredo Silva, Sérgio Menuzzi and Giorgia Brazolla.

2001. Ordem VS e sujeito nulo em PE e PB. Colóquio PE/PB, Fortaleza

Costa, João. 1996. Positions for subjects in European Portuguese. in WCCFL XV Proceedings.

CSLI: Stanford

Costa, João. 1997. On the behavior of adverbs in sentence-final context. in The Linguistic

Review

Costa, João. 1998. Word Order Variation: a constraint-based approach. PhD dissertation,

HIL/Leiden University

Costa, João. 2000. Multiple focus, apparent optionality and subject positions. paper presented at

Workshop on Topic and Focus in Romance, Utrecht

Costa, João. 2001. Spec,TP in European Portuguese conditioned by morphology. ms, Universidade

Nova de Lisboa

Duarte, Eugênia. 1995. A perda do princípio ‘evite pronome’ no português brasileiro. PhD

dissertation, University of Campinas

Duarte, Inês. 1987. A construção de topicalização na gramática do português. PhD

dissertation, Universidade de Lisboa

Emonds, Joseph. 1978. The verbal complex V’-V in French. Linguistic Inquiry 9

Figueiredo Silva, Cristina. 1996. A posição do sujeito no português brasileiro. Unicamp

Galves, Charlotte. 1983. Pronomes e categorias vazias no Português do Brasil. in Cadernos de

Estudos Lingüísticos, 7: 107-136

Galves, Charlotte. 1987. A sintaxe do português brasileiro. in Ensaios Lingüísticos. 13: 31-50

Galves, Charlotte. 1994. V-movement, levels of representation and the structure of S. Letras de

Hoje, 96. Porto Alegre, 35-58

Galves, Charlotte. 1998. Tópicos e sujeitos, pronomes e concordância no Português do Brasil.

in Cadernos de Estudos Lingüísticos, 34: 19-32

Galves, Charlotte. 2000. Agreement, predication and pronouns in the history of Portuguese. in

J. Costa (ed) Portuguese Syntax. New Comparative Studies. Oxford University Press

Kato, Mary. 1993. The distribution of null and pronominal objects in Brazilian Portuguese. in

Linguistic Perspectives on the Romance Languages. Amsterdam: John Benjamins

Martins, Ana Maria. 1994. Os clíticos na história do português. PhD dissertation, Universidade

de Lisboa

Pollock, Jean-Yves. 1989. Verb Movement, Universal Grammar and the Structure of IP. in

Linguistic Inquiry, 24

Rizzi, Luigi. 1986 “Null objects in Italian and the theory of pro”, Linguistic Inquiry, 1986, pp. 501-557, vol. 17.

Williams, Edwin. 1994. A reinterpretation of the evidence for verb movement in French. in D.

Lightfoot and N. Hornstein (eds) Verb Movement. Cambridge University Press

---

[1] Note that (10b) is possible only in EP, which may be due to the different status of FQs in the two varieties, as argued in Galves (1994). For the purposes of the arguments, it is sufficient to note that the FQ may be used to show the lack of adjacency between the verb and both the subject and the object.

[2] Note that ‘la’ being a clitic, it is left-adjoined to Infl, yielding the order S laV.

[3] This does not mean that null subjects do not show up in BP. This is a very complex matter, which goes beyond the limits of this paper. For our present purposes, it is sufficient to say that, since null subjects in BP are not identified by the verbal agreement, theirdistribution is more constrained and, consequently, lexical pronouns show up more frequently.

[4] See Costa (2001) for the availability of Spec,TP being conditioned by the syntax-morphology interface.

[5] We leave aside the constructions in which Spec,TP is filled by a non pronominal subject, and the DP adjoined to AgrSP is co-indexed with a pronoun in another argumental position, as illustrated in (i):

(i) Essa menina o João saiu com ela ontem

This girl João went out with her yesterday

That this type of construction has an underlying structure similar to that of sentences like (15) is supported by the observation that they show no restrictions of embedding , in contrast with what is observed in EP (cf. Duarte 1987).

[6] As pointed out by an anonymous reviewer, this analysis predicts that adverbs show up before the doubling pronoun, like in “O Pedro provavelmente ele viu a Maria” Pedro probably he saw Mary. However, the reverse order “O Pedro ele provavelmente viu a Maria” is not impossible. Interestingly enough, the interpretation of the latter sentence is different , and comparable with what we observed in the EP sentence (37). In this case, we can therefore conclude that ele is adjoined to AgrSP, and o Pedro in a higher position, arguably adjoined to CP, where it receives a contrastive topic interpretation. In this case Spec,TP is filled by pro.

[7] As for the D-features of T, they can be considered to be checked via the agreement relationship between the verbal agreement morpheme (f_i) and AgrS. This makes more sense if, as in Galves (2000), D-features are interpreted not as EPP-features but as Agr-features.

[8] An anonymous reviewer raises the question of how similar kinds of constructions are derived in languages like EP (see for instance “o copo quebrou”, the glass broke, or “o João quebrou a perna” João broke the leg.) The claim that some constructions are syntactically derived in some languages and dependent of lexical properties of verbs in other languages is not new. See for instance Rizzi (1986) on the difference between the licensing of null objects in Italian and in English. Furthermore, the fact that certain constructions are syntactically derived in a given language does not imply that no lexical restrictions apply on those constructions in this language, see again the case of null objects in Italian as analyzed by Rizzi (1986).

[9] See also Barbosa, Duarte and Kato (2000) and Coelho et alii (2001) for a similar proposal.