WH-extractions and relative clauses in Brazilian Portuguese

Extrações-WH e orações relativas no português brasileiro

Abstracts

Este artigo tem por objetivo descrever e buscar uma explicação para o comportamento de extrações de sintagmas QU-- do português brasileiro (PB) para fora de contextos que ficaram conhecidos pelo nome de ilhas, utilizando os princípios da Teoria Gerativa que explicam as propriedades dessas construções nesta e em outras línguas. Argumenta-se que o PB utiliza uma versão da estratégia de concordância em CP, estabelecida por meio de uma relação de predicação, para a extração de sintagmas QU-- a partir da posição de sujeito. A análise proposta para essas extrações evidenciou a existência de dois processos sintáticos diferenciados na construção das orações relativas. Um em que que é um operador que transforma sentenças em predicados, ocupa a posição de núcleo de CP em cujo especificador pode-se encontrar o sintagma que funciona como sujeito da predicação. Outro em que que é um elemento QU--, o que tradicionalmente poderia ser caracterizado como um pronome, que ocupa a posição de especificador de um sintagma quantificado, e introduz as relativas tal como tradicionalmente descritas.

Extrações de Sintagmas-QU; Sentenças Relativas; Predicação

Wh-Extractions and Relative Clauses in Brazilian Portuguese

(Extrações-WH e Orações Relativas no Português Brasileiro)

Esmeralda Vailati NEGRÃO

(Universidade de São Paulo)

ABSTRACT: This paper aims to describe and explain WH-extraction patterns out of island contexts in Brazilian Portuguese (BP), by means of the principles established by Generative Theory. I claim that BP uses a strategy for the extraction of subjects which involves a special case of Agreement. Extractions out of relative islands are possible when the extracted WH-phrase ends up in the specifier position of the higher CP and from there it behaves as the subject of the predication. The subject-predicate relationship established under agreement makes Comp a proper governor for traces in subject position under its scope. The analysis proposed makes a distinction between two processes of relative clause formation. One in which que is an operator that transforms sentences into predicates and sits in the Comp position of a CP whose specifier can be occupied by a QP functioning as the subject of the predication. The other, in which que is a WH-word, traditionally treated as a pronoun, occupying the specifier position of a QP and introducing relative clauses as we know them.

KEY WORDS: WH-Extractions, Relative Clauses, Predication.

RESUMO: Este artigo tem por objetivo descrever e buscar uma explicação para o comportamento de extrações de sintagmas QU¾ do português brasileiro (PB) para fora de contextos que ficaram conhecidos pelo nome de ilhas, utilizando os princípios da Teoria Gerativa que explicam as propriedades dessas construções nesta e em outras línguas. Argumenta-se que o PB utiliza uma versão da estratégia de concordância em CP, estabelecida por meio de uma relação de predicação, para a extração de sintagmas QU¾ a partir da posição de sujeito. A análise proposta para essas extrações evidenciou a existência de dois processos sintáticos diferenciados na construção das orações relativas. Um em que que é um operador que transforma sentenças em predicados, ocupa a posição de núcleo de CP em cujo especificador pode-se encontrar o sintagma que funciona como sujeito da predicação. Outro em que que é um elemento QU¾, o que tradicionalmente poderia ser caracterizado como um pronome, que ocupa a posição de especificador de um sintagma quantificado, e introduz as relativas tal como tradicionalmente descritas.

PALAVRAS-CHAVE: Extrações de Sintagmas-QU, Sentenças Relativas, Predicação.

1. Introduction

One of the ways by which Generative Theory captures the observation that a certain phrase may exhibit behavior compatible with two different syntactic positions is by assuming movement of the phrase in question from one to the other position involved. WH-questions and relative clauses are two examples of such constructions. Treating them in this way made it possible to observe general patterns and asymmetries in the distribution of those phrases within the same language and across languages, which, in turn, are explained by general principles of the grammar.

This paper aims to offer an account of the behavior of WH-question and relative clause extractions in Brazilian Portuguese (BP) by means of the principles already established by the theory as playing a role in explaining the properties of these constructions in other languages. In doing so, the paper will also offer an analysis for relative clauses in this language, showing that work being done on relative clauses in BP have treated two different constructions under the same analysis: restrictive relative clauses and another construction that I will call pseudo-relative clauses, following McCawley (1988).

A theory explaining the behavior of WH-questions and relative clauses through movement has to take into consideration the following questions:

1) what elements move from where?

2) to where do they move?

3) how do they move?

4) why do they move?

In this paper I will address the two first questions.

2. The BP facts

In pursuing an answer for the first question, we have to examine the paradigms involving the structural position occupied at D-structure by the WH-phrase within its clause, plus the kind of clause itself. The paradigm in (1) shows the behavior of WH-phrases when extracted from subject (1a), object (1b) and adjunct (1c) positions out of that-complement clauses:

(1) a. Quem₁ você acha [que t₁ saiu da festa mais cedo]?

who₁ do you think that t₁ left the party earlier

b. Quem₁ você acha [que a Maria encontrou t₁ na festa]?

Who₁ do you think that Maria met t₁ at the party

c. Por que₁ você acha [que a Maria foi até a casa do Pedro t₁] ?

why₁ do you think that Mary went to Pedro's house t₁

With that-complement clauses, extraction from the three positions is grammatical in BP. Unlike English, extraction from the subject can be done even though the complementizer is always overtly present (Complementizer deletion is not allowed in BP).

However, an asymmetry shows up when we try to extract out of WH-complement clauses:

(2) a. Quem₁ a secretária não sabe [se t₁ já deixou o currículo para ser analisado]?
who₁ the secretary do not know whether t₁ left the curriculum to be analysed b. O que₁ a secretária não sabe [se a Cristina deixou t₁ no departamento]?
who₁ the secretary do not know whether Cristina left t₁ in the Department c. *Por que₁ a secretária não sabe [se a Cristina deixou estes documentos no Departamento t₁]?
why₁ the secretary do not know whether Cristina left this documents at the Department t₁ (3) a. Que aluno₁ você não sabe [quando₂ t₁ vai entregar o trabalho t₂ ]?
Which student₁ you do not know when₂ t₁ is going to hand in the paper t₂ b. Que aluno₁ você não sabe [quando₂ o professor vai entrevistar t₁ t₂ ] ?
which student₁ you do not know when₂ the professor is going to interview t₁ t₂ c. *Por que₂ você não sabe [que trabalho₁ o aluno não entregou t₁ t₂ ]?
why₂ you do not know which paper₁ the student did not hand in t₁ t₂

Sentences (2) and (3) show that extraction out of indirect question clauses (WH-islands) is fine when the extracted WH-phrase occupies either the subject ((2a) and (3a)) or the object position ((2b) and (3b)) . However, if it is sitting on an adjunct position, extraction creates an ungrammatical sentence ((2c) and (3c)).

The pattern exhibited by extraction out of complement clauses contrasts with the one exhibited by extraction out of adjunct clauses in (4) and subject clauses in (5):

(4) a. *Quem₁ o João saiu da festa antes que t₁ tivesse entregado o presente?
Who₁ John left of the party before that t₁ had given the present b. *Quem₁ o João saiu da festa antes que a Maria tivesse encontrado t₁?
who₁ John left of the party before that Mary had met t₁ c. *De que modo₁ o João saiu da festa antes que a Maria tivesse tratado o Pedro t₁ ?
in which manner₁ John left of the party before that Mary had treated Peter t₁

(5) a. ?*Quem₁ [que t₁ recebeu doações ilegais para sua campanha] é óbvio?

Who₁ that t₁ received illegal contributions to his campaign is obvious

b. ?*Quem₁ [que a Academia vai escolher t₁ para ganhar o prêmio] é óbvio?

Who₁ that the Academy will choose t₁ to win the prize is obvious

c. *Quando₁ [que a Maria vai entregar o trabalho t₁] é óbvio?

When₁ that Maria will hand in the paper t₁ is obvious

The sentences in (4) and (5) show that in adjunct and subject clauses, the syntactic position occupied by the WH-phrase does not matter. Extraction out of them produces ungrammatical sentences. Nonetheless, if the subject clause is extraposed, the ungrammaticality disappears. WH-extractions out of extraposed subject clauses patterns WH-extractions out of that-complement clauses:

(6) a. Quem₁ é óbvio [que t₁ recebeu doações ilegais para sua campanha]?

Who₁ it is obvious that t₁ received illegal contributions to his campaign

b. Quem₁ é óbvio [que a Academia vai escolher t₁ para ganhar o prêmio]?

Who₁ it is obvious that the Academy will choose t₁ to win the prize

c. Quando₁ é óbvio [que a Maria vai entregar o trabalho t₁]?

When₁ it is obvious that Maria will hand in the paper t₁

It seems interesting to observe the pattern of extractions out of complement and adjunct clauses of nouns, that is, the set of data which in traditional terms was captured by the Complex Noun Phrase Constraint. The sentences in (7) show the pattern of extraction out of sentences which are complements of nouns. The asymmetry "argument versus adjunct" ((7a-b)) vs. (7c)) is also found in the extraction out of a clause which is the complement of a noun:

(7) a. ?Quem₁ o Pedro ficou chocado com a notícia de que t₁ está namorando a Maria?
Who₁ Peter got chocked with the news of that t₁ is dating Mary b. ?Quem₁ o Pedro ficou chocado com a notícia de que a Maria está namorando t₁ ?
who₁ Peter got chocked with the news of that Mary is dating t₁ c. *Como₁ o Pedro ficou chocado com a notícia de que o Pedro tinha morrido t₁ ?
how₁ Peter got chocked with the news of that Peter have died t₁

In the case of extraction out of relative clauses, a different pattern is observed:

(8) a. Que animais₁ o diretor do Zoo disse que a comida₂ que t₁ comeram t₂ estava estragada ?¹
which animals₁ the director of the Zoo said that the food₂ that t₁ ate t₂ was deteriorated b. ???Que animais₂ o diretor do Zoo disse que a comida₁ que t₁ matou t₂ estava estragada?
Which animals₂ the director of the Zoo said that the food₁ that t₁ killed t₂ was deteriorated c. *De que modo₂ o diretor do Zoo disse que os animais₁ que as crianças alimentaram t₁ t₂ eram da África?
In which way₂ the director of the Zoo said that the animals₁ that the children fed t₁ t₂ were from Africa d. *De que modo₁ o diretor do Zoo disse que as crianças₂ que t₂ trataram dos animais t₁ estavam muito felizes
in which way₁ the director of the Zoo said that the children₂ that t₂ treated the animals t₁ were very happy

Extraction of a subject over a relativized object (8a) is perfect, but extraction of an object over a relativized subject (8b) is not as good. Extraction of an adjunct either over a relativized subject (8d) or a relativized object (8c) results in ungrammaticality. However, some further remarks about the data in regard to extraction out of relative clauses is in order.

First, building up the data for testing the extraction possibilities has to be done very carefully in order to control for other variables that may be playing a role in the results obtained. For example, sentences (9) exhibit a different pattern:

(9) a. ??Que comida₁ o diretor do Zoo disse que os animais₂ que t₂ comeram t₁ foram dormir?²
Which food₁ the director of the Zoo said that the animals₂ that t₂ ate t₁ went to sleep b. Que comida₁ o diretor do Zoo disse que os animais₂ que t₂ comeram t₁ passaram mal?
Which food₁ the director of the Zoo said that the animals₂ that t₂ ate t₁ got sick

Sentence (9b) may sound perfect at first, because we can attribute it a different analysis. In as much as anteposed PPs may have their preposition deleted in BP, the WH-phrase que comida 'which food' may be interpreted as an argument of the predicate passaram mal 'got sick', meaning that 'the animals got sick with the food', in which case it has been extracted out of a that-complement clause.

Second, extractions out of relative clauses whose head is in subject position are much better than the ones in which the head is in object position. This claim can be attested if we compare the sentences in (8a-b) with the sentences in (10a-b):

(10) a. ?*Que animais₁ o diretor do Zoo disse que a televisão mostrou as crianças₂ que t₁ atacaram t₂?
Which animals₁ the director of the Zoo said that the television showed the children₂ that t₁ attacked t₂ b. ?*Que animais₁ o diretor do Zoo disse que a televisão mostrou as crianças₂ que t₂ alimentaram t₁?
Which animals₁ the director of the Zoo said that the television showed the children₂ t₂ fed t₁

In sentences (10) the relative clause is attached to the DP as crianças 'the children' which occupies that object position of the complement clause, whereas in sentences (8) the relative clause is attached to the DP a comida 'the food', which is in subject position.

If the grammatical results achieved with extraction from the subject out of that-complement clauses and WH-islands is something that requires some adjustment in the principles of the theory, nevertheless it is a fact already accounted for in regard to a variety of languages. What is unexpected, as far as I know, is the asymmetry between subject extraction versus object extraction out of relative clauses. In this paper, I will propose that the same strategy of subject extraction used by BP to escape that-complement clauses and WH-islands plays a role in subject extraction out of relative clauses.

3. The Literature and the BP Data

Asymmetries of the kind observed in the BP data have been dealt with under the Generative framework by a principle that regulates the licensing of empty categories resulting from movement (traces), namely, the Empty Category Principle (ECP). This principle has received different formulations along the history of the theory. Rizzi (1990) revises Chomsky's (1986) proposal, postulating that ECP could be formulated as comprising a formal licensing and an identification requirement. He ends up proposing that ECP is a principle dealing only with the formal licensing of traces:

(11)ECP: a nonpronominal empty category must be properly head-governed. (p. 87)

The identification requirement of traces is subsumed either under binding of referential chains, where by referential he means a chain formed by segments sharing a referential index attributed at D-structure under theta-role assignment; or under government, a more local relationship. This system gives the possibility of accounting for the two kinds of asymmetries: 1) in terms of its formal licensing, subject traces pattern with adjunct traces against object traces, in as much as for a trace to be properly head governed, it must be governed by a head within its immediate projection, namely, X'; 2) in terms of its identification, subject traces pattern with object traces against adjunct traces, because subjects and objects receive a referential index, in as much as they may receive a referential theta-role, whereas that is not the case with adjuncts.

The extraction pattern showed in the BP data described above may be explained in part by the principles of the theory as already proposed for other languages. For example, the ungrammaticality in the case of subject, object and adjunct extractions out of subject and adjunct clauses fall under the generalization expressed by the Condition on Extraction Domains (CED) proposed by Huang (1982), and as such can be captured by any of the proposals dealing with this kind of extraction in the literature (see Chomsky, 1986). Other than that, the analysis proposed for the Italian data in regard to the grammaticality of object and adjunct extractions out of that-complement clauses, in regard to the grammaticality of object extractions out of WH-islands, and in regard to the ungrammaticality of adjunct extractions out of WH-islands may account for equivalent BP sentences (see Rizzi, 1990). The problem seems to be the grammaticality of subject extractions out of that-complement clauses and WH-islands. This kind of extraction is also possible in Italian, but the attempt to extending to the BP data, the analysis proposed for this kind of extraction in the former language, requires discussion.

Rizzi discusses the symmetrical behavior in regard to subject and object extractions out of that-complement clauses and WH-islands, and claims that languages of the world seem to diverge in terms of the strategies they use to license subject traces. He describes three major strategies used in order to license subject traces. The first consists of transforming C^o into a proper governor when it hosts Agreement features. This is the case of English, which deletes the complementizer to avoid incompatibility with Agr features raised to C^o. This is also the case in French, a language in which the form of the complementizer changes, showing its agreement features. This is also true of V2 languages, in which the inflected verb moves to Comp. The second strategy is to eliminate the gap through the insertion of a resumptive pronoun. This is true of languages such as Swedish and Vata. The third strategy, used by Null Subject languages, consists of extracting the subject from a post verbal position, a strategy available because those languages allow for free inversion of the subject. This is the case of Italian.

In the literature about BP we can find defenders of all three strategies. Moreira da Silva (1983) and more recently Vitral (1992) argued that the empty category is in fact a null resumptive pronoun. Figueiredo Silva (1994), who argues that empty categories in BP result from movement, claims that this language adopts the Agreement strategy, in the case of epistemic verbs, and the extraction from post-verbal position, in other cases. This latter conclusion is a little surprising, because in her dissertation she argues convincingly for the fact that BP, having lost the possibility of assigning Case under government, does not allow free inversion of the subject, a fact also recognized by other Brazilian linguists.

Based on the claims already made in the literature that BP has lost the possibility of free inversion of the subject, therefore it does not count on a post verbal position for subject extractions; and that empty categories in extraction contexts are traces of movement, in as much as they obey islands, as shown in this paper with adjunct islands; in the next section, I will develop the claim that BP uses a strategy for the extraction of subjects which involves a special case of Agreement.

4. The strategy of Subject Extraction in BP

Rizzi (1990) explains the asymmetries in subject traces when extraction took place out of complement clauses in regard to extraction out of subject relative clauses in English by establishing a typology of complementizers in terms of the features [+/¾ wh] and [+/¾ pred(icative)]:

(12) a. +wh -pred: (I wonder) what 0 [ you saw t ]

b. +wh +pred: The thing which 0 [ you saw t ]

c. -wh +pred: The thing Op that [ you saw t ]

d. -wh -pred: (I know) that [ you saw t ] (p. 68)

That in English is incompatible with COMPs having a [+wh] feature. Therefore, if a trace of a WH-moved phrase has to be in COMP, that must be deleted. Relative clauses have the feature [+pred], because they establish with their heads a predication relationship. That can cooccur with the null operator, which has a [+pred] feature, because Op is underspecified for the [+/¾ wh] features. Therefore, that-deletion does not have to take place in relative clauses.

To account for the data regarding subject extractions in BP two assumptions have to be made. First, the complementizer que is underspecified for the [+/¾ wh] features. As a consequence, que can occupy the head of a [+wh] COMP which has a WH-phrase in its specifier position. An evidence for this claim is the fact that in sentence (13), a grammatical sentence, que cooccurs with the WH-phrase quem under the same functional projection:

(13) Quem₁ que João espera encontrar t₁ na festa?

Who₁ that John expects to meet t₁ at the party

Second, que functions as an operator which transforms the clause under its scope into a predicate of the phrase occupying its specifier position, which in turn will function as the subject of the predication. Having established this predication relation under Spec-Head agreement, Comp becomes a proper governor for the trace in subject position.

The use of the agreement in Comp strategy to explain the grammaticality of subject extractions in Portuguese is not new. Zubizarreta (1982) claimed that agreement in Comp is the process taking place in European Portuguese when subject WH-phrases are extracted out of non-factive complement clauses. She says: "We may assume that Portuguese has recourse to the same strategy as French. The only difference is that in Portuguese the result of the morphological rule is phonologically identical to the complementizer. That is, the complementizer que and the trace of the nominative WH-morpheme in Comp rewrite as que." (p. 85-86).

The analysis proposed in this paper differs from Zubizarreta's proposal not only in its motivation and implementation, but also in its scope. The agreement in Comp strategy takes place in a large number of extraction contexts, even in the case of extractions out of relative clauses.

4.1. Agreement in Comp: The Relative Clause Cases

Let's start by investigating two observations already presented: 1) extraction of a subject over a relativized object produced better results than extraction of an object over a relativized subject; 2) extraction out of relative clauses whose head is in subject position are much better than the one in which the head is in object position. The paradigm in (14)-(17) was built to test those observations. Each sentence exemplifies one of the two observations combined. Sentence (14) is an examle in which the head of the relative clause is in the subject posiion of the matrix clause and is anaphorically³ related to the object of the relative clause. In this case WH-movement extracts the subject of the relative:

(14) [Que animais₁] [a comida₂ [que t₁ comeram t₂] estava estragada]?⁴

[which animals₁] [the food₂ [ that t₁ ate t₂ ] was deteriorated ]

(14') A comida₂ [que que animais₁ comeram t₂] estava estragada ?

the food₂ [that which animals₁ ate t₂] was deteriorated

Sentence (15) is an example in which the head of the relative clause is in the subject position of the matrix clause and is anaphorically related to the subject of the relative clause. In this case WH-movement extracts the object of the relative:

(15) *?[Que animais₂] [a comida₁ [que t₁ matou t₂] estava estragada]?

[which animals₂] [ the food₁ [ that t₁ killed t₂ ] was deteriorated]

(15') A comida₁ [que t₁ matou que animais₂ ] estava estragada ?

the food₁ [ that t₁ killed which animals₂ ] was deteriorated

Sentence (16) is an example in which the head of the relative clause is in the object position of the matrix clause and is anaphorically related to the object of the relative clause. In this case WH-movement extracts the subject of the relative:

(16) *?[Que animais₁] [a televisão mostrou as crianças₂ [que t₁ atacaram t₂]]?

[which animals₁] [the television showed the children₂ [that t₁ attacked t₂]]

(16') A televisão mostrou as crianças₂ [que que animais₁ atacaram t₂] ?

the television showed the children₂ [that which animals₁ attacked t₂]

Sentence (17) is an example in which the head of the relative clause is in the object position of the matrix clause and it is anaphorically related to the subject of the relative clause. In this case WH-movement extracts the object of the relative:

(17) *?[Que animais₂] [a televisão mostrou as crianças₁ [que t₁ alimentaram t₂]]?

[which animals₂] [the television showed the children₁ [that t₁ fed t₂ ]]

(17') A televisão mostrou as crianças₁ [que t₁ alimentaram que animais₂]?

the television showed the children₁ [that t₁ fed which animals₂]

The pattern showed in the paradigm (14 ¾ 17) leads to the conclusion that extraction of a subject WH-phrase out of a relative clause whose head is the subject of the matrix clause generates the best results. Nevertheless, this generalization does not seem to hold in sentences (18):

(18) a. *?[Que menina₁] [os alunos₂ [que t₁ convidou t₂ para a festa] se arrependeram]?
[Which girl₁] [the students₂ [ that t₁ invited t₂ to the party] felt sorry] a'. Os alunos₂ [que que menina₁ convidou t₂ para a festa] se arrependeram ?
the students₂ [that which girl₁ invited t₂ to the party] felt sorry b. ??[Que menina₂][os alunos₁ [que t₁ convidaram t₂ para a festa] se arrependeram]?
[which girl₂ ] [the students₁ [that t₁ invited t₂ to the party] felt sorry] b'. Os alunos₁ [que t₁ convidaram que menina₂ para a festa] se arrependeram
the students₁ [that t₁ invited which girl₂ to the party] felt sorry

The judgments in sentences (18) are reversed if compared with sentences (14 ¾ 17). Subject extraction out of a relative clause whose head is the subject of the matrix clause (18a) generates an unacceptable sentence whereas object extraction out of a relative whose head is the subject of the matrix clause (18b) is more acceptable. Things get even more complicated in sentences (19):

(19) a. ??[Que rapaz₁][a comissão premiou o conto₂ [ que t₁ escreveu t₂ ]]?

a'. A comissão premiou o conto₂ [que que rapaz₁ escreveu t₂ ]

b. *[Que comissão₁][o rapaz escreveu o conto₂ [que t₁ premiou t₂ ]]?

b'. O rapaz escreveu o conto₂ [que que comissão₁ premiou t₂ ]

In both sentence (19a) and (19b) the WH-phrase is extracted from the subject position out of a relative clause whose head is the object of the matrix clause. Although both are not perfect, (19a) is acceptable whereas (19b) is ungrammatical. At this point, an observation is in order. Although the resumptive pronoun strategy may be used to improve those sentences, their acceptability varies in the same direction than the acceptability of the sentences in (18) and (19):

(20) a. ?[Que menina₁][os alunos₂ [que ela₁ convidou t₂ para a festa] se arrependeram]?
[which girl₁] [the students₂ [that she₁ invited t₂ to the party] felt sorry] ? b. [Que menina₂][os alunos₁ [que t₁ convidaram ela₂ para a festa] se arrependeram]?
[which girl₂ ] [the students₁ [that t₁ invited her₂ to the party] felt sorry ]? c. [Que rapaz₁][a comissão premiou o conto₂ [ que ele₁ escreveu t₂]]?
[which boy₁][ the committee gave a prize to the short story₂ [that he₁ wrote t₂ ]]? d. ???[Que comissão₁][o rapaz escreveu o conto₂ [que ela₁ premiou t₂]]?
[which committee₁][ the boy wrote the short story₂ [that she₁ gave a prize t₂]]?

The sentences in (20) are evidence for the fact that although BP can build interrogative sentences with resumptive pronouns, this is not the strategy being used for WH-extractions in this language. Moreover, the data presented so far is counterevidence for analyses that claim that it is because BP has resumptive empty pronouns that islands can be violated in this language. If overt or empty resumptive pronouns were the strategy adopted by BP to avoid island effects, all those sentences would have to be equally acceptable.

The examination of all the cases involving subject and object extraction out of relative clauses leads to the conclusion that, in order for the sentence to be acceptable, some kind of relation must hold between the eventualities expressed by the two sentences related through relativization. If we build an example in which the two sentences express independent eventualities, even if we relate both sentences through relativization, WH-extraction will produce an ungrammatical sentence. The comparison between the sentences in (21) with the sentences in (22) and (23), both cases in which the extraction is of a subject WH-phrase out of a relative clause whose head is the subject of the matrix clause, shows a contrast in acceptability:

(21) a. Meu aluno encontrou alguns amigos

my student met some friends

b. Os amigos sabem falar inglês

the friends know to speak English

c. Os amigos₁ [que meu aluno₂ encontrou t₁] sabem falar inglês

the friends₁ [the my student₂ met t₁ ] know to speak English

d.*[Que aluno₂][os amigos₁ [que t₂ encontrou t₁] sabem falar inglês]]?

[which student₂][ the friends₁ [that t₂ met t₁] know to speak English]]?

(22) a. Um motorista desconhecido atropelou uma criança an unknown driver run over a child b. A criança ficou muito machucada the child got very hurt c. A criança₁ [que um motorista desconhecido₂ atropelou t₁] ficou muito machucada the child₁ [that an unknown driver₂ run over t₁] got very hurt d. ?[Que motorista₂][a criança₁ [que t₂ atropelou t₁] ficou muito machucada]]?
[which driver₂][ the driver [ that t₂ run over t₁ ] got very hurt]]?

(23) a. A cozinheira fez a comida

the cook made the food

b. A comida matou os animais

the food killed the animals

c. A comida₁ [que a cozinheira₂ fez t₁] matou os animais

the food₁ [that the cook₂ made t₁] killed the animals

d. [Que cozinheira₂][a comida₁ que t₂ fez t₁] matou os animais]]?

[which cook₂][the food₁ that t₂ made t₁] killed the animals]]?

The contrast between (21), on one hand, and (22) and (23), on the other, is evidence that the relationship between the two eventualities expressed in the matrix and the relative clause makes WH-extractions more acceptable. Moreover, although it is very hard to characterize this relationship between eventualities, it does not seem unsound to claim that in order for this greater acceptability to take place, the eventuality expressed by the VP of the relative clause takes place in a time prior to the time of the event corresponding to the matrix clause. In other cases, it is an antecedent-consequence relationship that seems to hold, in other words, the eventuality expressed by the main clause seems to close the sequence of eventualities.

An observation has to be made in regard to the sentences in (22). Sentence (22d) is ambiguous: it has a first interpretation in which the child run over the driver and end up hurt. Even though pragmatically odd, this interpretation seems to be the preferable one. If we build the structures corresponding to this interpretation, we will see that it is a case of object WH¾ extraction out of a relative clause whose head is the subject of the matrix clause. The second interpretation corresponds to the structure presented in (22d), which is also possible. If so, why is sentence (15), repeated here as (24), much worst?

(24) *?[Que animais₂] [a comida₁ [que t₁ matou t₂] estava estragada]?

[which animals₂] [ the food₁ [ that t₁ killed t₂ ] was deteriorated]

Being a sentence in which an object extraction out of a relative clause whose head is the subject of the matrix clause was performed is not the problem with (24). The unacceptability of (24) comes from the fact that the relationship between eventualities is reversed: killing the animals should be the eventuality closing the sequence. Sentence (25) in contrast with (26) support the claim even further:

(25) a. Os ladrões roubaram o banco

the robbers robbed the bank

b. Os ladrões foram presos

the robbers were put in jail

c. Os ladrões₁ [que t₁ roubaram o banco₂] foram presos

the robbers₁ [that t₁ robbed the bank₂] were put in jail

d. [Que banco₂] [os ladrões₁ [que t₁ roubaram t₂] foram presos]]?

[which bank₂] [the robbers₁ [that t₁ robbed t₂] were put in jail]]?

(26) a. Os ladrões roubaram o banco

the robbers robbed the bank

b. Os ladrões moram em outra cidade

the robbers live in another city

c. Os ladrões₁ [que t₁ roubaram o banco₂] moram em outra cidade

the robbers₁ [that t₁ robbed the bank₂] live in another city

d. ?*[Que banco₂] [os ladrões₁ [que t₁ roubaram t₂] moram em outra cidade]]?

[which bank₂] [the robbers₁ [that t₁ robbed t₂] live in another city]]?

Sentence (26d), in which the eventualities to rob the bank and to live in another city are independent, is ungrammatical, whereas sentence (25d), in which the eventualities to rob the bank and to be put in jail express a cause-effect relationship is perfect.

The contrast between (19a) and (19b), a case of extraction out of relatives whose head is in object position, also reinforce this conclusion. To write a short story and to win a prize are related eventualities and the eventuality closing the sequence is the eventuality expressed by the main clause.

The claim that the unacceptability of (24) comes from the fact that the eventuality corresponding to the consequence should be the closing sequence eventuality is related to one of the properties proposed by McCawley (1988) to characterize a third type of relative clauses: "There is a class of cases in which what appears at first glance to be restrictive relative clauses behave more like the cleft clauses than the restrictive relatives ( ) The apparent relative clauses, which I will henceforth refer to as pseudo-relative clauses, also differ from ordinary restrictive relatives with regard to a constraint ( the Complex NP constraint) that excludes extraction of material from a relative clause construction"(p. 428).

According to McCawley, pseudo-relatives usually occur in the coda of existential sentences. The sentences in (27), McCawley's (34a) and (35a), show the contrast between extraction out of a restrictive relative (27a) and extraction out of a pseudo-relative (27b):

(27) a. *Which books did John praise the person who wrote f?

b. ?Which persons do you think there are many Americans who distrust f?

Other properties associated with pseudo-relatives are: they appear in final position in regard to the sentence in which they are embedded and they can be paraphrased by an existential clause in which the relative is the main clause. Moreover, based on work done by Prince (apud McCawley), he points out that pseudo-relatives are the only relatives that allow deletion of the subject relative pronoun, that is, they allow that-deletion, which is the strategy used by English to render grammatical sentences containing extraction out of a subject position in that-complement clauses. The sentences in (28), his examples (3), are examples of that-deletion in relative clauses:

(28) a. I have a friend f called me yesterday

b. We got a lot of fancy cadillacs f don't tip

McCawley also points out that those sentences correspond to a class of Chinese sentences, studied by Huang (apud McCawley, 1988), which shows up in the coda of existential sentences, in final position, and are not introduced by any particle. They differ from Chinese relative clauses, which precede their heads and are closed by the -de particle.

Brazilian Portuguese has a very used strategy for giving indefinite quantifier phrases wide scope, which correspond to existential sentences having a relative clause in its coda. Following McCawley, we can characterize them as pseudo-relatives. The interesting is that the acceptable sentences exhibiting extraction out of relative clauses may have an existential version, as shown in (29):

(29) a. Tem uns animais que a comida que comeram estava estragada.

there are some animal that the food that they ate was deteriorated

b. Tem uma cozinheira que a comida que fez matou os animais

there is a cook that the food that cooked killed the animals

c. Tem um rapaz que a comissão premiou o conto que escreveu

there is a boy that the committee gave a prize to the short story wrote

The proposal developed is that extraction out of relative islands is possible when the extracted WH-phrase ends up in the specifier position of the higher CP and from there it behaves as the subject of the predication. The subject-predicate relationship established under agreement makes Comp a proper governor for traces in subject position under its scope. Independent eventualities do not allow the establishment of a predication relationship. Consequently, in cases like that, Comp will not be a proper governor for traces in subject position and ECP will be violated. The possibility of having an overt complementizer que 'that' occupying the head position of CP renders visible the predication relationship between the WH-phrase and the sentence transformed into a predicate by the operation of que:

(30) a. [_CP [Que animais₁] [_C' que a comida₂ [que t₁ comeram t₂] estava estragada]?
[which animals₁] [ that the food₂ [ that t₁ ate t₂ ] was deteriorated ] b. [_CP[Que rapaz₁][_C' que a comissão premiou o conto₂ [ que t₁ escreveu t₂]]?
[which boy₁][that the committee gave a prize to the short story₂[that t₁ wrote t₂]]?

The analysis proposed in this article in regard to the strategy used in BP for extracting WH¾ phrases out of islands gives motivation to one of the properties raised by McCawley to characterize pseudo¾ relatives, namely, the property of appearing in final position in regard to the sentence in which they are embedded. This property is recaptured in terms of the need to establish a predication relationship.

The proposal that subject extraction out of relatives is licensed by a predication relationship which establishes agreement in Comp and renders C^o a proper governor has to be extended for the subject extraction cases out of that-complement clauses and WH-islands.

4.2. The Analysis Implementation

Subject WH-Extractions out of that-complement clauses, as in the case of sentence (1a) repeated here as (31) is derived in the following way:

(31) Quem₁ você acha [que t₁ saiu da festa mais cedo]?

who₁ do you think that t₁ left the party earlier

The WH-phrase quem moves to the specifier position of the embedded CP, whose head is occupied by the operator que. A predication relation takes place and consequently, agreement between the referential features of the WH-phrase renders Comp a proper governor for the trace. From there, the WH-phrase moves to Comp in the matrix sentence. The licensing conditions are fulfilled and the interpretation of the empty category is given by binding by an antecedent.

Let's see now the cases of subject extraction out of WH-islands, as in (3a), repeated as (32a) and (32b):

(32) a. Que aluno₁ você não sabe [quando₂ t₁ vai entregar o trabalho t₂ ]?

Which student₁ you do not know when₂ t₁ is going to hand in the paper t2

b. Que moça₁ você não sabe [com que rapaz₂ t₁ saiu t₂ ontem ]?

Which girl₁ you do not know with which boy₂ t₁ went out t₂ last night

In (32a) the WH-phrase que aluno moves first to the specifier (Spec) position of the embedded COMP, leaves a trace there and goes on to the Spec of the matrix COMP. Being in a Spec-Head relation, Agreement takes place and the embedded COMP receives the referential index from the subject trace left there. Consequently, COMP becomes a proper head licenser for the trace in the subject position.

It is time now to address the second question raised in the introduction of this paper, namely, where do WH-Phrases move to? We assumed that WH-phrases are the 'subject' of a predicate created by the operator que. Therefore, they move to the specifier position of the phrase headed by que. However, in (32a-b) we are dealing with two moved WH-phrases. BP is a language that does not allow for two WH-phrases to surface in the same COMP system, as shown in (33):

(33) a. *Que aluno₁ quando₂você não sabe [t₁ vai entregar o trabalho t₂ ]?

Which student₁ when₂ you do not know t₁ is going to hand in the paper t2

b. *Que moça₁ com que rapaz₂ você não sabe [t₁ saiu t₂ ontem ]?

Which girl₁ with which boy₂ you do not know t₁ went out t₂ last night

I do not want to discuss in this paper the internal structure of the CP system. Since we need two specifier positions available for phrases to move to, either we propose that we have CP adjunction or COMP has two specifier positions. So, in sentence (32a) the WH-phrase quando also moves to a specifier position within the COMP system. The trace of the adjunct phrase is locally antecedent governed by the WH-phrase quando, which stays in the embedded COMP, otherwise ungrammaticality would result. The WH-phrase quando being an adjunct, is not theta-marked and does not have a referential index. Although quando is in a specifier position of the embedded COMP system, it does not have a referential index to turn COMP into a proper governor for traces. It is the trace of the subject moved WH-phrase that gives its index to COMP turning it into a proper governor for the trace in subject position. This explains the grammaticality of (32a).

However, things may be more complicated with sentence (32b). In (32b) the object WH-phrase surfaces in the Spec position of the embedded COMP and the subject WH-phrase is in the Spec of the main clause. Since the object WH-phrase is referentially theta-marked , it has a referential index which, in turn, can be transferred to the embedded COMP. The question is how is the trace in subject position properly head governed? Probably we could say that the subject extraction in this sentence is a case of long distance binding (in terms of its interpretation) and having a referential index suffices for a COMP to be a proper head licenser of the trace. Rizzi argues against this hypothesis, requiring that Agreement in COMP takes place when the AGR features of COMP are the subject trace features.

However, another route may be pursued. We can keep the assumption that the subject WH-phrase has first moved to the Spec position of the embedded COMP, being in a Spec-Head agreement relation with que it gives it its index, and then moves on to the matrix COMP. The object WH-phrase also moves to another Spec position of the embedded COMP system, but cannot transfer its index to the whole COMP, because it already has the subject trace's index. At this point of the investigation I do not have evidence to decide which is the best alternative.

The last case to be treated is the subject extraction out of relative clauses, here represented by our very known sentence:

(34)[Que animais₁] [a comida₂ [que t₁ comeram t₂] estava estragada]?

[which animals₁] [the food₂ [ that t₁ ate t₂ ] was deteriorated ]

If we adopt the raising analysis of relative clauses proposed by Kayne (1994), following Vergnaud, we can say that the WH-phrase que comida 'which food' in the object position of the embedded clause, raises to one of the most embedded specifier positions of Comp and the NP comida moves to a specifier position of the quantified phrase introduced by the WH-word que. Then the WH-phrase que animais moves to the higher specifier position, enters in a predication relation, renders C^o a proper governor for the trace, and moves again to the specifier Comp position of the matrix clause. The same proposal made for the cases of extraction out of WH-islands applies here in regard to being the WH-phrase establishing the predication relation with the whole clause the one to transfer the referential features that renders the embedded clause a proper governor for traces. This is why only interrelated eventualities, which in turn allow the establishment of the predication relationship, that generate acceptable sentences.

5. WH-Extractions and Relative Clauses

The analysis proposed made a distinction between two processes of relative clause formation. One in which que is an operator that transforms sentences into predicates and sits in the Comp position of a CP whose specifier can be occupied by a QP functioning as the subject of the predication. It introduces the kind of clause called pseudo-relative, which appears in existential constructions, WH-extraction constructions and probably, cleft sentences. The pseudo-relative is the prototypical construction showing the strategy used by BP to extract subject WH-phrases, namely, agreement in Comp established through predication.

The other, in which que is a WH-word, traditionally treated as a pronoun, occupying the specifier position of a QP and introducing relative clauses as we know them.

Kato (1993) proposes an insightful analysis for relative clauses in BP. She starts reviewing Tarallo's (1983) classical proposal by which BP has three different strategies for relative clause formation: 1) the standard relative, in which the WH-phrase moves from its base position to a specifier position in the CP system; 2) the resumptive pronoun strategy in which que is a complementizer occupying the head position of CP; 3) and the gap-leaving strategy, formed by the ellipsis of the constituent containing the resumptive pronoun. Kato shows that in the three cases raised by Tarallo, the que introducing the clause is a relative pronoun. For her, the difference among them resides in the fact that relativization takes place from the left-dislocated position occupied by the relative pronoun in the case of the last two strategies.

The analysis developed in this article shares Tarallo's idea that que in relative clauses may occupy either the specifier or the head position of CP. It also shares Kato's idea that a predication relation is established between the relative clause and its head, which is similar to the one observed in left-dislocation constructions.

• CHOMSKY, N. (1986) . Barriers Cambridge, Mass.: MIT Press.
• FIGUEIREDO SILVA. M.C. (1994). La position sujet en Portugais Brésilien Thčse de doctorat, Université de Genčve.
• HUANG, C.-T. J. (1982). Logical relations in Chinese and the theory of grammar Ph. D. dissertation, MIT, Cambridge, Mass.
• LOBATO, L.M.P. (1986). Sintaxe gerativa do portuguęs: da teoria padrăo ŕ teoria da regęncia e ligaçăo. Belo Horizonte, Vigília.
• KATO, M. A.(1993). Recontando a história das relativas em uma perspectiva paramétrica. In: I. Roberts & M. Kato (orgs). Portuguęs Brasileiro: uma viagem diacrônica. Campinas: UNICAMP: 223-261.
• KAYNE, R.S. (1994). The Antisymmetry of Syntax Cambridge, Mass.: MIT Press.
• MCCAWLEY, J.D. (1988). The syntactic phenomena of English Chicago: The University of Chicago Press, v.2.
• MOREIRA DA SILVA, S. (1983). Études sur la symétrie et l'asymétrie SUJET/OBJET dans le Portugais du Brésil Doctorat de Troisičme Cycle, Université de Paris VIII, Paris.
• RIZZI, L. (1990). Relativized Minimality Cambridge, Mass.: MIT Press.
• TARALLO, F. (1983). Relativization strategies in Brazilian Portuguese, Ph.D. dissertation, University of Pennsylvania.
• VITRAL, L. (1992). Structure de la proposition et syntaxe du movement du verbe em portugais brésilien Université de Paris VII, tese de doutorado.
• ZUBIZARRETA, M.L. (1982). Theoretical implications of subject extraction in Portuguese. The Linguistic Review, 2: 79-96.

¹

This sentence corresponds to example (30a) used by Lobato (1986:419) to show that extractions out of some islands are possible in BP.

²

The strangeness of sentence (9a), in which the subject is extracted over a relativized object, will be explained later in the text.

³

The word

anaphorically is taken in a very generic sense to express the relationship with an antecedent.

⁴

Sentence (14') represents sentence (14) before WH-extraction. This observation is also valid for all the examples having a (x') counterpart.

Publication Dates

• Publication in this collection
  11 Dec 2001
• Date of issue
  2000