Aspectual Verbs in European and Brazilian Portuguese

This paper is about the semantics and the syntax of aspectual verbs in European and Brazilian Portuguese. Some of these verbs select a+Infinitive in the European variety whereas the same verbs select that structure and Gerund in the Brazilian variety. We show that this discrepancy can be explained in semantic and syntactic grounds due to the different nature of the semantic ‘input’ (stative or eventive), the availability of tense constructions and the functional structure of the embedded domain. Another group of aspectual verbs do not differ in their syntactic structure in the two varieties as their structure (de+Infinitive) is the same, although syntactically different from a+ Infinitive, but they do differ in their semantics and also in the selection of different verbs.


Introduction
The main goal of this paper is to articulate the semantic and syntactic characterization of aspectual verbs in European Portuguese (EP) and Brazilian Portuguese (BP).These verbs occur in two different structures in both varieties: 1. a + Infinitive (Inf) / Gerund (Ger); 2. de + Infinitive 1 .The first pattern is associated with semantic and syntactic distinctions in both varieties, whereas the dissimilarities in the second one are fundamentally semantic, the syntax being basically uniform.In brief, in the first case, as Ger is not available in standard contemporary EP in the relevant constructions, these can differ with respect to the availability of tense values and forms, the functional structure of the embedded domain and the stative/eventive nature of the 'inputs' at stake.In the de + Inf constructions, the aspectual verbs have a quite similar 'output' but their 'inputs' are of a different aspectual nature.
In section 2 we present the semantic analysis of aspectual verbs that select a + Inf or Ger structures in both varieties.Section 3 will analyse the syntax of the same verbs.Section 4 will deal with the second type of aspectual verbs, those that select de + Inf in both varieties, from both a semantic and a syntactic point of view.

Aspectual verbs as semantic operators
We assume that the main semantic feature of aspectual verbs is to modify the inner temporal properties of the eventuality with which they co-occur, being true aspectual operators, as defined in Moens (1987).In this respect, the main function of verbs like começar, passar, continuar, estar, andar, ficar, deixar, parar or acabar, when they occur in one of the structures mentioned above, is to select an aspectual class as its 'input' and convert it into another one with (possibly) different temporal features as its 'output' (cf., among others, Cunha, 1998).
Having said this, we must elucidate what can be considered as an 'input'.In fact, for an aspectual transition to obtain, the 'input' is constrained by the necessary conditions that must be met, in particular the existence of one or more phases of the eventuality that must have occurred before the application of the operator.This means that, in many cases, the 'input' is not of the same aspectual nature as the basic predication, since derived situations may also be at issue, as long as they meet the required conditions to serve as 'inputs' for a given aspectual operator.
The aspectual nature of the 'output' corresponds to the final reading of the sentence itself and its aspectual class is determined by the familiar tests (Dowty, 1979, Moens, 1987, Cunha, 1998).

Aspectual verbs in EP: a + Infinitive
In order to illustrate and clarify what we said about the nature of the 'input' and the 'output' of aspectual verbs, we will briefly analyse the behaviour of some of these in EP2 .
The Progressive form, estar + a + Inf (be + A + Inf) for instance, is used to convert a process in progress into a stative predication that is, a progressive state.This conversion involves the presence of a basic or derived process as its 'input'.This would explain why the Progressive hardly applies to states (Vendler, 1967, Dowty, 1979, Vlach, 1981).However, the basic states do not form a homogeneous class and, following Cunha (1998), we must admit a fundamental distinction between non-phase states and phase states as the latter can be integrated in the Aspectual Network and converted into a process whereas the first ones cannot undergo any kind of transition, remaining always stative.This distinction allows one to understand why some basic states can appear in a progressive form and others cannot, as shown by the contrast between (1) and ( 2) where a phase-state behaves very similarly to events in this respect (see examples (3)-( 5)) (1) *A Maria está a ser alta.(non-phase state) the Maria is A to-be tall 'Maria is being tall' (2) O meu cão está a ser agressivo.(phase state) the my dog is A to-be aggressive 'my dog is being aggressive' (3) O Pedro está a correr.(process)  the Pedro is A to-run 'Pedro is running' (4) A Rita está a comer uma maçã.(culminated process)  the Rita is A to-eat an apple 'Rita is eating an apple' (5) O gato está a morrer.(culmination)  the cat is A to-die 'the cat is dying' The analysis of these examples shows that only non-phase states cannot occur in the Progressive.The other aspectual classes seem to combine straightforwardly with this operator.However, to obtain such acceptable results, they have to be previously converted into processes.So, culminated processes like the one in (4) have to lose their culmination in order to meet the conditions required by the Progressive form. 3The same transition would obtain with culminations like (5), provided that a process had previously been added to them.Phase states have to be converted into processes in order to combine with the Progressive: such assumption explains why only this kind of states can occur in the above-mentioned context, contrasting with others that cannot.
As we said, the 'output' of the Progressive is a state.This can be confirmed by the behaviour of progressive sentences with respect to temporal adverbials and to several structures expressing agentivity, that is, basic agentive structures like ler o jornal (to read the newspaper) become generally non agentive in the context of a progressive construction. 4) O Rui esteve a ler o jornal durante 2 horas / ?às 2 horas / *em 2 horas.the Rui was A to-read the newspaper for 2 hours/at 2 pm/in 2 hours (7) *Rui, está a ler o jornal!Rui, be A to-read the newspaper!(8) *A mãe obrigou o Rui a estar a ler o jornal.
the mother forced the Rui to be A to-read the newspaper A similar analysis can be given to the 'inputs' of andar and ficar, as their combinatory conditions are fairly the same as those for estar. 5The most significant differences arise with their 'outputs': andar gives as its 'output' a habitual (or iterative) state, while ficar triggers a punctual event associated with a process. 6The reason why we advocate a process in the 'output' of ficar is due to the fact that sentences with this aspectual verb are acceptable in contexts like the ones in ( 7) and ( 8), that is, ler o jornal (read the newspaper).doesnot loose its agentivity, in contrast to estar a (to be + A + Inf).
Consider now the aspectual verb começar (to begin/start).Informally, we can say that its basic semantic function is to indicate the beginning of a given situation.But, what are, precisely, the necessary conditions that must be met in order for this verb to be appropriately used?We hypothesise that the 'input' for this verb is identified with a period leading up to, but not including, the initial portion of a basic or derived process.We consider then that such an 'input' is a kind of a preliminary state inasmuch as it should be viewed as a stable period of time preceding a (basic or derived) event. 7This explains why 4 For a much more detailed discussion about the arguments in favour of this point, see, e.g., Dowty (1979), Vlach (1981) and Cunha (1998). 5It is difficult to give an accurate translation of andar + a + Inf or ficar + a + Inf.
We suggest for a sentence like ele anda a ler este livro the following translation: he is (currently) reading this book.For a sentence like ele ficou a ler este livro we suggest the following translation: he stayed reading this book. 6However, it should be mentioned that andar and ficar do not occur with all basic culminations (coerced into processes).
(ii) Ele está/ *anda /*fica a chegar a casa.See Oliveira, Cunha & Matos (2001) for a more detailed discussion of the data.It should also be pointed out that the punctual event (associated to a process) of ficar, although not generally explicit, can be seen in examples like the following: (i) Quando saí, o Pedro ficou a ler um livro.when I left, the Pedro 'stayed' A to-read a book 7 Preliminary states are needed in the Aspectual Network not only in order to account for the 'input' of começar but also to explain the semantic behaviour of other aspectual structures like estar para (cf.Cunha, 1998).
non-phase states are incompatible with começar, like in (9), and phase states (because they can be previously coerced into processes) and most events are acceptable (see ( 10)-( 14)).The reason why we consider a preliminary state is due to the impossibility of detecting any successive phases in the situation preceding the aspectual operation carried out by começar.This explains again why non-phase states cannot co-occur with this aspectual operator, that is, começar requires a change in the aspect profile of the situation which does not exist either in the preliminary state or in the non-phase one.The contrast between ( 13) and ( 14) shows that some culminations resist co-occurrence with começar.As we can see, ( 14) is not acceptable whereas ( 13) is.This is due to the fact that only some culminations admit the adding of a sufficiently extended preparatory phase and this is, as we have seen before, the required condition for the operator to obtain.What we have just said confirms our hypothesis about the nature of the 'input', that is, that we are dealing with a preliminary state of an event.
The 'output' brought about by começar is a punctual event that establishes the beginning of a (basic or derived) process as shown in (15).The reason why we restrict the 'output' to processes is related to a kind of 'Imperfective Paradox' triggered by começar, since a sentence like O Rui começou a comer a maçã ('Rui COMEÇAR-Past A to eat the apple') does not entail, in itself, the truth of O Rui comeu a maçã ('Rui ate the apple').
In case we accept a sentence like (16) as some speakers do, the only possible reading includes the final part of the preliminary state preceding it, meaning that 'it took 5 minutes for Mary to start working'.
the Maria COMEÇAR-Past A to-work / to-read the book at five (16) ?A Maria começou a trabalhar em cinco minutos.
the Maria COMEÇAR-Past A to-work in five minutes' We would presume that an aspectual operator like passar shares a great number of similarities with começar, since it points to the beginning of a given situation.However, it combines only with basic or derived (mostly habitual) states.
Examples in ( 17)-( 18) show that passar is compatible with both non-phase and phase states. 8Moreover, it can combine with events whenever they can be previously coerced into habitual states, as in ( 19) -(22), where ( 21 the Pedro PASSAR-Past A to-eat this apple Concerning its 'output', passar equally seems to receive a stative interpretation, of the same nature of the 'input'.But a kind of a punctual event is associated with this state in order to mark the transition. 9This is due to the fact that some restrictions apply to the co-occurrence with some temporal locating adverbials.Examples ( 23)-( 24) reveal similar restrictions to those exhibited by individual-level non-phase states as in a Maria foi alta *às 2 horas/*em 2 horas/*durante 2 horas (Maria was tall at 2 o'clock/in 2 8 It is difficult to translate passar.When the basic predications are states, the approximate meaning is to become (Portuguese / aggressive); when the basic predications are events the meaning is similar to to begin to (work) in a habitual manner. 9As one of the referees mentioned, there are some apparent counter-examples to this generalization: (i) Depois dos assobios, passou a defender ideias menos radicais na sua exposição.after the whistles, 'passou' to-argue for less radical ideas in his talk.However, this is not a truly counter-example as it shows a different behaviour: (i) without depois dos assobios (after the whistles), the sequence is not possible unless we remove na sua exposição (in his talk), acquiring, in this case, a habitual state reading.This means that, contrary to the cases under discussion in this paper, which can occur without any adverbial, there is in (i) some kind of inference of a situation already going on which suffers some transition (or even some kind of interruption).
What these examples do have in common is a kind of a punctual event to mark a transition.
hours/for 2 hours).10But, at the same time, there are some adverbials that can occur with this aspectual verb, as in (24'): ( We can, then, conclude that both the 'input' and the 'output' of passar convey stative predicates.However, although the main function associated with this aspectual verb is that of expressing the bringing about of a new stative predication of the same nature of the 'input', there is a kind of a punctual event associated to it for marking the transition. 11et us, finally, look at the semantic behaviour of the aspectual verb continuar (to continue).Informally, we can say that this verb adds a new temporal portion of a given durative situation.This is the reason why culminations are unacceptable in sentences with continuar (see ( 29)), although they can occur if interpreted iteratively, e.g., as derived processes like in (30): This verb takes as its 'input' a state, apparently without any restrictions, or a (basic or derived) process.One of the reasons why this is so is that once again we face a kind of Imperfective Paradox (that is, from the truth of A Maria continuou a comer o bolo ('Maria continued to eat the cake') we cannot infer in any way that A Maria comeu o bolo ('Maria ate the cake') is also true).Taking into account that continuar does not affect the internal structure of the eventualities in its scope, since it only adds a new temporal portion of the same type of situation to the eventuality with which it combines, we have to postulate that continuar 1 takes as its 'input' a state and as its 'output' another state of the same type, and continuar 2 takes as its 'input' a (basic or derived) process and as its 'output' a process.So, although it carries some aspectual information, continuar does not fit our definition of operator because it only adds a new temporal portion of the same type of the 'input'. 12he following table sums up what we have just said about various aspectual verbs.

Process
In the following sections we will see how this analysis of aspectual verbs can help to explain discrepancies observed in the structures involving these verbs in European and Brazilian Portuguese.

Aspectual Verbs in EP and BP: a + Infinitive/Gerund
As we have just seen in the previous section, the aspectual verbs given in Table II co-occur with a + Inf in standard contemporary EP. 13 However, interestingly enough, in BP we find that these aspectual verbs can appear with Gerund or with a + Inf contrasting, in this way, with EP.This means that the EP structures in which a + Inf is consistently required by aspectual verbs correspond to BP sequences in which both a + Inf and the Ger can take place.The examples ( 31)-( 36) illustrate standard EP sentences where these aspectual verbs consistently select a + Inf, whereas examples ( 37)-( 42) exhibit an unquestionable alternation between expressions taking a + Inf and Ger in BP14 .
the Rui COMEÇAR-Past A to-write/writing the dissertation (35) A Rita passou a viver/*vivendo na Holanda.
the Rita PASSAR-Past A to-live/living in the Netherlands (36) O Guilherme continuou a beber/ # bebendo o leite.
the Guilherme CONTINUAR-Past A to-drink/drinking milk ( 37 Would it then be possible to find some adequate explanation for this?First of all, we would like to point out that the distribution of a + Inf / Ger with BP aspectual verbs is far from homogeneous.In fact, while some operators prefer, with very few exceptions, the Gerund construction, others, in contrast, require almost exclusively the presence of a + Inf.The following table, based on the examination of examples from the Corpus NILC/S.Carlos, is very significant in this respect.A first look at the BP data presented in this table suggests that there are significant divergences concerning the different configurations in which the aspectual verbs are involved.In fact, aspectual verbs like estar, andar, ficar occur most preferably with the Ger structure, contrasting, in this regard, with começar and passar, which rather select a + Inf, in accordance to what happens with the EP southern dialects (see ( 34)-( 35)).Continuar, on the other hand, seems to be the only operator that can occur with both structures in a balanced manner.
Before we continue the analysis, we would like to point out some important practical issues.Given the corpus extension and complexity, we decided to restrict the scrutiny of the forms in the third person (singular and plural) to the past tense (Pretérito Perfeito (PPerf)).The choice of this tense was mainly determined by the fact that PPerf preserves almost all the aspectual properties of the situations inside its scope.It contrasts, in this respect, with tenses like Present or Imperfeito, which may modify the internal temporal structure of the eventualities in their scope.
It is worth stressing that the misleading scantiness of examples with estar + Ger is due to the fact that the PPerf is not very common in progressive forms, since this tense imposes an obligatory termination on the situations in its scope, contrasting with the Progressive, which expresses chiefly aspectual progress or continuity.However, if we consider the corresponding data with Present tense, we see that the Ger structure is very common in BP, contrasting with the a + Inf one: while está / estão + Ger appears 24340 times, the corresponding construction with a + Inf comes out only 47 times.
Notice also that, in Table II, we have only recorded the figures concerning unequivocally aspectual constructions, ignoring cases in which the verbs under analysis express other semantic values.For instance, começar + Ger or passar + Ger can get interpretations other than the aspectual one: começar + Ger receives very often a strict temporal interpretation, corresponding to começar por in EPsee ( 43); passar + Ger, on the other hand, seems to be only possible with passar as a full verb, expressing either temporal or spatial "movement"see ( 44) and ( 45 At first sight, one could contemplate explaining the two possible constructions in BP saying that a + Inf is restricted to contexts in which the presence of the Gerund gives rise to differentiated readings, that is, non--aspectual ones.In other words, a + Inf would be a valid alternative only for those cases in which the presence of the Gerund explicitly carries out a non--aspectual interpretation; otherwise, the Gerund would be necessarily selected by the aspectual verbs.
However, as we will see, this hypothesis comes easily into trouble.If it can explain the behaviour of aspectual verbs like começar and passar, as the examples just presented suggest, it misleads the whole story about continuar.In fact, this aspectual verb supports either the a + Inf or the Ger construction, both expressing unequivocally an aspectual meaning.This is to say that both Ger and a + Inf structures can occur, in the relevant contexts, having an unambiguous aspectual reading.So, we are forced to reconsider that hypothesis in order to accommodate the relevant data.
A significant feature which can decisively influence the distribution of a + Inf and Ger within the BP aspectual verbs constructions relies on their own semantic as well as syntactic structure.Pursuing this line of thought, we will investigate in the first place the relation between the selection of a + Inf or Ger and the aspectual classes involved in the 'inputs' required by the relevant aspectual verbs.This will be followed by their syntactic analysis.
Comparing the results presented in Tables I and II, we can formulate some hypotheses about the relationship between the aspectual class of the 'inputs' and the relevant structures in BP sentences.While estar, andar and ficar, whose 'inputs' are of an eventive type, take preferably the Ger, começar and passar, which combine exclusively with 'inputs' of a stative nature, rather select the a + Inf form.On the other hand, continuar, which occurs without problems both with eventive and stative 'inputs', appears naturally with any of the above-mentioned structures.
This hypothesis is strengthened by the observation that continuar a + Inf appears mostly with stative contexts (basic or derived, mostly habitualcf.( 46)-( 49)), contrasting with continuar + Ger, which is much more frequent with (basic or derived) eventive 'inputs' (cf.( 50 (NILC, par 295191) Wanderley not paid attention to-the wife and continued watching the game Taking into account this analysis, we propose the following conclusions about BP aspectual verbs with a + Inf / Ger: a. Whenever an aspectual verb takes as its 'input' a (basic or derived) process, it combines preferably with Gerund.
b. Whenever an aspectual verb takes as its 'input' a (basic or derived) state, it occurs essentially with a + Infinitive.
The generalization just presented describes a common behaviour of the BP aspectual verbs under consideration but it is not an unrestricted principle.Actually, we believe that this is a zone subject to change in a more clear way in BP than in EP, although the latter includes some southern dialects where Ger is also used.It is possible to find in BP some exceptional cases in which estar and ficar occur with a + Inf and, conversely, examples in which começar allows, in particular circumstances, the presence of a Ger in its aspectual reading.We can also find sentences in which continuar + Ger takes place in a stative environment and continuar a + Inf appears within eventive 'inputs' (see ( 52)-( 53)).
(NILC, par 1276928) [he] drank a sip and continued A to-eat

a + Inf and Ger constructions: the syntactic facts
In this section, we will show that the semantic differences between EP and BP in what concerns aspectual verbs taking a + Inf and Ger complements are intrinsically related to syntactic structural discrepancies.In this sense, the main goals of the next subsections are as follows: (i) to provide a syntactic distinction between a + Inf and Ger constructions; (ii) to account for the differences between EP and BP on syntactic grounds; (iii) to relate syntactic and semantic facts.

a + Inf constructions 3.1.1. General syntactic properties
The a + Inf construction is possible both in EP and BP, as illustrated by the following examples, in both of which it is selected by an aspectual verb: 15 (54) As patrulhas estão a fazer rusgas (Natura/Público, par 227) (EP)  the patrols are A to-make swoops 'the patrols are swooping' (55) Muitas meninas começaram a chorar.(NILC, par 8774) (BP) many girls COMEÇAR-Past A to-crie 'many girls started crying' This constructiongerundive-infinitive, for Sten (1953), or Prepositioned Infinitival Construction, for Raposo (1989)presents two main properties (see Raposo, 1989, among others.):(i) it occurs in syntactic isolation, as in ( 56) and (57), and (ii) it occurs as the final subject in the context of predicative verbs, as in ( 58) and ( 59): (56) [As patrulhas a fazer rusgas!]Que estranho!(EP) the patrols A to-make swoops!How strange!(57) [Manoel a dizer isso!] Que estranho!(BP) Manoel A to-say that!How strange!(58) [As patrulhas a fazer rusgas] é estranho!(EP) the patrols A to-make swoops is strange (59) [Manoel a dizer isso] é estranho!(BP) Manoel A to-say that is strange Notice that, in these contexts, the a + Inf structure co-occurs with the final subject of aspectual verbs.This is so, because we are assuming, along with Raposo (1989) and Gonçalves (1992), among others, that these verbs are raising predicates, in the sense that they do not -mark the subject, this being an argument of the embedded infinitival verb.So, if the aspectual verb is absent, the final subject co-occurs with the predicate that selects it.
Based on these facts, Raposo (1989) claims that a is a preposition which heads a clausal domain, whose distribution is identical to the distribution of canonical small clauses.Yet, the characterization of a as a preposition poses several problems.First, prepositioned infinitival domains selected by verbs other than the aspectual ones do not occur in contexts like those in (56)-(59).Consider, for example, a verb like insistir (to insist), which takes a complement headed by the preposition em (in): Second, under certain conditions, the a + Inf domain selected by these aspectual verbs is distributionally equivalent to Ger constructions, as it was shown in section 2.2.
Third, no lexical material can interrupt the adjacency between a and the infinitival verb, contrary to what happens in contexts where the infinitival domain is selected by a non-aspectual verb.The relevant contrast is presented in ( 63) and ( 64): (63) *(O) Manoel está a realmente dizer isso.
(EP/BP) (the) João forced (the) Ana A really to-say that Finally, the complement of a cannot be replaced by the demonstrative isso (that) in the context of an aspectual verb (see ( 65)); on the contrary, replacement by isso is allowed in (64), as illustrated in (66): (65) *(O) Manoel está a isso.
(EP/BP) (the) João forced (the) Ana A that The specific behaviour of a in aspectual constructions led Gonçalves  (1992) and Duarte (1993) to recategorize this element as an aspectual (Asp) head that constitutes a discontinuous morpheme with the infinitival marker, -r. 16The aspectual value is obtained when Asp amalgamates with T.

The structure of a + Inf constructions: Differences between EP and BP
As we have shown in the previous section, the a + Inf domain exhibits identical syntactic properties in EP and in BP.The crucial fact is that the preposition a is recategorized as an Asp head in both varieties.Now the question is to discuss the categorial status of the infinitival complement.
Following Gonçalves (1992), Duarte (1993) and Duarte & Gonçalves  (2002), we assume that in the construction under analysis aspectual verbs select a TP domain.One of the most important empirical arguments for this analysis comes from negation.In fact, assuming, as Zanuttini (1996) and Matos (1999), that there is a strong correlation between sentential negation and T, the data presented in ( 67) and ( 68) suggest that the embedded infinitival domain is TP:17 (67) muitos continuam a não acreditar (Natura/Público, par 62255) (EP) many continue A NOT to-believe (68) Rubião continuava a não ouvir nada (NILC, par 111799) (BP) Rubião continued A NOT to-hear nothing Assuming that a is an Asp head and that T projects in the majority of cases, the partial representation of the infinitival domain in aspectual contexts of EP and BP is as follows: In both varieties, the embedded T is dependent on the higher T, that is, the embedded domain cannot introduce a new temporal reference (see, for example, Newmeyer, 1975; Stowell, 1982; Raposo, 1987 and Ambar, 1992).Thus, the occurrence of modifiers with opposite temporal values produces ungrammatical sequences: 19 (70) *Muitos, ontem, continuavam a não acreditar nisso, hoje.
(BP) Rubião, yesterday, continued A NOT to-hear nothing, today On the contrary, whenever the embedded T is independent, the infinitival domain admits a modifier introducing a new temporal reference; as shown by the grammaticality of ( 72): (72) Os professores, ontem, obrigaram os alunos a apresentar o trabalho, hoje.the teachers, yesterday, forced the students to-present the work, today In (72), the adverb ontem (yesterday) refers to the moment when the teachers forced the students to present their work, whereas hoje (today) refers to the moment when the students have to present their work.
Although the structure of the infinitival complement selected by aspectual verbs seems to be identical in EP and BP, the two varieties differ w.r.t. the possibilities of clitic placement.In fact, in EP clitics selected by the infinitival verb can either occur in adjacency to this verb (see ( 73) and ( 74)) or freely climb to the aspectual verb (see ( 75) and ( 76)): (73) outros estão a fazê-lo agora (Natura/Público, par 983669) others are A to-do-CL now (74) (...) as pernas começaram a queixar-se (Natura/Público, par 68832) the legs COMEÇAR A to-complain-CL (75) ainda se estão a realizar os estudos preliminares (Natura/Público, par 13662) still CL are A to-do the studies preliminaries (76) que se começam a dar os primeiros passos (Natura/Público, par 5293) that CL COMEÇAR A to-take the first steps However, Clitic Climbing is very restricted in BP; the preferred option is to keep the clitic adjacent to the infinitival verb, proclisis being the more natural position, according to the pattern of clitic placement in this variety: 20 (77) estava a se queixar [s/he] was A CL to-complain Since there is evidence in favour of T projection both in EP and in BP, the question now is how to derive the attested difference between the two varieties in what concerns Clitic Climbing.
Following Bartra & Suñer (1997) and Gonçalves (1999), let us assume that functional heads can be either active or inert/defective.Assume, also, that the relevant functional head in infinitival complements is the embedded T, which can be active or defective; if active, it checks the verb features, and the following facts are naturally derived: (i) sentential negation is allowed in the embedded domain, as in ( 67) and (68), repeated in ( 78) and (79): 21 (78) muitos continuam a não acreditar(Natura/Público, par 62255) (EP) many continue A not to-believe (79) Rubião continuava a não ouvir nada (NILC, par 111799) Rubião continued A not to-hear nothing (BP) (ii) clitics are licensed in the embedded domain (see Duarte et al., 2001), enclisis or proclisis being the available options, according to the general pattern of clitic placement in each variety, as shown in ( 73) and (77).
Notice that the two arguments also apply to infinitival embedded domains whose T is independent of the higher T and to finite embedded sentences.In ( 80), the embedded T is independent of the higher T and, in such a case, sentential negation (see ( 82)) and clitics (see ( 83)) are admitted in the embedded domain: 22 (82) Os deputados decidiram [não aprovar a lei].
the representatives decided to-approve-CL The same behaviour is attested in embedded finite contexts like the one in ( 81 the teachers decided that the students CL would-hand in today Let us now assume that in EP, the embedded T of aspectual constructions can also be defective in what concerns its V-features.In this case, the embed-22 In (80), the embedded T is independent of the higher one since the tense of the infinitival domain is autonomous from the tense of the matrix; consequently, the infinitival domain can introduce new temporal information.This fact is illustrated in (i), where the co-occurrence of two modifiers with non-overlapping temporal values is allowed: (i) Os deputados, ontem, decidiram aprovar a lei hoje.the representatives yesterday decided to-approve the law today ded V cannot check its features against the embedded T; consequently, the embedded V must rise to the matrix T, as a last resort operation in order to prevent the derivation from crashing.Raising of the embedded V results in the formation of a complex predicate, that is this verb and the matrix one behave like a syntactic unit and they check their features against the same T headthe matrix one. 23The empirical data that support the complex predicate analysis are essentially the following: (i) clitics depending on the embedded verb occur in adjacency to the matrix verb, since there is no verbal host in the embedded domain; this is illustrated in ( 75) and ( 76), repeated in ( 86) and ( 87): (86) ainda se estão a realizar os estudos preliminares (Natura/Público, par 13662) still CL are A to-do the studies preliminaries (87) que se começam a dar os primeiros passos (Natura/Público, par 5293) that CL begin A to-take the first steps (ii) if clitic climbing applies, sentential negation is no longer allowed in the infinitival complement; recall that we have assumed a close relation between active T and negation: (88) *O Presidente disse que muitos o i começaram a não apreciar t i .
the President said that many CL i COMEÇAR-Past A not to-appreciate t i (iii) SE passive is allowed, and, in consequence, the Object of the embedded verb becomes the subject of the matrix; this can be interpreted as a transparency effect, that is, the embedded TP domain does not qualify as a barrier for NP extraction: (89) que se começam a dar [os primeiros passos] SUB (Natura/Público, par 5293) that CL COMEÇAR-Pres A to-take the first steps (90) ainda se estão a realizar [os estudos preliminares] SUB (Natura/Público, par 13662) still CL are A to-do the studies preliminaries From the syntactic point of view, the complex predicate formation is the main difference between EP and BP in a + Inf constructions selected by aspectual verbs.In fact, in the NILC corpus we have found only two occurrences of Clitic Climbing24 and no occurrence of SE passive in adjacency to the aspectuals estar, começar or continuar.This means that in BP the infinitival complement contains the necessary and sufficient functional heads to license the features of the embedded verb, thus avoiding, in a high percentage of cases, the formation of the complex predicate Assume, again, that in BP T projects in the embedded domain.As in EP, this head is dependent on the matrix T, since the temporal specifications of the infinitival complement are dependent on the temporal specifications of the matrix.Following Duarte et al. (2001), we claim that in BP T can also be defective or active, as in EP.However, whereas in EP the defectivity of T forces the complex predicate formation, in BP it has consequences on clitic placement.According to the authors, in BP: (i) if T is active (and in the absence of a proclisis trigger), enclisis is the pattern of clitic placement, as shown by ( 91)-( 93): (91) (…) este estava a dar-lhe notícias do ataque(...) (NILC, par 125019)  this was A to-give-CL news of-the attack (92) (...) comecei a copiá-lo (NILC, par 44824) [I] COMEÇAR-Past A to-copy-CL (93) (…) ela continuava a fugir-me (NILC, par 103867) she continued to-escape-CL (ii) if T is defective, proclisis to the embedded verb is obtained, as illustrated by ( 94)-( 96): (94) (…) ele estava a se queixar (NILC, par Opinião-94 a -opi-2) he was A CL to-complain (95) a burguesia urbana começou a se organizar (NILC, par 131630) the middle-class começar-Past A CL to-organize (96) (...) o Universo continuará a se expandir (NILC, par 97227) the Universe will-continue A CL to-spread In synthesis, in a + Inf constructions dependent on aspectual verbs Syntax meets Semantics.On the one hand, the semantic facts led us to conclude that EP is distinct from BP in the sense that in the latter the aspectual nature of the 'input' generally restricts the occurrence of the a+ Inf construction, contrary to what happens in EP.On the other hand, the syntactic facts allow us to conclude that the infinitival complement is structurally distinct in the two varieties.In fact, although T can be either active or defective in these varieties, the defectivity of this head has distinct consequences: in EP, it forces the complex verb formation, and Clitic Climbing is attested; in BP, the defective T does not prevent clitics from being licensed within the embedded domain but it restricts their position w.r.t. the verb.This fact suggests the following hypotheses about BP (see Duarte et al., 2001): (i) a defective T licenses clitics or (ii) a functional head [-T] non distinct from V (namely Asp, in the constructions under analysis) can also license clitics.The fact that proclisis to participles is the only available option in the context of auxiliary verbs, where no T is projected in the domain of the main verb (see ( 97)), constitutes empirical evidence in favour of the second hypothesis.

Gerund constructions
The data presented in section 2.2.show that in BP, when the aspectual verb takes as its 'input' a (basic or derived) process, it occurs, in a high percentage of cases, with Gerund.On the contrary, whenever the aspectual verb takes as its 'input' a (basic or derived) state, it appears preferably with a + Inf.Recall that in EP, the Gerund construction is productive in some southern varieties, but it is not the preferred option in the standard contemporary variety.
A+Inf and Gerund domains share a significant number of properties.In fact, both domains occur in syntactic isolation (see ( 98) and ( 99)), and as the final subject in the context of predicative verbs (see ( 100) and ( 101)); moreover, the aspectual value is the same in both cases.
(98) [Manoel a dizer isso!] Que estranho!Manoel A to-say that!How strange!(99) [Manoel dizendo isso!] Que estranho!Manoel saying that!How strange!(100) [Manoel a dizer isso] é estranho!Manoel A to-say that is strange (101) [Manoel dizendo isso] é estranho!Manoel saying that is strange The identical behaviour of the two constructions suggests that the Gerund domain is also a projection of Asp.However, in the a + Inf construction this head is lexicalized by a, whereas in Gerund constructions the same head is empty and checks the V-features of the Gerund, after V movement. 25The fact that sentential negation is allowed in some contexts (namely with continuar) suggests that T projects in some gerundive contexts (see, for instance, (102)): (102) o Carrefour e o Sé continuam não vendendo the Carrefour and the Sé continue NOT selling However, we did not find in the NILC corpus the possibility illustrated in (102) with the aspectual verbs estar, ficar and começar.Thus, we propose that two different structures are involved, as we partially represent in ( 103) and ( 104): In (103), T is active, and sentential negation in the embedded domain is allowed, as expected.In this case, proclisisbut not enclisisis obtained in virtue of the negation operator.However, if T is defective or does not project as in (104), proclisis is obtained. 26If this is true, we account for (105): 27 (105) As pessoas estão [me dando apoio nas ruas] (NILC, par 10046) the people are CL giving support in the streets

de + Infinitive constructions
The aspectual verbs having this construction are related to interruption, cessation, conclusion or culmination of an eventuality.Those verbs are the following: deixar de, parar de (stop), acabar de (finish/to come to an end), in EP and BP, and terminar de (terminate) in BP.In both varieties their dissimilarities are fundamentally semantic as they exhibit identical syntactic properties.

The semantics of 'de + Inf' constructions 4.1.1. 'Acabar de' in EP
According to the aspectual nature of the 'inputs', acabar de in EP may be either an aspectual operator or may have a temporal reading.In BP this distinction is very much associated (although not always) with two different verbs: terminar de and acabar de.
As an aspectual operator, acabar de (to finish) selects processes and culminated processes as basic predications.However, its 'input' must be 26 Recall that proclisis is also obtained in a + Inf constructions if T is defective. 27Once again, we follow Duarte et al. (2001) in assuming that in BP either a defective T or a functional head [-T] non-distinct from V (namely, Asp) license clitics in BP.
coerced into a process and the 'output', as it marks the end of a situation, is a culmination (or possibly a culminated process) taking into account the adverbials with which it can co-occur (see ( 106)-( 107)).This aspectual verb cannot combine with states (see ( 108)-( 109)) and when the basic predication is a culmination it has a temporal reading (110).(111) O Rui acabou de ser simpático: ofereceu-me um livro.
the Rui ACABAR-Past DE to-be nice: [he] offered me a book 'Rui has just been nice: he offered me a book' (112) O Rui acabou de sair (neste momento / há pouco).
the Rui ACABAR-Past DE to-knock at the door (in this moment / a while ago) However, it should be noticed that the temporal perspective point is in the relevant interval adjacent and immediately subsequent to the terminus of the eventuality.If this is not the case, with culminations, states and points that localization does not take place as we can see, for instance, in the contrast between ( 114) and ( 115 In BP, apart from acabar de, there is another verb, terminar de (to finish/terminate), which does not act as an aspectual verb in EP.According to the data in the NILC corpus, the first of these two verbs has mainly a temporal reading (see ( 117)) whereas the latter is aspectual (see ( 116)).So, the difference between the two varieties seems to be of a lexical nature, as EP has only one verb for both cases. 28  4.1.3.'Parar de' and 'deixar de'  4.1.3.1.In EP In EP, parar de (to stop) is in some respects similar to the aspectual verb acabar de: the basic predications are processes and culminated processes (see ( 118)-( 119)) and the 'input' is a process.But the 'output' is different, since it does not indicate the end of a situation but its (temporary) interruption.So, its 'output' is a punctual event associated with a ceasing state.However, the 'output' is somehow unstable as it can focus either the punctual event (see (120)) or the ceasing state as in example (121). 28In the 84 occurrences analyzed in NILC corpus with terminar de, only four may be temporal and in 382 with acabar de only twelve are aspectual and forty two seem to be ambiguous. 29In BP acabar may co-occur with Ger, but in this case it is a different verb, parallel to the EP construction acabar por (to end up by).Terminar por is also possible but again it is not an aspectual operator as it is similar to acabar por in EP.
(i) acabam prejudicando a suavidade de sua pele.the Rui stopped DE to-read the book for an hour Deixar de in EP can select states and events as basic predications but its 'input' is different depending on what it selects: when the basic predication is a state, the 'input' is also a state, and when the basic predication is an event the 'input' is a habitual state (see ( 122)-( 123)).Nevertheless, this aspectual verb admits a different kind of 'input', a process, in some restricted cases of processes (possibly some culminated processes), like chover (to rain), rir (to laugh), and ver (to see) where there is not coercion to a habitual state.In this case, it is close to parar de like in ( 124)-( 125).The somehow rare combination with a process as 'input' shows that in EP there is a hesitation between deixar de and parar de.Actually, the 'output' is very similar, that is, a punctual event plus a ceasing state, but while parar de focuses mainly on the punctual event, deixar de focuses on the ceasing state.
she DEIXAR-Past DE to-be nice/to-smoke/to-go out/to-read the paper (124) Deixou de chover.
in that moment the driver DEIXAR-Past DE to-see the car ahead If we compare acabar de, parar de and deixar de, we can say that the first one indicates the culmination of a process, the second one its (possibly temporary) interruption and the third one the cessation of a habitual state, as in ( 126)-( 128).

In BP
There is no noticeable variation in the behaviour of deixar de in BP and EP.As for parar de, it exhibits some differences, inasmuch as it can share EP contexts of both parar de and deixar de.In (129) parar de selects a process whereas in (130) it selects a state.
(129) (…) pediu a um policial que parasse de atirar (NILC, par 9520) asked a policeman that [he] PARAR-Past DE to-shoot (130) aos quais pediu para pararem de ter medo (NILC, par 19554)  to whom [he] asked to PARAR-Past DE to-have fear The possibility of EP parar de of highlighting the ceasing state is much more evident in BP.In the first variety, apart from its fundamental meaning focusing the punctual event, it also allows (although rarely) to focus the ceasing state when supported by measure temporal adverbials.On the other hand, deixar de (associated to a basic or derived state) focuses mainly the ceasing state but it also shows a somehow weak instability when it occurs with processes not coerced into habitual states.These are possibly the reasons why some BP examples with parar de correspond to deixar de in EP, as it happens, for instance, in BP example (130).So, it seems, according to the data, that parar de in BP instantiates two different types: one very similar to EP deixar de, combining with states (see ( 130)) and events, having in this latter case a habitual state as 'input' (see ( 131)), and another one similar to parar de in EP, like in (132).We see then that these two verbs are somehow unstable in both varieties but in a different way.In EP parar de mainly focuses the punctual event, as it occurs very naturally with 'punctual' adverbials, whereas deixar de prefers the ceasing state.In BP parar de seems to compete with deixar de in many cases, although the outcome is still not very clear, according to the data.The following table sums up what has been said:

The syntax of 'de + Inf' constructions
In the preceding section, we have shown that interesting semantic contrasts between EP and BP can also be found in the group of aspectual verbs that select de + Inf.In this section, we aim to show that (i) de + Inf is distinct from a + Inf complements in both varieties and (ii) de + Inf complements exhibit identical syntactic properties in both varieties.

General syntactic properties
Consider the following sentence, illustrative of de + Inf constructions in both EP and BP: (136) Os meninos deixaram de ler o livro the children DEIXAR-past DE to-read the book The infinitival domain of this sentence is syntactically distinct from a + Inf constructions in the two varieties.First, contrary to a + Inf, de + Inf does not occur in syntactic isolation, as illustrated by the contrast between ( 137) and ( 138 The contrast between the two constructions suggests that de cannot be seen as another lexicalization of the Asp head.In turn, if we assume that de keeps its prepositional nature in both varieties, we account for the similarity of de + Inf complements selected by aspectual and non-aspectual verbs. 30Thus in both cases, (i) the infinitival complement does not occur in syntactic isolation (compare ( 137 The fact that proclisis is preferred over enclisis in the context of de + Inf (the EP and the BP corpora contain a higher percentage of proclisis) means that de is a heavy word, as it is usually assumed for prepositions.

The structure of 'de + Inf' complements
The data presented in the previous section suggest that de + Inf complements have the structure partially represented in ( 154 The embedded T is dependent on the higher one, since it cannot introduce a new temporal reference: (155) *O João, ontem, deixou de ler o livro, hoje.
the João, yesterday, DEIXAR DE to-read the book, today In this case, non-finite T is active both in EP and in BP.Thus, the embedded verb checks its features within the embedded domain and, so do the clitics.As expected, and contrary to what happens in a + Inf construction, Clitic Climbing is disallowed even in EP: 32 (156) *Os meninos deixaram-no i de ler t i .
the children DEIXAR-CL to-read t i In synthesis: (i) de is distinct from a in both varieties, in the sense that the former is a preposition whereas the latter is the lexicalization of an Asp head; (ii) de + Inf constructions have essentially the same syntactic structure in both varieties; the differences are fundamentally semantic and/or lexical distinctions, such as the temporal vs. aspectual one, found in terminar de and acabar de in BP in contrast to acabar de in EP or the different choice made by both varieties of parar de /deixar de.

Concluding Remarks
The analysis of the two major types of aspectual verbs in EP and BP, exhibiting a + Inf /Ger and de+Inf structures, showed that the first type is of a great interest on semantic and syntactic grounds as the almost exclusive selection of a + Inf in EP is due to the fact that in this variety the embedded T can be either active or defective.In the case that T is defective, the embedded verb raises to the higher T in order to check its features, thus forming a complex predicate with the aspectual verb.On the contrary in BP a + Inf and Ger structures may compete, depending on the class of the selected aspectual verb.In this variety the infinitival complement is structurally distinct from the one in EP as it contains the necessary and sufficient functional heads to license the features of the embedded verb, which avoids the formation of the complex predicate.At the same time we can see that some verbs occur mainly with Ger, some others with a + Inf and one verb (continuar) selects both structures.This corresponds in an extensive way to differences in the semantic nature of the 'input' of these verbs before they operate the aspectual transition.In fact, the selection of Ger is very much related to an eventive (process) 'input', and the selection of a + Inf to a stative 'input'.The verb continuar, which selects both structures, has mainly an eventive 'input' when it selects Ger and a stative one when it selects a + Inf.
The structure de+Inf, although different from a + Inf in its syntactic analysis, does not show any syntactic discrepancies in the two varieties.These are left for the lexical and semantic domain as there are two pairs of verbs in BP that show some significant differences.The verb acabar de in EP corresponds to the pair acabar de /terminar de in BP exhibiting a temporal /aspectual contrast.Parar de and deixar de, show in the two varieties some distinct 'inputs' as the first one in BP is similar in some respects to the latter one in EP, having some semantic overlapping.These BP aspectual verbs seem to be much more unstable than the EP ones.

( 9 )
*O Pedro começou a ter um B.M.W.. (non-phase state) the Pedro COMEÇAR-Past A to-own a B.M.W. (10) A escalada começou a ser perigosa.(phase state) the climbing COMEÇAR-Past A to-be dangerous (11) O cavalo começou a correr.the horse COMEÇAR-Past A to-run (12) Os operários começaram a construir a casa.the workers COMEÇAR-Past A to-build the house' (13) O Pedro começou a ganhar a corrida.the Pedro COMEÇAR-Past A to-win the race (14) ??/*O comboio começou a chegar.the train COMEÇAR-Past A to-arrive

( 60 )
Os meninos insitiram em ver o jogo.the children insisted EM to-watch the game In this case, the prepositioned domain does not occur either in syntactic isolation (see (61)) or as the final subject in predicative contexts (see (62)): (61) *Os meninos em ver o jogo!Que estranho!the children EM to-watch the game!How strange!(62) *Os meninos em ver o jogo é estranho!the children EM to-watch the game is strange!

Table I :
Semantic Effects of Aspectual Verbs in EP

Table II :
a + Inf and Ger: distribution of BP Aspectual Verbs ).

Table IV :
Deixar de and parar de in EP and BP