Agreement patterns with a gente in Portuguese

This paper aims to provide a description of different agreement patterns with ‘a gente’ (we) in European Portuguese (EP) and Brazilian Portuguese (BP). We consider the verbal agreement patterns and the gender-number agreement patterns in predicative structures. It will be shown that the behavior of ‘a gente’ is quite similar in both varieties of Portuguese, the differences being mainly in quantitative than in grammatical terms. The data analyzed provide evidence that, differently from the binding domain, the agreement domain seems to present no restriction to the actuation of grammatical and semantic features of ‘a gente’. The theoretical claim is that EP and BP have the same grammar of ‘a gente’, which can be described from a single set of phi-features.

According to Lopes (1999), the grammaticalization process of the noun gente (people, the folks) to the pronoun a gente (we) has been triggered by the fact that the noun gente presupposes "a group of people".Depending on the context of use, it could include or not the speaker who is necessarily a "person".In (1), we can see an example that allowed this dual interpretation: (1) E os tigres, em tanta cantidade (por não haver descampados), que, em se metendo a rês no mato, não sae, e o mesmo risco corre a gente, se não anda acompanhada, e pelos rios e lagos dos jaguarés... (Noticiário Maranhense, 18 th century) a.And the tigers, in such great amount (due to lack of open fields), which, in hiding in the bushes, don't leave, and the same risk we are at, if not alone, and by the rivers and lakes of the skunks... b.And the tigers, in such great amount (due to lack of open fields), which, in hiding in the bushes, don't leave, and the same risk the people are at, if not alone, and by the rivers and lakes of the skunks...
Although we can interpret a gente as "the people, everybody," the reading "including the speaker" is available.In (1), any person is in danger in the bush with tigers, including the 1 st person.Another favoring aspect of the grammaticalization is the fact that the pronoun nós (we) introduces the notion of "an amplified first person", i.e. the speaker and one or more non-speakers (Benveniste, 1988).The pronoun nós (we) can refer to a determined group (eu + você~tu/ele (1SG + 2SG/3SG), as in (2), or to an indeterminate group (eu + todo mundo (1SG + everybody)), as in (3).In this context, a gente (we) was inserted as a pronominal form: (2) Nós, meu irmão e eu, gostávamos de brincar no quintal quando éramos crianças.
We, my brother and I, liked-1PL to play in the backyard when we were kids.
(3) Nós precisamos resolver o problema da educação.We need-1PL to solve the problem of education.
In accordance with Lopes (1999Lopes ( , 2003)), the pronoun a gente (we) conserves, on the one hand, some historical morphosyntactic features of the noun gente (people); on the other hand, it acquired some intrinsic properties of the personal pronouns.Considering the person and number features, the pronoun a gente shows mixed features: it is grammatically specified for 3SG and semantically for 1PL.
These two sets of features -grammatical (3SG) and semantic (1PL) -are available in grammar and would be responsible for generating different patterns of agreement in both EP and BP.Regarding the verbal agreement, a gente (we) can occur with verbs that present morphology of 3SG (4), 1PL (5) and, less often, 3PL (6) : (4) A gente canta muito.
We sing-3SG a lot.
(6) A gente cantam muito.We sing-3PL a lot The agreement patterns in predicative structures, with adjectival and participial forms, can be singular (7) or plural (8).Alike the pronoun nós (we), a gente does not present a formal specification for gender, but it may trigger an interpretation of gender according to the referent, as the masculine and feminine data show.
We are tired-FEM.PLAccording to Menuzzi (2000), both sets of features are active, but there seems to be some restriction regarding their activation in the binding contexts: in local domains, grammatical features are used in anaphoric forms (9), whereas semantic features are activated in the non-local domains (10): (9) a.A gente i viu-se i no espelho.
We-3SG i looked (…)self.3SGi in the mirror.Costa & Pereira (2012) show that Menuzzi's analysis (2000) for binding contexts can be extended to patterns of verbal and predicative structures agreement with a gente in EP.Therefore, a constraint is expected on the actuation of grammatical and semantic features of a gente depending on the domain: grammatical features are activated in the local domain and semantic features in the non-local domain.Thus, Costa & Pereira (2012) consider the results of Pereira (2003) and verify which patterns are found, if they are dialectally restricted1 or if they are actually optional.
Regarding the verbal agreement, Costa & Pereira (2012) show that the patterns of 3SG (third person singular) and 1PL (first person plural) are found in PE, as in ( 4) and ( 5).The agreement pattern with the 3SG is the most common in all Portuguese dialects.In these cases, since the subject is in [Spec,IP] and the verb is in I(nflection), the operation occurs locally.Thus, the grammatical features are expected.This hypothesis would explain the most frequent pattern (3SG), but not the pattern with verbs in 1PL.
Since the patterns with 3SG and 1PL coexist in all dialects of EP, we cannot postulate a case of grammar or structure competition.In this case, the two patterns must be generated by the same grammar.In this sense, in order to explain the constructions in (5), in which a gente agrees with 1PL, the authors argue that the pronoun a gente is generated in apposition to a null 1PL pronoun2 (11).Therefore, the verbal agreement, which takes place in local domain, occurs between pro [1PL] and the verb.
Concerning predicative structures in EP, the pattern illustrated in (8), with plural adjectives, is the most frequent one, both in the standard EP as in dialectal areas.The small clause is a full-phi domain and, therefore, one phase.Thus, the relation between predicative and subject, in ([Spec,IP]), is non-local.In this case, the activation of the semantic features of a gente justifies the plural number feature in the adjective.
The subject moved and I(nflection) are in a local relationship, which justifies the actuation of grammatical features of a gente in the verb in the third person singular, as seen in ( 8).As a gente is a pronoun, such as nós (we), the manifestation of gender in predicative is directly associated with the gender of the referent, not being the gender a syntactically relevant feature.
In summary, Costa & Pereira (2012) reconcile the actuation of the phi--features in the agreement patterns to the hypothesis formulated by Menuzzi (2000) for binding contexts.The results can be summarized in the following table: As it was seen, it is assumed that the verbal agreement patterns with 3SG and 1PL, not being dialectally restricted, should be analyzed internally to the same grammar.Therefore, the same orientation should be given to the patterns with number agreement in predicative structures3 .The pattern with singular adjectives (7), found mainly in BP4 (Vianna, 2006(Vianna, , 2011)), cannot be explained by Costa & Pereira's analysis, since in the non-local domain the actuation of grammatical features is not expected.
Accordingly, we claim that the case described above is an evidence to postulate that the restriction observed by Menuzzi (2000) would work only in binding contexts in EP and BP, not in the cases of agreement.In addition, other issues show us that this analysis, which dispenses the existence of distinct grammars of a gente in EP and BP, may be feasible: (i) the pattern with verbs in the 3PL, as shown in (6), found in EP and BP (Costa & Pereira 2012;Vianna 2006), in which we verify the activation of the grammatical person feature (3P) and the semantic number feature (PL), could not be generated; (ii) the grammatical property that motivates the apposition of the DP a gente to a null 1PL pronoun needs to be explained; and, finally, (iii) in terms of lexicon, it is necessary to explain, due to the existence of two sets of features (grammatical and semantic), if there are two distinct pronouns a gente or if the mixed pronominal compositionality is provided in a single lexical item.
Considering the analysis of patterns of verbal agreement and agreement in predicative structures with a gente, in EP and BP, our goal is to demonstrate that the differences, especially in terms of frequency, between EP and BP, do not suggest two different grammars.We interpret the quantitative differences as a result of the implementation of the pronoun a gente in both varieties of Portuguese.Therefore, we argue that the agreement patterns found in both EP and BP can be taken as an argument in favor of the hypothesis of autonomous domains, as Costa, Moura & Pereira (2001) initially proposed: there is a (i) binding domain, in which the grammatical features are active in local contexts and the semantic features act in non-local contexts, corroborating with the analysis of Mennuzi (2000); and (ii) an agreement domain, in which there is no restriction for the activation of features, since both grammatical as well as semantic features could be activated.Finally, we show that the grammar of a gente both in EP and in BP can be described in a unified and more economical way without the need for the actuation of grammatical or semantic features.The same set of hierarchically organized features (Béjar, 2008) may explain the different agreement patterns produced.
In the following section, we present the results of the analysis of verbal agreement patterns and agreement in predicative structures with a gente, in spoken samples and written tests, in EP and BP.With such description of the data, we seek to answer the following questions: (i) What patterns of verbal agreement and agreement of gender and number in predicative constructions are found and what patterns are more productive with the pronoun a gente in EP and BP? (ii) How to analyze the similarities and differences between EP and BP? (iii) Do the agreement patterns found in EP and BP pose a difference in terms of grammar?How to propose an economic description for the grammar of a gente?

Results in agreement patterns with a gente
We analyzed in spoken samples and written tests in EP and BP all instances in which the pronoun a gente established an agreement relation with a verb, adjective or participial form.
Concerning the corpora used, the samples of spoken language come from the Estudo comparado dos padrões de concordância em variedades africanas, brasileiras e europeias project, available at http://www.concordancia.letras.ufrj.br/,and includes representative interviews of European Portuguese (represented by three different geographical areas: Cacém, Oeiras and Funchal) and Brazilian Portuguese (represented by two cities in Rio de Janeiro: Copacabana and Nova Iguaçu).Written tests of gap filling, in turn, have been applied exclusively to Portuguese informants, and the results obtained were compared with the results of Vianna (2006), who applied the same type of test among Brazilian informants.In such forms, the informants should fill in the blanks with verbal and adjectival/participial form in agreement with a gente, as if they were in a casual conversation, as illustrated in ( 12): (12) Lá no bairro, a toda hora há festa!A gente _____________ (comprar) doces, salgados e muito vinho.No final, a gente _____________ (chegar) sempre bêbad___ a casa.There in the district, we party all the time!We _____________ (to buy) sweets, salty snacks and a lot of wine.In the end, we always _____________ (to arrive) drunk ___ at home.
In this paper, we do not intend to carry out a stricto sensu description within the variationist framework (Weinreich et al., 1968;Labov 1972Labov , 1994)), due to the small number of predicative structures with a gente that are found in samples of spoken language and even in written tests.It would be impossible to use the methodological tools of the Variation Theory, due to the large number of empty cells.Therefore, we used the program of variant rules Goldvarb 2001 (Robinson et al., 2001) only to control the percentage of occurrences.
In section 2.1, we present the results obtained in the analysis of spoken language and written tests, comparing European and Brazilian varieties, with regard to the patterns found in verbal agreement.In section 2.2, we will present the results concerning the variable behavior in predicative structures, in spoken samples and written tests, in EP and BP.II, considering the samples of spoken and written tests, we found, in the two varieties of Portuguese, three patterns of verbal agreement with a gente: (i) with the verbs in 3SG ( 13), (ii) in 1PL ( 14) and (iii) in 3PL (15), pattern found only in BP5 .As we can see, the difference between BP and EP, except for the very little productive pattern with 3PL, concerns only to the productivity of each pattern of agreement.Both in spoken samples and in written tests, the pattern with 3SG is more productive (EP spoken samples: 82%; BP spoken samples: 99%; EP written tests: 65%; PB written tests: 81%).Comparatively, we noticed a higher proportion of 1PL with a gente in EP.

Verbal agreement patterns with a gente in spoken samples and written tests: EP and BP As indicated by the results in Table
Regarding the verbal agreement patterns found, we observe, in the table III, how the actuation of person and number features is: the CORDIAL-SIN corpus (Dialect Corpus for the study of syntax), this pattern also occurs in EP, being locally restricted to the insular dialect of S. Miguel, in Azores.According to Costa & Pereira (2012), the only pattern generated would be with verbs in 3SG, since the verbal agreement is an operation that occurs in the local domain (with the activation of grammatical features of number and person of a gente).The pattern with 1PL could only be produced, according to the authors, if we assume an alternative structure to a gente, as a DP generated in apposition to a pro that contains the 1PL features.However, neither of the two possibilities described above is applicable to the pattern with 3PL, in which we verify the activation of the grammatical person feature and the semantic number feature.
In this sense, our proposal has to be able to explain how a single pronominal compositionality can generate, in the same domain, the three patterns found.

Predicative structures agreement patterns with a gente in spoken samples and written tests: EP and BP
We found four agreement patterns in predicative structures with a gente: feminine-singular, feminine-plural, masculine-singular and masculine-plural: written tests 46/344 -13% 9/344 -3% 242/344 -70% 47/344 -14% In EP, both in spoken samples and written tests, the masculine-plural pattern is predominant, with 43% and 64%, respectively, as shown in the example ( 16).In spoken data, the feminine-plural pattern ( 17) is also productive, with 36% of the total data, followed by cases of masculine--singular, with 21% of the occurrences.In written tests, the other patterns were also not significant: feminine-singular (18), masculine-singular ( 19) and feminine-plural, with 16%, 13% and 7% of the total data, respectively.( 16 We were small-MASC.SG Regarding BP, the most productive pattern in spoken samples and written tests is masculine-singular, with 90% and 70% of the data, respectively, as shown in the example (20).Despite the low productivity, the spoken samples also register the feminine-singular pattern (21), with 10% of the occurrences.In written test, masculine-plural (22), feminine-singular and feminine-plural patterns (23) are also verified, with 14%, 13% and 3% of the total data, respectively.
(20) ... a gente sempre teve uma relação muito boa ... de  We were more neat-FEM.PL Considering the overall results, we can say that the patterns of masculine and feminine in the plural are more frequent in EP, whereas in BP patterns of masculine and feminine in the singular predominate.In order to refine this analysis, we will now present the role of gender and number features separately.

Patterns of gender agreement in predicative structures with a gente in spoken samples and written tests: EP and BP
To analyze the gender, we adopt the approach used by Lopes (1999Lopes ( , 2003) ) and control four possible references the pronouns can take: (i) generic reference: when the referent is a general category or an indeterminate group, such as the people; (ii) mixed reference: when, in discursive context, it is clear that we have in mind one specific group, necessarily encompassing men and women; (iii) women-exclusive: when the referent is a group composed exclusively of women; and (iv) men-exclusive: when the referent is a group comprised only by men.
Table V shows the results for the control of the gender due to the semantic reference in the spoken samples and written tests in EP: In EP, the feminine pattern was categorical in the spoken samples and more productive in the written tests, with 74% of women-exclusive category.On the other hand, the masculine pattern was more productive in men--exclusive, mixed and generic groups in both samples.
The results for BP are shown below: In BP, the results related to the actuation of gender feature confirm, in general, the results found in EP, with a predominance of the feminine in women-exclusive groups in both samples.The masculine pattern, in the same way as EP, seems to be more productive in men-exclusive, mixed and generic groups, with frequencies ranging between 89% and 100%.
In summary, the actuation of the gender feature shows similar patterns in the two varieties of Portuguese.The gender in predicative structures seems to be conditioned by the type of reference: feminine in groups composed exclusively by women and masculine in men-exclusive, mixed and generic groups.

Agreement patterns in predicative structures and the pronominal status of a gente: resuming a discussion
It is interesting to note that the results described above for the behavior of a gente in predicative structures could be an argument in favor of the pronominal status of a gente.Since a gente is a 1PL pronoun, the same way as nós (we), it is not formally specified for gender, but can present different gender agreement patterns depending on the referent.
On this issue, there are two different points of view: on one side, a gente is argued to have a pronominal status; on the other, a gente does not behave as a pronoun, by as a complex DP.We will resume this discussion below.
According to Costa & Pereira (2012), the linguistic literature converges to assign a pronominal behavior to a gente, since, due to the grammaticalization process it has gone through, this form behaves today the same way as other pronominal phrases, as argued by Lopes (1999Lopes ( , 2003)).
In syntactic terms, Menuzzi (2000) uses the following arguments for the pronominal status of a gente: (i) a gente obeys the Condition B of the Binding Theory, because it cannot be locally bounded by an antecedent, and may, however, present and antecedent in a matrix sentence (24); (ii) in embedded clauses, a gente occurs preferably null when is co-indexed with the subject of the matrix sentence ( 25); (iii) a gente cannot be modified by adjectives (26); and, at last, (iv) a gente may have an arbitrary reading (27): For this reason, Taylor (2009) does not consider a gente as a pronominal form, but as a complex DP with a pluringular status, within the terms of Den Dikken (2001), i.e., a grammatically singular expression that may be referentially plural7 .Therefore, the complex structure of a gente is composed by a DP in apposition to a null 1PL pronoun.It is interesting to note that the grammatical features of the noun head are not compatible with the grammatical features of the DP: (30) [ DP pro [1PL] [ DP a [ NP gente [SG] ]]] This proposal would account, according to Taylor (2009), to the two cases that are counter-arguments to the pronominal status of a gente.Once a gente is not a pronoun and would be in apposition to a pro head, the structures in (28a) and (29a) could not be generated, because numerals and bare plural nouns occupy the same appositive position of a gente.
However, Costa & Pereira (2012) argue that the analysis of Taylor ( 2009) is not incompatible with the pronominal status of a gente.Considering the proposal of Taylor (2009), the authors assume that a gente is an autonomous DP, with the difference that the determiner and the noun do not occupy different positions (31), but they form a complex allocated in D, position occupied by the personal pronouns (32) (Abney, 1987) and that may, in cases of verbal agreement with 1PL, be generated in apposition with a 1PL pro (33): Following Chomsky (2001) that the DPs are phi-complete domains or phases, Costa & Pereira (2012) argue that the agrammaticality of (28a) and (29a) can be explained by the fact that the pronoun a gente followed by numerals and bare plural nouns, constructions which show a non-coincidence of number features, belong to the same phase, i.e., the same local domain where there is the activation of the grammatical features of 3SG (34).It is important to note that if a new phase is introduced, there is no conflict between the number features of a gente and the numerals and bare plural nouns, since the correlation would occur in a non-local domain, and the semantic features of 1PL are activated ( 35 Thus, Costa & Pereira (2012) conclude that the analysis proposed by Taylor ( 2009) may be reconciled with a pronominal status for a gente.The agrammaticality of the cases above (co-occurrence with numerals and bare plural nouns with a gente) are determined due to a restriction of agreement and not related to the categorial status of a gente.This means that the agrammaticality of "*a gente dois vai ao cinema" (*we two go to the movies) and "*a gente portugueses canta muito" (*we Portuguese sing a lot)constructions perfectly grammatical with nós (nós dois vamos ao cinema / we two go to the movies; nós portugueses cantamos muito / we Portuguese sing a lot) -cannot be seen, under no circumstance, as an argument to consider that a gente does not have a pronominal status, as stated by Taylor (2009).
In favor of the position taken by Costa & Pereira (2012) against the proposal of Taylor ( 2009), we argue here that the analysis of gender agreement patterns in predicative structures and of the verbal agreement patterns, rather than the tests of Taylor (2009), can quite effectively clarify the pronominal status of a gente.We claim that the variable gender patterns in predicative constructions is a particular property of the personal pronouns.Besides that, the patterns of verbal agreement with verbs that show morphological marks of 1PL are also an irrefutable empirical evidence for the presence of a person feature in a gente, an exclusive information of the pronominal categories.The co-occurrences of the quantifiers (nós dois/*a gente dois) and the appositive structure with bare plural nouns (nós portugueses/*a gente portugueses) are agrammatical for a gente and comprise, in our view, exceptional and not regular cases with pronominal items.As shown by Costa & Pereira (2012), the coexistence test with bare plural nouns would also result agrammatical for 3PL (*Eles portugueses gostam de cantar / They-MASC Portuguese-MASC.PL like-3PL to sing).
Different from the noun phrases that may present determiners (esta gente / these people) and have adjective modifiers (gente bonita / beautiful people), the pronouns are almost categorically comprised by an isolated head (eu cantei;*o eu cantei;*eu bonita cantei / I sang; *the I sang; *I nice sang).
The most relevant point of this proposal is that the pronominal status of a gente is given by (i) the insertion of the semantic person feature [1P] and (ii) the variable gender agreement in predicative constructions when a gente is interpreted as a pronoun.This property is common even in personal pronouns which suffer gender flexion such as ele/ela/eles/eles (he/she/they-MASC/they--FEM).Likewise, the pronouns eu, tu/você, nós, vós/vocês (I, you-SG, we, you--PL), that are not formally specified for gender, activate semantic interpretation of gender in predicative constructions, according to the referent: (36) Eu estou incrivelmente irritada com você.

Patterns of number agreement in predicative structures with a gente in spoken samples and written tests: EP and BP
Table VII shows the actuation of the number feature in predicative structures in EP and BP: Regarding the actuation of the number feature in agreement operations with a gente in predicative structures, data from EP and BP point to divergent results, even though the two patterns -SG and PL -are registered.The pattern in the plural is predominant in EP.BP, in turn, shows the pattern in the singular preferably in both samples.
For Costa & Pereira (2012), the only pattern that can be explained is the plural, given that the actuation of semantic features is in non-local relations, as the relation between the subject and the predicate.Thus, once the two patterns are found in the two varieties of Portuguese, our proposal has to be able to explain how these two patterns can be generated from an economic description of a gente.
The table below gives an overview of the agreement patterns found in EP and BP: Our results show behaviors that, although quantitatively different, points to grammatical similarities between the two varieties of Portuguese.
With respect to verbal agreement, although the combination of a gente with verbs in 1PL is more frequent in EP than in BP, in both varieties the verbal agreement pattern is more productive with verbal forms of 3SG.
The control of semantic referents showed that the agreement of a gente occurs both with the masculine pattern and the feminine pattern in EP and BP.The masculine pattern predominates in men-exclusive, mixed and generic groups, whereas the feminine pattern prevails in structures that refer to groups formed exclusively by women.These results are consistent with the expected behavior of a 1PL pronoun.Assuming a pronominal behavior for a gente, the distribution patterns of gender depends on the nature of referent, since the 1PL pronouns a gente and nós (we) are not specified for the gender feature, but can activate masculine or feminine semantic gender.
The major difference between the two varieties seems to be in number agreement in predicative structures.While the plural is more frequent in EP, the agreement with the singular predicative characterizes the behavior of a gente in BP.

Towards an economic description: the pronominal compositionality of a gente
The results presented in the previous section show that, although there are quantitative differences, which may be related, in our view, to a matter of implementation of a gente in the two varieties of Portuguese, the grammar of a gente seems to work the same way.
The only problematic case to sustain this view was the fact that it is more frequent in BP the presence of adjectives and participles in the singular predicative constructions, which, according to the proposal of Costa & Pereira (2012), would not be expected since, being a non-local domain, there is an activation of the semantic features of a gente, not their grammatical features.
In general, the possible patterns in EP and BP are exactly the same.Therefore, we will work with the idea that linguistic evidence found in EP and BP are not robust enough to assert that there is a constraint, depending on the domain, to the actuation of grammatical features or semantic features.
These evidences could be taken in favor of a distinction between domains, as proposed by Costa, Moura & Pereira (2001).The hypothesis of autonomous domains could be used to consider a unified grammar of a gente in EP and BP: in the (i) binding domain, the grammatical features are overt in local contexts and the semantic features act in non-local contexts, according to the analysis of Mennuzi (2000); in the (ii) agreement domain, there is no clear restriction for activation of features, since both grammatical and semantic features could be activated.In other words, the restriction observed in the actuation of grammatical and semantic features only apply to cases related to the Binding Theory.
Although we know that this hypothesis lacks a more refined description, we assume it on the basis of linguistic evidence found in EP and BP.Taking into consideration that the patterns with 3SG and 1PL are found in cases of verbal agreement and patterns of singular and plural predicative structures are also registered in both varieties, our aim is to provide a more economical description, in terms of grammar, for the pronominal compositionality of a gente which accounts for the possible patterns.
Up here, we have worked with the pronominal compositionality of a gente that considers the joint action of grammatical and semantic features.Sometimes some features are activated, other times another are 8 .As a gente goes through a grammaticalization process, it loses some original nominal properties and acquires some pronominal properties.This proposal explains the different patterns found in the samples, regarding verbal agreement as well as gender and number agreement in predicative structures.
In order to propose a more economic model that takes into account the cases of a gente discussed, we assume that a pronoun must be specified, in the lexicon, with a single set of phi-features of person and number, and this set must be able to express both the grammatical and semantic content of the pronoun.The joint action of grammatical and semantic features of person and number would not make sense for other personal pronouns, since there is a full correspondence of features9 : To do so, we use the feature geometry proposed by Béjar (2008).Besides organizing hierarchically the pronominal features, this proposal also predicts the agreement patterns that may be produced.
The proposal of Béjar ( 2008) is motivated by some restrictions that occur in some cases in verbal agreement, when there is no complete matching between the features of the goal and the probe.For this proposal, the traditional person and number pronominal phi-features are assumed to be non-primitive, being represented by different types of feature bundles.As regards the person feature, the inventory adopted is {[π], [participant], [speaker]}; for number feature, the inventory is {[ω], [plural]}.These features are private, which means that their absence is interpreted as a negative value.
Considering the person feature, all persons -1 st , 2 nd and 3 rd -are considered [π].Since only the 1 st and the 2 nd are participants of the speech act, they also have the feature [participant].Finally, the feature [speaker] will be responsible for differentiating the 1 st and the 2 nd .These features are in an entailment relationship.So, what is specified for the feature [participant] is specified for [π], and what is specified for [speaker] is also specified for [participant].The same procedure is applied to the number feature: singular and plural are considered [ω], the least specified number feature; plural is also specified for [plural] feature.In other words, everything that is specified for [plural] is also specified for [ω].In this sense, the compositionality of the pronoun a gente could thus be defined: In order to have the matching of grammatical information, overt on the verbal morphology, between the goal (subject) and the probe (verb), there must be the matching of at least one of the features, and the probe features must be equal to or a subset of the goal features.Tables X e XI summarize the possible matching between the goal and the probe, regarding person feature and number feature, respectively: Considering this proposal, in which there is a unified treatment for the pronominal compositionality of a gente, the three patterns of verbal agreement found in the samples with verbs in 3SG, 1PL and 3PL can be explained: Regarding the patterns found in predicative structures, in which there are agreement patterns with singular or plural, we have: The same analysis can be applied for the gender feature.It is known that the primitive personal pronouns in Portuguese, inherited from Latin forms ego, tu, nos and vos are not formally marked for gender.The gender is, thus, possibly expressed in adjectival and participial forms of predicative structures.Although the morphological expression depends on the gender of the referent, the different possibilities to express masculine or feminine have to be previously provided in the pronominal compositionality 10 .In this sense, there is some identity between the patterns found in the two varieties of Portuguese, although there are slight quantitative differences, with a prevalence of morphological expression of masculine in men-exclusive, mixed and generics groups; and of feminine in women-exclusive groups.
In summary, although linguistic evidences indicate quantitative differences, we argue that the similarity of the agreement patterns found with a gente enables us to provide a unified treatment to the varieties of Portuguese.Using the pronominal compositionality that replaces the joint action of grammatical and semantic features is to provide a more economic treatment to the behavior of a gente in Portuguese.The pronoun a gente, with specification [π participant speaker] [ω plural] for person and number features, respectively, can generate different patterns of verbal agreement, as the patterns with verbs which show morphology marks of 3SG, 1PL and 3PL.Moreover, its configuration with regards to the number feature would also generate the patterns in the singular and plural found in predicative structures.Since a gente, as well as nós (we), does not present a specification for the gender feature, the two possibilities, masculine and feminine, are 10 By the proposal of Duarte et al (2002), inspired by the pronominal compositionality of Harley and Ritter (2002), a gente is defined by the following features: [+speaker +proximity, Group, Class].It is interesting to observe that the "class" node, responsible for coding the gender, is not specified in a gente, as occurs with the pronouns eu (I), tu/você (you-SG), nós (we) and vocês (you-PL), in opposition to the pronoun ela (she), that presents the following configuration: [Referential, Group, +feminine].
provided for the pronominal compositionality, and the referent of the pronoun is the factor that determines the morphological expression of gender in predicative structures.

Final considerations
Although quantitative differences are found in the two varieties of Portuguese, the analysis of the behavior of a gente in predicative structures and patterns of verbal agreement, considering the person, number and gender agreement, points to the existence of the same agreement patterns in both EP and BP.These results weaken the argument of Costa & Pereira (2012) that there is, as in the binding contexts, a restriction of the actuation of grammatical and semantic features depending on the domain.
In this sense, we argue that (i) there are no differences in the grammar of a gente in EP and BP; and that (ii) the restriction observed by Menuzzi (2000) is restricted to the binding domain, but cannot be applied to the agreement domain.In the binding domain, the grammatical features are active in local relations and the semantic features act in non-local relations.In the agreement domain, in turn, grammatical and semantic features are active in both local relations and non-local relations.
With regard to the pronominal compositionality of a gente, the application of the proposal of Béjar (2008) allows us to explain how the different agreement patterns can be generated by the same pronominal geometry, following a feature hierarchy.
Applying this proposal to the restrictions in the binding contexts described by Menuzzi (2000), it is still necessary, on the one hand, to investigate the reason why the least specified features ([π]; [ω]) are activated in local domains and the most specified features ([π participant speaker]; [ω plural]) act in the non-local domains.On the other hand, new studies could examine more closely whether these restrictions in fact exist, since data such as "A gente i vimo-(*se/nos) i no espelho" (We i saw-1PL …self i [*3SG/1PL] in the mirror), in which there is the actuation of more specified features in the local domain, are registered in varieties of Portuguese, as shown by Costa & Pereira (2012).
The differences between EP and BP with respect to the pronoun a gente would be related to a matter of implementation of the new pronominal form in the two territories, faster in Brazil than in Portugal, and not to the pronominal status of the new form or to a difference in the grammatical level.

Table II .
Patterns of verbal agreement with a gente in spoken samples and written tests: EP and BP.

Table III .
Person and number features of a gente according to the patterns of verbal agreement.

Table IV .
Patterns of agreement in predicative structures with a gente in spoken samples and written tests: EP and BP.

Table V .
Patterns of gender agreement for a gente in predicative structures in spoken samples and written tests: EP.

Table VI .
Patterns of gender agreement in predicative structures with a gente in spoken samples and written tests: BP.

Table VII .
Patterns of number agreement in predicative structures in spoken samples and written tests with a gente: EP and BP.

Table VIII .
Summary of results.

Table IX .
Control of grammatical and semantic person and number features in primitive pronouns.

Table X .
Success or failure of Agree with probe-goal pairs (person features).

Table XI .
Success or failure of Agree with probe-goal pairs (number features).

Table XII .
Patterns of verbal agreement with a gente.

Table XIII .
Patterns of number agreement with a gente in predicative structures.