Estudos-Ling 11 (3 P)

ESTUDOS LINGUÍSTICOS
LINGUISTIC STUDIES
Este número da Revista Estudos Linguísticos / Linguistic Studies
foi subsidiado pelo FAAC (ref.ª 440.02, Proc. 11/1/803) e pelo
Projeto Estratégico (PEst-OE/LIN/UI3213/2014) atribuído ao CLUNL
ESTUDOS LINGUÍSTICOS
LINGUISTIC STUDIES
Revista do Centro de Linguística

da Universidade Nova de Lisboa
(CLUNL)
ESTATUTO EDITORIAL
O objetivo primeiro desta revista será o de dar visibilidade à produção cientí-

fica na área de Linguística, desenvolvida quer no âmbito do Centro de Lin-
guística da Universidade Nova de Lisboa (CLUNL), quer por investigadores,
nacionais e estrangeiros, pertencentes a outras instituições.
A estrutura de cada número dependerá do planeamento editorial preven-
do-se as seguintes possibilidades:
 Números regulares, a incluírem:

o apresentação dos resultados – parciais ou finais – da investiga-
ção em curso levada a cabo pelos diferentes Grupos de investi-
gação do CLUNL
o textos submetidos a arbitragem científica propostos por investi-
gadores em Linguística, nacionais ou estrangeiros
o recensões e notícias
 Números especiais, eventualmente de caráter temático, dedicados a

iniciativas do CLUNL (workshops, colóquios ou conferências)
Em todos os números (regulares e temáticos) a publicação de trabalhos

poderá ser feita em português, francês, inglês ou espanhol.
ÍNDICE / TABLE OF CONTEN TS
Nota editorial / Editorial note

Clara Nunes Correia & Maria Antónia Coutinho........................................... 9
Introduction to this special volume

Timothy Gupton, Pilar Chamorro & Chad Howe ......................................... 11
On the emergenge of topicalisation in European Portuguese:

A study at the syntax-information structure interface
Aroldo de Andrade ........................................................................................ 13
Discussing parametric variation:

is there dative shift in Brazilian Portuguese?
Julio William Curvelo Barbosa, Paula Roberta Gabbai Armelin
& Ana Paula Scher ........................................................................................ 35
The pear stories film: simple presents and present progressives

in Portuguese, Spanish and English
Stephen Fafulas ............................................................................................. 61
Focus movement as PF movement and other peripheral positions in BP

Marcello Modesto .......................................................................................... 83
O possessivo de 2ª PL no dialeto mineiro: DP e CP em análise

Bruna Karla Pereira.....................................................................................111
Don’t put a label on me. Labels and interpretation of é que-clefts

Aleksandra Vercauteren ...............................................................................129
The effects of lexical frequency on object clitic placement

in European Portuguese
Hannah B. Washington .................................................................................157
Pragmatics and indexicality: the case of favelado

Hannah B. Washington & Mary Elizabeth Beaton .......................................177
NOTA EDITORIAL
De acordo com o estatuto editorial da Revista Estudos Linguísticos/

Linguistic Studies em que se prevê a publicação “(...) de trabalhos conside-
rados relevantes no domínio da linguística (…)” a partir de propostas feitas
por investigadores nacionais e estrangeiros (http://www.clunl.edu.pt/) temos
o prazer de publicar neste número os Textos Selecionados da 1ª conferência
‘Portuguese Linguistics in the United States (PLUS)’.
Ao publicar estes textos estamos a contribuir para divulgar a investigação
em linguística do Português, reforçando laços internacionais que ajudem a
conhecer e a divulgar o que se faz em diferentes perspetivas linguísticas e
em diferentes contextos académicos e científicos.
Os artigos que integram este volume foram avaliados por um painel de
especialistas, o que satisfaz os padrões de qualidade que a Revista definiu
como orientação primeira para os textos que publica.
Aos editores deste número – Timothy Gupton, Pilar Chamorro e Chad
Howe – queremos deixar o nosso agradecimento.
Clara Nunes Correia

Maria do Céu Caetano
EDITORIAL NOTE
According to the Journal Estudos Linguísticos/Linguistic Studies editorial

statute, which foresee the publication of "(...) work considered relevant in the
field of Linguistics (...)" from proposals made by national and foreign re-
searchers (http://www.clunl.edu.pt/) we are pleased to publish in this number
the Selected Texts of the first Conference ‘Portuguese Linguistics in the
United States (PLUS)’.
By publishing these texts we are contributing to disseminate research in
Portuguese Linguistics, strengthening international ties that help to know
and divulge what is done in different linguistics perspectives and in different
academic and scientific contexts.
The articles included in this volume were evaluated by a panel of experts,
which meets the quality standards that the Journal has defined, as first orien-
tation, for publishing texts.
To the editors of this number – Timothy Gupton, Pilar Chamorro and
Chad Howe – we want to give our thanks.
Clara Nunes Correia

Maria do Céu Caetano
THE SELECTED PROCEEDINGS
OF THE PORTUGUESE LINGUISTICS
IN THE UNITED STATES (PLUS) CONFERENCE
TIMOTHY GUPTON
(University of Georgia)
PILAR CHAMORRO
CHAD HOWE
1. Introduction to this special volume of Estudos Linguísticos

The inaugural Portuguese Linguistics in the United States (PLUS) conferen-
ce was held November 14-16, 2013 on the campus of the University of
Georgia in Athens, Georgia. This conference attracted academic researchers
working on Portuguese and Lusophone Linguistics from all over the world.
All told, there were 45 presentations and 4 keynote presentations at the con-
ference. The following collection of selected papers from PLUS represents
recent and ongoing research in a wide diversity of linguistic subfields from a
plurality of theoretical perspectives. The collection of papers presented here
is just a sampling of the breadth and depth of scholarly activity currently
taking place in Portuguese linguistics.
The editors would like to extend special thanks to Patrícia Amaral,
Meghan Armstrong, Robert Moser, and Lamar Graham. We would like to
thank the The Willson Center, The Portuguese Flagship, The Latin American
and Caribbean Studies Institute, and the Department of Romance Languages
at UGA. The inaugural PLUS conference program was also supported in part
by the President's Venture Fund and the generous gifts of University of
Georgia Partners. The editors would also like to thank our abstract revie-
wers, proceedings reviewers, presenters, invited speakers, and graduate stu-
dent volunteers for their collaboration in making this event such a success.
The editors of the Selected Proceedings of PLUS

Timothy Gupton, Pilar Chamorro, Chad Howe
ON THE EMERGENCE OF TOPICALISATION
IN EUROPEAN PORTUGUESE:
A STUDY AT THE SYNTAX-INFORMATION
STRUCTURE INTERFACE
AROLDO DE ANDRADE
(University of Campinas)
ABSTRACT: This text aims at accounting for the emergence of Topicalisation in the
history of European Portuguese from a corpus-based analysis of plays written by
Portuguese authors born either in the 15th-16th centuries or in the 19th-20th centuries.
The results support the hypothesis that the construction arises with the loss of V-to-
-C movement, by means of a replacement of the older V2 Topicalisation, which is
maintained nowadays only when a contrastive topic is used. This result shows that
syntax is able to trigger changes in information structure, and not the other way
around. An immediate implication of the presented analysis is that the loss of Topi-
calisation in other Romance languages does not obligatorily follow from the loss of
V-to-C, as usually proposed.
KEYWORDS: information structure; topicalisation; left dislocation; verb move-

ment; subject position.
1. Introduction*
The passage from medieval into modern stages of Romance languages has
typically entailed, among other syntactic and morphological shifts, the loss
of V2 Topicalisation (V2T) in favour of the use of Clitic Left Dislocation
(CLLD). Against this backdrop, Modern European Portuguese (henceforth
MEP) displays (Modern) Topicalisation (TOP) alongside other strategies to
*
Many thanks to the audience of the first conference ‘Portuguese Linguistics in the United
States’, to Charlotte Galves and to an anonymous reviewer for important comments. All
remaining errors are my own. I acknowledge support by Fapesp (project nr. 2011/19235-2).
This work is related to the thematic project “Portuguese in time and space”, also funded by
Fapesp (project nr. 2012/06078-9).
Estudos Linguísticos/Linguistic Studies, 11, Edições Colibri/CLUNL, Lisboa, 2015, pp. 13-34
14 Aroldo de Andrade
express marked topics, such as the already mentioned CLLD, Hanging Topic
Left Dislocation (HTLD) and even some remnants of the older V2T con-
struction. The following examples illustrate these constructions, which differ
syntactically regarding verb position, as well as regarding the availability
and form of a resumptive element (in boldface; topics are underlined and
their referents are elucidated between brackets, whenever necessary):1
(1) a. TOP
Esse nunca tinha ouvido. (provérbio)
this.M.SG never have.PST.1SG hear.PTCP
‘This (proverb), I have never heard of.’ (MENDES [20]:158)
b. CLLD
…e essa hei-de defendê-la… (honra)
… and this.F.SG have.PRT.1SG-of defend.INF-3SG.F.ACC
‘… and this (honour) I will defend it…’ (REBE [20]:653)
c. HTLD
O homem cabe-lhe alguma razão.
the man fit.PRT.3SG-3SG.DAT some.F reason
‘The man, he shall be entitled some reason.’ (REGIO [20]:90)
d. V2T
O que ele quere sei eu.
what he want.PRT.3SG know.PRT.1SG I
‘What he wants, (this) I know.’ (LUCCI [19]:234)
A possible compromise with the broader Romance development would

involve hypothesising that a change in European Portuguese (EP) would
have occurred in two stages, the first one spreading the use of an overt re-
sumptive, following the trend observed in the other Romance languages
around the 16th century, thus V2T  CLLD (Kroch, 2010, for French data),
whereas the second stage would erase the sentence-internal copy (leading to
CLLD  TOP), in a much later shift. However, I will argue that such an
approach completely diverges from the observed data for Portuguese.
The implementation of a successful research design required the adoption
of the following tenet: the existence of correlations between marked con-
structions and discourse functions. More specifically, I assume that a change
in marked constructions is possible provided that a given discourse function
is kept (quantitatively) unaltered through time.2 The choice among possible
1
References to corpus examples include code, the author’s century of birth and page number.
2
This assumption has been tacitly adopted in a number of papers on the syntax-information
structure interface, sometimes without proper motivation. I assume that it stems from the
notion of ‘equivalence’ used in contrastive linguistics (Krzeszowski, 1990), which derives
the fact that studies on diachronic pragmalinguistics need to assume some sort of speech
function as the tertium comparationis.
On the emergence of topicalisation in European Portuguese 15
source constructions for TOP is thus radically shortened due to both empiri-
cal and theoretical reasons, as I shall discuss.
The empirical research involved the creation of a corpus of theatre plays
by fifteen different Portuguese writers born either in the 15th-16th centuries or
in the 19th-20th centuries. (Data from the 17th and 18th centuries could not be
included due to the influence of Spanish theatre in Portugal.) This choice
offered data from a uniform genre, relatively prolific with regard to the use
of marked topic constructions, with explicit indication of contextual infor-
mation (didascalia) and usually including excerpts of spoken language, once
they are conceived as ‘to be spoken’ texts.3 The ensuing database includes
about 500 sentences with a marked topic, classified according to syntactic
and pragmatic criteria, from which some questions around the emergence of
TOP could be assessed.4
The text is organized as follows. Section 2 gives an overview of marked
syntactic constructions of the preposing type in EP. From there section 3
discusses possible source constructions for Portuguese TOP. In section 4 I
establish the variation contexts between V2T and CLLD, previously argued
to be the constructions to be given emphasis to, as possible sources for TOP.
From there section 5 presents the corpus results on this variation over time.
Section 6 discusses qualitative data related to the finding that V2T is the
source for TOP, and some of its implications. Finally section 7 displays the
concluding remarks.
2. Preposing Constructions in European Portuguese

In this section I propose a classification for marked syntactic constructions in
EP, where diachronic aspects are taken into account.5 I follow the seminal
work on EP marked topic constructions (Duarte, 1987) by considering the
criteria resumption and connectedness, relevant to comprehend the relation
established between the topic and a position inside the sentence (a copy, be it
fully realized as an NP, partially realised as a pronoun or demonstrative, or
null). However, I also consider a criterion related to the semantic relation
3
Following general criteria for corpus design, I have separately compared three major text
types: ‘private’, ‘public’ and ‘to be spoken’ texts. In general, the use of marked construc-
tions is gradually bigger, following the mentioned order. However, no significant bias be-
tween construction types has been noticed between these texts. Stylistic biases are more
prone to occur in dissertative texts (included among those of the ‘public’ type) as a tool to
convince the reader. Thanks for an anonymous reviewer for making me explain this issue.
4
The database is available for any interested reader at the following internet address:
https://dl.dropboxusercontent.com/u/1097038/Corpus-Theatre_ClP_MEP.xlsx.
5
This use of ‘preposing’ is less restricted than in Birner and Ward (1998) and related work,
where it refers only to Focus Preposing and Topicalisation.
between the topic and the eventuality it specifies, foregrounding (cf. Petrova
& Solf, 2008; Jensen & Christensen, 2013), thus helping to observe a formal
and functional distinction between V2T and TOP. Furthermore, I retrieve
delimitation and focus as further concepts encoded in preposing construc-
tions.
The resumption criterion sets apart constructions with a gap (null copy)
in sentence-internal position, such as TOP and V2T, from those with an
overt copy, such as CLLD and HTLD. This basic distinction was already
illustrated in (1a/d) versus (1b/c).
Connectedness (a term coined in Higgins, 1973) informs whether the top-
ic and its sentence-internal copy share syntactic features, such as Case and
thematic selection, setting apart CLLD from HTLD. In this latter construc-
tion the resumptive is not casually or thematically connected to the topic,
thus being either a clitic with distinct casual features–cf. (2a)–a pronoun
(overt or null) or an NP (identical or similar to the topic, such as an epithet)–
cf. (2b):
(2) a. …e a elesi acuso-osi… (“colegas de trabalho”)

… and to them.M blame.PRT.1SG-3PL.M.ACC…
‘… and them (my colleagues), I blame…’ (MENDES [20]:29)
b. Esta mulheri, por exemplo, tomemos o exemplo
this woman, for example, take.IMP.1PL the example
d[esta mulher]i que foi contratada …
of.this.F woman that be.PST.3SG hired.PTCP.F …
‘This woman, for example–let us take the example of this woman
that was hired.’ (CALDAS [20]:23)
Foregrounding is an effect of verb movement to a head position in the

CP, following den Besten (1983).6 Although verb movement is usually taken
as a mere formal condition holding in V2 languages, high verb position vis-
-à-vis sentence adverbs marks a foregrounded event, as recent research on
spoken Danish complement clauses, a variation context between V2 and V3,
has shown (Jensen & Christensen, 2013). This may explain why a V2 system
tends to have the clausal first constituent locally anchoring the eventuality in
what Carroll & Lambert (2003) call a ‘bounded language’. Old/Classical Por-
tuguese shows relevant pieces of evidence for a V2-like status, with basis on
6
Notice that in many works the event is considered to be foregrounded only if the verb is in
first position of the sentence. Following Petrova & Solf (2008), I consider that V1 and V2 in
ClP is regulated by the syntax-rhetorical structure interface. This allows me to suppose that
event foregrounding is usually grammaticalised is a general requirement in V2 and V2-like
languages.
relative adverb position and VS order (Ribeiro, 1995; Antonelli, 2011).7 In

other words, subject postposition is usually implied by a non-subject topic:
(3) … isso chamo eu falar ao pé da letra.

… this call.PRT.1SG I talk.INF to.the foot of.the letter
‘… this I call to say (it) word for word.’ (MIRA-E [15]:38’)
Two types of analyses have been proposed in the literature for this type of
language: the weak analysis supposes that verb movement occurs only as a
last resort, in the lack of XP movement to the CP domain; the strong hypoth-
esis assumes that V-to-C occurs across the board. In this paper I assume the
second proposal.
These three criteria give the following distribution of preposing construc-
tions with marked topics in EP (where foregrounding is not used with CLLD
and HTLD because it cross-cuts connectedness):
Criterion/Construction V2T TOP CLLD HTLD

Resumption - - + +
Connectedness + +* + -
Foregrounding + - ± ±
* With the apparent exception of Null Preposition Topicalisation
Table 1: Preposing constructions with marked topics in EP
Before going along, I observe the existence of subtypes of TOP that are
sometimes taken as special constructions in the literature: (i) The null object
construction–cf. (4); (ii) The indefinite se construction–cf. (5); (iii) Hyper-
-raising topicalisation–cf. (6); (iv) Null preposition topicalisation–cf. (7):
(4) …  gosto muito __ (o baton)

… (ec) like.PRT.1SG much __
‘… I like (the lipstick) a lot.’ (SANT [20]: 95)
(5) … mas isso já não se usa __.

… but this already not SE use.PRT.3SG __
‘… but this, people do not use anymore.’ (MENDES [20]:41)
7
“V2-like” means that Old/Classical Portuguese has V-to-C movement, but does not require
that one and only one constituent moves to the preverbal position. Matrix clauses may in-
deed show V1 and V>2, given different informational-structural requirements (Galves &
Paixão de Sousa, 2013).
(6) … os olhos parece [que __ lhe ficam atrás]…

… the eyes seem.PRT.3SG [that __ 3SG.DAT stay behind]…
‘… (that man’s) eyes, it seems that (they) stay behind him…’
(MIRA-E [15]: 35)
(7) E de pasta japonesa não precisa __?

and of folder japanese NEG need.PRT.3SG __?
‘And (a) japanese folder, you do not need (it)?’ (LUCCI [20]:90)
Example (4) shows that the interpretation of null objects in EP is fixed

vis-à-vis a null marked topic, which corresponds to the discourse topic
(Raposo, 1986). In (5), the underlined element sits in a left-peripheral posi-
tion, thus a marked topic, which may also occur in postverbal position, as
Raposo & Uriagereka (1996) argue with various pieces of evidence, such as
the impossibility of bare NPs in the subject position of neutral (active or
passive) clauses, but their possibility in preverbal position combined with an
indefinite se. In (6), the gap corresponds to the subject position of a finite
clause, a construction usually receiving a hyper-raising analysis, where the
element in the matrix preverbal position can only receive a topic interpreta-
tion, once the subject is a (usually null) expletive. In (7), the topic appears
without the preposition required by the verb precisar (‘to need’). This case,
referred to in the literature on EP as topicalização selvagem (cf. Brito, Du-
arte & Matos, 2003: 501-2), is distinct from Delimiter Preposing in that it
exhibits connectedness with a gap inside the clause.
Now I turn to other concepts encoded in preposing constructions that
surpass topichood. Delimitation is described as the expression of a limited
contribution for the communicative needs at the current point of discourse
(Krifka, 2007). We include in this relatively novel concept frame-setters,
which give special or temporal information, and hanging topics. This last
type is illustrated in (8), where the generic noun praias (‘beaches’) does
not manage content storage in the common ground once, as a delimiter, it
cannot initiate or maintain a topic chain. Therefore, cases similar to this
one were excluded from the database.
(8) Bom… praias, adoro a Arrábida.

well… beaches, love.PRT.1SG the Arrábida
‘Well, (as for) beaches, I love Arrábida.’ (Duarte, 1987:73)
A further type of preposing construction involves focus: Focus Preposing,

which necessarily has a contrastive value and, as usual with focus elements,
activates a set of alternatives. Costa & Martins (2011) present a series of
defining characteristics of this construction, among which preverbal clitic
placement and VS order, both illustrated in (9) below:
(9) De notícias se faz o nosso mundo.

of news SE make.PRT.3SG the our world
‘It’s news that make up our world.’ (Costa & Martins, 2011:223)
This presentation, made necessary due to the proliferation of conflicting

terminology in the area of information structure, allows me to consider the
the specific task to be undertaken in the following sections, starting with the
possible source constructions for TOP.
3. Possible Source Constructions for Topicalisation in Modern Europe-

an Portuguese
In section 1 I have presented the subtypes of constructions encoding
topichood in EP.8 From the outset, one can assume that the possible source
constructions for TOP in the grammatical system of Classical Portuguese
(ClP)9 are HTLD, CLLD and V2T.
Let us now consider whether any of these candidates can be discarded
due to independent reasons. I argue that, once resumptive deletion in HTLD
does not generate TOP, because of the lack of casual and thematic connect-
edness between topic and resumptive, as shown in (10), HTLD cannot be
maintained as a possible source construction for TOP.10 (This example and
the following ones were formulated to show possible paths of change.)
(10) HTLD  *TOP

O capitãoi, ninguém sabe delei.
the captain, nobody know.PRT.3SG of.him
‘The captain, nobody knows about him / his whereabouts.’
Resumptive deletion  *O capitão, ninguém sabe __.
On the other hand, CLLD and V2T remain as possible candidates. They
differ regarding the syntactic change involved: the first presupposes resump-
tive deletion–cf. (11)–whereas the second involves a syntactic change (loss
of verb movement)–cf. (12).
8
The assumption that Focus Preposing and Delimiter Preposing have different informational-
-structural content from topicalisation and dislocation constructions leads me to discard
them as possible sources for TOP.
9
ClP encompasses the 15th-17th centuries, according to Galves, Namiuti & Paixão de Sousa
(2006).
10
From a quantitative standpoint, it is also possible to dismiss HTLD as a source construction
for TOP, once it is used less than 10% of the cases, very far from the frequency of TOP in
MEP, above 50%.
(11) CLLD  TOP

O capitãoi, vi-oi ontem.
the captain, see.PST.1SG-3SG.M yesterday
‘The captain, I saw him yesterday.’
Resumptive deletion  O capitãoi, vi __i ontem.
(12) V2T  TOP
O capitão vi eu__ ontem.
the captain, see.PST.1SG I __ yesterday
‘The captain, I saw him yesterday.’
Loss of verb movement  O capitão, eu vi __ ontem.
The investigation can therefore be constrained to two source construc-

tions for TOP: V2T or CLLD.
4. The variation between Topicalisation and Left Dislocation

Once the research must now emphasise the variation between V2T,TOP and
CLLD, I establish the variation contexts between Topicalisation and Left
Dislocation. This will prevent me from comparing data of marked topics that
cannot be ‘transposed’ from CLLD into TOP. To do so, I mention key points
regarding the distribution of TOP and Left Dislocation (LD) in English, and
apply them to EP.
Following Birner & Ward (1998), I assume that function-to-form map-
pings among marked constructions are generally similar crosslinguistically.
Consider the variation between TOP and LD in English:
(13) English TOP

“He couldn’t deal with pressure. Just the slightest pressure he found diffi-
cult to handle...” (Philadelphia Inquirer, Nov. 13, 1984, In Birner & Ward
1998:44)
(14) English LD
I bet she had a nervous breakdown. That’s not a good thing. Gallstones,
you have them out and they’re out. (Roth, 1969:162 In Birner & Ward,
1998:93)
The most accepted proposal for this variation considers three discourse
functions (DFs) for the latter construction (Prince, 1997):
a. DF(LD1): Simplifying discourse processing (by removing a dis-
course-new entity from a disfavourable position, i.e. subject);
b. DF(LD2): Triggering a poset inference (by marking contrast of the
item vis-à-vis others forming a list);
c. DF(LD3): Amnestying an island violation (by avoiding a long-

-distance dependency that would be created by TOP).
I briefly discuss in which way the discourse functions presented above
can be also relevant to explain the variation between CLLD/TOP in Portu-
guese, and in turn to enlighten the question whether a pragmatic change from
CLLD into TOP has taken place.
From the DFs presented above, the first one presents a discourse-new
constituent, which is usually considered impossible for TOP.11 In fact,
DF(LD1) is couched in Lambrecht’s (1994) Principle of Separation between
Role and Reference (PSRR), according to which pointing to a referent and
indexing a previous topic should be distinguished in terms of syntactic posi-
tion. It is important to remember that the use of a resumptive in subject posi-
tion implies a HTLD, once the element cannot be a clitic pronoun in EP:
(15) … de quem te menos temesi essei te

… of whom INH.CL less fear.PRT.2SG this.M 2SG.ACC
engana…
deceive.PRT.3SG
‘… the one you fear less, (this one) deceives you…’ (FERR-F [16]:74)
Once HTLD was previously excluded as a possible source construction for

TOP, LD1 is not under our concerns.
LD2, on the other hand, would be possibly interchangeable with TOP in
terms of discourse function, once both share the requirement that “… the ref-
erent or denotation of the preposed constituent be anaphorically linked to the
preceding discourse” (Birner & Ward, 1998:32). According to Prince (1997),
unlike TOP, though, LD2 does not present a further requirement that entails an
open proposition inferable from the anaphoric link to be salient in the hearer’s
mind, this lack in saliency being thus signalised with resumption. The example
below shows an example of LD2 expressed by CLLD in EP:
(16) Anselmo, podes trazer o chá… E os biscoitosi

A., can.PRT.2SG bring.INF the tea… and the biscuits
deixa-osi por ora na lata.
leave.IMP.2SG-3PL.ACC by now in.the can
‘Anselmo, you can bring the tea… and (regarding) the biscuits, leave them
for now in the can.’
(MENDES [20]:265)
11
This entails that a preposed brand-new element (Prince, 1981) would be either resumed by
a subject in HTLD or be contrastively focused itself.
In this dialogue, the poset {snack products} was not clearly stated, but the
hearer can build it in his/her discourse model due to the linking relation es-
tablished with a previous item pertaining to it: o chá (‘the tea’).
Finally LD3 displays a resumptive connected to a topic outside of a syn-
tactic island, which is necessarily expressed by CLLD in EP, once the topic
must have a local connexion with its gap in the case of TOP:
(17) … e ao meu filhinhoi só tinha

… and to my little.son only have.PST.1SG
[para lhei dar] a água das bicas…
[for 3SG.DAT give.INF] the water from.the taps…
‘… and to my little son I only had tap water for him to give…’
(BRUN [19]:33)
The constituent between brackets is an adjunct island; so the presence of an

overt resumptive is required. Hence this type of example shall not be includ-
ed among the variation cases between CLLD and TOP.
I conclude this section by observing that the following data shall be dis-
missed in order to grasp the variation contexts between V2T/TOP and
CLLD:
a. Topics connected to a subject position, an oblique position or a nomi-
nal complement position, because these are grammatical functions that
cannot be resumed by clitic pronouns, thus hit must be a HTLD.
b. Cases of CLLD in which the resumptive occurs inside a syntactic is-
land.
In other words, what is left are topics connected to an argumental posi-
tion: either a direct or an indirect object, thus potential contexts for
LD2/TOP variation.
5. Corpus results
Before presenting the results, I would like to summarise the possible change
scenarios for the emergence of topicalisation: (i) a change in two stages
(V2T  CLLD; CLLD  TOP), the first following from a common Ro-
mance shift; (ii) a change from CLLD into TOP, due to informational-
-structural shifts that would have altered the correlation between discourse
function and syntactic form, and (iii) a change from V2T into TOP, due to a
syntactic change, restricting the amount of verb movement to the CP do-
main, a movement with consequences for foregrounding. The following
subsections present the results for the two first scenarios, first looking at the
general results per century, then on different syntactic and pragmatic criteria.
5.1. General results and evidence for the loss of verb movement
Figure 1 below shows the results on the relevant preposing constructions
per century, plus the ambiguity cases between TOP and V2T. Subject expres-
sion is the criterion for distinguishing among the three cases at hand: a pre-
posed topic with a postverbal subject was coded as V2T; a preposed topic with
a null subject or with a preverbal subject that is clearly dislocated was coded
as TOP/V2T (ambiguous); a preposed topic with a preverbal subject that sits
in a bona fide Spec,TP position was coded as TOP.12 The chart includes all
data, irrespectively of the grammatical function of the topic constituent.
100% 0 0 0,02 0,05
75%0,47 0,37 0,55 0,49 TOP

50% TOP/V2T
0,14 0,26 0,08 V2T
0,12
25% CLLD
0,38 0,3 0,29 0,36
0%
15th 16th 19th 20th
Figure 1: Use of marked topic preposing constructions (excluding HTLD)
12
Nominal and pronominal non-quantified subjects in declarative clauses in the order Topic-
-S-V seem to be found only in MEP, and by hypothesis occupy Spec,TP. We have consid-
ered as occupying a dislocated position:
– S in the configuration S-Topic-V, as they are mapped as hanging topics or foci – cf. (i);
– S expressed as a quantifier in the configuration Topic-S-V, following Bocci (2007)
– cf. (ii);
– S expressed as a preverbal 1st or 2nd pronominal subject in yes-no questions. See observati-
ons in Vanrell Bosch & Fernández Soriano (2013:20-24) – cf. (iii).
(i) E eu a Lívia buscava mais honra que apetite.

and I to L. search.PST.3SG more honour than appetite
‘And in Livia I looked mostly for honour than for appetite.’ (FERR-C[16]:132)
(ii) Do não cobiçoso poucos se queixam.
of.the not greedy few 3SG.REFL complain
‘Few ones complain of the non-greedy (people).’ (VASC[16]:62)
(iii) E a todos esses tu queres meter em casa?
and to all these you want.PRT.2SG put.INF in house?
‘So you want to put all these ones inside home?’ (MIRA-V[15]:28’)
The figure shows that the percentage of CLLD keeps around 30% for
texts from all the studied centuries. Instead of a decrease in CLLD, one ob-
serves that V2T is even slightly more frequent than CLLD in the 16th century
(numbers for CLLD/total per century are 32/85; 47/154; 19/65; 55/148).
Therefore, no preference for the use of CLLD is observed in late ClP, unlike
in other Romance languages, thus denying scenario (i). Notice that a slight
decrease in the number of TOP/V2T between the 15th and the 16th century
seems a natural fluctuation.
The figure also shows that a great amount of data express ambiguity be-
tween TOP and V2T, and only in MEP can one find unambiguous cases of
TOP. This is consistent with the hypothesis according to which ClP was a V-
-to-C language, and therefore displayed only cases of V2T, which are dis-
guised due to the high quantity of null subjects.13
5.2. Results on pragmatic criteria
In this and in the following subsection I discuss only the data in variation
contexts between CLLD and TOP, according to the criteria exposed in sec-
tion 4. This reduces the sample to 341 tokens, a fact that led me to display
the results by language period again (ClP versus MEP). Here I examine the
dependent variable presence of resumptive element against two different
pragmatic criteria, givenness and salience (–res. refers to both V2T and
TOP, and +res. indicates CLLD).
Table 2 presents the results on givenness, taking into account the distinc-
tion between new, inferable and evoked (Prince, 1981).
PERIOD New Inferable Evoked Total

n % n % n % n
CLP -RES. 15 50 31 45 65 79 111
+RES. 15 50 38 55 17 21 70
MEP -RES. 7 50 23 38 57 79 87
+RES. 7 50 38 62 15 21 60
Table 2: Topic givenness and presence of resumptive element
The results indicate that new elements (which besides brand-new ele-
ments, also include elements anchored to the context by a possessive or simi-
lar type of anaphor) distribute more or less evenly regarding [±resumption].
13
Notice that according to Galves & Paixão de Sousa (2013), the use of null subjects com-
pete with VS during ClP, whereas null subjects compete with SV in MEP. This seems to
comply with the hypothesis stated before, according to which null subjects occupy a post-
verbal position and preverbal subjects are always dislocated.
Inferable elements (either situationally or textually) slightly favour resump-

tion. Finally, evoked (given) elements strongly favour non-resumption. Cru-
cially, the results are very similar in ClP and MEP.14
On the other hand, the results on salience suggest a change between ClP
and MEP, as shown in Table 3 below. This result is specifically relevant for
contrastive topics, where resumption would be favoured in ClP, but strongly
disfavoured in MEP. Shifting topics (very salient, due to situational infer-
ence or focus at the previous sentence) and familiar topics (less salient, due
to repeated retrieval from previous utterances) favour non-resumption in
both periods (cf. the classification for marked topics put forward in Fras-
carelli & Hinterhölzl, 2007).15 However, these data cannot be used as evi-
dence for a pragmatic change, due to statistical measures.16
PERIOD Contrastive Shifting Familiar Total

n % n % n % n
CLP -RES. 21 60 87 61 3 75 111
+RES. 14 40 55 39 1 25 70
MEP -RES. 2 29 60 53 25 93 87
+RES. 5 71 53 47 2 7 60
Table 3: Topic salience and presence of resumptive element
These results contradict scenario (ii), according to which a pragmatic

change took place whereby TOP in MEP would have a similar discourse
function to CLLD in ClP. All the other way around, what the results show is
a clear stability of the correlation between discourse function and syntactic
form during time.
14
The results are significant for the two studied periods: ClP (2=20.5762, p=3.404e-05), and
MEP (2=24.0404, p=6.021e-06).
15
The reader can observe each one of these types of topics in the examples below: in the
preverbal position, in (ia), a shifting topic; in (ib), a contrastive topic preverbally, and, in
the immediate postverbal position, a familiar topic (cf. also Galves & Gibrail, 2012):
(i) a. ... e desses mimos vêm todas as ousadias.
... and of.these gifts come.PRT.3PL all the cockinesses
‘...and all cockinesses come from these gifts.’ (VASC[16]:12)
b. Mas as alheas sinto eu mais que as minhas.
but the foreign feel.PRT.1SG I more than the mine
‘But I feel other people’s longings more than mine own.’ (FERR-C[16]:123)
16
From Fisher’s Exact Test, the results are not signficant for ClP (p=1), but are indeed signif-
icant for MEP (p=5.385e-06).
5.3. Results on morphosyntactic criteria

The only possible scenario at this point consists in a change from V2T in-
to TOP, thus entailing a syntactic change whereby verb movement (and con-
sequently event foregrounding) came to be optional in MEP. To do so, I
show in this section only the relevant morphosyntactic criteria, observed
from multivariate analyses conducted separately for ClP and MEP.
Tables 4 and 5 show the results of a series of multivariate analyses using
Goldvarb X (Sankoff, Tagliamonte & Smith, 2005). Considering the stronger
effect of givenness, I have ignored the results from salience. The data from
ClP and MEP were analysed separately, and the same independent variables
(person, givenness, subject status and sentence type (context)) were available
for the two of them, and eliminated by the program.
Corrected mean = .63

Log likelihood = -106.249
n % weight
Givenness Range = 39
Given 65/ 82 79.3 .70
New (anchored) 15/ 30 50.0 .40
Inferable 31/ 69 44.9 .31
Subject status Range = 28
Postverbal 53/ 73 72.6 .63
Null 41/ 73 56.2 .44
Preverbal 17/ 35 48.6 .35
Table 4: Multivariate analysis for non-resumption in ClP
Corrected mean = .59

Log likelihood = -88.433
n % weight
Referent person Range = 37
+person 82/ 114 71.9 .61
-person 10/ 45 22.2 .24
Givenness Range = 29
Given 61/ 77 79.2 .63
Inferable 23/ 67 34.3 .39
New (anchored) 8/ 15 53.3 .34
Table 5: Multivariate analysis for non-resumption in MEP

The expectation, if V2T competed with CLLD, is to find a greater influ-

ence of subject status (preverbal, null, postverbal) in ClP for the choice of
non-resumed preposing constructions, as Table 5 indeed shows. On the other
hand, in MEP the hypothesis is that, once TOP competes with CLLD, other
factors will be more relevant for the same choice, which is indeed obtained:
referent person alongside givenness.
The results in Table 4 are compatible with the idea that the choice be-
tween marked constructions is basically guided by pragmatic principles, but
there is a clear connexion between the occurrence of postverbal subjects and
non-resumption. Notice that the more given the topic element is, the bigger
is its probability of being topicalised instead of being left-dislocated (new
anchored elements seem to behave together with given elements).17
In Table 5, on the other hand, the role of subject status becomes com-
pletely irrelevant, and givenness is superseded by referent person, where
third person (-person) favours non-resumption, whereas first and second
person favours resumption. I tentatively suggest that this result can be under-
stood as a coding of presupposition connected to indexicality, but I leave a
proper analysis for a future work.
These results confirm scenario (iii), whereby a syntactic change has led to
the non-obligatory status of verb movement in MEP. In other words, (Mod-
ern) Topicalisation is the mere reanalysis of V2 Topicalisation (except for a
relatively small number of vestigial data of the latter construction). Eventual
readjustments regarding the importance of the factor groups analysed shall
be attributable to the effects of a syntactic change.
6. Qualitative analysis
I lay out in this section some observations regarding the clausal periphery of
ClP and MEP that may explain their differences regarding marked construc-
tions. From there I consider how some cases of V2T in MEP can be accom-
modated in the present analysis.
First of all, consider the close connexion between the syntax of the left-
-periphery and changes in other domains of the syntax of Portuguese, such as
clitic and subject placement. The results shown here are in accordance with
the hypothesis of a change in two steps, the first step being a prosodic shift
from a syllable-based to a rhythm-based system around the beginning of the
17th century, without a clear language-external trigger (Frota, Galves &
Vigário, 2007). As a consequence of this change, the second step involved a
syntactic change involving the loss of V-to-C movement around the begin-
ning of the 18th century. According to Galves, Britto & Paixão de Sousa
17
Except in Table 5, probably due to its small token number.
(2005), there are two clear pieces of evidence for the latter shift: (i) variation
contexts for clitic pronoun placement, which became obligatory contexts for
enclisis (i.e. post-verbal clitic placement); and (ii) VS order, whose numbers
decreased from 20% into 10% around this time.18
The assumption around the loss of a V-to-C grammar would not only ex-
plain the phenomena previously mentioned, but also provide a natural expla-
nation for the emergence of TOP in EP out of a grammar with V2T. To do
so, I assume an analysis of the left periphery elaborating on proposals laid
out in Frascarelli & Hinterhölzl (2007) and Neeleman et al. (2009), shown in
(18), with clear indication of the position of the verb and of different types of
topics (Con=Contrastive Topic, Fam=Familiar Topic, Shi=Shifting Topic):
(18) a. [ForceP [DelimP [TopicP Con/Shi [FocusP (Con) [FinP V [TP Fam …
b. [ForceP [DelimP [TopicP Con/Shi [FocusP (Con) [FinP [TP Fam V …
First of all, consider the stable features holding in (18a-b). In line with the
observations in section 2, delimiter phrases would occur in Spec, DelimP,
and Con and Shi, in Spec,TopicP; the difference between these two elements
would reduce to the presence of a [+contrastive] feature in the first one, due
to its base-generation in Spec, Focus0 and further movement to check its
topic feature. Finally, Fam would always be encoded in Spec,TP.
This proposal can explain the change from V2T into TOP in the follow-
ing way. The structure in (18a) shows that ClP had obligatory movement of
the (finite) verb into a head of CP, which I consider to be Fin 0, following
Antonelli (2011), entailing the verb’s precedence with respect to Fam, en-
coded by the subject. After the loss of V-to-C movement, MEP exhibits Fam
in preverbal position, as in (18b). I assume here that the preverbal subject of
MEP occurs in Spec,TP, following evidence presented in Costa & Duarte
(2003).
In this scenario, data with post-verbal subjects combined with a preverbal
topic are of special interest, because by hypothesis they should have disap-
peared in MEP, contrary to the facts. The change seems to have involved a
restriction in the types of topic elements requiring verb adjacency: a shifting
topic could be found in Topic-V-S order in ClP, as the contrast in (19a-b)
shows:
18
Among different analyses for clitic placement, I subscribe to a PF account, in line with
Galves & Sandalo (2012). I assume that clitics are X0 elements found to the left of the verb,
thus proclisis is the non-marked placement case; enclisis would be obtained post-
-syntactically. Thus verb movement is not required as a last resort operation, so as to derive
enclisis.
(19) Shifting topics in ClP and MEP

a. … a essas virtuosas solicita o mundo mais…
… to these virtuous.F.PL require.PRT.3SG the world more…
‘…to these virtuous women the world requires more…’ (VASC [16]: 8’)
b. … e disso as testemunhas farão prova…
… and of.this the witnesses make.FUT.3PL proof…
‘… and this, the witnesses will demonstrate evidence…’ (REBE [20]:539)
On the other hand, contrastive topics can be found in both ClP and MEP
in Topic-V-S order:
(20) Contrastive topics in ClP and MEP

a. … isso não ousaria ela…
… this NEG dare.COND.3SG she…
… this she would not dare…’ (FERR-F [16]:20)
b. … já o mesmo não posso eu dizer …
… already the same NEG can.PRT.1SG I say.INF…
‘… but the same I cannot say…’ (REBE [20]:115)
The use of a ‘strong’ anaphor, usually expressed by a demonstrative pro-

noun referring to a whole proposition, is very common in this configuration.
The analysis of this constituent as a contrastive topic is detailed in pragmatic
terms in Light (2012) for a set of Germanic languages. I assume that this
analysis can be extended to Romance languages as well.
Finally, I suggest that a criterial analysis can be invoked so as to explain
these lingering cases of Topic-V-S sentences in MEP without having to en-
rich the left periphery, much in the spirit discussed in Neeleman et al.
(2009). Besides contrastive topics, preposed focus constituents–cf. (9)
above–and non-D-linked wh-phrases–cf. (21) below, from Amaral
(2009:66)–also require VS order in EP:
(21) Que posso eu dizer sobre isso?

what can.PRT.1SG I say.INF about this?
‘What can I say about this?’
Although a detailed analysis of this phenomenon would go far off the

goals of this paper, it seems plausible that the three constructions at hand
involve the calculus of alternatives of a more constrained type, thus some-
how they share a [+contrastive] feature, much in line with the Focus Criteri-
on by Rizzi (1997):
(22) Topic-V-S in MEP

[ForceP [DelimP [TopicP Con [FocusP (Con)/Focus/Wh V [FinP [TP ...
As a result of the application of this criterion (instead of VS as the non-

-marked word order), the amount of sentences with Topic-V-S order also
decreases: 9% in MEP, thus much lower than the 22% in ClP. (Cf. however
Galves & Gibrail, 2012 for a whole analysis of VS in transitive sentences.)
7. Concluding remarks
This paper supports the hypothesis that the emergence of TOP in EP is the
mere consequence of a syntactic change involving the loss of verb move-
ment into the CP domain, which has occurred in the transition between ClP
and MEP, in the beginning of the 18th century. This idea is backed up by the
corpus results showing a considerable stability regarding the influence of
givenness on the choice between non-canonical syntactic constructions. On
the other hand, the quantitative analysis has also pointed out that V2T was
an unconstrained phenomenon in ClP, even if the high quantity of sentences
with null subjects hinders this analysis.
These findings undermine the role of an autonomous informational-
-structural change as a trigger of TOP in Portuguese grammar, so as to relate
it to a later development stemming from the CLLD construction. In other
words, I maintain that changes in the expression of informational-structural
concepts are due to changes in the grammar. This is expected if the use of
information structure concepts is stable over time, a crucial assumption for
this work. In the case at hand, the loss of V-to-C movement has triggered the
loss of unmarked V2T, a construction that became restricted to mark con-
trastive topics in MEP.
On a par with that, this way of interpreting the shifts in the left periphery
of EP is coherent with Birner & Ward’s (1998) proposal for variation affect-
ing non-canonical constructions, which should be well-constrained cross-
-linguistically. In fact, the data analysed here show that non-canonical con-
structions are relatively similar over time regarding their pragmatic value.
Though new function-to-form mappings may be theoretically possible, they
should be seen primarily as consequences of independent changes occurring
in the syntactic domain.
If the emergence of TOP is a natural development in the history of Portu-
guese, a broader explanation for the availability of preposing constructions is
in order, so as to explain why this construction is not found in other Ro-
mance languages (except in colloquial French), where CLLD seems to take
over the functions of Portuguese (and English) TOP. This task is left for
future research.
References
A. Primary sources (Preceded by reference code with author’s century
of birth, and year of birth)
BRUN [20] (1881) Brun, André 1916. Teatro. Lisboa: Guimarães & Cia.
CALDAS [20] (1974) Caldas, Miguel Castro 2008. Comida/Casas/Repartição. Lis-
boa: Artistas Unidos/Cotovia.
DANTAS [19] (1876) Dantas, José 1912. O reposteiro verde. Lisboa: Livraria Clás-
sica.
FERR-C [16] (1528) Ferreira, António 1622. Comédia do Cioso. In. António Álvares
(org.) Comédias famosas portuguesas dos Doutores Francisco de Sá de Mi-
randa, & António Ferreira, dedicadas a Gaspar Severim de Faria. Lisboa:
Oficina de António Álvares, pp. 117-155.
FERR-F [16] (1518) Ferreira, António 1562. Comédia do Fanchono. Coimbra: Ofi-
cina de João de Barreyra.
LUCCI [19] (1860) Lucci, Eduardo Schwalbach 1923. Os postiços. Porto: Lello &
Irmão.
MENDES [20] (1976) Mendes, José Maria Vieira 2008. Teatro. Lisboa: Cotovia.
MESQ [19] (1856) Mesquita, Marcelino 1897. O regente: tragédia em 12 quadros.
Lisboa: António Maria Pereira.
MIRA-E [15] (1481) Miranda, Francisco Sá de 1561. Comédia intitulada os Estran-
geiros, ao Ifante Cardeal Dom Anrique. Coimbra: Antonio de Maris.
MIRA-V [15] (1481) Miranda, Francisco Sá de 1560. Comédia intitulada os Vilhal-
pandos, ao Ifante Cardeal Dom Anrique. Coimbra: Antonio de Maris.
REBE [20] (1924) Rebello, Luís Francisco 1999. Todo o teatro. Vol. 1. Lisboa:
INCM.
REGIO [20] (1901) Régio, José 1968. A salvação do mundo. 2.ed. Lisboa: Portugá-
lia.
ROQUET [19] (1875) Roquette, Vitório Chagas 1924. Sete comédias num acto. Lis-
boa: Guimarães & Cia.
SANT [20] (1924) Santareno, Bernardo 1962. Anunciação. Lisboa: Ática.
VASC [16] (1515) Vasconcelos, Jorge Ferreira de 1618. Comédia Ulissipo. [Nesta]
segunda impressão apurada, e correcta dos erros da primeira. Lisboa: Pedro
Craesbeck.
B. Secondary sources
Amaral, Diana Travado (2009). Interrogativas-Wh: Periferia Esquerda e Fases. Tex-
tos Seleccionados. XXIV Encontro da APL. Lisboa: Colibri, pp. 61-79.
Antonelli, André (2011). Posição do verbo no português clássico: evidências de um
sistema V2. Alfa 55 (2), pp. 501-522.
den Besten, Hans (1983). On the interaction of root transformations and lexical
deletive rules. In. Werner Abraham (Ed.). On the Formal Syntax of the West-
germania. Amsterdam: John Benjamins, pp. 47-131.
Birner, Betty J. & Gregory L. Ward (1998). Information status and non-canonical
word order in English. Amsterdam/Philadelphia, PA: John Benjamins.
Bocci, Giuliano (2007). Criterial positions and Left Periphery in Italian: Evidence
for the syntactic encoding of contrastive focus. Nanzan Linguistics 3 (1), pp.
35-70.
Brito, Ana Maria; Inês Duarte & Gabriela Matos (2003). Tipologia e distribuição
das expressões nominais. In. Maria Helena Mira Mateus et al. Gramática da
Língua Portuguesa. 5. ed.rev.aum., Lisboa: Caminho, pp. 795-867.
Carroll, Mary & Monique Lambert (2003). Information structure in narratives and
the role of grammaticised knowledge: A study of adult French and German
learners of English. In. Christine Dimroth & Marianne Starren (Eds.) Infor-
mation structure and the dynamics of language acquisition. Amster-
dam/Philadelphia, PA: John Benjamins, pp. 267-287.
Costa, João & Inês Duarte (2003). Preverbal subjects in null subject languages are
not necessarily dislocated. Journal of Portuguese Linguistics, 1 (2), pp. 159-
-176.
Costa, João & Ana Maria Martins (2011). On focus movement in European Portu-
guese. Probus 23 (2), pp. 217-245.
Duarte, Maria Inês Pereira da Silva (1987). A construção de topicalização na gra-
mática do português: regência, ligação e condições sobre movimento. Dis-
sertação de doutoramento, Universidade de Lisboa.
Frascarelli, Mara & Roland Hinterhölzl (2007). Types of topics in German and Ital-
ian. In. Susanne Winkler & Kerstin Schwabe (Eds.) On information struc-
ture, meaning and form. Amsterdam/Philadelphia, PA: John Benjamins, pp.
87-116.
Frota, Sónia, Charlotte Galves & Marina Vigário (2007). Ler a Fonologia: do Portu-
guês Clássico ao Português Moderno. Textos Seleccionados. XXIII Encontro
da APL. Lisboa: Colibri, pp. 193-206.
Galves, Charlotte, Helena Britto & Maria Clara Paixão de Sousa (2005). The change
in clitic placement from Classical to Modern European Portuguese: Results
from the Tycho Brahe Corpus. Journal of Portuguese Linguistics 4 (1), pp.
39-67.
Galves, Charlotte & Alba Gibrail (2012). Subject inversion in transitive sentences
from Classical to Modern European Portuguese: a corpus-based study. Ma-
nuscript submitted for publication.
Galves, Charlotte, Cristiane Namiuti & Maria Clara Paixão de Sousa (2006). Novas
perspectivas para antigas questões: revisitando a periodização da língua por-
tuguesa. In. Annette Endruschat, Rolf Kemmler & Barbara Schäfer-Prieß
(Orgs.). Grammatische Strukturen des europäischen Portugiesisch: Syn-
chrone and diachrone Untersuchungen zu Tempora, Pronomina, Präprosi-
tionen und mehr. Tübingen: Calepinus Verlag, pp. 45-74.
Galves, Charlotte & Maria Clara Paixão de Sousa (2013). The loss of V-to-C in the
history of Portuguese: subject position, clitic placement and prosody. Manu-
script submitted for publication.
Galves, Charlotte & Filomena Sandalo (2012). From intonational phrase to syntactic
phase: The grammaticalization of enclisis in the history of Portuguese. Lin-
gua 122 (8), pp. 952-974.
Higgins, Francis Roger (1973). The pseudo-cleft construction. Dissertação de dou-
toramento, Massachusetts Institute of Techonology.
Jensen, Torben Juel & Tanya Karoli Christensen (2013). Promoting the demoted:
The distribution and semantics of ‘‘main clause word order’’ in spoken Dan-
ish complement clauses. Lingua 137 (1), pp. 38-58.
Krifka, Manfred (2007). Basic notions of information structure. Interdisciplinary
Studies on Information Structure 6, pp. 13-56.
Kroch, Anthony (2010). Prosody, topicalisation and verb-second: how they interact
in the history of English and French. PPT file. Retrieved February 5, 2014,
from http://www.ling.upenn.edu/~kroch/handouts/udel-wpsi4.pdf
Krzeszowski, Tomasz P. (1990). Prototypes and equivalence. In. Jacek Fisiak (Ed.)
Further insights into contrastive linguistics. Amsterdam/Philadelphia: John
Benjamins, pp. 29-46.
Lambrecht, Knud (1994). Information structure and sentence form. Cambridge:
Cambridge University Press.
Light, Caitlin (2012). The syntax and pragmatics of fronting in Germanic. Dis-
sertação de doutoramento, University of Pennsylvania.
Neeleman, Ad et al. (2009). A syntactic typology of topic, focus and contrast. In
Jeroen van Craenenbroeck. (Ed.) Alternatives to Cartography. Berlin: Mou-
ton de Gruyter, pp. 15-52.
Petrova, Svetlana & Michael Solf (2008). Rhetorical relations and verb placement in
the early Germanic languages: a cross-linguistic study. In Cathrine Fabricius-
-Hansen & Wiebke Ramm (Eds.) ‘Subordination’ and ‘Coordination’ in Sen-
tence and Text: a cross-linguistic perspective. Amsterdam/Philadelphia: John
Benjamins, pp. 329-351.
Prince, Ellen (1981). Toward a taxonomy of given-new information. In Peter Cole
(Ed.) Radical Pragmatics. New York: Academic Press, pp. 223-256.
Prince, Ellen F. (1997). On the functions of left-dislocation in English discourse. In
Akio Kamio. (Ed.) Directions in Functional Linguistics. Amster-
dam/Philadelphia, PA: John Benjamins, pp. 117-143.
Raposo, Eduardo (1986). On the null object in European Portuguese. In Oswaldo
Jaeggli & Carmen Silva-Corvalán (Eds.) Studies in Romance Linguistics.
Dordrecht: Foris, pp. 373-390.
Raposo, Eduardo & Juan Uriagereka (1996). Indefinite se. Natural Language and
Linguistic Theory 14 (4), pp. 749-810.
Ribeiro, Ilza (1995). Evidence for a Verb-Second Phase in Old Portuguese. In Adri-
an Battye & Ian Roberts (Eds.) Clause Structure and Language Change. New
York: Oxford University Press, pp. 110-139.
Rizzi, Luigi (1997). The fine structure of the left periphery. In Liliane Haegeman
(Ed.) Elements of grammar: Handbook of generative syntax. Dordrecht:
Kluwer, pp. 281-337.
Sankoff, David, Sali A. Tagliamonte & Eric Smith (2005). Goldvarb X: A variable
rule application for Macintosh and Windows. Retrieved May 6, 2013, from
http://individual.utoronto.ca/tagliamonte/Goldvarb/ GV_index.htm
Vanrell Bosch, Maria del Mar & Olga Fernández Soriano (2013). Variation at the
Interfaces in Ibero-Romance: Catalan and Spanish Prosody and Word Order.
Catalan Journal of Linguistics 12, pp. 1-30.
DISCUSSING PARAMETRIC VARIATION: IS THERE
DATIVE SHIFT IN BRAZILIAN PORTUGUESE?
JULIO WILLIAM CURVELO BARBOSA*

(Universidade Estadual do Paraná)
PAULA ROBERTA GABBAI ARMELIN*
(Universidade Federal de Juiz de Fora)
ANA PAULA SCHER*
(Universidade de São Paulo)
ABSTRACT: The main goal of this paper is to derive two different parametric dis-
tinctions that explain the distribution of dative constructions in two dialects of Bra-
zilian Portuguese – the main dialect, spoken by most speakers (BP), and a dialect
spoken in the Zona da Mata Mineira region (MBP). Within a Distributed Morpholo-
gy view, we follow Barbosa’s (2012) parallel between double object constructions
and datives to compounds and N+de+N compound-like expressions. They differ by
means of a macroparameter that regulates the phonological presence of the preposi-
tion and the order of the arguments it mediates. In MBP, however, it is possible to
have datives with a surface order that resembles double object constructions we
argue with Scher (1996) and Armelin (2011) that this resemblance is a byproduct of
a microparameter regulating the absence of the preposition – and the surface order
– in contexts of default Case marking (Kato, 2012) at vP’s left periphery (Belletti,
2002).
KEYWORDS: Parameters; Double Object Constructions; Distributed Morphology;

Brazilian Portuguese.
*
We would like to thank CNPq (Processes 140146/2011-3 (Ph.D.) and 312610/2013-0 (Re-
search Productivity)), CAPES (PRODOC program, processes 1267797 (PNPD) and
4102/13-4 (BEX)), and the Linguistics Department from the Universidade de São Paulo for
financially supporting the research resulting on this paper. We also thank the audience at
PLUS 2013 for the discussion of a previous version of this work, as well as an anonymous
reviewer for the comments and suggestions provided to us. Any shortcomings and remain-
ing errors are of our entire responsibility.
36 Julio W. Curvelo Barbosa, Paula R. Gabbai Armelin & Ana Paula Scher
1. Introduction
This paper intends to shed light into the debate between the properties de-
fining micro- and macroparametric variation (Baker, 1996, 2008; Kayne,
2008, inter alia). In order to discuss these properties, we will focus on the
dative constructions from two dialects of Brazilian Portuguese (henceforth
BP), making a contrastive approach to English dative shift phenomena. The
key data set that instigates us to pursue this comparison is one of BP’s dia-
lects, spoken at Zona da Mata Mineira, which we will call Minas Brazilian
Portuguese (henceforth MBP, in contrast to standard BP).
We will contrast the data from BP to the data from English known as
double object constructions (DOC; cf. Barss & Lasnik, 1986; Larson, 1988,
inter alia). The DOC would differ from standard dative constructions such
as (1)-(2) in presenting the order verb-theme-goal, with the goal argument
bearing a prepositional marker:
(1) John gave [a book] [to Mary].
(2) Peter sent [a letter] [to Sue].
The DOC is a variation of the dative construction – hence, the dative shift
terminology –, with some effects of asymmetry; the goal argument (i.e., the
recipient) precedes the theme and the dative preposition (to) may be absent
from the sentence:
(3) John gave [Mary] [a book].
(4) Peter sent [Sue] [a letter].
There is a long discussion in the literature1 regarding the status of these

constructions and their properties, but one important factor to notice here is
that there is a consensus in the many works on the subject, claiming these
constructions do not occur in Romance – therefore, it would seem odd that
Brazilian Portuguese could present any kind of DOC. In this paper, we will
follow Scher (1996) and Armelin (2011), who argue that, even though the
goal can sometimes be the closest argument to the verb in BP, the absence of
the preposition ((6)) is only allowed in the MBP dialect:
1
The discussion of DOC’s and datives’ properties can be found in Barss & Lasnik (1986),
Larson (1988), Pesetsky (1995), Harley (1997), Bresnan et alli (2007), among other works.
Discussing Parametric variation 37
(5) A Maria deu [PP ao/para o Pedro goal] [DP um livro theme]. (BP)
(6) A Maria deu [DP o Pedro goal] [DP um livro theme] (MBP/*BP)
(7) Mary gave [DP Peter goal] [DP a book theme].
Although this dialect allows for the same order of arguments seen in the
Double Object Constructions (DOCs) from English, Scher (1996) and Ar-
melin (2011) claim that this order is a mere epiphenomenon, while distinct
properties account for the surface resemblance to English2. The three main
questions we will try to answer are:
(a) How similar are, in fact, English and MBP datives?
(b) What are the properties that make MBP resemble English?
(c) Can we explain the properties from MBP with the same parametric
distinction that separates English and BP?
The hypothesis that guides our work is the following: while apparently
similar to the contrast observed between English and MBP, the dialect varia-
tions detected in the comparison between BP and MBP are to be accounted
for by a different parameter type than the one that excludes dative shift from
BP in general. The parameter contrasting BP and English relates compounds
and datives, as suggested by Barbosa (2012), while accounting for the varied
behaviors seen within the pair English versus BP, as well as the pair BP ver-
sus MBP. From the empirical and theoretical consequences presented
throughout the paper, we consequently question the accuracy of the defining
characteristics of (micro-/macro-) parametric classification, especially when
a non-lexicalist model of grammar (i.e., Distributed Morphology, henceforth
DM; cf. Halle & Marantz, 1993; Marantz, 1997; Embick & Noyer; 2007), is
taken into consideration.
The structure of the paper is the following: in section 2, we present the
basic characteristics of datives in BP and MBP, comparing these properties
to English dative shift phenomena, in order to make clear the empirical dif-
ferences there are between this language and the BP dialects. Section 3
brings the analysis from Barbosa (2012) for English versus BP, while section
4 discusses the micro-/macroparametric properties and the analysis for ex-
plaining the surface effect that makes MBP’s datives look like dative shift.
Section 5 presents the concluding remarks.
2
For an alternative take on the subject, cf. Torres-Morais (2001, 2007), and Torres-Morais &
Lima-Salles (2010).
2. Characteristics of datives in BP and MBP

In this section, we highlight the differences between datives in both BP and
MBP. Crucially, two central aspects to deal with in these two dialects are: (a)
the status of the prepositions a (‘to’) and para (‘for’) – which may alternate
in the introduction of the goal argument in BP, but not in MBP3 – and (b) the
different informational aspects conveyed by the sentences when under dif-
ferent circumstances. These circumstances are, namely: (i) when the order of
the complements of the verb is rearranged and (ii) when the preposition is
absent (specifically in MBP).
The first important characteristic to point out is that, in BP, the absence of
the preposition on the goal argument ungrammatical under any circum-
stance. We confirm this fact by the ungrammaticality detected in (9), when
compared to (8), below:
(8) A Maria deu [DP um livro] [PP ao/pro Paulo].

‘Maria gave [DP a book] [PP to+the Paulo].’
(9) * A Maria deu [DP um livro] [DP o Paulo].

‘Maria gave [DP a book] [DP the Paulo].’
In addition, there is an additional order in which the arguments of the

verb can be arranged. In the first order, the goal argument linearly follows
the theme, as in the example (8), above. In the other possible ordering, the
goal is linearly at the left of theme. It is important to notice that the differ-
ence in the order of the complements does not correlate to the absence of the
preposition in any sense: in both orderings, there must be a preposition in-
troducing the goal argument. However, the order goal+theme is marked for
informational aspects, in the sense that it usually triggers a contrastive focus
reading4:
3
The goal argument is usually related to the element receiving the object affected by the
action of giving, also known as a recipient. In dative constructions, it is commonly marked
by a (morphological) dative case, or by a preposition, receiving oblique (structural) Case. In
the DOC, the goal argument may appear without the preposition, and as such, it is consid-
ered to receive (structural) accusative Case, just like the theme (a.k.a., the object). Hence,
the name double object construction. Our goal in this paper is to show that the data in
(M)BP is dissociated from the so-called “accusative” properties of the goal argument when
closer to the verb. This independent motivation does not need to associate dative semantics
to oblique Case marking, since the latter is a well-formedness requirement for the computa-
tional component.
4
Although there is no distinction between a and para in BP, it should be noted that speakers
of BP who use para with the goal PP tend not to have a as an option in their grammars. In
the more recent generations, the use of a is restricted to formal contexts, and is not as pro-
ductive as the use of a in MBP.
(10) a. A Maria deu [PP ao/pro Paulo focus] [DP um livro].

‘Maria gave [PP to+the Paulo] [DP a book]’
b. *A Maria deu [DP o Paulo focus] [DP um livro].
‘Maria gave [DP the Paulo] [DP a book]
The picture is different in MBP, however. In this dialect, the preposition

introducing the goal argument may be omitted, as in (11)b. Crucially, as
noted in Scher (1996), the absence of the preposition is closely related to
verbs that allow a as a possible preposition. In the context of verbs that only
allows para, the V+DP+DP construction is ruled out (12)c. This might be an
argument towards the fact that the preposition a, in this dialect, is, in fact,
specialized for the goal (dative) semantics:
(11) a. A Maria deu [DP um livro] [PP ao Paulo].

‘Maria gave [DP a book] [PP to+the Paulo].’
b. A Maria deu [DP um livro] [DP o Paulo].
(12) a. A Maria construiu [DP uma casa] [PP para o Paulo].

‘Maria built [DP a house] [PP for+the Paulo].’
b. *A Maria construiu [DP uma casa] [PP ao Paulo].
‘Maria built [DP a house] [PP to+the Paulo].’
c. *A Maria construiu [DP uma casa] [DP o Paulo].
‘Maria built [DP a house] [DP the Paulo].’
As in BP, the order between the theme and the goal argument may vary in
MBP and, importantly, the order goal+theme is marked in terms of informa-
tional aspects as well. In this sense, the same contrastive focus reading ex-
plicated in (10)a for BP may emerge in MBP when the goal is at the left of
the theme (13):
(13) a. A Maria deu [DP um livro theme] [DP o Paulo goal].

b. A Maria deu [DP o Paulo goal-focus] [DP um livro theme].
‘Maria gave [DP the Paulo] [DP the book Paulo].’
When specifically dealing with the status of the preposition that may in-
troduce the goal argument, it is important to note that para and a differ in
important aspects. First, there can be two elements bearing the preposition
para in the same sentence. In such case (14), the first argument has to be
interpreted as the goal, while the second PP, an adjunct, is interpreted as the
beneficiary5.
(14) A Maria deu [um livro theme] [pro Paulo goal] [pro Pedro benef.].
‘Maria gave [DP a book] [DP for the Paulo] [DP for the Pedro].’
On the other hand, no ambiguity is associated to the preposition a. Cru-

cially, the arguments bearing a can only be interpreted as goal – a dative
reading. Simply put, the benefactive reading is not available when the PP is
headed by a. This is confirmed by the impossibility that having two PPs
headed by a in the same sentence.
(15) *A Maria deu [um livro theme] [ao Paulo goal] [ao Pedro benef.].
‘Maria gave [DP a book] [DP to the Paulo] [DP to the Pedro].’
The only way to have a grammatical sentence with a and an argument

with the benefactive semantics is to mark the second (benefactive) PP with
para. Notice that if the benefactive element is marked with a, the sentence is
ungrammatical (with the intended reading, cf. (17)):
(16) A Maria deu [um livro theme] [ao Paulo goal] [pro Pedro benef.].
‘Maria gave [DP a book] [DP to the Paulo] [DP for the Pedro].’
(17) *A Maria deu [um livro theme] [pro Paulo goal] [ao Pedro benef.].
‘Maria gave [DP a book] [DP for the Paulo] [DP to the Pedro].’
Fundamentally, the only possible interpretation associated to the absence

of the preposition in MBP is the a (dative/goal) reading. The DP without the
preposition may not be interpreted as a benefactive. This ensures that, when-
ever the preposition is absent, it is the preposition a, not para, that are being
omitted. This is confirmed by the fact that, in MBP, it is not possible to elide
the preposition from both the goal and the benefactive PPs in the same sen-
tence. Otherwise, a sentence such as (18) would be grammatical in MBP.
(18) *A Maria deu [um livro theme] [o Paulo goal] [o Pedro benef.].
‘Maria gave [DP a book] [DP the Paulo] [DP the Pedro].’
It is an interesting fact that the absence of the preposition in MBP may

result in a surface order that resembles the DOC of the English dative alter-
5
Despite the ambiguity seen in para, the gloss throughout the paper will remain as for, in
order to keep the contrast with a (translated as to).
nation. The question that emerges form that observation is the following: do
the linear orders for the internal arguments in MBP result from the very
same phenomenon or do they underline different processes? As we will
show in the next section, the properties that derive the V+DP+DP order in
MBP are not exactly the same than those generating it in English.
2.1. Does MBP behave like English, or does it look like English?
It is a widely discussed fact that, in English, the absence of the preposi-
tion correlates directly to displacement: when the preposition is absent, the
goal element is on the left of the theme (20)a, (20)b, except in cases of heavy
NP shift (21):
(19) Mary gave [a book] [to Paul].
(20) a. Mary gave [Paul] [a book].

b. *Mary gave [to Paul] [a book].
(21) Mary gave [to Paul] [the heaviest book in the store].
However, as we have already pointed out in the previous section, MBP

does not present a correlation between the absence of the preposition and the
property of displacement, a crucial aspect of the DOCs in English. In MBP,
both the theme+goal and the goal+theme orders are possible, regardless of
the presence or absence of the preposition:
I. Preposition Absence II. Goal before theme

BP No Free
MBP Yes Free
English Conditioned to II Conditioned to I
Table 1: Complement order and presence/absence of the preposition
As we have seen in the previous section, the properties of English, BP

and MBP have shown to be different in what concerns goal displacement
and preposition absence. This is already an indication that different phenom-
ena may be leading to the order V+DP+DP in English and in MBP.
Investigating this question, Scher (1996) argues extensively against the
hypothesis that MPB presents DOCs. An important difference between MBP
and English pointed out by the author regards the impossibility of a passive
sentence in which the indirect object becomes the subject of the sentence in
either MBP or English; in the DOC version however, English allows this
passivization, while MBP still yields ungrammatical results. We can observe
this pattern in the examples in (22), below:
(22) a. A Maria deu um livro aos/pros meninos.

b. Mary gave a book to the boys.
c. *Pros/Aos meninos foram dados um livro. *MBP
d. *To/For the boys were given a book. *English
e. *Os meninos foram dados um livro. *MBP
f. The boys were given a book. English
(Scher, 1996: 28)
In the same sense, in MBP, but not in English, it is possible to omit the
preposition when no explicit direct object is present in the sentence.
(23) a. O que ela fez com o retrato? Ela deu o Pedro.

b. What has she done with the portrait? *She gave it Peter.
c. O que você vai fazer com o livro? Eu entrego o pai dela.
d. What will you to do with the book? *I’ll give it her dad.
(Adapted from Scher, 1996: 112)
Another important difference between English and MBP is observed

through relativization data6:
(24) O menino que eu dei o caderno não está aqui.MBP

*The boy who I gave the book is not here. *English
‘The boy to whom I gave the book’
(Scher, 1996: 29)
Based on the differences between MPB and English, it is plausible to

conclude that (i) MBP has no double object and that V+DP+DP sentences in
this dialect will need a different explanation, since it presents different em-
pirical facts than the English ones. In the next section, we investigate what
the structural differences behind these contrasting patterns are.
2.2. The informational approaches in English and MBP: a connection
between two different phenomena
As we have shown so far, the dative structures in MBP have a wider
range of structural possibilities, since the preposition a does not make itself
obligatory – neither when the goal precedes the theme, nor when the theme
precedes the goal, which is the standard dative order. If we think about free-
dom of order in a language that has a relatively fixed order, such as (M)BP,
we can come up with two explanations: (i) different interpretations for every
structure or (ii) movement within (syntactic) structure.
6
The same is true for data relating to interrogative pronouns and fronting.
So far, there is no evidence that allows us to postulate more than one se-
mantic interpretation for dative structures in MBP, especially when we con-
sider that the preposition a has a semantic restriction in favor of dative (re-
cipient/goal) arguments. When it comes to English, however, there is a de-
bate between two theoretical standpoints. The first one suggests that datives
and DOCs are formed by different structures in syntax (e.g., Pesetsky, 1995;
Harley, 1997; Levin, 2006; Bruening, 2010, inter alia). The second one
claims one of these constructions derives from the other (Bresnan & Nikiti-
na, 2009; Ormazabal & Romero, 2012; inter alia). We believe the answers
given by the second group of works bring consequences that are more inter-
esting to cross-linguistic work, Based on that, we will justify why we rule
out the first group in the discussion ahead. In the following sections, we will
discuss the properties for English and BP, as well as the parametric conse-
quences for the choices we make from here on.
2.2.1. DOCs do not have an interpretation that is excluded from datives

As we have said earlier, the studies on dative shift diverge as to the rela-
tion between DOCs and dative constructions, with a debate too long to re-
produce in this paper. As a comparative work, our intent is to follow the
most fruitful ideas, especially if these ideas can handle the differences be-
tween (M)BP and English. If we were to follow the approaches that claim
datives and DOCs are two separate constructions, we would have to figure
out (i) what is the meaning conveyed by these structures and (ii) whether the
interpretation associated to DOCs is completely ruled out from any other
construction in a language that does not have DOCs. Works such as Pesetsky
(1995) and Harley (1997) claim that the DOCs are a different structure from
the datives for the fact that they entail a transfer of possession reading. This
would be a problem for our analysis of MBP, once we have seen this reading
is entailed in any constructions with the preposition a, exclusively associated
to that type of semantic relation.
Levin (2006) shows that the dative-like semantics manifests, crosslinguis-
tically, in three types of constructions associated with the expression of re-
cipients. The description she gives can be seen below:
– Double Object Construction (recipient as possessor): theme–second object,

recipient–first object
– Dative Construction (recipient as possessor): theme–object, recipient–dative
NP
– Allative Construction (recipient as goal): theme–object, recipient–
allative/locative NP/PP
A given language’s mode of realizing recipients depends on its morphosyntac-

tic resources, but a given language will have only one of the first two syntactic
constructions.
(Levin, 2006: 2)
According to Levin (op. cit.), while the English DOCs has a recipient as
possessor reading, the English dative constructions are not like the “true”
morphological datives, since they present the recipient as goal reading. On
that note, the author also says that “Many languages which lack a double
object construction still have a core (i.e., nonadjunct) grammatical relation,
distinct from subject and object, used to express recipient” (Levin, 2006: 2).
When looking at the MBP data discussed above, we saw the preposition a
can only fulfill a goal/dative semantics, i.e., a recipient as possessor reading.
If there is no DOC in MBP (or BP), neither is there dative case marking in
these dialects, there would be no explanation to why the recipient as posses-
sor reading is available. Hence, we should not be able to have separate con-
structions yield separate meanings, since the recipient as possessor seman-
tics is not exclusive of DOCs and morphological datives.
Once we have empirically dismissed the possibility of having DOCs in
both BP and MBP and justified the theoretical point of view we will (not)
base our analyses on, we can move on, motivating a parallel question to-
wards solving the puzzle that the MBP data brings us. That is the goal of the
next section.
2.2.2. The motivation for argument movement in MBP datives

If we follow the reasoning behind the choice of our theoretical path of
choice, we will have to consider the two possibilities for the dative construc-
tions in BP and MBP: either (i) they are two separate constructions, or (ii)
one of them derives from the other. Both possibilities sound unfeasible to
handle these dialects, once we consider there is no dative shift in MBP. For
(i), we would have to make sure that, if BP and MBP present different struc-
tures for their datives constructions, then they would also need to have dif-
ferent semantic interpretations. This possibility is ruled out by the fact that
the data in BP express the same semantic content than their dative counter-
parts in MBP. As for the case of (ii), we would need to acknowledge these
sentences are not far apart at all, and there may be a smaller role played by
the presence/absence of the preposition, as well as for the displacement of
the goal argument.
We claim that the difference in these constructions is due to a few factors.
While the thematic roles of goal and benefactive (or even locative) are di-
vided between a (goal) and para (benefactive/locative) in MBP, para in BP
can represent either goal, benefactive and locative semantics. In order to
support that fact, we can recall that a is not even present in many dialects of
BP speakers, especially the younger speakers and the ones with less experi-
ence through formal instruction – moreover, these variables do not need to
be simultaneous. Scher (1996) provides an analysis under Minimalist terms,
suggesting that, whenever we have the verb+goal+theme sequence, the goal
argument is interpreted as a topic, being left-dislocated above vP in order to
get its discursive features. Armelin (2011) expands Scher’s (op. cit.) analysis
even further, contextualizing it into the cartographic approach, more specifi-
cally, Belletti (2002). As for the absence of the preposition a in such con-
texts, Scher (1996) claims that the phonological contexts allows for a syl-
labic restructuring, while Armelin (2011) briefly suggests a DM-based solu-
tion related to feature incompatibility between the terminal node and the
preposition’s Vocabulary Item (henceforth VI). In section 4, we will try to
address this question, while maintaining the ideas for topic dislocation.
Having in mind the discussion presented in the two previous subsections,
the question we want to address is the following: can we explain, in terms of
a parametric approach, the distinctions seen in the DOC/dative constructions
from English, as well as the DOC-like surface expression of datives in MBP,
contrasted to standard BP datives? If so, can we account for all the differ-
ences by a single (type of) parameter?
3. The (macro-) parametric constraint:

prepositions in English versus BP
3.1. Compounds and dative shift derived from the same structure
According to what we have shown in the previous sections, it seems that
the dative shift phenomena occurring in English is a distinct empirical fact
than the ones seen on either BP or MBP. If we take these differences to be
the result of a parametric variation, we should consider (i) what type of pa-
rameter allows dative shift in English, but not in BP, and (ii) whether the
contrasts between English and BP are byproducts of the same type of varia-
tion that sets BP and MBP apart.
In order to make that distinction, we should think about the current para-
metric variation approaches. The approach we will consider for the differ-
ences between English and BP is that of marcroparameters (Baker, 1996,
2008). Baker (1996) begins that enterprise by questioning the range of ef-
fects and the fragmentation (i.e., the suggestion of more construction-
-specific parameters) of the Pro-Drop parameter (Rizzi, 1982). Going the
opposite direction from most parametric analyses in the literature, Baker
(2008) advances the proposal that we should expect macroparameters,
roughly defined by (a) showing strong effects in grammar variation, (b) its
variation being realized in the syntax (i.e., not in the lexicon), and (c) being
found when distantly related languages are compared.
At a first moment, let us consider (c). Although BP and English are not as
typologically distant from each other as, for instance, English and Mohawk,
we believe that it would be implausible to consider English and BP as close-
ly related as MBP and BP. Therefore, we will advance the idea that BP and
English behave differently in what concerns dative shift, thus showing char-
acteristics of macroparametric variation.
To present evidence towards this choice, we will follow the analysis
made by Barbosa (2012), a new take on Snyder’s Compounding Parameter
(1995, 2001). Snyder claims that the acquisition by a child of N+N com-
pounds forms in a given language triggers the acquisition of complex predi-
cates such as resultative, verb-particle, double object, causative, perceptual
verb, dative and locative constructions. Snyder also presents cross-linguistic
evidence to show that, if a language has no productive N+N compounds,
then it will not have the complex predicates that are related to them. Barbosa
(2012) uses BP data to show that Snyder’s Compounding Parameter should
not be so pervasive, since not all complex predicates are actually absent in
BP – causative, locative, perceptual and dative constructions are quite pro-
ductive, and occur very early in BP speaking children’s grammar (cf. Bar-
bosa & Simioni, 2011; Barbosa, 2012).
From this perspective, Barbosa tries to find a new explanation for the ab-
sence of productive N+N in BP. He observes that N+de+N form of com-
pound seems to be quite productive in that language. Semantically, he claims
that they are similar to English N+N, once one considers the two forms pho-
nologically analogous. Thus, these constructions would share the same inner
properties, yet realizing distinct surface outputs:
(25) dog food vs. food for a dog

(Olsen, 2008:10, apud Barbosa, 2012: 102)
(26) comida de cachorro vs. comida para cachorro(s)

food of dog food for dog(s)
(Barbosa, 2012: 102)
Barbosa (2012) claims that the preposition de is a linking element, which

does not need phonological content in English. The author explains these
claims with an analysis based on DiSciullo (2005), according to which a
functional projection places the compound members in specifier and com-
plement positions. If the semantics of English and BP’s compounds is the
same, their structure in LF should not differ. The structure for English or-
ange juice and for suco de laranja (lit. “juice of orange”) in BP is as (27):
(27)
As we can see, the structure above inverts the order of the compound in
English, since the complement orange should appear before the semantic
head juice. Considering the model of DM, Barbosa proposes that, since BP
and English have the same LF outputs, these different outputs should be
derived after Spell-out, on the way to PF. With the rule in (28), Barbosa ex-
plains the reversed order of English compounds, with its resulting morpho-
logical structure in (29):
(28) Compulsory Dislocation Rule

A noun phrase occupying the position [Comp, P] must adjoin to [Spec, P]
whenever P has no phonological content.
(Barbosa, 2014: 8)
(29)
The compulsory dislocation operation applies only after VI insertion –

the last Morphological Structure operation, considering the late insertion
hypothesis (cf. Halle & Marantz, 1993; Embick & Noyer, 2007) – still, oc-
curring prior to linearization. Therefore, if BP or English inserts a VI at P,
the order between the specifier and the complement will remain the same in
both PF and LF.
Yet, Barbosa (2012) still needs to account for why BP is not productive
at compulsory dislocated structures. First, there is no robust data for N+N
compounding in BP, in contrast to N+de+N. Even considering this margin-
al type of formation, it rarely has the reverse order seen in English (right-
-side headedness), and most of these cases are byproducts of affixes, such
as ferrovia (‘railroad’, lit. iron + path), also seen as via de ferro; (cf. Bar-
bosa, 2012: 120), and cannot be easily retrieved to their fully-fledged
counterparts.
Based on these arguments, Barbosa (2012) suggests that this difference is
parametrical, and that it applies over two of the complex predicates Snyder
(1995, 2001) say should not be possible in a language such as BP – datives
and DOCs. We claim, along with Barbosa (op. cit.) that the dative shift con-
structions share with compounds the same characteristics that sets them apart
when we compare English and BP: the possibility of omitting the phonologi-
cal output of P.
(30) Preposition Phonological Realization Parameter

A language allows prepositions to be absent of phonological content at PF.
{yes} = English
{no} = BP (including MBP)
(Adapted from Barbosa, 2012: 198)
Once we consider the parameter in (30), the relation between N+N versus
N+de+N compounds and dative shift phenomena can be explained by two
main properties, namely (i) no phonological content for the prepositional
node, and (ii) the “reversed” order of the relevant nominal elements being
related by this prepositional node. It is important to point out that the node c-
-commanding the preposition – either n or v – will derive the distinct behav-
iors form each construction, and that, since English allows both positions
(conditioned to the presence of P) in both constructions, this analysis cap-
tures a desirable pattern.
As for the structure representation for the dative shift phenomena, we
claim that (31), below, is the way datives get to the LF component in both
English and BP. Just like compounds, the surface realization is what will
differ in these two languages: the parameter in (30), above, triggers com-
pulsory dislocation of the complement of P at (31), resulting in the mor-
phological structure (32) ((31) and (32) are adapted from Barbosa,
2012:166-167):
(31)
(32)
If correct, the structures for dative shift in (31) and (32), above, compare
to the same type of asymmetries seen on English’s versus BP’s compounds.
One could argue that these structures do not accurately represent the asym-
metry effects from Barss and Lasnik (1986), often used in the literature to
justify two sets of independent structure for DOCs and datives. Despite that
fact, we will follow Bresnan et alli (2007), and say these asymmetries are an
epiphenomenon. Once they seem to deal with questions of phonological
requirements during VI insertion, we will postulate the rule in (33) below,
considering that the c-command restrictions are satisfied at LF (once both
interpretations share the same structure) and at PF (since compulsory dislo-
cation occurs only after VI insertion):
(33) The binding/quantification hierarchy constraint

VI insertion at PF must respect c-command (i.e., no>any, referential ex-
pressions>bound
pronoun/anaphor, each>the other, every>bound pronoun, etc.)
An argument towards this constraint is Larson’s claim: “the asymmetries

observed with V-NP-NP structures occur with V-NP-PP structures as well.”
(Larson, 1988: 338).
If we recall the three factors that distinguish microparameters and macro-
parameters shown in the beginning of this section, we can see that the pro-
posal from Barbosa (2012); along with any other parameter suggested within
the DM framework) will not comply with either side of the parametric mod-
els on the market. If Barbosa’s analysis is correct, the macroparametric fac-
tors (a) (showing strong effects in grammar variation) and (c) (being found
when distantly related languages are compared) seem to be compatible with
the parameter in (30), above. As for (b) (its variation being realized in the
syntax (i.e., not in the lexicon)), there are important theoretical consequences
we should consider. Barbosa’s proposal does not comply with the Borer-
-Chomsky conjecture, presented in (34), below, since it relies on functional
projection variation outside of the lexicon, with no lexical (i.e., word-
-formation component) considered at all. This should be a good thing, how-
ever, if we want this analysis to be a macroparametric one.
(34) The Borer-Chomsky Conjecture

All parameters of variation are attributable to differences in the features of
particular items
(e.g., the functional heads) in the lexicon.
(Baker, 2008: 253)
However, the relevant steps for this parameter occur after syntactic op-
erations (or, at least, after Spell-out), which is not a concern for the macro-
parametric model either. Thus, it seems that the main assumption behind DM
– namely, that the lexical operations are “distributed” to other places of the
derivation – is not compatible with either macro- or microparametric as-
sumptions. That does not mean we should discard it, but that the conse-
quences for parametric variation within this model should be considered
with more attention, if this analysis proves fruitful. In the next section, we
will present an analysis that explains the differences between MBP and BP,
contextualizing them in terms of a microparameter.
4. The (micro-) parametric constraint

During the presentation of the empirical facts in section 2, we introduced
two separate patterns for English versus BP, and for BP versus MBP, respec-
tively. Considering the analysis in section 3, we claimed that the contrast
between BP and English relates to a macroparameter, since the dative con-
struction patterns in these two languages (i) strongly affect the grammar
variation and (ii) occur in distantly related languages.
What about dative-shift-like data from MBP? According to a micropara-
metric approach (e.g., Kayne, 2008), the three main factors that classify a
parametric difference within this approach are (a) weak effects in grammar
variation, (b) variation in the lexicon (thus, following the Borer-Chomsky
conjecture), and (c) comparison between dialects, or closely related lan-

guages.
From the empirical discussion shown in section 2, we could assert that,
even though MBP and English share a somewhat similar surface realization
of dative sentences, some other properties from MBP keep us from saying
these surface manifestations are the byproduct of the same underlying LF
structure. Once again, we will explain this difference in terms of DM, sug-
gesting an analysis mainly derived from the empirical characteristics we
have discussed in section 2, complementary to the one separating BP and
English.
If the properties that distinguish BP from English seen above are taken
into consideration, micro- and macroparameters can only be distinguished by
(i) the proximity (microparameters)/distance (macroparameters) between
language families, and by (ii) the effects on grammar they may cause (small-
er on microparameters, bigger on macroparameters). The factor yet to be
accounted for is the locus of the variation (lexicon versus syntax). That is a
direct reflection of the theoretical model of Grammar chosen in this paper,
which we will discuss more thoroughly later on.
Macroparmetrically, both BP and MBP should not present dative shift
phenomena. Nevertheless, we still have to explain the surface resemblance
seen between English and MBP. Our proposal is that, while we distinguish
BP and English by a macroparameter, thus leading to broader effects (such
as the compounding realization in both languages), the lack of phonological
content in MBP must be a consequence of other forces over the computa-
tional component.
First, we have to recall that both English and MBP may have no preposi-
tion phonology for the goal argument, but only English is free to keep this
VI from being inserted, according to the parameter in (30). Therefore, we
need to explain what occurs in MBP. There are two possible approaches to
this fact: either (i) the preposition’s phonological content is inserted, then
erased, or (ii) the preposition’s phonology is not inserted at all. Both ap-
proaches have problems. The first one has to deal with an extra operation,
and this operation has to occur as late as possible in the derivation, consider-
ing VI insertion is one of the latest steps prior to PF linearization. The ap-
proach in (ii), however, presents an even more interesting challenge, espe-
cially if we consider that, if MBP did not insert phonological content into P,
the parameter proposed to distinguish English and BP would be at stake.
Against all odds, we will suggest that (ii) is the most elegant solution to
the parametric question at hand. At a first glance, the choice seems arbitrary
and contradictory to what we have said so far. Yet, the fact that the order
V+DP+DP in MBP occurs only in a set of specific contexts, with the possi-
bility of the order V+DP+DP corresponding to both the sequences
verb+goal+theme and verb+theme+goal, gives us enough evidence to say a

simple erasure of the preposition would be even harder to explain.
In order to justify the choice we have made, we will claim there is a dif-
ference in the VIs between BP and MBP. This, associated to the possibility
of Case-less P in MBP, triggers a vP periphery displacement (Armelin, 2011,
Belletti, 2012). In MBP, the movement of the P+DP sequence to those posi-
tions neutralizes the need to fill P with phonological content. A consequence
for this movement is the topic/focus interpretation we can get from the data
in MBP, as previously shown in section 2.
A first argument that we can give is that, if the order verb+goal+theme is
not a byproduct of dative shift in MBP, the only plausible explanation for
this sort of argument scrambling is the movement to any information-
al/discourse-related structural position. This seems to be a fruitful approach
– as both Scher (1996) and Armelin (2011) have already presented solid
argument towards this view. While BP does allow a free order of the com-
plements of the verb, it depends on the presence of the preposition. On the
other hand, MBP is not restrictive at all when it comes to argument order,
either with or without the preposition. This shows us that MBP is even less
restrictive than English, since the order verb+goal+theme is conditioned to
the absence of to (cf. Table 1, above).
Once we establish our base grounds for the BP/MBP distinction, it be-
comes necessary to specify what features are relevant for allowing MBP to
have phonologically null P in dative contexts. We claim that a (oblique)
Case feature, along with semantic/thematic features are enough to account
for the distribution of a and para in both MBP and BP. Once these VI fea-
tures are settled, we will show that a microparameter, specific to MBP and
BP will derive the preposition-less sentences, in a way compatible to the
features specified by the VIs in MBP.
As we have demonstrated in section 2, the dynamics of the prepositions
in BP and MBP seem to play an important role – or, at least, indicate a great
diagnostic tool to understand – the surface realization of the dative argu-
ments in these two dialects. Thus, it is quite relevant that a and para are
differentiated: while para is ambiguously used in BP both as introducing a
goal or introducing a beneficiary (cf. (14), above), a in MBP can only intro-
duce goals7. Torres-Morais & Berlinck (2006) bring evidence that the pro-
nominal system in BP might be going through a simplification process, in
which dative case is disappearing. Except for some regional variants, the use
of a keeps reducing in a widespread manner. If MBP is the dialect in which
the semantic feature [dative] still plays a role in its distribution, we will
7
It is important to notice that the ambiguity in para does not concern MBP, once the only
cases in which it is used are the ones with benefactive elements.
claim this feature is exclusive of this MBP VI – a dialect where this distinc-
tion is still active.
We believe that the order V+DP+DP in MBP is restricted to the non-
-insertion of VIs in the contexts where it occurs. In order to determine how
that happens, it is necessary to define the features of VI, so their (non-) in-
sertion can be explained. The relevant VI feature list is presented in (35):
(35) Dative Vocabulary Items for MBP:

a _ [dative]
para _ [oblique Case]
With the features in (35), we guarantee that the P nodes where its com-
plement needs to be interpreted as goal receive a, but not para. Note that,
with this configuration, a presents little restrictions, and could appear in
Case-less contexts8. However, that would violate any Case requirements in
the grammar. Later, we will give an explanation as to why there is no such
occurrences, even maintaining this VI feature set.
The VI description for BP, on the other hand, has the [oblique Case] fea-
ture for a being obligatory, while para will be inserted in nodes where the-
matic role is unspecified, but Case is present:
(36) Dative Vocabulary Items for BP

a _ [dative], [oblique Case]
para _ [oblique Case]
In the traditional sense in the literature, the arrangement for MBP’s VI

features suggests that, in MBP, para behaves like a “lexical” preposition
(thematically less restricted, Case-bound), while a is more akin to a “func-
tional” preposition (thematically fixed, Case-“free”). Once we consider a
defective for Case, there may be two situations in which a can appear at a
given head P: P = [dative], and P = [dative], [oblique Case]. For the dative
contexts in BP and MBP where P has the [oblique Case] feature, the VI in-
sertion possibilities are the following:
8
An anonymous reviewer pointed out that one could argue that a is not a preposition, but a
Spanish-Style case-marker. It seem that the analysis of the Vocabulary Item gives us the
opposite result. Instead of marking Case, the insertion of a is restricted by the semantics,
rather than its Case-marking abilities. Rather than implying that Case is in the preposition
per se, the Case marking features are present in the functional structure P. If that node is
specified with a dative semantics, the VI a is inserted.
(37)
When it comes to BP, the structure in (38), below – a version of (37) de-
prived of the [oblique Case] feature in P –, is ungrammatical, since no VIs
lack the [oblique Case], blocking insertion according the Subset Principle,
which states the following:
“The phonological exponent of a vocabulary item is inserted into a position if

the item matches all or a subset of the features specified in that position. Inser-
tion does not take place if the vocabulary item contains features not present in
the morpheme. Where several vocabulary items meet the conditions for inser-
tion, the item matching the greatest number of features specified in the termi-
nal morpheme must be chosen.”
(Embick & Noyer, 2007: 298)
(38)
(39)
Having that question in mind, we suggest that the distinction on the dis-
tribution of datives in BP and MBP should be made in the following manner:
in the MBP contexts where P lacks phonological content, a is not inserted.
This is accomplished due to the parameter in (33).
(40) The Prepositional Case-marking Parameter

Case marking by P is obligatory
BP = {Yes}
MBP = {No}
Why is this parameter necessary? As Scher (1996) and Armelin (2011)

have already pointed out, the goal+theme order in MBP seems to trigger
some sort of vP left periphery displacement of arguments (cf. Belletti, 2002).
In order to account for the phenomena, we suggest that, whenever P lacks
Case, the sequence P+DP moves to either a topic (for verb+goal+theme) or a
focus position (for verb+theme+goal, in V+DP+DP contexts). These
Top/Foc projections mark the moved element with default Case – which is
nominative in BP, according to Kato (1999, 2012). Because of this special
condition on Case-marking, the insertion of the VI a does not occur, what
generates the surface resemblance to dative shift in MBP.
Therefore, whenever P on MBP lacks the [oblique Case] feature, move-
ment occurs, resulting in the structures below:
(41)
Yet, we still need to account for the goal+theme contexts where the prep-
osition’s phonological content is actually there. We will say that these sen-
tences have the [oblique Case] feature, and VI insertion at P occurs normal-
ly, either in BP or in MBP. This displaced argument is a byproduct of
movement by informational requirements – namely, contrastive focus, as
shown in (10), in section 2, above. That is attested by a (sometimes required)
intonational contour for pronouncing the goal argument. The resulting struc-
ture would be as in (42)9:
(42)
An interesting piece of independent evidence supports the current analy-

sis: Kato (2012) shows that CP topic positions allow the absence of func-
tional prepositions, what may indicate that default Case also appears to man-
ifest in CP positions.
9
It is important to notice that this structure would also be adequate for the cases in which
there is contrastive focus in English, generating the goal PP in the order verb+goal+theme.
(43) a. O Pedro não gosta de mimdat/obl.

The Pedro no likes of medat/obl.
‘Pedro does not like me’.
b. De mimdat/obl o Pedro não gosta.
Of medat/obl the Pedro no likes
c. EUdefault, o Pedro não gosta (de mim)
Idefault, the Pedro no likes (of me)
d. *Mimdat/obl, o Pedro não gosta.
e. *D’EUdefault, o Pedro não gosta. a. Da Maria, eu gosto.
(Kato, 2012: 88)
Finally, we can provide answers to the questions formulated in section 1.

We have shown that, although there is a superficial resemblance to English,
MBP presents no “real” dative shift (DOCs). We have also argued in favor
of the idea that both English and MBP have post-syntactic (PF) dislocation
effects, but each for a different reason (defective phonology in English, in-
formational factors in BP/ Case in MBP). Finally, we have given arguments
that show that, even though the parametric distinctions between BP and Eng-
lish and between BP/MBP occur in features within P, there are different
parameters that explain the distribution of arguments in the dative construc-
tions in each language/dialect.
So far, we have seen that the question of parametric variation is not at all
easy to handle. When we observe the dative phenomena in BP, MBP and
English, it may look like the surface phenomena gives a clear-cut distinction
of what happens. Yet, the empirical tests and the structure we proposed
based on the evidence gathered suggests that MBP and BP are mere dialects.
Due to that, we claim that MBP and BP variation is microparametric. When
it comes to the distinction between BP and English, however, the dative shift
question seems to be much deeper effects into the syntactic/grammatical
structure in general. That allows us to suggest that those effects are the con-
sequence of a macroparameter, namely the one proposed in Barbosa (2012),
which affects the distribution of compounds, as well as the surface realiza-
tion of the prepositional node. Still, there is a lot to account for, such as the
main characteristics claimed to distinguish micro- and macrovariation phe-
nomena. Within DM – and the analysis we have proposed – the component
responsible for handling the parametric differences is the post Spell-out,
morphosyntactic one, not the syntactic nor the lexical one alone. We also
claim that the differences between the languages/dialects do not affect mean-
ing in most cases, but the surface disparities are byproducts of phonological
restrictions given by the parameters in question – more especially Case,
when it comes to MBP/BP. If our proposal is on the right track, we have

explained the “myth” of the DOCs in BP, and raised important questions
about the structuring of parametric variation, a field of many fruitful results
to come.
References
Armelin, Paula Roberta Gabbai (2011). Sentenças Bitransitivas do Português do
Brasil Revisitas à Luz da Teoria de Núcleos Funcionais Aplicativos. M.A.
Thesis, Universidade de São Paulo.
Barbosa, Julio (2012). Predicados Complexos e o Parâmetro de Composição: Um
Estudo Translinguístico. Ph.D. Dissertation, Universidade de São Paulo.
Barbosa, Julio & Simioni, Leonor (2011). Aquisição do Parâmetro de Composicio-
nalidade: uma Análise Translinguística. In. Rozana Reigota Naves & Heloísa
Maria Moreira Lima-Salles (Eds.), Estudos Formais da Gramática das Lín-
guas Naturais. [CD]. Brasília: Cânone Editorial, pp. 280-287.
Baker, Mark C. (1996). The polysynthesis parameter. Oxford: Oxford University
Press.
Baker, Mark C. (2008). The Macroparameter in a Microparametric World. In There-
sa Biberauer (Ed.), The Limits of Syntactic Variation. Amsterdam: John Ben-
jamins, pp. 351-374.
Barss, Andrew & Lasnik, Howard (1986) A Note on Anaphora and Double Objects.
Linguistic Inquiry 17, pp. 347-354.
Belletti, Adriana (2002). Aspects of the low IP area. In Luigi Rizzi (Ed.), The struc-
ture of IP and CP. The Cartography of Syntactic Structures, Vol. 2. Oxford:
Oxford University Press, pp. 16-51.
Bresnan, Joan; Cueni, Anna; Nikitina, Tatiana. and Baayen, Harald (2007). Predict-
ing the Dative Alternation. In Gerlof Bouma, Irene Krämer, and Joost Zwarts
(Eds.), Cognitive Foundations of Interpretation. Amsterdam: Royal Nether-
lands Academy of Science, pp. 69-94.
Bresnan, Joan & Nikitina, Tatiana (2009). The Gradience of the Dative Alternation.
In Linda Ann Uyechi and Lian-Hee Wee (Eds.), Reality Exploration and Dis-
covery: Pattern Interaction in Language and Life. Stanford, CA: CSLI Publi-
cations, pp. 161-184.
Bruening, Benjamin (2010). Double Object Constructions Disguised as Preposition-
al Datives. Linguistic Inquiry 41, pp. 287-305.
Di Sciullo, Anna Maria (2005). Decomposing Compounds. SKASE Journal of
Theoretical Linguistics 3, pp. 14-33.
Embick, David & Noyer, Rolf (2007). Distributed Morphology and the Syntax-
-Morphology Interface. In Gillian Ramchand & Charles Reiss (Eds.), The
Oxford Handbook of Linguistic Interfaces. Oxford: Oxford University Press,
pp. 289-324.
Halle, Morris & Marantz, Alec (1993). Distributed Morphology and the Pieces of
Inflection. In Ken Hale & Samuel Jay Keyser (Eds.), The View from Building
20: Essays in Linguistics in Honor of Sylvian Bromberger. Cambridge: The

MIT Press, pp. 111-176.
Harley, Heidi (1997). If you have, you can give. In Brian Agbayani and Sze-Wing
Tang (Eds.), Proceedings of the 15th West Coast Conference on Formal Lin-
guistics. Stanford, CA: CSLI Publications, pp. 193-207.
Kato, Mary Aizawa (1999). Strong and weak pronominals in the null subject pa-
rameter. Probus 11, pp. 1-37.
Kato, Mary Aizawa (2012). Caso inerente, caso default e ausência de preposições.
In: Adeilson Pinheiro Sedrins, Ataliba Teixeira de Castilho, Marcelo Amorim
Sibaldo, & Rafael Bezerra de Lima (Eds.), Por amor à Linguística: miscelâ-
nea de estudos linguísticos dedicados à Maria Denilda Moura, 1st ed. Ma-
ceió: Edufal, pp. 83-99.
Kayne, Richard S. (2008). Some Notes on Comparative Syntax, with Special Refer-
ence to English and French in Guglielmo Cinque & Richard Kayne (Eds.),
Handbook of Comparative Syntax. New York: Oxford University Press, pp.
3-69.
Larson, Richard K. (1988). On the double object construction. Linguistic Inquiry 19,
pp. 335-391.
Levin, Beth (2006). First Objects and Datives: Two of a Kind? Parasession on Theo-
retical Approaches to Argument Structure, 32nd Annual Meeting of the
Berkeley Linguistics Society. [Handout]. Berkeley: University of California.
Marantz, Alec (1997). No Escape from Syntax: Don’t Try Morphological Analysis
in the Privacy of Your Own Lexicon. University of Pennsylvania Working
Papers in Linguistics 4, pp. 201-225.
Olsen, Susan (2012). Semantics of Compounds. In: Claudia Maienborn, Klaus von
Heusinger & Paul Portner (Eds.), Semantics: an international handbook of
natural language meaning. Vol. 3. Berlin, Germany; Boston: De Gruyter
Mouton, pp. 2120-2150.
Ormazabal, Javier & Romero, Juan (2012). PPs without Disguises: Reply to
Bruening. Linguistic Inquiry 43, pp. 455-474.
Pesetsky, David (1995) Zero Syntax: Experiencers and Cascades. Cambridge: MIT
Press.
Rizzi, Luigi (1982). Issues in Italian syntax. Dordrecht: Foris.
Scher, Ana Paula (1996). As Construções com Dois Objetos no Inglês e no Portu-
guês do Brasil: um Estudo Sintático Comparativo. M.A. Thesis, Universida-
de Estadual de Campinas.
Snyder, William B. (1995). Language Acquisition and Language Variation: The
Role of Morphology. Ph.D. Dissertation, MIT. Cambridge: MIT Press.
Snyder, William B. (2001). On the nature of syntactic variation: Evidence from
complex predicates and complex word-formation. Language 77, pp. 324-342.
Torres-Morais, Maria Aparecida (2001). A Preposição e a Caracterização do Objeto
Indireto: Aspectos Sincrônicos e Diacrônicos. São Paulo: Universidade de
São Paulo. [Manuscript].
Torres-Morais, Maria Aparecida (2007). Os Dativos. Habilitation Thesis, Universi-
dade de São Paulo.
Torres-Morais, Maria Aparecida & Lima-Salles, Heloísa Maria Moreira (2010).

Parametric change in the grammatical encoding of indirect objects in Brazili-
an Portuguese. Probus 22, pp. 181-209.
Torres-Morais, Maria Aparecida & Berlinck, Rosane de Andrade (2006). A caracte-
rização do objeto indireto no português: aspectos sincrônicos e diacrônicos.
In: Tânia Lobo, Ilza Ribeiro & Zenaide Carneiro (Eds.), Novos Dados, Novas
Análises. Vol. VI, Tome I. Salvador: EDUFBA, pp. 73-106.
THE PEAR STORIES FILM: SIMPLE PRESENTS
AND PRESENT PROGRESSIVES IN PORTUGUESE,
SPANISH AND ENGLISH
STEPHEN FAFULAS
(East Carolina University)
ABSTRACT: The current study examines simple present and present progressive
forms in Brazilian Portuguese, Argentinean Spanish, and American English. It of-
fers a comprehensive cross-linguistic analysis of progressive aspect in the present
for a specific genre: simultaneous film narrations. In total, 30 participants watched
the Pear Stories Film and were asked to simultaneously tell the story with as much
detail as possible. Each present-time token was coded for the linguistic factors of
lexical aspect, adverb, clause-type, polarity, animacy of the referent, and whether
intervening material appeared between the auxiliary and gerund. Significant differ-
ences were found between each language leading to the conclusion that the progres-
sive has followed a somewhat different evolutionary trajectory in each language.
KEYWORDS: present progressive; film narration; Portuguese; Spanish; English.
1. Introduction
Croft (2010) analyzed 20 native-speaker English retell narratives of the ‘Pear
Stories Film’ (Chafe, 1980; http://www.pearstories.org/) and found extensive
morphosyntactic variation. He concludes that this method is appropriate for
viewing language change in synchrony. As Croft suggests, cognitive/func-
tional linguists are in need of ways to document grammatical change in real-
-time. The present investigation extends Croft’s analysis to simultaneous
narrations of the ‘Pear Stories Film’ and observes use of simple present and
present progressive forms in Brazilian Portuguese (BP), Argentinean Span-
ish and American English.
Diachronic grammaticalization paths for tense-aspect-mood expressions
display cross-linguistic similarities (Bybee, Perkins, & Pagliuca, 1994). Evo-
lutionary pathways of progressives and habituals are reflected in present-day
variation of form and function. Progressives exhibit variation of form within
62 Stephen Fafulas
the functional domain of imperfective aspect through the availability of mul-

tiple constructions for the expression of action simultaneous with speech
time. In addition, the simple present may compete with the present progres-
sive for the expression of this same function, displaying multifunctionality
through its various uses.
It has been documented (Bybee, Perkins, & Pagliuca, 1994) that the estar
+ V-ndo and be + V-ing progressives have developed from a locative to a
form denoting progressive aspect in many of the world’s languages. Spanish,
English and BP progressives all evidence a remarkable diachronic increase
in frequency, fusion (i.e., reduction of intervening material between the aux-
iliary verb and gerund), and expansion of contexts of use (Hundt, 2004;
Mendes, 2008; Torres Cacoullos, 2000). However, the possibility that in
each of these languages the progressive has expanded its contexts of use to
occupy specific functions has yet to be empirically in a single study. Similar-
ly, we know little about the impact of this diachronic increase of progres-
sives on other forms, such as the simple present. Moreover, previous anal-
yses limit their envelope of variation to the estar + V-ndo/be + V-ing form,
and not much is known about variation and use of other progressive con-
structions. The current study analyzes a range of progressive constructions,
including estar + V-ndo/be + V-ing, ir + V-ndo/go + V-ing, andar + V-
-ndo/go around + V-ing, and continuar/seguir + V-ndo/keep on + V-ing. In
addition, the present investigation explores the availability of simple present
forms for the narration of events simultaneous with speech time.
BP, Spanish, and English differ in their availability of forms for the ex-
pression of progressive aspect (Bybee, Perkins, & Pagliuca, 1994). The sim-
ple present and present progressive can both be used to express overt, ongo-
ing action in Spanish (Butt & Benjamin, 2000; Comrie, 1976). However, in
BP and English the use of the simple present is more restricted to habitual
reference and generic actions/states (Bybee, Perkins, & Pagliuca, 1994;
Cunha, 2004; King & Suñer, 1980; Whitlam, 2010). This is illustrated in the
following example which reflects the idea that in Spanish the simple present
and present progressive can both be used to encode the meaning ‘action in
progress’ (example 1), while in BP and English only one form (estar + V-
-ndo/be + V-ing) predominates for this function (example 1b).
(1) a. Simple present verb form:

BP *Olha, agora sai o sol.
Spanish Mira, ahora sale el sol.
English *Look, the sun comes out now.
b. Present progressive verb form:
BP Olha, agora está saindo o sol.
Spanish Mira, ahora está saliendo el sol.
English Look, the sun is coming out now.
The pear stories film 63
Empirical analyses suggest that a number of linguistic variables constrain

selection of simple presents and present progressives, including: lexical as-
pect, co-occurring adverbs, clause type, polarity and animacy (Cunha, 2004;
Mendes, 2010; Römer, 2005; Torres Cacoullos, 2000). However, the extent
to which these factors predetermine use of simple presents and present pro-
gressives similarly across BP, Spanish and English is not known. For exam-
ple, within the factor of lexical aspect, previous accounts of the Spanish es-
tar progressive (i.e., estar + V-ndo) reveal that this form is disfavored with
stative verbs (Fafulas, 2013; Torres Cacoullos, 2000) while in BP the estar
progressive is more widely employed with stative verbs (Bastos, 2004;
Cunha, 2004; Mendes & Howe, 2013). In English, the progressive is not
accepted with all stative verbs but it is expanding to more aspectual classes
(Aarts, Close, & Wallis, 2010). Example 2 displays the canonical use of the
estar/be progressive with stative verbs in each language.
(2) Progressive w/stative verbs: BP Estou gostando desse pão de queijo.

Spanish *Me está gustando este pan.
English ?I’m liking this cheese roll.
Authors such as Cunha (2004) highlight the difficulties in assigning a

single meaning or function to the progressive and simple present. Cunha
points to the availability of the progressive in Portuguese for the expression
of habitual states of affairs, as in “Patrícia está chegando tarde ao escritório
todos os dias (?Patricia is arriving late to work everyday).” Fafulas (2012)
makes a similar claim for the use of the progressive with repetitive adverbi-
als in Spanish, pointing to the more general acceptability of this function in
Spanish as compared to English. As well, Cunha (2004) offers examples of
the simple present to express “action simultaneous with speech time”, as in
the narration of sports events; what is commonly referred to as the reporta-
tive function: “Ronaldinho passa a bola para Tevez, Tevez corre em direção
ao trave de gol, ele chuta a bola...(Ronaldinho passes the ball to Tevez,
Tevez runs toward the goalpost, he shoots the ball...).” This narra-
tive/discourse function allows for the perspective of the speaker, who at-
tempts to relate the events as punctual rather than occuring over an extended
period of time (Kamp & Rohrer, 1983). Thus, viewpoint aspect or personal
perspective is an important factor to be considered in the narration of events.
2. Grammaticalization of simple presents and present progressives

Responsible for much of the morphosyntactic variation observable in syn-
chrony is grammaticalization. Grammaticalization is defined as the parallel
set of processes whereby a specific lexical item, or string of items, in a con-
64 Stephen Fafulas
struction becomes autonomous from other instances of the construction,

resulting in new pragmatic and semantic uses distinct from the source con-
struction (Bybee, 2010: 106-7). As a construction takes on new meaning it
may slowly encroach upon the domain of other forms serving similar func-
tions, thereby producing variation (layering) (Hopper, 1991; Torres
Cacoullos, 2009). Grammaticalizing constructions continue to exhibit specif-
ic patterns and uses of their original forms (retention), adding to the compet-
ing variants for a given function (Torres Cacoullos & Walker, 2009).
Bybee, Perkins, and Pagliuca (1994) studied seventy-six languages repre-
senting the major typological groupings of the world and found that progres-
sives commonly derive from locative expressions and begin with the mean-
ing ‘animate subject is located somewhere doing something’. In addition,
progressives can further grammaticalize into imperfectives or presents. Par-
ticular instances of these patterns are estar ‘be located’, ir ‘go’, or andar ‘go
around/walk’ + V-ndo which have grammaticalized from lexical spatial ex-
pressions into grammatical aspectual morphemes, following the cross-
-linguistic imperfective grammaticalization path set forth by Bybee, Perkins,
and Pagliuca (1994) in Figure 1.
locative or movement > progressive > general imperfective/present
Figure 1. Cross-linguistic progressive grammaticalization path

(Bybee, Perkins, & Pagliuca, 1994)
In the evolution of lexical into more grammatical material, constructions

may become obligatory. The creation of obligatory categories is reflected
through a rise in frequency as forms come to be used in redundant and non-
-redundant categories (Bybee, Perkins, & Pagliuca, 1994). While the gram-
maticalizing construction increases in frequency it undergoes parallel seman-
tic and structural changes. In turn, the construction generally becomes less
marked and expands its contexts of use. Dramatic frequency rises have been
witnessed for English and Romance progressives. Visser (1973) estimates
that the English be + V-ing form has doubled every century since the 1500s.
Torres Cacoullos (2000) finds a similar tendency for the Spanish estar + V-
-ndo form. Squartini (1998) reports that Portuguese has a widely employed
progressive that developed from the Latin verb stare. Mendes (2008) con-
firms that the BP estar + V-ndo has undergone the same grammaticalization
process described for Spanish and English. Note that in European Portuguese
estar + V-ndo is less frequently used than the infinitival gerund (e.g. estar a
‘be at’ + infinitive) but in BP it is overwhelmingly favored (Hutchinson &
Lloyd, 2003; Thomas, 1969; Whitlam, 2010).
The current study examines synchronic data to explore whether previous

diachronic accounts of the linguistic patterning of simple presents and pre-
sent progressives still hold. This is particularly important because the major
assumptions of grammaticalization theory and the cross-linguistic tendencies
of tense-aspect morphemes are largely on written data from some twenty or
more years ago. For example, Dahl’s (1985) groundbreaking work on the
universality of developing morphemes consisted of data obtained with a
questionnaire that prompted speakers to translate approximately 160 sen-
tences into their native language. Results of the study provided strong evi-
dence that the elements (i.e. grams/morphology) of which tense-aspect sys-
tems are constructed in human languages are reducible to a small set of
cross-linguistic types that can be classified by their prototypical functions.
Bybee (1985) used reference grammars for her analysis and found similar
results regarding the development of tense-aspect grams in the world’s lan-
guages. Bybee, Perkins, and Pagliuca (1994: 35) analyzed data from written
reference materials in seventy-six languages, all produced by authors with
varied training in linguistics. The present investigation builds on this previ-
ous work by exploring the extent to which cross-linguistic morphosyntax can
be observed in synchrony through simultaneous film narrations in three dis-
tinct languages.
3. The progressive: definitions and cross-linguistic differences

Attempts at defining the progressive abound in linguistic references. Quirk,
Greenbaum, Leech, and Svartvik (1985) argue that the progressive renders a
situation as: (1) durative, (2) limited, and (3) incomplete. Römer (2005)
holds that the progressive is made of two core notions ‘continuous’ and ‘re-
peated’. Jespersen (1931) points to the animate and subject-involved nature
of the progressive. Mindt (2000) adds to this list the functions of temporari-
ness and predication (future). Binnick (1991) speaks of the “temporal fram-
ing” function of the progressive. A precise definition of the progressive is
unattainable given that each language has come to encode progressive aspect
in its own way. Furthermore, aspect is dependent upon subjective interpreta-
tion of the boundedness of events (i.e., view point; Smith, 1997).
Languages such as English, Japanese, Brazilian Portuguese and Spanish
use aspectual markers (morphology) to indicate progressive meaning in the
present (nonpast). However, a number of languages, including German,
Modern Greek and Russian, do not rely on overt morphemes for indicating
progressive action and instead use the present tense form (see Binnick, 2012;
Comrie, 1976; Dahl, 2000). It is held that in BP and English, ongoing action
must be overtly expressed with the estar/be progressive (PP) while in Span-
ish also exists the possibility of using simple present (SP) morphology in
66 Stephen Fafulas
contexts of ongoing action (see King & Suñer, 1980; Klein, 1980; Whitlam,
2010). Table 1 presents an overview of the general uses of each form for BP,
Spanish and English.
Language Action in Progress Form(s)

Nesse momento Rafael está falando/
BP PP
*fala ao telefone com sua mãe
En este momento Rafael está hablando/
Spanish PP/SP
habla por teléfono con su mama
At this moment Rafael is talking/
English PP
*talks on the phone with his mom
Table 1. Availability of Simple present (SP) and Present progressive estar/be form (PP)
for expression of action in progress in Brazilian Portuguese (BP), Spanish and English
Table 1 initially supports claims that Spanish shows an optionality of forms

in referencing ongoing action while BP and English do not. However, empiri-
cal data is needed. Moreover, the actual situation is more complex than this,
especially when one considers other-based progressive forms (ele an-
da/vai/vem pensando que… “he goes around/goes thinking that…”), lexical
aspect (stative vs. dynamic predicates), and adverbs (ahora/agora/now). Gen-
re effects may also be crucial in this distinction. Torres Cacoullos (2000) finds
that the Spanish estar + V-ndo progressive is more than three times greater in
popular oral varieties than in formal literary works. One must also consider the
less canonical functions of these forms, such as the progressive in frequenta-
tive contexts and the simple present in narrating sports and historical events.
Thomas (1969: 198) notes that the distinction between the SP and the PP
in spoken BP has reached obligatory status. Whitlam (2010) adds that the BP
SP and PP function similarly to these forms in English. Thomas (1969: 206)
notes that the estar progressive is the most frequent and neutral of the avail-
able progressive constructions in BP, not adding additional semantic content
to the progressive form, although it does retain some traces of its locative
origins, as seen by its preference for intervening locative adverbs between
the auxiliary and gerund (i.e., V-ndo). Thomas lists the ir + V-ndo construc-
tion as the second most frequent progressive in spoken colloquial BP.
Thomas asserts that andar + V-ndo is less frequent than the ir + V-ndo con-
struction. The author also comments that the vir + V-ndo construction ex-
presses progress in time or space toward a specific location or referent.
Whitlam (2010) highlights among the available progressive constructions in
BP: acabar + gerund ‘end up doing something’, andar + gerund ‘have been
doing’, ir + gerund ‘keep doing, get on with doing, go along doing’, vir +
gerund ‘have been doing’, and viver + gerund ‘to always be doing some-
thing.’ Fafulas (2015) holds that multiple progressive constructions are
available in Spanish. His results also point to, albeit more limited than in
Spanish, availability of multiple progressive forms in English. The current
study tests the availability of these forms and their linguistic predictors of
use in American English, BP, and Argentinean Spanish.
4. Current study
4.1. Methodology
This investigation explores a corpus of simultaneous oral narratives in
three languages and asks:
(i.) What is the frequency of use of present progressives (aux + V-ndo/-ing)
and simple presents?
(ii.) Are there differences between American English, Argentinean Spanish, and BP?
(iii.) Which linguistic factors are correlated with the use of each form in each of
these languages?
The participants for the current study were all native speakers of their re-
spective languages. The relevant information for each participant group is
summarized in Table 2.
Each participant watched the ‘Pear Stories Film’, a movie that contains
sound but no dialogue. Participants were asked to simultaneously tell the
story with as much detail as possible as if telling it to a friend on the phone
that could not see the video. Participants completed a brief background ques-
tionnaire (either written or orally) from which the participant information
presented in Table 2 was derived.
Mean
Group Participants Inclusion Criteria
Age
All living in Rio de Janeiro, BP-speaking
Brazilian 10 parents, limited English and Spanish
38.5
Portuguese (3M/7F) proficiency, limited experience in Eng-
lish and Spanish speaking countries
All living in Buenos Aires, Spanish-
Argentinean 10 -speaking parents, limited experience in
38.8*
Spanish (5M/5F) English and Portuguese-speaking coun-
tries, majority from River-plate region
All born in USA, English-speaking par-
American 10
19.6 ents, no significant time abroad in Span-
English (4M/6F)
ish or Portuguese-speaking countries
*= A handful of participants in this group did not give their age; calculations are based on
approximate/average ages.
Table 2. Participant Groups of the current study
68 Stephen Fafulas
First, each digital audio recording obtained from the simultaneous narra-
tion of the ‘Pear Stories Film’ was transcribed. Subsequently, all verb forms
were tabulated and the range and frequency by group calculated. Next, each
token of a verb produced in a context with present time reference was sepa-
rated for coding. The initial analysis concerned frequency and use of the
variants: present tense and estar/be progressive forms. Then similar counts
and analyses were conducted for other-based auxiliaries with V-ndo/-ing
(e.g. andar/to go around, venir/vir/come along, ir/to go, seguir/conti-
nuar/keep on + V-ndo/-ing). Forms such as existential + V-ndo/-ing and
elliptical/bare gerunds were excluded, given these did not match the criterion
of a progressive construction as a periphrastic formed of an auxiliary verb
(e.g. estar/be, ir/go, andar/go around) and a present participle (Verb + -an-
do, -endo, -iendo, -ing, etc.). In addition, all tokens of estar/ser ‘to be’ as a
main verb and ter / hay ‘there is/are’ functioning as existentials were exclud-
ed given these were invariantly produced with the simple present. BP infini-
tival forms, such as ‘continua a tirar’ (4 tokens total) were coded as simple
present given their lack of progressive morphology.
After the frequency of forms was tabulated, each form was coded for
these linguistic variables:
(a) Adverbial phrase (locative, temporal, other (e.g., sequential), none)
(b) Lexical aspect (statives, activities, achievements, accomplishments)
(c) Clause type (subordinate vs. main)
(d) Animacy of the subject (animate vs. inanimate)
(e) Polarity of the clause (negative particle present?)
(f) Intervening material between the auxiliary verb and the gerund (yes/no?)
Example of coding:
O homem está realmente colhendo pêras da árvore
‘The man is really gathering pears from the tree’
[estar progressive: locative, activity verb, main clause, animate subject, no
negative, yes intervening]
4.2. Examples of narrations from each language

A representative sample of tokens from the corpus of narrations for each
language follows.
4.2.1. Brazilian Portuguese
Simple Present
O menino coloca a cesta na bicicleta
“The boy places the basket on the bicycle”
Estar + V-ndo
(Es)tá realmente colhendo pêras com uma vestimenta…
“(He) Is really collecting pears with an apron…”
Other-base + V-ndo
O homem não se da conta e continua colhendo seus frutos
“The man does not notice and continues collecting his fruits”
4.2.2. Argentinean Spanish

Simple Present
Viene otro, otro campesino, bueno, alguien
“Comes another, another farmer, well, someone”
Estar + V-ndo
Estamos viendo un paisaje, tranquilo
“We are seeing a landscape, tranquil”
Other-base + V-ndo
…pero va medio tambaleándose
“…but (he) goes kind of wobbling”
4.2.3. American English

Simple Present
The kid steals a pear
Be + V-ing
And now the man is dumping pears out of his apron type bag
Other-base + V-ing
The man keeps picking pears and all the sudden there’s another man who
rides by….
5. Results
5.1. Distribution of forms
The overall distribution of forms can be seen in Table 3. In total, participants
produced 2314 verbal predicates for analysis. As anticipated, participants pro-
duced mostly present tense forms. However, the languages differ in their use of
other tensed verbs and infinitives. This could mean that each language allows
for different verbal tenses and moods in the narration of present-tense actions.
American Argentinean Brazilian

Forms
English Spanish Portuguese
Present Tense (619) (703) (574)
Verbs 93.2% 81.9% 72.5%
Other Tensed (45) (155) (218)
Verbs + Infinites 6.8% 18.1% 27.5%
(664) (858) (792)
Total Verbs
100% 100% 100%
Table 3. Distribution of all forms in simultaneous oral narrative, Total Tokens: 2314
70 Stephen Fafulas
As the object of analysis centers on the variation of simple presents and

present progressives, the remainder of the results will include present tense
tokens only. This distribution is displayed in Table 4.

Forms
(350) (617) (422)
Simple Present
56.5% 87.8% 73.5%
Estar + V-ndo (247) (29) (73)
Be + V-ing 39.9% 4.1% 12.7%
Other-base + (22) (57) (79)
V-ndo/-ing 3.6% 8.1% 13.8%
(619) (703) (574)
Totals
100% 100% 100%
* all groups show significance (p < .001) for distribution of forms
Table 4. Distribution of present tense forms by group*
in simultaneous oral narratives, Total Tokens:1896
Table 4 reveals that the simple present was the most frequently used form
by the groups in the current study. However, the frequency of use of each
form is noticeably different across language groups. Of the three languages,
Argentinean Spanish allows for the most use of the simple present in the
corpus of simultaneous narrations. This finding might affirm claims that the
simple present is more available for the expression of action in progress in
Spanish than in BP and English. At the same time, the simple present is also
frequent in BP. It is known that the simple present is available for other
functions such as historical narrations and in the reporting of sequential
events. Thomas (1969: 117) observes that in Carioca Portuguese narrations,
the indicative may replace the preterite or imperfect. Cunha (2004: 230)
holds that the simple present is frequent in sports narrations and as a rhetori-
cal device to project past events into the present and add a quality of “liveli-
ness” to the discourse. While future research is necessary, the results of the
current study indicate that the simple present is more available for ongoing
action with speech time than previously claimed, and this function is strong-
er in Argentinean Spanish than in BP or American English. Equally as re-
vealing is the use of the estar/be and other-based progressives by each
group. For the American English group, be + V-ing constitutes 39.9% of
their total forms. Among the BP group the estar + V-ndo form makes up
12.7% of their total forms. Finally, the Argentinean Spanish group uses estar
+ V-ndo at a rate of only 4.1% but they nearly twice the amount of other-
-based progressives. The BP group also uses more other-based progressives
than estar + V-ndo. In summary, be + V-ing in American English is three
times more prevalent in simultaneous narrations than the estar + V-ndo form
in BP, and almost ten times more common than estar + V-ndo in Argentinean
Spanish. Equally as notable, other-based progressives are more frequent over-
all than estar + V-ndo in the Romance languages, but not in English.
5.2. Linguistic variables
The remaining analyses are concerned with the linguistic variables asso-
ciated with the use of each form by each language group. In each table, SP
stands for Simple Present, EPP Estar/Be Present Progressive, and OPP for
Other-Based Present Progressives. Each table displays the results of the
across-category distribution for each form and the linguistic variable in ques-
tion. That is, the percent and token count of each form used across all cate-
gories of the linguistic factor are presented. The across-category distribution
reveals the total use of each form (approximately 100%) across all categories
of the factor in question. For example, consider the American English group
in Table 5; the across-category division is: SP 39.1% with stative predicates,
3.4% with activity predicates, 38.0% with accomplishment predicates, and
19.4% with achievement verbs. This is the total distribution of simple pre-
sent forms across the four categories of the factor lexical aspect for this
group. The same holds for the present progressive be + V-ing and other-
-based progressive. The results of the individual chi-squares appear in the
discussion section.
5.2.1. Lexical aspect
As seen in Table 5, American English and BP allow for the pairing of
progressive morphology with stative verbs while Argentinean Spanish does
not. American English also allows for more progressives with accomplish-
ments than do BP or Argentinean Spanish. In fact, Argentinean Spanish dis-
allows the estar progressive with accomplishment predicates. While the es-
tar progressive is highest with activity verbs for BP and Argentinean Span-
ish, the ‘be’ progressive is much lower in American English.
Aspect Stative Activity Accomplishment Achievement

Predicates Predicates Predicates Predicates
Group SP EPP OPP SP EPP OPP SP EPP OPP SP EPP OPP
American (137) (18) (0) (12) (101) (7) (133) (117) (12) (68) (11) (3)
English 39.1 7.3 0 3.4 40.9 31.8 38.0 47.4 54.5 19.4 4.5 13.6
Argentinean (208) (0) (0) (92) (26) (50) (103) (0) (2) (214) (3) (5)
Spanish 33.7 0 0 14.9 89.7 87.7 16.7 0 3.5 34.7 10.3 8.8
Brazilian (99) (4) (2) (81) (54) (52) (141) (14) (11) (101) (1) (14)
Portuguese 23.5 5.5 2.5 19.2 74.0 65.8 33.4 19.2 13.9 23.9 1.4 17.7
Note: SP=simple present; EPP=estar/be progressive; OPP=other-based progressive
Table 5. Distribution of Lexical Aspect by form and group in simultaneous narrations
72 Stephen Fafulas
Each language shows a different affinity for pairing of morphology and

lexical aspectual type. While the prototypical [activity verb + progressive]
paring is strongest in Argentinean Spanish and BP, American English allows
for a much wider distribution of ‘be’ progressives across lexical aspectual
predicates, potentially indicating that it is more grammaticalized in this lan-
guage than in the Romance languages. The simple present form as well as
other-based progressives spread more evenly across aspectual types in Bra-
zilian Portuguese than in the other two languages. There are also similarities
across these languages in regard to prototypical morphological-lexical aspec-
tual parings. For instance, the simple present is more commonly paired with
stative verbs than with activity verbs and this is true for all three languages.
5.2.2. Clause Type

Table 6 reveals that the estar progressive is more likely to appear in sub-
ordinate clauses in Argentinean Spanish (48.3%) than in BP (16.4%), and
more than the ‘be’ progressive in American English (10.1%). Still, all groups
show a higher use of estar/be progressives in subordinate clauses than they
do simple presents. A main function of the progressive is to create a tem-
poral frame around another situation (Jespersen, 1931). This coincides with
the functions of subordinate clauses, in addition to the functions of the pro-
gressive to provide background information in narrations, as simple presents
are typically used in foregrounded clauses to move the storyline forward
(Bardovi-Harlig, 2000). While the results concur with these affirmations, it
appears that there is cross-linguistic variation in the degree to which this
strategy is applied in simultaneous film narrations. Future analyses will need
to observe this finding in more detail.
Clause Type Simple Clause Subordinate Clause

Group SP EPP OPP SP EPP OPP
American (331) (222) (20) (19) (25) (2)
English 94.6 89.9 90.9 5.4 10.1 9.1
Argentinean (518) (15) (47) (99) (14) (10)
Spanish 84.0 51.7 82.5 16.0 48.3 17.5
Brazilian (384) (61) (78) (38) (12) (1)
Portuguese 91.0 83.6 98.7 9.0 16.4 1.3
Table 6. Distribution of clause type by form and group in simultaneous narrations
5.2.3. Animacy
Table 7 reveals that the majority of subject tokens in the ‘Pear Stories
Film’ were animate. The results show a stronger favoring of animates with
progressive morphology (both estar/be and other-based progressives) than
with simple present morphology. While both simple present and present
progressive forms are favored with animates, there is a stronger pairing of
animate and progressive morphology than there is between inanimate and
progressive. This is in line with the proposal that progressive aspect is heavi-
ly associated with dynamic situations, which are generally acted out by ani-
mate agents (Bybee, Perkins, & Pagliuca, 1994). Differences are also visible.
BP allows for the most estar progressives with inanimates (5.5%) and Ar-
gentinean Spanish the most other-based progressives with inanimates
(5.3%).
Animacy Animate Inanimate

American (280) (242) (22) (70) (5) (0)
English 80.0 98.0 100 20.0 2.0 0
Argentinean (526) (29) (54) (91) (0) (3)
Spanish 85.3 100 94.7 14.7 0 5.3
Brazilian (385) (69) (78) (37) (4) (1)
Portuguese 91.2 94.5 98.7 8.8 5.5 1.3
Table 7. Distribution of animacy of the agent by form and group in simultaneous narrations
5.2.4. Polarity
Table 8 displays the results for polarity and form by language group.
According to Torres Cacoullos (2000) negative clauses are more conserva-
tive and display older patterns of the language. Negative polarity contexts
are more resistant to change and the extension of new uses of a given lin-
guistic form. If the present progressive is the form undergoing change,
expanding its use to more contexts, it is predicted that the simple present
will be preferred in negative polarity contexts. Torres Cacoullos (2000)
adds that negative polarity situations will disfavor the estar + V-ndo pro-
gressive precisely because positive polarity is consistent with overtness
which is one of the core features of the progressive. As predicted, the data
show that the estar + V-ndo, as well as other-based progressives, are pro-
hibited in negative polarity contexts by the BP and Argentinean Spanish
groups. Although not favored, the ‘be’ progressive is possible in negative
polarity contexts for the American English group. This could potentially
signal that the progressive is more grammaticalized in American English
than in these other two languages.
74 Stephen Fafulas
Polarity Negative Not-Negative

American (11) (6) (0) (339) (241) (22)
English 3.1 2.4 0 96.9 97.6 100
Argentinean (29) (0) (0) (588) (29) (57)
Spanish 4.7 0 0 95.3 100 100
Brazilian (18) (0) (0) (404) (73) (79)
Portuguese 4.3 0 0 95.7 100 100

Table 8. Distribution of negative clause by form and group in simultaneous narrations
5.2.5. Intervening elements

As the degree of intervening elements only concerns progressive forms,
the simple present is excluded from this analysis. The gradual decline of
intervening material in diachronic observations of the progressive has been
used as a criterion for evaluating the degree of fusion between emerging
grammatical morphemes and the verbs they attach to in the construction
(Bybee, Perkins, & Pagliuca, 1994). In Spanish, the degree of intervening
material is predicted to be less than in BP (Mendes & Howe, 2013). The
results presented in Table 9 confirm the prediction that BP (20.5%) should
allow for more intervening elements (i.e., adverbs) between the auxiliary
estar and the gerund (V-ndo) as compared to Argentinean Spanish (6.9%).
American English falls in the middle (12.6%) of these rates. Interestingly,
the amount of intervening material allowed in other-based progressives is
greater than that allowed for in the estar/be progressive, and this result is
consistent across all languages. This is initial evidence that these construc-
tions are not as tightly fused as the estar/be progressive (see Bybee & Torres
Cacoullos, 2009).
Intervening Intervening material No intervening material

SP EPP OPP SP EPP OPP
Group
American (31) (4) (216) (18)
NA NA
English 12.6 18.2 87.4 81.8
Argentinean (2) (17) (27) (40)
NA NA
Spanish 6.9 29.8 93.1 70.2
Brazilian (15) (36) (58) (43)
NA NA
Portuguese 20.5 45.6 79.5 54.4

Table 9. Distribution of intervening elements in the periphrasis by form and group
5.2.6. Adverb
Lastly, the results and distribution of simple present and present progres-
sive forms based on adverbial type in the context are presented in Table 10.
All three groups display a similar pattern in that the be/estar progressive is
highest with no accompanying adverb, followed by locative adverbs. In fact
this pattern holds for other-based progressive constructions as well. Thus, in
the majority of contexts observed in the current study, morphological mark-
ing and the semantics of the verb are sufficient for expressing the nature of
the predicate. This is quite possibly a result of the task, which presented
action clearly identifiable in the video and subsequently may have required
fewer adverbial tools for specification. The finding that estar and ‘be’ pro-
gressives are common with locative adverbs is linked to their diachronic
development, as they originated out of verbs that were used to locate a sub-
ject in a place (Bybee, Perkins, & Pagliuca, 1994).
Other
Locative Temporal No
Adverb Types of Ad-
Adverb Adverb Adverb
verbs
Group SP EPP OPP SP EPP OPP SP EPP OPP SP EPP OPP
American (59) (59) (6) (36) (45) (2) (10) (2) (0) (245) (141) (14)
English 16.9 23.9 27.3 10.3 18.2 9.1 2.9 0.8 0 70.0 57.1 63.6
Argentinean (122) (7) (15) (37) (3) (6) (30) (0) (3) (428) (19) (33)
Spanish 19.8 24.1 26.3 6.0 10.3 10.5 4.9 0 5.3 69.4 65.5 57.9
Brazilian (127) (24) (22) (44) (14) (11) (44) (8) (9) (207) (27) (37)
Portuguese 30.1 32.9 27.8 10.4 19.2 13.9 10.4 11.0 11.4 49.1 37.0 46.8
Table 10. Distribution of adverbs in the context by form and group in simultaneous narrations
6. Discussion
The present investigation offers an empirical analysis of simple present and
present progressive forms in simultaneous narrations of the ‘Pear Stories
Film’. Croft (2010) used the ‘Pear Stories Film’ to show how morphosyntac-
tic variation might be observed in real-time, analyzing retells of the film by
native speakers of English. The current study extends Croft’s analysis with a
cross-linguistic observation of one linguistic function: the expression of on-
going action simultaneous with speech time. The current study takes a first
step in resolving previous uncertainties of the values and possible uses of
simple present and present progressive forms in Spanish, English, and BP
(see sections 1-3).
Results indicated both similarities and divergences across language types.
The overall distributions in the corpus of oral narratives revealed that the
76 Stephen Fafulas
American English group employed the ‘be’ progressive much more than the
Argentinean Spanish or BP speakers used the estar progressive. At the same
time, the Argentinean Spanish group used more simple presents than either
other group, and the BP group displayed the highest overall percentage of
other-based progressives. Furthermore, BP and Argentinean Spanish both
used more other-based progressive types than estar + V-ndo, while the
American English speakers used vastly more be + V-ing than other-based
progressives. Thus, it appears that in Argentinean Spanish and BP other-
-based progressives are much more commonly used in film narrations than
they are in American English. These finding add to those of Fafulas (2015)
who considered a range of progressive constructions in monolingual and
bilingual Spanish oral narrations and found that speakers from Spain used
more other-based progressives than estar + V-ndo while the opposite was
true for speakers from Mexico. The current study as well adds results for
different progressive constructions in BP. Future analyses will need to test
whether a similar diversity of progressive constructions is used in other gen-
res and speech contexts.
The linguistic factors observed in the current study also uncover signifi-
cant disparities between the three groups. Table 11 displays the chi-squares
for each and the distribution of simple presents and present progressives by
group. The results in Table 11 show that Argentinean Spanish and BP be-
have more similarly than either group does in comparison to American Eng-
lish. This is likely due to the historical origins of Portuguese and Spanish
(Clements, 2009).

Lexical Aspect .000*** .000*** .000***
Clause Type .094 .000*** .004**
Animacy .000*** .013* .051
(Negative) Pola- .631 .121 .035*
rity
Intervening Ele- .452 .015* .001**
ment
Adverb .005** .415 .365
Note: * p<.05, ** p<.01, *** p<.001

Table 11. Summary of results of chi-squares for linguistic factors in the current study
All three groups show a strong significance for lexical aspect. This lends
support to claims that inherent semantics of the verb predict morphological
pairing, making this a potential universal property of languages. Regarding
clause type there are cross-linguistic differences. American English does not
reach significance while Argentinean Spanish and BP do show significance
in their distribution of forms across main and subordinate clauses. Table 11
reveals differences in the cross-linguistic strength of the effects of animacy
on morphological distribution and tense/aspect marking. One can also note
in Table 11 that the distribution of forms in the corpus of simultaneous nar-
ratives is significantly affected by polarity for the BP group only. Table 11
illustrates that American English does not reach significance for the percent-
age of intervening elements and progressive type while in BP and Argen-
tinean Spanish there are significant differences. The results in Table 11 dis-
play further cross-language differences in that the American English group
reached significance for distribution of simple presents and present progres-
sives with adverbial types, while the BP and Argentinean Spanish groups did
not render significant results.
The current study holds that synchronic studies of grammatical variation
in semi-spontaneous speech should be compared to previous classifications
of cross-linguistic prototypical gram types in order to explore the extent to
which diachronic changes may have resulted in language-specific form-
-function mappings. It is through this method that new language-specific
changes and the expansion of previous usages of gram types and linguistic
forms can be uncovered. With this approach, the present investigation was
able to further advance our understanding of recent developments such as the
spread of progressives to stative verbs in BP and American English, but not
in Argentinean Spanish. Michaelis (2011: 1386) explores the stativity of
progressive phrases in English, citing this language as typologically rare in
that the simple present does not allow an “in-progress” reading of present-
-tense predications in which the event denoted by the verb is ongoing at
speech time. Michaelis also points out the fact that currently in other Ger-
manic languages, and in earlier stages of English, the simple present is used
to report an event ongoing at speech time. The present study found that the
simple present is used in narrative discourse to depict events “simultaneous
with speech time”, with all lexical aspectual classes in English, however this
specific function varies by language type and between typologically similar
languages (Spanish and Portuguese).
The differences prescribed by grammarians and researchers in as far as
the uses of the simple present and present progressive are at least partly
traceable to the specific diachronic developments of form-function parings
across languages. The present study holds that methodologies such as those
of Dahl (1985) and Bybee, Perkins, and Pagliuca (1994) should be comple-
mented by cross-linguistic studies of oral data. Usage-based researchers have
exposed the mechanisms by which once prevalent forms in a language,
which might still be taught as standard and cited as frequently used in
78 Stephen Fafulas
grammar references, are replaced by competing variants. For example, Pop-

lack and Dion (2009) found that there was little resemblance between ac-
counts offered in the literature and observations of speaker use of French
future forms in their corpus.
Importantly, the current study supports Croft’s (2010: 7-8) claim that alt-
hough morphosyntactic innovation and change is harder to capture in syn-
chrony than phonological innovation, it is, nonetheless, possible:
If first-order morphosyntactic variation is manifested in different ways of
‘‘saying the same thing’’, then we need to be able to identify ‘‘the same
thing’’… The best way to address this problem is to employ the experimental
method … and examine the same situation being verbalized by a relatively
homogeneous group of speakers under similar circumstances.
7. Conclusions
Through a corpus of simultaneous oral narratives of the ‘Pear Stories Film’,
the current study provided an overview of important similarities and differ-
ences in the use of simple present and present progressive forms by speakers
of BP, Argentinean Spanish, and American English. While each form is pos-
ited to have followed a similar grammaticalization pathway (Bybee, Perkins,
& Pagliuca, 1994), observable differences in frequency of use and linguistic
constraints lead to the conclusion that in each language these forms may
have developed somewhat differently and may currently be at different
points in the grammaticalization process. This idea is not new among usage-
-based linguists (see Bybee, 2010). If all languages followed the same pre-
cise path without divergence there would be much less grammatical varia-
tion and change. Thus, while diachronic accounts of the development of
tense-aspect forms offer the global patterns through which grammatical
forms may change, the current study follows in the work of Croft (2010) in
providing evidence of the necessity to complement these findings with syn-
chronic analyses of cross-linguistic grammatical variation in semi-
-spontaneous speech. Specifically, the findings of the present analysis sup-
port the claims of Mendes and Howe (2013) who hold that previous cross-
-linguistic comparisons of the progressive have obscured important lan-
guage-specific behaviors of these forms. The present investigation has ex-
tended this to more language types and included the simple present form to
explore the extent to which it competes with progressives in depicting “on-
going action simultaneous with speech time”. Importantly, the present inves-
tigation posited that other-based progressive constructions, aside from the
frequent be/estar progressive, compete in the domain of imperfective aspect,
and are essential in the analysis of the development and present-day use of
progressive and present forms. Without the inclusion of these constructions
our accountability of the forms available for the expression of simultaneous

action in discourse is lessened, and our emphasis on a few forms is subse-
quently inflated.
References
Aarts, Bas, Joanne Close & Sean Wallis (2010). Recent changes in the use of the
progressive construction in English. In Bert Cappelle & Naoaki Wada (eds.)
Distinctions in English grammar, offered to Renaat Declerck. Tokyo:
Kaitakusha, pp. 148-67.
Bardovi-Harlig, Kathleen (2000). Tense and aspect in second language acquisition:
Form, meaning, and use. Malden, MA: Blackwell.
Bastos, Ana Claudia Pinto. (2004). Progressive constructions in Brazilian Portu-
guese and English. Revista Letras 63, Curitiba, pp. 41-59.
Binnick, Robert (1991). Time and the verb: A guide to tense and aspect. New York:
Oxford University Press.
Binnick, Robert (ed.) (2012). The Oxford Handbook of Tense and Aspect. Oxford:
Oxford University Press.
Butt, John & Carmen Benjamin (2000). A New Reference Grammar of Modern
Spanish. London: Arnold.
Bybee, Joan (1985). Morphology: A study of the relation between meaning and form
(Vol. 9). John Benjamins Publishing.
Bybee, Joan (2010). Language, Usage, and Cognition. Cambridge: Cambridge Uni-
versity Press.
Bybee, Joan, Revere Perkins & William Pagliuca (1994). The evolution of grammar:
tense, aspect, and modality in the languages of the world. Chicago: Universi-
ty of Chicago Press.
Bybee, Joan & Rena Torres Cacoullos (2009). The role of prefabs in grammaticiza-
tion: how the particular and the general interact in language change. In Rob-
erta Corrigan, Edith A. Moravcsik, Hamid Ouali & Kathleen Wheatley (eds.)
Formulaic language, volume 1 Distribution and historical change. Amster-
dam: John Benjamins, pp. 187-217.
Chafe, Wallace (ed.) (1980). The Pear Stories: Cognitive, cultural, and linguistic
aspects of narrative production. New York: Ablex.
Clements, J. Clancy (2009). The linguistic legacy of Spanish and Portuguese: colo-
nial expansion and language change. Cambridge University Press.
Comrie, Bernard (1976). Aspect. New York: Cambridge University Press.
Croft, William (2010). The origins of grammaticalization in the verbalization of
experience. Linguistics 48, pp. 1-48.
Dahl, Östen (1985). Tense and aspect systems. New York: Basil Blackwell.
Dahl, Östen (ed.) (2000). Tense and aspect in the languages of Europe. Berlin/New
York: Mouton de Gruyter.
80 Stephen Fafulas
Fafulas, Stephen (2012). Nuevas perspectivas sobre la variación de las formas pre-
sente simple y presente progresivo en español y en inglés. Spanish in Context
9, pp. 55-87.
Fafulas, Stephen (2013). First and second-language patterns of variation: Acquisi-
tion and use of simple present and present progressive forms in Spanish and
English. Doctoral dissertation, Indiana University.
Fafulas, Stephen (2015). Progressive constructions in native-speaker and adult-
-acquired Spanish. Studies in Hispanic and Lusophone Linguistics 8 (1), pp.
85-133.
Hopper, Paul (1991). On some principles of grammaticalization. In Elizabeth C.
Traugott & Bernd Heine (eds.) Approaches to grammaticalization. Amster-
dam: John Benjamins, pp. 17-35.
Hundt, Marianne (2004). Animacy, agentivity, and the spread of the progressive in
Modern English. English Language and Linguistics 8, pp. 47-69.
Hutchinson, Amélia P. & Janet Lloyd (2003). Portuguese: An Essential Grammar.
New York: Routledge.
Jespersen, Otto (1931). A modern English grammar on historical principles. Part
IV: Syntax, Vol.3. London: George Allen & Unwin.
Kamp, Hans & Christian Rohrer (1983). Tense in Texts. In Rainer Bauerle, Chris-
toph Schwarze & Arnim von Stechow (eds.) Meaning, Use and Interpreta-
tion of Language. Berlin, Walter de Gruyter, pp. 250-269.
King, Larry & Margarita Suñer (1980). The meaning of the progressive in Spanish
and Portuguese. The Bilingual Review/La Revista Bilingüe 7, pp. 222-238.
Klein, Flora (1980). A quantitative study of syntactic and pragmatic indications of
change in the Spanish of bilinguals in the U.S. In William Labov (ed.) Locat-
ing language in time and space. Academic Press, New York, pp. 69-82.
Mendes, Ronald Beline (2008). Contraction, grammaticalization, and the phonolo-
gy/syntax interface: Progressives in Brazilian Portuguese. In Susie Jones
(ed.) 2008 CLA Conference Proceedings. Vancouver: Canadian Linguistics
Association, pp. 1-10.
Mendes, Ronald Beline (2010). Aspectual periphrases and syntactic variation in
Brazilian Portuguese. In James A. Walker (ed.) Aspect in Grammatical Vari-
ation. London: Routledge, pp. 27-47.
Mendes, Ronald Beline & Chad Howe (2013). “Progressive pathways and the estar
+ gerúndio periphrasis in Brazilian Portuguese”. Oral presentation: Inaugural
Portuguese Linguistics in the United States (PLUS) conference. Athens,
Georgia. November 14.
Michaelis, Laura A. (2011). Stative by construction. Linguistics 49 (6), pp. 1359-
-1399.
Mindt, Dieter (2000). An empirical grammar of the English verb system. Berlin:
Cornelsen.
Poplack, Shana & Nathalie Dion (2009). Prescription vs. praxis: The evolution of
future temporal reference in French. Language 85 (3), pp. 557-587.
Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech & Jan Svartvik (1985). A
comprehensive grammar of the English language. London: Longman.
Römer, Ute (2005). Progressive, patterns, and pedagogy: A corpus-driven approach

to English progressive forms, functions, contexts, and didactics. Amsterdam:
John Benjamins.
Smith, Carlota (1997). The parameter of aspect. Dordrecht: Kluwer.
Squartini, Mario (1998). Verbal periphrases in Romance: Aspect, actionality, and
grammaticalization. Berlin: Mouton de Gruyter.
Thomas, Earl W. (1969). The syntax of spoken Brazilian Portuguese. Vanderbilt
University Press.
Torres Cacoullos, Rena (2000). Grammaticization, synchronic variation, and lan-
guage contact: A study of Spanish progressive -ndo constructions. Amster-
dam: John Benjamins.
Torres Cacoullos, Rena (2009). Variation and grammaticisation: the emergence of
an aspectual opposition. In Stavroula Tsiplakou, Marilena Karyolemou &
Pavlos Pavlou (eds.) Studies in Language Variation: European Perspectives
II. Amsterdam/Philadelphia: John Benjamins, pp. 215-224.
Torres Cacoullos, Rena, & James A. Walker (2009). The present of the English
future: Grammatical variation and collocations in discourse. Language 85,
pp. 321-54.
Visser, Frederic Theodor (1973). An historical syntax of the English language 3/2:
Syntactical units with two and with more verbs. Leiden: Brill.
Whitlam, John (2010). Modern Brazilian Portuguese grammar: a practical guide.
Routledge.
FOCUS MOVEMENT AS PF MOVEMENT
AND OTHER PERIPHERAL POSITIONS IN BP
MARCELLO MODESTO
(Universidade de São Paulo)
ABSTRACT: This paper examines the interaction between topics and focused
phrases and their behavior with respect to reconstruction and minimality in Brazili-
an Portuguese; and shows that a topic may be moved across a moved focused
phrase, but may not move over a base generated focus or another (moved or base
generated) topic. Such state of affairs is explained if focused phrases (when moved)
are moved in the PF component of grammar, as argued by Aoun and Benmamoun
1996. Wh-phrases are argued to behave differently (being moved in the syntax) due
to the presence of a wh-feature, which is moved to the Spec of CP during the syntac-
tic derivation.
KEYWORDS: Topicalization; Focalization;Focus Movement; PF movement; Brazil-

ian Portuguese.
1. Introduction
In this work, we examine the interaction between topics, focused phrases,
wh-phrases and their behavior with respect to reconstruction and minimality
effects in Brazilian Portuguese (BP). It will be argued that both topics and
focalized phrases may move from their argument position to a dislocated
position in the left periphery of the sentence, or be merged already in that
position, binding a null or an overt pronoun at the thematic position. Moved
elements will display reconstruction, as expected. We assume, as commonly
done (cf. Lebeaux, 1990; Chomsky, 1995; Aoun et al., 2001), that recon-
struction implies movement. Since strong pronouns, not clitics, are used as
resumptives in BP, whenever there is resumption, it will be assumed that no
movement occurred. The interaction between topics, focused phrases and
wh-phrases with respect to reconstruction reveals the following generaliza-
tion: a topic may move across a moved focused phrase, but may not move
over a base generated focus or another (moved or base generated) topic.
84 Marcello Modesto
We argue that these interventions are instances of minimality effects

(Rizzi, 1990). Moreover, reproducing the argument in Aoun and Benma-
moun (1996), we conclude that focused phrases (when moved) are moved in
the PF component (in BP).
The argument is organized as follows. Section 2 contains the relevant da-
ta on BP. In subsection 2.1, we characterize Topicalization, Left Dislocation
and Focalization constructions, showing that both topics and focused phrases
can be moved or base generated in the dislocated position; section 2.2 argues
that only moved elements display reconstruction effects; subsections 2.3 and
2.4 examine the interaction between topics and focused phrases and topics
and wh-phrases, respectively. Section 3 explains those interactions as in-
stances of minimality. Section 4 argues that minimality is to be taken as a
condition on derivations (or on the application of move-) because Focaliza-
tion involves a non-syntactic movement, which takes place at the PF compo-
nent; section 5 offers a conclusion.
2. BP data
2.1. Topic and focus
Many authors have tried to define what is usually called Topicalization,
(Clitic-) Left Dislocation and Focalization (Chomsky 1981; Rizzi 1986;
Cinque 1990, a.o.). In this discussion, I will assume the cartographic view of
Rizzi (1997) in which topics and focused phrases need to be in some relation
to a specialized functional projection in the complementizer region of the
clause. Such a relation may come to be by external or internal merge, charac-
terizing what we will call movement. In other words, a topic may be merged
in its thematic position and then remerged in Spec TopP, or it may be
merged from the initial array in Spec TopP and be related to a thematic posi-
tion (occupied, in some languages, by a clitic, a weak, a strong, or a null
pronoun). I will assume that the projection of TopP is only triggered when
some constituent contains a Topic feature; Agree between the Top head and
the head of that constituent is necessary for convergence. Assuming also that
left peripheral positions such as topic and focus are “criterial” or “EPP” po-
sitions (i.e. they force the presence of some constituent in their specifier
position; by containing an EPP feature or some other technical machinery),
either the topic constituent is remerged in Spec TopP or another phrase from
the array is merged there. In this way, topic constructions in BP will present
the two possible structures depicted below:
Focus movement as PF 85
(1) a. [C [TopP esse livro [TP você devia dar esse livro pra Maria
this book you should give this book to Maria
(não pro Paulo)]].
not to Paulo
b. [C [TopP esse livro [TP você devia dar ele pra Maria
this book you should give it to Maria
(não pro Paulo)]].
not to Paulo
In (1), both sentences contain a dislocated element that may be character-

ized as a topic, it expresses old information and it is separated from the rest
of the clause (which expresses new information) by “comma intonation” (in
Rizzi’s (1997) terms). The structure in (1a) is usually called Topicalization;
the structure in (1b) is usually called Left Dislocation.1 Rizzi also recognizes
a formally similar but interpretively very different construction in English, in
which a focused phrase, possibly containing new information, appears at the
beginning of the sentence. We will call those structures, such as in (2), Fo-
calization:
(2) ESSE LIVRO você devia dar _ pra Maria (não uma agenda).2
this book you should give to Maria (not a calendar)
In this case, the fronted element, bearing focal stress, usually introduces
new information,3 whereas the open sentence that follows it expresses old
information, given by the discourse. In English, just like in BP, this structur-
al option is restricted to a specific kind of contrastive focus: (2) presupposes
that you believe that you should give a calendar to Mary, and tries to correct
that belief.
1
The analysis of Topicalization in English may not be as simple as described by (1a) for BP;
it may involve movement of a null operator. The assumption of a null operator, opposed to
actual movement of the focused phrase in Focalization, is used in Rizzi (1997) to highlight
his claim that Focus is quantificational while Topic is not, explaining for instance why only
Focus movement (and Topicalization in English) gives rise to WCO effects. I will disregard
these language particular idiosyncrasies.
2
Throughout this paper, capitalized letters indicate phrases receiving Extra High (Ex-H)
pitch, as defined in Zubizarreta (1998): a nuclear pitch accent that is higher relatively to the
other pitch accents in the intonational contour than in the case of “unmarked” main promi-
nence. Also throughout, an underscore marks the argumental position related to the dislo-
cated focused phrase.
3
One instance in which the fronted focused phrase constitutes old information, pointed out
by Zubizarreta (1998), is the case of answers to wh-questions where the alternatives are ex-
plicit in the discourse, as in the extract below:
(i) Q: What does John read every day: the news or the horoscope?
A: The NEWS John reads every day.
86 Marcello Modesto
Romance languages usually express the topic-comment articulation

(equivalent to English Left Dislocation) with the construction in (3), which
Cinque (1990) calls Clitic Left Dislocation (CLLD). Cinque argues CLLDed
phrases are not moved, but merged in their dislocated position, although the
construction is sensitive to strong islands and show connectivity effects.
Cecchetto and Chierchia (1999), however, argue that DPs are indeed moved
and only PPs are merged in their dislocated positions (however, it is unclear
how Cecchetto and Chierchia would deal with data involving subjacency
discussed by Cinque). In BP, third person clitics have been lost (at least in
the colloquial oral variety), so nothing like CLLD exists. BP seems to resort
to the same strategies used in English to express both the topic-comment and
the focus-presupposition articulations, i.e. movement (possibly of a null op-
erator) or base generation in the left periphery, possibly accompanied by a
resumptive pronoun, (see (4)).
(3) Il tuo libro, lo ho letto.

Your book I have read it
(4) a. Pra Maria, (eu acho que) você devia dar um livro _
to Maria (I think that) you should give a book
(não uma agenda).
(not a calendar)
‘To Maria, (I think) you should give a book’
b. Pra Maria, (eu acho que) você devia dar um livro pra ela
to Maria (I think that) you should give a book to her
(não uma agenda).
(not a calendar)
‘To Maria, (I think) you should give a book to her (not a calendar).’
c. PRA MARIA (eu acho que) você devia dar um livro _
to Maria (I think that) you should give a book
(não pra Joana/*não uma agenda).
(not to Joana/*not a calendar)
‘To Maria, (I think) you should give a book (not to Joana).’
A focused structure like (4c) is usually taken to involve movement (of

some phrase) to a focus-related position in the left periphery. This is due to the
fact that resumption is impossible with focalization. Structures involving top-
ics, on the other hand, are trickier. If BP topic structures are analyzed in the
same manner as in English, we may assume that the topic in (4b) is merged
already in the periphery, since it is accompanied by resumption; and the one in
(4a) may have been moved to that position. However, either (4a) is like (4c),
i.e. it involves movement, or it may have the same structure as (4b), but with a
null resumptive PP. In other words, a sentence like (4a) is ambiguous between
the structure (5a), in which the PP was moved, and (5b), in which the PP is
merged already in the periphery. In the examples in (4) and (5) we use PPs
instead of DPs to make clear that the topics (dislocated phrases) discussed here
are not “hanging topics” as defined by Benincà and Poletto 2004. However,
using PPs raises the question of what is the null resumptive category in the-
matic position, in such cases. I will simply assume that a null pronoun may
resume a PP, since its Case will be valued by binding of the dislocated topic.
Alternatively, the resumptive may be elided because it contains a subset of the
features present in the topicalized PP (see Roberts 2010), making the uttering
of it redundant. We have no space to discuss such matters. We will simply
assume that the structures in (5) both exist in BP: one in which there is move-
ment of some argument or adjunct to the left periphery, and one in which there
is no movement and a (possibly null) pronoun occupies the theta position re-
lated to the topic (be it a DP or a PP). I will also assume that overt resumption
implies lack of movement (although that may not be true of languages like
Italian, in which clitics are used for resumption).
(5) a. [Pra Maria, [eu acho que você devia dar esse livro pra Maria]
to Maria I think that you should give this book to Maria
b. [Pra Maria, [eu acho que você devia dar esse livro pro]
‘To Maria, I think you should give this book.’
That apparent Topicalization can be in fact Left Dislocation with a null

resumptive pronoun is shown by the fact that apparent cases of Topicaliza-
tion violate strong island contexts, from where wh-movement is impossible,
in the same way Left Dislocated elements do:4, 5
(6) a. (Eu acho que) (d)o Pedro, eu conheço uma mulher que gosta
(I think that) of.the Pedro I know a woman that likes
(dele).
(of.him).
‘I think that Pedro, I know a woman who likes him.’
b. *(De) quem voce conhece uma mulher que gosta?
(of) who you know a woman that likes
‘*Who do you know a woman who likes?’
4
There is a preference, in BP, to use the silent pronoun when referring to inanimate things
and a full pronominal when referring to animate beings. Disregard to that preference does
not lead to ungrammaticality.
5
The topics in (6-8) cannot be the hanging topics discussed by Benincà and Poletto 2004,
since they follow the complementizer in embedded clauses. In fact, BP does not have the
kind of hanging topic discussed by the Italian authors. BP has a kind of hanging topic found
also in Chinese, which is a topic that has no thematic place in the sentence, discussed later
in this work.
88 Marcello Modesto
(7) a. (Eu acho que) filmes de arte, ver (eles) em vídeo é perda de tempo.
(I think that) films of art watch (them) in video is waste of time
‘I think that art movies, to watch them on video is a waste of time.’
b. *O que ver em video é perda de tempo?
what to watch in video is waste of time
‘*What to watch on video is a waste of time?’
(8) a. (Eu acho que) o Pedro, a Maria só foi embora depois que
(I think that) the Pedro the Maria only went away after that
encontrou (ele).
saw (him)
‘I think that Pedro, Maria only went away after seeing him.’
b. *Quem a Maria só foi embora depois que encontrou?
who the Maria only went away after that saw
‘*Who did Maria only go away after seeing?’
Since extraction from strong islands is the relevant context, from (6-8),
we conclude that topic-comment structures in BP are ambiguous between
Topicalization, a syntactic derivation which involves movement, or Left
Dislocation, which involves no movement but external merge of a phrase in
a dislocated position related to a (silent or overt) pronoun in argument posi-
tion. When the topic is related to a position inside a (strong) island, as in (6-
-8), the structure is unambiguously a case of Left Dislocation, even if no
overt pronoun is present, since movement from strong islands is barred.
2.2. Island contexts and reconstruction
As expected, if Topicalization involves movement, reconstruction of top-
icalized phrases is possible:
(9) O seu melhor aluno, todo professor convidou _ pra fazer pós.
the his best student every professor invited (him) to do graduate
‘His best student, every professor invited him to enter the graduate program.’
In (9), we see that the pronoun seu (‘his’) may be bound by the quantifier
todo (‘every’). However, to have this bound reading, there must not be an
overt resumptive pronoun in the argument position related to the topicalized
phrase. With the overt pronoun, the only possible readings are the ones in
which seu assumes second person interpretation6 or it refers to some student
6
In BP, as the pronoun você (which is syntactically third person) replaced the second person
pronoun tu ‘you’, the third person possessive pronouns seu ‘his’ became ambiguous be-
tween second and third person interpretations, so the analytical form dele (literally of+him)
is used for third person in most cases, in order to avoid ambiguity. Therefore, a phrase such
in particular. This is what we expect: the version of (9) with an overt pro-
noun in argument position is a case of Left Dislocation, which involves no
movement and therefore cannot reconstruct the topic into the argument posi-
tion. The version of (9) with a gap is ambiguous between Topicalization and
Left Dislocation (using a silent pronoun). We can assume that, if the intend-
ed meaning of (9) requires reconstruction, the ambiguous sentence will be
analyzed only as Topicalization.
Following this reasoning, we predict that, in island contexts, reconstruc-
tion should be never possible. An ambiguous sentence between a topicalized
and a left dislocated structure will be analyzed as Left Dislocation only,
since no movement can happen from inside of a (strong) island. This predic-
tion is borne out:
(10) a. O seu melhor aluno, todo professor que convidou (ele) pra fazer
the his best student every professor that invited (him) to do
pós se arrependeu.
graduate refl regretted
‘Your best student, every professor who invited him to enter the grad-
program regretted it.’
b. O seu melhor aluno, que todo professor tenha aborrecido (ele)
the his best student that every professor have bothered (him)
é duvidoso.
is doubtful
‘Your best student, that every professor bothered him is doubtful.’
c. O seu melhor aluno, todo professor cumprimentou a família dele
the his best student every professor congratulated the family his
depois de sabatinar (ele).
after of questioning (him)
‘Your best student, every professor congratulated his family after ques-
tioning him.’
In either version of (10a-c), with an overt pronoun or a gap, the bound

reading is impossible and so the possessive pronoun seu is interpreted as
second person.7 The dislocated phrase cannot have been moved since it is
related to a position inside an island. Consequently, no reconstruction takes
place and the bound reading is impossible.
as “o seu melhor aluno” may be translated either as ‘his best student’ or as ‘your best stu-
dent’. This ambiguity will be crucial in telling when reconstruction is allowed.
7
The sentences in (10) are somewhat marginal if seu is interpreted as third person when it is
not bound. This is due to the fact, pointed out in Negrão and Müller (1996), that, in BP,
there has been a specialization of pronominal forms: seu (interpreted as third person) is used
in bound contexts (when the pronoun is bound by a quantifier) whereas dele (of+him) is
used otherwise.
90 Marcello Modesto
With respect to Focalization, it is shown in (11) that it is impossible to

have a resumptive pronoun in the “original” place of the focalized phrase,
which argues for a movement analysis:8
(11) a. O LUIZ eu vi _ ontem (não o Pedro).

b. *O LUIZ eu vi ele ontem (não o Pedro).
the Luiz I saw (him) yesterday (not the Pedro)
‘LUIZ, I saw him yesterday (not Pedro).’
However, based on (12a-c), we can assume that Focalization also violates

island constraints, which argues against a movement analysis:
(12) a. (Eu acho que) O PEDRO, eu conheço uma mulher que viu _ .
(I think that) the Pedro I know a woman that saw _
‘I think that Pedro, I know a woman who knows him.’
b. (Eu acho que) FILMES DE ARTE, ver _ em video é
(I think that) films of art watch _ in video is
uma perda de tempo.
a waste of time
‘I think that art movies, to watch on video is a waste of time.’
c. (Eu acho que) O PEDRO, a Maria só foi embora depois
(I think that) the Pedro the Maria only went away after
que viu _.
that saw _
‘I think that Pedro, Maria only went home after seeing him.’
There seems to be, then, two contradictory facts about Focalization: on

one hand, it does not allow for a resumptive pronoun in the argument posi-
tion related to the dislocated (focalized) phrase, which seems to indicate that
that phrase was moved from the argument position; on the other hand, Focal-
ization violates island constraints, which indicates that no movement is in-
volved. Interestingly, although an overt resumptive pronoun is excluded in
most of the cases, as seen in (11), they may appear if the position related to
the dislocated phrase is inside a (strong) island (13):
(13) a. (Eu acho que) O PEDRO,eu conheço uma mulher que viu ele
(I think that) the Pedro I know a woman that saw him
na festa .
at.the party.
‘I think that Pedro, I know a woman who saw him at the party.’
8
There may be some BP speakers who accept (11b), focalization with resumption; my own
judgment is only representative of the BP spoken in big cities, generally by educated peo-
ple.
b. (Eu acho que) FILMES DE ARTE, ver eles no video

(I think that) films of art watch them in video
é perda de tempo.
iswaste of time
‘I think that art movies, to watch them on video is a waste of time.’
c. (Eu acho que) O PEDRO, a Maria só foi embora depois
(I think that) the Pedro the Maria only went away after
que viu ele.
that saw him
‘I think that Pedro, Maria only went home after seeing him.’
The contrast between (13) and (11) shows that sentences in which a fo-
calized phrase is related to a position occupied by an overt resumptive pro-
noun inside an island are possible, whereas sentences in which the pronoun
is not in an island are not. It is possible, then, to assume that focalized con-
stituents may be generated in either position. In other words, Focalization
may or may not involve movement, just like Topicalization/Left Dislocation.
When the constituent is moved, an overt resumptive pronoun is impossible
in the (argument) position related to the focalized phrase; when it is base
generated in the left periphery position (in the case where insertion in the
argument position and movement to the dislocated position would violate
some island constraint, for instance), an overt or null pronoun occupies the
position to which the focalized phrase is related to. The sentences in (13)
indicate that this possibility is empirically correct. Theoretically, the assump-
tion is also not problematic since the two possible derivations are not compa-
rable in terms of economy: if there is no resumptive pronoun in the initial
array, the focalized phrase is inserted (merged) in its thematic position and
moves. If there is a resumptive pronoun in the initial array, the pronoun is
inserted in the -position and the focalized phrase in the dislocated position.
The only open question is why it should be the case that, in sentences that
do not involve islands, an initial array containing a resumptive pronoun is
excluded. Although we do not have an answer for that question, the general-
ization we can extract from (12-13) is that the focalized phrase will be gen-
erated in the dislocated position (binding a null or overt resumptive pronoun)
only if related to an argument position inside an island. Informally speaking,
the numeration without the resumptive pronoun is always preferred, if a
convergent derivation can be obtained.9 This resembles the behavior of wh-
-phrases in BP: wh-phrases are (not obligatorily, but usually) moved overtly;
however, if the wh-phrase is inside an island, it obligatorily stays in situ. In
9
I do not see how this generalization should be formalized since it clearly cannot be stated as
a condition on the well-formedness of numerations.
92 Marcello Modesto
this way, both focalized phrases and wh-phrases are preferably moved (for
unclear reasons). As for violation of islands effects, we now see that no
movement is involved when the focalized phrase is related to a position in-
side an island: it actually binds a (null or overt) resumptive pronoun in that
position.
We should expect, following the considerations above, that focalized
phrases reconstruct only when moved, and that seems to be what we get, i.e.
sentences “b”, “c” and “d” in (14) only have the interpretation where seu is
taken as a second person pronoun:
(14) a. O SEU DEPARTAMENTO todo professor acha o melhor.10

the his department every professor considers the best
‘His department, every professor considers the best.’
‘Your department, every professor considers the best.’
b. COM O SEU DEPARTAMENTO todo professor que se preocupa
about the his department every professor that refl worries
foi embora.
went away
‘*About his department, every professor who worries himself went away.’
‘About your department, every professor who worries himself went away.’
c. SOBRE O SEU DEPARTAMENTO que todo professor fale é natural.
about the his department that every professor talks is natural
‘*His department, that every professor talks about is natural.’
‘Your department, that every professor talks about is natural.’
d. O SEU DEPARTAMENTO todo professor foi embora antes que
the his department every professor went away before that
fosse reconhecido.
was recognized
‘*His department, every professor got out before it was recognized.’
‘Your department, every professor got out before it was recognized.’
Sentence (14a) is derived by movement since the position related to the

focalized phrase is not in an island. Sentences (14b-d), conversely, are not
derived by movement. As we expect, the former is ambiguous, allowing a
bound reading of the pronoun seu. In the latter sentences, however, seu is
10
Mary Kato (p.c.) suggested that the absence of the bound reading in the (b) example, for
instance, in contrast with (a) could be caused by the complexity of the former. It is easy to
show, however, that increasing the complexity of (14a) would not affect the possibility of
the bound reading:
(i) O SEU DEPARTAMENTO todo professor que chegou da Europa acha o
the his department every professor that arrived from Europe considers the
melhor.
best.
interpreted as second person, so we conclude that both topics and focused

phrases reconstruct only when they are derived by movement.
Summarizing, we have concluded that both focalized and topicalized
phrases may be generated in the dislocated position or moved, according to
the restrictions imposed on movement. It seems that movement is the default
operation in the case of Focalization, which is intriguing but, as argued be-
fore, does not cause problems related to economy, since derivations with and
without movement are not comparable as they present different initial arrays
(containing a resumptive pronoun or not).
2.3. Interaction between focalized and topicalized phrases
In (15a, b), we see that a focalized phrase may appear to the right, but not
to the left, of a topic, either in matrix or embedded contexts. In (15c, d), we
see that the same ordering applies even when the topic and the focused
phrase are in different clauses:
(15) a. (Ele disse que) proLuiz, ESSES CDS a Maria recomendou.

he said that to Luiz these CDd the Maria recommended
‘(He said that) to Luiz, theses CDs, Maria recommended.’
b. *(Ele disse que)PRO LUIZ esses CDs, a Maria recomendou.
he said that to Luiz these CDs the Maria recommended
‘(He said that) to Luiz, these CDs, Maria recommended.’
c. Pro Luiz, ele disse que ESSES CDS a Maria recomendou.
to Luiz he said that these CDs the Maria recommended.
‘To Luiz, he said that these CDs Maria recommended.’
d. *PRO LUIZ ele disse que esses CDs, a Maria recomendou.
to Luiz he said that these CDs the Maria recommended
‘To Luiz, he said that these CDs Maria recommended.’
In the spirit of Rizzi (1997), we will assume a topic projection TopP

dominating a focus projection FocP between CP and the Inflectional com-
plex, where topics and focused phrases are moved to or generated at in order
to check their T(opic) or F(ocus) features. Such assumption explains the
order of constituents in (15a) and the impossibility of (15b) but does not
explain why (15d) is also impossible since, in that case, the dislocated
phrases occupy different clauses. That fact might be due to prosodic con-
straints since we know that a pause must follow a topic but no pause may
follow a focused phrase, causing (15d) to have problems at PF. I will also
assume that the TopP projection allows for multiple Spec positions, as in
Kayne (1994), deriving the fact that in BP, and in Romance in general, there
is no (theoretical) limit to the number of dislocated topics (but see Frascarelli
2007 for a different view). The FocP projection, on the other hand, cannot
allow for multiple Spec positions since no more than one dislocated focal-
94 Marcello Modesto
ized phrase is observed. Rizzi (1997) tries to explain the “non-recursiveness”

(in his terms) of FocP as an interpretative clash. He takes the Spec FocP
position to be the focus, which could contain only new information, and the
complement of the FocP to be the presupposition associated with that focus,
which would contain only old information. In this way, no focalized phrase
could be placed inside the presupposition of another focus since new infor-
mation could not be contained inside the presupposition. However, Rizzi’s
explanation would disallow all in situ focused phrases co-occurring with a
dislocated focus and, if wh-phrases are considered as F-marked elements, as
Rizzi himself assumes, we should not expect to find sentences with multiple
wh-phrases. Besides, Zubizarreta (1998) convincingly argues that the no-
tions of focus and presupposition on one hand and new and old information
on the other should be kept apart, since their overlapping is only partial. The
ban against two focused phrases moving overtly applies even if they sit in
different clauses, seemingly having two different FocP projections as availa-
ble landing sites. This is illustrated by (16):11
(16) a. Roupas, pra mim, naquela loja, o Luiz nunca comprou.

clothes to me in.that shop the Luiz never bought
‘Clothes, to me, in that shop, Luiz never bought.’
b. *ESSES CDs PRO LUIZ a Maria recomendou.
these CDs to Luiz Maria recommended
‘These CDs, to Luiz, Maria recommended.’
c. *ESSES CDs eu acho que PRO LUIZ a Maria recomendou.
these CDs I think that to Luiz the Maria recommended
‘These CDs, I think that to Luiz Maria recommended.’
d. *ESSE FILME a Maria recomendou PRO LUIZ (não aquele vídeo
this movie the Maria recommended to Luiz (not that video
pro Paulo)
to Paulo)
‘This movie, Maria recommended to Luiz (not that video to Paulo).’
e. Quem recomendou o que pra quem?
who recommended what to whom
‘Who recommended what to whom?’
In (16), we see that, contrary to what happens to topics (16a), no more

than one focused phrase can be dislocated, even if the two focused phrases
11
I intentionally placed the direct object in the higher dislocated position in (16b, c, d) to
show that the impossibility of moving two focalized phrases is not related to superiority ef-
fects. This can also be seen by the fact that (16c), for instance, is bad with both orders:
(i) *PRO LUIZ eu acho que ESSES CDS a Maria recomendou.
to Luiz I think that these CDs Maria recommended
are in different clauses. There can be no more than one F-marked constituent
even if one of them stays in situ, unless in the case of multiple wh-phrases,
to which we will return. I will try to explain those facts by assuming, with
Zubizarreta, that a focused phrase cannot be deaccented. At PF, dislocated
phrases will receive ex-H pitch by virtue of occupying the Spec FocP posi-
tion, and what follows the dislocated Ex-H pitch phrase is deaccented (this
last part according to Zubizarreta). If two phrases occupy the Spec position
of (the same or different) FocP projection, one of these two phrases would
have to be both deaccented and Ex-H, causing the derivation to crash at PF.
Even if the F-marked phrase stays in situ, it cannot be deaccented, explain-
ing the impossibility of (16d). On the other hand, (16e) is possible because,
as we independently conclude below, moved wh-phrases do not sit at Spec
FocP but at Spec CP. In this way, moved wh-phrases are not marked Ex-H at
PF and the rest of the sentence need not be deaccented. It is then predicted
that a non-wh-phrase when focus-moved cannot co-occur with any F-marked
phrase, including a wh in situ, and the prediction is borne out, as seen in (21)
below.
It also seems problematic that both topics and focalized phrases can either
stop at the embedded CP complex or move up to the matrix clause even if
related to a position in the lower clause. We will assume that, in the former
case, TopP and FocP are simply not present in the structure of the matrix
clause. If they were, then topics and focused phrases should move to that
position, in order to check the features of the Top and the Foc heads. How-
ever, having checked their features in the lower clause would prevent those
phrases from moving any further and the derivation would not converge. In
the latter case, when topics and/or focus move to the matrix CP complex
from an embedded clause, it must be the case that there are no TopP and
FocP in the lower CP complex, for the reasons just described.
With these considerations out of the way, it is possible now to examine
the interaction between moved topics and focalized phrases. Consider (17):
(17) a. O seu melhor aluno, PRA FAZER PÓS (eu tenho impressão
the his best student to do graduate (I have impression
que) todo professor convidou.
that) every professor invited
‘His best student, to enter the graduate program, I have the impression
that every professor invited.’
b. O seu melhor aluno, eu tenho impressão que PRA FAZER PÓS
the his best student I have impression that to do graduate
todo professor convidou.
every professor invited
‘His best student, I have the impression that, to enter the graduate pro-
gram, every professor invited him.’
96 Marcello Modesto
In (17a, b), the focalized phrase was moved since it is not related to a po-
sition inside an island. Both sentences in (17) allow a bound reading of the
pronoun in the topic phrase, so the topic must have been derived by move-
ment if that interpretation is considered. We see, then, that Topicalization
can apply over a focused phrase derived by movement. However, if the fo-
cused phrase is base generated in the dislocated position, the bound reading
of the pronoun in the topic becomes unavailable:
(18) a. O seu melhor alunoi, (eu tenho impressão que) GEOGRAFIAj

the his best student I have impression that geography
todo professor que ensina _j convidou _i pra jantar.
every professor that teaches invited to dinner
‘Your best student, I have the impression that, Geography, every
professor who teaches it invited him to dinner.’
b. O seu melhor aluno, todo professor que ensina geografia
the his best student every professor that teaches geography
convidou _ pra jantar.
Invited to dinner
‘His best student, every professor who teaches Geography invited him
to dinner.’
c. O seu melhor alunoi, (eu tenho impressão que) EMPREGOj todo
the his best student I have impression that job every
professor apresentou _i a um empresário que ofereceu _j
professor introduced to one entrepreneur who offered
‘Your best student, I have the impression that, a job, every professor
introduced him to an entrepreneur who offered it.’
d. O seu melhor alunoi, todo professor apresentou _i a um
the his best student every professor introduced to one
empresário que ofereceu emprego.
entrepreneur that offered job
‘His best student, every professor introduced him to an entrepreneur
who offered him a job.’
In (18a, c), the focused phrase must be generated in the dislocated posi-
tion since it is related to a position inside a strong island. The topic, on the
other hand, could in principle be moved or base generated since it is not
related to a position inside an island. Those sentences, however, do not allow
for the bound reading of the pronoun inside the topic, so the topic must have
been left dislocated (generated at the dislocated position). Sentences (18b,
d), without the intervention of the focused phrase, do allow for the bound
reading. We conclude, then, that a topic can be topicalized over a moved
focused phrase but not over a base generated focused phrase. This conclu-
sion contrasts with the interaction between two topics:
(19) a. O seu melhor aluno, pra fazer pós, (eu tenho impressão
the his best student to do graduate I have impression
que) todo professor convidou.
that every professor Invited
‘His best student, to enter the graduate program, I have the impression
that every professor invited him.’
b. O seu melhor aluno, eu tenho impressão que pra fazer
the his best student I have impression that to do
pós, todo professor convidou.
graduate every professor invited
‘Your best student, I have the impression that, to enter the graduate
program, every professor invited him.’
c. Pra fazer pós, o seu melhor aluno, todo professor
to enter the grad-program his best student every professor
convidou.
invited
‘To enter the graduate program, his best student, every professor
invited.’
d. Pra fazer pós, eu tenho impressão que o seu
to enter the grad-program I have the impression that his
melhor aluno, todo professor convidou.
best student every professor invited
‘To enter the graduate program, I have the impression that, his best
student, every professor invited him.’
In (19), the topic containing the pronoun seu might have been moved or
generated in the dislocated position, since we are assuming that those struc-
tures are usually ambiguous between Topicalization and Left Dislocation.
The second topic, however, must have been moved, since a pronominal ele-
ment cannot refer to the VP fazer pós inside the PP.12 The sentence (19b),
however, does not allow a bound reading of the possessive pronoun in the
higher topic, whereas the ones in (19a,c,d) do. The generalization then seems
to be that, in contrast with the facts in (18), a topic can be moved over an-
other moved topic only if they occupy Spec positions of the same projection.
The same pattern can be seen with respect to a base generated topic. Topical-
ization over a base generated topic is not possible, so sentences (20b, d) do
12
Left dislocation of a PP with a VP complement related to a position inside an island will
be, then, also impossible and those sentences will never be possible since neither the Topi-
calization or Left Dislocation strategies are possible:
(i) *Pra fazer pós, o professor que convidou o seu melhor
to enter the grad-program, the professor that invited the his best
aluno t/pro se arrependeu.
student self regretted
‘To enter the graduate program, the professor who invited his best student regretted doing it.’
98 Marcello Modesto
not allow for the bound reading. Sentences (20a, c), however, allow for the
bound interpretation since both topics are Specs of the same Top projection:
(20) a. O seu melhor aluno, geografia, todo professor que ensina

the his best student, geography, every professor that teaches
convidou pra jantar.
invited to dinner
‘His best student, geography, every professor who teaches it invited him
to dinner.’
b. O seu melhor aluno, eu tenho impressão que geografia, todo
the his best student, I have the impression that geography, every
professor que ensina convidou pra jantar.
professor that teaches invited to dinner
‘Your best student, I have the impression that, geography, every
c. O seu melhor aluno, emprego, todo professor apresentou a um
the his best student job every professor introduced to one
empresário que ofereceu.
entrepreneur that offered
‘His best student, a job, every professor introduced him to an
entrepreneur who offered.’
d. O seu melhor aluno, eu tenho impressão que emprego, todo
his best student I have impression that a job every
professor apresentou a um empresário que ofereceu.
Professor introduced to one entrepreneur that offered
‘Your best student, I have the impression that, a job, every professor
introduced him to an entrepreneur who offered.’
The fact that reconstruction is impossible in (19b) and (20b, d) resembles

the Arabic cases treated by Aoun and Benmamoun (1996) as Minimality
effects. However, it seems that the relevant notion here is not one of gov-
ernment as in Rizzi’s (1990) Relativized Minimality, since all moved topics
in those cases are complements, therefore being head governed by some
[+V] head. The notion of “shortest movement” of Chomsky (1993, 1995)
seems more adequate. Something like the Minimal Link Condition (MLC)
and the notion of equidistance13 of Chomsky (1993) could handle (19b) and
(20b, d): in those cases, if the topics are moved from the lower clause, they
have to reach the closest TopP. If there is a TopP projection in the embedded
clause, both topics move to the Spec positions of that projection. Further
movement, as in (19b) and (20b,d) will be barred because the topics have
already checked their features and movement is last resort. If no TopP is
13
If ,  are in the same minimal domain, they are equidistant from . (Chomsky 1993: 17).
projected in the embedded clause, then both topics move to the higher
clause, as in (20c). In both cases, the two topics are equidistant from their
argument position, deriving the fact that no intervention will ensue if the
topics are Specs of the same head. The only way to check features of both
topic heads if two TopP are present in the structure is resorting to base gen-
eration of one of the topics in the Spec position of the higher TopP, which is
the strategy used in (19b) and (20b, d) explaining the lack of reconstruction.
This approach runs into problems, however, in trying to explain why recon-
struction is also impossible in (18a, c), where a base generated focused
phrase intervenes between the topic and its original position. In those cases,
there is, presumably, no feature that a topic could check in the domain of the
head of the Focus projection, so movement of the topic does not “skip” any
position where its features could have been checked. In this sense, Relativ-
ized Minimality is descriptively more adequate since any A’-specifier will
intervene between the topic and its trace (unless the focused phrase is de-
rived by movement, for reasons that will become clear as we proceed). We
will keep to the descriptive generalization, assuming that those cases involve
some kind of Minimality effects. In this way, a topic cannot be moved over a
base generated focused phrase or another (moved or base generated) topic
(unless the two topics are equidistant from the trace positions) because a
specifier position of a “similar kind” intervenes between the moved phrase
and its trace position.
2.4. Interaction with wh-phrases
Still following Rizzi (1997) and Zubizarreta (1998), we assume that a
wh-phrase always bears the feature F and is moved to Spec FocP. Thus, we
predict not only that a wh-phrase cannot co-occur with a dislocated focalized
phrase in the same CP complex in any order, since they compete for the
same position, but also that a moved wh-phrase cannot co-occur with a
fronted focused phrase at all (even if in different clauses), since prosodic
constraints do not allow a Foc projection inside the presupposition of another
FocP. In the same way, in situ focused phrases and wh-phrases cannot co-
-occur with a dislocated focused phrase because F-marked constituents can-
not be deaccented:
(21) a. *O que PRO JOÃO você deu?

what to João you gave
b. *PRO JOÃO o que você deu?
to João what you gave
c. *PRO JOÃO você deu o quê?
to João you gave what
d *O que você acha que PRO JOÃO eu dei?
what you think that to João I gave
100 Marcello Modesto
e. *PRO JOÃO você acha que o que eu dei?

to João you think that what I gave
f. *PRO JOÃO você acha que eu dei o quê?
to João you think that I gave what
We should also expect, following our discussion, that a wh-phrase could

not be fronted across a topic in matrix contexts since TopP dominates FocP,
to where we are assuming wh-phrases are moved (see (22)). Unexpectedly,
however, we see, in (23), that wh-phrases cannot be moved over a (moved or
base generated) topic even if the topic sits in the embedded CP:
(22) a. *Quem o João, viu (ele) ontem?

who the João saw (him) yesterday
b. *?O que pro João, você deu?
what to.the João you gave
c. *?O que o seu melhor aluno, todo professor convidou pra fazer?
what his best student each professor invited to do
(23) a. *Quem você acha que o João, viu (ele) ontem?

who1 you think that João t1 saw (him) yesterday
b. *O que o Pedro acha que pro João, você deu (pra ele)?
what Pedro think that to João you gave (to him)
c. *O que você acha que o seu melhor aluno, todo professor
what you think that the his best student every professor
convidou (ele) pra fazer?
invited (him) to do
(22-23) show that a wh-phrase is never moved over a topic. Nevertheless,

a topic can precede a wh-phrase, which is expected if wh-phrases sit in Spec
FocP:
(24) a. O João, quem viu (ele) ontem?

the João who saw (him) yesterday
‘As for João, who did see him yesterday?’
b. O João, o que você deu pra ele?
the João what you gave to him
‘To João, what did you give to him?’
c. O seu melhor aluno, o que todo professor convidou pra fazer?
the his best student what every professor invited to do
‘Your best student, what did every professor invited to do?’
Notice that we are following Rizzi in assuming exactly the same deriva-
tion for sentences (24c) and (25). However, there is an asymmetry between
these two sentences left unexplained: as mentioned before, (25) allows re-
construction of the topic to the argument position; (24c), on the other hand,
does not allow that reading (the topic must be left dislocated):
(25) O seu melhor aluno, PRA FAZER PÓS (eu tenho

his best student to enterthe grad program (I have
impressão que) todo professor convidou.
the impression that) every professor invited
‘His best student, to enter the graduate program, I have the impression that
every professor invited.’
If both focused phrases and wh-phrases are moved to Spec Foc, as Rizzi
assumes, this asymmetry cannot be explained. Nevertheless, Rizzi’s argu-
ments, involving adjacency effects and possible word orders in Italian, are
very convincing to the effect that wh-phrases are moved to Spec FocP. The
impossibility of having a wh-phrase and a focused phrase both fronted in the
same sentence in BP, in any order, seems to argue that, in this language, it
also should be the case that wh-phrases and focused phrases compete for the
same position. However, BP differs from other Romance languages in an
important aspect: a wh-phrase can always be followed by the complementiz-
er que, even when there is a preceding topic. Consider the sentences below:
(26) a. O João, o que que você deu pra ele?

the João what that you gave to him
‘To João, what did you give to him?’
b. Da Maria, quem que gosta dela?
of.the Maria who that likes (her)
‘As for Maria, who likes her?’
c. O seu melhor aluno, o que que todo professor convidou
the his best student what that every professor invited
pra fazer?
to do
‘Your best student, what did every professor invited him to do?’
The possibility of having a complementizer following the wh-phrase indi-

cates that the latter is not in Spec FocP but in Spec CP. This is in fact what we
expect, considering that wh-phrases have two features that must be checked
during the derivation: a F(ocus) feature and a wh-feature. In languages like
Italian, it is plausible to assume, as Rizzi does, that the wh-feature is generated
under T in main questions, and I to C movement applies to bring that feature
to the C system, where the wh-phrase can check it (or, in Rizzi’s terms, the
wh-Criterion can be satisfied). This assumption explains the asymmetry be-
tween wh-phrases and focused phrases in Italian seen in (27):
(27) a. QUESTO Gianni ti dirá.

this Gianni to.you will.say
b. *Che cosa Gianni ti dirá?
what Gianni to.you will.say?
Sentence (27b) is ungrammatical since, besides its F-feature, the wh-

-phrase in Spec Foc must also check its wh-feature but I to C has not ap-
plied. In BP, where there is no I to C movement, we can assume that wh-
-features are inherently possessed by C0. In BP, then, a wh-phrase moves to
Spec Foc to check its F-feature and then moves further to Spec CP to check
its wh-feature, explaining why it can always be followed by an overt com-
plementizer.14 This conclusion has two important corollaries: the sentences
in (23), where a wh-phrase precedes a topic are ungrammatical not because
TopPs are higher than FocPs but because wh-phrases cannot be moved over
(moved or base generated) topics in any context; secondly, there is a second
topic position higher than CP for left dislocated topics.15
The fact that wh-phrases are never moved over a topic explains the
asymmetry between (25), where reconstruction of the topic is possible, and
(24c) and (26c) where the bound reading of the pronoun inside the topic is
never possible: in the former, the topic might have been topicalized (moved)
to Spec TopP, from where it can reconstruct, but in the latter two sentences,
the topic can only be at the left dislocated position above CP, since the pres-
ence of the wh-phrase in Spec CP prevents any topic from being moved to or
generated at Spec TopP (since any topic, moved or base generated, induces
what we have been calling a Minimality effect between the wh-phrase and
its base position). That a left dislocated topic can occupy a different position
than a topicalized topic is corroborated by the fact that, in BP, some topics
are not related to any position inside the sentence:
(28) a. Peixe, eu gosto muito de namorado.

fish, I like lots of.the red.snapper
‘As for fish, I like red snapper a lot.’
b. A Maria, o Pedro não quer nem saber de casamento.
the Maria, the Pedro not want even know of wedding
‘As for Maria, Pedro doesn’t even want to hear about a wedding.’
14
We also assume that the head of CP is optionally null when a wh-phrase occupies its Spec.
15
That only base generated topics can appear in that position is also clear since movement is
only driven by feature checking considerations (as it is commonly assumed). In this way,
no topic could be moved to this adjoined position because there is no relevant (topic) fea-
ture that can be checked there.
In such cases, it would be problematic to assume that there is a TopP

where the left dislocated topic is merged in, since no element in the sentence
bears the feature Topic. Some left dislocated phrases, then, must be adjoined
to the sentence, explaining why they can appear in a position higher than a
wh-phrase in Spec CP. This is corroborated by the fact that the kind of topic
in (28) cannot appear in embedded contexts:
(29) a. *Eu acho que peixe, a Maria gosta muito de namorado.

I think that fish, the Maria likes lot of red snapper
‘I think that, as for fish, Maria likes red snapper a lot.’
b. *Eu acho que a Maria, o Pedro não quer nem saber
I think that the Maria, the Pedro not want even know
de casamento.
of wedding
‘I think that about Maria, Pedro does not even want to hear of marriage.’
It also becomes clear now why multiple wh questions are possible, con-
trasting with the sentences in (21). In (30), for instance, the wh-phrase is in
Spec CP position and so does not receive Ex-H at PF. The sentence need not
be deaccented in this case and another wh in situ is possible:16
(30) Quem recomendou o que pra quem?

who recommended what to whom
3. Minimality
Let us recapitulate what we have concluded so far: a) movement of a topic is
possible over a moved but not over a base generated focused phrase; b)
movement of a topic is impossible over another topic (either moved or base
generated) and movement of wh-phrases is impossible over topics. These
conclusions are summarized below (where elements related to traces indicate
movement and elements related to pro indicate base generation):
(31) a. OKtopic1 ………. focus2.. ………..t2………….…………..t1…

b. *topic1 ………. focus2……..[island....pro2…]……….…t1…
16
Following this logic, we should expect that something like (i) would be possible. However,
although (i) would have no problem at PF, it is semantically awkward. For one thing, the
sentence asserts that João is the correct value for the variable in “Pedro recommended
something to x” but at the same time it questions what is the x such that Pedro recommend-
ed x to João. The sentence is then impossible since one cannot assert that João is the cor-
rect value for a certain variable if one does not know which variable that is:
(i) O que o Pedro recomendou PRO JOÃO?
what the Pedro recommended to João
c. *topic1 ………. topic2……….. t2/pro2………………….t1…

d. *wh1…………..topic2………..t2/pro2…………………..t1…17
As we said before, the facts in (31) seem to be related to Minimality ef-

fects. Two choices immediately present themselves: either Minimality is a
constraint applying to LF representations or a constraint on derivations (see
Chomsky 1995, Aoun and Benmamoun 1996). Taking Minimality as a con-
straint on LF representations could explain the fact that a topic can be moved
over a focused phrase since, at LF, both phrases are reconstructed to their
argument position and no intervention takes place. However, it would fail to
explain why a topic could not be moved over another topic, since reconstruc-
tion could apply here as well. On the other hand, if we take Minimality as a
constraint on derivations, the intervention of a topic by another topic is ex-
plained but the non-intervention of a topic by a focused phrase is left out.
We then reach the paradox that Minimality cannot be taken either as a con-
straint on LF representations or on derivations. In the next section, we argue
that this paradox can be solved if Focalization is taken to involve movement
in the PF component, and not in the syntax.
4. Focalization as PF-movement
There is an asymmetry between reconstruction of topics and focused phrases
that has not yet been explained: reconstruction must be total just in the case
of Focalization, but not Topicalization. We see that reconstruction is total in
(32a) since the pronoun in the relative clause of the focused phrase can be
bound by the quantified sentential subject but, in (32b), if there is recon-
struction, it must be partial, because the pronoun in the topic phrase cannot
be bound by the subject, which is not predicted by our analysis so far, since
the topic apparently could be moved in that case:
(32) a. O ALUNO QUE ELE ORIENTOU, cada professor indicou

the student that he advised each professor recommended
pra um emprego.
for a job
‘The student that he advised, each professor recommended for a job.’
b. O aluno que ele orientou, cada professor indicou
the student that he advised each professor recommended
pra um emprego.
for a job
‘The student that he advised, each professor recommended for a job.’
17
This structure will always derive an ungrammatical sentence (irrespectively of its reading),
since the wh-phrase does not have the option of being generated already in CP and move-
ment would cause a Minimality violation.
The same pattern can be seen in (33) concerning binding facts discussed
by Lebeaux (1989). In (33a, c), coreference between the pronoun and the R-
-expression inside the focalized phrase is impossible, while, in (33b, d), co-
reference is possible:18
(33) a. A PROPOSTA QUE O JOÃOi FEZ ele*i/j aprovou _ na

the proposal that the João made he approved in.the
mesma reunião.
same meeting
b. A proposta que o Joãoi fez, elei/j aprovou na mesma
the proposal that the João made he approved in.the same
reunião.
meeting.
‘The proposal that João made, he approved in the same meeting.’
c. A AFIRMAÇÃO DE QUE O JOÃOi GOSTA DA MARIA
the affirmation of that the João likes of.the Maria
ele*i/j negou.
he denied
d. A afirmação de que o Joãoi gosta da Maria, elei/
the affirmation of that the João likes of.the Maria he
negou
denied
‘The claim that João likes Maria, he denied.’
Lebeaux treats similar “anti-reconstruction” effects in English by assum-

ing that principle C of the Binding Theory applies throughout the derivation
and by ordering two different operations: adjoin- and move-. In this way,
non-subcategorized elements (adjuncts) would be generated as separated
phrase markers and then adjoined after move- has applied. This mechanism
would explain the asymmetry between arguments and adjuncts shown in
(34):
(34) a. *Hei denied the claim that Johni made.

b. *Hei denied the claim that Johni likes Mary.
c. Which claim that Johni made did hei latter deny t?
d. *Whose claim that Johni likes Mary did hei deny t?
In (33), however, arguments and adjuncts pattern together: Focalization

shows anti-reconstruction effects, and Topicalization does not, when the
focused/topic phrase includes both subcategorized and non-subcategorized
18
The most embedded element inside the focused phrase receives main prominence in (33a,
c): the verb fez and Maria respectively. However, varying the main prominence inside the
focused phrase does not seem to alter the possibility of coreference.
relatives. Let us first look at Topicalization. Sentence (33d) should not be

possible in the coreferential reading, just like (34d). It seems to be the case,
as it is natural, that generation of topics in the dislocated position does not
happen only to avoid island effects, but to avoid any violation of a constraint
or principle that would ensue with a movement operation. In other words, as
all topic-comment structures are ambiguous between a Left Dislocation and
a Topicalization structure in BP, if one of those structures would violate a
principle or constraint, the other alternative is preferred. In this way, the
coreferential reading in (33d) is grammatical because the topic is interpreted
as left dislocated: as the topic was generated already in the dislocated posi-
tion, there is no violation of principle C. Topicalization (base generation in
the argument position and later movement), on the other hand, would violate
principle C. Sentence (33b), however, is still ambiguous between Topicaliza-
tion and Left Dislocation, but shows only partial reconstruction since the
relative clause, being non-subcategorized, is adjoined to the topic only after
movement if the topic is topicalized. Turning now to Focalization, the prob-
lem is reversed: (33a) should be possible with the coreferential reading and
(32a) should not show radical reconstruction. It seems, then, that both (32a)
and (33a) show that the focused phrase is moved only after the adjunct is
inserted in the derivation.
This fact can be explained if the operation move- in Focalization is not
a syntactic operation, i.e. it applies in the PF component. As a post-Spell Out
movement, Focalization is expected not to interfere with LF operations: fo-
cused phrases show total reconstruction even if an adjunct is contained in it,
as in (32a), and Focalization does not alter binding relations, as in (33a).
Going back to the facts summarized in (31) above, we saw that a moved
focused phrase does not prevent a moved topic from reconstructing (see
(25), repeated below), while a base generated focused phrase (35a), a
(moved or base generated) topic (35b) or a (moved) wh-phrase (24c) does:
(25) O seu melhor aluno, PRA FAZER PÓS (eu tenho

his best student to enter the grad program (I have the
impressão que) todo professor convidou.
impression that) every professor invited
‘His best student, to enter the graduate program, I have the impression that
every professor invited.’
(35) a. O seu melhor aluno, (eu tenho impressão que) GEOGRAFIA

the his best student (I have impression that) geographyj
todo professor que ensina _j convidou _i pra jantar.
every professor that teaches _j invited _i to dinner
‘Your best student, (I have the impression that) Geography, every
b. O seu melhor aluno, pra fazer pós, todo

the his best student to enter the grad-program every
professor convidou.
professor invited
‘Your best student, to enter the graduate program, every professor invited.’
(24) c. O seu melhor aluno, o que todo professor convidou pra fazer?
the his best student what every professor invited to do
‘Your best student, what did every professor invited to do?’
If we take Focalization as an operation in the PF component, those data

can be handled in a simple way: (moved) focused phrases differ from topics
and wh-phrases in that both latter phrases are crossed by a topic on its way to
Spec TopP; focused phrases, on the other hand, will not be crossed since
they are moved only at the PF component. Base generated focused phrases,
however, will be crossed by the topic, giving rise to Minimality effects. In
the same way, if a topic is moved to (or base generated at) Spec TopP,
movement of a wh-phrase from FocP to CP will cross the topic. The only
possibility of having a topic co-occurring with a wh-phrase, then, is if the
topic is base generated at the left dislocated position adjoined to CP, which
is a different position available for topics in BP, as seen above.
If focalized phrases are moved in the PF component, there is no argument
left against treating Minimality as a constraint on derivations: a (moved or
base generated) topic intercepts another topic on its way to Spec TopP, but a
(moved) focused phrase does not since it will only be moved at PF. The
same facts could also be handled if Minimality is taken to be a condition on
the application of the operation Move-, as argued by Aoun and Benma-
moun, since, in Arabic, a PF-moved constituent intercepts another PF-moved
constituent. This cannot be seen in BP due to the independent fact that there
can be only one dislocated focused phrase. Nevertheless, as the second
option handles both the data in BP and Arabic it should be preferred.
5. Conclusion
We have shown that topics in BP can be moved, characterizing the construc-
tion we have been calling Topicalization, or generated already in a dislocat-
ed position, characterizing Left Dislocation. As expected, only the former
will show reconstruction effects. A left dislocated topic which is related to a
(argument) position occupied by a null or overt resumptive pronoun in the
argument position can be generated as the Spec of a TopP in the C system or
in a position adjoined to CP. A topic that is not related to any position in the
sentence can only be adjoined to CP.
In the same way, a focalized phrase can be moved to the Spec of a FocP
or be generated in that position, being related to a null or overt resumptive
pronoun in the sentence. When moved, the focused phrase is moved in the
PF component and so will not affect any LF operation: it shows total or radi-
cal reconstruction and it does not alter binding relations. When it is generat-
ed already in the dislocated position, it must be related to a position inside an
(strong) island.
Wh-phrases in BP are moved to Spec FocP and then to Spec CP in order
to check its focus and wh-features, which are inherently possessed by C0.
The various intervention facts holding between topics and focused
phrases in BP are explained as Minimality, although it is unclear which
common feature among topics and focused phrases would be responsible for
such effects. We also see that Minimality should be taken as a condition on
derivations (or on the operation move-) and that PF movement is con-
strained in the same ways as syntactic movement is: it is driven by feature
checking considerations, observes islands, etc.
References
Aoun, Joseph. & Elabbas Benmamoun (1996). Minimality, Reconstruction, and PF-
-Movement. Ms., USC, Los Angeles.
Aoun, Joseph, Lina Choueiri & Norbert Hornstein (2001). Resumption, Movement
and Derivational Economy. Linguistic Inquiry, 32, 371-403.
Benincà, Paola & Cecilia Poletto (2004). Topic, Focus and V2: Defining the CP
sublayers. In Rizzi, L. (org). The structure of CP and IP. New York: Oxford
University Press, pp. 52-75.
Cecchetto, Carlo and Gennaro Chierchia (1999). Reconstruction in Dislocation Con-
structions and the Syntax/Semantics Interface. In S. Blake, E. Kim and K.
Shahin (orgs.), Proceedings of the 17th West Coast Conference on Formal
Linguistics. Stanford: CSLI publications, pp. 132-146.
Chomsky, Noam (1981). Lectures on Government and Binding. Dordrecht: Foris.
Chomsky, Noam (1993). A Minimalist Program for Linguistic Theory. In K. Hale &
S. J. Keyser, (eds.), The View from the Building 20. Cambridge, Mass: MIT
Press, pp. 1-52.
Chomsky, Noam (1995). The Minimalist Program. Cambridge, Mass: MIT Press.
Cinque, Guglielmo. 1990. Types of A’ Dependencies. Cambridge, Mass: MIT Press.
Frascarelli, Mara (2007). Subjects, Topics and the interpretation of referential pro.
An interface approach to the linking of (null) pronouns. Natural Language
and Linguistic Theory, 25, pp. 691-734.
Kayne, Richard (1994). The Antisymmetry of Syntax. MIT Press, Cambridge, Mass.
Lasnik, Howard & Mamoru Saito (1992). Move-: Conditions on its Application
and Output. Current Studies in Linguistics, 22, Cambridge, Mass: MIT Press.
Lebeaux, David (1989). Relative clauses, licencing, and the nature of the derivation.
NELS 20.
Negrão, Esmeralda & Ana Lúcia Müller (1996). As mudanças no sistema pronomi-
nal do Português Brasileiro: substituição ou especialização de formas?,
D.E.L.T.A., 12 (1), pp. 125-152.
Rizzi, Luigi (1986). On the status of subject clitics in Romance. In Osvaldo Jaeggli
and Carmen Silva-Corvalán, (eds.), Studies in Romance linguistics. Dor-
drecht: Foris, pp. 391-419.
Rizzi, Luigi (1990). Relativized Minimality. Cambridge, Mass: MIT Press.
Rizzi, Luigi (1997). The Fine Structure of the Left Periphery. In: L. Haegeman,
(org.) Elements of Grammar, A Handbook in Generative Syntax. Dordrecht:
Roberts, Ian (2010). Agreement and Head Movement: Clitics, Incorporation, and
Defective Goals. Cambridge, Mass: MIT Press.
Zubizarreta, Maria Luisa (1998). Prosody, focus and word order. Cambridge, Mass:
MIT Press.
O POSSESSIVO DE 2ª PL NO DIALETO MINEIRO:
DP E CP EM ANÁLISE
BRUNA KARLA PEREIRA

(Universidade Federal dos Vales do Jequitinhonha e Mucuri)
ABSTRACT: This paper aims to analyze structures, such as “um aluno seus” (a-SG
student-SG your-PL = ‘your student’), that occur in dialectal Brazilian Portuguese. In
these structures, the determiner and the noun are inflected for singular, whereas the
possessive is marked with the plural morpheme ‘-s’. In this case, the possessive with ‘-
s’ refers specifically to 2nd person plural and is often postnominal. Considering these
properties, this paper: firstly, reviews the literature about possessives showing that
this issue has not been addressed in previous studies; secondly, presents several data
that have been collected from spontaneous utterances in the dialect spoken in Minas
Gerais; thirdly, investigates what determines this “split agreement”; and, finally,
explains how the postnominal position is derived.
KEYWORDS: 2nd Person Possessives, Number Agreement, DP, CP, Dialectal Brazili-
an Portuguese
1. Introdução
No português padrão (1a) e em outras línguas românicas, como o espanhol
(1b), pronomes possessivos concordam com o nome em número. No entanto,
têm sido frequentes, em certas variantes do português do Brasil, como no dia-
leto mineiro, ocorrências em que o pronome possessivo de segunda pessoa,
seja pós-nominal (2a) seja pós-cópula (2b), não estabelece concordância em
número com o nome. Desse modo, observa-se uma cisão na concordância em
número, porque, enquanto o determinante e o nome carregam traços de singu-
lar, o possessivo recebe marca morfológica ‘-s’ de plural.
(1) a. as casas suas b. las casas suyas
(2) a. a alegria suas (a alegria de vocês)1 b. a caneta é suas (a caneta é de vocês)2
1
“Fico imaginando a alegria suas quando viram o trabalho pronto” (De professora universitária
a membros de um projeto, por ocasião de conversa informal, em Lavras, setembro de 2012).
2
“A caneta é suas” (De cliente de restaurante a garçonete por ocasião de atendimento comer-
cial, em Belo Horizonte, março de 2014).
112 Bruna Karla Pereira
Embora o possessivo no português brasileiro seja tema de vários estudos

(Kato, 1985; Perini, 1985; Silva, 1996; Cerqueira, 1996; Müller, 1997), além
de ser tratado em gramáticas tradicionais (Cunha; Cintra, 2001; Bechara,
2003) e outras (Neves, 1993; 2000; Castilho, 2010), não se realizou ainda
uma pesquisa que tomasse como objeto as construções mencionadas em (2).
Nesse sentido, com este trabalho, objetiva-se investigar ocorrências como
aquelas tipificadas em (2) nas quais o possessivo pós-nominal se flexiona no
plural enquanto o nome está no singular. Assim, do ponto de vista empírico,
serão considerados dados efetivos coletados no cotidiano e em bancos de
dados disponíveis na internet, além de dados de intuição. Do ponto de vista
teórico, teremos como base o programa minimalista (Chomsky, 1995) a par-
tir do qual será averiguada a aplicação de propostas formalistas de análise da
estrutura do DP, tais como Cinque (2005) e Bernstein (2005).
2. Revisão da literatura
Esta seção apresenta inicialmente pesquisas já desenvolvidas sobre o posses-

sivo ‘seu/sua’ e posteriormente pressupostos teóricos sobre a estrutura do
DP. Sendo assim, a primeira seção (2.1) tem como objetivo tratar, de modo
panorâmico, de pesquisas prévias sobre possessivos a fim de mostrar que
seus interesses diferem daquele que será foco do presente trabalho: a cisão
na concordância em número, em DPs com possessivo posposto de 2 a pessoa.
A segunda seção (2.2), por sua vez, destaca os fundamentos teóricos para a
análise deste fenômeno dentro de uma abordagem gerativista.
2.1. Pesquisas sobre possessivos no português

O possessivo ‘seu’ tem sido tema de diversos trabalhos no português do
Brasil, tais como Perini (1985), Kato (1985), Müller (1997), Cerqueira
(1996) e Rocha (2009). Esses pesquisadores têm se preocupado, dentre
outros aspectos, com: (i) o tipo de referente retomado pelo possessivo, se
genérico ou referencial; (ii) a substituição do pronome ‘seu’ de 3ª pessoa
pelo genitivo ‘dele/dela’, que teria desencadeado uma suposta redução no
uso daquele em favor deste; (iii) a ambiguidade no uso de ‘seu’ como se-
gunda e terceira pessoas; e (iv) a relação entre posição e função sintática
do possessivo.
Por exemplo, Silva (1996: 171) observa que, após introdução do ‘você’
no sistema pronominal, o possessivo ‘seu’, antes utilizado apenas para ter-
ceira pessoa, passou a ser utilizado também para segunda pessoa, conforme
quadro abaixo, uma adaptação de Silva (1996: 171).
O possessivo de 2ª PL no dialeto mineiro 113
Antes da introdução de ‘você’ Após a introdução de ‘você’

pessoal possessivo pessoal possessivo
eu meu eu meu
tu teu tu teu
ele seu você seu
ele seu
Em consequência, ‘seu’ torna-se ambíguo em (3a), podendo se referir

tanto a Joana, pessoa com quem se fala (2ª pessoa), quanto a Stella, pessoa a
respeito de quem se fala (3ª pessoa). Nesse caso, algumas estratégias são
utilizadas para evitar ambiguidade, dentre elas, a utilização de ‘dela’, em
(3b), e a utilização de uma oração relativa restritiva, em (3c). A utilização de
‘teu’, em (3d), apesar de ser uma alternativa pouco provável no dialeto mi-
neiro, é possível em outros dialetos do português.
(3) a. “Joanai, vi Stellaj beijando seui/j namorado” (Silva, 1996: 172).

b. Joana, vi Stellaj beijando o namorado delaj.
c. Joanai, vi Stella beijando o namorado que é seui.
d. Joanai, vi Stella beijando o teui namorado.
Diante dessa ambiguidade e da frequente utilização de ‘dele’ em ambientes

nos quais ‘seu’ de terceira pessoa seria possível, pesquisadores como Silva
(1996) e Rocha (2009) começaram a investigar se ‘dele’ estaria substituindo
‘seu’, em um processo de mudança. Essa hipótese foi contrariada por Neves
(1993) e questionada por Müller (1997), conforme será discutido adiante.
A partir da análise de dados do projeto NURC, coletados em diversas ci-
dades do país (incluindo Salvador, São Paulo, Rio de Janeiro, Porto Alegre e
Recife), Neves (1993: 207) afirma que,
para a 3ª pessoa, é altamente preferida a forma ‘seu’ (68,3%) à forma ‘dele’
(31,7%); essa conclusão, que contrariou as expectativas (especialmente em se
tratando de língua falada), foi confirmada pela análise de um corpus de con-
fronto (6 EF de São Paulo), que apresentou 75,6% de uso do ‘seu’ contra
24,4% de uso de ‘dele’.
Diferentemente, a partir de um corpus da cidade do Rio de Janeiro, Silva
(1996: 180) afirma estar “praticamente enterrada na língua oral a forma ‘seu’
para terceira pessoa, exceto nos casos de possuidores gerais, reservando-se a
forma ‘seu’ para segunda pessoa semântica”. Segundo Silva (1996: 175),
‘dele’ teve uma frequência de 75% no corpus oral e 14,1% no corpus escrito.
Também Rocha (2009: 103) afirma que a “forma possessiva ‘dele’ é a prefe-
rida entre os falantes do português de Belo Horizonte, com 78% dos casos”.
Müller (1997), em observância à disparidade nas conclusões apresenta-
das, defende a existência de uma especialização no uso de ‘seu’. Não se tra-
ta, portanto, de uma substituição de ‘seu’ por ‘dele’, mas de uma especiali-
zação no uso de ‘seu’ para sintagmas mais genéricos e de ‘dele’ para mais
referenciais. Por exemplo, em (4), o sintagma “todos os homens” inviabiliza
a retomada por ‘deles’ (4b), havendo preferencialmente a retomada por
‘seus’ (4a).
(4) a. Todos os homens têm os seus problemas.

b. ?Todos os homens têm os problemas deles.
Assim sendo, “essas formas estariam se especializando segundo o eixo

semântico da referencialidade” (Müller, 1997: 58-59), isto é, haveria uma
“relação entre o tipo de denotação do sintagma nominal antecedente e a es-
colha de uma ou de outra forma possessiva” (Müller, 1997: 62). Diante desse
dado, a autora sugere que a predominância no uso de ‘dele’ no corpus de
Silva (1996), por exemplo, ocorreria porque “os assuntos dessas entrevistas
são bastante presos aos contextos imediatos em que vivem os entrevistados,
evitando a discussão de assuntos mais abstratos” (Müller, 1997: 72-73).
Conclui a autora então que não são fatores extralinguísticos tais como condi-
ção social, região, grau de formalidade ou modalidade (fala versus escrita)
que determinam a escolha da forma pronominal ‘seu’ ou genitiva ‘dele’.
Tendo explicitado essa questão, a autora parte para um estudo da posição
dos possessivos no português. Müller (1997: 152) explica que, anteposto, o
pronome possessivo funciona como um determinante. Por sua vez, posposto
ao núcleo, ele funciona como um adjunto. Alguns testes sintáticos são eleitos
para justificar essa diferença: (i) apenas o pronome posposto aceita modifi-
cação por um advérbio (5) e (ii) apenas o pronome posposto aceita coorde-
nação com um adjetivo (6).
(5) a. Lá eu tinha um jardim gostosamente meu (Müller, 1997: 174).

b. *Lá eu tinha um gostosamente meu jardim (Müller, 1997: 174).
(6) a. Foi uma vitória gerativista e minha quando o grupo de Teoria Gramati-
cal conseguiu as primeiras classificações no concurso de monografias
(Müller, 1997: 174).
b. *Foi uma grande e minha vitória quando o grupo de Teoria Gramatical
conseguiu as primeiras classificações no concurso de monografias
(Müller, 1997: 174).
Há, portanto, diferenças sintáticas entre o pronome possessivo anteposto e o

pronome possessivo posposto. Segundo Müller (1997), essas diferenças refletem
a posição estrutural desses itens na hierarquia sintática, sendo o primeiro um
determinante, à semelhança de artigos, e o segundo um adjunto, à semelhança
de adjetivos. Nas seções seguintes, acompanhando a proposta de Cinque (2005),
veremos que o possessivo pós-nominal será analisado não como adjunto, mas
como especificador de uma projeção funcional (PossP) no DP.
2.2. Quadro teórico

Nesta seção, podemos destacar, inicialmente, a proposta de Cinque
(2005) segundo a qual adjetivos, possessivos, numerais e outros ficariam na
posição de especificador de categorias funcionais licenciadas no interior da
estrutura do DP. Posteriormente, abordaremos o estudo de Bernstein (2005)
sobre possessivos no inglês e no espanhol segundo o qual, no espanhol, o
possessivo pós-nominal não pertenceria à estrutura interna do DP, mas do
CP a ele subordinado.
2.2.1. O possessivo no DP
Segundo Abney (1987), Giorgi e Longobardi (1991), Cinque (2005) e
Aboh et al. (2010), a estrutura nominal reflete a estrutura sentencial, dentre
outras razões, por disponibilizar uma hierarquia de projeções funcionais, que
é determinada pela Gramática Universal. De acordo com Cinque (2005), os
modificadores são gerados em uma ordem fixa pré-nominal, qual seja, Dem
> Num > A > N. Assim, as diferentes ordens atestadas nas línguas resultam
do movimento da projeção máxima NP (e não do núcleo) para posições de
Spec em categorias funcionais (AgrP), geradas acima dos modificadores,
conforme mostrado no diagrama (7).
(7)
Cinque (2005: 317)
As categorias AgrP, geradas acima de cada projeção funcional, são justi-

ficadas por Cinque (2005: 325-326, tradução nossa) da seguinte maneira:
Suponhamos que cada sintagma (aquele contendo um Sintagma Adjetival, ou-

tro contendo o Sintagma de Número, outro contendo o Sintagma Demonstrati-
vo, etc.) precise ser dotado com um traço nominal para ser licenciado (i.e., para
ser contado como parte da projeção estendida do NP) e que isso possa ser reali-
zado inserindo, por merge, acima dele, um núcleo Agr(eement) cujo Spec, em
última instância, venha a ter esse traço nominal.3
Assim, o licenciamento desse traço que caracteriza o item como perten-
cente à estrutura nominal pode se dar simplesmente a partir do merge de Agr
ou a partir do movimento do NP para Spec,AgrP. O movimento, por sua vez,
pode decorrer de duas formas: o NP pode alçar sozinho ou conjugado com
um XP, em pied-piping. O primeiro tipo justifica as ordens (8a – c), enquan-
to o segundo justifica as ordens (8d – f), havendo ainda outras ordens possí-
veis.
(8) a. Dem Num N A c. N Dem Num A e. N A Dem Num

b. Dem N Num A d. A N Dem Num f. Dem A N Num
Em resumo, a variação na ordem dos modificadores na estrutura interna

do DP é explicada a partir do movimento do NP, como projeção máxima,
para Spec,AgrP. Além disso, cada categoria é licenciada com um AgrP, co-
mo indicativo de pertencimento à estrutura nominal.
2.2.2. O possessivo no CP
Bernstein (2005) estabelece uma comparação entre os pronomes posses-
sivos no inglês, no espanhol e no francês. Destacaremos adiante a relação
feita entre inglês e espanhol apenas, visto que o francês não possui pronome
possessivo pós-nominal (un ami à toi).
Segundo Bernstein (2005), com base em Cardinaletti (1998), os prono-
mes possessivos pós-nominais são formas fortes, que se caracterizam por
apresentarem uma complexidade morfológica maior que os pré-nominais.
Por exemplo, no espanhol, além de número e pessoa, o possessivo pós-
-nominal (9a) apresenta marca de gênero, enquanto o possessivo pré-
-nominal (9b) não apresenta concordância em gênero, sendo por isso menos
robusto em termos morfológicos.
(9) a. casas suyas b. sus casas
3
“Suppose that each phrase (the one containing an Adjective Phrase, the one containing the
Number Phrase, the one containing the Demonstrative Phrase, etc.) needs to be endowed
with a nominal feature to be licensed (i.e., to count as part of the extended projection of
NP), and that this can be brought about by merging above it an Agr(eement) head whose
Spec ultimately comes to have such a nominal feature” (Cinque, 2005: 325-326).
Além de serem mais complexos morfologicamente, os pronomes posses-

sivos pós-nominais apresentam um agrupamento de propriedades, algumas já
exemplificadas na seção anterior com dados de Müller (1997), que os carac-
teriza como formas fortes, tais como:
(10) a. podem ser coordenados: “vitória gerativista e minha”;

b. podem ser modificados: “jardim gostosamente seu”;
c. podem se realizar como predicativos: “esse jardim é seu”.
Por sua vez, uma semelhança entre inglês e espanhol é que ambas as lín-
guas podem expressar posse não apenas por meio de pronomes, mas também
por meio de DPs pós-nominais precedidos por preposição, como em (11).
Contudo, enquanto a construção possessiva em inglês permite a presença de
pronomes possessivos seguidos de preposição (12a).
(11) a. un amigo de Maria b. a friend of Mary’s

(12) a. un amigo de *suyo b. a friend of mine/yours
Essa diferença entre inglês e espanhol é decisiva para Bernstein (2005)

propor uma distinção no sistema de concordância dos possessivos nestas
duas línguas. Segundo Bernstein (2005: 66), a não existência de construção
com possessivo precedido de preposição no espanhol (12ª) reflete o fato de
que o pronome possessivo pós-nominal não pertence à estrutura do DP, mas
à estrutura de um CP reduzido. Sendo assim, em vez da preposição ‘de’, o
espanhol apresentaria um complementizador ‘que’, realizado como forma
zero, nas estruturas com possessivo pós-nominal. Nesta hipótese, D tomaria
CP como complemento [D[CP]], nos moldes da análise de Kayne (1993
apud Bernstein, 2005:66). Em outros termos, o possessivo pós-nominal seria
um predicativo: “um amigo (que é) seu”.
Portanto, o sistema de concordância do possessivo pós-nominal no espa-
nhol consistiria em um tipo de Agree sentencial (Bernstein, 2005: 67), como
se observa em (13).
(13) [D el [CP [amigo viejo]i [C Ø [IP/AgrP suyo [I/Agr [NP [ ]i ] ] ] ] ] ] (Bernstein,

2005: 66).
Sendo assim, segundo Bernstein (2005: 68), a diferença do espanhol para

o inglês seria o fato de que, no inglês, a concordância do possessivo pós-
-nominal é interna ao DP, enquanto, no espanhol, ela é sentencial.
3. Metodologia
Os procedimentos metodológicos empreendidos para obtenção dos dados
compreendem tanto a utilização de entrevistas e diálogos espontâneos, trans-
critos e disponibilizados para download em bancos de dados,4 quanto a utili-
zação de exemplos efectivos, coletados em diferentes eventos do cotidiano
em Minas Gerais, tais como: conversas espontâneas, debates acadêmicos,
reunião de negócios, atendimento comercial e outros. É a partir desta segun-
da fonte que se obtém a maior parte dos exemplos do fenômeno estudado.
Finalmente, para análise, o trabalho se desenvolve dentro das premissas me-
todológicas da sintaxe comparativa na qual se espera o estabelecimento de
correlações entre o português brasileiro e outras línguas, para enfim se pro-
por uma descrição formal.
4. Descrição e proposta de análise

4.1. Descrição
Adiante, pretende-se descrever o fenômeno em estudo e delimitá-lo, a
partir de um contraste com diferentes ocorrências do possessivo “seus/suas”.
Sendo assim, mostraremos: a que pessoas o possessivo ‘seus/suas’ pode se
referir (4.1.1); exemplos em que o possessivo está posposto e anteposto
(4.1.2); exemplos em que o possessivo apresenta concordância não padrão
com o nome e ocorrências em que o possessivo de 2ª pessoa, flexionado no
plural, é ambíguo, podendo se referir tanto a 2ª pessoa do singular quanto do
plural (4.1.3). Com esse contraste, poderemos finalmente compreender as
especificidades dos dados que nos interessam (4.1.4).
4.1.1. Referência
Começamos por ressaltar que o pronome ‘seus/suas’ no plural pode ter
referentes de:
(14) 2ª pessoa do plural: “Caros eleitoresi, preciso de seusi votos!”
4
Em projeto desenvolvido na UFVJM de agosto de 2013 a julho de 2014, foi feita uma sele-
ção de dados com ‘seus’ e ‘suas’ a partir de corpora dos projetos: Mineirês (UFMG),
NURC (UFRJ), CEDoHS (UEFS), ALIP (UNESP) e PHPB (UFRJ/USP/UEFS). Esses da-
dos foram gerados por meio da utilização do AntConc. Como resultado, observou-se que,
em um total de 3.406.316 palavras, sendo a maior parte provinda de corpora de Minas Ge-
rais, São Paulo, Bahia e Rio de Janeiro, e uma pequena parte de Rio Grande do Norte, Per-
nambuco, Paraná e Paraíba, encontraram-se duas ocorrências do fenômeno apenas no banco
de dados do Estado de Minas Gerais (Projeto Mineirês – UFMG): “só pra saber um pouquin
da vida suas” e “se ocêis é inimigo pobrema é seus...”. No entanto, como será visto na seção
4.1.4, a produtividade dessas ocorrências se manifesta claramente em dados anotados de
fala espontânea.
(15) 2ª pessoa do singular: “o que que vocêi conversa com as suasi amigas?”
(Projeto Mineirês, MAR65).
(16) 3ª pessoa do plural: “alguns moradoresi, principalmente da rua São José,
abandonam suasi casas por causa da bagunça” (Projeto Mineirês, OP03).
(17) 3ª pessoa do singular: “elei é um homi qui trabalha na na prefeitura [...]
mais é um homi qui é apaixonadu pur mato, di ficá pescanu, levá suasi pa-
nelinha [...] fritá seu pêxim” (Projeto Mineirês, BH06).
Sendo assim, pode-se observar, no quadro (1), a relação de referência do

possessivo da seguinte maneira:
Exemplo Possessivo Referência a: Referente

14 seus 2PL “caros leitores”
15 suas 2SG “você”
16 suas 3PL “alguns moradores”
17 suas 3SG “ele”
Quadro 1: Quadro de referência pessoal do possessivo ‘seus/suas’
Como já mencionado, o interesse deste trabalho está em dados nos quais

o possessivo se refere a 2ª pessoa, não sendo foco do estudo dados como
aqueles em (16) e (17).
4.1.2. Posição
Partimos agora para uma amostra de ocorrências do possessivo de segunda
pessoa anteposto (18) e posposto (19), que são possibilidades do português:
(18) “E você continuou os seus estudos?” (Projeto Mineirês, BH15).

(19) “e se você fosse sua mãe assim e as duas filhas suas tivessem brigan[d]o, aí
você ia por de castigo” (Projeto Mineirês, MAR55).
Como já mencionado em seções anteriores deste texto, o interesse deste

trabalho está em dados nos quais o possessivo é posposto. Não serão parte
desse estudo dados como aquele em (18), pois ocorrências com cisão na con-
cordância, como mostrado em (2), são raras com o possessivo anteposto (20).
(20) “O seus carro não pode ficar estacionado aqui” (De porteiro a passagei-
ros de um carro, em Ouro Preto, 10 de junho de 2015).
4.1.3. Concordância
Será vista agora uma amostra de ocorrências do possessivo de segunda pes-
soa com concordância padrão (21) e com concordância não padrão (22), típica
de variantes coloquiais em geral no PB e não restrita a dialetos regionais.
No primeiro caso (21), o morfema aditivo ‘-s’, em ‘suas’, está em con-

cordância com o nome ‘avós’ também no plural. No segundo caso (22), exis-
te concordância, mas o nome ‘coleguinha’ apresenta uma realização zero do
morfema indicador de plural, conforme explica Scherre (1997).
(21) “convivi com meu avô quando eu era pequeno. E as suas avós? Eram brasi-
leiras tamém?” (Projeto Mineirês, ARC06).
(22) “e que que cê faz com as suas coleguinha? brinca de quê?” (Projeto Mi-
neirês, Fita nº 3).
Como já mencionado em seções anteriores deste texto, o interesse deste

trabalho não está em dados nos quais o possessivo concorda em número com
o nome, seja por morfema aditivo (21), seja por realização zero deste mor-
fema (22). Por isso, é importante ter clara a distinção entre concordância não
padrão (com o nome) e ausência de concordância em número (com o nome).
Na concordância não padrão, como ocorre em (22), o nome não apresenta
marca de plural, mas ele se refere a mais de uma entidade (a coleguinhas, no
caso). Neste tipo de concordância, é esperado que apenas o primeiro ou os
primeiros elementos do DP apresentem marca de plural. Na ausência de con-
cordância em número com o nome, como ocorre em (23), o possessivo é o
único que recebe marca ‘-s’ de flexão, e o nome se refere a uma entidade
singular (à vida (23), à decisão (29), à caneta (30), ao projeto (31), etc.). Por
isso, enquanto o ‘-s’ de ‘suas’ em (22) se configura como concordância com
o nome, o ‘-s’ de ‘suas’, em (23) e em (20, 27-34, cf. seção 4.1.4), não tem
esta propriedade.
(23) “só pra saber um pouquin da vida suas. Cê tava falano que sua mãe vende,
cê num vende biscoito tambein não [...]?” (Projeto Mineirês, SJP01).
Quanto à concordância não padrão, Costa e Figueiredo Silva (2003: 27,

tradução nossa) afirmam que “possessivos pré-nominais apresentam concor-
dância em número em PB1, mas possessivos pós-nominais não apresentam
[...] a. o meus livro b. uns livro meu [...]”5. Se a marca de pluralidade tende a
ocorrer em possessivos pré-nominais e não em pós-nominais, o que se pode
dizer dos dados apresentados em (23) e em (27-28, 31-32)? Esses dados apre-
sentam marca de plural justamente no possessivo pós-nominal, sendo todos os
outros elementos do sintagma nominal singulares. Em princípio, os dados em
(23) e em (27-28, 31-32) não atendem a esta previsão, porque evidenciam
5
“Prenominal possessives show number agreement in BP1, postnominal ones do not [...] a. o
meus livro b. uns livro meu.”
uma outra gramática na qual o possessivo posposto de 2a pessoa não concor-

da em número com o nome.
Ainda quanto à questão da concordância, relacionada à referência, ressalta-
-se que, quando nome e possessivo estão no plural, não é possível identificar,
com clareza, a referência de 2ª pessoa, se plural ou singular, o que gera uma
estrutura ambígua6 (24). A frase (24) pode ser dirigida tanto a um interlocutor
apenas (25a) quanto a mais de um interlocutor (25b). No primeiro caso (25a),
fica claro que o ‘-s’ de ‘seus’ é um morfema de plural que estabelece concor-
dância com o nome plural. Como o vocativo ‘amigo’ é singular, não seria
possível dizer que o ‘-s’ de ‘seus’ compartilha traços de plural com seu ante-
cedente referencial. No entanto, em (25b), pode-se considerar que ‘-s’: em
uma leitura, estabelece concordância em número com o nome ‘favores’; em
outra leitura, compartilha traços de plural compatíveis com seu antecedente
referencial “amigos”.
(24) Preciso de dois favores seus! (‘seus’ = ‘de você’7ou ‘de vocês’?)
(25) a. Amigo, preciso de dois favores seus! (‘seus’ = ‘de você’)

b. Amigos, preciso de dois favores seus! (‘seus’ = ‘de vocês’)
(26) Amigos, preciso de um favor seus!8
A diferença entre (25b) e (26) é que, em (25b), devido ao fato de ambos,

nome e possessivo, estarem no plural, a cisão na concordância fica neutralizada,
isto é, não aparece9. Diferentemente, em (26), como nome está no singular e
possessivo no plural, esta cisão se evidencia.
6
Para uma reanálise ocorrer, é necessário que a mesma construção “esteja aberta à possibili-
dade de análises estruturais múltiplas, sendo que uma das análises potenciais é a antiga (...)
e a outra análise potencial é a inovadora (...)” (Harris; Campbell, 1995: 72, tradução nossa).
Pensando em termos de reanálise, se pudéssemos escolher uma estrutura ambígua em que a
marca de plural indicar, em uma leitura, o número do nome e, em outra, o número da pes-
soa, compatível com o do antecedente (vocativo), esta estrutura seria (25b). Por sua vez, a
estrutura inovadora seria (26), pois, neste caso, a marca de plural só poderia indicar o núme-
ro da pessoa.
7
Para Perini (1985: 5), ‘de você’ possessivo é agramatical (*pai de você). Por outro lado,
Neves (2000: 473) mostra a seguinte ocorrência: “sei os podres de todos, DE VOCÊ e de
seus amigos”, evidenciando sua gramaticalidade.
8
Exemplo adaptado de “Preciso de um favor seus urgente” (De engenheiro de uma construto-
ra a dois engenheiros de outra companhia, em Belo Horizonte, fevereiro de 2014).
9
Conforme observação de parecerista anônimo, parece haver uma coexistência de duas
gramáticas: uma em que o possessivo concorda em número com o nome e outra em que o
possessivo concorda em número com o possuidor. Apesar de nos parecer clara a existência
de duas gramáticas, teríamos uma ressalva quanto à noção de "concordância com o possui-
dor", que não se aplica. Em Pereira (2016), argumenta-se que, neste dialeto, o pronome
possessivo posposto de 2ª pessoa sofre uma reanálise em seus traços de número. Nesta re-
Em resumo, esta seção pontua a diferença entre concordância não padrão

com o nome (22) e ausência de concordância em número com o nome (23 e
27-34), sendo este o fenômeno analisado no trabalho. A seção também mos-
tra que esse fenômeno só pode ser plenamente identificado quando o nome
está no singular e o possessivo posposto no plural (26)10.
4.1.4. Produtividade das ocorrências
Outros dados que revelam a cisão na concordância podem ser vistos nos
seguintes exemplos. Trata-se de ocorrências espontâneas, coletadas11 por
meio de anotações, em cidades do Estado de Minas Gerais.
(27) “Fico imaginando a alegria suas quando viram o trabalho pronto” (De pro-
fessora universitária a membros de um projeto, por ocasião de conversa in-
formal, em Lavras, setembro de 2012).
(28) “Sorte suas que só tem o Galo que dá prensa em Minas” (Mensagem de
What’s App a amigos após vitória do Atlético sobre o Cruzeiro, em Belo
Horizonte, 19 de abril de 2015).
(29) “A decisão final é suas” (De economista a dois engenheiros, por ocasião de
reunião de negócios, em Belo Horizonte, fevereiro de 2014).
(30) “A caneta é suas” (De cliente de restaurante a garçonete, por ocasião de

atendimento comercial, em Belo Horizonte, março de 2014).
(31) “O projeto seus prevê aplicação no ensino?” (De professora universitária a

uma palestrante, por ocasião de evento acadêmico, em Diamantina, junho
de 2013).
(32) “O sensor de rotação seus raspa na roda fônica do motor” (Canal High
Torque. Disponível em: <https://www.youtube.com/watch?v=
q0yzIdIydmo>. Acesso em: 31 de julho de 2016. A ocorrência está em
1’50” do vídeo).
análise, o número passa a estar associado à categoria de pessoa e vem do léxico como traço
interpretável deste pronome. Sendo interpretável, o traço não desencadeia concordância,
visto que já está valorado. Portanto, nesta gramática, em posição posposta, ‘seu’ refere-se a
2ª pessoa do singular, e ‘seus’ a 2ª pessoa do plural.
10
Em Pereira (2016), observa-se que o fenômeno pode ser identificado também quando o
nome está no plural e o possessivo posposto no singular: “Amanhã, ele verá dois serviços
seu” (Belo Horizonte, setembro de 2015).
11
Agradeço ao Marcus o envio de dados coletados em Belo Horizonte.
A produtividade do fenômeno se evidencia não apenas pela sua frequên-

cia na dinâmica da oralidade em Minas Gerais, como também pela possibili-
dade de elipse do nome (33), além da possibilidade de o possessivo se pos-
por NP com artigo definido (31) e indefinido (26)12 ou a NP sem artigo (28),
e com pronomes, como ‘nada’ (34).
(33) Sobre interfone: “O meu está com defeito, e o Pedro me disse que o seus
também” (De professora aposentada a vizinha, por ocasião de conversa es-
pontânea, em Belo Horizonte, março de 2014).
(34) “Eu não contratei nada seus, não!” (De cliente a engenheiro, em Belo Hori-
zonte, fevereiro de 2014).
Além disso, observa-se que, em termos de referência, o possessivo pode

estar em uma sentença dirigida a apenas um interlocutor e, ainda assim, refe-
rir-se a 2ª pessoa do plural. Em (30), a afirmação é dirigida a apenas uma
garçonete, mas o possessivo plural se refere ao grupo de responsáveis pelo
restaurante. Em (31), a pergunta é dirigida a apenas uma palestrante, mas se
refere ao grupo de pesquisadores atuantes no projeto. Por último, em (33), a
afirmação é dirigida a apenas uma vizinha, mas se refere aos moradores do
apartamento.
Em suma, observaram-se diversos dados com possessivo posposto de 2ª pes-
soa, que, marcado com plural, não concorda com os outros elementos do
DP, no singular.
4.2. Propostas de análise
Em estruturas como (27), repetida abaixo (Pereira, 2016) como (35a), o
possessivo plural pós-nominal concorda em gênero com o nome, mas não em
número. Portanto, ‘suas’ em (35a) compartilha propriedades do espanhol
(35b), por um lado, que é a concordância em gênero com o nome, e do inglês
(35c), por outro lado, que é a coincidência de traços de número com os do
referente/antecedente.
(35) a. “Fico imaginando a alegria suas quando viram o trabalho pronto”.

b. la(s) casa(s) suya(s)
c. a friend of his/theirs
12
Segundo Castro (2001 apud Costa; Figueiredo Silva, 2003: 25), se possessivos ocorrem em
posição anteposta, o DP é definido, mas, se ocorrem em posição posposta, o DP é indefini-
do. Os dados (27) e (31) não atendem a esta previsão, pois o possessivo pós-nominal ocorre
com artigo definido.
Ainda comparando as duas línguas, observa-se que, no português (36a) e

no espanhol (37a), o pronome pessoal (‘vocês’ e ustedes) diferencia morfo-
logicamente a referência à segunda pessoa do plural.
(36) a. um favor de vocês

b. um favor seu
c. “Preciso de um favor seus urgente” (De engenheiro de uma construtora
a dois engenheiros de outra companhia, em Belo Horizonte, fevereiro de
2014).
(37) a. un favor de ustedes

b. un favor suyo
c. *un favor suyos
Contudo, ‘seu’, em (36b) não torna clara a referência à 2ª pessoa do plu-

ral, o que nos leva à seguinte conclusão: em vez de um pronome possessivo
como ‘vosso’, em desuso no PB, que seja capaz de significar exatamente o
que ‘de vocês’ significa, tem-se utilizado o pronome no plural (36c), mesmo
com nome no singular. Nota-se, porém, que a contraparte de (36c) no espa-
nhol (37c) é agramatical, o que aponta para o surgimento, no PB, especial-
mente no dialeto mineiro, de um padrão diferente daquele esperado para
línguas românicas, que merece ser investigado.
Vale destacar ainda um fenômeno semelhante que ocorre no inglês. O
pronome you pode se referir tanto a 2ª pessoa do plural quanto a 2ª pessoa do
singular. Em certos dialetos, registram-se novas estruturas como yous e you-
-uns (Maynor, 2000), em que se percebe justamente o acréscimo de ‘-s’ para
deixar clara a referência à 2ª pessoa do plural. Com este mesmo objetivo, o
possessivo de 2ª pessoa recebe o acréscimo de guys’s , em your guys’s.
Finalmente, uma última propriedade a ser mencionada se refere a restri-
ções na posição do possessivo, nas construções em estudo (38a). Enquanto o
genitivo “de vocês” jamais ocorre anteposto (38b), o possessivo (indicando
“de vocês” em estruturas com cisão na concordância) também é bastante
incomum anteposto.
(38) a. a alegria de vocês/a alegria suas

b. *a de vocês alegria/a suas alegria13
13
A estrutura em (38b) “a suas alegria” é ambígua, como descrito em A e B abaixo. Quando
afirmamos que a posição anteposta do possessivo é incomum, referimo-nos ao padrão des-
crito em B.
A) Pode ser equivalente a “as suas alegrias” em PB padrão. Nesse caso, o DP como um todo é
plural, e a marca de plural no possessivo está associada à concordância nominal , que não é
marcada morfologicamente em todos os elementos do DP, em PB não padrão.
Considerando essas propriedades, averiguam-se duas hipóteses concor-

rentes para análise: uma em que o possessivo seria parte da estrutura interna
do DP e outra em que ele seria entendido como parte de um CP subordinado
ao DP, em posição predicativa, conforme será tratado nas seções seguintes.
4.2.1. O possessivo no DP
A primeira proposta pode ser formulada com base em Cinque (2005). Em
(39), uma derivação hipotética de (2a/35a), observa-se que o possessivo seria
gerado em uma posição baixa na hierarquia do DP, mas ainda acima do NP.
Por isso, o NP deveria ser alçado a Spec,AgrPPossP de modo a gerar a posição
pós-nominal do possessivo.
(39) [DP D a [AgrPossP [NPalegriai] Agr FEM [PossP suas Poss 2ªPL [NP ti]]]]
Quanto à concordância, poder-se-ia sugerir, que Agrº, categoria nominal

inserida junto a PossP, tem apenas o traço de gênero (FEM) compatível com
o do nome. O núcleo da categoria possessiva, por sua vez, abrigaria traços de
pessoa e número (2PL) do pronome.
Em suma, o movimento do NP para Spec,AgrP fornece uma explicação
para a posição pós-nominal do possessivo. No entanto, a proposta acima
descrita não oferece respostas para as seguintes perguntas: (i) O que causa a
cisão na concordância nesses dados, que se dá em gênero com o nome, mas
não em número? (ii) Qual é o papel dos traços-φ do possessivo neste
fenômeno? (iii) Por que este fenômeno é produtivo com o possessivo em
posição pós-nominal, mas não em posição pré-nominal? (iv) Existe algum
tipo de restrição no que se refere à posição pré-nominal ou motivação no que
se refere à posição pós-nominal?
4.2.2. O possessivo no CP
Diante das dúvidas apresentadas, parte-se para a segunda proposta de
análise. Bernstein (2005) mostra que, no espanhol, possessivos pós-nominais
não são precedidos por preposição. Em vez de uma preposição, o espanhol
teria um complementizador zero introduzindo uma oração relativa reduzida.
Aplicada a (2a/35a), esta proposta poderia ser justificada porque o posses-
sivo pós-nominal, além de não ser precedido por preposição (*a alegria de
suas), poderia estar em CP, devido à possibilidade de posposição a um verbo
(2b, 29 e 30), em estruturas predicativas.
B) Pode ser equivalente a “a alegria de vocês”. Nesse caso, o DP como um todo é singular, e a
marca de plural no possessivo está associada aos traços pronominais de pessoa. Este é o caso de
(20). Em “o seus carro”, ‘s’ indica o número da pessoa (2ª PL) e não dos elementos nominais (D
e N), que estão todos no singular.
Nesta hipótese, D tomaria CP como complemento [D[CP]], conforme

(40), uma derivação hipotética de (2a/35a). Então, o possessivo ocuparia a
posição Spec,AgrP, correspondente ao IP da oração, enquanto o NP se move
para Spec,CP, gerando assim a ordem pós-nominal. Segundo Bernstein
(2005), em orações relativas, é esperado o alçamento do NP, como em “A
[alegriai] que ti é suas”, o que impediria a permanência do NP in situ. Esta
proposta permitiria uma análise unificada das estruturas em (2), sendo que,
em (2a), o possessivo estaria em um CP reduzido (“a alegria suas”) e, em
(2b, 29 e 30), em um CP com a presença explícita de verbo (“a caneta é
suas”).
(40) [DP Da [CPalegriai C°[AgrP suas Agr° [NP ti]]]]
No entanto, já foi observado que, embora o possessivo em posição pré-

nominal, com este tipo de cisão na concordância nominal, seja fato inco-
mum, houve ocorrência nos dados (20). Dessa forma, impedir a permanência
do NP in situ não seria condizente para a análise destas ocorrências.
Sendo assim, diferentemente da proposta de Bernstein (2005), a proposta
de Cinque (2005) é viável para explicar as diferentes posições lineares do pos-
sessivo. Contudo, questões específicas sobre a cisão na concordância, como
aquelas listadas de (i) a (iv) acima (seção 4.2.1.), não são respondidas com
base em nenhuma dessas propostas, mas em estágio posterior desta pesquisa
(Pereira, 2016).
Em suma, nesta seção 4, observou-se que o possessivo, nas construções
analisadas, apresenta as seguintes propriedades: i) é uma forma de 2ªPL; ii)
pode ocorrer em posição predicativa (2b); iii) é nitidamente mais frequente
em posição posposta; iv) dispõe de uma forma analítica (‘a alegria de vo-
cês’); e v) apresenta cisão na concordância, que se dá em gênero com o no-
me, mas não em número.
5. Considerações finais
Nesta pesquisa, temos o propósito de analisar construções nominais e predi-
cativas com possessivo de segunda pessoa do plural, como “o projeto seus” e
“a caneta é suas”, que se mostram recorrentes no dialeto mineiro. Foram
levantadas duas hipóteses de análise, sendo considerada mais adequada a de
Cinque (2005) segundo a qual se justifica a posição pós-nominal do posses-
sivo a partir do movimento do NP para Spec,AgrP, acima de PossP. Esta
investigação, por um lado, demanda a realização de uma série de empreen-
dimentos futuros, tais como: explorar a relação entre possessivos pós-
-nominais e outras categorias no DP; discutir as propriedades oracionais no e
do DP, especialmente orações relativas e estruturas predicativas; verificar os

parâmetros que permitem este tipo de cisão na concordância no PB, mas o
impedem em outras línguas românicas; e principalmente explicar qual é exa-
tamente o papel dos traços-φ do pronome possessivo nestes fatos (Pereira,
2016). Por outro lado, o estudo, em carácter pioneiro, foi capaz de revelar
dados inéditos na literatura sobre possessivos, descrevê-los, levantar ques-
tões investigativas e apontar caminhos de análise, à luz do modelo formal.
Referências
Abney, Paul (1987). The English noun phrase in its sentential aspect. Dissertação de
doutoramento, MIT.
Aboh, Enoch; Corver, Norbert; Dyakonova, Marina; Koppen, Marjo van (2010).
DP-internal information structure: some introductory remarks. Lingua 120
(4), pp. 782-801.
Bechara, Evanildo (2003). Moderna gramática portuguesa. Rio de Janeiro: Lucerna.
Bernstein, Judy (2005). On the morpho-syntax of possessive constructions. Re-
cherches linguistiques de Vincennes, 34, pp. 55-76.
Cardinaletti, Anna (1998). On the deficient/strong opposition in possessive systems.
Working Papers in Linguistics 8 (1), pp. 65-111.
Castilho, Ataliba (2010). Os possessivos. In: _______. Nova gramática do portu-
guês brasileiro. São Paulo: Contexto, pp. 501-505.
Castro, Ana. (2001). Os possessivos em português europeu e português brasileiro:
unidade e diversidade. Textos Seleccionados. XVI Encontro Nacional da
APL. Lisboa: Colibri, pp. 599-613.
Cerqueira, Vicente (1996). A sintaxe do possessivo no português brasileiro. Disser-
tação de Doutoramento, Universidade Estadual de Campinas, Campinas.
Chomsky, Noam (1995). The Minimalist Program. Cambridge, Mass.: MIT Press.
Cinque, Guglielmo (2005). Deriving Greenberg’s Universal 20 and its exceptions.
Linguistic Inquiry 36 (3), pp. 315-332.
Costa, João; Figueiredo Silva, Maria Cristina (2002). Notes on nominal and verbal
agreement in Portuguese. Revista di Grammatica Generativa 27, pp. 17-29.
Cunha, Celso; Cintra, Luiz (2001). Nova gramática do português contemporâneo.
Rio de Janeiro: Nova Fronteira.
Giorgi, Alessandra; Longobardi, Giuseppe (1991). The syntax of noun phrases:
configuration, parameters and empty categories. Cambridge: Cambridge Uni-
versity Press.
Harris, Alice; Campbell, Lyle (1995). Historical syntax in cross-linguistic perspec-
tive. Cambridge: University Press.
Kato, Mary (1985). A complementaridade dos possessivos e das construções geniti-
vas no português coloquial: réplica a Perini. DELTA 1 (1 e 2), pp. 107-120.
Maynor, Natalie (2000). Battle of the pronouns: y’all versus you-guys. American
Speech 74 (4), pp. 416-418.
Müller, Ana (1997). A gramática das formas possessivas no português do Brasil.

Dissertação de Doutoramento, Universidade Estadual de Campinas.
Neves, Maria Helena (2000). O pronome possessivo. In: _____. Gramática de usos
do português. São Paulo: UNESP, pp. 471-489.
Neves, Maria Helena (1993). Possessivos. In: Castilho, Ataliba (org.). Gramática do
português falado. Campinas: UNICAMP. v. 3, pp.149-211.
Pereira, Bruna. Number features on second person possessives in dialectal Brazi-
lian Portuguese. Florida Linguistics Papers (Proceedings of the Florida Lin-
guistics Yearly Meeting – FLYM 3), 2016 (In press).
Perini, Mário (1985). O surgimento do sistema de possessivo coloquial: uma inter-
pretação funcional. DELTA, 1 (1 e 2), pp. 1-16.
Rocha, Fernanda (2009). A alternância nos pronomes possessivos do português de
Belo Horizonte. Dissertação de Mestrado, Pontifícia Universidade Católica
de Minas Gerais.
Scherre, Marta (1997). Concordância nominal e funcionalismo. ALFA 41 (esp), pp.
181-206.
Silva, Giselle (1996). Estertores da forma seu na língua oral. In: Silva, Giselle;
Scherre, Maria (orgs.). Padrões sociolinguísticos: análise de fenômenos vari-
áveis do português falado na cidade do Rio de Janeiro. Rio de Janeiro: Tem-
po Brasileiro, pp. 170-181.
Universidade Federal de Minas Gerais. Faculdade de Letras. Projeto mineirês: a
construção de um dialeto. Coordenação de Jânia Ramos, Belo Horizonte.
Disponível em: <http://www.letras.ufmg.br/mineires/>. Acesso em: 30 mar.
2014.
DON’T PUT A LABEL ON ME.
LABELS AND INTERPRETATION OF É QUE-CLEFTS. 1
ALEKSANDRA VERCAUTEREN
(Ghent University & Universidade Nova de Lisboa)
ABSTRACT: Through the analysis of the discourse properties and the distribution of
European Portuguese é que-clefts in root and embedded contexts, this paper aims at
contributing to the discussion in cartographic approaches to Generative Grammar
on how much of discourse properties should be codified in syntax. Building on Cec-
chetto & Donati (2015), I argue that a contrastive interpretation is due to the label
of é que-clefts, which is determined by probing. However, when the é que-cleft has
no label, which is possible in root contexts, the cleft constituent is not necessarily
contrastive but rather referentially controlled. Label-less é que-clefts are the result
of unprobed movement of the cleft constituent. Different types of minimality apply in
labelled and unlabelled é que-clefts. Relativized Minimality causes the incompatibil-
ity of labelled é que-clefts with contexts in which Q-movement takes place. Gross
Minimality in unlabelled clefts causes the low frequency of object clefts.
KEYWORDS: left periphery, focus, é que-clefts, European Portuguese, intervention,

labelling
1. Introduction
The question of how much of discourse properties should be codified in syn-
tax has been widely discussed since Rizzi (1997). Some authors defend a
discourse-free syntax (Chomsky, 2001; Costa, 2010; a.o.), while others ar-
gue that (some) discourse features are active in syntax (Rizzi, 1997, 2013a;
Belletti 2004; a.o.). This paper focuses on European Portuguese (EP) é que-
clefts, for which it has been argued that the cleft constituent (CC henceforth)
1
Research funded by FWO13/ASP/258. I thank Liliane Haegeman, Maria Lobo and Lobke
Aelbrecht for comments on previous versions of this paper, as well as an anonymous re-
viewer of Estudos Linguísticos. I also thank the audience of PLUS 2013 for their useful
comments, in particular João Costa and Patrícia Amaral.
130 Aleksandra Vercauteren
carries a focus feature and moves to the left peripheral FocP (Modesto, 1995;
Ambar, 2005). An example of an é que-cleft is given below:
(1) O queijo é que o corvo comeu.

the cheese be-PRES-3SG that the raven eat-PRET-3SG
Cleft constituent (CC) – copula – complementizer – cleft clause
If this analysis is correct, we expect that the CC will always have a con-
trastive focus interpretation. We also expect it to interact with elements in
the left periphery with similar features, analogous to for instance English
argument fronting. Such fronting operations are restricted in embedded con-
texts with operator movement (see Haegeman, 2012 for a detailed descrip-
tion).
I will illustrate that neither prediction is borne out. On the one hand, the
CC can have interpretations other than contrastive focus, at least in root con-
texts, and on the other hand, although é que-clefts are restricted in embedded
contexts, they are less restricted than for instance English argument fronting.
One prerequisite on embedding is however that the CC has a contrastive
focus interpretation. These observations indicate that é que-clefts have dif-
ferent properties in embedded and in root contexts.
I account for these observations building on the idea that the root can re-
main label-less, while embedded structures need a label in order to be visible
in the computation, as was argued by Cecchetto & Donati (2015). More pre-
cisely, I propose that the lack of a label on root é que-clefts, due to unprobed
movement, allows them to be interpreted rather freely, as long as the CC can
be linked to an element in the Common Ground. In contrast, embedded é
que-clefts, which enter further computation, do need a label. This is only
possible if the CC undergoes probed movement. I propose that a focus fea-
ture on C probes the CC and triggers movement. The probe provides the
label of the new syntactic object, which will be labelled CP, with a visible
focus feature on it. Because of this label, the CC will have a contrastive fo-
cus interpretation. Additionally, since focus features are Q-features (Rizzi,
2004), the structure is ungrammatical in contexts with other Q-movement
operations, such as relative clauses. Regarding the sequence é que, I will
adopt the standard view that it is a grammaticalized unit without internal
structure, lexicalizing one single left peripheral head (Ambar, 2005; Soares,
2006; Lobo, 2006; Costa & Lobo, 2009).
The paper is organized as follows: in section 2 I describe the relevant da-
ta. The first part of the section focuses on the discursive value of é que-
-clefts, showing that the CC is not always contrastively focalized, but rather
referentially controlled. The second part discusses the restrictions on embed-
ding. Section 3 focuses on the role of contrast, showing first that contrastive
Don’t put a label on me 131
CCs do not need to be referentially controlled, and that contrast is needed in

some embedded contexts for the é que-cleft to be felicitous. In section 4 I
develop the account sketched out above.
2. Data
This section presents two aspects of é que-clefts about which the FocP-
analysis makes predictions. First I show that, contrary to standard assump-
tions, the CC of é que-clefts is not always interpreted as a contrastive focus.2
When examining data from spontaneous speech, we find a variety of discur-
sive values associated with é que-clefts. In the second part of this section I
demonstrate that é que-clefts cannot be embedded in contexts where Q-
movement takes place.
The data analyzed in this paper is mainly from the Cordial-SIN corpus, a
corpus of spontaneous and semi-directed speech, obtained from native
speakers of EP with little or no scholarly education, from rural areas.3 The
corpus examples cited in this paper are followed by the code of the locality
and the number of the file, according to the Cordial-SIN norms. Constructed
examples were all judged for acceptability by native speakers from Lisbon.
The source of any other examples will be indicated in the text. Examples
illustrating discourse properties will be paraphrased in order to illustrate the
relevant interpretation as accurately as possible. Word-by-word glosses will
be provided for é que-clefts only, or whenever discussing syntactic proper-
ties of other structures.
2.1 Discourse value
Traditionally it is assumed that the CC of é que-clefts is interpreted as a
contrastive focus (Casteleiro, 1979; Modesto, 1995; Costa & Duarte, 2001;
Ambar, 2005; a.o.). Indeed, the corpus analyzed for this research contains
2
By contrastive focus I mean the same as identificational focus. It differs from informational
focus in that the constituent marked with contrastive focus is not necessarily the new por-
tion of the proposition. The definition of contrastive focus I assume throughout this paper is
approximately the one proposed by Kiss (1998). The only difference is that I do not make
any assumptions on the degree of givenness of the set of alternatives, since givenness can
refer to several concepts as well (see Prince 1981 and Gundel 2003 for a discussion). I as-
sume the set of alternatives needs to be logically deducible:
(i) An identificational focus represents a subset of a logically deducible set of elements for
which the predicate can potentially hold; it is identified as the exhaustive subset of this
set for which the predicate actually holds.
3
The corpus used is Martins, A. M. (coord.) [2000- ] 2010. CORDIAL-SIN: Corpus Dialec-
tal para o Estudo da Sintaxe / Syntax-oriented Corpus of Portuguese Dialects. Lisboa, Cen-
tro de Linguística da Universidade de Lisboa. URL: http://www.clul.ul.pt/en/resources/411-
cordial-corpus
several instances of é que-clefts in which the CC is explicitly (2) or implicitly

(3) contrasted with an element in the surrounding discourse. The relevant é
que-cleft is boldfaced, the element the CC is contrasted with is underlined:
(2) INF1A cagarra onde é que vai ver o peixe é de dia, não é de noite.
‘The shearwater, it looks for fish by day, not by night.’
INQ2 Sim senhora.
‘Ok.’
INQ1 É de dia. Mas é uma que faz pa-pa-pa de noite.
‘It is by day. But it is one that goes pa-pa-pa by night.’
INQ2 A cagarra não é a mesma coisa.
‘The shearwater is not the same thing.’
INQ1 Espera lá.
‘Wait.’
INF2 A de noite não é esta.
‘The one by night is not this one.’
INF3 Gorguja.
‘Gorguja.’
INQ1 Não...
‘No...’
INQ2 A cagarra cá não é a mesma coisa.
‘The shearwater is not the same thing here.’
INF3 A gorguja é que é de noite.
the gorguja be.PRES.3SG that be.PRES.3SG of night
‘The gorguja is by night.’ (CLC05)
In this dialogue, the inquirers are asking for the name of a seabird. The
first informant says that the cagarra ‘shearwater’ hunts by day. The first
inquirer replies that they need the name of a bird that hunts by night, hence,
they are not looking for the name cagarra. Then the second informant con-
firms that this night hunting bird is not the cagarra, and finally the third
informant says it is the gorguja that hunts by night, thus identifying the bird
they are looking for and contrasting it with the cagarra.
Now consider the following example:
(3) INF É uma injustiça mesmo. A gente tem o prejuízo (...) e, no fim, com a
licença de caça e tudo, não se podemos caçar desde que a gente não
pague ali à sociedade.
‘It is really unfair. We suffer the losses and in the end, even with a
hunting license, we cannot hunt if we don’t pay the association.’
INQ Pois é.
‘I see.’
INF Temos que pagar dez contos de entrada; temos que pagar de seis em
seis meses cinco contos... Isto (...) é uma roubadeira!
‘We have to pay ten thousand quid to enter; we have to pay five thou-
sand quid every six months…That…is a rip off!’
INQ Pois é.
‘It is.’
INF Não é? É só uma roubadeira. Não tem lucro nenhum. O caçador não
tem lucro nenhum.
‘Isn’t it? It is just a rip off. There’s no profit at all. The hunter has no
profit at all.’
INQ Pois.
‘True.’
INF Pois então! Pois se as fazendas são nossas,
‘Well then! And if the farms are ours,’
a gente é que tem os prejuízos,
the people be.PRES.3SG that have. PRES.3SG the losses
a gente é que está a criar a caça!
the people be.PRES.3SG that be.PRES.3SG breed the game
‘it is us who suffer the losses, it is us who is breeding the game!’
(EXB03)
In the example above, a gente ‘us’ is implicitly contrasted with a sociedade

‘the association’: if you pay for something, in this case for hunting, you as-
sume that eventual costs will be covered by the institution you pay, the rele-
vant costs being those for breeding game. However, it is the people owning the
grounds that have to pay these costs, and not the association earning money
with the hunt. The é que-cleft is used to contrast the actual situation with the
expected one, it highlights the unexpectedness of the actual situation.
The data discussed so far confirm the generalization that the CC is a con-
trastive focus. However, we find several instances of é que-clefts in which
the CC lacks this interpretation. The following examples illustrate this:
(4) INQ1 E dá uma?… Uma espécie de um malmequer?

‘And it gives a?... Some kind of marigold?’
INF1 Sim. É, é.
‘Yes. That’s it.’
INQ1 Cor-de-laranja?
‘Orange?’
INF1 Cor-de-laranja.
‘Orange.’
INQ1 Já sei o que é.
‘I know what it is.’
INF1 É. Que cheira até um cheiro não…
‘Yes. It has a smell that’s even not...’
INQ1 Feio? Mau?
‘Ugly? Bad?’
INF1 É, é. Não é assim muito agradável. É bom para o jardim.

‘Yes, yes. It’s not that nice. It’s good for the garden.’
INF2 Esse amarelo,
‘This yellow one,’
INQ1 É. É.
‘Yes, yes.’
INF2 esse amarelo é que não é a
this yellow one be.PRES.3SG that not be.PRES.3SG the
bonina?
four o’clock flower
‘this yellow one, isn’t that the four o’clock flower?’
Que a gente chama cor de bonina? (MIG29)
‘The one we call four o’clock flower?’
In this dialogue, the participants are talking about flowers and their col-
ors. The é que-cleft is a polarity question in which the CC introduces the
topic of this question. It is not contrasted with the previous flower; it simply
introduces a new topic.
Another discourse function often associated with the CC of é que-clefts is
that of a scene-setter:
(5) E de forma que, depois, quando era para o ano, aí

and of form that after when be.PST.3SG for the year there
ao São Miguel, ao São Miguel, aí (...) em Fevereiro, agora
at.the São Miguel at.the São Miguel there inFebruary now
neste tempo, mais ou menos, Fevereiro, Março, é
in.this time more or less February March be.PRES.3SG
que ele ia outra máquina
that EXPL go.IMP.3SG other machine
‘In such a way that, afterwards, when it was a year later, around Saint
Michael, on Saint Michael’s day, around (…) February, in that period,
more or less, February, March, another machine went’
– outra máquina ou à mão- abria-se outra vala, tudo assim alinhado e
plantava-se o bacelo, como a gente lhe chama, o bacelo. (AAL02)
‘– another machine or by hand- another trench was opened, everything was
aligned, and the rootstocks were planted, as we call them, the rootstocks.’
In some cases, the CC does not have a specific referent and seems to
function as a discourse connector, indicating how the information in the cleft
clause should be interpreted in relation to the previous discourse. In the ex-
ample below, the CC indicates that what is described is a succession of
events, depois ‘afterwards’ does not refer to any moment in particular:
(6) Quando era de trigo da terra, peneirava-se a

when be.IMP.3SG of wheat of.the earth sift.IMP.3SG-CL the
farinha, duas vezes e depois é que se
flour two times and afterwards be.PRES.3SG that CL
deitava dentro da vasilha. (PST16)
pour.IMP.3SG inside of.the canister
‘When it was made of local wheat, the flour was sifted, twice, and then it
was poured into the canister.’
The examples discussed so far illustrate that, although the CC of é que-

clefts can be a contrastive focus (see (2) and (3)), this is not always the case.
The CC can have several other discursive functions, namely it can be an
aboutness topic (4), a scene setter (5) or a discourse connector (6).
The question arises whether there is one particular discourse property
shared by all é que-clefts. The answer is affirmative: the CC needs to be
related to an element that is already present in the common ground. This
relation can be established in several ways, which I specify in the rest of this
section.
In order to characterize the ‘linkedness’ of the CC of é que-clefts, I adopt
the notion referential control (Reinhart, 1980). A constituent is linked if it is
referentially controlled by an element in the Common Ground. Referential
control is defined as follows:
(7) a. An expression is directly controlled by another expression when two

expressions co-refer.
b. An expression is indirectly controlled when an expression contains a
direct mention of a previously mentioned referent.
c. An expression is implicitly controlled when an expression contains a
membership relation to a class previously mentioned.
Data from question-answer pairs provide evidence for the hypothesis that
the CC in é que-clefts has to be referentially controlled: unlike canonical
clefts (8), é que-clefts cannot constitute the answer to a wh-interrogative
(Soares, 2006; Vercauteren, 2010), as illustrated in (8a).4 In the following
contexts, # indicates that the sentence is grammatical but inappropriate in the
context, boldface indicates intended information focus:
4
Canonical clefts are the Portuguese counterpart of English it-clefts. Even though canonical
clefts are not the default answering strategy, the contexts where they can be used to consti-
tute an answer to a wh-interrogative are much less restricted than for é que-clefts.
(8) Quem cortou a árvore?

Who cut the tree?
a. #O João é que cortou a árvore.
the João be.PRES.3SG that cut.PRET.3SG the tree
‘João (is that) cut the tree.’
b. Foi o João que cortou a árvore.
‘It was João that cut the tree.’
Nevertheless, é que-clefts are felicitous to answer d-linked interrogatives,

where the range of answers is limited to a set of alternatives that is contextu-
ally given or that the participants of the conversation have in mind (Pesetsky,
1987: 108). In this case, the CC is appropriately controlled by an element in
the common ground, both by set-membership and coreference:5
(9) (talking about alternative forms of energy)

E neste momento, qual é que é mais bem visto pela população?
‘And at this moment, which one is the most accepted by the population?’
Não posso dizer com certeza, mas a energia solar
not can.PRES.1SG say with certainty but the energy solar
é que me parece mais bem vista.
be.PRES.3SG that CL seem.PRES.3SG most well see.PTCP
‘I cannot say with certainty, but solar energy (is that) seems to be the most
accepted.’
É que-clefts can also answer non d-linked interrogatives if the cleft con-
stituent is properly controlled. In (10), the constituent is a member of a set of
time intervals that has been identified in the previous discourse:
5
Note that the alternatives of the set have to be identified, a generic set is not enough to
license é que-clefts. This corroborates the definition of implicit control, in which it is ex-
plicitly mentioned that the set should be previously mentioned:
(i) [S]f:= {x: x is a car}
A. Que carro é que estavas a conduzir no momento do accidente?
‘What car were you driving at the moment of the accident?’
B. #O meu Porche é que estava a conduzir.
‘My Porche (is that) I was driving.’
(ii) [S]f:={my Mitsubishi, my Toyota, my Porche,…}
A. Qual dos teus carros é que é mais rápido?
‘Which one of your cars is the fastest?’
B. O meu Porche é que anda mais depressa.
‘My Porche (is that) is the fastest.’
(10) A. Quando é que vamos visitor a avó?

‘When are we going to visit grandmother?’
B. Vamos ver… primeiro tenho de ir ao médico buscar uma prescrição.
Depois acho que é melhor irmos ao supermercado antes que feche.
‘Let’s see… first I have to go to the doctor to pick up a prescription.
Then I think it is better that we go to the supermarket before it closes.’
E só depois é que podemos ir
and only afterwards be.PRES.3SG that can.PRES.3PL go
visitar a avó.
visit the grandmother
‘And only then (is that) can we visit grandmother.’
The other types of relation with the common ground are also all attested
in the corpus. As such, the CC can be directly controlled by an element in
the common ground. In the next example, the cleft constituent refers to an
element that has been previously introduced:
(11) INQ1 Mas os almocreves vinham com quê? Com machos?

‘But the muleteers came with what? With mules?’
INF1 Vinha com machos.
‘They came with mules.’
INQ1 Não vinha com carros?
‘Didn’t they come with chariots?’
INF1 Era com machos. Os carros não vinham cá na altura!
‘It was with mules. Chariots didn’t come here back then!’
INF2 Eram homens, os donos, com os odres.
‘It were men, the owners, with the water skins.
Ele assim uns odres é (...) que eles
EXPL such some water skins be.PRES.3SG that they
levavam a transportar o azeite lá (...) para donde o
take.IMP.3PL to transport the oil there to where CL
queriam vender. (PVC27)
want.IMP.3PL sell
‘Something like a water skin they used to transport the olive oil to
the place where they wanted to sell it.’
The CC can also be indirectly controlled, i.e., é que-clefts are appropriate

if the CC contains a previously mentioned expression. In (12), the CC con-
tains a mention of leite ‘milk’, hence the cleft is licensed:
(12) (as crias) Andam com as mães a comer. Comer, come, com o pau, mas
mamar, não mama. Agora, as ovelhas não põem isso. Deixam lá as crias à
parte, à noite vêm, tiram o leite à ovelha, e depois (...)
(the offspring) walks with their mother to eat. It does eat, with the stick, but
suckle, it does not. Now, they don’t put this on the sheep. They keep the
offspring apart, by night they come, they milk the sheep, and then (…)
o que resta de leite é que vai para lá
what remain.PRES.3SG of milk be.PRES.3SG that go.PRES.3SG to there
ele (...) escorropichar. (ALC22)
EXPL drink up
‘what is left of milk (is that)(the offspring) goes there to drink up.’
In sum, the CC of é que-clefts has to be referentially controlled, which is

the case when the CC co-refers with an element in the common ground (11),
contains a mention of an element in the common ground (12), or is a mem-
ber of a set of alternatives in the common ground as in (9) and (10).
I have shown in this section that the CC of é que-clefts does not neces-
sarily carry a contrastive focus interpretation, but can also function as a top-
ic, as a scene setter or as a discourse connector. The only requirement seems
to be that the CC should be referentially controlled.
2.2. Restrictions on embedding
In the standard analyses of é que-clefts the CC occupies a Left Peripheral
position (Modesto, 1995; Ambar, 2005; Costa & Duarte, 2005; Lobo, 2006;
Soares, 2006). Such an analysis predicts that é que-clefts are not compatible
with contexts lacking a left periphery altogether, such as some non-finite
complements. This prediction is borne out:
(13) *Algumas pessoas afirmam nós é que

some persons affirm.PRES.3PL we be.PRES.3SG that
termos sido enganados
have.INF.1PL misled6
(Casteleiro, 1979: 114)
Another prediction of this analysis is that the CC interacts with other left
peripheral elements. As has been discussed extensively in the literature since
Emonds (1970), some patterns involving the Left Periphery of the clause are
highly restricted in embedded contexts. These patterns are called Root
Transformations or Main Clause Phenomena (MCP) and have been widely
discussed in the literature, see for example Aelbrecht et al. (2012) for some
recent proposals concerning these phenomena and an extensive bibliography.
6
Portuguese has marks of personal agreement on the infinitive in certain contexts. As an
anonymous reviewer points out, some infinitive complements have been argued to have a
(reduced) left periphery. Infinitival complements with inflected infinitive do not seem to be
one of those complements (Gonçalves & Duarte, 2001).
On a par with MCP such as English argument/VP-fronting, é que-clefts

cannot be embedded in temporal adverbial clauses (14), event conditionals
(15),7 wh-interrogatives (16) and relative clauses (17):
(14) *Quando o João é que chegou, a Maria

when the João be.PRES.3SG that arrive.PRET.3SG the Maria
ficou contente.
stayed happy
‘When João (is that) arrived, Maria was happy.’
(15) *Se a relva é que estiver seca, os meninos

if the grass be.PRES.3SG that be.SUBJ.3SG dry the boys
podem brincar no jardim.
can.PRES.3PL play in.the garden
‘If the grass (is that) is dry, the boys may play in the garden.’
(16) *O Pedro perguntou quando o João é que

the Pedro ask. PRET.3SG when the João be.PRES.3SG that
chegou.
arrive.PRET.3SG
‘Pedro asked when João (is that) arrived.’
(17) *O homem que ontem é que vimos no

the man that yesterday be.PRES.3SG that see.PRET.3PL on.the
Rossio é o meu irmão.
Rossio be.PRES.3SG the my brother
‘The man that yesterday (is that) we saw in the Rossio square is my brother.’
It is standardly assumed that movement takes place in wh-interrogatives

and in relative clauses (Chomsky, 1977; Kayne, 1994; Bianchi, 1995; Rizzi,
1997; a.o.), and, building on a long tradition starting with Geis (1970), Hae-
geman (2010, 2012) argues for the existence of operator movement in tem-
7
I adopt the terminology of Haegeman (2003). She distinguishes between event conditionals
(i), which modify the event expressed by the predicate of the main clause, and premise con-
ditionals (ii), which introduces a context in which the main clause is evaluated:
(i) If it rains we will all get terribly wet and miserable.
(ii) If [as you say] it is going to rain this afternoon, why don’t we just stay at home and watch a
video?
Both types of conditionals do not only differ in interpretation, but also in internal and exter-
nal syntax. Only the first blocks Main Clause Phenomena such as argument preposing.
Premise conditionals occupy a more peripheral position than event conditionals. See Hae-
geman (2003) and, for EP, Lobo (2003).
poral and event conditional adverbial clauses as well.8 Following proposals

by Rizzi (2004) and Haegeman (2012), I will assume that the element that
moves in these contexts carries a Q-feature. The Q-feature is to be seen as a
‘superfeature’, i.e. a class of features that gives rise to very similar syntactic
properties. These properties include the following: the moved constituent
leaves a gap, it licenses parasitic gaps, it creates weak islands, the movement
operation is sensitive to islands and is not compatible with other Q-
movement operations. The Q-feature includes at least foc(us), wh, neg(ation)
and rel(ativization), as was proposed by Rizzi (2004). The presence of
movement of a Q-constituent in the four contexts above makes them incom-
patible with é que-clefts, because the Q-constituent cannot cross the CC
which also carries a Q-feature. Indeed, the CC has properties of a Q-moved
constituent. First, it licenses parasitic gaps (Soares, 2006: 225, fn. 96):9
(18) A sopa é que a Maria fez sem provar [-].

the soup be.PRES.3SG that the Maria made without taste
‘Maria made the soup without tasting it.’
Second, in standard EP, like other Q-constituents, the CC cannot be dou-

bled by a (clitic) pronoun, it has to leave a gap:
(19) a. *O queijo é que o corvo comeu-o.

the cheese be.PRES.3SG that the raven eat.PST.3SG-CL
b. *O que é que o corvo comeu-o?
what be.PRES.3SG that the raven eat.PST.3SG-CL
Third, there can only be one Q-moved element in the CP of the clause:10
8
See Kayne (1975) and Hulsey and Sauerland (2006) a.o. for an analysis of relative clauses
without movement.
9
An anonymous reviewer points out that the example in (18) can also be analysed as involv-
ing a null object. However, the data in (19-21) support the hypothesis that the CC undergoes
Q-movement.
10
In some non-standard varieties of EP, there can be two CCs. Note that these structures are
ungrammatical for most speakers:
(i) A gente é que lá é que fazia as cangas
the people be.PRES.3SG that there be.PRES.3SG that made the oxbows
para a gente. (STJ57)
for the people
‘The people made the oxbows for us there.’
It is possible that in these varieties, the CC does not necessarily undergo Q-movement, as I
will suggest is section 4.
(20) *Quando é que o João é que vai

when be.PRES.3SG that the João be.PRES.3SG that go.PRES.3SG
chegar?
arrive
And finally, the CC cannot be extracted from strong islands:11
(21) *O João é que vi a menina que ama [-].

the João be.PRES.3SG that see.PST.1SG the girl that love.PRES.3SG
The incompatibility of é que-clefts with wh-interrogatives, relative claus-

es and temporal and event conditional adverbial clauses can thus be ascribed
to an intervention effect, in the line of what Haegeman (2010, 2012) propos-
es for English MCP. I assume, following Rizzi’s (2004) definition of featural
Relativized Minimality, that a constituent with a certain feature cannot cross
a constituent with a feature of the same class. In the cases under discussion
here, the Q-feature on the CC blocks movement of other Q-constituents
across it:
(22) *[CP OpQ CCQ … [ [TP V OPQ CCQ… ]]]
Additional evidence for this analysis is that in non-standard EP, é que-

clefts can be embedded in relative clauses, if the head of the relative clause
is doubled in its base position:
(23) Tinha uma roda que a força da água é

have.IMP.3SG a wheel that the force of.the water be.PRES.3SG
que fazia andar a roda e com os baldes porque
that make.IMP.3SG go the wheel and with the buckets because
tem esses 'tales' baldes, como era o estanca rio
have.PRES.3SG these ‘such’ buckets since was the river.stagnator
e regava aquele campo grande (...) que está além.(MIN31)
and irrigated that field big that be.PRES.3SG there
11
The CC can undergo long movement, hence the ungrammaticality of (21) is not due to a
general ban on long movement of the CC.
(i) [O João]i é que eu acho que a Maria
the João be.PRES.3SG that I think.PRES.1SG that the Maria
beijou ti.
kiss.pret.3sg
‘I think that Maria kissed João.’
‘There was a wheel that the force of the water put [the wheel] into motion,
with the buckets, because it has these buckets since it stanches the river and
irrigates that big field over there.’
Since a full DP occupies the base position of the head of the relative
clause, it is hard to assume it has been Q-moved to the left periphery.12 It is
notable that all three instances of é que-clefts embedded in relative clauses
found in the Cordial-SIN corpus are cases of resumptive relative clauses,
with a full DP doubling the head. The resumptive strategy would thus be a
way of avoiding an intervention effect.
Unlike English MCP, é que-clefts can occur in complements of factive
verbs (24), in clausal subjects (25) and in clausal complements of nouns
(26). In section 3.2 I will show that their appropriateness in these contexts
depends on the discourse context.
(24) O João lamenta que só com cunhas é que

the João regret.PRES.3SG that only with connections be.PRES.3SG that
se consiga arranjar emprego.
CL manage.PRES.3SG find work
‘John regrets that it is only possible to find a job if you have connections.’
(25) Que o João é que tenha ganho o prémio

that the João be.PRES.3SG that have.CONJ.3SG won.PRTC the prize
não me surpreende nada.
not CL surprise.PRES.3SG nothing
‘That John won the prize does not surprise me at all.’
(26) Platão mudou alguns aspectos deste modelo político, mas

Plato change.PST.3SG some aspects of.this model political but
manteve sempre a ideia que a razão é que
maintain.PST.3SG always the idea that the reason be.PRES.3SG that
devia governar, sendo a única que podia proporcionar aos
should govern being the only that could provide to.the
cidadãos a justiça e a felicidade.13
citizens the justice and the happiness
12
Also if we assume that it is not the head of the relative clause itself that moves, but a (cov-
ert) operator (see Chomsky, 1977), the lack of a gap makes it hard to argue for a movement
operation in these relative clauses. We could assume movement in combination with multi-
ple spell-out, but as far as I know, doubled elements are never quantificational in EP, and
as such we do not expect any intervention effects.
13
http://afilosofia.no.sapo.pt/platao1.htm
‘Plato changed some aspects of this political model, but he maintained

always the idea that reason should govern, being the only thing that could
provide the citizens with justice and happiness.’
Summing up, é que-clefts are restricted in some embedded contexts, as

the left-peripheral analysis predicts. Still, the pattern differs from the one
observed for English MCP: é que-clefts are less restricted than for example
fronted arguments in English.14
3. The role of contrast

In the previous section I have sketched the standard analysis of é que-clefts
according to which they activate an LP position (Costa & Duarte, 2005; Lo-
bo, 2006; Soares, 2006; FocP according to Modesto, 1995 and Ambar,
2005). First, I have shown that the CC is not always a contrastive focus con-
stituent. The crucial discursive restriction for the CC is that it needs to be
referentially controlled. I also illustrated that, although there are some re-
strictions on embedding, é que-clefts do not pattern with fronting operations
that are argued to make use of the same left-peripheral FocP as the CC does.
This section deals with the role of contrast in é que-clefts. On the one
hand, I show that contrastive CCs do not need to be referentially controlled.
On the other hand I show that, in certain contexts, é que-clefts with an overt-
ly contrastive CC are more easily embedded.
3.1. Contrast and referential control
Question-answer pairs illustrate that a contrastive constituent does not
need to be linked in the sense I described in section 2.1.. I illustrated that é
que-clefts can only be used to answer wh-interrogatives if there is a given set
of alternatives in the common ground (see (8)-(10)). However, if the CC is
contrasted with another (implicit) referent, the CC of é que-clefts can be
discourse new without any problem. In (27), a mãe ‘mother’ is contrasted
with the null subject ‘you’ in the question:
(27) Cortaste a árvore?

Did you cut the tree?
Não, a mãe é que cortou a árvore.
no the mother be.PRES.3SG that cut.PST.3SG the tree
‘No, mother cut the tree.’
14
As an anonymous reviewer correctly points out, fronting is generally much more restricted
in English than it is in Romance, also in root contexts.
Contrast does not have to be explicit. In the following example, the CC a

mãe ‘mother’ is contrasted with the subject of the implied answer ‘I don’t
know’.
(28) Quando é que é o próximo feriado?

when is the next holiday?
A mãe é que deve saber.
the mother be.PRES.3SG that should know
‘Mother probably knows.’
What seems to happen is that, when the CC is not referentially controlled,

the hearer deduces a plausible set with which the CC can be contrasted. The
set of alternatives is thus not necessary a priori, as in examples (9) and (10),
but a posteriori.
When neither referential control nor contrast is involved, é que-clefts are
inappropriate:
(29) O nosso novo vizinho é lindo!

Our new neighbour is gorgeous!
#Sim, e a sua mulher é que é também!
yes, and the his wife be.PRES.3SG that be.PRES.3SG also
‘Yes, and his wife is gorgeous as well!’
In the conversation above, the CC is not referentially controlled: it does

not co-refer with an element in the common ground, it does not contain a
mention of a previously mentioned referent and it is not an element of a set
that has previously been identified. There is also no contrasting set that can
be deduced, since both the new neighbor and his wife are considered to be
gorgeous.
I stress the fact that set-membership does not have the same effect in the
case of contrast as it has in the case of referential control: when there is con-
trast, there is a clear opposition to one or more alternatives. With referential
control this is not necessarily the case. Also, as is clear in examples such as
(28), previous identification of the set is not necessary. The set involved in
referential control on the other hand, is contextually determined, see exam-
ples (9) and (10). Of course, a referentially controlled element can also be
contrastive.
As argued by Mólnar & Winkler (2010: 1396), “like topic, contrast plays
an important role in information linking and contributes to the integration of
the utterance into a larger discourse context”. This is particularly relevant for
the analysis I elaborate in section 4, in which I argue that non-contrastive
clefts are label-less, and hence depend on the discourse for their interpreta-
tion.
3.2. Contrast and embedding

As mentioned in section 2.2, the discourse context has an influence on the
felicity of embedding é que-clefts in clause types that resist MCP. The gen-
eral observation is that é que-clefts seem to be more easily embedded when
the CC is contrastive. Compare the following examples:
(30) a. ?O João lamenta que a Maria é que

The João regret.PRES.3SG that the Maria be.PRES.3SG that
tenha tido um acidente.
have.CONJ.3SG have.PRTC an accident.
‘Jon regrets that Maria had an accident.’
b. O João lamenta que só com cunhas é
the João regret.PRES.3SG that only with connections be.PRES.3SG
que se consiga arranjar emprego.
that CL manage.CONJ.3SG find work
‘John regrets that it is only possible to find a job if you have connec-
tions.’
Both examples contain an é que-cleft inside of a factive complement, a

context generally incompatible with MCP. Without any context, the first
example is less acceptable than the second one. This can easily be explained:
the presence of the adverb só ‘only’ makes the CC inherently contrastive,
contrasting it with any possible alternative. The CC in (30a) on the other
hand is not contrasted with anything.
The same effect is found in peripheral adverbial clauses:
(31) a. ??O João vai ao dentista, já que ele é

the João go.PRES.3SG to.the dentist since he be.PRES.3SG
que precisa.15
that need.PRES.3SG
‘João goes to the dentist, since he needs it.’
b. Continuo sem perceber, porque é que
continue.PRES.1SG without understand why be.PRES.3SG that
nos lares da Santa Casa, só aceitam o casal
in.the homes of.the Santa Casa only accept.PRES.3PL the couple
e não só ele, já que ele é que precisa!16
15
Liliane Haegeman (p.c.) points out that this sentence might be odd because of the avoid
pronoun principle in null subject languages. However, in the absence of é que, the pronoun
can be overt without giving rise to the same oddness:
(i) O João vai ao dentista, já que ele precisa.
the João go.PRES.3SG to.the dentist since he need.PRES.3SG.
16
http://alzheimerdepapie.blogs.sapo.pt/
and not only he since he be.PRES.3SG that need.PRES.3SG

‘I continue not to understand, why it is that in the homes of Santa Casa, they
only accept the couple and not only him, since he is the one who needs it!’
Again, the second example, where o casal ‘the couple’ is contrasted with
the CC ele ‘he’, is much better than the first one, where there is no contras-
tive interpretation provided by the context. Note that referential control is
not enough; otherwise we would expect example (31a) to be felicitous, since
the CC co-refers with a given referent.
The same pattern can be observed in complements of nouns:
(32) a. ?A ideia de que o João é que devia liderar

The idea of that the João be.PRES.3SG that should lead
a empresa é uma aberração.
the company be.PRES.3SG an aberration
‘The idea that João should lead the company is an aberration.’
b. Platão mudou alguns aspectos deste modelo político,
Plato change.PST.3SG some aspects of.this model political
mas manteve sempre a ideia que a razão
but maintain.PST.3SG always the idea that the reason
é que devia governar, sendo a única que podia
be.PRES.3SG that should govern being the only that could
proporcionar aos cidadãos a justiça e a felicidade.
provide to.the citizens the justice and the happiness
‘Plato changed some aspects of this political model, but he maintained
always the idea that reason should govern, being the only thing that
could provide the citizens with justice and happiness.’
Once again, the contrastive (32b) is much better: from the context we can
deduce that a razão ‘reason’ is contrasted with all other alternatives, while in the
previous example the CC is not contrasted with any alternative. Also in com-
plements of non-assertive verbs, é que-clefts occur more easily if the CC is con-
trastive:
(33) a. ?Duvido que a Maria é que tenha

doubt.PRES.1SG that the Maria be.PRES.3SG that have.CONJ.3SG
tido um accidente
have.PRTC an accident.
‘I doubt that Maria had an accident.’
b. Duvido que só o Homem Aranha é que
doubt.PRES.1SG that only the Spiderman be.PRES.3SG that
saiba resolver este problema.
know.CONJ.3SG solve this problem
‘I doubt that only Spiderman knows how to solve this problem.’
Finally, the same pattern can be observed in subject clauses. The (b) ex-
ample is contrastive in the sense that generally there can only be one winner,
so all alternatives to João are excluded:
(34) a. ??Que o bolo é que o João tenha

that the cake be.PRES.3SG that the João have.CONJ.3SG
comido, é mesmo estranho.
eat.PRTC be.PRES.3SG really strange
‘That the cake João ate is really strange.’
b. Que o João é que tenha ganho o
that the João be.PRES.3SG that have.CONJ.3SG won.PRTC the
prémio não me surpreende nada.
prize not CL surprise.PRES.3SG nothing
‘That John won the prize does not surprise me at all.’
I have no clear explanation as for why the absence of contrast on an em-

bedded é que-cleft gives rise to different degrees of degradation (compare
(32a) with (31a)). A possible line of reasoning is that such differences are
due to the difference in accessibility of a contrastive interpretation in the
absence of any context explicitly providing such an interpretation. In relation
to (32a) for instance, it is easy to imagine a context in which several persons
compete for the leadership position, and João would be contrasted with all of
these. In (31a) on the other hand, there is no plausible contrastive set availa-
ble, since an interpretation in which João goes to the dentist because some-
one else needs to does not make sense.
The generalization seems to be the following: the more embedded the é
que-cleft, the more contrast it needs on the CC to be felicitous. However,
contrast cannot ‘save’ all embedded é que-clefts. No matter how contrastive
the CC, é que-clefts are always banned in contexts with Q-movement, be-
cause of the intervention effect discussed in section 2.2.
We thus seem to have two groups of embedding contexts: the first group
contains temporal and event conditional adverbial clauses, wh-interrogatives
and relative clauses and do not allow embedding of é que-clefts, inde-
pendently of their interpretation. The second group contains factive com-
plements, peripheral adverbial clauses, complements to nouns, non-asserted
complements and clausal subjects. This group allows é que-cleft embedding,
with the additional restriction that the CC must be contrastive.
The question arises why contrast is able to ‘save’ embedded é que-clefts
in some contexts but not in others. Based on English MCP, Haegeman
(2010, 2012) argues that an intervention account similar to the one devel-
oped in section 2.2 applies to all contexts blocking MCP, including factive
complements, and possibly complements of nouns and clausal subjects.
However, the data from EP show that we have to assume that not all MCP-
blocking contexts are identical. At least two groups of contexts should be

distinguished: the ones that block both English MCP and é que-clefts, and
the ones that block English MCP only.
I propose that in the first group of contexts, Q-movement of a (covert)
operator does take place, as has been argued by several authors, while this is
not the case in the second group. I will not elaborate on the reason why Eng-
lish MCPs are blocked in the second group of contexts, since that is outside
the scope of this paper. I refer the reader to Aelbrecht et al. (2012) for sever-
al proposals concerning these phenomena in English and other languages. In
the next section I propose an account for the need of contrast in embedded é
que-clefts, based on labeling.
4. Discourse and labeling

In this final section, I develop an analysis of é que-clefts that accounts for
the dichotomy we observe when it comes to the interpretation of the CC: in
root contexts, the interpretation seems to be rather free. In embedded con-
texts (in which é que-clefts are not ruled out because of intervention) the CC
can only be contrastive. I propose that the solution lies in the way Cecchetto
& Donati (2015, hencefort C&D) argue syntactic objects are labelled: labels
are only absolutely necessary when further computation is needed. When the
computation is complete, i.e. at the root, a label may be absent (see also Ott
2014). Also, labels are determined by probing relations. I argue that the in-
terpretation of the CC is restricted in case the CC undergoes probed move-
ment, resulting in a cleft with a label. When movement is unprobed, the re-
sulting object is label-less and interpretation is more free.
4.1. The Probing Algorithm (Cecchetto & Donati, 2015)
In recent work, C&D provide the following definition of label:
(35) When two objects α and β are merged, a subset of the features of either α or
β become the label of the syntactic object {α, β}. A label
(i) can trigger further computation
(ii) is visible from outside the syntactic object {α,β}
(C&D: 32)
They argue that labelling of syntactic objects is a core property of syntax

(contra Chomsky, 2013 who argues that they are only needed at the interfac-
es), since (i) they can trigger merge, both external and internal; (ii) features,
or labels, determine the distribution of constituents; (iii) they play a role in
selection and (iv) they are needed to explain locality restrictions on move-
ment.
The idea that labels are needed to trigger syntactic computation predicts
that every syntactic object will need a label, except those that do not enter
any further operations (see Citko, 2008: 941 for a similar suggestion). Since
the computation ends at the root of sentences, root sentences do not need a
label, although nothing prevents them from having one. In short, root sen-
tences differ from embedded ones in that they do not need a label.17
C&D argue that labels are determined by the probing algorithm (also see
Citko, 2008):
(36) The label of a syntactic object {α, β} is the feature(s) which act(s) as a
Probe of the merging operation creating {α, β}.
This predicts that each time merge is not probed, the resulting syntactic
object has no label, and the computation cannot proceed, since labels are
needed for a syntactic object to be visible in the computation and to trigger
further computation. Together with the previous prediction, this means that
unprobed movement can only take place in a very restricted set of contexts
in which labels are not necessary. Since the root can remain label-less, un-
probed movement to the root is possible. This is the second ingredient of the
analysis.
A third ingredient is that probed movement gives rise to different inter-
vention effects than unprobed movement. The former type is restricted by
Relativized Minimality (Rizzi 1990, 2004): if an element with the same fea-
ture as the probed feature sits in a position c-commanding the foot of the
movement chain, the result is ungrammatical. This was illustrated in section
2.2: a Q-feature on the CC blocks movement of another element, if this
movement is probed by a Q-feature. Unprobed movement on the other hand
is restricted by Gross Minimality (C&D): any shared feature can create a
minimality effect. The more features are shared with the intervening ele-
ment, the more difficult movement becomes. I refer to C&D for a detailed
discussion and several examples that support this view on minimality and
labeling.
17
Lobke Aelbrecht (p.c.) points out that this might imply look-ahead. I refer to C&D for a
discussion of this issue. An anonymous reviewer raises the question whether the idea that a
syntactic object can remain label-less is compatible with Minimalism. In the framework
adopted here, labeling comes ‘for free’, in the same manner merge does: since a label is de-
termined by probing (see (37)), the presence or absence of a label will depend on whether
merge is probed or not. Within the Minimalist framework, merge is seen as a costless oper-
ation that does not need any trigger. However, other requirements of the computation will
rule out unprobed merge in the majority of the cases. I refer to C&D for further details.
4.2 É que-clefts and probing

In the previous sections I illustrated that the interpretation of the CC is re-
stricted in embedded contexts. This can be accounted for if we assume that
embedded contexts need a label, and hence movement to the left periphery
of the embedded clause needs to be probed.
I propose that in embedded contexts, feature(s) on the embedded C
probe(s) the CC. The question now is what features are probed. Locality
restrictions can shed a light on this question. I illustrated that é que-clefts are
not compatible with contexts with Q-movement, so one of the probed fea-
tures is definitely a Q feature. However, incompatibility in local contexts can
also be due to the fact that both constituents target the same position. To be
sure that an intervention effect causes the ungrammaticality, extraction in
non-local contexts have to be tested (Abels, 2012). In the examples below, it
is clear that a wh-constituent cannot be extracted across the CC. However,
the CC can (marginally) be extracted from a wh-island. Assuming featural
RM, if the features of the intervener are properly included in the set of fea-
tures of the target, no RM problem arises (Starke, 2001; Haegeman, 2010).
The pattern below indicates that the CC has more features than the wh-
constituent.
(37) *O João perguntou quem disseste que a Maria é

the João asked who say.PRET.2SG that the Maria be.PRES.3SG
que beijou?18
that kiss.PRET.3SG
‘João asked who you said that Maria (is that) kissed.’
(38) ?Acho que [só o Rui]i é que a Maria

think. PRES.1SG that only the Rui be. PRES.3SG that the Maria
se lembra [wh quando chegou ti à Bélgica].
CL remember.PRES.3SG [wh when arrive.PRET.3SG ti to.the Belgium
‘I think that only for Rui Maria remembers when he arrived to Belgium.’
The question arises what these exact features might be. First of all, I as-
sume that CCs of embedded é que-clefts carry a focus feature. This feature
gives rise to the obligatory contrastive focus interpretation. Additionally,
since foc is a Q-feature (Rizzi, 2004), apart from giving rise to the interven-
18
In the absence of é que, this structure is perfectly grammatical:
(i) O João perguntou quem disseste que a Maria beijou.
the João ask.PRET.3SG who say.PRET.2SG that the Maria kiss.PRET.3SG
An anonymous reviewer suggests that it might be the case that é que creates an island, in
way similar to secondary que in Spanish recomplementation structures (see Villa-García,
forthcoming). This is not the case, as can be seen in example (39b).
tion effects described in section 2.2, it creates a weak island for extraction.
As such, non-specific wh-constituents cannot be extracted across the CC, as
in (39a), but specific wh-constituents can marginally move across the CC
(39b), as is the case for extraction of a wh-constituent from a wh-island (see
(40)).
(39) a. *Quemi dizes que a Maria é que beijou t i?

Who say.PRES.2SG that the Maria be. PRES.3SG that kiss.PST.3SG
b. ?[Que assunto]i dizes que o Rui é que
what problem say.PRES.2SG that the Rui be. PRES.3SG that
resolveu ti?
solve.PST.3SG
‘Which issue do you say that Rui solved?’
Additionally, since the extraction of the CC seems to be insensitive to

weak islands, it could be the case that the CC carries a specificity feature. It
has often been noted that specificity has an influence on the extractability of
constituents from weak islands (see Szabolcsi, 2006 for an overview). As
such, specific wh-constituents can be extracted from a weak island, while
non-specific ones cannot:
(40) a. ?[Which problem]i do you wonder how to solve ti?

b. *Whoi do you wonder when you met ti?
The CCs of é que-clefts exhibit a similar pattern. As can be seen in ex-

ample (38), a specific CC can be extracted from a wh-island. Non-specific
CCs on the other hand cannot be extracted from wh-islands (41b), while they
can be fronted in the absence of an intervening island (41a):19
(41) a. ?[Alguém desconhecido]i é que o João diz

someone unknown be. PRES.3SG that the João say.PRES.3SG
que viu entrar na casa ti.
that see.PST.3SG enter in.the house
‘Someone unknown John says he saw enter the house.’
b. ?? [Alguém desconhecido]i é que o João não
someone unknown be. PRES.3SG that the João not
se lembra quando viu entrar na casa ti.
CL remember.PRES.3SG when see.PST.3SG enter in.the house
‘Someone unknown John does not remember when he saw enter the
house.’
19
The felicity of é que-clefts with a non-specific CC is context-dependent. See Vercauteren
(2015) for details.
The CC thus behaves as other Q-constituents, in particular wh-elements:

it creates a weak island for extraction, and can be extracted from weak is-
lands if it is specific. This confirms that the CC carries a Q-feature, arguably
foc, and optionally a specificity feature. I will assume that the probed feature
is the focus feature, on a par with standard assumptions. I leave the exact
role of the specificity feature for future research.
Since features on C are the probing features, C provides the label, and the
cleft can be selected for further computation:
(42) [CP [DP o Joãofoc ] Cfoc é que [ o João comeu o bolo]]
Returning to the differences in interpretation in root and embedded claus-

es, I would like to suggest that the label, obligatory in embedded clauses and
on which the focus feature is visible since it has been probed, gives the CC a
contrastive focus interpretation. In root clauses on the other hand, the CC
can move for free, since no further computation takes place. In this case, the
syntactic object has no label, and cannot be selected:
(43) [Ø [DP o João ] C é que [ o João comeu o bolo]]
When the structure reaches the interfaces, there is no label that indicates
how it should be interpreted, hence the previous discourse is searched for a
plausible interpretation. This is not possible when the CC is not referentially
controlled, giving rise to infelicitous clefts such as the ones discussed in
section 2.1. However, nothing prevents the CC from undergoing probed
movement, even in root contexts, giving rise to a contrastive interpretation.
Both derivations co-exist.
One prediction this analysis makes is that in root contexts, movement will
be restricted by Gross Minimality. This is borne out by the fact that object
clefts are quite rare. In the corpus, the number of subject clefts is over-
whelmingly bigger than the number of object clefts (629 vs. 28). This is
because in object clefts, the CC has to cross the subject which, being a DP as
well, has several features in common with the CC, while in subject clefts, the
CC does not have to cross any constituent with the same features. 20 I refer to
20
An anonymous reviewer suggests that the difference between subjects and objects might be
due to the fact that (dislocated) subjects can be base generated in their surface position,
with a doubling null pronoun in the base position, as has been argued by for instance Bar-
bosa (1995) for EP. Although it might be the case that in some EP varieties subject CCs in
é que-clefts can be base generated, this is certainly not generally true, since subject CCs
cannot be extracted from a strong island:
(i) *o Joãoi é que vi a menina que ti ama.
the Joãoi be.PRES.3SG that see.PST.1SG the girl that ti love.PRES.3SG
C&D for more details concerning Gross Minimality, and a discussion of

similar subject-object asymmetries in other contexts.
A second prediction is that it should be possible to have two CCs in an é
que-cleft: if one of the CCs undergoes unprobed movement and the other
one probed movement, we do not expect there to be any intervention, since
different types of intervention are at stake. As I mentioned in footnote 10,
this is possible in some varieties of EP:
(44) A Fábia é que quase sempre é que ia.(GRJ41)

the Fábia be.PRES.3SG that almost always be.PRES.3SG that go.PST.3SG
‘Fábia went almost always.’
In short, the analysis in terms of labeling can easily explain why there is
an interpretative difference between root and embedded é que-clefts, and it
makes accurate predictions concerning intervention effects and recursivity.
5. Conclusion
In this paper I showed that the CC of é que-clefts is not always contrastive,
contrary to standard assumptions. The only prerequisite seems to be that the
CC is referentially controlled. Contrastive CCs do not need to be referential-
ly controlled.
Second, although é que-clefts are restricted in embedded contexts, they
pattern differently than English MCP. Based on the EP data, I distinguish
two groups of contexts: temporal and event conditional clauses, relative
clauses and wh-interrogatives do not allow English MCP or é que-clefts be-
cause the Q-operator cannot cross the fronted argument/the CC, which also
carries a Q-feature. The second group consists of factive/non-asserted com-
plements, clausal subjects, peripheral adverbial clauses and complements of
nouns. These allow for contrastive é que-clefts but block English MCP (with
the exception of peripheral adverbial clauses).
The analysis I proposed to account for the interpretative differences be-
tween root and embedded é que-clefts is based on the labeling theory devel-
oped by C&D. I argue that, contrary to embedded é que-clefts, which need a
label in order to be visible for further computation, root é que-clefts can re-
main label-less. As such, the CC can undergo unprobed movement to the left
periphery, giving rise to a label-less syntactic object. Since there is no label
that indicates how the structure should be interpreted, the previous discourse
is searched for a possible interpretation. That is why the CC needs to be ref-
erentially controlled.
Embedded é que-clefts need a label in order to be visible for further com-
putation, hence the CC needs to be probed. The features on the label of the
cleft give rise to a contrastive interpretation. This derivation is also available

in root contexts.
References
Abels, Klaus (2012). The Italian left periphery: A view from locality. Linguistic
Inquiry 43 (2), pp. 229-254.
Aelbrecht, Lobke; Liliane Haegeman & Rachel Nye. (Eds.) (2012). Main Clause
Phenomena: New Horizons. Amsterdam: John Benjamins.
Ambar, Manuela (2005). Clefts and Tense Asymmetries. In Anna Maria Di Sciullo
(Ed.), UG and external systems. Language, brain and computation.
Amsterdam: John Benjamins, pp. 95-127.
Belletti, Adriana (2004). Aspects of the low IP area. In Luigi Rizzi (Ed.), The
structure of CP and IP. New York: Oxford University Press, pp. 16-51.
Bianchi, Valentina (1995). Consequences of Antisymmetry for the Syntax of Headed
Relative Clauses. PhD dissertation, University of Pisa.
Casteleiro, João Malaca (1979). Sintaxe e Semântica das construções enfáticas com
é que. Boletim de Filologia, XXV.
Cecchetto, Carlo, & Caterina Donati (2015). (Re)labeling. Cambridge, MA: The
MIT Press.
Chomsky, Noam (1977). On Wh-movement. In Peter Culicover, Thomas Wasow, &
Adrian Akmajian (Eds.), Formal syntax. New York: Academic Press, pp. 71-
-133.
Chomsky, Noam (2001). Derivation by phase. In Michael Kenstowicz (Ed.), Ken
Hale: A Life in Language. Cambridge: MIT Press, pp. 1-52.
Chomsky, Noam (2013). Problems of Projection. Lingua 130, pp. 33-49.
doi:http://dx.doi.org/10.1016/j.lingua.2012.12.003
Citko, Barbara (2008). Missing labels. Lingua 118, pp. 907-944.
doi:10.1016/j.lingua.2008.01.001
Costa, João (2010). Discourse-free syntax. Talk presented at the 20th Colloquium on
Generative Grammar. Barcelona.
Costa, João & Inês Duarte (2001). Minimizando a Estrutura: uma Análise Unificada
das Construções de Clivagem em Português. Actas do XVI Encontro
Nacional da APL. Lisboa: APL/Colibri, pp. 627-638.
Costa, João & Inês Duarte (2005). Cleft strategies in Portuguese: a unified approach.
Unpublished manuscript. Lisbon.
Costa, João & Maria Lobo (2009). Estruturas clivadas: evidência dos dados do
Português Europeu não standard. Anais do Congresso Internacional da
Abralin 2, pp. 3800-3806. João Pessoa: Universidade Federal do Paraná.
Emonds, Joseph (1970). Root and structure-preserving transformations. PhD
dissertation, MIT.
Geis, Michael. (1970). Adverbial subordinate clauses in English. PhD dissertation,
MIT.
Gonçalves, Anabela & Inês Duarte (2001). Construções Causativas em Português

Europeu e Português Brasileiro. In Clara Nunes Correia & Anabela
Gonçalves (Eds.), Actas do XVI Encontro Nacional da Associação
Portuguesa de Linguística. Lisboa: APL & Colibri, pp. 657-671.
Gundel, Jeanette K. (2003). Information structure and referential givenness/newness:
how much belongs in the grammar? Proceedings of the HPSG03 Conference.
East Lansing, MI: CSLI Publications.
Haegeman, Liliane (2003). Conditional Clauses: External and Internal Syntax. Mind
and Language 18, pp. 317-399.
Haegeman, Liliane (2010). Locality and the distribution of main clause phenomena.
Accessible at http://www.gist.ugent.be/file/79.
Haegeman, Liliane (2012). Adverbial clauses, main clause phenomena and
composition of the left periphery. New York: Oxford University Press.
Hulsey, Susan, & Uli Sauerland (2009). Sorting out relative clauses. Natural
Language Semantics 14, pp. 111-137.
Kayne, Richard (1975). French Syntax. Cambridge, MA: MIT Press.
Kayne, Richard (1994). The Antisymmetry of Syntax. Cambridge, MA: MIT Press.
Lobo, Maria (2003). Aspectos da Sintaxe das Orações Subordinadas Adverbiais do
Português. PhD dissertation, Universidade de Lisboa.
Lobo, Maria (2006). Assimetrias em construções de clivagem do português:
movimento vs. geração na base. XXI Encontro Nacional da APL Textos
seleccionados. Lisbon: APL, pp. 457-473.
Modesto, Marcello (1995). As construções clivadas no PB: relações entre
interpretação focal, movimento sintáctico e prosódia. Master thesis,
Universidade de São Paulo.
Molnár, Valéria & Susan Winkler (2010). Edges and gaps: contrast at the interfaces.
Lingua 120, pp. 1392-1415.
Ott, Denis (2014). Labels, linearization, and syntactic satellites. DGfS 2014,
workshop on Labels & Roots. Marburg, Germany.
Pesetsky, David (1987). Wh-in-Situ: Movement and Unselective Binding. In Eric
Reuland, & Alice G. ter Meulen (Eds.), The Representation of
(In)definiteness. Cambridge, MA: MIT Press, pp. 98-129.
Prince, Ellen (1981). Toward a taxonomy of given-new information. In Peter Cole
(Ed.), Radical pragmatics. New York: Academic Press, pp. 223-256.
Reinhart, Tanya (1980). Conditions for text coherence. Poetics Today 1 (4), pp. 161-
-180.
Rizzi, Luigi (1997). The Fine Structure of the Left Periphery. In Liliane Haegeman
(Ed.), Elements of Grammar. Handbook in Generative Syntax. Dordrecht:
Rizzi, Luigi (2004). Locality and the Left Periphery. In A. Belletti (Ed.), Structures
and Beyond. The Cartography of Syntactic Structures. New York: Oxford
University Press, pp. 223-251.
Rizzi, Luigi 2013. Notes on cartography and further explanation. Probus, 25(1)
pp.197-226.
Soares, Carla (2006). La syntaxe da le périhpérie gauche Portugais Européen et son

acquisition. PhD dissertation, Université Paris 8.
Szabolcsi, Anna (2006). Strong vs. weak islands. In Martin Everaert, & Henk van
Riemsdijk (Eds.), The Blackwell Companion to Syntax. Malden, MA:
Blackwell Publishing.
Vercauteren, Aleksandra (2010). Estruturas com "é que" em variedade não-standard
do PE. Textos selecionados do XXVI Encontro Nacional da APL. Lisboa:
APL, pp. 574-588.
THE EFFECTS OF LEXICAL FREQUENCY
ON OBJECT CLITIC PLACEMENT
IN EUROPEAN PORTUGUESE
HANNAH B. WASHINGTON
(The Ohio State University)
ABSTRACT: European Portuguese (EP) displays a general trend of postverbal pro-

nominal object placement, with preverbal placement licensed by proclisis triggers.
These triggers include contexts involving negation, subordinating conjunctions,
quantifiers, WH operators, and certain adverbs (Barrie, 2000; Cunha & Cintra,
2002; Perini, 2002; inter alia), although variation is still found in clitic placement
in these contexts (cf. Andrade, 2010a; Washington, 2012). The present study seeks to
evaluate whether variation in clitic placement in complex predicates–or restructur-
ing contexts–is correlated with verbal frequency. Frequency measures for the gov-
erning verbs are obtained from the Corpus do Português (Davies & Ferreira, 2006-
-) following three representative proclisis triggers and compared against the rate of
non-normative enclisis for each verb. The analysis demonstrates that lexical fre-
quency of the verbal host plays a crucial role in the synchronic trend toward gener-
alized enclisis in these environments, providing evidence for analogical change
prompted by predicate complexity and verbal frequency.
KEYWORDS: European Portuguese; clitics; frequency; variation; gradience
1. Introduction
In European Portuguese (EP), pronominal clitic objects display a general
trend of postverbal–or enclitic–placement, with proclisis licensed by specific
triggers.1 These proclisis triggers have traditionally been characterized as
giving rise to obligatory proclisis, and included in the set of triggers are ne-
gation, subordinating conjunctions, quantifiers, WH operators, and certain
1
Many thanks to the audiences of the inaugural meeting of the PLUS conference and NWAV
43, as well as to Scott Schwenter, Andrea Sims, and the anonymous reviewers of this vol-
ume for their invaluable feedback. All remaining shortcomings are strictly my own.
158 Hannah B. Washington
adverbs (Barrie, 2000; Cunha & Cintra, 2002; Perini, 2002; Galves & Sanda-
lo, 2012; inter alia). However, in these so-called obligatory proclisis con-
texts, a great deal of variation is still found in clitic placement. Building on
recent variationist work (Andrade, 2010a; Washington, 2012) that has un-
covered previously undocumented contexts that allow for non-normative
clitic in EP, the present study shows that prior formal analyses alone are
inadequate to explain the variation in usage, which both Andrade (2010a)
and Washington (2012) found to be present across a wide variety of linguis-
tic registers. In the present paper, I offer a new analysis that focuses on clitic
placement in complex predicates–that is, restructuring environments involv-
ing two verbs (cf. Andrade, 2010a)–which differ in terms of the degree of
grammaticalization of the governing verb as either auxiliary-like or semanti-
cally additive (as suggested by Washington, 2012). The present study care-
fully considers the variation as a result of the lexical frequency of the finite
or governing verb. Specifically, it will be argued that the lexical frequency
of the verbal hosts plays a crucial role in the current synchronic trend toward
generalized enclisis in the presence of proclisis triggers.
2. Background
2.1. Clitics in Romance and EP
The pronominal object clitic system in Romance languages has been de-
scribed as more affix-like than clitic systems in other languages. One reason
for this description derives from the fact that pronominal object clitics in
Romance do not have the freedom of host selection and must surface directly
adjacent to verbal hosts (Spencer & Luís, 2012). Generally speaking, pro-
nominal clitic objects in Romance languages (e.g. Spanish, Italian) are
placed before finite verb forms, and they can be placed postverbally in the
presence of nonfinite and command forms (ibid.). In consideration of these
facts, the synchronic tendency across Western Romance varieties is clearly
one of generalized proclisis of object clitics. In Medieval varieties of these
languages, however, we find a pattern of second-position (Wackernagel)
clitics, such that the clitics occupy the second position in the phrase with the
associated verb typically in first or third position. This second-position
placement holds in both main and subordinate clauses, with clitics generally
appearing immediately prior to the verb form unless the verbal constituent
came sentence-initially. In the context of a verb-initial clause, clitics tended
to be placed immediately following the verb (Spencer & Luís, 2012).
European Portuguese, however, diverges from other Romance varieties
with respect to clitic objects. While objects in Brazilian Portuguese (BP) dis-
play generalized preverbal clitic placement like other Romance varieties, EP
clitics are typically postverbal units. Unlike Spanish and Italian, BP strongly
The effects of lexical frequency on object clitic placement in European Portuguese 159
prefers clitic placement prior to the lexical or main verb in complex predicates,
as in (2b) below. The examples in (1) show the typical placement found in
simple and complex predicates in EP, with enclisis to the main or lexical verb
as the norm. The situation shown in (1c) reflects what has been described as
‘clitic climbing’ in EP (cf. Andrade, 2010a). The corresponding examples in
(2) show the general placement patterns for simple and complex predicates in
BP, both of which reflect proclisis to the lexical verb.
(1) a. Eu vejo-te.
1sg see.1s-CL.2sg.acc
‘I see you.’ (EP)
b. Eu vou ver-te.
1sg go.1sg to see.inf-CL.2sg.acc
‘I am going to see you.’ (EP)
c. Eu vou-te ver.
1sg go.1sg-CL.2sg.acc see.inf
‘I am going to see you.’ (EP)
(2) a. Eu te vejo.
1sg CL.2sg.acc-see.1s
‘I see you.’ (BP)
b. Eu vou te ver.
1sg go.1sg CL.2sg.acc-see.inf
‘I am going to see you.’ (BP)
These patterns in BP, which are also found in French, are quite different
from the available options in EP, which itself differs from the rest of the
Romance family due to its generalized pattern of enclisis in main clauses.
However, BP is thought to be relatively stable in its preference for proclisis
to the lexical verb throughout the grammar, while EP places clitic objects in
different positions based on the morphological and syntactic conditions of
the phrase, as outlined below. That is, in EP certain verbal forms diverge
from the typical placement patterns, such that object pronouns are placed
between the stem and the inflectional morphology for person/number and
tense/aspect. Consider the following:
(3) Eu dar-lhe-ei uma prenda.

1sg give-CL.3sg.dat-fut.1sg a.fem present
‘I will give her a present.’
(4) Eu dar-lhe-ia uma prenda.

1sg give-CL.3sg.dat-cond.1sg a.fem present
‘I would give her a present.’
The placement of clitics in (3) and (4) exemplify the patterns observed
with future or conditional stems and the inflectional affix for person and
number of the subject and tense and aspect. The placement of clitics between
the verbal stem and the inflectional affix, along with various other facts of
the EP clitic system, has led some scholars to conclude that EP has a mixed
clitic/affix system. Specifically, morphophonological alternations occur
across the boundary between the stem and the enclitic in EP (e.g. ver + o >
vê-lo ‘to see him’), while proclitics in contrast with enclitics can have scope
over coordinating clauses. With these facts in mind, numerous authors have
argued that pronominal object clitics function as affixes when attached as
enclitics to the right edge of the verbal host, as in (3) and (4), and as clitics
when attached as proclitics to the left edge of the verb phrase (cf. Luís &
Otoguro, 2004; Galves & Sandalo, 2012; Spencer & Luís, 2012). However, a
mixed system of this sort presents interesting theoretical problems for both
morphological and syntactic accounts for placement, and the differences in
behavior between proclitics and enclitics has led other authors to argue
against treating the pronominal object system as either only affixal or only
syntactic (e.g. Vigário, 1999).
The placement of object clitics in preverbal position in EP requires the
presence of certain trigger words. As described in §1, these triggers include
preverbal negation (não, nunca, ninguém, etc), subordinating conjunctions
(que), quantifiers and demonstratives (este, isto, muitos, etc.), WH operators
(quando, onde, quem, qual, etc.), and certain adverbs (talvez, mal, etc.) (Bar-
rie, 2000; Cunha & Cintra, 2002; Perini, 2002; inter alia). The examples
below show normative clitic placement (proclisis) following the subordinat-
ing conjunction que ‘that’ with simple (5a) and complex (5b) verb phrases.
(5) a. Ele disse que a menina me

3sg.masc.nom say.past.3sg that the.fem.sg girl CL.1sgacc
viu.
see.3sg.past
‘He said that the girl saw me.’
b. Ele disse que a menina me
3sg.masc.nom say.past.3sg that the.fem.sg girl CL.1sg
vai ver.
acc go.3sg see.inf
‘He said that the girl is going to see me.
While these contexts are often described as requiring proclisis, some prior
descriptions of object clitic placement in EP attest the variation in placement
patterns following these triggers (e.g. Cunha & Cintra, 2002). In the follow-
ing sections, I will summarize the relevant work on variation in clitic place-
ment within these contexts and discuss some analyses that attempt to unify
these so-called obligatory proclisis contexts. And although variation in

placement can be found in both simple and complex predicates, the present
study will focus exclusively on clitic placement following these proclisis
triggers in the presence of complex predicates (i.e. as seen in 5b).
2.2. Variationist approaches to clitic placement
Despite strong evidence for variation in pronominal object clitic place-
ment in EP, the majority of prior work on clitic placement in EP focuses on
other questions. These studies have often sought to address either: a) the
development of a unified analysis that accounts for proclisis and enclisis
placement rules within formal syntactic, morphological, or prosodically-
-based frameworks (Crysmann, 1997; Vigário & Frota, 1998; Luís & Sadler,
2003; Luís & Otoguro, 2004; Luís & Spencer, 2005; Luís, 2009a; Luís,
2009b; Luís & Otoguro, 2011; inter alia); or b) the diachronic tracing of
changes in the clitic placement rules in EP (Martins, 1993; Galves & Sanda-
lo, 2004; Galves et al., 2005; Andrade, 2010b). While many of the historical
studies consider proclisis and enclisis contexts together, most formal accounts
that separate the two contexts do not make mention of variation in proclisis
contexts. Recent variationist work, however, has uncovered some of the fac-
tors conditioning the variable placement of object clitics in EP.
Andrade (2010a) provides a preliminary analysis of the variation with re-
spect to complex predicates in syntactic restructuring environments, with a
focus on the effect of register. Using data from the Corpus do Português (45
million words, Davies & Ferreira, 2006-) and a selection of novels, he ex-
tracts 913 EP tokens and codes for register, adjacency of the two verbs in the
verbal sequence, clitic function, syntactic context (proclisis vs. enclisis), and
verbal frequency. He finds that “clitic climbing”2 is found more commonly
in informal registers than in formal ones, and that verbs with no intervening
elements show a higher rate of clitic climbing than when intervening ele-
ments are present. Of particular interest are Andrade’s findings with respect
to case and function–showing that clitic groups and datives undergo clitic
climbing more regularly than do accusative and reflexive clitics–and fre-
quency–such that more frequent verbs show a higher rate of clitic climbing
2
Andrade (2010a) looks specifically at clitic climbing–that is, the movement of pronominal
clitics leftward to a higher position in the syntax. Because clitic climbing in EP can result in
either proclisis (in triggered contexts) or enclisis to the auxiliary verb (in non-triggered con-
texts), Andrade (2010a) distinguishes between placement and position of the clitic. Howev-
er, since I am considering only triggered contexts involving two-verb predicates in this pa-
per, I will use placement and position interchangeably, with enclisis referring to placement
following the nonfinite verb (V V=CL) and proclisis referring to placement before the gov-
erning or finite verb (CL=V V).
than less frequent ones. The frequency effects reported by Andrade (2010a)
will be discussed in more detail in §3.
Washington (2012, 2015) builds on Andrade’s (2010a) work by present-
ing a variationist analysis that considers only environments following a) the
epistemic adverb talvez ‘perhaps’; b) the subordinate conjunction and rela-
tive pronoun que ‘that, which’; and c) the preverbal sentential negation não
‘not’. These three environments were chosen as representative sample of the
variety of triggers available (an adverb, a conjunction, and negation), and
they reflect differences in Vigário & Frota’s (1998) phonological schemati-
zation of Type I (talvez, não) and Type II (que) proclisis triggers available in
EP, which have been posited to differ in the admissibility of the change in
progress toward generalized enclisis. All three environments under consider-
ation are described as indisputably and normatively triggering proclisis by
grammarians and linguists alike (Cunha & Cintra, 2002; Perini, 2002; Mar-
tins, 1993; Barrie, 2000; Luís & Otoguro, 2004; Galves & Sandalo, 2012;
among others). With a total of 1554 tokens of personal object pronouns (me,
te, nos, lhe, lhes) extracted from the European Portuguese portion of the
Corpus do Português (45 million words, Davies & Ferreira, 2006-), Wash-
ington (2012, 2015) employs a mixed effects logistic regression model to
determine which factors are correlated with non-normative clitic placement
following the three proclisis triggers. She finds that the person/number refer-
ence of the subject, the object clitic form, the mood of the verb, the style or
register of the source text, and the complexity of the verb phrase or verbal
construction [V or V V sequence] all play important roles in non-normative
clitic placement. Like Andrade’s (2010a) findings, Washington’s (2012) data
show variation across all registers of language use available in the corpus,
ranging from informal sociolinguistic interviews to literature and journalistic
writings to scholarly articles. The disparity in their results are indicative of
differences in methodologies between the two studies: Andrade (2010a) does
not limit token extraction to normative proclisis contexts, in contrast with
Washington’s (2012) highly restricted contexts involving only normative
proclisis environments. Although Washington (2012) finds significant dif-
ferences in rates of non-normative clitic placement between registers, the
fact that variation in placement occurs in all registers suggests that this varia-
tion cannot adequately be explained as performance errors (cf. Galves &
Sandalo, 2012). Rather, the use of postverbal clitics following proclisis trig-
gers must indeed be part of language users’ grammatical competence.
Washington’s (2012, 2015) results point to a crucial correlation between
verbal constructions–that is, multi-verb verb phrases (V V sequences)–and
non-normative clitic placement. Specifically, the author finds that simple
predicates, as well as two-verb predicates containing perfect (ter/haver +
past participle ‘to have Xed’) and progressive (estar a + infinitive ‘to be
Xing’) forms, favor normative proclitic placement in these obligatory pro-

clisis contexts. These patterns follow the trend seen for simple predicates:
when only one verb is available in the predicate, a total of 7% of the clitic
tokens are found post-verbally in the presence of one of the proclisis triggers
(que, não, talvez). Similarly, with the perfect and progressive verb sequenc-
es, 12% enclisis is observed. For these more cohesive verbal sequences with
progressive and perfect meanings, the low rate of enclisis in triggered con-
texts reflects lower than average rates of this non-normative pattern than are
found in the data at large. Other kinds of two-verb predicates, however,
show a strong tendency to attach pronominal object clitics to the right edge
of the second verb. The examples below are representative of Washington’s
(2012) findings:
(6) Para que é que me estão a filmar?

for what is.it that CL.1sg.acc be.3pl.pres prog film.inf
‘What are you all filming me for?’
[RUSSIA:971290]
(7) Fiquei contente ao saber que podes visitar-nos

get.1sg.past happy to find.out.inf that can.2sg visit.inf-CL.1pl.acc
nesta primavera.
in.this spring
‘I was happy to discover that you can visit us this spring.’
[TrdeCorr:Viegas]
As explained above, the progressive construction tends to follow more

normative proclisis trends following the trigger word que, as seen in (6). In
contrast, modal constructions involving verbs like poder ‘can’ and dever
‘should’ are more likely to allow for non-normative enclitic placement,
which is what we find in (7). Washington (2012) hypothesizes that such
trends reflect the importance of the degree of grammaticalization of verbal
constructions for clitic placement, given the breakdown of the data presented
in Table 1:
First verb in sequence Proclitic Enclitic

Resolver ‘to resolve/decide to X’ 0/6 6/6 (100%)
Pretender ‘to attempt to X’ 0/6 6/6 (100%
Procurar ‘to seek to X’ 0/5 5/5 (100%)
Tentar ‘to try to X’ 0/6 6/6 (100%)
Conseguir ‘to manage to X’ 1/26 25/26 (96%)
Querer ‘to want to X’ 12/57 45/57 (79%)

Dever ‘should X’ 6/23 17/23 (74%)
Vir ‘to come to X’ 8/26 18/26 (69%)
Poder ‘to be able to X / can X’ 46/151 105/151 (69%)
Ir ‘to be going to X’ 22/48 26/48 (54%)
Estar a ‘to be Xing’ 11/13 2/13 (15%)
Table 1: Rates of pro- vs. enclitic placement in normatively proclitic contexts in two-verb
sequences using data from the Corpus do Português. Reproduced from Washington (2012,
2015). Only contexts with more than four token occurrences are included in the table.
Table 1 shows the rate of pre- and post-verbal clitic placement by the
first, or governing, verb in the verbal sequence. The 367 tokens presented
above reflect the majority of the 399 tokens involving multi-verb predicates
found in the dataset. The 32 tokens missing from Table 1 involve governing
verbs showing four or fewer tokens in the data. The remaining 1155 tokens
extracted from the corpus for Washington’s (2012, 2015) work are not re-
flected in the table because they involve a single verb or perfect construction
(ter/haver + past participle) following the trigger word.
Washington (2012) understands the ordering in Table 1 to be a case of
least to most “grammaticalized” verbs when coupled with an infinitival verb
form. In essence, the verbs closer to the bottom of the table provide less se-
mantic content to the verb phrase than those further up, and these more
grammaticalized units instead add to the temporal, aspectual, or modal con-
tent of the verb phrase. Grammaticalization, then, is understood in this con-
text as a reduction in constituent structure and the reclassification as fixed
expressions that have taken on a grammatical, morphological, or modal role,
having largely lost their semantically additive meanings. For example, estar
a + infinitive and ir + infinitive are used to indicate progressive aspect and
future tense, respectively, without also indicating action on the part of the
speaker beyond what is expressed by the infinitival verb that follows. Wash-
ington’s (2012) analysis depends on the assumption that more grammatical-
ized two-verb sequences like estar a + infinitive ‘to be Xing’ show rates of
enclisis that are slightly lower than the average found in the data at large,
while those that add greater semantic content to the proposition show much
higher rates of non-normative enclisis. Under this analysis, the adjacency (cf.
Torres Cacoullos & Walker, 2011) of these verbs with a following infinitive
indicates increased unithood over time, which would encourage the reanaly-
sis of the verb phrase and the subsequent usage of syntactic patterns found in
simple verb phrases containing a single verbal element.
As explained by Torres Cacoullos (2013), proclisis in the corresponding
estar + gerund constructions in Spanish indicates a high degree of cohesion
between the two verb forms. When proclisis is found in these sequences, the
verbal construction is understood to function more like a single verbal unit
rather than as two adjacent predicates, suggesting that the auxiliary has be-
come grammaticalized as a tense-mood-aspect marker. Accordingly, Wash-
ington’s (2012) assertion that higher rates of proclisis in EP in the presence
of complex predicates reflects a higher degree of grammaticalization will be
assessed in the present study through the consideration of frequency effects.
Higher frequency lexical items that have become associated with certain
types of units can result in more grammaticalized meanings and are also
typically more resistant to analogical changes (cf. Bybee, 2002). According-
ly, we would expect to see more normative clitic placement in the presence
of high frequency verbal constructions (i.e. V V[+infinitive] sequences) that re-
flect more grammaticalized or cohesive meanings if grammaticalization is
indeed the mechanism promulgating a change in progress in EP.
Schwenter & Torres Cacoullos (2014) present similar clitic placement
patterns with respect to grammaticalized constructions in two-verb sequenc-
es in Mexican Spanish. In this language, both proclisis to the governing verb
and enclisis to the lexical verb are acceptable regardless of preverbal con-
stituents. Their data show that constructions like estar + gerund ‘to be Xing’
(progressive) and ir a + infinitive ‘to be going to X’ (future) show much
lower rates of enclisis (18% and 9%, respectively) than do more modal se-
quences like poder + infinitive ‘to be able to X’ and querer + infinitive ‘to
want to X’ (26% and 44% respectively). To account for these differences by
governing verb, these authors suggest the general pattern for clitic placement
in Mexican Spanish–that is, proclisis in simple predicates–is actually “dif-
fusing construction by construction” with respect to complex predicates
(Schwenter & Torres Cacoullos, 2014:528). Furthermore, they believe this
pattern to have begun with more grammaticalized, cohesive expressions,
spreading to other verb+infinitive constructions that contain a pronominal
clitic object through analogical change. This pattern of diffusion has its roots
in the placement patterns of simple predicates and spreads to the two-verb
constructions that function with greater unithood, and it only then finally
affects complex predicates that function as separate semantic units. The
similarities between the results in Washington (2012) and Schwenter &
Torres Cacoullos (2014) for two-verb constructions point to a cross-
-linguistic trend of pronominal object clitic movement propelled by gram-
maticalization effects.
3. Frequency effects
The study of frequency effects in phonetic and morphosyntactic variation
and change has been of recent interest to researchers (cf. Bybee, 2007;
Bybee, 2010; Bybee, 2011). It has been widely noted that phonological
changes are due to phonetic reduction in highly frequent tokens, as the result
of automation of articulation (Bayley, Greer & Holland, 2013; Bybee, 2002).
Under this view, forms that are produced more often offer more opportuni-
ties for reduction through the automation of articulatory pathways. In con-
trast, it has been argued that morphosyntactic changes are the result of ana-
logical change, which first affects low frequency forms (cf. Bybee, 2002).
Within Bybee’s (2002) model, high frequency morphosyntactic construc-
tions have a conserving effect against analogical leveling in the system due
to greater repetition and thus entrenchment of the forms.
Recent studies of morphosyntactic variation have sought to include fre-
quency information into variationist models. Erker & Guy (2012) and Bay-
ley, Greer & Holland (2013), for example, offer two such analyses of subject
personal pronoun expression for North American speakers of Spanish. These
studies use binary frequency measures (high vs. low, where high frequency
corresponds to forms that make up over 1% of the total data) to analyze the
effects of verbal frequency on subject expression, resulting in divergent out-
comes. While Erker & Guy (2012) find that high frequency verbs amplify
the independent effects of tense/mood/aspect and switch reference, Bayley,
Greer & Holland (2013) do not find significant differences in rates of subject
expression between frequent and infrequent verbs and instead find that ver-
bal frequency is better accounted for as an independent effect on the varia-
tion. Recent studies of this sort demonstrate the value of considering the
effects of lexical frequency on morphosyntactic variables.
With respect to clitic positioning, Davies’ (1997) and Andrade’s (2010a)
studies on variable clitic placement in complex predicates in diachronic and
synchronic Portuguese, respectively, take verbal frequency into account. In
these studies, the governing verb (i.e. the finite verb) in the complex predi-
cate is taken as the verb of interest with respect to frequency. Using binary
measures of frequency, they classify the verbs into two categories: frequent
and infrequent. Davies (1997) takes a sample of eight verbs in his corpus,
and Andrade (2010a) looks a larger number of verbs. Both studies find that
more frequent verbs show a higher rate of clitic climbing, while less frequent
verbs show a lower rate, meaning that the clitics tend to surface in postverbal
position with respect to the main verb more often in the presence of low
frequency governing verbs.3
3
As mentioned previously, clitic climbing can also mean the enclitic or postverbal placement
of the clitic object with respect to the governing verb. In the present study, postverbal or
enclitic placement refers to attachment to the right edge of the lexical or main verb.
Because neither of these studies accounts for the preverbal placement in-
duced by proclisis triggers, thus potentially confounding the results due to
novel distributions in the data, the present study considers only contexts in
which such triggers are present. Furthermore, frequency measures in the
current study will be analyzed as a continuous factor that ranges from 0 to
the maximum rate of occurrence per million words in the corpus rather than
as a binary factor in order to capture gradient patterns. That is, Andrade
(2010a) and Davies (1997) analyze frequency measures as binary measures
either within the corpus or within their particular datasets, defining each verb
as either high frequency or low frequency. In contrast, the present study con-
siders individual governing verbs, which each reflect a single number for the
frequency per million words in the corpus. This numeric frequency measure
is then matched with the rate of non-normative enclisis for a given governing
verb. Accordingly, the frequency measures are not subjected to arbitrary
divisions or binning of perceived high and low frequency, and the frequency
measures reflect the rate at which each verb is found in the corpus from
which tokens were originally extracted, regardless of any coding decisions
that might have influenced the total number of tokens extracted involving
any given governing verb.
4. Methodology
I examine frequency effects in clitic placement in the complex predicates
that Washington (2012) considered for EP, presented above in Table 1, in
the presence of the three proclisis triggers que, talvez, and não. As previous-
ly mentioned, these three triggers are taken as representative of Vigário &
Frota’s (1998) division between Type I and Type II triggers, with phonolog-
ical differences found between the three. Talvez displays the lowest rate of
non-normative enclisis, while que and não are quite similar in their rates.
Considering the similarities to Mexican Spanish (Schwenter & Torres
Cacoullos, 2014) with respect to verb sequences that have higher rates of the
less preferred enclitic placement, the question of verbal frequency remains
highly relevant for both Spanish and EP. That is, could the generalization of
enclisis in EP (as put forth by Vigário & Frota, 1998) be spreading in pro-
clisis contexts based on verbal frequency? And, is verbal frequency the cor-
rect operationalized measure for Schwenter & Torres Cacoullos’ (2014) and
Washington’s (2012) claims that placement patterns are (at least partially)
based on degree of grammaticalization?
To address these questions, frequency counts for each of the verbs pre-
sented in Table 1 are obtained from the Corpus do Português (Davies &
Ferreira, 2006-). Verbal frequencies from only the European Portuguese
portion of the corpus are extracted, thus providing faithful frequencies that
match not only the corpus but also the specific language variety within the
corpus from which Washington’s (2012) data were extracted. Two types of
frequency measures are considered:
(8) Frequency Measures

a. General verbal frequency of the governing verb: This measure is deter-
mined by considering the number of occurrences of a single verb per
million words in the corpus. Essentially, this is a frequency count of a
given verb in all of its potential forms (token frequency across all forms).
b. Frequency of governing verb in V + infinitive sequences: This measure
is determined by considering the occurrences of a given verb followed
by another verb (in infinitival form) per million words in the corpus.
Like the general frequency measure above, this includes all potential
person/number/tense/aspect inflections of the verb in question, followed
by any infinitive. For verbs like estar that are always followed by a be-
fore an infinitive, frequency is determined by including the required
preposition.
These measures are used to evaluate three specific questions related to

frequency: 1) Are there patterned frequency differences between verbs that
show higher and lower rates of non-normative clitic placement?, 2) If so,
what do these differences look like?, and 3) Do verbs that are highly fre-
quent as a sort of auxiliary show pronominal clitic placement patterns that
differ from those verbs that are less commonly found in two-verb sequences?
Although the type frequency of the infinitival verbs following each of the
verbs in Table 1 is understood to be a reflex of the degree of entrenchment or
fixedness of the construction, this measure has been left for future considera-
tion. The following section provides a summary of the results.
5. Results & discussion

For the sake of convenience, a revised version of Table 1 has been inserted
below with added frequency information and revised organization based on
the frequency measures as they relate to the rate of enclisis. The frequency
information presented in Table 2 shows a number of general trends with
respect to the rate of enclisis in contexts with proclisis triggers. For starters,
we see that the general frequency of the verbs in question increases as the
rate of enclisis decreases. Accordingly, less frequent verbs tend to disfavor
normative behavior, or in other words, they tend to show more enclisis in the
context of proclisis triggers than more frequent verbs. The negative correla-
tion between non-normative behavior and frequency is shown in Figure 1
below, thus illustrating a positive correlation between frequency and norma-
tive behavior: as the governing verb’s frequency increases, so does the rate
of normative object clitic placement in these proclisis-triggered context.
General Frequency
frequency of governing
First verb in Procli-
Enclitic (per mil- verb+ infinitive
sequence tic
lion (per million
words) words)
Resolver ‘to 0/6 6/6 (100%) 170.92 31.42
resolve/
decide to X’
Pretender ‘to 0/6 6/6 (100%) 183.25 119.62
attempt to X’
Group I Procurar ‘to 0/5 5/5 (100%) 258.14 94.27
seek to X’
Tentar ‘to 0/6 6/6 (100%) 270.67 194.04
try to X’
Conseguir ‘to 1/26 25/26 (96%) 549.46 297.59
manage to X’
Querer ‘to want 12/57 45/57 (79%) 1402.66 699.13
to X’
Dever ‘should 6/23 17/23 (74%) 1072.97 659.48
X’
Vir ‘to come 8/26 18/26 (69%) 1327.68 185.89
Group II
to X’
Poder ‘to be able 46/151 105/151 2992.59 2275.45
to / can X’ (69%)
Ir ‘to be going 22/48 26/48 (54%) 2854.18 1135.33
to X’
Estar a ‘to be 11/13 2/13 (15%) 4289.53 605.55
Group III
Xing’
Table 2: Rates of pro- vs. enclitic placement in complex predicates in normatively proclitic
contexts (from Washington, [2012] and Washington, [2015]), with added frequency counts
for each verb per million words in the corpus.
Davies (1997) and Andrade (2010a) report similar findings in their studies of
clitic climbing in all (proclisis and enclisis) contexts: more clitic climbing–
typically resulting in proclisis (cf. [5b]), though sometimes involving enclisis
(cf. [1c])–is found with higher frequency governing verbs. However, both
authors miss the gradience of the pattern by classifying verbs categorically
as either high frequency or low frequency, based on the number of tokens in
the presence of each verb within the data set.
The results for frequency of the governing verb when followed by an infini-
tive show a similar pattern, albeit one that is a bit less clear. That is, the same
negative correlation between frequency and non-normative behavior can been

seen below in Figure 2. It must be highlighted, though, that general verbal fre-
quency, rather than frequency of the governing verb when followed by an infini-
tive, appears to be a better predictor for clitic placement in two-verb sequences.
Figure 1: Rate of enclisis following proclisis triggers que, talvez, and não by general verbal
frequency of the governing verb. All verbs from Table 2 are included, as well as others with
lower token counts. Multiple R-squared = 0.914.
Figure 2: Rate of enclisis by frequency of governing/auxiliary verb when followed by an

infinitive. Multiple R-squared = 0.3189.
As can be seen, the regression line is not as well fitted to the

verb+infinitive frequency-to-rate of enclisis data. The R-squared value of
0.914 in Figure 1 indicates that a very large portion of the variation in rate of
enclisis is explained by verbal frequency. In contrast, the R-squared value of
0.3189 for Figure 2 is not as well fit, with less of the variation explained by
this frequency measure.
These two figures suggest a number of patterns related to verbal frequen-

cy and clitic placement. Since the estar a form can be interpreted as the in-
flectional information for the lexical verb, the resulting syntax is more likely
to treat this construction as a single unit in the lexicon, without conflict be-
tween the prosodic and syntactic constraints on placement. Whereas estar a
+ infinitive truly functions like a simple predicate, all of the other verbs dis-
play much higher rates of enclisis than expected. The hypothesized change
toward generalized enclisis, as suggested by Vigário & Frota (1998), has had
a more pronounced effect on the verbs in Group I, as well as those in Group
II that are lower in both frequency measures. In essence, dever and querer,
which have reached 74% and 79% enclisis following proclisis triggers and
which show generally lower verbal frequency than other verbs in Group II,
are likely to result in complete generalization of enclisis before the other
verbs in Group II (i.e. poder and ir, which are more frequent and show
somewhat lower rates of enclisis).
The results presented above suggest that frequency is a factor in clitic
placement in two-verb sequences in EP. While the most frequent and highly
grammaticalized form estar a largely maintains normative placement pat-
terns in the presence of proclisis triggers, verbs with lower frequency that are
also generally less grammaticalized diverge considerably. These results pro-
vide a motivation for previously posited theoretical approaches to clitic
placement in EP. Specifically, Vigário & Frota (1998) and Galves & Sandalo
(2012) both argue that the trend in EP is toward a less phonologically- or
prosodically-dependent system of placement. Although Galves & Sandalo
(2012) indicate that the overgeneralization of enclisis results from speech
errors by younger speakers, their general argument for a morphologized
pronominal object clitic system is in line with Vigário & Frota’s analysis.
The results presented here offer evidence for the generalization of enclisis
following proclisis triggers in the presence of complex predicates, with low-
er frequency finite verbs leading the change.
Speakers’ experience with respect to less frequent verbs that have associ-
ated object pronouns is much more limited, thus providing the necessary
ingredients for analogical, generalized linguistic behavior (cf. Bybee, 2010).
This pattern is thus in line with established theories about analogical change
in morphosyntactic constructions, which affects low frequency forms first
(Bybee, 2002). If indeed proclisis in the presence of proclisis triggers is a
remnant of a system that maintained dual phonological and syntactic re-
strictions on placement, as suggested by Vigário & Frota (1998), it is not
surprising that proclisis is more common in the presence of frequent forms
which share the necessary conditions to maintain non-conforming, unpro-
ductive patterns. Such trends point to the interconnectedness of frequency
and grammaticalization, such that grammaticalization occurs due to the high
frequency and adjacency of certain verbs followed by nonfinite forms, while

a construction functioning as a single unit or as “chunked” information (cf.
Bybee, 2010) may appear in less productive morphosyntactic configurations.
In essence, analogical change occurs when low frequency forms are affected
first and show a tendency toward preferring the stronger patterns in the lan-
guage, while grammaticalization occurs when high frequency forms are af-
fected first (cf. Bybee, 2002). The data provided in this paper thus points
toward analogical change rather than grammaticalization with respect to the
placement of clitics in proclisis-triggered contexts in EP. More grammatical-
ized verbal constructions, in contrast, are the most robustly resistant to the
stabilization of generalized enclisis in EP, maintaining a conservative irregu-
larity in the syntax.
Furthermore, in complex predicates, the conflict arising from the dual re-
quirement for clitic hosting is naturally present in all cases. That is, the clitics
can either lean on the trigger or be adjacent to the verbal host with which they
associate semantically. In these cases, the tendency toward resolution is, as
Vigário & Frota (1998) assert, in favor of the syntax. The exception is with
estar a + infinitive, for which estar is functioning simply as an auxiliary to
provide tense and aspect information for the lexical verb. Here, the conflict is
less likely to require a resolution resulting in enclisis, since the estar form can
be interpreted as the inflectional information for the lexical verb. Because the
presence of proclisis triggers que, não, and talvez largely lead to proclisis in
simple verb phrases involving a single verb (7% enclisis in triggered contexts,
as reported by Washington [2012]), the high rate of this proclisis placement in
complex verb phrases containing highly frequent governing verbs suggests an
approximation of the normative pattern.4 As Bybee (2010:71) explains with
respect to frequency, “even after a construction has lost its productivity, spe-
cific exemplars of the construction may live on because they have accrued
strength through repetition and so continue to be used”. Estar a + infinitive
could be one such outlying exemplar, with other high frequency forms also
clinging to a lesser extent to the older construction. Such an analysis–with
4
An anonymous reviewer questioned whether estar a CL=V[+infinitive] is possible in EP, with
reference to examples (3) and (4) in which the clitic is placed between the verbal host and
the inflectional affixes, given that the proposal presented in the text suggests that estar a
functions more as inflectional information than as semantically additive content. To my
knowledge, this form is not available in EP, due to the fact that EP strongly prefers enclitics
(estar=CL a + V or estar a + V=CL) in unmarked contexts, although some variation is
found following prepositions and other trigger-like particles (e.g. ter que CL=V) in other-
wise untriggered contexts. While clitic climbing to a postverbal position, such as in es-
tar=CL a + V, is possible in the triggered contexts included in this investigation, it is quite
uncommon. In general, the triggered contexts show clitics that have ‘climbed’ to absolute
preverbal position (trigger CL=V V) or clitics that remain attached postverbally to their lex-
ical host (trigger V V=CL).
frequency as a motivating factor for non-normative trends that become possi-

ble because of changes to the phonological and syntactic restrictions in gram-
mar–is suggestive of the importance of lexical frequency in usage patterns as
changes occur at the phonology-syntax interface.
The results presented above reflect gradience in the synchronic grammar
based on verbal frequency. Prior studies on variation in clitic climbing (Da-
vies, 1997; Andrade, 2010a) and clitic placement (Washington, 2012) fail to
capture this fact by either classifying verbs into binary frequency categories or
by excluding frequency information altogether. Davies (1997) and Andrade
(2010a) present preliminary findings for the effect of frequency on the varia-
tion: in their studies, high frequency verbs show greater rates of clitic climb-
ing. By limiting the study of frequency effects to only proclisis contexts and
by considering frequency as a continuous rather than categorical measure, the
present study has been able not only to illustrate the gradient nature of the
variation in clitic placement in complex predicates but also to better motivate
the trend toward generalized enclisis following proclisis triggers.
This paper has provided evidence for frequency effects related to modern
pronominal object clitic placement in the presence of two-verb sequences in
EP. If the “regression of proclisis” is indeed related to changes in the re-
quirements of clitics with relation to their hosts, and phonologically weaker
triggers are leading the change, as Vigário & Frota (1998) suggest, then this
frequency account offers predictions about which verbal (or syntactic) hosts
will be found to lead the change toward generalized enclisis in all environ-
ments. The analysis also presents some interesting questions about cross-
-linguistic variation, namely the issue of pronominal clitic object placement
in Spanish complex predicates. In spite of Washington’s (2012) findings
showing similarities between Mexican Spanish and EP clitic placement in
two-verb sequences, it appears that the movement of the patterns with re-
spect to verbal frequency reflects divergent trends. As suggested here, EP is
moving toward generalized enclisis, and this trend includes contexts with
proclisis triggers. In contrast, Mexican Spanish–and indeed Spanish at large
(cf. RAE, 2009-2011) – is a generalized proclisis language, with required
proclisis in the presence of simple finite predicates. Similarities in the per-
centages of enclisis between the two languages led Washington (2012) to
conclude that verbal restrictions on clitic placement are a function of the
degree of grammaticalization of the verbal construction. However, if we
think of the similarities in enclisis rates (i.e. with querer and poder construc-
tions having higher rates of enclisis than estar and ir constructions in both
languages) as a reflection of similarities in relative frequency cross-
-linguistically, then it appears that related but divergent changes are in pro-
gress. While higher frequency constructions in European Portuguese have a
tendency to hang on to “remnant”, unproductive placement patterns follow-
ing certain environmental restrictions (proclisis), the same higher frequency
constructions in Mexican Spanish result in the more generalized pattern
found consistently throughout the system (proclisis). Interestingly, the diver-
gent directions of the frequency effect in the two languages have the same
functional result of preverbal placement of the pronominal object clitics. But
the relationship between Mexican Spanish and EP provides evidence for
Bybee’s (2010:75) assertion that “high-frequency forms are less likely to
undergo analogical change than low-frequency forms”. That is, the Spanish
patterns seem to show analogical behavior at a higher rate in the presence of
high frequency forms. However, if we account for behavior related to ad-
vanced unithood or increased grammaticalization in the Spanish data (cf.
Torres Cacoullos, 2013), we can see that the divergent patterns are likely
resulting from different processes. Whereas proclisis environments with
complex predicates in EP are moving toward enclisis through analogical
change disseminating through lower frequency verbs, Mexican Spanish is
showing an increase in proclisis as complex predicates become more unit-
-like due to grammaticalization processes.
The frequency findings of this paper leave a number of open questions.
First and foremost, the verbal constructions from Schwenter & Torres
Cacoullos’ (2014) Mexican Spanish data should be analyzed for frequency
to confirm the parallels drawn here regarding cross-linguistic similarities in
frequency of various two-verb predicates. Beyond this, the frequency of
verbal hosts in simple, single-verb predicates in EP should be analyzed for
frequency to determine whether the frequency effects reported in this study
are at work across all verbal hosts. In essence, Washington (2012) finds a
much higher rate of non-normative postverbal clitic placement in complex
predicates than in simple ones. Because the clitic in a complex predicate is
semantically associated with the second verb, the fact that these construc-
tions generally show higher rates of enclisis (to the second verb) is expected.
This is because the constraints in the syntax put pressure on the system to
place the clitic adjacent to its verbal host, which in these cases could be in-
terpreted as a semantic host. If indeed the frequency patterns found for com-
plex predicates reflect patterns seen for single-verb predicates, this would
provide strong evidence for a lexically-driven change in the morphosyntax
resulting from changes to the prosodic restrictions. These questions are left
for future study.
References
Andrade, Aroldo Leal de (2010a). The Application of Clitic Climbing in European
Portuguese and the Role of Register. In Selected Proceedings of the 12th His-
panic Linguistics Symposium. Somerville, MA: Cascadilla Proceedings Pro-
ject, pp. 97-108.
Andrade, Aroldo Leal de (2010b). A subida de clíticos em português: Um estudo
sobre a variedade europeia dos séculos XVI a XX. Doctoral dissertation,
Universidade Estadual de Campinas.
Barrie, Michael (2000). Clitic Placement and Verb Movement in European Portu-
guese. MA Thesis, University of Manitoba.
Bayley, Robert, Kristen Greer & Cory Holland (2013). Lexical frequency and syn-
tactic variation: A test of a linguistic hypothesis. U. Penn Working Papers in
Linguistics, Vol. 19.2, pp. 21-30.
Bybee, Joan (2002). Word frequency and context of use in the lexical diffusion of
phonetically conditioned sound change. Language Variation and Change 14
(3), pp. 261-290.
Bybee, Joan (2007). Frequency of use and the organization of language. Ox-
ford/New York: Oxford University Press.
Bybee, Joan (2010). Language, Usage and Cognition. New York: Cambridge Uni-
versity Press.
Bybee, Joan (2011). Usage-based theory and grammaticalization. In The Oxford
Handbook of Grammaticalization. New York: Oxford University Press, pp.
69-78.
Crysmann, Berthold (1997). Cliticization in European Portuguese using parallel
morpho-syntactic constraints. Proceedings of the LFG97 Conference. Stan-
ford: CSLI Publications, pp. 1-14.
Cunha, Celso & Lindley Cintra (2002). Nova Gramática do Português Contemporâ-
neo. Lisboa: Edições João Sá da Costa.
Davies, Mark (1997). A corpus-based approach to diachronic clitic climbing in Por-
tuguese. Hispanic Journal 17, pp. 93-111.
Davies, Mark. & Michael Ferreira (2006-). Corpus do Português: 45 million words,
1300s-1900s. Available online at http://www.corpusdoportugues.org.
Erker, Daniel & Gregory R. Guy (2012). The role of lexical frequency in syntactic
variability: Variable subject personal pronoun expression in Spanish. Lan-
guage 88, pp. 526-557.
Galves, Charlotte, Helena Britto & Maria Clara Paixão de Sousa (2005). The change
in clitic placement from classical to modern European Portuguese: Results
from the Tycho Brahe Corpus. Journal of Portuguese Linguistics 4 (1), pp.
39-67.
Galves, Charlotte & Filomena Sandalo (2004). Clitic-placement in modern and clas-
sical Portuguese. MIT Working Papers in Linguistics 47, pp. 115-128.
Galves, Charlotte & Filomena Sandalo (2012). From intonational phrase to syntactic
phase: The grammaticalization of enclisis in the history of Portuguese. Lin-
gua 122, pp. 952-974.
Luís, Ana R. (2009a). Para uma (re)definição da sufixação no Português Europeu: a

adjunção prosódica de enclíticos pronominais. Biblos: Revista da Faculdade
de Letras, Universidade de Coimbra, pp. 451-470.
Luís, Ana R. (2009b). Patterns of clitic placement: Evidence from 'mixed' clitic
systems. In New Challenges in Typology: Transcending the Borders and Re-
fining the Distinctions. Berlin: Walter de Gruyter, pp. 11-34.
Luís, Ana & Ryo Otoguro (2004). Proclitic contexts in European Portuguese and
their effect on clitic placement. Proceedings of the LFG04 Conference. Can-
terbury, England: CSLI Publications, pp. 335-352.
Luís, Ana R. & Louisa Sadler (2003). Object clitics and marked morphology. Em-
pirical Issues in Formal Syntax and Semantics 4, pp. 133-154.
Luís, Ana & Andrew Spencer (2005). A paradigm function account of ‘mesoclisis’
in European Portuguese. Yearbook of Morphology 2004, pp. 177-228.
Martins, Ana Maria (1993). Clitic placement from old to modern European Portu-
guese. Historical Linguistics, 1993: Selected Papers from the 11th Interna-
tional Conference on Historical Linguistics, Los Angeles, California. Phila-
delphia: John Benjamins, pp. 295-308.
Perini, Mário A. (2002). Modern Portuguese: A Reference Grammar. New Haven:
Yale University Press. Real Academia Española (RAE): Nueva gramática de
la lengua española, 2 vols 2009-2011 (2009-2011). Madrid: Espasa Libros.
Schwenter, Scott A. & Rena Torres Cacoullos (2014). Competing factors on the
variable placement of direct object clitics in Mexico City Spanish. Revista
Española de Lingüística Aplicada 27 (2), pp. 514-536.
Spencer, Andrew & Ana R. Luís (2012). Clitics: An Introduction. New York: Cam-
bridge University Press.
Torres Cacoullos, Rena (2013). Gradual loss of analyzability: diachronic priming
effects. MS. Revised July 2013, Submitted to System, Usage, and Society,
FRIAS Workshop, Nov. 2011.
Torres Cacoullos, Rena & James A. Walker (2011). Collocations in grammaticaliza-
tion and variation. In The Oxford Handbook of Grammaticalization. New
York: Oxford University Press, pp. 225-238.
Vigário, Mariana (1999). Pronominal cliticization in European Portuguese: a post-
lexical operation. Catalan Working Papers in Linguistics, pp. 219-237.
Vigário, Mariana & Sónia Frota (1998). Between syntax and phonology: on phrasal
weight effects in European Portuguese. Unpublished manuscript, presented at
8th Colloquium on Generative Grammar, Pamela, Portugal.
Washington, Hannah B. (2012). Shifting tendencies: Variable object clitic placement
in European Portuguese. Paper presented at New Ways of Analyzing Varia-
tion (NWAV 41), Bloomington, Indiana.
Washington, Hannah B. (2015). Variable object clitic placement: Evidence from
European and Brazilian Portuguese. PhD dissertation, The Ohio State Uni-
versity.
PRAGMATICS AND INDEXICALITY:
THE CASE OF FAVELADO
HANNAH B. WASHINGTON
MARY ELIZABETH BEATON

ABSTRACT: In this paper, we utilize Gricean implicatures within the indexicality

framework to explain the differences between the synonymous terms favelado and
morador de favela ‘slum-dweller’ in Brazilian Portuguese. Using internet data, we
show that indexical meanings of favelado are generalized conversational implica-
tures (GCIs), while the use of morador de favela blocks those GCIs. We propose that
common indexes for favelado (such as traficante, pobre, ladrão) form a GCI cluster.
Thus, when part of the cluster is canceled, other meanings within the cluster of im-
plicatures are also canceled. In certain contexts, favelado has been reclaimed,
providing conditions for positively-valenced particularized conversational implica-
tures (PCIs). We argue that speakers manipulate implicatures to create and attempt
to destroy place-based ideologies. We conclude that discourse-dependent social
indexes–whether phonetic, morphosyntactic, or lexical–are in fact pragmatic devic-
es that fit within existing theoretical frameworks.
KEYWORDS: indexicality; pragmatics; conversational implicatures; Brazilian Por-

tuguese; favelado.
1. Introduction
Recent work on indexical relationships between linguistic forms and social
structures has illuminated the connection between social information and
speech patterns (Eckert, 2008; Beaton & Washington, 2014; Sinnott, 2013).1
Most previous studies on indexical relationships focus on speakers’ selection
1
We are grateful to Scott Schwenter, Sarah Sinnott, the audience of the inaugural PLUS
conference, and an anonymous reviewer for their insight and comments on earlier versions
of this work. Any remaining shortcomings are strictly our own.
178 Hannah B. Washington & Mary Elizabeth Beaton
of a specific phonological or morphosyntactic variant because of the associ-

ated social indexes; however, little work has been done on lexical items.
Beaton & Washington (2014) use indexicality to explain the differences in
meaning between a pejorized lexical item and its non-pejorized synonym,
showing that lexical items also index social meanings. In this work, they
consider favelado and morador de favela, both meaning ‘slum-dweller’ in
Brazilian Portuguese (BP). These items co-exist in semantic space–meaning
that they represent two ways of referring to the same entity–despite the ten-
dency for languages to avoid synonymy by enforcing lexical blocking effects
(Horn, 2007; Blutner, 1998; Clark, 1990; Kiparsky, 1982). In essence, if a
language develops two words that refer to the same entity or concept, the
meaning for one or both words will shift to avoid total synonymy. Beaton &
Washington (2014) follow Silverstein’s (2003) orders of indexicality and
utilize Eckert’s (2008) indexical field to model meanings associated with
favelado. In their analysis, morador de favela maintains a basic (Silver-
stein’s [2003] nth order) reading, whereas favelado indexes multiple social-
ly-determined (n+1 order) meanings. This analysis formalizes the dictionary
definition of favelado as the nth order meaning of ‘slum-dweller’, while
other socially layered meanings that develop from the dictionary definition
are reflected in the n+1 order meanings. These n+1 order meanings are re-
presented within a cloud of commonly associated ideas–that is, an indexical
field (Eckert, 2008). Morador de favela, in contrast, does not have socially-
charged n+1 order meanings in addition to its dictionary definition.
In this paper, we use the data and analysis in Beaton & Washington
(2014) as a foundation to show how lexical indexicality can be modeled
within a pragmatic framework. Specifically, we build on Sinnott’s (2013)
layered indexical model, in which pragmatically implicated meanings are
modeled within an indexical field. The research questions motivating our
study are the following: (1) What sorts of conventional and conversational
implicatures arise from morador de favela and favelado, and how do these
implicatures influence speakers’ decisions about which variant to use to refer
to someone who lives in the slum? (2) How does pragmatic theory align with
concepts from social order theories, such as indexicality? In other words,
how do the two overlap, and can they account for the same things? And, (3)
In what ways do conversationally implicated meanings (i.e. generalized con-
versational implicatures and particularized conversational implicatures, from
Grice, 1989[1975]) interact with Eckert’s (2008) concept of the indexical
field? Utilizing the notion of Gricean implicatures within the framework of
indexicality, we provide a model that bridges the semantic and social per-
spectives, thereby offering an account of what the two forms mean and how
speakers choose between them.
Pragmatics and indexicality 179
This paper is organized as follows: section 2 first explores the indexicali-

ty framework and its application to phonetic, morphsyntactic, and lexical
variables and concludes with an explanation of how indexicality can be un-
derstood within pragmatic theory. Section 3 presents our analysis of the
pragmatic devices that create and cancel the meanings found in the indexical
field, and we put forth our conclusions in section 4.
2. Background
2.1. Indexicality
Indexicality and the indexical field offer frameworks for the analysis of
clusters of social meanings inherent in linguistic forms. Indexicality was first
proposed for linguistic anthropology (Silverstein, 2003) and expanded for
phonetic (Eckert, 2008), morphosyntactic (Sinnott, 2013), and lexical (Bea-
ton & Washington, 2014) variation. The application of the theory of indexi-
cality to each of these cases of linguistic variation is explained in the sec-
tions that follow. We use this prior work on indexicality as a point of depar-
ture for the present pragmatic study, which adds a deeper understanding of
how speakers access and move between meanings within the indexical field.
The indexicality framework assigns a neutral base-level meaning to a lin-
guistic form, called nth order meaning, and maps macro social structures
onto it, forming n+1 order meanings. The n+1 meanings form a living struc-
ture and reinforce each other through each utterance in which the n+1 order
meanings are used. In this way, people create meaning in language, and lan-
guage creates meaning for society. As Ochs (1990: 287-8) explains, “lan-
guage must be studied not only as a symbolic system that encodes local so-
cial and cultural structures, but also as a tool for establishing (i.e. maintain-
ing, creating) social and psychological realities”. More succinctly, each in-
stantiation of a linguistic form that indexes a social meaning is necessarily
said in a context that differs slightly from the prior context. Thus, the repeti-
tion of a sound, word, or construction solidifies its social meaning and each
new context allows speakers to make new associations with the form.
2.2. Phonetic and morphosyntactic variation within indexicality
Eckert (2008) expands on Silverstein’s (2003) concept of meanings de-
rived from indexical orders by introducing the indexical field–a “constella-
tion” of meanings that are interrelated and can give rise to new meanings.
She creates three types of indexical meanings for /t/ release, i.e. hyper-
aspiration of /t/, in American English: social types, permanent qualities and
stances. Stances are moment-to-moment characterizations of the speaker. If a
speaker frequently displays a certain stance, she undergoes a process of
“stance accretion” (cf. Rauniomaa’s [2003] presentation, cited in Bucholtz &
Hall [2005]) whereby she becomes characterized as a person who possesses

certain permanent qualities (for example, a stance as ‘clear’ could become a
‘clear/educated person’). Social types represent types of people in the world,
e.g. a person who is temporarily clear or articulate may then be perceived as
educated (a permanent quality) and may therefore be perceived as a ‘school
teacher’ (a social type). Figure 1 below shows Eckert’s field for /t/ release.
Figure 1: Eckert’s (2008) indexical field for /t/ release. Boxes denote social types, bold de-
notes permanent qualities, and grey denotes stances.
Eckert proposes the indexical field for phonetic variables with little men-
tion of its use for morphosyntactic or lexical forms. Sinnott (2012, 2013)
broadens this work through her discussion of the indexicality of T/V address
forms in Spanish (tú, usted), extending the possibilities for indexical work to
morphosyntactic forms.
2.3. Lexical items in indexicality

Beaton & Washington (2014) work with Eckert’s (2008) framework to
introduce the indexical field as a tool to analyze a lexical item, favelado
‘slum-dweller’. It is important to note that different linguistic forms–
phonetic, morphosyntactic, and lexical–behave differently in terms of who or
what exactly the indexes point to. Phonetic variables index something about
the speaker of the utterance (e.g. a speaker with noticeable /t/ release can be
perceived as ‘emphatic’), whereas the morphosyntactic variable of T/V ad-
dress form usage in Spanish studied by Sinnott (2012, 2013) points to the
relationship between the speaker and the hearer. Lexical items do not neces-
sarily index something about the speaker or hearer directly since they are
used to name conversation-external things or concepts in the world. In fact,
Beaton & Washington (2014) argue that the lexical item favelado points
externally to an ideologized concept of place and of the people that inhabit

the favela.
Figure 2: Indexical field for favelado from Beaton & Washington (2014). Lower case =
permanent qualities; upper case = social types; dotted line separates the positive
reappropriated meanings from the negative qualities and social types.2
The indexical field for favelado proposed by Beaton & Washington (2014)
includes permanent qualities and social types, much like Eckert’s (2008) in-
dexical field for phonetic variables. Since lexical items usually point to the
external world, these authors do not include stances as part of the field. Ano-
ther factor that comes into play with a socially-charged lexical item like favelado
is the polarization of meanings. Favelado functions much like queer, dyke, gay
(McConnell-Ginet, 2001), nigger (Croom, 2011), and the Cantonese word
tongzhi ‘comrade’ (Wong, 2005; Wong, 2008), in that the reclaiming of a
negatively-valenced term has occurred in a particular context. Figure 2 above
shows Beaton & Washington’s (2014) indexical field.
Beaton & Washington (2014) suggest that the modern-day usage of fave-
lado reflects a grammaticalization process by which a noun comes to be used
as an adjective and then as an adverb. Through this process, the characteriza-
tion of inanimate objects as favelado becomes possible. The internet abounds
with examples of adjectival uses of favelado, such as in the following exam-
ple captioning a picture of a newly constructed mansion:
(1) O chão tava muito favelado. A construção tava sendo finalizada.

‘The ground was really favelado. The construction was being finalized.’
[http://www.fotolog.com/annemaniacs/62916800]
2
In the indexical field, the social type funkeiro refers to people who listen to funk, which is a
popular style of music in Brazil.
The picture being described resembles the slums in only one way: the
ground was not landscaped, suggestive of the dirt floors common in favelas.
Favelado in this context cannot index meanings associated with people, such
as illiterate, violent, etc.
Another example is found in descriptions of new and expensive athletic
shoes or certain kinds of name-brand clothing3, which have been described
as favelado through association with a perceived sense of the style prefer-
ence of slum-dwellers. The mansion and expensive clothing obviously have
little to do with the favela itself, and the use of favelado as an adjective ab-
stracts over qualities associated with slums and slum-dwellers.
2.4. Indexicality as pragmatics
In his work on lexical pragmatics, Blutner (1998:9) defines systematic
polysemy as “the phenomenon that [sic] one lexical unit may be associated
with a whole range of senses, which are related to each other in a systematic
way”. This concept of systematic polysemy, whereby meanings are intimate-
ly and systematically linked to each other, is delineated in the sociolinguistic
literature by movement between meanings, which Eckert (2008) describes as
“side-stepping”. Side-stepping is the process by which interlocutors shift
between indexical meanings, allowing for the creation of new meanings
through close association with existing indexes. In indexicality, contextual
information helps a hearer locate meanings within the indexical field. For
instance, the interactional context helps a hearer determine the social mean-
ing of /t/ release or of favelado.
One fundamental device for discourse pragmatic enrichment is the Grice-
an notion of implicature (Grice, 1989[1975]), which has been split into two
primary divisions of implicated meanings: conventional implicatures and
conversational implicatures. If a linguistic form has a conventional implica-
ture, the meaning it conveys is inseparable from the form that expresses it
and is not cancelable, meaning that these implicatures arise invariably.
Meanwhile, an important test for conversational implicatures is whether the
implicature is cancelable by linguistic or nonlinguistic material within the
context: conversational implicatures are necessarily cancelable, meaning that
these implicatures can be explicitly denied. For example, to say ‘I am not
unhappy’ implicates that the speaker is less than happy. However, this impli-
cature can be canceled if the speaker states, ‘I am not unhappy–actually, I’m
quite happy!’ These implicatures can be either generalized (GCI) or particu-
3
For example, speakers turn to Yahoo Answers to help them determine whether certain kinds
of clothing are favelado: http://br.answers.yahoo.com/question/index?
qid=20111030120034AAkhv9C and http://br.answers.yahoo.com/question/index?
qid=20120915085932AA8vZGO
larized (PCI), with the primary difference between the two being that GCIs
arise unless blocked by contextual information, while PCIs depend heavily
on the context. For example, in the sentence ‘If it starts to rain, I’m going
inside’, the GCI is that if it does not start to rain, the speaker will not go
inside. In this example, a PCI that can arise is that the speaker cannot get
wet, if the context allows for it due to specific conditions that are mutually
understood between the speaker and the hearer (e.g. the speaker is wearing
water-soluble hair dye).
A recent study has modeled an indexicality-pragmatics interface with ref-
erence to morphosyntactic variables. In this work, Sinnott (2013) uses prag-
matic notions to construct an indexical model for the use of T/V address
forms in Madrid Spanish (in essence, tú vs. usted). She employs a layered
model, reproduced below, to illustrate the relation between conventionally
and conversationally-implicated meanings.
Figure 3: Sinnott’s (2013) layered model of the indexical field.
In her model, the innermost layer contains the factors that a speaker con-
siders when deciding between the use of one form over the other. That is,
this layer reflects the most basic levels of meaning related to the hearer (age,
sex, profession) and the situation (relation, location); these foundational
pieces of information point to “membership in a population” (Eckert, 2008)
and thus make up the nth order meaning. The selection of T/V then reflects
the relationship between the speaker’s and the hearer’s membership within
each of the categories, with reference to the culturally-determined and con-
ventionally-implicated meaning of social distance between speakers. The
second and third layers in the field, represented as the outer layers, reflect
the n+1 order meanings, or the social structures that are activated by usage.
When expressed by use of the V form, distance activates certain indexes in
the second layer, e.g. formality and respect. The third layer, in contrast, is
used when speakers choose an address form against social norms, e.g. in
disrespecting an elder. In short, the first layer is indicative of conventionally-
implicated meanings, while the indexes in the second and third layers reflect
conversational implicatures that are contextually created and canceled to
accomplish pragmatic goals. In the second layer, we find the generalized
conversational implicatures (GCIs), whose meanings arise unless blocked by
contextual information. This contrasts with the social meanings in the third
layer, for which more calculation and contextual information is required to
access these indexes, or PCIs. Accordingly, access to certain kinds of mean-
ings depends largely on contextually available information. Sinnott’s (2013)
work broke ground in understanding different indexical values as calculable
pragmatic phenomena, and we build on her approach in the section that fol-
lows.
3. Analysis
3.1. Favelado vs. morador de favela: Testing for implicatures
The data for this study come from online discussions about slums and
what it means to be a slum-dweller (morador de favela/favelado) and online
commentary characterizing inanimate objects as favelado. A wide cross-
section of society participates in online forums, and these forums provide a
space for the negotiation of meaning. Modern search engines make access to
this abundant, constantly updated material readily available. Using this data,
we argue that the use of morador de favela conventionally implicates the
meaning of a person from a slum. This meaning has a direct connection to
inhabitants of the places that are delimited as slums while avoiding ideolo-
gized meanings. In example (2) below, three native speakers of Portuguese
debate the difference between the terms morador de favela and favelado:
(2) sambistapt: ‘Eu li isto na parte de esportes da BBC sobre as olimpíadas em

2016 no Rio. Dizia que 1/5 da população do Rio era formada de “Slum-
-dwellers”. Não sei se concluí corretamente, quer dizer Favelado?
Dom Casmurro: Sim.
Joca: Mas não seria melhor dizer “morador de favela”? Parece que existe
uma diferença sutil entre “favelado” e “morador de favela”. Nem todo mo-
rador de favela é favelado. E também favelado tem uma conotação mais
ampla, não só econômica, significando, por vezes, uma pessoa de gosto
cultural duvidoso, mesmo que não seja necessariamente pobre.
Dom Casmurro: Acho que todo morador de favela é favelado. Há mil ou-
tras conotações, mas serão sempre metáforas com intenções pejorativas,
normalmente racistas, que se aplicam a uma grande variedade de situações

– até mesmo aos insultos futebolísticos como este, destinado a provocar a
torcida do Flamengo.
‘sambistapt: I read in the sports part of the BBC about the 2016 Olympics
in Rio. It was saying that one-fifth of the population of Rio is made up of
“Slum-dwellers”. I don’t know if I understood correctly, does that mean
Favelado?
Dom Casmurro: Yes.
Joca: But wouldn’t it be better to say “morador de favela”? It seems that

there exists a subtle difference between “favelado” and “morador de fave-
la”. Not every “morador de favela” is a “favelado”. And also favelado has
a wider connotation, not only an economic one, meaning, sometimes, a per-
son of dubious cultural taste, even if he isn’t necessarily poor.
Dom Casmurro: I think every morador de favela is favelado. There are a

thousand other connotations, but there will always be metaphors with pejo-
rative intentions, normally racist, that are applied to a wide variety of situa-
tions – even to the soccer insults like this one, aimed to provoke the chant
of Flamengo fans.’
[http://forum.wordreference.com/showthread.php?t=1550077]
Dom Casmurro captures a semantic generalization about shared meanings

between the two lexical items. However, this meaning arises through differ-
ent pragmatic calculations. This speaker’s interpretation of shared meaning
between the two forms (favelado and morador de favela) is validated by the
context of politically correct usage, as is exemplified in the terminology used
by government-funded organizations that serve slum-dwelling populations
(cf. the website for Movimento de Defesa do Favelado [Movement for the
Protection of Slum-dwellers, http://www.mdf.org.br/]). While morador de
favela always carries the conventional implicature, favelado can also be used
in this context; however, favelado does not arrive at this meaning through a
conventional implicature. Because of the socially-charged nature of fave-
lado, the neutral meaning as ‘slum-dweller’ can only be arrived at by way of
GCI. This will be explained in more detail below.
In example (3), a blogger differentiates between someone who is a mora-
dor de favela and someone who is favelado:
(3) Soa mal quando dizemos “Você mora na favela...”. Às vezes fica uma situ-
ação chata mas se eu mesmo morasse na favela, eu diria que sim, eu moro
na favela. Quando dizem que uma pessoa habita uma comunidade, a pri-
meira palavra que vem na cabeça é: Favelado. Mas fiquem calmos, existe
uma diferença grotesca entre ser favelado e apenas residir.
‘It sounds bad when we say “You live in the favela...” Sometimes this is
annoying, but if I lived in the favela, I would say “yes, I live in the favela”.
When they say a person lives in a ‘community’ (i.e. a favela), the first word
that comes to mind is: Favelado. But stay calm, there’s a grotesque differ-
ence between being favelado and residing in the favela.’
[http://blogdapostgeral.blogspot.com/2010/09/grande-diferenca-entre-o-
-favelado-e-o.html]
This blogger clearly explains that one can live in a favela without being
favelado. This is indicative of GCI status, since the ‘slum-dweller’ implica-
ture can be dissociated from favelado. This is not case for morador de fave-
la, which will be shown below. The following question forum discussion
provided in (4) from a thread on YahooAnswers (http://br.answers.yahoo.
com/) shows that speakers also understand the inverse of what is demon-
strated by (3), i.e. that one can be favelado without living in the favela:
(4) Pergunta: Para ser favelado tem que morar em favela?
CaRol: Mas tem gente que eu conheço que escuta funk, fala mal dos outros,
xinga todo mundo, fala mal de onde mora, não da [sic] a minima [sic] a es-
cola, fala alto, é chato, inconveniente e fala palavrão até 5 horas da manhã,
e mora em um prédio bom. Para mim isso que eu disse é ser favelado, por-
que tem gente que é normal e discreto e mora no fim do mundo.
Maria Claudia: Não pois um favelado, mesmo depois que deixa a favela,
continua [sic] se comportando como um favelado. É como dizem, a pessoa
sai da favela, mas a favela não sai da pessoa.
‘Question: To be favelado do you have to live in a favela?
CaRol: But there are people that I know who listen to funk, speak badly
about others, cuss all the time, complain about where they live, don’t give a
shit about school, speak loudly, are annoying and say bad words until 5 in
the morning, and live in a nice building. For me, what I said means being
favelado because there are people who are normal and keep to themselves
and live at the ends of the earth.
Maria Claudia: No because a favelado, even after having left the favela,
continues acting like a favelado. It’s like they say, you can take the person
out of the favela, but you can’t take the favela out of the person.’
[http://br.answers.yahoo.com/question/index?qid=20111204144203AA6Oivj]
From example 4, we extrapolate the following:
(5) a. #Uma pessoa é moradora de favela mas não mora na favela.

‘#A person is morador de favela but does not live in the favela.’
b. Uma pessoa é favelada mas não mora na favela.
‘A person is favelado but does not live in the favela.’
In (5a), the utterance is infelicitous because one cannot dissociate or can-

cel a linguistic form from its conventional implicatures (Grice, 1989[1975]).
Thus, the infelicity of (5a) suggests that the slum-dweller implicature must
be conventional because it both consistently arises and cannot be canceled.
In contrast, (5b) is acceptable because the implicature of living in the favela
that arises automatically with favelado can be canceled, which is clearly
suggestive of its GCI status. In the case of (5b), the implicature of being a
slum-dweller is also non-conventional, which Birner (2013:70) explains to
mean that “the implicature is not consistently carried by the particular lin-
guistic expression (which is why it can be canceled)”. Given that conversa-
tional implicatures require defeasibility and non-conventionality, and GCIs
specifically arise unless blocked, when the ‘slum-dweller’ index/implicature
is canceled from favelado, other social meanings must necessarily be ac-
cessed instead in order for the speaker to adequately communicate her mean-
ing. This suggests that what favelado actually “means” is ‘slum-dweller’
plus the myriad ideologized meanings that speakers associate with slum-
dwellers; and when the GCI of ‘slum-dweller’ is canceled in context, other
GCIs and potentially also PCIs are made available to hearers without the
direct connection to the favela. Since morador de favela and favelado are
synonymous in that they can both be used to describe a person who lives in
the favela, the proposition in (5b) seems contradictory.4 However, the multi-
tude of meanings for favelado, along with the possibility of canceling its
‘slum-dweller’ implicature, enable speakers to access and exploit socio-
pragmatic information that allows for the felicity of (5b).
3.2 GCI clusters
We propose that the sociopragmatic meanings for favelado function in
clusters, i.e. as multiple associated meanings. That is, when favelado is ut-
tered, a multitude of possible meanings can be activated. Consider (6) below:
4
And, while some might argue that there is an entailment relationship between morador de
favela and favelado, such that <morador de favela…favelado>, it is not the case that every
morador de favela is favelado due to the implicatures inherent in favelado. Nor is there an
entailment relationship in the other direction, given that not every person or thing described
as favelado comes from or resides in the favela. Similarly, the relationship between these
two terms does not involve hyponymy or hyperonymy.
(6) Em outros termos, “favelado” é tudo aquilo que rejeitamos pela falta de
prosperidade, de elegância, de ordem, de beleza ou de polidez, entre outros
aspectos, no qual são ressaltadas as ausências.
‘In other words, “favelado” is everything that we reject for lack of prosper-
ity, elegance, order, beauty or politeness, among other characteristics, in
which absences are highlighted.’
[http://dialnet.unirioja.es/servlet/articulo?codigo=3166013]
In this example, the writer attempts to define favelado by explaining a

range of qualities associated with it. In this way, the cluster of meanings (or
GCIs) is explicitly constructed for the reader. It follows that the act of can-
celing implicatures necessarily involves cluster cancellation. This process is
made clear in the following examples:
(7) Eu sou favelado, mas não sou traficante!

‘I am favelado, but I’m not a drug-dealer!’
[http://www.osabetudo.com/ei-voce-para-com-isso-de-me-julgar-sem-me-
-conhecer/]
(8) Tem muito favelado que é pobre mas é limpinho.

‘There are many favelado people who are poor but are clean.’
[http://comments.gmane.org/gmane.comp.db.postgresql.brasil/14865]
(9) Sou favelado mas sou humilde.

‘I am favelado but I am humble.’
[http://www.orkut.com/Main#Community?cmm=22134954]
(10) Sou favelado mas tenho muita dignidade.

‘I am favelado but I have a lot of dignity.’
[http://letras.mus.br/bezerra-da-silva/sou-produto-do-morro/]
(11) Sou favelado mas gente boa.

‘I am favelado but (I’m) a good person.’
[https://twitter.com/vagnerdossant20]
In all of these cases, the information following the adversative conjunc-

tion mas ‘but’ is included as the speaker/author attempts to cancel certain
GCIs that arise through the use of favelado. According to Anscombre &
Ducrot (1976), an adversative such as but (mas) carries an argumentative
force that affects the hearer’s conclusion about the propositions. The two
argumentative principles (for example, sou favelado and mas sou humilde)
are at odds, as seen in the examples above, such that the latter topos–or the
argument containing the adversative conjunction–is stronger and thus over-
rules the former. This can be seen in (7)-(11), where the former topos of sou
favelado is weaker than the argument beginning with mas, leading the hearer
to draw a conclusion about what is contextually meant by favelado. In the
examples in (7)-(10), the latter topos targets a specific GCI, while (11) tar-
gets the entire GCI cluster. However, even when a single implicature is “tar-
geted” for cancellation, the result is cancellation of multiple implicatures,
since each lexical item or expression (i.e. traficante, limpinho, humilde, ter
dignidade) comes with its own conversationally implicated, indexical mean-
ings. A speaker’s motivation for targeting one GCI for cancellation is in line
with Grice’s (1989[1975]) Maxim of Quantity and Horn’s (1984) R principle
(‘say no more than you must’). Accordingly, speakers can be efficient with
their language use by canceling more than what has been explicitly uttered.
By rejecting an implicature associated with favelado in (7), the speaker can-
cels the implicatures that arise with that particular lexical item (e.g. trafican-
te +> violento, deshonesto, etc.). Similarly, by asserting a certain property
that is contrary to typical associations made with favelado in (7)-(10), the
speaker is also making use of the pragmatic tool of implicature to cancel
multiple indexical meanings (limpinho +> orderly, not chaotic, responsible,
chaste, etc.). Thus, speakers make use of cancellation strategies to cancel
large clusters, or otherwise limit the possible implicated meanings of fave-
lado in context by removing certain potential meanings or implicating other
meanings that directly conflict with GCIs associated with favelado. Figure 4
below illustrates a potential indexical field for traficante. We posit that the
overlapping indexes (or implicated meanings) between Figure 4 for trafican-
te and Figure 5 for favelado are the ones that are canceled by an utterance
like (7).
Figure 4: Hypothetical indexical field for traficante ‘drug-dealer’
Accordingly, the potential meanings for favelado in (6) would be similar

to what is seen in Figure 5, by eliminating the GCIs that overlap between
traficante (Figure 4) and favelado:
Figure 5: Hypothetical implicatures for favelado in the context of example (7). Meanings that
are shared between the indexical fields for traficante (Figure 4) and favelado (this figure),
shown in bold, become canceled or removed from potential implicatures arising from the use
of favelado when (7) is uttered.
It is important to note that Figures 4 and 5 are possible representations for

the GCI meanings that are present for a given speaker/hearer. Due to differ-
ences in social and ideological experience, associations with words like fave-
lado or traficante necessarily vary from person to person. Depending on the
context and the relation between speaker and hearer, the interlocutors may
accommodate to each other. So, speakers assume common ground when they
use socially charged lexical items, but negotiation of meanings is often nec-
essary. The examples in (7)-(11) above represent this meaning negotiation
process, such that the speaker assumes that the hearer makes certain infer-
ences that align with the speaker’s GCIs. In this process of meaning negotia-
tion, the speaker seeks to cancel at least one implicature and potentially an
attached cluster from the discourse model. The speaker therefore not only
cancels the GCI(s) in the particular interaction but also reinforces and per-
petuates the social ideology that certain concepts (e.g. traficante) “should”
be associated with favelado. We suggest that if the speaker and hearer have
more common ground, the interlocutors’ indexical fields are more likely to
align. When the conversational participants differ in their ideological social-
ization, their indexical fields and thus the implicatures that arise from them
may be mismatched, potentially impairing communication.
Participants in speech communities tend to agree about which words are
socially charged due to shared language socialization. While socially-
charged words may vary slightly in their force from individual to individual,
some words are inherently more charged than others. Taboo terms and those
that divide people into social categories are inherently otherizing and polar-
izing, whether these categories are gender, race, sexual orientation, socio-
economic class, or something else. These taboo terms are fertile grounds for
creating new meanings, which co-exist with older meanings. As Traugott &
Dasher (2002: 11-12) state, “Despite what is often thought, the loss of an
earlier meaning is relatively rare. What is typical is the accretion of more
and more meanings over time [...]”. From this, we propose that the more
socially charged the lexical item, the more GCIs emerge. That is, traficante
necessarily implicates more social meanings than limpinho or humilde.
Terms that are pregnant with socially charged meanings have more extensive
indexical fields and therefore have a more complex pragmatic structure. In
other words, pragmatic creativity rests in the social complexity of lexical
entries in the collective mind of the speech community.
3.3 PCIs and reclaimed meanings
Particularized conversational implicatures (PCIs) differ from GCIs in that
the former are only accessible in specialized contexts. With respect to fave-
lado, we argue that the PCIs are related to competitive contexts. Beaton &
Washington (2014) show that soccer fans who cheer for the Rio de Janeiro
soccer team Flamengo have been characterized as favelado. These fans have
reappropriated or reclaimed the slur used against them, changing the nega-
tively-charged term to a positively-charged one. Consider the following
example:
(12) Não moro na favela…mais [sic] torço por um time CHAMADO

FLAMENGO…E esse time não é favelado…e se for, qual o preconceito?
E se ser FLAMENGO é ser FAVELADO, eu sou então…
‘I don’t live in the favela...but I cheer for a team CALLED

FLAMENGO... And this team isn’t favelado... and if it was, then why the
prejudice?
And if being FLAMENGO is being FAVELADO, then that’s what I am...’
[http://www.blablagol.com.br/ser-flamengo-4718]
In the example above, a positively-valenced PCI arises as a result of the

context of the specific soccer team. This contrasts with GCIs and their clus-
ters that arise unless blocked. In (12), the GCIs for favelado are canceled
(‘and this team isn’t favelado...’) and then reclaimed by the speaker (‘and if
being FLAMENGO is being FAVELADO, then that’s what I am...’) as a
positive concept with which to align himself. That is, the Flamengo fan is
taking the negative meaning for favelado and, by identifying with it himself
even though he does not live in the slums, turns the negative GCI meanings
(essentially, “bad”) into positive meanings (e.g. “badass”). Similar PCI use
can be found in other game contexts, such as the following comment on a
video clip from a videogame, in which the player uses a gun to kill ‘bad
guys’:
(13) caralho, tu é muito favelado meu

‘shit, you’re really favelado dude’
[http://www.youtube.com/watch?v=kr0H_77FLNE]5
In this usage, the implicature is interpreted positively, in terms of the skill

of the gamer, but could also be interpreted negatively by an outsider that
observes the violent nature of the game. The speaker in (13) might be ex-
pressing that the gamer is street-smart or ‘badass’. Both of these meanings
could be positively-valenced PCIs that arise in specific contexts–that is, if
one is commenting on the gamer’s competence, as is the case in (13), the
implicatures obtained are PCIs. When favelado is used in a context outside
of gaming or soccer, negatively-valenced indexes or clusters of GCIs will
arise necessarily unless blocked or canceled by the context. Accordingly, the
PCIs found in (12) and (13) do not arise in examples (7) through (11).6 We
hypothesize that positively-valenced PCIs will arise at some point in con-
texts other than gaming or soccer, but this does not require that the GCIs be
unavailable in the context. For example, we can imagine a context where the
‘street-smart’ implicature could be positive or negative: it could mean that
the referent is competent in specific ways, or it could implicate that the ref-
erent is a dangerous or shady character; in a given context, both of these
meanings could potentially arise. The availability of this flexible interpreta-
tion is common with socially-charged items, since in-group usage creates
positive PCIs without discarding the underlying GCIs used by society at
large. The simultaneous GCI and PCI package can be creatively exploited
for humor or to indicate membership within a stigmatized population (cf.
tongzhi from Wong [2005, 2008] and queer, fag from McConnell-Ginet
[2002]; for further discussion of these reclaimed terms, see Beaton & Wash-
ington [2014]). Given that the GCIs always arise and that, in specific con-
texts, PCIs can be layered on top of them, speakers can use the two types of
implicatures together for special effect.
4. Conclusions
In this paper, we have shown that speakers of Brazilian Portuguese have the
choice between two apparent synonyms, morador de favela and favelado.
5
This comment has been deleted from the source; originally accessed in June 2012.
6
PCIs, like GCIs, are cancelable. For instance, if the speaker in (13) had uttered, ‘tu é muito
favelado, mas não acho isso bom’ (‘you’re really favelado, but I don’t think that’s a good
thing’), the PCI of ‘badass’ that arises due to the competitive gaming context is canceled out
by the speaker’s explicit denial of any positive accomplishment. This kind of cancelation of
meaning is not available in the present data due to the specificity of the original contexts
required for the PCIs.
We propose that speakers use morador de favela in an attempt to avoid the

negative GCIs that arise with the use of favelado. However, if speakers want
to implicate positive values associated with ‘slum-dweller’, they have to use
favelado to access the PCIs through the negatively-valenced GCIs. Morador
de favela, in contrast with favelado, does not carry strong social ideologies
and therefore can be used neither as a slur nor as a positive term. As noted in
the previous section (3.3), words with strongly negative social ideologies
attached to them can be used positively only in very specific contexts; the
use of these words in situations in which the term has been reclaimed,
although perhaps empowering to the group in question, is accessed through
their pejorative roots with the help of layered implicatures. In terms of
Grice’s Maxims, speakers choose to use morador de favela to fulfill the
Maxim of Quantity (‘don’t say more than is necessary but say enough to
fulfill the communicative goals’). In other words, a speaker who uses this
more ‘neutral’ term does not want to activate the GCIs attached to favelado.
Grice’s Maxim of Quantity also comes into play when speakers choose to
use favelado but want to cancel certain GCIs or parts of the indexical field
that might be invoked for the hearer. For instance, a speaker can fulfill this
maxim by saying ‘I am favelado, but I’m not a drug dealer’ (cf. example 7),
which effectively cancels a whole cluster of associated meanings.
This paper has contributed to the study of indexicality by shedding light
on the pragmatic devices that enable speakers to make associations among
the various social meanings of favelado. The indexical approach developed
by Silverstein (2003) and fleshed out by Eckert (2008) for /t/ release pro-
vides a foundation for understanding the plurality of meanings attached to a
single linguistic form. However, the process of “side-stepping” to access
various meanings is not fully explained in the indexical framework. That is,
this theory implicitly suggests that only a single n+1 order meaning can be
highlighted at a time. Even though these approaches attempt to explain var-
iation in meaning, they inherently assume a one-to-one mapping between
form and meaning in a given utterance. We show that, in actuality, a cluster
of indexes within a field can be activated by a single instantiation of a lin-
guistic form. Furthermore, the indexical framework does not consider the
common ground of the interlocutors or the conversational context in a sys-
tematic way. In this paper, we discuss a variety of examples in which the
linguistic, discourse, and social context allow the speaker to convey the in-
tended meanings of favelado. The context-dependent nature of social index-
es calls for a discourse pragmatic approach, whether for phonetic, morpho-
syntactic, or lexical indexes.
Sinnott’s (2013) layered model proposed for morphosyntactic variation
could be adapted for lexical items. Using such an approach, we propose a
separate model for each lexical item. Morador de favela would consist of a
single layer involving the non-defeasible conventional implicature of ‘slum-

dweller’. Favelado, on the other hand, would have two layers, with the inner
layer representing the GCIs and the outer layer containing the PCIs.
Sinnott’s model improves upon indexical models by showing that indexical
meanings are highly context-specific and by indicating whether these mean-
ings are part of the GCI core or PCI fringe. This analysis can also be extend-
ed to other socially-charged lexical items.
Finally, we believe that linking linguistic forms with social meanings
through the use of indexical relationships is inherently pragmatic in nature
due to the interrelatedness between a linguistic form, context, and meaning.
With this in mind, we propose that our pragmatic approach could be expand-
ed to all kinds of indexical research, including morphosyntactic and phonetic
variables. For example, the types of relationships shown in Eckert’s (2008)
indexical field reflect social meanings implicated by the use of hypoarticu-
lated /t/ by speakers of American English. While some of these meanings
arise automatically, others require specific contexts in order to arise and be
accessed by speakers and hearers. The pragmatically-based analysis present-
ed in this paper highlights the role of pragmatics in social deictics and
demonstrates that speakers manipulate implicatures in the creation and at-
tempted destruction of place-based ideologies.
References
Anscombre, Jean-Claude & Oswald Ducrot (1976). L’argumentation dans la langue.
Languages 10 (42), pp. 5-27.
Beaton, Mary Elizabeth & Hannah B. Washington (2014). Slurs and the Indexical
Field: The pejoration and reclaiming of favelado ‘slum-dweller’. Language
Sciences. DOI: 10.1016/j.langsci.2014.06.021.
Birner, Betty J. (2013). Introduction to Pragmatics. Malden, MA: Blackwell.
Blutner, Reinhard (1998). Lexical Pragmatics. Journal of Semantics 15, pp. 115-
-162.
Bucholtz, Mary & Kira Hall (2005). Identity and interaction: A sociocultural linguis-
tic approach. Discourse Studies 7 (4-5), pp. 585-614.
Clark, Eve V. (1990). On the pragmatics of contrast. Journal of Child Language 17,
pp. 417-431.
Croom, Adam (2011). Slurs. Language Sciences 33, pp. 343-358.
Eckert, Penelope (2008). Variation and the indexical field. Journal of Sociolinguis-
tics 12, pp. 453-476.
Grice, H. Paul (1989[1975]). Logic and conversation. In Studies in the way of words.
Cambridge, MA: Harvard University Press, pp. 22-40.
Horn, Lawrence R. (1984). A new taxonomy for pragmatic inference: Q-based and
R-based implicature. In Meaning, Form and Use in Context (GURT '84).
Washington: Georgetown University Press, pp. 11-42.
Horn, Lawrence R. (2007). Neo-Gricean pragmatics: a Manichean manifesto. In

Pragmatics. London: Palgrave, pp. 158-183.
Kiparsky, Paul (1982). Word-formation and the lexicon. Proceedings of the
1982 Mid-America Linguistic Conference.
McConnell-Ginet, Sally (2002). ‘Queering’ semantics: Definitional struggles. In
Language and Sexuality: Contesting Meaning in Theory and Practice. Stan-
ford: CSLI, pp. 137-160.
Ochs, Elinor (1990). Indexicality and socialization. In Cultural psychology: The
Chicago symposia. Cambridge: Cambridge University Press, pp. 287-308.
Silverstein, Michael (2003). Indexical order and the dialectics of sociolinguistic life.
Language and Communication 23, pp. 193-229.
Sinnott, Sarah (2012). Toward a Connection between Generalized Conversational
Implicatures and Indexical Fields. Paper presented at The Ohio State Univer-
sity Congress on Hispanic and Lusophone Linguistics (OSUCHiLL 2012).
Columbus, OH, April 2012.
Sinnott, Sarah (2013). Generalized Conversational Implicatures and Indexical
Fields: The Case of Address Forms. Selected Proceedings from the 2012
Workshop on Spanish Sociolinguistics. Somerville, MA: Cascadilla Proceed-
ings Project, pp. 53-62.
Traugott, Elizabeth Closs & Dasher, Richard B. (2002). Regularity in Semantic
Change. Cambridge: University Press.
Wong, Andrew D. (2005). The reappropriation of tongzhi. Language in Society 34,
pp. 763-793.
Wong, Andrew D. (2008). On the actuation of semantic change: The case of tongzhi.
Language Sciences 30, pp. 423-449.
REVISTAS EM PERMUTA
1. Revista Portuguesa de Humanidades (Universidade Católica Portu-

guesa – Faculdade de Filosofia de Braga, Portugal)
2. Revista de Estudos Linguísticos da Universidade do Porto (Depar-
tamento de Estudos Portugueses e de Estudos Românicos da Facul-
dade de Letras da Universidade do Porto / Centro de Linguística da
Universidade do Porto, Portugal)
3. Lingue & Linguaggi (Università del Salento, Itália)
4. Verba Hispânica (Filozofska Faculteta Univerze v Ljubljana, Eslo-
vénia)
5. Revista de Estudos da Linguagem (Universidade Federal de Minas
Gerais, Brasil)
6. Revista Brasileira de Linguística Aplicada (Associação de Linguís-
tica Aplicada do Brasil)
7. Diacrítica – Revista do Centro de Estudos Humanísticos (Universi-
dade do Minho)
8. Journal of Portuguese Linguistics (AEJPL, Universidade de Lisboa)
Contactos para permuta
Endereço postal:
Estudos Linguísticos / Linguistic Studies
Revista do Centro de Linguística da Universidade Nova de Lisboa
Av. de Berna 26 C
1069-061 Lisboa
Portugal
Endereço eletrónico: revling@fcsh.unl.pt

Estudos-Ling 11 (3 P)

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Estudos-Ling 11 (3 P)

Hochgeladen von

Copyright:

Verfügbare Formate

ESTUDOS LINGUÍSTICOS

Revista do Centro de Linguística

O objetivo primeiro desta revista será o de dar visibilidade à produção cientí-

 Números regulares, a incluírem:

 Números especiais, eventualmente de caráter temático, dedicados a

Em todos os números (regulares e temáticos) a publicação de trabalhos

Nota editorial / Editorial note

Introduction to this special volume

On the emergenge of topicalisation in European Portuguese:

Discussing parametric variation:

The pear stories film: simple presents and present progressives

Focus movement as PF movement and other peripheral positions in BP

O possessivo de 2ª PL no dialeto mineiro: DP e CP em análise

Don’t put a label on me. Labels and interpretation of é que-clefts

The effects of lexical frequency on object clitic placement

Pragmatics and indexicality: the case of favelado

De acordo com o estatuto editorial da Revista Estudos Linguísticos/

Clara Nunes Correia

According to the Journal Estudos Linguísticos/Linguistic Studies editorial

Clara Nunes Correia

1. Introduction to this special volume of Estudos Linguísticos

The editors of the Selected Proceedings of PLUS

KEYWORDS: information structure; topicalisation; left dislocation; verb move-

A possible compromise with the broader Romance development would

2. Preposing Constructions in European Portuguese

(2) a. …e a elesi acuso-osi… (“colegas de trabalho”)

Foregrounding is an effect of verb movement to a head position in the

relative adverb position and VS order (Ribeiro, 1995; Antonelli, 2011).7 In

(3) … isso chamo eu falar ao pé da letra.

Criterion/Construction V2T TOP CLLD HTLD

(4) …  gosto muito __ (o baton)

(5) … mas isso já não se usa __.

(6) … os olhos parece [que __ lhe ficam atrás]…

(7) E de pasta japonesa não precisa __?

Example (4) shows that the interpretation of null objects in EP is fixed

(8) Bom… praias, adoro a Arrábida.

A further type of preposing construction involves focus: Focus Preposing,

(9) De notícias se faz o nosso mundo.

This presentation, made necessary due to the proliferation of conflicting

3. Possible Source Constructions for Topicalisation in Modern Europe-

(10) HTLD  *TOP

(11) CLLD  TOP

The investigation can therefore be constrained to two source construc-

4. The variation between Topicalisation and Left Dislocation

(13) English TOP

c. DF(LD3): Amnestying an island violation (by avoiding a long-

(15) … de quem te menos temesi essei te

Once HTLD was previously excluded as a possible source construction for

(16) Anselmo, podes trazer o chá… E os biscoitosi

(17) … e ao meu filhinhoi só tinha

The constituent between brackets is an adjunct island; so the presence of an

100% 0 0 0,02 0,05

75%0,47 0,37 0,55 0,49 TOP

Figure 1: Use of marked topic preposing constructions (excluding HTLD)

(i) E eu a Lívia buscava mais honra que apetite.

PERIOD New Inferable Evoked Total

Table 2: Topic givenness and presence of resumptive element

Inferable elements (either situationally or textually) slightly favour resump-

PERIOD Contrastive Shifting Familiar Total

Table 3: Topic salience and presence of resumptive element

These results contradict scenario (ii), according to which a pragmatic

5.3. Results on morphosyntactic criteria

Corrected mean = .63