Null Complement Anaphora and Null Objects in Brazilian
Portuguese
*Sonia Maria Lazzarini Cyrino Universidade Estadual de Londrina/CNPq
It is argued in this paper that Brazilian Portuguese (BP) does not show Null Complement Anaphora (NCA) as it has been classically described, since the phenomenon appears not to be a deep anaphor in this language but it appears to be similar to other surface anaphora . It is proposed that in BP the null complement in the NCA-like structures in fact has internal structure, which can be thought of as reconstruction of the antecedent, and the relationship between these structures with the null object in BP will be pointed out.
1. Introduction
Hankamer and Sag (1976) and Sag and Hankamer (1984) propose a distinction between two types of elliptical constructions based on the two ways anaphoric elements can be assigned interpretations: surface and deep anaphora. Deep anaphors have their interpretation based on contextually provided information, even in the absence of a linguistic antecedent. Surface anaphors, on the other hand, require some form of syntactic representation.
According to this distinction, ellipsis phenomena such as VP ellipsis, gapping, stripping, and sluicing have been proposed to belong to the surface anaphora realm while other phenomena as ‘do it’ anaphora and Null Complement Anaphora (NCA) would be deep anaphoric constructions. Williams (1977) and Chao (1987), however, propose that gapping, VP ellipsis, and sluicing should be considered deep anaphora.
In this paper, I propose that Brazilian Portuguese (BP) does not show NCA as it has been classically described, since the construction in this language appears not to behave as deep anaphora, but it is similar to other surface anaphora - it apparently allows extraction, missing antecedents and occurrence in antecedent-contained deletion structures. I will also relate the absence of NCA in BP to the occurrence of null objects and the absence of the neuter clitic o ‘it’.
I will propose that in BP the null complement in the NCA-like structures in fact has internal structure, which can be thought of as reconstruction of the antecedent, that is, the ellipsis site contains a syntactic
*This research has been supported by a grant from CNPq/Brazil (Process 301137/2003-0). I have
profited from presentations of parts of the material contained in this paper at Universidade de São Paulo (USP), Universidade Federal da Bahia (UFBa) and Universidad de Buenos Aires. I also thank an anonymous reviewer for valuable comments. Any errors that remain are of my responsability.
copy of its antecedent that is unpronounced. The proposal will relate the NCA to the null object in BP.
The paper is organized as follows: in section 2, I will present some studies on NCA; in section 3, the question of the existence of NCA in BP is addressed–I review some of the arguments for the distinction between deep and surface anaphora in Spanish and English and its relationship to NCA, and the facts of BP. Based on Cyrino (1997), I present the null object in BP as reconstruction and ellipsis in section 4 and, in section 5, I argue that NCA in BP can be thought of as being a similar phenomenon, hence its surface anaphora status detected in the data presented in section 3.
2. On NCA
There seems to be little disagreement in relation to the deep anaphora status of NCA. Hankamer and Sag (1976) first proposed this classification and Depiante (2000) argues for the deep anaphora character of this construction for English, Spanish and Italian, on the evidence that this null element: a) allows a pragmatic antecedent; b) cannot contain the antecedent of a pronoun (the missing antecedent argument); and c) does not require strict syntactic parallelism with its antecedent.
NCA is an ill-understood process, in which the complement of a predicate is phonologically null:
(1) a. I asked Bill to leave, but he refused ___.
b. Sue was attempting to kiss a gorilla, and Harry didn’t approve ___. c. We needed somebody to carry the oats down to the bin, but
nobody volunteered ___. ((56) in Hankamer and Sag, 1976:411) Besides occurring not only in English but also Italian and in Spanish,1 the
interesting thing about this process is that it is restricted to some predicates (Zubizarreta, 1982, Radford 1977, Bosque 1984, among others -apud Depiante 2000; Brucart 1999); that is, it is lexically determined.
Brucart (1999) lists the predicates with which NCA may occur in Spanish: modals, aspectual verbs, verbs that express predisposition, attitude or purpose, causatives that express permission, collaboration or influence on the attitude of others, and some verbs that govern a prepositional complement. He also states that the NCA is in complementary distribution with the neuter clitic lo ‘it’, that is, verbs that allow lo do not allow NCA (2):
(2) Comenzó a escribir el artículo. Lo comenzó a escribir. *Lo comenzó.
1 And perhaps in French - see note 38 in Cyrino and Reich (2000) and Zribi-Hertz (1986), the latter mention the possibility of ellipsis in French, licensed by aspectual verbs and modals, a phenomenon which seems very similar to NCA. Further study is necessary on French data (see Cyrino, 2003).
‘(He) began to write the article. (He) began to write it. (He) began
it.’ ((92) in Brucart 1999)
Some examples of NCA in Spanish are in (3) below (Brucart 1999:2838-9)–they show that the null complement is an infinitive or subjunctive clause (although this fact has not always been pointed out), whose content is expressed either in a linguistic antecedent or in the pragmatic context:
(3) a. Luis fue al acto; María, en cambio, no pudo ___. ‘Luis went to the act; María, in turn, couldn’t (go).’ b. La invitaron a visitar la exposición y rehusó ____.
‘(They) invited her to visit the exposition and (she) refuse (to visit
the exposition).’
c. Le gusta bailar el charlestón, pero su marido no sabe ___. ‘(She) likes to dance the charleston, but her husband doesn’t know (how to dance the charleston).’
d. El chico estudia porque su madre quiere ___. [ ___ = que estudie] ‘The boy studies because his mother wants (him to study).’
(Bosque,1986:176)
Since Brazilian Portuguese is a Romance language, related to Spanish, Italian and French, we would expect the phenomenon also to occur. In fact, NCA2 is
possible in BP as we may see in the translation of some of the examples above, in (4):
2 BP also has VP ellipsis (i) and other kinds of ellipsis, like stripping (ii), gapping (iii), sluicing, (iv):
(i ) João comprou as maçãs no supermercado, mas a Maria não comprou ____. ‘João bought the apples in the supermarket, but Maria didn’t.’
(ii) a. João deu chocolates para a Maria, e Pedro ___ também. ‘João gave chocolate to Maria, and Peter too.’
b. João deu chocolates para a Maria e ___ flores ____ também. ‘João gave chocolate to Maria and flowers too.’
c. João não deu chocolates para a Maria, mas ____ para Ana. ‘João didn’t give chocolate to Mary , but to Ana.’ (iii) João gosta de cinema e Pedro ____ de shows musicais. ‘João likes cinema and Pedro, musical shows.’
(iv) João foi para algum lugar no verão, mas não sei aonde ____. ‘João went somewhere in the summer, but I don’t know where.’
(4) a. Eu pedi ao João para partir, I ask-past-3sg to-the João to leave
mas ele (se) recusou___.
but he (CL-reflexive) refuse-past-3sg___ ‘I asked John to leave, but he refused.’
b. Luis foi ao ato público; Maria, ao contrário, Luiz go-3sg-past to-the act public Maria to-the contrary
não pôde ___. not can-past-3sg ___
‘Luis went to the public act; Maria , in turn, couldn’t.’ c. Ela gosta de dançar o charleston, She like-pres-3sg of dance-inf the charleston
mas seu marido não sabe ___. but her husband not know-pres-3sg ___
‘She likes to dance the charleston, but her husband doesn’t know (how to dance the charleston/ that she likes to dance the
charleston).’
d. O menino estuda porque sua mãe quer ___. The boy study-pres-3sg because his mother want-pres-3sg___
‘The boy studies because his mother wants (him to study).’
In this paper, however, I question the existence of NCA in BP as it has been classically described in the literature. In what follows I show that the data do not fit the arguments that it is a case of deep anaphora; in fact, we are led to the conclusion that what looks like NCA in BP is a process similar to surface anaphora3.
3. NCA in BP
As far as I know, besides Cyrino (in press) there are no studies about NCA in Portuguese. As a first step towards that investigation, I will assume in this paper that the NCA process as studied for English and Spanish can be applied to Portuguese; therefore, I will look at Depiante’s description of NCA and see how the BP data fit that characterization. In other words, I will temporarily assume NCA exists (or existed, see below) in BP, since the BP allows null complements with the same range of verbs that do so in other languages which present NCA.
3 In this paper, I take VP ellipsis as the “model” for a surface anaphora. Other surface anaphora constructions, e.g. stripping and gapping, also share with VP ellipsis the characteristic properties of this type of anaphora, such as the possibility for missing antecedents and the possibility for operation across a speaker boundary (cf Hankamer & Sag, 1976).
3.1 Deep anaphora status
Depiante (2000) argues that in Spanish (and Italian), NCA is a deep anaphor because it may have a pragmatic antecedent (Surface anaphors only have linguistic antecedents, deep anaphors may have either a pragmatic or a linguistic antecedent):
(5) [Javier jumps into the icy cold sea] Juan says: You también puedo ___!
‘I also can!’
(6) Context: [ Mary is going out to play] Ana says: Io non posso ___
‘I can’t.’
In BP, either a pragmatic (7) or a linguistic antecedent (8) is possible with a “NCA verb”:4
(7) [Maria attempts to stuff a 9-inch ball through a 6-inch hoop. Ana notices this and says]:
A: Eu ajudo ___!
‘I help ___!’
(8) Maria insiste em ir ao teatro Maria insist-pres-3sg in go-inf to-the theater neste final de semana,
in-this end of week
mas sua amiga não concorda ___. but her friend not agree-pres-3sg ___
‘Maria insists on going to the theatre this weekend, but her friend doesn’t agree (on going to the theatre).’
However, as pointed out by Chao (1987:119-120), and noted above, this may not be a conclusive argument for the distinction between deep and surface anaphora, since VP ellipsis, a surface anaphora, may also have a pragmatic antecedent:
(9) [In a situation where both participants hesitate in doing something, as jumping into an icy cold swimming pool]:
A: Don’t __!
Here, the correct interpretation for the gap is given by the context.
Nevertheless, Depiante checks for another property of NCA, namely, the fact that in Spanish NCA lacks strict parallelism as any deep anaphor:
4 A “NCA verb” in this paper is a verb which allows a NCA construction. See Depiante (2000) and Cyrino (in press) for a list of those verbs.
(10) Juan las quiere ver Juan them(cl,fem,pl) want-pres-3sg see
y María también quiere ___ . and María also want-pres-3sg ___
‘Juan wants to see them and María also wants (to see them).’ ((2c) in Depiante 2000:23) Here the first conjunct has undergone restructuring (clitic climbing has occurred), but the same does not happen in the second conjunct, that is, there is no strict parallelism and the sentence is grammatical. This would argue for the deep anaphora status of NCA in Spanish.
However, we also have a problem with this argument: a surface anaphor such as VP ellipsis may also lack strict parallelism, as we can see in sentence (11):
(11) John has always liked Mary, but Peter never will ___
Therefore, this piece of evidence cannot be conclusive for the status of NCA as a deep anaphor. As in Spanish, in BP, there is no need for strict parallelism in “NCA constructions”, as we can see in (12):
(12) Os pacientes do terceiro andar devem ser The-pl patients of-the third floor must-pres-3pl be removidos ainda que a enfermeira mais forte removed-pl still that the nurse more strong
não consiga ___.
not manage ___
‘The patients in the third floor must be removed even if the strongest nurse doesn’t manage to (remove them).’
The missing complement is removê-los ‘remove them’, but the antecedent of the gap is ser removidos ‘be removed’, a non-parallel structure. But, in view of what was said about VP ellipsis and strict parallelism, this cannot be taken as evidence for a deep anaphora status of NCA in BP.
Although Depiante used both arguments to show the deep anaphora status of NCA in Spanish, we saw that they are not exclusive of this phenomenon – they apply both to deep and surface anaphora as seen above (assuming VP ellipsis is surface anaphora). We cannot rely on them for the distinction.
However, Depiante also provides a strong argument, the “missing antecedent” test, whereby only surface anaphors can contain an element that can serve as antecedent to a pronoun, contrary to what occurs with deep anaphors: (13) ?? Juan no pudo asesinar a Pablo com um cuchillo
Juan not can-past-3sg kill to Pablo with a knife pero Pedro si pudo ___ y pro estaba oxidado. but Pedro yes could ___ and pro was rusty
‘Juan couldn’t kill Pablo with a knife, but Pedro could and (it) was
In this example, as expected for Spanish/English, the NCA, a deep anaphor, does not contain an antecedent for the null pronoun (pro) in any point in the derivation (deep anaphors have no internal structure). Compare (13) to (14), where we see that there is an overt antecedent for the null pronoun and to (15), where we have a VP ellipsis structure, which should have internal structure at the ellipsis site.
(14) Juan no pudo asesinar a Pablo con un cuchillo pero Pedro sí pudo asesinar a Pablo con un cuchillo y pro estaba oxidado.
(15) Jack couldn’t kill Peter with a knife, but John could, and it was rusty.
However, this is not what happens in BP. Sentences like (13), whose ungrammaticality both in English and in Spanish provided an argument for the deep anaphora status of the NCA, are grammatical5 in BP, as shown in (16)6.
With this evidence, it looks like NCA does not behave like deep anaphora in BP. (16) João não pôde assassinar Pedro com uma faca, João not can-past-3sg assassinate-inf Pedro with a knife
mas Mário pôde ___e ela estava enferrujada. but Mario can-past-3sg ___ and it was rusty
‘João couldn’t kill Pedro with a knife, but Mario could, and (it)
was rusty.’
We may argue that this is a VP ellipsis structure in BP because of the identity of the verbs, since that is a characteristic of VP ellipsis in Portuguese (see Matos 1992, and below). But changing the verb also renders the sentence grammatical, cf. (17):
(17) João não pôde assassinar Pedro com uma faca, João not can-past-3sg assassinate-inf Pedro with a knife
mas Mário conseguiu ___e ela estava enferrujada. but Mario manage-past-3sg __and it was rusty. ‘João couldn’t kill Pedro with a knife, but Mario succeeded ___, and it was rusty.’
3.2 Lack of internal structure
Depiante presents a series of arguments – clitic climbing, auxiliary change phenomenon, long object preposing – to point out the lack of internal structure in NCA constructions in Spanish and English.
5 This happens with VP ellipsis as well (see Zocca 2003:15).
6 As pointed out by a reviewer, sentences (15) and (16) also differ from their Spanish counterpart (13) by having an overt pronoun, while in Spanish we have pro. This is an additional difference between Spanish on one side and English and BP on the other in this example, but the argument here is for the possibility of missing antecedents in surface anaphora, as VP ellipsis in English and, as will be argued below, NCA in BP.
However, in the face of the absence of these phenomena in BP, the only test which is possible to be applied to BP is the extraction test. Depiante proposes that in Spanish we cannot move a constituent out of a NCA, due to the fact that there is no internal structure in the NCA, so that constituent is never there at any point of the derivation:
(18) * Juan sabe qué libro Maria quiere/puede Juan know-pres-3sg which book Maria want/can-pres-3sg leer y Pedro sabe
read-inf and Pedro know-pres-3sg
qué revista Ana quiere/puede ___. which magazine Ana want/can-pres-3sg ___
‘Juan knows which book Maria wants/can read and Pedro knows which magazine Ana wants/can (read).’
Another argument for this position comes from the ungrammaticality of the following sentence in English:
(19) * This novel, Bill agreed to read and this biography, Peter agreed __. The ungrammaticality of these sentences is explained if we assume that there is no internal structure in the null complement, so the wh-phrase and the topicalized NP have no trace to bind.7
In BP, however, it is possible to extract out of a NCA in BP, contrary to what happens in Spanish:
(20) a. O João sabe que livro a Maria pode
The João know-pres-3sg which book the Maria can-pres-3sg ler e o Pedro sabe que revista
read-inf and the Pedro know-pres-3sgwhich magazine a Ana pode ___.
the Ana can-pres-3sg ___
‘João knows which book Maria can read and Pedro knows which
magazine Ana can (read).’
b. Esta novela, o João concordou em ler This novel the João agree-past-3sg in read-inf e este conto o Pedro concordou ___. and this short-story the Pedro agree-past-3sg ___
‘This novel, John agreed to read and this short-story Pedro agreed
(to read).’
One may argue again that these are VP ellipsis structures in BP (the verbs in both clauses are the same), and, therefore, there is internal structure (see Zocca
7 That does not happen, for example, in VP ellipsis, since, according to this line of argument, there is internal structure in that case (Depiante 2000:28):
2003:16). But if we change the verbs in the examples (the examples above were translated from Depiante’s Spanish examples), it is still possible to extract out of a NCA-like structure in BP:
(21) O João sabe que livro a Maria quis
The João know-pres-3sg which book the Maria want-past-3sg ler e o Pedro sabe que revista
read-inf and the Pedro know-pres-3sg. Which magazine a Ana recusou-se ___.
the Ana refuse-past-3sg-CLreflexive ___
‘João knows which book Maria wanted to read and Pedro know which magazine Ana refused (to read).’
(22) Esta novela, o João começou a escrever, This novel the João begin-past-3sg to write-inf mas este conto, ele nunca acabou ___. but this short-story he never finish-past-3sg ___
‘This novel, João began to write, but this short story, he neve
finished (to write)’
3.3 On other characteristics of NCA
NCA cannot appear in Antecedent Contained Deletion (ACD) structures (23) in English (see Depiante, 2000):
(23) a. * I always eat anything that he volunteers ___ b. * I refused to do everything that he refused ___
This is a consequence of its lack of internal structure, as it is a deep anaphor in her analysis. In contrast, VP ellipsis, a surface anaphor, allows ACD:
(24) a. I always eat anything that he does. b. I refused to do everything that he did.
However, NCA does allow sloppy identity readings (25), and in this aspect it resembles a surface anaphora, like VP ellipsis (26):
(25) John refused to talk to his mother and Peter also refused ___ (26) John saw his mother, and Peter did ___ too.
The possibility for sloppy reading is usually taken as a test for surface anaphors. Since NCA allows sloppy readings, Depiante (following Hankamer and Sag), who we saw proposed NCA as an instance of deep anaphora, is forced to conclude that “sloppy identity is not a necessary or sufficient condition for
determining whether a certain element constitutes surface anaphora or not” (Depiante, 2000: 34)8.
Depiante (2000) proposes that NCA is represented in the syntax (contrary to Grimshaw, 1979) as a null pro-form with no internal structure. As such, it can be interpreted as a proposition, a property or a question, and it is analogous to a pronoun, getting its reference from some salient individual in the linguistic or in the non-linguistic context. Therefore, it cannot serve as an antecedent for a pronoun, and the lack of missing antecedents is explained. NCA would be the null counterpart of a sentential pro-form as ‘it’ or ‘so’ in English and lo in Spanish (27). She proposes the generalization in (28):
(27) Mary believes that Anne is pregnant but I don’t believe it/so.
((42)in Depiante2000: 43)
(28) A predicate that selects NCA cannot take an over predicative/propositional pro-form.
As for BP, contrary to what happens in English, ACD is allowed in a NCA construction (29), as well as in VP ellipsis constructions (30):
(29) a. Eu sempre digo qualquer coisa always say-pres-1sg any thing
a que o João se negue ___. to that the João CL-reflexive refuse-subjunct.-3sg ___
(João se nega a dizer algo em particular e eu digo isso) (João refuses to say something in particular and I say it) ‘I always say anything that João refuses (to say).’
b. Eu me recuso a fazer
I CL-reflexive refuse-pres-1sg to do-inf
qualquer coisa com que ele concorde ___. any thing with that he agree-subjunct.-3sg ___ ‘I refuse to do anything with which he agrees (to do).’ (30) a. Eu me nego a dizer
I CL-reflexive refuse-pres-1sg to say-inf
qualquer coisa a que ele se negue ___. any thing to that he CL-reflexive refuse-subjunct ___ ‘I refuse to say anything that he does.’
8 Andrés Saab (personal communication) points out that Spanish does allow NCA in ACD structures as in BP (see below). Both facts, NCA in ACD and the availability of sloppy readings, are arguments
b. Eu me recuso a fazer I CL-reflexive refuse-pres-1sg to do-inf
qualquer coisa a que ele se recuse___. any thing to that he CL-reflexive refuse ___ ‘I refuse to do anything that he does.’
Moreover, along with English and Spanish, BP also allows sloppy readings with NCA constructions (besides the strict reading), in sentences like (31):
(31) José se recusou a conversar José CL-reflexive refuse-past-3sg to talk-inf com sua mãe, mas Pedro concordou ___. with his mother but Pedro agree-past-3sg ___ ‘José refused to talk to his mother, but Pedroi agreed (to talk to hisi mother).’
Since the tests which can be applied in BP show that the NCA has internal structure, we will have to say that it appears not to be a deep anaphor in this language, but it is similar to other surface anaphora - it apparently allows extraction, missing antecedents and occurrence in antecedent-contained deletion structures, besides the availability of sloppy reading.
Additionally, we have another difficulty in identifying NCA constructions in BP, if we are to follow Depiante (2000)’s and Brucart (1999)’s complementary distribution proposal that verbs which allow NCA do not allow the clitic lo.
According to the Spanish data, the verbs querer and saber are NCA verbs; however, contrary to the complementary distribution proposal, these verbs may have lo as a complement, as shown by an example with the verb saber, in Depiante, 2000:52 (her example (57c), my emphasis):
(32) Los pacientes del tercero tienen que ser llevados al segundo piso, aunque el director del hospital no lo sepa.
‘The patients of the third (floor)m have to be taken to the second floor, even though the hospital’s director doesn’t know it.’
So the problem with the complementary distribution proposal is that it is not reliable for the identification of “NCA verbs”.
Besides that, in some examples, if we have the clitic lo, the interpretation of the clitic will be different from the NCA interpretation. For instance, notice example (3c), repeated here as (33). If there is ellipsis (NCA), the interpretation is as given in the gloss:
(33) Le gusta bailar el charlestón, pero su marido no sabe ___. ( ___ = bailar el charlestón)
‘(She) likes to dance the Charleston, but her husband doesn’t know (how to dance the charleston).’
If lo is used in this sentence (cf. Le gusta bailar el charlestón, pero su marido no lo sabe), the interpretation of the clitic will be the proposition “she likes to
dance the Charleston”, that is, her husband doesn’t know her likes on dancing matter.9 This is an additional complication if we are to rely on the
“complementary distribution” criterion.
For European Portuguese, Matos (2003) proposes that the NCA is the gap corresponding to a propositional complement, the clitic o ‘it’, or the demonstrative pronoun isso ‘this’, which, in this context, would denote a given situation, not an entity—the distribution of o/isso would be lexically determined. Anyway, we can not rely on the complementary distribution of o in order to identify the NCA in BP because this neuter clitic is not used anymore– therefore, it is very difficult to distinguish between the verbs which could allow (or, have once allowed) the clitic and the verbs that disallow it, in order to point out which are the NCA verbs.
I suggest, as a first step in this investigation (see Cyrino, 2003), that the only way we could identify NCA in BP is by relying not only on the classification of verbs given by Depiante (2000) and Brucart (1999) but also on the nature of the gap, that is, if it corresponds to an infinitive or a subjunctive clause. This fact is not focused in Depiante’s work: she proposes the NCA as a null pro-form interpreted as a proposition, a property or a question.
But, as we shall see below, perhaps this identification is not even necessary–I will propose that we subsume the NCA-like structures in BP into the range of possible null complements in the language. Another phenomenon in this range is the null object.
4. The Null Object in Brazilian Portuguese
As is well known, BP exhibits null objects in any syntactic context, as opposed to other languages, for example, European Portuguese (EP). According to Raposo (1986), EP does not allow null objects in islands—sentences like (34) or (35) are ungrammatical in EP, but grammatical in BP:10
(34) Comprei o casaco depois que experimentei ___. Buy-past-1sg the coat after that try_on-past-1sg ___. ‘I bought the coat, after I tried (it) on.’
9 Zribi-Hertz (1986) in note 3, chapter XII, distinguishes modal “savoir” (quasi-synonym of “pouvoir”), ex. “savoir jouer du piano” which allows null complements, from “savoir” which governs an objective complement (“savoir qu’on joue du piano”). In example (33), if we use the clitic lo, it looks like the verb saber is no longer a modal, but a lexical transitive verb.
10 For a comprehensive review on the null object in BP, see Cyrino (1997), Cyrino and Reich (2001), and references therein.
(35) O rapaz que trouxe ___agora mesmo The boy that bring-past-3sg ___now just da pastelaria era o teu afilhado. of-the pastry shop was the your godson
‘The boy that brought (it) just now from the pastry shop was your godson.’
Although the null object occurs quite freely in BP, there are constraints on its use according to the semantic features of the antecedent. These constraints involve the interplay between animacy and specificity: an animate antecedent never occurs as a null category (see (36)), unless it is non-specific (see (37b)): (36) A Clara não quer que o filho veja TV, The Clara not want-pres-3sg that the son watch-subjun-3sg TV
então ela sempre leva *___/ele no parquinho. so she always take-pres-3sg *___/him in-the park-little
‘Clara doesn’t want her son to watch TV, so she always takes (him) to the playground.’
(37) a. O policial insultou o preso The policeman insult-past-3sg the prisoner
antes de torturar *___/ele.
before of torture-inf *___/him
‘The policeman insulted the prisoner before torturing (him).’ b. O policial insulta presos
The policeman insult-pres-3sg prisoners
antes de torturar ___/eles.
before of torture-inf ___/them
‘The policeman insults prisoners before torturing (them).’
The default null object cases replace [- animate] antecedents, as in (38). It should be noticed that the pronominal realization of an object is guaranteed through the use of strong pronouns, since 3rd person clitics are no longer available in the language.
(38) O Pedro perdeu a carteira e The Pedro loose-past-3sg the wallet and
não consegue achar ___/ela em lugar nenhum. not can-pres-3sg find-inf ___/ it in place none ‘Pedro lost his wallet and can’t find (it) anywhere.’
A striking effect of the null object is that it makes strict and sloppy readings available, as seen in (39), whereas the strong pronoun makes the ambiguity go away, leaving only the strict interpretation option, as in (40):
(39) De noite, João liga seu aparelho de som, At night, João turn-on-pres-3sg his system of sound
mas Pedro desliga ___.
but Pedro turn-off-pres-3sg ___
‘At night, João turns on his sound system, but Pedro turns (it) off.’ (40) De noite, João liga seu aparelho de som, At night, João turn-on-pres-3sg his system of sound
mas Pedro desliga ele. but Pedro turn-off-pres-3sg it
‘At night, João turns on his sound system, but Pedro turns it off.’ In (39) we have two interpretations: Pedro turns off his own sound system or he turns off João’s sound system, while in (40) only the latter interpretation is possible.
Some researches have proposed that the null object in BP is an empty pronoun, pro, but the recurrent problem with these proposals is that there is no agreement on the requirements of identification and licensing of the empty category (cf. Farrell, 1990; Galves, 1987, 1989; Kato 1993; Bianchi and Figueiredo, 1994; Kato 2000; Barra Ferreira, 2000, among others). Besides, these proposals are unable to capture the animacy constraints we have discussed, invariably offering a stipulation on this aspect of the null object. In order to fully appreciate its effects, a comparison between (41a) and (41b) can be helpful: (41) a. Eu nunca vejo o meu pai. .
I never see-pres-1sg he my father. Não me lembro d[a cara dele]i Not CL-reflexive remember of [the face of-he]i . Vou esquecer ___i
Go-pres-1sg forget-inf ___i
‘I never see my father. I don’t remember [his face]. I’m going to
forget (it).’
b. Eu nunca vejo [o meu pai]i. I never see [the my father ]i
Não me lembro da cara dele. Not CL-reflexive remember of-the face of-he. Vou esquecer elei
Go-pres-1sg forget-inf himi
‘I never see my father. I don’t remember his face. I’m going to
forget him.’
The only possible interpretation for the null in (41a) is the [- animate] his face, contrary to the one in (41b) where the use of the strong pronoun leads to the interpretation of the [+ animate] my father.11
11 Examples from Schwenter and Silva (2002).
I depart from the proposals that treat the null object as pro, not only because they cannot satisfactorily explain the facts, but also because they may be hard to maintain if we consider diachronic facts and acquisition data (on the latter, see Cyrino and Lopes, 2004). In Cyrino (1997), I propose that the null object is the result of ellipsis, after analyzing diachronic data.
The analysis follows the work on VP ellipsis constructions in Fiengo and May (1994), who propose that ellipses are subject to a more general principle at LF, reconstruction of the antecedent, together with inaudibility (ellipsis) at PF. The null object in BP, according to Cyrino (1997), is a similar phenomenon—it is a nominal ellipsis.
Fiengo and May also propose a Dependency Theory that may be integrated in their theory of reconstruction and thus account for the strict and sloppy reading phenomena present in ellipsis constructions. They propose that the strict or sloppy reading is a consequence of the type of occurrence of a pronoun, that is, pronouns may have independent or dependent occurrences. A pronoun is an alpha occurrence if it is independent of an antecedent in the same phrase marker, whereas it is a beta occurrence if it is dependent of an antecedent. If the occurrence is independent, reconstruction copies the occurrence of the index. If it is dependent, reconstruction copies the dependency. Reference for α-occurrences is established independently for each occurrence, even if they are coindexed. β-occurrences are indexical dependencies, being well formed if there is another occurrence with the same index value that they can depend on, thus, a pronoun with a β-type index gets its reference from the element it is connected to.
(42) John1α told his1β wife2α that she2β is beautiful. <[NP John]α, [NP his]β, 1>
<[NP his wife]α, [NP she]β, 2>
The analysis for the null object in BP as ellipsis comes up due to historical facts, but also due to the possibilities for the interpretation of the empty category. Cyrino (1997) assumes that some pronouns which have low semantic value, such as it in English and the neuter clitic o in Portuguese can also be thought of as reconstruction at LF, and, because of that, they can be null in languages which allow nominal ellipsis. Such a pronoun depends on its antecedent for the interpretation of its contents, as we can see in paycheck sentences like (43) for the pronoun it, or sentences like (44) with the neuter clitic in Portuguese, in which we can have strict and sloppy readings:12
(43) The man who gave his paycheck to his wife was wiser than the man who gave it to his mistress.
(44) Pedro pediu para ser o professor da turma Y, Pedro ask-past-3sg to be-inf the teacher of-the class Y,
antes de Jane o solicitar.
12 For a distinction between VP-ellipsis and cases of null objects in Portuguese, see Cyrino and Matos (2002).
before of Jane it solicit
'Pedro asked to be the teacher of class Y, before Jane asked for it.’ Sentence (44) sounds old fashioned in BP. Its current counterpart would be (45), with a null complement:
(45) Pedro pediu para ser o professor da turma Y, Pedro ask-past-3sg to be-inf the teacher of-the class Y,
antes de Jane solicitar ___. before of Jane solicit ___
'Pedro asked to be the teacher of class Y, before Jane asked for it.’ Both sentences (44) and (45) allow for strict and sloppy readings: for both, we understand either that Jane asked for Pedro to be the teacher of class Y (strict reading) or that Jane asked for herself to be the teacher of class Y (sloppy reading). We have to pose there is reconstruction in these structures in order to account for the possible interpretations. The reconstructed sequences are in (46) – we have to have different strings, involving different gender agreement:13
(46) a. antes de Jane solicitar [ele ser o professor da turma Y] before of Jane solicit-inf he be the teacher -masc of-the class Y ‘before Jane asked that he be the teacher of class Y.’
b. antes de Jane solicitar [ela ser a professora da turma Y] before of Jane solicit-inf she be the teacher-fem of-the class Y ‘before Jane asked that she be the teacher of class Y.’
Examples like (39) and (45) above shows we cannot assume the null object is a pro in BP, as many studies propose (see above). If we did, we should be able to explain why a pro behaves exactly as ellipsis in this case, with respect to the ambiguity of readings, but not in other structural positions.
In this paper, I propose that in BP the null complement in the NCA-like structures in fact has internal structure, which can be thought of as reconstruction of the antecedent, if we assume that the ellipsis site contains a syntactic copy of its antecedent that is unpronounced.
The use of the null object above in (45), the propositional ellipsis (cf. Cyrino 1997) has always been possible in European Portuguese, in variation with the neuter clitic (see also Matos 2003). In BP, however, the neuter clitic has disappeared, and ellipsis is the only option now. Furthermore, the null nominal object with the same ([-animate]) type of features has appeared. The hypothesis in Cyrino (1997) for the diachronic shift is that there has been a change in which the child extended the possibility for the ellipsis of the structures with the neuter
13 It is important to notice that in Fiengo and May’s account lexical content is not relevant for identity of dependencies, but categorial identity is. In their terms, the strict reading involves reconstruction of α-occurrences and the sloppy one, reconstruction of β-occurrences.
clitics to other structures that were similar to them – the structures in which the clitic also had the [-animate] features14. Below are the relevant figures:
Century Null object ([+ani]NP antecedent) Null object ([-ani]NP antecedent) Propositional ellipsis (absence of neuter clitic ‘o’) XVI 2% (2/86) 6% (5/87) 23% (23/99) XVII 5% (3/55) 13% (17/130) 21% (14/68) XVIII 5% (1/22) 8% (10/131) 45% (41/90) XIX 1% (1/79) 44% (38/87) 83% (81/98) XX 14% (4/28) 88% (91/103) 91% (97/107)
Table 1 Null objects and propositional ellipsis according to type of antecedent, adapted from (Cyrino (1997), (Numerator = null; Denominator = null + overt objects)
The proposal in this paper is based on the studies in Cyrino (1997) and Cyrino (2003) for null objects in BP and on Cyrino and Matos (2002, in preparation) for VP ellipsis.15 In these studies, these processes are all instances
of reconstruction at LF.
14 Incidentally, European Portuguese still has the neuter clitic o (see Matos, 1985, for a complete description of the occurrence of this clitic in EP. It also allows null objects, as we saw above, but of a different kind, if compared to BP. Besides not being allowed in island structures, null objects in EP do not have the animacy restriction that we see in null objects in BP.
15 European and Brazilian Portuguese (EP and BP) pattern with English in presenting VP ellipsis (cf. (i)).
(i) Perguntamos se eles já tinham chegado e, efectivamente, já asked if they already had arrived and, indeed, already had
tinham __
‘we asked if they had arrived already and, indeed, they already had.’
In fact, EP and BP exhibit differences concerning the licensing of the elliptical constituent whenever verbal sequences formed by an auxiliary and a main verb occur. While (iia) below is interpreted as VP ellipsis both in EP and BP, (iib) in BP allows for the recovering of all the complements of the verb; in opposition, (iic) in EP favours indefinite Null Object reading and misses to recover the indirect object.
(ii) a. Ela está a ler/ lendo livros às crianças mas ele não está __ EP,BP
she is to read/reading books to-the children but he not is __
'She is reading books to the children but he is not.' __ = [VP (is) reading the books]
b. Ela está lendo livros às crianças mas ele não está lendo__. BP
she is reading books to-the children but he not is reading __ __= [VP (reading) books to the children]
We saw above that the null object in BP (47a) shows the strict/sloppy ambiguity. This happens similarly in a VP ellipsis structure (47b).
(47) a. O João dobra seu jeans, The João fold-pres-3sg his jeans
mas a Maria amassa ___
but the Maria crumple-pres-3sg ___
‘João folds up his jeans, but Maria crumples___’ b. O João dobra seu jeans,
The João fold-pres-3sg his jeans
e a Maria também dobra ___. and the Maria too fold-pres-3sg ___ ‘João folds up his jeans, and Maria does too’
In one interpretation in (47a), we can have the sloppy reading, in which Maria crumples her own jeans whereas in the other interpretation, we can have the strict reading, in which she crumples John’s jeans. Likewise, in (47b), in one interpretation Maria folds her own jeans, and in the other interpretation, she folds João’s jeans.
As we also saw above, NCA also allows this strict/sloppy reading ambiguity—it was one of the arguments presented for its status as surface anaphora in BP. Therefore, I propose that we extend the reconstruction analysis to this structure (reconstructed sequence in italics):
(48) Joãoi beijou sua amiga, mas Pedroj não quis ___ . ( ___ = [beijar suai/j amiga])
‘Johni kissed his friend but Peterj didn’t want to ___. (___ = [kiss hisi/j friend]).’
In this way, the proposal in this paper relates the NCA in (49) below to the null object in (50), this latter structure being possible in BP, but not in English (cf. Cyrino and Matos, in preparation, for licensing conditions present in Portuguese but not in English):
c. Ela está a ler livros às crianças mas ele não está a ler __.PE
she is to read books to-the children but he not is to read__ 'She is reading books to the children but he is not reading.’ __ = [VP (reading)[DP anything]]
Cyrino and Matos (in preparation) claim that the differences exhibited by the two varieties of Portuguese are mainly due to the distinct functional projections involved in interaction with the parallelism requirements for the recovering/deletion of the elliptical category.
(49) a. John asked me to go to the party, and I accepted ___. ( __ = to go to the party)
b. João me pediu para ir à festa, João me ask-past-3sg to go-inf to-the party e eu aceitei ___. ( __ = ir à festa) and I accepted ___ (___ = go-inf to-the party) (50) a. * John gave me an invitation, and I accepted ___.
( __ = the invitation)
b. João me deu um convite, João me give-past-3sg a invitation
e eu aceitei ___. ( __ = o convite) and I accepted ___ ( ___ = the invitation)
Here we see that both structures are very similar, and we may consider the NCA in BP as an epiphenomenon of a larger process of ellipsis possible in the language.
5. NCA in BP?
In this paper, I argue that BP does not have NCA as classically described. Instead, the phenomenon is another instance of surface anaphora (with internal structure) in this language, as is the null object (and VP ellipsis).
Depiante (2000) proposes that one of the characteristics of NCA is that it is never interpreted nominally (only as a property or proposition) (51): (51) a. The teacher told the children that it was time to leave even though
they already knew ___ (proposition: “that it was time to leave”) b. * The children learned the song on Monday but by Friday they no
longer knew ___ (nominal: “the song”) Compare with (52):
(52) a. * Ana quiere una manzana y María también quiere ___ ‘Ana wants an apple and María also wants (an apple)’ b. Ana quiere retirarse temprano y María también quiere __
‘Ana wants to leave early and María also wants (to leave early)’ When we try to see how this would be in BP, we have an indication that perhaps we are not facing the same kind of phenomenon. As we have seen, it is difficult to apply this test in BP because of the existence of the null object construction. The BP correspondents of (51) and (52) are grammatical, see (53) and (54) below—we interpret the gap as a null ‘it’, propositional ellipsis, but
remember that the null object in BP could also be understood as a null ‘it’, nominal ellipsis:
(53) a. O professor disse para as crianças The teacher say-past-3sg to the-pl children
que era hora de ir embora, mesmo que that was time of go-inf away even that elas já soubessem ___
they already know-subj-3pl ____
(proposition: “que era hora de ir embora”) (proposition: ‘that it was time to go away’)
‘The teacher told the children that it was time to leave, even though they already knew ___
(proposition: “that it was time to leave”)
b. As crianças aprenderam a música The-pl children-pl learn-past-3pl the song na segunda-feira mas na sexta-feira in-the Monday but in-the Friday
elas já não sabiam ___ (nominal: “a música”) they already not know-past-3pl ___(nominal: ‘the song’)
‘The children learned the song on Monday but on Friday they already didn’t know ___’ (nominal: ‘the song’)
(54) a Ana quer uma maçã
Ana want-past-3sg a-fem apple
e Maria também quer ___. and Maria too want-pres-3sg ___ ‘Ana wants an apple and María also wants (an apple)’ b. Ana quer ir embora cedo
Ana want-pres-3sg go-inf away early e Maria também quer ___. and Maria too want-pres-3sg ___
‘Ana wants to leave early and Maria also wants (to leave early)’ (53), however, could also be characterized as a VP ellipsis structure in BP.
Depiante also points out that the NCA in English and in Spanish has a definite interpretation given by the context (55):
(55) I eat meat and go late to bed, and my brother doesn’t approve ___ Depiante makes it clear that this is different from the intransitive use of verbs, like eat—(55) doesn’t mean my brother doesn’t approve of anything the speaker does, it means he doesn’t approve of the definite fact of the speaker’s eating meat and going to bed late.
Again, this is also possible in BP, see (56). As stated above, in BP we do not have the occurrence of a neuter clitic o ‘it’. This clitic has been
diachronically lost—now we have a null complement in this position, cf. Cyrino (1997):
(56) Eu costume comer carne e dormir tarde, I use-pres-1sg eat-inf meat and sleep-inf ate mas meu irmão não aprova ___
but my brother not approve ___
‘I use to eat meat and go to bed late, but my brother doesn’t approve (my eating meat and going to bed late)’
6. Conclusion
The NCA is a poorly understood phenomenon, which occurs in various languages, invariably related to auxiliary and/or aspectual verbs. There is a restriction on the nature of the null complement as well, but very few accurate descriptions or analyses of the process have been advanced.
My proposal in this paper is that we do not have NCA in BP anymore, at least not as a deep anaphora phenomenon. Instead, the possibility for reconstruction and ellipsis, present in null objects and VP ellipsis, has been extended16 to the complements of NCA verbs. Consequently, the NCA structures
have the surface anaphora characteristics which are present in those ellipsis constructions.
References
Barra Ferreira , M. (2000). Argumentos nulos em português brasileiro. MA Thesis, UNICAMP, Brasil.
Bianchi, V. and M.C. Figueiredo Silva (1994). On some properties of agreement-object in Italian and Brazilian Portuguese. In M. Mazzola (ed.), Issues and theory in Romance linguistics, Washington, DC: Georgetown University Press, pp. 181-197.
Bosque, I.(1984). Negación and ellipsis. Estudios de Linguistica 2: 171-199.
Brucart, J. M. (1999). La elipsis. In I. Bosque and V. Demonte (eds.), Grammatica descriptiva de la lengua española, Madrid: Espasa Calpe, pp.2787-2863. Chao, W. (1987). On ellipsis. Doctoral Dissertation, University of Massachussetts,
Amherst.
Cyrino, S. (1997). O objeto nulo no português do Brasil: um estudo sintático-diacrônico.
Londrina: Editora da UEL.
Cyrino, S. (2003). Para a história do português brasileiro: investigando a Anáfora do Complemento Nulo, a categoria Aspecto e o Objeto Nulo no português brasileiro. Research Project, CNPq.
Cyrino, S. (in press). On the existence of null complement anaphora in Brazilian Portuguese. Letras.
Cyrino, S. and R. Lopes (2004). Animacy as a driving cue in change and acquisition in Brazilian Portuguese. Paper presented at the International Conference on Linguistic Evidence, University of Tübingen.
Cyrino, S. and G. Matos (2002). VP ellipsis in European and Brazilian Portuguese: a comparative analysis. Journal of Portuguese Linguistics. 1(2): 177-195. Cyrino, S. and G. Matos (in preparation). Local licensers and recovering in VP ellipsis
construction: variation across languages and language varieties.
Cyrino, S. and U. Reich (2002). Uma visão integrada do objeto nulo no Português Brasileiro. Romanistiches Jahrbuch. Band 52: 360-386.
Depiante, M. (2000). The syntax of deep and surface anaphora: a study of null
complement anaphora and stripping/bare argument ellipsis. Doctoral
Dissertation, University of Connecticut.
Farrell, P. (1990). Null objects in Brazilian Portuguese. The linguistic review 8: 325-346. Fiengo, R. and R. May (1994). Indices and Identity. Cambridge, Mass: MIT Press. Galves, C. (1987). A sintaxe do português brasileiro. Ensaios de lingüística 13: 31-50. Galves, C. (1989). Objet nul et la structure de la proposition en Portugais Brésilien.
Review des Langues Romanes 93: 305-336.
Grimshaw, J. (1979). Complement selection and the lexicon. Linguistic Inquiry 2: 269-313.
Hankamer, J. and I. Sag. (1976). Deep and surface anaphora. Linguistic Inquiry. 7(3): 391-426.
Kato, M. (2000). Pronomes fortes e fracos na sintaxe do português brasileiro, Ms., UNICAMP.
Matos, G. (1985). Clítico verbal demonstrativo. MA thesis. Universidade de Lisboa. Matos, G. (2003). Construções elípticas. In M.H.M. Mateus et al. (eds.) Gramática da
língua portuguesa. Lisboa: Caminho, pp. 869-913.
Radford, A. (1977). Italian syntax: transformational and relational grammar. Cambridge: Cambridge University Press.
Raposo, E. (1986). On the null object in European Portuguese. In O. Jaeggli and C-S. Corvalán (eds.) Studies in Romance Linguistics, Dordrecht: Foris, pp. 373-390. Sag, I. and J. Hankamer (1984). Toward a theory if anaphoric processing. Linguistics and
Philosophy 7: 325-345.
Schwenter, S. and G. Silva (2002) Overt vs. null direct object in spoken Brazilian Portuguese: a semantic/pragmatic account. Hispania 85: 577-86
Zubizarreta, M-L. (1982). On the relationship of the lexicon to the syntax. Doctoral Dissertation, MIT.
Zocca, C. (2002). O que não está lá? Um estudo sobre morfologia flexional em elipses
MA Thesis, UNICAMP.
Zribi-Hertz, A. (1986). Relations anaphoriques en Français: esquisse d’une grammaire générative raisonné de la réfléxivité et de l’ellipse structurale. Doctoral Dissertation, University of Paris.
Departamento de Letras Vernáculas e Clássicas Centro de Letras e Ciências Humanas
Universidade Estadual de Londrina Caixa Postal 6001 CEP 86051-990 Londrina, PR BRAZIL [email protected]