VP ellipsis in European and Brazilian Portuguese – a comparative analysis

European and Brazilian Portuguese present VP ellipsis licensed by auxiliary and main verbs. However, these varieties of Portuguese show up some differences concerning the licensing and identification of the elliptical constituent whenever sequences of verbs are involved. We will claim that these differences are mainly due to the properties of the functional projections involved.


Introduction
In contrast with other Romance languages, European and Brazilian Portuguese (EP and BP) present VP ellipsis. In both varieties of Portuguese, this construction displays the same general pattern of behaviour: it may occur both with auxiliary and main verbs and it requires that the verb which licenses the elliptical constituent also appears in the antecedent VP. Yet, whenever sequences of verbs are involved, specially when they include the main verb, standard EP and BP exhibit differences concerning the identification of the elliptical constituent as VP ellipsis or not. This contrast correlates with the distribution of também 'too' within these verb sequences: in BP this adverb may asymmetrically c-command the whole sequence of verbs, or just the last verb, and still keep the VP ellipsis interpretation, usually lost in EP in the latter case. We will try to show that these differences are mainly due to the intrinsic properties of the functional categories that license the elliptical constituent and constitute a clue for its recovering. This work is organised as follows: in the first section we present empirical evidence for VP ellipsis in European and Brazilian Portuguese, as an elliptical construction distinct from Stripping and Null Object; in section two we will try to characterise the configuration minimally required for the licensing of the elliptical VP; in the third section we will deal with the differences in both varieties of Portuguese; finally, in the final remarks, we will summarise the main points of this study.

1.VP ellipsis in Portuguese an autonomous elliptical construction
In many Romance languages, VP Ellipsis is not an available option. So, some studies on predicate ellipsis tend to relate it across Romance to other elliptical constructions. In this section we will provide some empirical arguments to distinguish VP ellipsis in European and Brazilian Portuguese from Stripping and Null Object. 1

VP ellipsis vs. Stripping
In contrast with other Romance languages, European and Brazilian Portuguese present VP ellipsis, as shown by the examples in (1) and (2): (1) a. A Ana já tinha lido esse livro à irmã mas a Paula não tinha [-]. The Ana already had read that book to-the sister but the Paula not had [-] 'Ana had already read that book to her sister but Paula had not.' b. O João é simpático para todas as pessoas e a Ana também é [-]. The João is nice for all the people and the Ana too is [-] 'João is nice to everybody and Ana is, too' c. Ela só vai visitar os amigos se a Ana também for [-] She go.Indicative.3sg visit the friends if the Ana too go.Subjonctive. 'She will visit her friends if Ana will, too.' (2) a. *Susana había leído Guerra y Paz pero María no había [-]. 2 Susana had read War and Peace but Maria not have b. *On a demandé si ils ont déjà mangé et ils ont [-]. 3 One has asked if they have already eaten and they have [-] c. *Claudine est une bonne étudiante, et Marie est aussi. Claudine is a good student and Marie is too In Spanish and French the usual strategy of predicate ellipsis is Stripping, illustrated in (3) and (4).  (1a) and (1b) vs. (6): (6) a. A Ana já tinha lido esse livro à irmã mas essa revista não [-]. The Ana already had read that book to-the sister but that magazine not [-] 'Ana had already read that book to her sister but Paula had not.' b. Por la mañana podría ir de compras, pero por la tarde no [-]. 8 By the morning I-could go of shopping, but by the afternoon not [-] 'I could go shopping in the afternoon not.' 4 Brucart 1987: 134. 5 López 1999:275. 6 Zribi-Hertz 1986:466. 7 Chao 1987:180. 8 López 1999 Additionally, VP ellipsis contrasts with Stripping in being able to appear in island domains (see (1c) vs. (7)): (7) a. *Ela só vai visitar os amigos se a Ana também [-]. She just will.Indicative visit the friends if the Ana too [-] b. *O João não vai ao cinema hoje mas perguntou quem sim [-] The João not goes to the movies but (he) asked who yes [-]. c. *Tendré que hacerlo yo porque Susana no [-]. 9 Will_have that to do it I because Susana not [-] Thus, the data from European and Brazilian Portuguese show that in this language VP ellipsis and Stripping are two different constructions.

VP Ellipsis vs. Null Object
One of the most significant differences between VP ellipsis in English and Portuguese is that, in the latter, in addition to the canonical cases where the elliptical category is licensed by an auxiliary (cf. (8a)=(1a)) or a copulative verb (cf. (8b)=(1b)), there are instances of VP ellipsis locally identified by a main verb (see (9)). The existence of these cases in both varieties of Portuguese, raises the problem of distinguishing VP ellipsis licensed by main verbs from Null Object. 10 A classical distinction between these constructions relies on the fact that typical cases of Null Object only involve the direct object of the verb, while VP ellipsis recovers all the complements of the verb and sometimes the VP adjuncts (cf. Raposo 1986). 11 Accordingly, while the example in (11) is ambiguous between a VP ellipsis or a Null Object interpretation due to the fact that the main verb selects only the direct object, the one in (12)  10 As far as Portuguese is concerned, see, among other, the studies of Raposo 1986, 1998, Costa & Duarte 2001, for EP and Farrel 1990, Kato 1993, Cyrino 1997 Summarising the properties presented in this section, the following descriptive pattern emerges for VP ellipsis in European and Brazilian Portuguese: VP ellipsis in EP and BP is an elliptical category, selected by a sentence functional projection, headed by an overt verbal element, which must be identical to (one of) the verb(s) of the antecedent VP. Chomsky & Lasnik 1993, Chomsky 1995, and Lasnik 1999a, 1999b consider that ellipsis must be analysed as a Deletion process operating at PF. A Parallelism Requirement (PR) applying at LF would ensure that the phrase to be deleted receive an interpretation similar to its antecedent. At the level of PF, the constituents to be deleted present a low-flat intonation (Chomsky & Lasnik 1993, Chomsky 1995 and are deaccented (Tancredi 1992), which constitutes an indication for Deletion to apply . Despite the relevance of the Parallelism Requirement for setting which linguistic expression may be deleted, it is not a sufficient condition as shown in (19a)  These contrasts show that some licensing condition must be met, whatever analysis we assume for ellipsis either Deletion at PF or Interpretation at LF. We can view these licensing conditions as structural clues for the identification of the constituent to be deleted or recovered. Within the Minimalist Program, some proposals have been advanced to account for the licensing and identification of VP ellipsis, both in English and in Portuguese (Lobeck 1999, López 1999, Martins 1994, Matos & Cyrino 2001). We will concentrate on the analyses which focus Portuguese. Martins (1994) and López (1999) argue that VP ellipsis is licensed by the functional category Sigma, proposed in Laka 1990, and the elliptical VP is characterised as a base generated null category. Yet, they differ in the configuration proposed: while Martins assumes that P heads the whole IP, as in (20a) ( López 1999) In both accounts, the licensing and identification of the null VP is achieved by checking of the features of the null VP in the domain of  with strong-V features. According to Martins (1994: 191), this licensing is obtained in a configuration akin to Spec-Head Agreement, by adjunction of the Null VP either to Spec, P or to P. The motivation for that movement would be the truth-value of the elided VP, which should be checked in the domain of Sigma. López (1999), proposes a different implementation of the licensing of the null VP. Accepting that the null VP is a pro-v, i.e., a category with no internal structure, presenting -features, López assumes that it overtly moves like a head and adjoins to º to check its strong -features. In English, this process is mediated by Aux  pro-v adjoins to Aux, which inherits its -features; then the complex head moves and adjoins to º to check its strong features, as illustrated in (21):

The licensing of VP ellipsis
9 The proposals of Martins 1994 and Lopez 1999 present a major problem: in opposition to the central hypothesis of these analyses, there is empirical evidence against a correlation between P and the licensing of VP ellipsis. In fact, as López 1999 remarks, the polarity items cannot by themselves license the elliptical VP. Thus, (22a) contrasts with (22b), due to the absence of a verb heading the licensing head. Similar contrasts occur in Portuguese, as shown in (23)  the polarity item in isolation is not able to license predicate ellipsis in island domains, while VP ellipsis is allowed to do it.
He asked who had eaten the cake and she asked who not b. Ele perguntou quem tinha comido o bolo e ela perguntou quem não tinha.
He asked who had eaten the cake and she asked who not had 'He asked who had eaten the cake and she asked who had not' López claims that the contrasts in (22) are due to the fact that in English the polarity item is a specifier of P; therefore, in (22a) º is lexically unfilled and cannot host the elliptical category, since elliptical constituents, like clitics, require phonetically realised hosts. Yet, taking into account the data of European Portuguese, this cannot be the correct explanation for the unacceptability of (22a): the sentence negative marker is a head, which may incorporate with the inflected verb in the sentence, as in (24b). Thus, these data strongly suggest that P is not involved in the licensing of VP ellipsis. 14 14 Notice that the relevance of P in the licensing of VP ellipsis in Portuguese is not systematically assumed in Martins 1994, since she claims that in embedded affirmative sentences and in negative sentences, the verb remains in AgrSº as illustrated in (i) and (ii). However, VP ellipsis is possible in these contexts: Moreover, López 1999 faces an additional problem in assuming that the elliptical VP must be a category with no internal structure This analyisi fails to account for Portuguese, accepting the usual claim that in this language, all the verbs, be they main or auxiliary, are originally merged in VP and then moved to Tense. 15 An alternative analysis has been presented in Matos & Cyrino 2002, which, with minor alterations will be adopted in this study. Developing previous work (Matos 1992, Cyrino 1997and Matos & Cyrino 2002), we will assume that the licensing of VP ellipsis is achieved in the configuration presented in (25) In fact she admits that in negative sentences the polarity head remains isolated in  º (Martins 1994:184), ignoring the evidence for Neg-incorporation onto the tensed verb , and assumes that in embedded sentences  º raises to Cº leaving the verb in Agrsº (Martins 1994:200). For Martins these configurations are crucial since she grounds on them her analysis of proclisis. For the author, enclisis (as VP ellipsis), arises from V movement to  º, whenever this movement is precluded, proclisis occurs. 15 Although the work by Lobeck 1999 does not deal with Portuguese, the same problem will arise if one tries to extended her proposal to this language. 16 In a framework that admits split Inflection the condition on the licensing of the VP ellipsis should be more accurately formulated as in (i): (i) VP ellipsis is licensed, under the local c-command, by the chain of the functional head with V-features that merges with elliptical constituent, when this head is lexically filled.
The condition proposed in (25) does not resort to Checking Theory. Instead, it assumes that the crucial licensing factor is local c-command. In fact, in the present framework only uninterpretable features have to be checked, and, as far as verbal elements are concerned, these features are restricted to the inflectional morphology associated to the verb, namely its -features. We admit, therefore, that in VP ellipsis just the verb moves to check uninterpretable features of the sentence functional projections, for reasons which are independent of the licensing of the elided category. On the contrary, the local c-command by a lexicalised V-functional head is required for the identification of the elliptical category. In particular, whenever the local c-commanding verbal head arises from Movement, it is enough to fully identify the elliptical category, as expect, according to the Bare Phrase hypothesis, which does not radically distinguish a Maximal Projection from its Head.

VP ellipsis with verbal sequences in European vs. Brazilian Portuguese
Though obeying to the same licensing condition, VP ellipsis in EP and in BP often behave differently when verbal sequences occur in the elliptical sentence. So, in sequences of Progressive Auxiliary + main verb, the following contrasts obtain: while in BP, both (26) and (27) are interpreted as cases of VP ellipsis, in EP (28) is preferably interpreted as a sentence with a null indefinite object (i.e., he is not reading anything) and no indirect object, as illustrated in the partial representations of the VP. For a sake of visibility we will use a trace for the moved verb instead of a copy: (31b) shows that the inclusion of the main verb in the verbal sequences is not a compelling factor for obtaining a Null Object reading, or it would not receive a VP ellipsis interpretation. Thus, we hypothesise that what really allows or precludes the interpretations of VP ellipsis is the functional projections where the verbal elements faces up at Spell-Out. The contrasts between European and Brazilian Portuguese also indicate that the functional projections that license VP ellipsis in these varieties may be not the same, possibly due to the idiosyncratic properties of the verbs involved in the verbal sequences.
Current studies on the syntax of verbal sequences in European Portuguese (Matos 1992, Gonçalves 1996 have emphasised that there is a scale of 13 gradation among auxiliaries in this language: 19 while the auxiliary verb of the Perfect Tense is a prototypical auxiliary, most of the other auxiliaries are best characterised as defective verbs which, in certain respects, behave like main verbs, selecting non-finite sentential projections as complement.
Taking the presence of pronominal clitics and sentence negation in European Portuguese as an evidence for the projection of Tense (see, for instance, Rouveret 1995, Ambar 1999, Duarte et alii 200, Gonçalves 1996, Matos 2001, the behaviour of the different auxiliaries show that they are in different stages of defectivity.
The most defective is the Perfect Auxiliary. It selects a Past Participle complement, which must be analysed as Verbal Projection distinct from TP. Thus, the Auxiliary and the main verb belong to the same tense domain, and the Past Participle may not be negated nor act as a host for clitic pronouns: (31) a. * Ele tem contado-lhes CL histórias. 'He has told themCL stories' b. * Ele tem não contado histórias aos miúdos. He has not told stories to the kids 'He has not told stories to the kids.' The same does not happen with the complement of the Progressive Auxiliary in (32): the infinitival verb may host the clitics, and the infinitival domain may accept negation.
(32) a. Ele (não) está a contar-lhes CL histórias He (not)is to tell them stories 'He is not telling them stories.' b. (?)Ele está a não lhes contar a história toda. He is to not them tell the story whole 'He is not telling them the whole story.' In fact, as argued in Raposo 1989, the Progressive aspectual verb occurs in a Prepositional Infinitival Construction, i.e., it selects a small clause headed by a Preposition which takes as its complement a TP projection, as shown in (33b): 19 Most of the criteria used for the classification have been analogous to the ones proposed in Pontes (1973) andLobato (1975) for PB.
However, there is a competing construction in EP, in which the progressive aspectual and the main verb restructure, forming a verbal complex and a single Tense domain occurs (cf. Matos 1992, Gonçalves 1996. In this case, the preposition is interpreted as an aspectual particle and the auxiliary subcategorize for an Aspectual Phrase (cf. Duarte 1993). In this construction, as expected, Clitic Climbing is possible and the sentence negation defines as its scope domain the tensed domain of the whole verbal complex ( It is this latter construction that canonically faces up in VP ellipsis in European Portuguese ( Matos 1992). As shown in (35), whenever VP ellipsis occurs, the splitting up of the verbal complex produces marginal results . This hypothesis seems to be corroborated by another contrast: the distribution of também 'too' within these verbal sequences is less restricted in BP than in EP. In EP, for the VP ellipsis interpretation obtain, this adverb must asymmetrically c-command the whole sequence of verbs, as in (41); in BP, the VP ellipsis interpretation occurs in (41), but also in (42) and in (43) where também c-commands just the main verb. In EP (42) as well as (44)  Assuming that também 'too' is a focussing adverb, and that focussing adverbs are heads which select as complements different projections, as claimed in Cinque 1999, we can explain the contrasts in acceptability between EP and BP for the sentence in (42) and (43) In contrast, the grammaticality of these sentences in PB, with the intended VP ellipsis reading, corroborates that in this variety of Portuguese the licenser of elliptical VP may be any sentence functional projection below TP.
We hypothesise that this change in the licenser of VP ellipsis in European vs. Brazilian  i.e. form Tº to any functional head that merges with the elliptical VP  correlates with another one: the loss of unrestricted Generalised Verb Movement in this variety of Portuguese. In fact, it has been often claimed that although Verb Raising to Inflection is available in BP, it is kept to a minimum.

Summarising
European and Brazilian Portuguese, differ from other Romance languages in presenting two distinct predicate ellipsis constructions: VP ellipsis and Stripping. In both varieties of Portuguese, the elliptical VP may be identified either by an auxiliary or a main verb. Nevertheless, in the latter case, there is empirical evidence to assume that the elliptical construction is not reducible to Null Object. Some analyses of VP ellipsis in Portuguese within the Minimalist Program have tried to explain its licensing in terms of the properties of , conceived as the functional category which is headed by the positive and negative polarity items in the sentence. However, the presence of these items is not enough to license the elliptical predicate in some contexts where the canonical configuration of VP ellipsis may occur. Alternatively we admitted that VP ellipsis is licensed under local c-command of the overt sentence functional head that merges with the elliptical VP. In EP this head must be Tº, or otherwise the VP ellipsis interpretation is lost; in BP it may be any functional head bellow Tº.