PhiN 36/2006: 23

Mary A. Kato (Campinas)

Generative grammar and variation theory: a happy marriage in the description of Brazilian Portuguese

Traditionally, most generative work dispenses with empirical data from E-language and most empiricist researches, in their turn, ignore formal theories of grammar, which have I-language as their object. Brazilian linguistics developed at the University of Campinas have a unique tradition in combining the two approaches in several projects. The deceased Fernando Tarallo and myself have initiated this line of research in Brazil, developing two projects: a diachronic study of Brazilian grammar and a description of spoken Brazilian Portuguese (SBP). After his premature death, I continued with this tradition, and at the moment I am coordinating a binational project on the comparison of European and Brazilian Portuguese grammars.1 In this presentation I will be reporting some of the results of the projects in which I was involved, which show the interesting questions and results that one can obtain when we combine both approaches in the analysis of corpora.

1. Introduction

It was our belief and faith that, while the Principles and Parameters theory could provide us with predictions as to what kinds of correlation to find in real data, a well conducted empirical analysis could reveal new unaccounted for facts and correlated phenomena. Methodologically, formal grammars help the researcher to constrain the universe of formal factors to be considered and to formulate more explicit questions. The variationist method, in its turn, allows us to use not only syntactic variables, but also constraints of other modules of the language faculty and also of other domains of the mind. It can also distinguish results due either to innate principles or to environmentally conditioned factors. After an overview of the history of each experience I will discuss some of the results obtained.

2. The projects

When Fernando Tarallo came to Brazil to collect data, during his PhD in Pennsylvania, he read an article of mine on the individual variation on relative clauses in BP, in which I posed an implicational hypothesis between the variation in the relative clauses and in the overall pronominal system. Tarallo decided to test this hypothesis empirically and diachronically and he found a statistical confirmation of its main points, namely, that a) BP lost the case-marked relative pronouns and developed the chopping relative along with the resumptive relative clause and b) that chopping relatives correlated with the increase of null objects. When he finished his doctorate, he was hired by my university (at the time, it was the Catholic University in São Paulo). Excited by our first experience, the first question he asked me was whether I had some new hypotheses to test. It happened that I had. It had to do with the loss of the indefinite "se" construction in BP. We wrote a paper together which was also a successful enterprise combining formal hypothesis and empirical data (cf. Kato / Tarallo 1986). However, when we first presented the paper to a Brazilian audience, we were attacked by both generativists and variationists, who claimed we were heretic linguists. But the article ended up being published in the States in a volume edited by a generativist (Oswaldo Jaeggli) and a sociolinguist (Carmen Silva-Corvalán).

PhiN 36/2006: 24

A diachronic project was then planned. UPenn was a good example to follow, since Anthony Kroch and collaborators did exactly what we intended to do, namely, study old corpora using a formal grammar. Generativists could not criticize us since we could not use the intuition of deceased people. Positive data was the only source in the judgement of grammaticality. No negative evidence could be used. But, according to Chomsky and other generative psycholinguists, that is exactly what happens when a child builds his/her grammar. If a child can discover his/her grammar being exposed just to positive data, why couldn't a linguist discover a grammar without negative evidence? Moreover, what was interesting to observe was that immersion in old corpus often made the researchers act as competent native speakers of the language of the period. People start having intuition about the grammar of the period in question, and the corpus often showed that the researcher was right in his/her intuitions We tried to base our study in genres that closely mirror the vernacular use of the language: personal letters and plays.

The partial results of the diachronic project can be seen in a volume edited by Ian Roberts and me (cf Roberts / Kato 1993), a Festshrift for Fernando Tarallo, who we unfortunately lost at an incredibly productive phase. It is quite exhaustive in topics as it covers from the realization and position of arguments, position of clitics, wh-constructions, agreement, possessive pronouns, periphrastic forms, besides two classical diachronic articles by Fernando Tarallo, in which he discusses the possible causes for the changes that gave birth to a new grammar: Brazilian Portuguese.

With the prestige acquired with the ongoing diachronic project, Fernando and I start a project on the grammar of synchronic oral corpora (the NURC, i.e.: Norma Urbana Culta 'educated urban norm').2 The first problem we faced was : what to do with the discursive elements. Should we sanitize the corpus the way Fries, the structuralist, did in the establishment of the English sentence patterns? We decided not to, as most studies on Brazilian Portuguese (SBP) are methodologically limited in that they provide a partial and idealized transcription of the data, being based only on portions of the corpus that are relevant for a specific syntactic or discursive phenomenon. The project had, therefore, the daring aim of studying how the syntax of predication and complementation integrated syntactic adjuncts and how these elements interacted with discursive ones. Thus, while in the diachronic project, the emphasis was basically on the expression and position of arguments, or the syntax of the main constituents of the sentence, the main object of the SBP project was adjunction.3

The third project, which I am presently involved in, is a bi-national comparative study of European (EP) and Brazilian Portuguse (BP). We are using data from written language, but trying to cover different "genres", like newspaper editorials, letters from readers, interviews, plays and novels.4 We intend to reveal not only the differences in the grammar, understood as I-language, better revealed in informal texts, but also the written norms prevalent in each country, which are better captured in formal texts. In this project, besides formal syntacticians, we also have formal semanticists. Many of the topics are being dealt with in co-authorship by both Portuguese and Brazilian linguists.

3. Some of the aspects studied in the diachronic Project

The project started investigating two well-defined Parameters of variation: a) the null subject and b) the null object, as there was a suspicion (cf. Tarallo's 1983) that a change was taking place in these domains.

The null subject (NS) parameter has been proposed to be a cluster of properties, including:
[a] the possibility of null subjects
[b] free inversion / postposed subjects (cf. Chomsky 1981, Rizzi 1982).
[c] clitic climbing (Kayne 1989)

PhiN 36/2006: 25

The correlation between NS and inversion has been challenged by Safir (1982), who found out that Northern Italian dialects, while disallowing the former, license the latter. In our project, we found out that BP has been exhibiting a change in progress in both properties [a] and [b], a fact that can be used in favor of a single parameter (cf. Tarallo / Kato 1989, Duarte 1993, Roberts 1993).

Duarte (1993/2000) shows that from a preference of null subjects in 80% of cases in the middle of the 19th century, the situation is found reversed in the end of the 20th century, namely a little more than 20%, a clear indication of a change in progress. As for VS order, Torres Morais (1993) finds that from the beginning of the 19th century to the present date, the XVS form in declaratives decreases from 20% to 0%, while the VSX, which was already rare in the 19th century (6%), becomes totally absent in modern plays. Andrade Berlinck (1995) shows that VS is possible, though, with unaccusative verbs. Lopes-Rossi (1993), on the other hand, shows the decrease of VS order in interrogatives, with an increase of cleft-interrogatives (37,6%) (example (1)a) and in-situ questions (32,4%) (example 1b), both of which require SV order.

The theory of parameters provides us with another correlation which has not been much explored, namely the fact that NS languages have clitic climbing (Kayne 1989). In our project Pagotto (1993) and Cyrino (1993), studying the position of clitics in the history of BP, finds that, like present day EP, the clitic in classic BP root clauses used to be enclitic to the finite verb, while in Modern BP it is proclitic to the main verb.5 In other words BP corroborates Kayne's theory, considering that it is also loosing the NS property, as pointed out by Roberts (1993) .

Though the empirical results corroborate the three properties (null subjects, VOS order and clitic climbing) as part of the same parameter, the facts regarding Brazilian Portuguese also present challenging problems for the theory of parameters. Thus, though Duarte (1993) and (2000) show a significant decrease in the use of null subjects,6 they show that the change affected mainly the first and second persons, leaving a still reasonably stable use of null subjects in the third person, specially the non-referential ones.

PhiN 36/2006: 26

Challenged by this problem, Brazilian researchers start to work on it. Today there are two theoretical lines to interpret the problem: a) the one that treats the present state as an intermediate stage, with the prediction that non-referential subjects will also end up filled (Duarte 2000, Kato 1999), b) the one that treats the existing null subjects as variables (Figueiredo Silva 2000, Modesto 2000, Ferreira 2000). The former line claims that parameters are morphologically dependent, and that the non-uniform behavior of null subjects derives from the fact that morphology is not regular (Kato 1999). The second line presupposes a more radical change : Brazilian Portuguese has completely lost the empty pronoun. Their argument is based on the fact that null subjects in PB cannot occur in islands, a property of variables, but intuitions regarding such constructions are not uniform. In Duarte's work she finds a low rate of overt subjects in relative clauses and interrogatives, but low productivity does not mean ungrammaticality.

We will now focus on the null object (NO) parameter, proposed by Huang (1984), who studied Chinese, and proposed also by Raposo (1986)for EP.7 In both studies the NO is considered an A'-bound empty category. For Raposo, the Portuguese NO is the trace of a null operator and as such it is presumed to present island effects. Galves (1984) and Farrell (1990) proposes that the NO in BP is a pronoun and as such does not exhibit island restrictions.

Our findings regarding the diachronic status of the NO in BP are the following:

a) NOs in BP are related to the loss of the phonological third person clitic; the paradigm is therefore me-te-Ø (Kato 1993).

b1) null clitics can be found since the first half of the 17th century, and in the 19th century it already appeared in islands (Cyrino 1993).

PhiN 36/2006: 27

b2) Brazilian NOs evolve from a non-referential antecedent (propositions and predicates in the 17th century to non-human DPs, in the 18th century and finally to referential, human antecedents in the 20th century (Cyrino 1997).

c) the appearance of the referential NOs is preceded by the loss of the third person clitic and of clitic climbing (Cyrino 1993, Pagotto 1993).8

Though we obtained interesting empirical generalizations, we were left with the following theoretical question:

a) is the NO parameter a uniform phenomenon?

b) how can clitic climbing be related to both the NS and the NO parameters ?

Thus the project fed formal grammarians with new questions and hypotheses, which, in turn, allow new empirical analyses. Both empiricists and formal grammarians profit in such debate.

4. Some findings of the comparative project

Since the assumption derived from the diachronic project is that BP is losing the NS properties, we decided to investigate the same topic comparatively, between EP and BP.

The starting point was Kato's (1999) theory of the NS parameter. She proposes that the grammatical subject of the NS languages is the agreement affix itself and the apparent lexical subject is in a dislocated position in a Clitic Doubling construction as in French (15):

Kato observes that in languages where doubling is visible, the strong pronoun in dislocation is always referential (+human).

BP created a paradigm of weak pronouns quasi-homophonous with the strong ones, and doubling today sounds like repetition:

PhiN 36/2006: 28

If the third person is going in the same direction, then doubling should also be possible, especially if the referent is [+human]:

Kato's theory predicted that:
a) EP rules out such doublings;
b) when the third person is expressed in EP it should be [+human];
c) when the third person is expressed in BP it does not have to be [+human]
All these predictions were confirmed (Barbosa / Duarte / Kato 2001).

Cyrino (1997) argues that loss of climbing can be interpreted as a reanalysis of clitics as object agreement afixes and Kato / Raposo (2001) assume that the unrestricted use of NOs in BP is a consequence of this reanalysis. Just like null third person agreement can occur with both human and non-human subjects in most Romance languages, the same holds for the NO in BP. Thus, before the 20th century the pronominal subject in BP were realized as agreement affixes, and the pronominal objects appeared as clitics. In the 20th century, we have the opposite: pronominal subjects are now free weak pronouns and pronominal objects are object agreement affixes.

5. Some results of the SBP project

In this project, we tried to obtain a global description of spoken data, by transcribing the linear flow of speech into two layers : a) the 'background' layer and b) the 'filler' layer. The 'background' layer was used to refer to the syntactic patterns of predication and complementation, or the syntax of the main constituents, which is definable in terms of grammatical principles and parameter choices. The 'filler' layer enriches the 'background' layer by adding elements that fill in the boundaries between the main constituents. Such elements can be syntactic in nature (adverbials, quantifiers and clitics) or play various discursive roles. Both types of elements are equally salient in the production of language and, therefore, equally important in the description of spoken data. We assumed that just as morphology (case and agreement) is necessary for lexical entries to be realized in a syntactic construction, fillers – whether adjuncts or punctuators – would interact at the level of phonetic output, making a sentence utterable, or part of discourse. Though it is assumed that the position of syntactic fillers, or adjuncts, may be constrained by some sort of syntactic or semantic constraint, they are considered the result of a generalized transformation process operating at a more superficial level of representation.9 In the case of SBP, their position seems to be much more sensitive to other factors, such as phonetic form, a hypothesis that was also tested in some parts of the project.

In Kato (2002), I summarize and interpret the main results of the Project. As the main research question was the interaction of arguments with adjuncts and discourse fillers, two interesting results called my attention. First, when the object is null, often an adjunct can appear in sentence-final position. But this can be due to focalization, as the last element in the sentence gets the focal stress. If the object is null and pressuposed, some other constituent receives the stress. A similar phenomenon was found with subjects. When it is null (recall that third person is still productively null), an adjunct often appears preceding the verb. In the absence of an adjunct what we have is a discourse filler. Since a sentence-initial position does not normally get the main stress, the explanation given for adjuncts in postverbal position cannot be used. My conclusion was that BP dislikes a pattern in which the verb is the sentence-initial element. In other words, BP has a sort of V2 effect, but this effect is different from Germanic V2, since any discourse filler can satisfy the first position. I assumed that it was a rhythmic constraint. The ongoing loss of the null subject may have been triggered by this rhythmic property.

PhiN 36/2006: 29

The three projects, a portion of which I presented here, made it possible for us to learn a lot about BP, but many questions and problems still remain unanswered.10


Andrade-Berlinck, Rosane (1995): La position du Sujet en Portugais. Université de Louvain: Ph.D.Dissertation.

Barbosa, Pilar / Duarte, M.Eugenia / Kato, Mary A. (2001): "A distribuição do sujeito nulo no português europeu e no português brasileiro", in: Actas do XVI Encontro Nacional da Associação Portuguesa de Lingüística (APL 2001), Lisboa, 539–550.

Chomsky, Noam (1981): Lectures on Government and Binding, Dordrecht.

Duarte, M.Eugenia L. (1993): "Do pronome nulo ao pronome pleno: a trajetória do sujeito no português do Brasil", in: Roberts & Kato (eds), 107–162.

Duarte, M.Eugenia L. (2000): "The loss of the 'Avoid pronoun' principle in Brazilian Portuguese", in: Kato / Negrão (eds), 17–36.

Cyrino, Sonia M.L. (1993): "Observaçoes sobre a mudança diacrônica no português do Brasil: objeto nulo e clíticos", in: Ian Roberts / Mary A. Kato (eds), 163–184..

Cyrino, Sonia M.L. (1997): O objetoNnulo no Português do Brasil – um estudo sintáticodiacrônico, Londrina: Editora da UEL.

Farrell, Patrick (1990): "Null objects in Brazilian Portuguese", in: NLLT 8, 325–346.

Ferreira, Marcelo Barra (2000): Argumentos Nulos em Português Brasileiro, UNICAMP: Dissertação de Mestrado.

Figueiredo Silva, M.Cristina (2000): "Main and embedded null subjects in Brazilian Portuguese", in: Kato / Negrão (eds), 127–146.

Galves, Charlotte. (1984): "Pronomes e categorias vazias em Português do Brasil", in: Cadernos de Estudos lingüísticos 7, 107–136.

Kato, Mary A. (1993): "The distribution of null and pronominal objects in Brazilian Portuguese", in: Linguistic Perspectives on the Romance Languages: Selected Papers from the XXI Linguistic Symposium on Romance Languages (Currents Issues in Linguistic Theory Series), Amsterdam, 225–235.

Kato, Mary A. (1999): "Strong and weak pronominals and the null subject parameter", in: PROBUS 11,1, 1–38.

Kato, Mary A. (2002): "La gramática del portugués hablado: reflexiones sobre el uso de la lengua", in: Curcó, C. / Colín, M. / Groult, N. / Herrera, L. (eds), Contribuciones a la lingüística aplicada en America Latina, México, 363–378.

Kato, Mary A. / Negrão, Esmeralda (eds) (2000): Brazilian Portuguese and the Null Subject Parameter, Frankfurt.

PhiN 36/2006: 30

Kato, Mary A. / Raposo, Eduardo (2001): O objeto nulo definido no português europeu e no português brasileiro: convergências e divergências. Actas do XVI Encontro Nacional da Associação Portuguesa de Lingüística (APL 2001), Lisboa, 673–685

Kato, Mary A. / Tarallo, Fernando (1986). "Anything YOU can do in Brazilian Portuguese", in: O. Jaeggli / C. Silva-Corvalán (eds), Studies in Romance Linguistics, Foris, 343–358.

Kayne, Richard (1989): "Null subjects and clitic climbing", in: O Jaeggli / K. Safir (eds), The Null Subject Parameter, Dordrecht, 239–262.

Kroch, Anthony / Joshi, A. (1985): The linguistic relevance of tree adjoining grammar, Technical report MS-CIS-85-16. Department of Computer and Informational Sciences, University of Pennsylvania.

Lebeaux, David (1988): Language Acquisition and the Form of Grammar.: University of Massachusets Ph.D.Dissertation.

Lopes-Rossi, M.Aparecida (1993): "Estudo diacrônico sobre as interrogativas do português do Brasil", in: Roberts / Kato (eds), 307–342.

Modesto, Marcelo (2000): "Null subject without 'rich' agreement", in: Kato / Negrão (eds), 147–176.

Pagotto, Emílio (1993): Clíticos, mudança e seleção natural. In Roberts / Kato (eds), 185–206.

Nunes, Jairo M. (1993): "Direção de cliticização, objeto nulo e pronome tônico na posição de bjeto em português brasileiro", in: Roberts / Kato (eds), 207–222.

Raposo, Eduardo (1986): "On the null object in European Portuguese", in: O. Jaeggli / C. Silva- Corvalán (eds) Studies in Romance Linguistics. Dordrecht: Foris. 373–390.

Raposo, Eduardo (1998): "Definite/zero alternations in Portuguese", in: A. Schwegler / B. Tranel / M.Uribe-Etxebarria (eds), Romance Lingusitics: Theoretical Perspectives, Amsterdam: John Benjamins.

Rizzi, Luigi (1982): Issues in Italian Syntax.Dordrecht: Foris.

Roberts, Ian (1993): "Posfácio: O Português Brasileiro no contexto das línguas românicas", in: Roberts / Kato (eds), 409–425.

Roberts, Ian / Kato, Mary. A. (eds) (1993): Português Brasileiro: uma viagem diacrônica, Campinas: Editora da Unicamp.

Safir, Ken (1982): Syntactic Chains and the Definiteness Effect, MIT Ph. D. Dissertation.

Tarallo, Fernando. (1983): Relativization Strategies in Brazilian Portuguese, University of Pennsylvania Ph.D. Dissertation.

Tarallo, Fernando / Kato, Mary. A. (1993): "Filling in syntactic boundaries in spoken Brazilian Portuguese", in: Language Variation and Change.5,1:91–112.

Torres Morais, M.Aparecida. (1993): "Aspectos diacrônicos do movimento do verbo, estrutura da frase e caso nominativo no português do Brasil", in: Roberts / Kato (eds), 263–306.


1 On the Portuguese side, we have the coordination under João de Andrade Peres, a formal semanticist from the University of Lisbon.

2 The project was part of a larger Project on Brazilian Portuguese, coordinated by Ataliba de Castilho (UNICAMP.

3 See Tarallo / Kato (1993) for a summary of the Project.

4 A good computerized data-base of Portuguese is available in internet, which helps the data analysis ( It contains data from both Portuguese and Brazilian varieties

5 Nunes (1993) attribute this difference to different directions of cliticization: while EP cliticizes to the left, BP cliticizes to the right.

6 See also Tarallo (1983).

7 For a more recent analysis of the NO by the author, see Raposo (1998)

8 Pagotto notices that the clitic climbs in passive constructions. It should be notices, however, that here what is involved is a dative clitic and not the accusative clitic. (i) Ele me foi prometido it 1stcl was promised

9 Cf Lebeaux (1988) and Kroch and Joshi (1985)

10 My colleagues Charlotte Galves and Bernadete Abaurre, from Unicamp,who participated in the SBP project, coordinate today a huge diachronic project in which formal syntactic and phonological theories are used in the analysis of Medieval and CalssicPortuguese. And our EP/BP comparative Project is only officially finished. All the partnerships of Portuguese and the Brazilians, formed during the project will certainly continue in the same line of research.