Subject Positions and Interfaces: The Case of European Portuguese

≥

Studies in Generative Grammar 73

Editors

Jan Koster
Harry van der Hulst
Henk van Riemsdijk

Mouton de Gruyter
Berlin

New York

Subject Positions and Interfaces:

The Case of European Portuguese

Joa˜o Costa

Mouton de Gruyter
Berlin

New York

Mouton de Gruyter (formerly Mouton, The Hague)
is a Division of Walter de Gruyter GmbH & Co. KG, Berlin.

The series Studies in Generative Grammar was formerly published by
Foris Publications Holland.

앪

앝 Printed on acid-free paper which falls within the guidelines

of the ANSI to ensure permanence and durability.

ISBN 3-11-018112-6

Bibliographic information published by Die Deutsche Bibliothek

Die Deutsche Bibliothek lists this publication in the Deutsche Nationalbibliografie;
detailed bibliographic data is available in the Internet at

⬍http://dnb.ddb.de⬎.

쑔 Copyright 2004 by Walter de Gruyter GmbH & Co. KG, D-10785 Berlin.
All rights reserved, including those of translation into foreign languages. No part of this
book may be reproduced in any form or by any means, electronic or mechanical, including
photocopy, recording, or any information storage and retrieval system, without permission
in writing from the publisher.
Cover design: Christopher Schneider, Berlin.
Printed in Germany.

Contents

Foreword . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.1. Preview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.2. Grammar model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.3. Assumptions on V-movement and adverbs . . . . . . . . . . . . . . . .

2. Preverbal subjects: Spec,IP or left-dislocated? . . . . . . . . . . . . . . . 11

Multiple preposing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

Unstressed negative QPs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

No minimality effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

Reconstruction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

Unmarkedness of SVO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

Raising constructions and definiteness . . . . . . . . . . . . . . . . . . . 17

Pronominal doubling contextually limited (Costa 2000) . . . . . . 18

Subject positions in C-less if-clauses (Costa and Galves 2002) . . 19

Preverbal and postverbal pronominal doubling . . . . . . . . . . . . . 19

3. Postverbal subjects: syntax and discourse –

VSO and VOS orders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

3.1. Post-verbal subjects in VSO context . . . . . . . . . . . . . . . . . . . . . 23
3.2. Postverbal subjects in VOS contexts . . . . . . . . . . . . . . . . . . . . . 30
3.3. Objects in VOS: Scrambling in European Portuguese . . . . . . . . 35

4. Inversion and information structure . . . . . . . . . . . . . . . . . . . . . . . 71

4.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.2. Focus-movement? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.3. Word order and focus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.4. Mapping syntax-discourse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.5. No focus-movement in Portuguese . . . . . . . . . . . . . . . . . . . . . . 91
4.6. Phases, locality and subjects in Spec, VP . . . . . . . . . . . . . . . . . . 97
4.7. Related evidence: agreement in copular constructions . . . . . . . 100
4.8. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

5. Optionality and left-dislocated subjects:

semantic and discourse properties . . . . . . . . . . . . . . . . . . . . . . . . . 107

5.1. SV and VS in unaccusative contexts . . . . . . . . . . . . . . . . . . . . . 107

5.1.1. Lack of agreement in postverbal position . . . . . . . . . . . . 109
5.1.2. Hypothesis and arguments . . . . . . . . . . . . . . . . . . . . . . . 112
5.1.3. Summing up: is there optionality? . . . . . . . . . . . . . . . . . 118

5.2. VSO and SVO in answers to multiple wh-questions . . . . . . . . . 119

5.2.1. Semantic properties of multiple-wh questions . . . . . . . . 120
5.2.2. Syntactic consequences . . . . . . . . . . . . . . . . . . . . . . . . . . 125

6. Subjects in Spec,TP and the interface with morphology . . . . . . . 129

6.1. Spec,TP available in I-to-C contexts only . . . . . . . . . . . . . . . . . 129
6.2. The interface with morphology . . . . . . . . . . . . . . . . . . . . . . . . . 131
6.3. The non-parametric availability of Spec,TP . . . . . . . . . . . . . . . 138

7. Syntactic outputs and the interfaces . . . . . . . . . . . . . . . . . . . . . . . . 141

7.1. The behaviour of ditransitives in English

and in European Portuguese . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142

7.2. The behaviour of possessives in Portuguese and Italian . . . . . . 153
7.3. The differences between subject-verb inversion

in Brazilian and European Portuguese . . . . . . . . . . . . . . . . . . . . 155

7.4. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157

8. Summary and conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159

Appendix: On the nature of agreement in European Portuguese . . . 163

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
1.

The facts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
1.1.

DP-internal number agreement . . . . . . . . . . . . . . . . . . . . 165

1.2.

Subject-verb agreement . . . . . . . . . . . . . . . . . . . . . . . . . . 167

1.3.

Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173

Proposal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174

Further predictions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182

Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183

Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209

Contents

Foreword

This book is a much revised version of my 1998 dissertation, integrating
some of the work developed afterwards.
The book shares with the original work the plan of analyzing patterns of
word order variation. However, it does not focus only on the interface with
discourse, but also on the interfaces between syntax and semantics and syn-
tax and prosody. By doing so, it is possible to discuss and analyze more
data, explaining not only the differences between subjects in Spec,Agr and
Spec,VP, but also accommodating data involving Spec,TP and left-dislo-
cated subjects.

I would like to thank all colleagues who have contributed to this research
with comments, questions and suggestions.

Special thanks to my family and friends for making life a really nice thing!

1. Introduction

European Portuguese, like other Romance languages, displays a great amount
of word order variation. Out of the six logically possible permutations
between Subject, Verb and Complement in a transitive sentence, five are
possible: SVO, VSO, VOS, OVS and OSV.

The primary goal of this book is to provide an analysis of the several

positions where the subject may surface in European Portuguese. Departing
from an architecture of the clause as sketched in early minimalist work,
containing two subject-related functional categories above VP (AgrP and
TP), it is shown that the subject may surface in all potential landing sites:
Spec,AgrP, Spec,TP and Spec,VP. Moreover, just like any other argument of
the clause, it is claimed that subjects also have the possibility of surfacing in
a left-dislocated position, arguably adjoining to the clause’s left periphery.

It is shown that there is no free variation. Each of these positions may be

occupied by the subject, only if two requirements are met:

i) The position is made available by syntax;

ii) The position does not violate any interface condition.

In other words, the following model is argued for: syntax generates legiti-
mate outputs. At the interface levels, each output may be selected or filtred
out, according to requirements of the interface. These interface licensing
conditions operate in the following way for each of the identified surface
positions:

1. Spec,VP – The subject may surface in Spec, VP, because it is able to

2. Spec,TP – This position provides an interesting puzzle. Looking at

adverb positions, it appears that Spec,TP is an available surface position
for subjects in I-to-C contexts. This observation is explained if the inter-
face with morphology is taken into account. It is argued that the subject
cannot be stranded in Spec,TP when the subject is in T, since it blocks
the merger of the heads Agr and T.

3. Left-dislocation and Spec,AgrP – Non-focused preverbal subjects are

shown to occupy the specifier of the topmost functional category of the
inflectional domain. This goes against recent claims in the literature that
preverbal subjects in null subject languages are left-dislocated. Never-
theless, it is shown that the fact that preverbal subjects occupy an A-
position does not imply that they necessarily must occupy an A-position.
Looking at contexts of apparent optionality in answers to multiple wh-
questions, it may be shown that in the appropriate context subjects may
be left-dislocated. For a subject to appear in adjunction to the clause, it
must meet semantic requirements such as non-exhaustivity.

The picture emerging from the proposal made in this book is the following:
syntax proper does not need to refer to conditions best placed at the inter-
face. All that is needed from syntax is that it generates an array of well-
formed outputs. Such outputs may be evaluated a posteriori by each of the
interfaces. If they meet requirements of the interface, they are selected as
legitimate. If, on the contrary, some interface condition is violated, they are
ruled out. Under this approach, three independent results are derived: i) an
explanation is found for the patterns of word order variation; ii) syntax
proper may be reduced to its own tools, not having to manipulate semantic,
discourse or prosodic variables; iii) the intuition that European Portuguese
is an SVO language is derived: this word order corresponds to the one in
which the subject occupies the only specifier position in which the other
interfaces play no role.

In this first chapter, a preview of the book is presented. Also, some general

assumptions necessary for the analyses to be developed are spelt out and
motivated.

Introduction

1.1. Preview

In the past few years, some authors have proposed that there is no preverbal
A-position for subjects in Null Subject languages (Barbosa 1995; Alexiadou
and Anagnostopoulou 1998; among many others). If this proposal is right, it
is pointless to investigate the difference between subjects in Spec,IP
(=Spec,AgrP) and subjects stranded in Spec, VP. Chapter 2 tests the proposal
that preverbal subjects are left-dislocated in European Portuguese, providing
evidence showing that this analysis makes wrong predictions in several
respects. The tests presented involve a comparison with the syntactic behavior
of left-dislocated (non-subject) constituents, an evaluation of definiteness
effects, and the observation of agreement facts. The conclusion drawn is that
preverbal subjects may occupy an A-position in European Portuguese, which
precludes a generalization stating that this is impossible in Null Subject
Languages.

The third chapter investigates the position of the subject in VSO and VOS

orders. Arguments are presented showing that postverbal subjects may be
analyzed as stranded in Spec, VP, which does not exclude an I-to-C analysis
for different sorts of inversion. It is further contended that the object in VOS
orders displays properties reminiscent of the ones found in Germanic short-
distance-scrambling. It is therefore claimed that VOS is analyzed as a
involving a scrambled complement and a subject in Spec, VP. This type of
analysis is compared to alternatives, in particular to remnant-movement
proposals. A first approximation to the contexts in which inversion is legiti-
mate is made, suggesting that the interface between syntax and discourse
may be at play.

The contexts in which inversion orders are produced is investigated, and

the conclusion reached is that subjects are in Spec, VP only if they are
focused. This observation will favor an implementation of Reinhart’s (1995)
analysis of scrambling in Germanic for these data in terms of an interface
between syntax, information structure and prosody. The type of interface
between Information Structure, syntax and prosody is further explored in
chapter 4. It is argued that there is no need to assume that focused constitu-
ents move to discourse-related functional categories, and that the relevant
data follow from an algorithm for sentence-stress assignment. The relations
found between word order and discourse context allow for clearly saying
that there basically is no free variation. In the final section of this chapter, it
is argued that syntax generates multiple outputs that can be used for differ-
ent discourse purposes, because Case can be licensed under Move or Agree.

Preview

Locality effects on inversion and sensitivity to phase-boundaries supporting
this analysis are presented.

Chapter 5 investigates a case of apparent optionality, which would con-

tradict the conclusion of the previous chapter. In answers to multiple wh-
questions of the type who did what?, both SVO and VSO are legitimate.
The conclusions drawn from a closer look at the two word orders are the
following: i) they are not semantically equivalent: the postverbal subject is
not a true focus, and it must be non-exhaustive; ii) the postverbal subject
behaves as if it is left-dislocated. These two conclusions combined show
that the interface with semantics is also relevant for deriving all word order
patterns found, and that, as defended in chapter 2, the claim that preverbal
subjects are in Spec,AgrP and the claim that preverbal subjects may be left-
dislocated do not exclude each other, provided that appropriate semantic
and discourse conditions are met.

Chapter 6 deals with the availability of Spec,TP as a surface position for

the subject. It is shown that, in declarative SVO sentences, this is not a legiti-
mate landing site. However, if there is I-to-C movement, as in wh-questions,
the subject may surface in Spec,TP. The explanation for this puzzle relates
the fact that there is short V-to-I movement and explores the interface
between syntax and morphology, adapting Bobaljik’s (1995) solutions for
the availability of Spec,TP in Germanic languages. It is argued that, since
the verb does not move to Agr, for Agr and T morphology to merge, these two
heads must be adjacent. For the adjacency requirement to be met, Spec,TP
must be empty. In I-to-C contexts, the verb must move from T-to-Agr for
locality reasons, and the adjacency is no longer relevant, which explains the
restricted availability of this position.

In the last chapter, I summarize the results obtained throughout the book,

discussing in further detail the relation between the syntactic component
and the interfaces. In particular, it is argued that the interfaces act as filters
or as selectors of multiple outputs generated in the syntactic component.
According to this view, syntax proper excludes interface considerations, and
does not have to integrate discourse, prosodic or morphological notions.

1.2. Grammar model

Since this book deals with interface issues, it is important to spell out the
basic assumptions concerning the relation established between the several
grammar components. I will be basically assuming a model of grammar, as

Introduction

depicted in Chomksy (1995), combining it with proposals made in Reinhart
(1995) for the interplay between syntax and discourse, and in Halle and
Marantz (1993) for the interface between syntax and morphology. According-
ly, I assume that the output of syntax after Spell-Out feeds a morphological
component of the grammar. I further assume that the level of Information
Structure, in which notions such as topic and focus play a role, has access to
the syntactic outputs. Obviously, these two considerations make it necessary
to free syntax from discourse-related rules or operations as well as from de-
riving all morphological aspects. These two consequences will be explored
in chapter 4 and 6, respectively. The architecture assumed is summarized in
the schema in (1):

(1)

Lexicon

Spell-Out

Morphology

Information Structure

As emphasized by many authors, in particular in the last decade (Adger
1994; Reinhart 1995; Zubizarreta 1998; Costa 1998), this type of articulated
model of the grammar opens up possibilities for dealing with apparently free
word order. If syntax generates multiple converging outputs, each one of such
outputs may be used in compliance with different requirements of one of the
interfaces. Alternatively, syntax may generate multiple outputs, and some of
them may be ruled out at one of the interface levels. The former situation
will be argued for to explain the alternation between SV and VS orders in
Portuguese. As mentioned above, I will argue that a subject may be licensed
both in Spec, VP and in Spec,IP. These two outputs may be used for satisfy-
ing different requirements of the interfaces with prosody and information
structure. The latter case will also be explored. It will be argued that some
subject positions are not available, because the presence of the subject in
that specific position violates some interface requirement. This will be the
argument built for subjects in Spec,TP.

The task I will be doing in this book is therefore the following. Let us

assume a clause structure as the one proposed in Pollock (1989) and revised
in Belletti (1990):

(2)

[

AgrP

[

]]]

Grammar model

According to this structure, in principle, there are three A-positions where
the subject can surface. Moreover, if there are processes of left-dislocation,
the subject can also surface in the left periphery. Since it is known that there
is no free variation, it must be established if all positions must be used, and
under what conditions.

1.3. Assumptions on V-movement and adverbs

In most cases to be dealt with, the position of the subject will be determined
by looking also at the position of the verb. This is very obvious for the alter-
nation between SV and VS. As was made clear above, I will explore the pat-
terns of word order variation found assuming that there are several surface
positions for the subject. Therefore, it must be previously established where
the verb is, in order to make it possible to take its position as a valid diagnos-
tic for detecting subject positions. Likewise, since adverbs are a traditional
diagnostic to detect V-movement and different landing sites for arguments,
it must be established which adverbs are used as a diagnostic, motivating
their usage. The goal of this section is therefore to spell out my assumptions
concerning V-movement and adverb placement.

In Costa (1996, 1998), it is argued that, in European Portuguese, verbs do

move out of VP, but do not target the highest functional head. Assuming the
clause articulation in (2), the proposal made is that V moves to T, not reach-
ing Agr. Let us review the evidence in favor of this analysis. The first fact to
be mentioned is the following: like French infinitives, verbs in European
Portuguese may either precede or follow an adverb:

(3)

a. O João ontem leu o livro.

João yesterday read the book

b. O João leu ontem o livro.

João read yesterday the book

In his (1989) paper, for dealing with similar French data, Pollock suggests
that the verb may either stay inside VP or move up to Agr. Given the current
trend to eliminate optionality from the grammar, Pollock (1994b) makes an
attempt to circumvent this problem by saying that when there is optional
movement, what is at stake is the occurrence of forms which are morpho-
logically ambiguous. This would be the case for French infinitives, which

Introduction

would allow an interpretation either as nominal forms or as verbal forms.
Since in French, only verbal elements do move to I, only when the infinitival
forms are interpreted as verbal will this movement be required. If the infini-
tival form is interpreted as nominal, movement is not induced. This analysis
is however not tenable for European Portuguese, since the case of apparent
optional verb movement just presented cannot be analyzed in terms of mor-
phological ambiguity. The form leu/read in (3) is unambiguously a verbal
form in the third person singular. The mechanism proposed by Pollock
(1994b) for the optional movement of French infinitives may not be involved
here, since this form is not subject to be analyzed as a nominal element.

An alternative view is to assume that adverbs may be adjoined to different

projections. Such an assumption, combined with the possibility of moving
the verb to intermediate functional projections (which is also assumed in
Pollock 1989), would then derive the word order facts in Portuguese. (4)
illustrates the relevant analysis:

(4)

[

AgrSP

O João [

leu [

o livro]]]

If the adverb ontem is adjoined to TP, we get the ‘English-like order’:
S-Adv-V-O:

(5)

[

AgrSP

O João [

ontem [

leu [

o livro]]]]

If the adverb is adjoined to TP, the surface word order will match the pattern
typical of French: S-V-Adv-O:

(6)

[

AgrSP

O João [

leu [

ontem [

o livro]]]]

The same type of analysis is also preferred if one takes into account sentences
containing sequences of auxiliaries and in which adverbs may surface in
several positions without radical meaning changes. (7) is an example of such
a case:

(7)

(provavelmente) O Paulo (provavelmente) tinha (provavelmente)
probably the Paulo probably had probably

lido (provavelmente) o livro (provavelmente) à Maria
read probably the book probably to Maria

(, provavelmente).
probably.

Assumptions on V-movement and adverbs

Adopting the rigid view on adverbs that there is a single position for adverb
attachment and trying to derive the multiple possibilities via optional move-
ments of all the other constituents is quite implausible, since it creates the
rather unlikely need of making as many optional movements as there are
constituents and positions for adverbs.

The method of attaching the same adverb to different positions is not new.

It has been used in the literature to derive the same effects of Pollock (1989)
by Iatridou (1990) and Williams (1994), among others. It has also been used
by Zwart (1993) to derive scrambling in Dutch, without resorting to optional
object movement.

Another potential approach to the order S-Adv-V in Portuguese would be

to follow Belletti’s (1990) approach to similar facts in Italian.

(8)

Gianni probabilmente sbaglierà.

Gianni probably fail-FUT-3ps

‘Gianni will probably fail.’

Belletti (1990) argues that the order presented in (8) is to be derived in terms
of left-dislocation of the subject, which would explain that the subject and
the verb (in I) would not be adjacent. This analysis is confirmed by her
observation that the order S-Adv-V is only possible with definite subjects
(indefinites may not be left-dislocated in Italian) or with indefinites bearing a
heavy stress (which may independently be A-bar moved, in contrastive focus
constructions). The contrast between the two types of elements is exemplified
in (9):

(9)

a. Ognuno probabilmente sbaglierà.

everyone probably fail(-future)

b. *Nessuno probabilmente sbaglierà.

no-one

c. NESSUNO probabilmente sbaglierà.

‘No-one will probably make a mistake’

Although, it seems to make the right predictions for Italian, this analysis
cannot be extended to Portuguese. The reason is that the counterpart of (9)
does not display any asymmetries: both types of subjects may occur in the
order S-Adv-V, independently of focal stress. This is exemplified in (10)
below:

Introduction

(10)

Todos provavelmente errarão.

Ninguém provavelmente errará.

NINGUÉM provavelmente errará.

Another argument for not adopting Belletti’s analysis for Portuguese comes
from the fact that there is no focus-movement of the Italian type in Portu-
guese, as I will show in chapter 4.

An additional argument for the claim that the verb only moves up to T

comes from the distribution of floating quantifiers. Since Sportiche (1988),
floating quantifiers are taken as a diagnostic for A-movement. More specifi-
cally, Sportiche argues that floating quantifiers indicate where a subject has
been on its way to Spec,IP. For instance, in the sentence in (11) from Portu-
guese, the floating quantifier may either move along with the subject to
Spec,IP (11a) or be stranded in the base position of the subject in (11b):

(11)

Todos os miudos foram ao cinema.

All the kids went to the cinema

‘All the kids went to the movies.’

Os miudos foram todos ao cinema.
The kids went all to the cinema

Note that the distribution of floating quantifiers is one of Pollock’s (1989)
arguments for saying that the verb does not move in English. He notes the
following contrast between English and French:

(12)

a. The kids all love Mary.
b. *The kids love all Mary.

(13)

a. Les enfants aiment tous Marie.

The kids love all Marie

b. *Les enfants tous aiment Marie.

In (12), the verb stays in situ, and as a consequence, the stranded floating
quantifier precedes it. In (13), the verb moves to I, hence the stranded float-
ing quantifier follows it.

Note that if the same argumentation is followed, there is again evidence

for short verb movement of the verb in Portuguese, since the two positions
for the floating quantifier are possible:

Assumptions on V-movement and adverbs

(14)

Os miudos amam todos a Maria.
The kids love all Maria

Os miudos todos amam a Maria.

As for adverbs, I will assume that low monosyllabic adverbs mark the left-
edge of VP. This assumption is based on the the following arguments pre-
sented in Costa (1996, 1998):

– Monosyllabic adverbs do not distribute as freely as other adverbs;

– Monosyllabic adverbs only surface in clause-final position when it is

possible to show that the internal arguments of the verb have moved out
of the VP.

Summing up, on the basis of the discussion above, I will be assuming that
there is short-V-movement in European Portuguese, that is, the verb moves
out of VP targeting T and not the highest head of the IP-domain.

Moreover, it

will be assumed that certain adverbs may be taken as diagnostics for move-
ment. In particular, it is argued, following the argumentation developed in
Costa (1996, 1998) that low monosyllabic adverbs adjoin to the left of VP.

Introduction

2. Preverbal subjects: Spec,IP or left-dislocated?

A well known fact about null subject languages is that several word orders
are allowed. European Portuguese behaves as expected, in displaying almost
all possible alternations between subject, verb and direct object. Considering
the possible orderings between Subject, Verb and Object, it can be observed
that Portuguese exhibits five out of six logically available orders:

(1)

SVO
?*SOV
VSO
VOS
OSV
OVS

(2)

a. O Paulo comeu a sopa.

Paulo ate the soup

b. ?*O Paulo a sopa comeu.
c. Comeu o Paulo a sopa.
d. Comeu a sopa o Paulo.
e. A sopa, o Paulo comeu.
f.

A sopa, comeu o Paulo.

In this chapter, I will discuss the status of preverbal subjects, providing tests
to identify the position it occupies. This is a relevant task, in light of recent
proposals suggesting that preverbal subjects in Null Subject Languages are
left-dislocated (Barbosa 1995; Alexiadou and Anagnostopoulou 1998).

The pattern of word-order variation presented in (2) is well known and

has been subject to several analyses in the Generative tradition, among which
Ambar (1992) and Duarte (1987) were the first studies considering the rela-
tion between these word orders and discourse. Both authors agree that, in
spite of the variation, the unmarked/canonical word order of European
Portuguese is SVO. It is interesting to investigate why it is the case that
SVO is considered unmarked, or putting it in other terms, why an unmarked
context does not allow for several word orders to co-occur. Duarte (1987)

presents an analysis of sentence-initial objects, showing that they may be
either topicalized, as in (3a) or clitic-left-dislocated, as in (3b):

(3)

a. O bolo, o Pedro comeu.

the cake, the Pedro ate.

b. O bolo, o Pedro comeu-o.

the cake, Pedro ate it

CL-ACC

I will assume her analysis for sentence-initial objects and concentrate on
preverbal subjects, and take the properties associated with left-dislocation
of objects in order to test the status of preverbal subjects.

Sentence-initial preverbal subjects, like in (4), are traditionally assumed

to occupy the Spec, IP position (cf. Duarte 1987; Ambar 1992; Martins 1994).

(4)

O Paulo comeu a sopa.
Paulo ate the soup

This assumption has been challenged by several authors (see e.g. Barbosa
1995, 2000 for Portuguese, Valmala Elguea 1994, Ordoñez and Treviño
1995 for Spanish and Alexiadou and Anagnostopoulou 1995, 1998 for
Greek). The assumption shared by these authors is that in Null-Subject-
Languages preverbal subjects are not in Spec,IP, but rather left dislocated.
As noted, for instance in Barbosa (1995), this has theoretical advantages in
a framework like the one proposed in Chomsky (1993). According to
Chomsky, derivations are uniform, and the overt or covert nature of a syn-
tactic operation is determined by the specification of formal features associ-
ated either with the categories to be moved or with the functional category
where they land. Strong features trigger overt movement and weak features
trigger covert movement only. Within such a framework, optionality in
word order of the type observed above is a problem, since it is theoretically
undesirable to have features that are at the same time weak and strong.
Barbosa and the other authors propose that N-features of the subjects in
Null Subject languages are uniformly weak. Hence, subjects are only
expected to move to Spec,IP in covert syntax or they do not move at all, if
agreement is pronominal and sufficient to check EPP-features. Now, noth-
ing prevents them from being topicalized or left-dislocated like other cate-
gories are, since overt movement of topics may be formalized by assigning
them some [topic] feature requiring checking.

Preverbal subjects: Spec,IP or left-dislocated?

In this section, I will review this analysis and argue that, although some pre-
verbal subjects may be left-dislocated, the traditional analysis according to
which preverbal subjects are in Spec,IP makes right predictions for European
Portuguese. This debate regarding the status of the preverbal subject is quite
controversial, since some data are not very clear. In any case, it is possible
to compare the two analyses, showing that some data must involve subject-
left-dislocation, while other data must involve assuming that the subject is
in Spec,IP. In Costa (1998), a more radical position was assumed: it was
claimed that the analysis involving subject-left-dislocation could not be right
for most cases. However, as it will be shown below, there is no clear reason
not to assume that both analyses are compatible.

In order to decide whether the subject is in Spec,IP or not, it is necessary

to find out whether preverbal subjects exhibit A- or A-bar properties. For
this reason, it will be necessary to compare preverbal subjects with A-bar
left-peripheral constituents. Let us then list some of the arguments in favor
of assuming that preverbal subject in European Portuguese are in Spec,IP.

A. Multiple preposing

When two complements of the verb are preposed, their relative order is not
rigid. This is illustrated in (5):

(5)

a. Aos alunos, sobre sintaxe, o Rui falou.

to the students, about syntax, Rui talked

b. Sobre sintaxe, aos alunos, o Rui falou.

about syntax, to the students, Rui talked

Assuming that the subject is left-dislocated, it is expected that the order
between a preposed complement and the subject is not rigid. However, as
shown in (6), there is a contrast between SOV and OSV:

(6)

a. Esse bolo, o Paulo comeu-o.

that cake Paulo ate it

b. ??O Paulo, esse bolo, comeu-o.

Paulo, that cake, ate it

Multiple preposing

Assuming that the preverbal subject is in Spec,IP, it is expected that it will
not be able to surface before the complement. The lack of ordering restric-
tions will only be expected if both sentence-initial elements are adjuncts.

B. Unstressed negative QPs

Another argument in favor of the analysis according to which preverbal
subjects are in Spec,IP comes from the fact that it is possible for a negative
QP to occur in sentence-initial position, non-adjacent to the inflected verb,
as in (7):

(7)

Ninguém provavelmente leu esse livro.
nobody probably read that book

This word order is impossible in Italian. Belletti (1990) shows that a nega-
tive QP may only occur before a sentence adverb if it bears heavy stress.
According to her analysis, in that case, the negative QP is A-bar moved. The
impossibility of non-adjacency between unstressed negative QPs and the
inflected verb is explained in terms of the impossibility of left-dislocating
negative QPs.

(8)

NESSUNO/*Nessuno probabilmente ha sbagliato.
nobody probably failed.

The fact that in European Portuguese there is no contrast like the one in
Italian argues for an analysis in which the preverbal negative QP is not nec-
essarily left-dislocated. In (8), the negative QP is in Spec,IP, which is not
inherently associated with topic properties; it may be unstressed, since it is
not an A-bar position.

C. No minimality effects

Like in English, in embedded contexts, wh-movement is not obligatorily
followed by I-to-C movement. This fact makes it possible to observe adja-
cency between the moved wh-phrase and the subject, like in (9a). The
analysis assuming that the subject is left-dislocated predicts that the wh-
constituent may be followed by some other left-dislocated constituent. This
prediction is however not borne out.

Preverbal subjects: Spec,IP or left-dislocated?

(9)

Perguntei que livro o Pedro leu.
I asked which book Pedro read

b. *Perguntei que livro, à Maria, lhe deram.

I asked which book, to Maria, her

CL-DAT

(they) gave

It is important to note that there seems to be variation concerning the gram-
maticality of (9b).

In any case, since there are speakers for whom the contrast

exists, the difference can only be explained by assuming a different status
for the subject and for the left-dislocated constituents. A straightforward
way of explaining this contrast is to assume that wh-movement to the left of
a left-dislocated constituent yields a problem of minimality. If the subject is
in Spec,IP, it is expected that it will not create any problem regarding mini-
mality, since it occupies an A-position.

D. Reconstruction

As it is well-known, A-bar movement obligatorily reconstructs, unlike A-
movement. If it is assumed that the preverbal subject is in Spec,IP, it is ex-
pected that there is no reconstruction of the subject, while a left-dislocated
complement may reconstruct. In fact, the preverbal subject in (10a) cannot
reconstruct, scoping obligatorily over the agent. In (10b), the left-dislocated
complement may reconstruct, and the sentence is ambiguous:

(10)

a. Três livros foram lidos por dois estudantes.

S>Ag; *Ag>S

three books were read by two students

b. Três livros, dois estudantes leram-nos.

S>O; O>S

three books two students read them

The fact that there is no ambiguity in (10a) and the contrast between the two
cases seems to indicate that the preverbal subject and the preverbal object
occupy different positions.

E. Unmarkedness of SVO

As it will be discussed later in this chapter, unmarkedness may be detected
by looking at answers to the question what happened?. This context is

Reconstruction

unmarked from the point of view of discourse, since all elements will yield
new information. In EP, the answer to this question is SVO:

(11)

A: O que é que aconteceu?

what happened

B: a. O Pedro partiu o braço.

Pedro broke the arm

b. #Partiu o Pedro o braço.

broke Pedro the arm

c. #O braço, o Pedro partiu-o.

the arm, Pedro broke it

The fact that SVO emerges in this context is problematic for the left-disloca-
tion analysis for two reasons. First, it is possible to observe that left-disloca-
tion is illegitimate in this context (11c), which makes it difficult to explain
why the subject can be left-dislocated if other elements cannot. Second,
even if one would assume that the subject could be left-dislocated, it would
be necessary to explain why it cannot stay in its base position (Spec, VP).
Note that the impossibility of (11b) is not due to the exhaustive nature of the
subject in inversion constructions (Costa 2000), since the behavior is the
same in mono-argumental intransitive contexts:

(11)

O que é que aconteceu?
what happened

a. O João espirrou.

João sneezed

b. #Espirrou o João.

sneezed João

a. O João viajou.

João travelled

b. #Viajou o João.

travelled João

Once again, if the only A-position for the subject were Spec, VP, inversion
should be found in this context.

Preverbal subjects: Spec,IP or left-dislocated?

F. Raising constructions and definiteness

In Barbosa (2000), cases of preverbal subjects are presented that can only be
analyzed as instances of left-dislocation. For instances, sentences like (12)
must involve subject left-dislocation, and not movement to Spec,IP, other-
wise they would be instances of super-raising:

(12)

a. O homem parece que viu um monstro.

the man seems that saw a monster

b. O João parece que está parvo.

João seems that is fool

Her analysis is corroborated by the fact that there is a definiteness effect in
this type of construction. If the preverbal subject is indefinite, the left-dislo-
cation is impossible:

(13)

a. *Umas meninas parece que estão doentes.

some girls seems that are sick

b. *Baleias parece que comem peixe.

whales seems that eat fish

This definiteness effect is expected, since left-dislocation typically affects
definite XPs. Now, the comparison with constructions that are not obligatorily
analyzed as instances of left-dislocation becomes crucial at this point. The
SVO word orders in (14) do not exhibit any definiteness effect:

(14)

a. O homem foi assassinado.

the man was murdered

a’. Um homem foi assassinado.

a man was murdered

b. As meninas estão doentes.

the girls are sick

b’. Umas meninas estão doentes.

some girls are sick

c. As baleias comem peixe.

the whales eat fish

c’. Baleias comem peixe.

whales eat fish

Raising constructions and definiteness

The assumption that the preverbal subjects in (14) are in Spec,IP derives the
lack of definiteness effects associated with left-dislocation.

G. Pronominal doubling contextually limited (Costa 2000)

As shown in Barbosa (1995), in some null subject romance languages, pre-
verbal DP subjects are obligatorily doubled by a pronoun, which strongly
argues in favor of her analysis. In EP, doubling is possible, for instance in
answer to a multiple wh-question (Costa 2002a):

(15)

A: Quem leu o quê?

who read what

B: a. O João, ele leu o livro.

João he read the book

b. O João leu o livro.

João read the book

The possibility of pronominal doubling in EP suggests that Barbosa’s analysis
is right for this language as well. There are, however, contexts in which dou-
bling the subject is ungrammatical. For instances, if the whole sentence is
focused, as in (16), doubling is marginal at best:

(16)

A: O que é que aconteceu?

what happened

B: a. O João leu o livro.

João read the book

b. ??*O João, ele leu o livro.

João he read the book

Barbosa’s analysis, assuming that the preverbal subject is always left-dislo-
cated, predicts that doubling by a pronoun will always be possible, like in
other Romance null subject languages, which is not the case. Assuming
that, in a context like the one given in (16), the subject is in Spec,IP derives
the ungrammaticality of pronominal doubling in this context.

Preverbal subjects: Spec,IP or left-dislocated?

H. Subject positions in C-less if-clauses (Costa and Galves 2002)

Complementizer-less if-clauses force I-to-C movement in EP, like in English.
As illustrated in (17), in these sentences the subject appears right after the
inflected verb. The sentence initial position is ungrammatical:

(17)

a. Tivesse o João ido ao Brasil...

had João gone to Brazil

b. *O João tivesse ido ao Brasil...

John had gone to Brazil…

If the subject is in Spec,IP, the word order in (17a) is predicted, since I-to-C
movement will make the verb cross the subject position. If the subject were
right-dislocated, one would expect to find the word order in (17b), since the
left-dislocated subject would be adjoined to the left of CP.

I. Preverbal and postverbal pronominal doubling

Returning to the cases in which the preverbal subject can be doubled by a
pronoun, it is possible to observe that the pronoun may be preverbal or
postverbal:

(18)

a. O João, leu ele o livro.

João, read he the book

b. O João, ele leu o livro.

João he read the book

The postverbal position of the pronoun is not problematic for the analysis
according to which the subject is left-dislocated. (18b) is problematic, since it
must be analyzed as a case of left-dislocation of both the DP and the pronoun.
Note that there is nothing wrong in assuming that a DP and a co-referring
pronoun are left-dislocated. Such a case is illustrated in (19), in which the
DP complement is doubled by the strong pronoun a ele. The strong pronoun
itself is left-dislocated and doubled by a clitic:

(19)

O João, a ele, vi-o no cinema.
João, to him, I saw him

CL-ACC

in the movie theater

Subject positions in C-less if-clauses

Now, if the clitic may cooccur with multiple fronting, as in (19), it is expected
that the postverbal pronoun, corresponding to the clitic in (19), may co-occur
with the preverbal DP subject and the preverbal pronoun. This prediction is
not borne out:

(20)

O João, ele leu (*ele) o livro.
João he read he the book

Assuming that the preverbal pronoun is in Spec,IP, it is expected that it will
not be possible to find another subject in a lower position, since the subject
in Spec,IP was generated in Spec, VP.

In Costa (2003) and Costa and Duarte (2003), some additional evidence

against the idea that null subject languages are languages that check EPP by
means of pronominal Agr was presented.

First, there are languages which are only semi-pro-drop. This is the case of

Brazilian Portuguese (Coelho, Costa, Figueiredo Silva and Menuzzi 2001)
and Cape verdean creole (Pratas 2002). In these two languages, referential
null subjects are ungrammatical, but expletive pro, available with weather
verbs and unaccusative inversions, is available. This fact, predicted under
Rizzi’s (1982) licensing conditions for pro, is illustrated in the examples
below:

(21)

a. Chegou o Pedro.

arrived Pedro

b. *Viajou o Pedro.

traveled Pedro

c. *pro viajou.

traveled

(22)

Está chovendo.

is raining

(23)

a. Txiga tres pesoa.

CVC

arrived three persons

b. *Papia tres pesoa.

CVC

talk three persons

c. *pro papia

CVC

talks

Preverbal subjects: Spec,IP or left-dislocated?

(24)

Txobi.

CVC

rains

Inversion is possible in BP and CVC, in contexts in which Spec,IP is occu-
pied by expletive pro. These languages do not allow referential pro. Crucially,
if the availability of null subjects were a consequence of the availability of a
pronominal Agr responsible for EPP checking, there should be no mixed
systems.

The existence of mixed systems suggests that what is at stake is an inter-

action between different constraints, as suggested by Rizzi (1982) and Coelho
et alii (2001), among others.

The relevant constraints are: structural (i.e. is

Spec,IP projected?), which has consequences for checking of EPP and Case-
features; lexical (i.e. does Spec,IP have to be lexically filled?), which is an
EPP matter; paradigmatic (i.e. does pro exist in the pronominal system?). A
language in which Spec,IP is not projected will be a language with both ref-
erential and expletive null subjects. A language in which Spec,IP is projected,
but it does not need to be lexically filled will be a language with both types of
null subjects or with expletive null subjects only, depending on the avail-
ability of pro.

Another domain providing counter-evidence for the claim that preverbal

subjects are left-dislocated is language acquisition. If preverbal subjects are
left-dislocated in null subject languages, it is expected that VSO is unmarked,
and that children will only produce SV sentences by the time they master left-
dislocation (Adragão and Costa 2003). A study of the acquisition of subjects
in a child in his second year of life (Adragão 2001) reveals that inversion is
highly marked and rare in the child’s early productions:

(25)

% of SV/VS utterances in the child’s productions:

SV – 93
VS – 7

(out of 1060 sentences)

From these, most VS structures corresponds to passives, unaccusatives (79%)
and predicative structures, which are contexts in which inversion is unmarked
in the target adult system as well. It is importante to note that, in her study,
Adragão was looking for contexts in which the subject should be inverted in
the adult counterpart. Notice, as well, that at this stage, there are very few OV
sentences, and that there is no evidence for strategies of clitic left-dislocation,
since the acquisition of clitics is quite late in EP (cf. Duarte and Matos 2000).
These data provide additional evidence for the unmarked status of the pre-

Preverbal and postverbal pronominal doubling

verbal position for subjects. The correlation between inversion and unac-
cusativity make it impossible to claim that children do not know whether
Spec,IP is projected.

Finally, the existence of inflected gerunds in Dialectal European Portu-

guese, and their behavior in identifying the reference of null subjects (Ribeiro
2002) casts some doubt on the pronominal status of Agr in this language. The
problem identified by Ribeiro is that there is no great difference between dia-
lectal and standard EP, as far as the licensing of null subjects is concerned.

Independently of the presence of Agr, a null subject in a gerund clause must
be controlled by the subject of the matrix clause, as shown in (26):

(26)

Dialectal EP:

Standard EP:

a. Estandes ec

cansado, tu

podes ir.

a. Estando ec

cansado, tu

podes ir.

being-2sg tired, you may go

being tired, you may go.

b. Estandes tu cansado, eu posso ir.

b. Estando tu cansado, eu posso ir.

being-2sg you tired, I may go

being you tired, I may go

c. *Estandes ec

cansado, eu

posso ir.

c. *Estando ec

cansado, eu

posso ir.

being-2sg tired, I may go

being tired, I may go

The ungrammaticality of (26c), and the lack of difference between the two
dialects casts some doubt on the idea that Agr is pronominal in European
Portuguese.

From the results of the tests listed above, I conclude that the traditional

analysis assuming that preverbal subjects in EP are in Spec,IP (cf. Ambar
1992, Duarte 1992, among others) cannot be dispensed with.

It is important to note that the results of the application of the tests above

to European Portuguese do not imply that the results may be extended to
other Null Subject Languages. Actually, in Costa (1998), I assumed that the
descriptions and analyses of Greek by Alexiadou and Anagnostopoulou
(1998) and Spanish by Ordoñez and Treviño (1995) are correct, and accounted
for the difference between these languages and Portuguese. As implicit in
the discussion above, stating that subjects may be in Spec,IP does not mean
that they are in Spec,IP in all constructions. As shown in Barbosa (2000),
some constructions must be analyzed as instances of left-dislocation of the
subject (see for instances the cases of apparent super-raising). In fact, since
all constituents may be left-dislocated, subjects do not behave differently: in
chapter 4, I will show that they can be left-dislocated in a specific context.

Preverbal subjects: Spec,IP or left-dislocated?

3. Postverbal subjects: syntax and discourse –

VSO and VOS orders

In this chapter, the properties of postverbal subjects will be studied. It will be
argued that VS word orders may arise not only in contexts of I-to-C move-
ment (Ambar 1992), but also if the subject is stranded in its base position,
Spec, VP. Two contexts for post-verbal subjects will be considered: VSO
and VOS, both in declarative contexts. For both contexts, the conclusion
will be that the subject is in Spec, VP.

The argumentation will be the following: first, I will show that the post-

verbal position of subjects is not necessarily a consequence of V-movement
across the subject from I° to another functional projection. This conclusion
will be based on similarities between root and embedded contexts. The
analysis will also be based on assumptions regarding the reliability of
adverb positioning as a diagnosis for tracing the left-edge of VP.

3.1. Post-verbal subjects in VSO context

In any theory admitting the existence of functional projections, which are
potential landing sites for V, and accepting the VP-Internal Subject Hy-
pothesis (Koopman and Sportiche 1991), there are at least three ways of
deriving postverbal subjects in VSO order. Such mechanisms are illustrated
in (1), (2) and (3):

(1)

[

Subject

[

I’

[

Object]]]]

(2)

[

Subject [

V’

Object]]]

(3)

[

Subject

[

Object]]]

(1) illustrates an analysis of VSO, according to which this word order arises
in the following way: both Subject and Verb move out of VP; the subject
stops in Spec,IP, but the verb is further moved to the head position of a
functional projection above IP.

In (2), an alternative is presented: the verb moves up to I° and stops there.
The subject does not precede it, because it has never been moved from its
base-position Spec, VP.

Finally, (3) illustrates another type of analysis, according to which the sub-

ject undergoes short-movement out of VP, in the sense that it does not move
all the way up to its ‘normal’ landing site, Spec,IP. The verb undergoes
movement to I°.

It is important to note that these three types of analyses do not exclude

each other. In fact, it may be seen that they are all independently instantiated
in different types of languages. For instance, (1) is the type of analysis argued
for Dutch and German Verb-second phenomenon and for English Subject-
Aux inversion in questions (cf. 4); (2) has been argued by Ouhalla (1991) to
be the correct analysis for Celtic VSO (5); (3) has been argued for Icelandic
Transitive Expletive Constructions by Bobaljik and Jonas (1996):

(4)

a. Dutch V2:

Gisteren heeft Jan het boek gelezen.
Yesterday has Jan the book read

b. English Subject-Auxiliary Inversion:

Who had Mary seen?

(5)

Celtic VSO:

Darnellod y plentyn y lyffr.
Read the child the book

(6)

Icelandic Transitive Expletive Constructions:
ﬁa› hafa margir jólasveinar bor›a› bú›ing
There have many Santa Clauses eaten puding

For European Portuguese, all three types of analyses have been proposed:
Ambar (1992) has argued that post-verbal subjects in this language are the
reflex of movement of I-to-C, along the lines of (1); Martins (1994) and
Duarte (1997) have proposed that under certain circumstances subjects may
be stranded in the Specifier of a functional projection below IP but higher
than VP, following the lines of (3); in this chapter, as in Costa (1997, 1998),
I argue for an analysis along the lines of (2).

The fact that the three types of analyses exist independently of each other

does not mean that they exclude each other. For instance, it is possible to

Postverbal subjects: syntax and discourse – VSO and VOS orders

assume that V moves to C in questions, as proposed in Ambar (1992), and
that the subject stays in Spec, VP. That is, we may have two analyses for the
sentence in (7):

(7)

Quem viu o João?
Who saw João

‘Who did João see?’

One analysis would assume that the subject is in Spec,IP and the verb has
raised up to C, as in (8a); another analysis would claim that the subject is in
Spec, VP, as in (8b):

(8)

a. [

Quem [

C’

viu [

o João [

I’

….

b. [

Quem [

C’

viu [

[

o João….

That the two analyses are necessary is confirmed by the fact that there are
two adverb positions, one before the subject and one after it, which may be
explained by assuming that the adverb is adjoined to VP, and adopting the
explanation in (9):

(9)

a. Quem viu o João ontem?

yesterday

a’. [

Quem [

C’

viu [

o João [

I’

[

ontem….

b. Quem viu ontem o João?
b’. [

Quem [

C’

viu [

[

ontem [

o João….

Bearing this idea in mind, I will now show that for simple declarative affir-
mative sentences, an analysis claiming that post-verbal subjects are in Spec,
VP is more attractive.

There are four main arguments for not deriving VS order in declarative

sentences necessarily in terms of an analysis involving I-to-C movement.
The arguments are listed in a) through c):

a) VS is possible in embedded sentences:

If postverbal subjects were only derived by V-movement to C, they should not
appear in embedded contexts, since C is occupied by the complementizer.
This prediction is not borne out by the data:

Post-verbal subjects in VSO context

(10)

O Paulo disse que comeu a Maria a sopa.
‘Paulo said that ate Mary the soup’

Note that this is not problematic for an analysis in which the subject is
stranded in the Specifier of a functional projection higher than VP and lower
than IP. Such an analysis will be considered below, when this word order is
confronted with possible readings of adverbs.

b) Sequences auxiliary-participle(s):

The second problem for the type of analysis defending that VS arises neces-
sarily by virtue of movement of the verb to a functional projection above
the landing site of the subject is the fact that in more complex verbal con-
structions (involving an auxiliary verb and one or more non-finite forms of
the verb), the subject may follow all of them:

(11)

a. Tinha comido o Paulo maçãs.

‘had eaten Paulo apples’

b. Tem estado a comer o Paulo maçãs.

‘has been eating Paulo apples’

If inversion could only be derived by moving the verb across the subject to
the functional projection above it, we would expect to find the subject
obligatorily following the auxiliary verb, in a construction similar to Rizzi’s
(1982) Aux-to-Comp. In order to assume that the subject in (31a) is in
Spec,IP, one has to postulate at least two heads above IP. If more auxiliaries
are present, more heads have to be postulated, as (11b) shows.

c) The distribution of adverbs:

The distribution of adverbs also supports the idea that postverbal subjects
do not reflect V-movement to a position higher than IP. In Costa (1998), I
have proposed that a proper characterization of the distribution of adverbs
with a subject-oriented reading requires that these adverbs are adjoined to
TP, and that this reading is only triggered whenever the subject moves up to
Spec,IP (=Spec,AgrSP). Assuming these results, there is a further diagnostic
for identifying the position of the subject: if postverbal subjects were in

Postverbal subjects: syntax and discourse – VSO and VOS orders

Spec,IP, adverbs with a subject-oriented reading should be expected to
adjoin to TP, thus intervening in between the subject and the object. This
prediction is incorrect:

(12)

*Comeu o Paulo inteligentemente maçãs.
‘ate Paulo cleverly apples’

(12) is grammatical, but only with a manner reading for the adverb. The
structure allowing (12) would be the one in (13):

(13)

[

Comeu [

o Paulo [

I’

[

inteligentemente [

…[

Subj

maçãs]]]]]]

On the basis of these arguments, I reject an analysis of inversion in declara-
tive sentences that is solely based on Verb movement across the subject in
Spec,IP. I stress nevertheless that such an analysis is still necessary e.g. for
interrogative sentences, as shown in Ambar (1992 and subsequent work).

Having pointed out some problems for an analysis of inversion solely in

terms of verb movement across a subject moved to Spec,IP, it is my task
now to develop the alternative analysis for VS orders. In this section I will
argue that VSO orders may be analyzed as the result of a representation in
which subjects stay in their base-position, Spec, VP, and do not move to
Spec,IP. Evidence for this analysis will come from the distribution of
adverbs.

In Costa (1996, 1998), I argued that the position of monosyllabic adverbs

is a good test for determining the left edge of VP. Although I emphasized
that the conclusions reached are valid for English, but not necessarily for
other languages, it seems that similar adverbs may be used as diagnostic for
the left-edge of VP in Portuguese as well.

Let us recall the two main properties of these adverbs:

a) First, they do not display the properties associated with right-adjunction,
which may be seen in the order PP-Adv, which, without any special intona-
tional mark on the adverb is ill-formed both in Portuguese and in English:

(14)

a. John looked hard at some pictures.

b. *John looked at some pictures hard.

Post-verbal subjects in VSO context

(15)

a. O Paulo olhou bem para alguns daqueles quadros.

Paulo looked well at some of those pictures

b. *O Paulo olhou para alguns daqueles quadros bem.

b) Secondly, monosyllabic adverbs have a more restricted distribution than
other types of adverbs:

(16)

a. John (carefully) has (carefully) looked (carefully) at some

pictures (carefully)

b. John (*well) has (*well) looked (well) at some pictures

(*well).

(17)

a. O Paulo (cuidadosamente) tinha (cuidadosamente) olhado

(cuidadosamente) para aqueles quadros (cuidadosamente).

b. O Paulo (*bem) tinha (*bem) olhado (bem) para aqueles

quadros (*bem).

Given the similarities between English and Portuguese, I will consider the
monosyllabic adverb as a reliable test for marking the left-edge of VP also
in Portuguese.

Let us then consider the possible positions for the monosyllabic adverb

bem ‘well’ in a VSO sentence:

(18) a. *Bem comeu o Paulo maçãs,

‘well ate Paulo apples’

b. ?*Comeu o Paulo bem maçãs,

c. *Comeu o Paulo maçãs bem.

d. Comeu bem o Paulo maçãs,

Example (18) shows that the only position for the adverb to surface in a
VSO sentence is in between the verb and the subject. All other positions are
excluded.

If the position of these adverbs is the same in English and

Portuguese, as the similarity of the data makes one suspect, it is legitimate
to conclude that in the VSO word order, the subject is in Spec, VP.

Note that it is not possible to adopt an analysis of this word order, in

which the subject would move out of the VP to some functional projection
above it but lower than AgrSP, as has been argued for Icelandic Transitive

Postverbal subjects: syntax and discourse – VSO and VOS orders

Expletive Constructions by Bobalijk and Jonas (1996), and for Portuguese
by Duarte (1997) and Martins (1994). The problem with such an analysis is
the fact that the adverbs considered here obligatorily follow the participial
form in a sequence consisting of the inflected auxiliary followed by the par-
ticipial form. This is exemplified in (19):

(19)

a. O Paulo tinha lido bem alguns livros.

Paulo had read well some books

b. *O Paulo tinha bem lido alguns livros.

The contrast in (19) not only shows that it is not possible to extend Bobaljik
and Jonas’ (1996) analysis to Portuguese, but it also provides further evidence
for the claim put forward here regarding the distribution of monosyllabic
adverbs. Note that the facts in (19) imply that the participial form moves out
of the VP to some functional head. This is possible and confirmed by the
distribution of the adverb itself. It should however be noted that the parti-
cipial form could remain in VP, as it does in French (M.T. Vinet, personal
communication):

(20)

a. Jean avait bien lu les livres.

Jean had well read the books

b. *Jean avait lu bien les livres.

The fact that (20b) is ungrammatical in French excludes an analysis of (19) in
terms of adjunction of the adverb to some lower level within VP. Assuming
that monosyllabic adverbs are adjoined to VP in the two languages, we may
explain the contrast between the order adverb-participle in French and
Portuguese with the usual tools deriving differences in head movement: parti-
ciples obligatorily move out of VP in Portuguese, while they never do so in
French.

I thus conclude that the distribution of adverbs seem to support the idea

that postverbal subjects are in Spec, VP. Further evidence for this claim
comes from the study of VOS orders.

Post-verbal subjects in VSO context

3.2. Postverbal subjects in VOS contexts

In the preceding section, I have investigated the position of postverbal sub-
jects in VSO contexts. In this section, I will look at another possible position
for post-verbal subjects, the one in which they follow the object. I will argue
for an analysis of this word order similar to that of VSO. In VOS, subjects
will be claimed to be in Spec, VP, but, differently from VSO, the object has
scrambled out of the VP, crossing the subject. The argumentation in defense
of the movement of the object will be presented in the second half of this
chapter (cf. Section 3.3). I will therefore assume the results of that section
in my discussion of the subject position in VOS, and refer the reader to sec-
tion 3 for the evidence supporting object movement in European Portuguese.

The VOS context is exemplified in (21):

(21)

Comeu a sopa o Paulo.
Ate the soup Paulo.

In the traditional analysis for this word order, subjects are claimed to be right-
dislocated (see e.g. Rizzi 1982 for Italian). I do not think that the subject-
right-dislocation analysis is to be dispensed with, but would nevertheless
like to clarify the data a little bit. In this section, I will be concerned with
sentence-final subjects that are not right-dislocated. How can the two types
of subjects be differentiated? I would like to show that right-dislocated sub-
jects and sentence-final base-generated subjects distribute differently.

The first distinguishing factor is intonation: while right-dislocated subjects

are preceded by a pause, sentence-final subjects are not:

(22)

a. Comeu a sopa # o Paulo.

Ate the soup Paulo

b. Comeu a sopa o Paulo.

This observation may seem at first sight rather awkward: it could be that, in
both cases, subject right-dislocation is involved, the pause being optionally
possible. As a result, different intonations would not be associated with dif-
ferent syntactic structures. However, if the paradigm in (23) is taken into
consideration, it is possible to see that, when a pronoun is inserted in
Spec,IP, the pause is obligatory:

Postverbal subjects: syntax and discourse – VSO and VOS orders

(23)

a. Ele comeu a sopa # o Paulo.

‘he ate the soup Paulo’

b. *Ele comeu a sopa o Paulo.

(24) shows that the pronoun is ungrammatical in a VSO context, which can-
not be treated as a case of subject right-dislocation:

(24)

(*Ele ) comeu o Paulo a sopa.
‘he ate Paulo the soup’

One way to interpret these facts is to assume with Kayne (1994) and
Zubizarreta (1998) that the right-dislocated subjects are clause-external
(tags in Zubizarreta 1998, the result of clause reduction for Kayne 1994).
Further evidence for this claim comes from the interaction between ques-
tion tags and subjects. Den Dikken (1995) shows that shifted heavy NPs,
traditionally analyzed as right dislocated, follow question tags:

(25) They have found, haven’t they?, the treasure buried on that island 100

years ago.

The same is true in EP: right-dislocated subjects follow question tags while
subjects in Spec, VP precede it. This is shown by the interaction between
these orderings and pronoun insertion:

(26)

Comeu a sopa o Paulo, não comeu?
‘ate the soup Paulo not ate’

b. *Ele comeu a sopa o Paulo, não comeu?

Comeu a sopa, não comeu?, o Paulo

Ele comeu a sopa, não comeu?, o Paulo.

The only possibility to obtain the order subject-tag is either to leave out the
pronoun, or to introduce a pause before the subject. This just shows that the
two sentence adjuncts, tag and right-dislocated subject are interchangeable:

(27)

(ele) comeu a sopa # o Paulo # não comeu?

Post-verbal subjects in VOS contexts

In the remainder of this section, I will ignore right-dislocated subjects, since
they appear to be extra-sentential. I will concentrate on the ones I claim to be
in Spec, VP, that is postverbal subjects not involving an intonational break
after the object.

Having established this difference, let us now look at the tests that enable

us to identify the position of the subjects in VOS sentences.

a) The distribution of adverbs:

Applying the same test as for VSO, it is now possible to check the paradigm
for the several potential positions for the monosyllabic adverb bem ‘well’ in
VOS contexts. Keeping in mind that cases of right-dislocation are to be
ignored and can be tested by inserting a nominative pronoun, the results are
the following:

(28)

a. Comeu bem a sopa o Paulo

‘ate well the soup Paulo’

b. *Comeu a sopa o Paulo bem.

c. Comeu a sopa bem o Paulo.

The results in (28) seem to indicate that, independently of what happens to
the object, the subject is VP -internal, surfacing obligatorily to the right of
the adverb. In section 3, I will return to these examples and explain why the
two word orders between the object and the adverb are possible.

b) Binding effects.

Binding effects seem to support the analysis of VOS I am presenting here.
As it is well know, A-movement feeds binding (cf. 29), while A-bar move-
ment does not (cf. 30).

(29)

a. *It seems to each other that the boys are tired.

b. The boys seem to each other to be tired.

(30)

a. *Paul introduced each other to Mary and Paul.

b. *Who did Paul introduce to each other?

Postverbal subjects: syntax and discourse – VSO and VOS orders

The same effects may be visible in Portuguese: in (31), the QP only binds
the possessive in the passive construction (31b), and not when it is wh-
extracted (cf. the Weak Cross -over effects in 31c):

(31) a. *O

seu

realizador viu todos os filmes

‘their director saw all the movies’

b. Todos os filmes

foram vistos pelo seu realizador

‘all the movies were seen by their director’

c. *Que filmes

viu o seu realizador

‘which movies saw their director’

Now consider the behavior of subjects and objects in a VSO order: if the
object is a QP and the subject contains a possessive anaphor, binding is
impossible. If object scrambling were A-movement binding would be accept-
able, on a par with (31b). This is not the case, as (32b) illustrates:

(32)

a. *Viu o seu

realizador todos os filmes

‘saw their director all the movies’

b. *Viu todos os filmes

o seu

realizador.

‘saw all the movies their director’

The impossibility of binding becomes even clearer in cases like (33), which
are only acceptable if interpreted as involving VSO order:

(33)

a. *Viu [

Obj

o Paulo

] [

Subj

o seu

irmão].

‘saw Paulo his brother’

b. Viu [

Subj

o Paulo

] [

Obj

o seu

irmão].

One could argue that these effects do not constitute conclusive evidence in
favor of the analysis of VOS order I am advocating, since binding would be
impossible anyway if the subject were moved rightwards and adjoined to a
position higher than the object, making it impossible for the object to c-
command it. However, the following examples prove that this is not true: in
the cases which were identified as instances of right-dislocation, i.e. those
cases in which a tag may intervene between the object and the subject, bind-
ing is possible:

Post-verbal subjects in VOS contexts

(24)

a. Viu [

Obj

o Paulo

i ]

, não viu?, [

Subj

o seu

irmão].

‘saw Paulo, not saw, poss brother’

b. Viu [

obj

todos os filmes

, não viu?, [

Subj

o seu

realizador].

‘saw all the movies, not saw, poss director’

c) The differences between definites and indefinites:

If VOS contexts indeed involve subjects in Spec, VP, there should not be
any difference between definites and indefinites, since both types of sub-
jects occur in VSO sentences, as illustrated in (25):

(25)

a. Comeu o Pedro a sopa.

ate Pedro the soup

b. Comeu um homem a sopa.

ate a man the soup

In fact, postverbal subjects in VOS may be either definite or indefinite:

(26)

a. Comeu a sopa o Paulo.

‘ate the soup Paulo’

b. Comeu a sopa um homem.

‘ate the soup a man’

The fact that definiteness does not play any role also confirms that VOS
word order is not necessarily derived in terms of right-dislocation of the sub-
ject, since in the true cases of subject-right-dislocation, indefinite subjects
are marginal, as illustrated in (27), in which the question tag helps to differ-
entiate the two types of constructions:

(27)

Comeu a sopa, não comeu?, o Paulo.

Ate the soup, not ate?, Paulo

b. *Comeu a sopa, não comeu?, um homem qualquer.

Ate the soup, not ate?, some man

The definiteness effect in (27) is expected, since right-dislocation may be
associated with pronominal doubling, as shown above. Since only definite
DPs can be doubled, right-dislocated subjects behave as expected.

Postverbal subjects: syntax and discourse – VSO and VOS orders

From the tests above, I conclude that in VOS context, subjects occupy the
Spec,VP position. The difference between VSO and VOS will be derived in
terms of object scrambling in the latter, in accordance with the evidence to
be presented in section 3. A comparison between the scrambling analysis
and an analysis involving remnant-IP movement, along the lines of Kayne
(1998) will provide further evidence in favor of the claim that subjects in
VOS are in Spec, VP.

At this point, it is possible to tell which position the subject occupies in

each of the attested word orders.

The conclusions reached so far are the following:

1. Preverbal subjects occupy the Spec,IP position; like all other arguments,

it can be left-dislocated in some contexts.

2. Postverbal subjects in VSO context may be in Spec, VP

3. Postverbal subjects in VOS without intonation break between the object

and the subject are in Spec, VP

In the next section, the position of objects in VOS sentences, which is cru-
cial for the analysis of this word order, will be discussed.

3.3. Objects in VOS: Scrambling in European Portuguese

In the preceding section, I have assumed without providing arguments that
Portuguese VOS orders involve object scrambling across the subject. In this
section, I will motivate that assumption, arguing that it is possible to identify
similarities between scrambling in European Portuguese and the well-known
scrambling configurations in Dutch and German. In order to achieve this
goal, I will first show some properties of scrambling in Dutch and German.
This will establish the grounds for a comparison between Portuguese and
Dutch/German, and determine how to trace scrambling in a VO language.
As mentioned in the previous section, if the scrambling analysis is on the
right track, there will be further evidence in favor of the claim that the sub-
ject may be stranded in Spec, VP in EP. In order to strengthen this claim, the
scrambling analysis will be compared with an analysis in terms of remnant
IP-movement, following proposals of Kayne (1998). Tests will be provided
for comparing the two analyses. I will further review three of the theories of
scrambling: case-driven movement (Zwart 1993 among others), semanti-

Objects in VOS: Scrambling in European Portuguese

cally-driven movement (Diesing and Jelinek 1995) and prosodically-driven
movement (Reinhart 1995, Zubizarreta 1998), showing that the Portuguese
data provide evidence in favor of the latter.

The results of this section will challenge the view advocated in Webelhuth

(1989) and Neeleman and Reinhart (1999) that scrambling is a property of
OV languages only. Moreover, showing that there is scrambling in Portu-
guese will add one more language to the group of languages in which this
phenomenon is visible, which makes it possible to broaden the empirical
coverage of the very debated issue of what the exact nature of scrambling is.

Scrambling was first discussed by Ross (1967) to refer to the syntactic

process that permits breaking the adjacency complement-verb by insertion
of an adjunct. The scrambled variant of (38a) is (38b):

(38)

a. Adverb Object Verb

No scrambling

b. Object

Adverb

Verb

Scrambling

In this section I will illustrate some of the properties of scrambling in German
and Dutch, just to establish a basis for comparison between these languages
and European Portuguese. I will sometimes use the term movement for
scrambling, but this does not yet mean that I am adopting a specific theory
of scrambling. At this stage, I will remain neutral with respect to the correct
analysis of scrambling (movement or base-generation of the complement in
the scrambled position). The properties that will be considered are the fol-
lowing:

a) Scrambling moves NPs and PPs

(39)

Dutch:
a. dat Jan in Amsterdam zijn vriendin ontmoet.

that Jan in Amsterdam his girlfriend meets

b. dat Jan zijn vriendin in Amsterdam ontmoet.

that Jan his girlfriend in A’dam meets

(40)

a. dat Jan waak op zijn vriend wacht.

that Jan often for his friend waits

b. dat Jan op zijn vriend waak wacht.

that Jan for his friend often waits

Postverbal subjects: syntax and discourse – VSO and VOS orders

b) Differently from the categories in (39) and (40), predicative APs and
some Small Clauses cannot be moved, as (41) and (42) show:

(41)

*dat Jan Marie aardig altijd vindt.
that Jan Marie always nice finds

(42)

*dat Jan ziek altijd is.
that Jan sick always is

c) Another property of scrambling noted by Bennis and Hoekstra (1984) is
that it licenses parasitic gaps:

(43)

a. Jan heeft die boeken zonder te lezen weggelegd.

Jan has these books without read put away

b. *Jan heeft zonder te lezen die boeken weggelegd.

Jan has without read these books put away

With respect to this property, it is important to note that some authors do not
consider these facts conclusive for determining the type of movement in-
volved in scrambling (see e.g. de Hoop 1992, Zwart 1993, Neeleman 1994,
among others), either because there are constructions in which subjects of
passives license parasitic gaps (cf. de Hoop 1992), or because parasitic gaps
are marginal in Dutch, or because there are cases in which parasitic gaps
appear to be licensed by a constituent in an A-position (other than subjects
of passives) (cf. (44) from Webelhuth 1989). The problem with (44), which
exemplifies what came to be known as Webelhuth’s paradox, is that the QP
jeden/everyone is in a position where it both licenses the parasitic gap and
A-binds a pronoun (contained in the same clause that contains the parasitic
gap). Note that the fact that the parasitic gap and the pronoun are in the
same clause is really paradoxical, and cannot be solved by means of two-step
movement, as proposed in Vanden Wyngaerd (1989) and Mahajan (1990).

(44)

German (Webelhuth’ paradox):

weil Maria jeden

ohne pg anzuschauen seinem Nachbarn t

because Maria everyone without to-look-at his neighbor

vorgestellt hat
introduced has

‘because Maria introduced everyone to his neighbor without looking
at him’

Objects in VOS: Scrambling in European Portuguese

Here, I will take the opposite approach to this problem, ignoring the A-bind-
ing facts. For a solution for this problem, I refer the reader to Lee and
Santorini (1994). One of the reasons for this option is that there is not a clear
description of what the WCO and WCO repair facts are, and, moreover, pure
A-positions (meaning absolute subjects and not subjects derived form object
positions never license parasitic gaps as (65) illustrates (see also Diesing
1995, Vikner 1994 and Bobaljik 1995 for similar ideas).

(45)

*Jan heeft zonder pg te zien het boek gelezen.
Jan has without to see the book read

I emphasize, though, that this is a purely methodological option, due to the
unclarity of binding as a diagnostic. It is very likely that whenever there is a
clear description of the facts, the methodology will have to be reviewed.

d) Though this property is not available in Dutch, but only acceptable in
German, it has been observed that scrambling may move an object across a
subject:

(46)

German:

weil den Patienten der Arzt besucht hat.
because the patient-ACC the doctor-NOM visited has

‘because the doctor visited the patient’

This difference between German and Dutch may be explained assuming
with Diesing (1992) that subjects may stay in Spec, VP in German. If this
option is not available for Dutch, the contrast between the two languages
follows. It is not a consequence of the availability of a different type of
scrambling in German, but rather an effect of the availability of Spec, VP as
a legitimate position for the subject only in German.

e) Another well-known property of Dutch and German scrambling are the
semantic/pragmatic requirements of this operation (the scrambled con-
stituent has to be specific or non-novel, cf. de Hoop 1992; Diesing 1995;
among others).

Postverbal subjects: syntax and discourse – VSO and VOS orders

Having described some of the properties of Dutch and German scrambling,
let me now move to a description of the Portuguese data and check out the
similarities with Germanic scrambling.

The problem with identifying scrambling in Portuguese is that this lan-

guage has V-to-I movement, which makes it difficult to identify the position
of the object. Since the verb is not in VP anymore, the position of the object
is ambiguous between its base position or the scrambled position. That is,
an SVO sentence may be analyzed as either (47a) without scrambling, (47b)
with the object adjoined to VP, or (47c) with the object moved to
Spec,AgrOP:

(47)

a. [

S [

I’

V [

O]]]

b. [

S [

I’

V [

O [

]]]]

c. [

S [

I’

V [

AgrOP

O [

]]]]

I will assume, once again, that given their properties monosyllabic adverbs
provide a good test to check whether the object is inside or outside VP.
Recall the distribution of monosyllabic adverbs in English:

(48)

a. John read the book well.

b. *John well read the book.

c. John looked hard at those pictures.

d. *John looked at those pictures hard.

In Costa (1996, 1998), the English pattern was analyzed by assuming, fol-
lowing Pesetsky (1989) and Johnson (1991), that both verbs and nominal
objects move out of the VP overtly: NPs move to Spec,AgrOP and the verb
to the first functional projection immediately above AgrOP. The sequence
PP-Adv is bad because these adverbs do not right-adjoin to VP and because
differently from nominal complements, PPs do not need to move to license
Case.

The behavior of these adverbs is not an idiosyncrasy of English. As

already mentioned in the discussion of subject positions, the same basic
generalizations hold for Portuguese: a paradigm almost equivalent to (48)
may be built for this language:

Objects in VOS: Scrambling in European Portuguese

(49)

a. O Paulo fala francês bem.

Paulo speaks French well

b. *O Paulo bem fala francês.

Paulo well speaks French

c. O Paulo olha bem para aqueles quadros.

Paulo looks well at those pictures

d. *O Paulo olha para aqueles quadros bem.

Paulo looks at those pictures well

e. O Paulo olha para aqueles quadros BEM.

Paulo looks at those pictures well

The crucial difference between English and Portuguese is that only in the
latter is the order V-Adv-NP possible:

(50)

a. O Paulo fala bem francês.

Paulo speaks well French

b. *Paul speaks well French.

The fact that the nominal complement may either precede or follow the
adverb that marks the left edge of the VP is quite similar to the distributional
pattern of nominal complements in scrambling languages: in Dutch and
German, the objects may appear either to the left or to the right of a VP-
adjunct.

The only difference is that these languages are V-final, hence the

adjacency between complement and verb is visible in the case of non-scram-
bling. In a language with V-to-I movement, there is only adjacency if there
is no other adjunct in between the inflected verb and the object, as in (51):

(51)

O Paulo fala sempre francês bem.
Paulo speaks always French well

Given the distribution of monosyllabic adverbs and the possible orderings
between the verbal complement and the adverb, I will assume that the word
order in (52a) traces a scrambling configuration, while the word order in
(52b) indicates that the object occupies its base-generated position:

(52)

a. [

V [ scrambled Object [

Adv [

Obj

b. [

V [

Adv [

non-scrambled Object

Postverbal subjects: syntax and discourse – VSO and VOS orders

Having established that adverb placement may be used as a test to identify
the structural position of the complements, let me now go through the prop-
erties identified for scrambling in Germanic, and see the similarities between
scrambled objects in the latter and pre-adverbial or pre-subject objects in
Portuguese.

a) Novelty effects (de Hoop 1992; Büring 1997; Webelhuth 1989; Diesing

1995; among others):

The first property of the XP-Adv order in European Portuguese I want to
consider is the novelty effects also observed in Dutch/German scrambling.
Checking for felicitous answers for questions, we see that in the answer to
(53aA) the complement yields new information, and the order of the con-
stituents has to be Adv-NP. On the contrary, if it is the adverb that intro-
duces new information, the NP that is repeated from the question has to be
scrambled, as (53b) shows.

(53)

a. A: Há alguém aqui que fale bem francês ou inglês?

Is there anyone here who speaks French or English well?

B: #Não, mas o Paulo fala alemão bem.

No, but Paulo speaks German well

B’: Não, mas o Paulo fala bem alemão.

b. A: Como é que o Paulo fala francês?

How does Paulo speak French

B: O Paulo fala francês bem.

the Paulo speaks French well

B’: #O Paulo fala bem francês.

This patterns like Dutch, where the elements that are new information, appear
rightmost. The only difference is that rightmost in Dutch means being to the
left of the non-inflected verb:

Objects in VOS: Scrambling in European Portuguese

(54)

Dutch:
a. A: Wanneer heeft Jan Marie gekust?

When has Jan Marie kissed

B: Jan heeft Marie gisteren gekust.

Jan has Maries yesterday kissed.

B’: #Jan heeft gisteren Marie gekust.

b. A: Wie heeft Jan gisteren gekust?

Who has Jan yesterday kissed

B: Jan heeft gisteren Marie gekust.

Jan has yesterday Marie kissed

B’: #Jan heeft Marie gisteren gekust.

b) Parasitic Gap licensing.

Another similarity between European Portuguese on the one hand and Dutch
and German on the other has to do with the fact that scrambling licenses
parasitic gaps. Prima facie, this appears to be hard to test, since it is known
that null objects exist in Portuguese (see Raposo 1986). However, Bianchi
& Figueiredo (1994) and Menuzzi (1994) have shown that null objects (of
the type treated in Raposo’s work) can only occur if the antecedent is inani-
mate. This is exemplified in (55) and (56) below:

(55)

(Talking of [the car we have just seen]

(from Menuzzi 1994)

a. O José conhece [

a mulher [

que comprou ec

José knows the woman who bought (it)

b. O José [

ficou nervoso] [

porque a Maria comprou ec

]

José got nervous because Maria bought (it).

(56)

(talking of [Paulo]

a. *O José conhece [

a mulher [

que beijou ec

José knows the woman who kissed (him)

b. *O José [

ficou nervoso] [

porque a Maria beijou ec

]

José got nervous because Maria kissed (him)

Postverbal subjects: syntax and discourse – VSO and VOS orders

Bianchi and Figueiredo (1994) and Menuzzi (1994) conclude from the sensi-
tivity to islands in (56) that only in these cases is operator movement in-
volved. They disagree with respect to the proper characterization of the
empty category in (56). Nevertheless, what is relevant for the present dis-
cussion is that this difference allows differentiating between null objects
and gaps left by movement.

A parasitic gap has to be involved when an

animate antecedent is used, since null objects may not be animate, as the
examples above show. Accepting this, it is now possible to test for parasitic
gap licensing, excluding the possibility of being misled by results involving
a null object construction. If scrambling in Portuguese is like in Germanic,
it is expected that an object be able to license the parasitic gap from the
scrambled position, but not from the base-position. This prediction is borne
out. Like in German and Dutch, the in-situ object does not license a parasitic
gap contained in a left-adjoined PP:

(55)

a. * O Pedro conhece [mesmo sem nunca ter visto pg ] bem a Maria.

b. O Pedro conhece a Maria [mesmo sem nunca ter visto pg ].

The post-adverbial object in (55a) cannot license the parasitic gap, unless it
is scrambled.

c) Predicative APs and Small Clauses do not scramble.

Like in German and Dutch, Small Clauses and predicative APs do not
scramble. This is shown in (56) and (57) below:

(56)

a. O Paulo acha sempre a Maria simpática.

Paulo finds always Maria friendly

b. * O Paulo acha a Maria simpática sempre.

(57)

a. O Paulo é sempre muito simpático.

Paulo is always very nice

b. * O Paulo é muito simpático sempre.

In (56), it can be seen that a small clause (complement of the verb achar ‘to
find’) may not be scrambled to the left of the adverb, which modifies the

Objects in VOS: Scrambling in European Portuguese

matrix verb.

In (57), it is shown that a predicative AP may never scramble

to the left of an adverb modifying a predicative verb.

d) Scrambling does not feed binding.

Although I have not exemplified this property for Dutch and German (see
Déprez 1989, Webelhuth 1989, Vanden Wyngaerd 1989, Müller and
Sternefeld 1994, Corver and Riemsdijk 1994, among others), in Portuguese,
scrambling does not feed binding, patterning like other cases of A-bar move-
ment and unlike A-movement. This is exemplified in (58): the sentence (58a)
in which the subject contains a possessive anaphor not bound by the QP in
object position does not improve if the object is scrambled across the subject.

(58)

a. * Viu o seu

pai cada

criança.

saw his father each child

b. * Viu cada

criança o seu

pai.

saw each child his father
‘His father saw each child’

As mentioned above, I am not taking into account the binding facts, since I do
not think that it is possible to establish a parallel with Germanic. I wanted
nevertheless to add these data, to show that, in Portuguese, scrambling an
object across a subject does not feed binding. Note that (58b) contrasts with
(59), where A-movement is involved and the binding relation can be estab-
lished:

(59)

Cada criança foi vista pelo seu pai.
Each child was seen by his father

e) Scrambling across the subject.

This property cannot be really used as an argument in favor of the scrambling
analysis, since it corresponds to the phenomenon I want to analyze: the order
VOS. Nevertheless, I would like to note that deriving VOS in EP in terms of
scrambling of the object across the in-situ subject is not anomalous, since this
word order also exists in German, being derived in the same way.

Postverbal subjects: syntax and discourse – VSO and VOS orders

From the evidence presented above and the similarities with German and
Dutch, I conclude that European Portuguese has scrambling of the Germanic
type. Moreover, given its properties, scrambling seems to be A-bar move-
ment, as has been argued for German and Dutch. I am aware that the
assumption that scrambling in German and Dutch is not A-bar movement is
quite controversial (see de Hoop 1992, Zwart 1993, Bobaljik 1995 for an A-
movement analysis). Here, I will follow Vikner’s (1994) conclusion that
there is a distinction between scrambling in West Germanic and Object
Shift in Scandinavian. Vikner (1994) compares the constructions in both
language groups, arguing that scrambling is A-bar movement, while object-
shift is A-movement. The table below lists the properties associated with
each construction:

Vikner’s (1994) observations summarized in the left-hand columns of (T1)
apparently provide robust evidence in favor of the claim that there are two
types of object movement out of VP: West Germanic scrambling, sharing
properties with O-Adv orders in EP displays A-bar properties, while Scandi-
navian object shift displays A-properties. I will assume these results, opting for
representing scrambling as adjunction to VP, as in (60a), and object-shift of
full NPs, like in Icelandic, as movement to a Case A-position (Spec,AgrOP),
as in (60b). These representations will account for the A- vs A-bar properties
discussed above.

Objects in VOS: Scrambling in European Portuguese

Table 1.

German/Dutch

Scandinavian

Parasitic gap licensing

yes

Movement of NPs

yes

(Icelandic and Faroese:

NPs; the other languages:

pronominal NPs)

Movement of PPs

yes

FQ-licensing?

no (?)

yes

Order Adv-NP-Adv?

yes

Interaction with

yes

V-movement?

(60)

a. [

[

Obj [

Adv [

Obj

b. [

[

AgrOP

Obj [

Adv [

Obj

The consequences of establishing the existence of scrambling in European
Portuguese are the following: First, scrambling may not be seen as an OV
phenomenon anymore (as suggested, for instance, in Vikner 1994). Further-
more, the lack of scrambling in other VO languages such as English becomes
an interesting question. Finally, European Portuguese can be considered
another language allowing for the investigation of what the reason for scram-
bling is. Below, an explanation will be offered as to why scrambling does not
occur in a language like English.

Apart from these considerations, note that I have already provided evi-

dence in favor of the analysis proposed in section 2 for Portuguese VOS. In
order to strengthen the analysis, it is important to compare it with an alterna-
tive analysis available in the literature, according to which, in VOS sentences,
the subject is in Spec,IP.

There are two competing analyses for VOS word orders that may be found

in recent literature. The Scrambling analysis argues that this word order is
derived via movement of the verb to I and scrambling of the object across
the subject left in its base-generated position. The configuration arrived at
may be as in (61a) or (61b), depending on the assumption regarding the land-
ing site of the object (Spec,AgrOP or adjunction to VP). This is the analysis
outlined above. Proponents of this analysis include Zubizarreta (1995),
Ordóñez and Treviño (1995), for Spanish, Alexiadou and Anagnostopoulou
(1998), for Greek, and Costa (1997, 1998) for Portuguese.

(61)

a. IP

I AgrOP

I VP

Obj AgrO’

Obj VP

AgrO VP

Subj V’

2 2

Subj V’

Obj

Postverbal subjects: syntax and discourse – VSO and VOS orders

This analysis raises the question of knowing how the subject in Spec,VP is
assigned or checks Case. So far, I offered no solution for this problem.

An alternative analysis for this word order in Romance languages is

found in work following Kayne (1998). Let us term this type of analysis
Remnant movement. According to the Remnant movement analysis, VOS
word orders are arrived at by moving the subject to Spec,IP, or to a left-
peripheral functional projection, and moving the remnant constituent TP or
VP, containing the trace of the subject to the specifier position of another
functional category higher than the one where the subject has moved to. The
configurations obtained are as in (62a,b). Proponents of this type of analysis
include Kayne and Pollock (1998), for stylistic inversion in French, Ordóñez
(1997), Zubizarreta (1998a), and Bok-Benema (1998), for VOS in Spanish,
and Ambar and Pollock (1998) for VOS in interrogative contexts in French
and Portuguese.

(62)

FP1

F’

F’1

2 2

5 2

Subj

V O F1

FP2

5 2

V O

Subj I’

Subj F’2

I t

T/VP

F2 t

As explicitly emphasized by Zubizarreta (1998a), this type of analysis is ad-
vantageous with respect to the scrambling analysis, since the Case problem
finds a solution. In either configuration in (62), the subject has moved or
landed in Spec,IP where it may be assigned nominative Case. However, this
analysis is not exempt of problems. I will now turn to presenting empirical
arguments listed in Costa (2002b) that seem to disfavor the remnant move-
ment analysis, comparing it with the scrambling analysis. It will be shown
that the problems raised to the remnant movement analysis do not arise
under a scrambling analysis for VOS.

The argumentation will be based on an examination of the following

aspects of VOS sentences:

Objects in VOS: Scrambling in European Portuguese

A – it will be tested whether it can be argued independently for the mobil-

ity of the remnant VP/TP;

B – Adverb positioning;
C – Floating quantifiers;
D – Pronominal doubling and question tags;
E – The discourse function of the subject;
F – Scope and c-command;
G – Clitics;
H – Properties of the object.

For each of these properties, it will be shown that the remnant movement
analysis either makes wrong predictions or offers no explanation.

A. TP \VP mobility

Consider a sentence like (63) below with VOS word order:

(63)

Leu o livro o Paulo.
read the book Paulo

According to the remnant movement analysis, the constituent TP or VP leu o
livro (read the book) is moved to the left of the DP subject. Arguably, it
should be possible to find independent evidence for the mobility of this con-
stituent. Indeed for a sentence like (83), it is possible to find this type of
independent motivation. The VP or TP containing the verb and the object
may be clefted (cf. 64), or preposed (cf. 65):

Cleft:

(64)

Foi ler o livro o que o Paulo fez.
it was read the book what Paulo did

Preposing:

(65)

O Pedro disse que leria o livro o Paulo, e ler o livro o Paulo leu.
Pedro said that would-read the book Paulo, and read the book Paulo did

However, if the verbal form is complex, the evidence for mobility of the
constituent TP or VP is not so compelling. For a sentence like (66), it must
be argued that the constituent containing the sequence Auxiliary-Main
Verb-Object is moved to the left of the subject:

Postverbal subjects: syntax and discourse – VSO and VOS orders

(66)

Tinha lido o livro o Paulo.
had read the book Paulo

However, there is no evidence for movement of this constituent. The con-
stituent containing the auxiliary verb, the main verb and the object cannot
be clefted (cf. 67) nor preposed (cf. 68):

Cleft:

(67)

*Era ter lido o livro o que o Paulo tinha feito.

it was have read the book what Paulo had done

Preposing:

(68) ??O Pedro disse que teria lido o livro o Paulo, e ter lido o livro ele tinha.

Pedro said that would-have read the book Paulo, and have read the
book he had.

Thus, there is no clear evidence that the material preceding the subject
forms a constituent that can be moved. The fact that some sequences VO or
Aux-V-O are not movable is not problematic for a scrambling analysis, since
no claims are made concerning the necessity of moving this constituent.

Under the scrambling analysis, the only claim that is made is that the

object is moved to the left of the subject. It can be argued that an object may
be moved independently of the complexity of the verbal form. For a pair of
sentences like those in (69), it can be shown that the object may be clefted
(cf. 70) or preposed (cf. 71) independently of the material that precedes the
subject.

(69)

a. Leu o livro o Paulo.

read the book Paulo

b. Tinha lido o livro o Paulo

had read the book Paulo

Cleft:

(70)

a. Foi o livro o que o Paulo leu.

it was the book that Paulo read

b. Era o livro o que o Paulo tinha lido.

it was the book that Paulo had read

Objects in VOS: Scrambling in European Portuguese

Preposing:

(71)

a. O livro, leu o Paulo.

the book, read Paulo

b. O livro, tinha lido o Paulo.

the book had read Paulo

In short, postulating that VOS word order is derived via movement of the
constituent containing V and O is problematic, since evidence for the mobil-
ity of this constituent is not very strong.

B. Adverb placement

The second problematic set of data for the remnant movement analysis
comes from the distribution of adverbs. Let us consider again the placement
of monosyllabic adverb in VOS sentences: as shown above, and repeated
under (72), the adverb may occur in two positions: either in between the
object and the subject or in between the verb and the object:

(72)

a. Leu aqueles livros mal o Paulo.

read those books bad Paulo

b. Leu mal aqueles livros o Paulo

read bad those books Paulo

It was observed that these adverbs have a very restricted distribution. Never-
theless, they appear in two possible positions in a VOS sentence. Under a
remnant movement analysis, there is no clear explanation for this pattern.
Why should the adverb have a less restricted distribution when there is move-
ment of the constituent containing it? The scrambling analysis offers a natural
explanation for this behavior. Under this analysis, the object is scrambled.
The list of properties associated with scrambling led us to the conclusion
that scrambled objects in EP adjoin to VP. If this is true, it is expected that
there is no fixed order between two VP-adjuncts. V-O-Adv-S and V-Adv-O-S
are the expected word orders. The two possibilities are illustrated in the
structures in (73) and (74):

Postverbal subjects: syntax and discourse – VSO and VOS orders

(73)

2
I VP

Obj VP

Adv VP

Subj V’

Obj

(74)

2
I VP

Adv VP

Obj VP

Subj V’

Obj

C. Floating quantifiers

Consider now the behavior of floating quantifiers. As it is well-known, float-
ing quantifiers are possible after subject movement to Spec,IP (Sportiche
1988, Koopman and Sportiche 1991). In EP, floating quantifiers behave as
expected: they may surface in any post-subject position, tracing the path of
the subject from Spec, VP to Spec,IP.

(75)

a. Os meninos todos tinham lido o livro.

the children all had read the book

b. Os meninos tinham todos lido o livro.

the children had all read the book

c. Os meninos tinham lido todos o livro.

the children had read all the book.

Objects in VOS: Scrambling in European Portuguese

Under the proposal made by the remnant movement analysis, the subject
moves to Spec,IP, or higher, and the remnant constituent containing its trace
moves to its left. Note that there is no problem in moving a VP or TP con-
taining a floating quantifier to the left of a subject. As shown in (76), when
a VP is clefted, it may contain a floating quantifier:

(76)

Foi ler todos o livro o que os meninos fizeram.
it was read all the book what the children did

Considering the assumption of the remnant movement analysis and the fact
that moved VPs or TPs may contain floating quantifiers, it is legitimate to
claim that the remnant movement analysis predicts that floating quantifiers
surface before the subject in VOS sentences. However, this prediction is not
borne out, as shown by the ungrammatical sentences in (77):

(77)

a. *Tinham todos lido o livro os meninos.

had all read the book the children

b. *Tinham lido todos o livro os meninos.

had all read the book the children.

c. Tinham lido o livro todos os meninos.

had read the book all the children

The only possibility for the floating quantifier is to occur adjacent to the
subject, that is, in a non-floating position. If we now check the predictions
made by the scrambling analysis, it is possible to see that this problem does
not arise. Under the scrambling analysis, the subject stays in Spec, VP,
therefore floating quantifiers are not predicted to surface.

D. Topic doubling and question tags

At this point, it is important to recall the difference between VOS with flat
intonation vs VOS with a right-peripheral subject. Recall that only the latter
can be doubled by a pronoun, and that they may surface after a question tag.
The behavior of subjects in the latter type of construction is not different
from the behavior of other constituents. Topic information appearing sen-
tence-finally after a pause can be doubled by a clitic or a pronoun, as shown
in (78). This is true for direct objects, indirect objects and subjects. The only

Postverbal subjects: syntax and discourse – VSO and VOS orders

difference between the latter and the internal arguments of the verb is that the
absence of subject clitics in Portuguese forces doubling by a strong pronoun.

(78)

a. Direct object:

O Paulo leu-o, esse livro.
Paulo read it, that book

b. Indirect object:

O Paulo deu-lhe o livro, à Maria.
Paulo gave her the book to Maria

c. Subject:

Ele leu o livro, o Paulo.
he read the book, Paulo

Clause-internal material cannot be doubled. This is true for direct objects, as
in (79a), in which the direct object appears right after the doubling clitic with-
out a pause, and left-adjacent to a marker of the right-edge of the sentence.
It is also true for the indirect object, as in (79b), in which the indirect object
appears before the direct object and adjacent to the doubling clitic, without
being preceded by a pause. The fact that doubling is only possible with
clause-external material enables a distinction between this pattern of clitic
doubling and the pattern of clitic doubling exhibited by other languages, in
which the clitic may double clause-internal elements. Based on the contrasts
between (78) and (79), it is legitimate to assume that the pronominal elements
occupy the thematic positions, and the DPs are peripheral to the sentence.

(79)

a. Direct object:

*O Paulo leu-o esse livro ontem.
Paulo read it that book yesterday

b. Indirect object:

*O Paulo deu-lhe à Maria o livro.
Paulo gave her to Maria the book

Let us now consider the case of subjects. Like the direct and indirect object,
subjects can only be doubled if they follow a question tag or any other marker
of the right edge of the sentence, no matter whether the pronoun is preverbal
or postverbal. Thus, they behave like the other sentence constituents.

Objects in VOS: Scrambling in European Portuguese

(80)

a. Ele leu o livro, não leu?, o Paulo

he read the book, didn’t he?, Paulo

b. Leu o livro ele, não leu?, o Paulo.

read the book he, not read, Paulo

c. *Ele leu o livro o Paulo, não leu?

he read the book Paulo, didn’t he?

Let us now consider the relevance of this behavior for the analysis of VOS
sentences. In a specific implementation of the remnant movement analysis
for accounting for stylistic inversion in French, Kayne and Pollock (1998)
argue that, in VOS sentences, the subject is in Spec,TopP. As we have seen
above, topic subjects can be doubled. Even if they are sentence-initial, they
can be doubled in some contexts (see section 2):

(81)

O Paulo, ele leu esse livro.
Paulo, he read that book

The question that arises for Kayne and Pollock’s analysis is why the alleged
topic cannot be doubled in VOS order without a break, as shown above and
repeated in (82).

(82) *Ele leu o livro o Paulo.

he read the book Paulo

The remnant movement analysis for VOS proposed in Kayne and Pollock
(1998) offers no clear answer for the ungrammaticality of (82). Under the
scrambling analysis, this pattern is expected. The subject is clause-internal,
it is in Spec, VP. Since it is not in a topic position, doubling is not expected.

E. Discourse function of the subject

So far, I have not explicitly discussed the relation between the several word
orders and discourse. It will be shown, however, that the sentence-final sub-
ject in VOS word orders is a focus. This may be tested in the following
question-answer pair:

Postverbal subjects: syntax and discourse – VSO and VOS orders

(83)

A: Quem leu o livro?

who read the book

B: Leu o livro o Paulo.

read the book Paulo

In Portuguese, contrastive foci are marked with stress (Frota 1998). Crucially,
a constituent may be simultaneously information and contrastive focus. In
other words, it may answer a question and be marked with strong stress.
The difference between (83) and (84) is that the answer in (84) implies that
only Paulo and no-one else read the book.

(84)

A: Quem leu o livro?

who read the book

B: Leu o livro O PAULO.

read the book Paulo

Having noted this discourse function of the subject for Spanish VOS orders,
Ordóñez (1997) proposes that, since the subject in VOS is focused, it is
moved to a left-peripheral functional projection codifying focus information:
Focus Phrase. The remnant constituent yielding old information (in (83) and
(84), the constituent containing the verb and the object) is moved to a topic
position to the left of the focus position.

Although it may account for the discourse function of the subject, this

analysis is problematic. In his study of the left-periphery, Rizzi (1997) shows
that wh-phrases and displaced focused constituents are in complementary
distribution. The prediction made by Ordóñez’s implementation of the rem-
nant movement analysis is then that wh-phrases in VOS word orders should
be ungrammatical. As shown below, this prediction is not borne out:

(85)

A quem deu o livro O PAULO?
to whom gave the book Paulo

Under Kayne and Pollock’s (1998) analysis, it is proposed that the subject is
in Spec,TopP. Assuming this, the problem raised by Ordóñez’s analysis does
not arise, but the relation with the actual discourse function of the subject is
left unexplained.

Ambar and Pollock (1998) provide a good argument in favor of Kayne

and Pollock’s analysis. They note that in interrogative contexts with VOS

Objects in VOS: Scrambling in European Portuguese

order, the sentence-final subject cannot be a pronoun or an indefinite pronoun.
Crucially, these elements cannot be topicalized in Portuguese. The similarity
between the type of subject that occurs sentence-finally in VOS interrogatives
and the type of subject that can be topicalized provides, thus, clear evidence
in favor of the analysis according to which the sentence final subject in
VOS is in Spec,TopP.

(86)

a. A quem deu o livro o João?

to whom gave the book João

b. *A quem deu o livro ele?

to whom gave the book he

c. *A quem deu o livro alguém?

to whom gave the book someone

Although the argument is a good one, the analysis proposed in terms of topi-
calization and remnant movement is not exempt of problems. First, note that
the sentence-final position for the subject is not reserved for topics. As shown
in (85), other discourse functions, such as contrastive focus, are available in
this position.

More problematic for the analysis proposed is that there is no additional

independent evidence for the claim that the subject is topicalized. One expects
to find properties of topicalization in VOS word orders in interrogative con-
texts. Duarte (1987, 1996) identifies several properties of topicalization in
European Portuguese that may be used to test Ambar and Pollock’s claim.

Duarte shows that topicalized elements license parasitic gaps. This is illus-

trated in (87a). Accordingly, the sentence-final subject in VOS interrogatives
should be able to license parasitic gaps, since it is assumed to occupy a topic
A-bar position. However, parasitic gaps are not found in this context (cf.
87b).

No parasitic gap licensing:

(87)

a. Esse artigo, o João elogiou sem ter lido.

that article, João praised without having read

b. *Quando foram criticados sem o Paulo ter lido os artigos?

when were criticized without Paulo have read the articles

If there is a main clause and an embedded clause, constituents from each of
these clauses may be topicalized, as shown in (88a’). The prediction made by

Postverbal subjects: syntax and discourse – VSO and VOS orders

Ambar and Pollock’s analysis is thus that both subjects of the main and the
embedded clause may be topicalized, as in (89b), and the remnant IPs move
to the functional categories to the left of the landing site of the subject, as in
(89c). This would derive a word order in which both subjects would appear
adjacent at the end of the sentence. However, this word order is ungrammati-
cal (cf. 88b).

No multiple topicalization:

(88)

a. O Pedro disse à Maria que o Paulo ia à praia nesse dia.

Pedro said to Maria that Paulo would-go to the beach on that day

a.’ À Maria, o Pedro disse que, à praia, o Paulo ia nesse dia.

to Maria Pedro said that to the beach Paulo would go on that day

b. *Quando disse que ia à praia o Paulo o Pedro?

When said that would-go to the beach Paulo Pedro

(89)

a. wh-movement

[

Quando [

[

TopP

[

o Pedro disse [

que [

[

TopP

[

o Paulo ia à

praia t

]]]]]]]]

b. Subject topicalization in both clauses

[

Quando [

[

TopP

o Pedro [

disse [

que [

[

TopP

o Paulo [

ia à praia t

]]]]]]]]

b. Remnant IP-movement:

[

Quando [

[

disse [

que[

[

ia à praia t

][

TopP

Paulo t

]]]] [

TopP

o Pedro t

]]]

A final problem for Ambar and Pollock’s analysis comes from the lack of
parallelism with another property of topicalization. As shown in (90), con-
stituents other than subjects can be moved to the topic position of the main
clause. This is an option available for the subject, as shown in (90b).

(90)

a. O Pedro contou à Maria que o Paulo vai ao cinema.

Pedro told Maria that Paulo goes to the movies

b. O Paulo, o Pedro contou à Maria que t vai ao cinema.

c. À Maria, o Pedro contou t que o Paulo vai ao cinema.

d. Ao cinema, o Paulo contou que o Paulo vai t.

Objects in VOS: Scrambling in European Portuguese

Since the topicalizations in (90) are possible, the prediction made is that in
an interrogative sentence, any of these constituents of the embedded clause
may surface in sentence-final position. However, as (91) illustrates, only the
(allegedly topicalized) subject may occur in this position. The remnant move-
ment analysis incorrectly predicts the grammaticality of (91b) and (91c).

(91)

a. Quando t contou à Maria que o Paulo vai ao cinema o Pedro?

when told Maria that Paulo goes to the movies Pedro

b. *Quando o Pedro contou à Maria que t vai ao cinema o Paulo?

c. *Quando o Pedro contou t que o Paulo vai ao cinema à Maria?

Under the scrambling analysis, the lack of parallelism with topicalization
constructions is not a problem. If the subject stays in Spec, VP, it may be
interpreted as the focus of the sentence, since it is in the rightmost/most
embedded position, where it may be assigned sentence nuclear stress (Nespor
and Vogel 1986, Cinque 1993, Zubizarreta 1998, among others), an issue we
will return to in the next chapter. The interpretation of the subject as focus
will follow from the interaction between prosody and syntax, rather than from
an association with a specific functional projection. The topical properties
of sentence-final subjects in interrogative contexts are not entirely unex-
pected nor constitute a problem for the scrambling analysis: the intonation
of interrogative sentences is different; hence no prediction regarding the
discourse function of the subject is made. As shown in Mateus et alii (1989),
wh-questions have an initial rising tone and lowering at the end, which pre-
dicts that constituents that can be downstressed (like topics) will tend to
emerge sentence-finally.

F. Scope and c-command

In VOS sentences, the object c-commands the subject. This statement is
confirmed by the fact that a quantified object may scope over the subject, as
in the ambiguous sentence in (92), and by the principle-C effects induced in
VOS sentence, as in (93) and (94). In both cases, an object preceding the
subject seems to c-command it.

Quantifier scope:

(92)

Leram um livro dois alunos.

(O>S, S>O)

read a book two students

Postverbal subjects: syntax and discourse – VSO and VOS orders

Principle-C effects:

(93)

SVO:
a. O irmão do Paulo

viu-o

the brother of Paulo saw him

VOS:
b. *Viu-o

o irmão do Paulo

saw him the brother of Paulo

(94)

SV IO DO:
a. O irmão do Paulo

deu-lhe

o livro.

the brother of Paulo gave him the book

V IO DO S:
b. *Deu-lhe

o livro o irmão do Paulo

gave him the book the brother of Paulo

These facts based on quantifier scope and violation of principle-C are prob-
lematic for the remnant movement analysis. According to this analysis, a
configuration is obtained in which the object is a constituent of the moved
constituent. The configuration obtained is as in (95). Note that, in such a
structure, the object does not c-command the subject:

(95)

[

TP\VP

V O ][

S....]]

The prediction made by the remnant movement analysis is, thus, that objects
in VOS sentences should not take scope over the subject, and that there
should be no violation of principle-C, since, in (95), the object does not c-
command the subject. Note that reconstruction would not help, since the
object would never c-command the subject.

The prediction made by the scrambling analysis is that object-wide scope

and violations of principle-C are found. The configuration obtained after
scrambling of the object across the subject is like the one in (96). In this con-
figuration, the object c-commands the subject:

(96)

[

V [

O [

S ]]]

The ambiguity of quantifier scope may be explained by the scrambling analy-
sis, assuming that the scrambled object is able to reconstruct into its base
generated position, in which it is c-commanded by the subject.

Objects in VOS: Scrambling in European Portuguese

G. Clitics: enclisis vs. proclisis

As it is well-known, the distribution of clitics in European Portuguese may
be dependent on syntactic properties of the sentence (cf. Duarte and Matos
2000, among others). A clitic cooccurring with a preverbal non quantified
DP subjects is enclitic, as shown in (97):

(97)

a. O Paulo viu-o.

Paulo saw him

b. *O Paulo o viu.

If the subject is quantified, as in (98), proclisis is triggered:

(98)

a. Ontem todos os meninos o viram.

yesterday all the children him saw

b. *Ontem todos os meninos viram-no.

Recall that the remnant movement analysis proposes that, in VOS sentences,
the subject is in Spec,IP or at least has passed through this position. The pre-
diction made by this analysis is that a quantified subject in VOS sentences
should trigger proclisis. However, this prediction is not borne out. Enclisis
is the pattern found in VOS sentences:

(99)

a. *Ontem o deram à Maria todos os meninos.

yesterday it gave to Maria all the children

b. Ontem deram-no à Maria todos os meninos.

Note that the argument based on the distribution of clitics also holds for wh-
phrases that remain in-situ. A moved wh-phrase triggers proclisis, as illus-
trated below:

(100) a. Quem o leu?

who it read

b. *Quem leu-o?

who read it

Postverbal subjects: syntax and discourse – VSO and VOS orders

A subject wh-phrase in VOS context does not trigger proclisis:

(101) a. *O deu à Maria quem?

it gave to Maria who

b. Deu-o à Maria quem?

gave it to Maria who

These patterns of clitic placement, that are problematic for the remnant move-
ment analysis, are expected under the scrambling analysis. According to the
latter, subjects in VOS contexts are in their base-generated position, Spec,
VP. Since proclisis is only triggered by quantified subjects in Spec,IP, it is
predicted that clitics are enclitic in this context.

There is an objection that may be raised to this argument. Let us assume

with Kayne and Pollock (1998) and Ambar and Pollock (1998) that the sub-
ject in VOS is in a topic position. In such case, it could be argued that procli-
sis is not expected, since topicalization does not trigger proclisis, independ-
ently of the position of the subject:

(102) a. Esse livro, o Paulo leu-o.

that book, Paulo read it

b. Esse livro leu-o o Paulo

that book read it Paulo

However, if the topic is quantified, proclisis is triggered, as argued in
Raposo (2000):

(103) Muitos livros lhe leu o Paulo.

many books him read Paulo

We are now able to test whether it may be argued that proclisis is not trig-
gered because the subject is topicalized. The test case is the use of a quanti-
fied subject in VOS context. As shown in (104), a quantified subject in VOS
similar to the object in (103) does not trigger proclisis:

(104) a. Leram-lhe livros muitos meninos.

read him book many children

b. *Lhe leram livros muitos meninos.

him read books many children

Objects in VOS: Scrambling in European Portuguese

The fact that enclisis is always obtained independently of the properties of
the subject permits discarding the objection to this argument based on the
idea that subjects in VOS sentences are topicalized.

H. Properties of the object

The acceptability of VOS sentences varies depending on the type of object
of the sentence. So far, almost all examples contain definite or strong DPs
as objects. However, if the object is indefinite, the acceptability of the sen-
tence is not as good, and it clearly degrades with non-specific indefinite
DPs:

(105) a. ?Viu um gato o Paulo.

saw a cat Paulo

b. *Viu um homem qualquer o Paulo.

saw some man Paulo

c. ??Leu algo o Paulo

read something Paulo

This relation between the properties of the object and the degree of grammati-
cality of the sentence is problematic for the remnant movement analysis.
According to this analysis, the object is just a part of the moved constituent.
There is thus no clear reason for there to be sensitivity to the definiteness of
a subconstituent of the moved XP.

This problem does not arise under the scrambling analysis. The constituent

that is moved in order to yield the VOS order is the object. It is therefore not
surprising that there is some sensitivity to the definiteness of the moved con-
stituent itself. That scrambling is better with definite DPs than with indefinite
DPs is a well-known fact in Germanic languages (see de Hoop 1992 among
others). We thus find just a similar pattern in Romance.

Summing up, although it solves the Case problem, the remnant movement
analysis faces a high number of empirical problems that do not arise under
the scrambling analysis. Incidentally, we ended up finding some additional
arguments in favor of the scrambling analysis by comparing the two. Taking
into consideration the arguments listed above, I will not consider the remnant
movement analysis for the derivation of VOS word orders in EP.

Postverbal subjects: syntax and discourse – VSO and VOS orders

So far, nothing has been said concerning the trigger for scrambling configu-
rations. In recent literature, three proposals have been made to account for
the nature of scrambling:

a) Scrambling is A-movement and it is case-driven

b) Scrambling is A-bar movement:

i) Semantically-driven
ii) Prosodically-driven

In what follows, I review these three hypotheses, arguing in favor of the
prosodic-driven approach by showing that it is the one that provides the best
account of the data.

Several authors have suggested that scrambling is movement to Spec,-

AgrOP, hence driven by the requirement for objects to check accusative fea-
tures (Vanden Wyngaerd 1989, Mahajan 1990, de Hoop 1992, Zwart 1993,
Bobaljik 1995 among others).

Under this hypothesis, there should be no difference between Scrambling

of the type just investigated and Object-shift in languages like Danish,
Faroese, Icelandic, Norwegian and Swedish (Holmberg 1986, Webelhuth
1989, Vikner 1994, Bobaljik 1995, among others).

However, as mentioned

above, object-movement in object-shift languages more clearly displays A-
movement properties that scrambled objects in Dutch, German and Portu-
guese.

Two additional arguments may be considered for not considering scram-

bling Case-driven movement.

a) Adverbs that are subcategorized by verbs scramble in Dutch (Costa 1995:
131a) shows that a verb like live in the relevant interpretation is not gram-
matical if no argument is selected. The argument may be either specific
(106b) or non-specific (106c).

Adding another adverb to check the posi-

tion of the subcategorized one shows that the adverb can be scrambled
(106e,h). If a non-specific adverb is scrambled (106g,j), the sentences
become ungrammatical (under a non-specific reading).

(106) a. *Ik woon.

I live

b. Ik woon daar.

I live there

Objects in VOS: Scrambling in European Portuguese

c. Ik woon ergens.

I live somewhere

d. Ik woon steeds daar.

all-the-time

e. Ik woon daar steeds.

f. Ik woon steeds ergens.

g. *Ik woon ergens steeds.

h. dat ik daar steeds woon.

i. dat ik steeds daar woon.

j. *dat ik ergens steeds woon.

k. dat ik steeds ergens woon.

This pattern is very similar to the behavior of NPs. The null hypothesis is to
subsume both categories under the same phenomenon. Assuming, as usual,
that adverbs do not require case, the movement of the adverbs may not be
driven by case.

b) Another argument against case-driven scrambling comes from the be-
havior of subjects of unaccusatives in Portuguese: as hypothesized above,
subjects in Portuguese may either stay in Spec, VP or move to Spec,IP.
Allegedly, subjects of unaccusatives do not stay in Spec, VP but in their
object base position. This is not possible to determine for Portuguese, but it
has been noted for Italian by Pinto (1994,1997) that subjects of transitives
and intransitives may not be low in wide focus context, while subjects of un-
accusatives may be. Crucially, subjects of unaccusatives can also scramble,
as shown in (107c) by the order NP-adv.

(107) a. O Paulo chegou depressa.

Paulo arrived fast

b. Chegou depressa o Paulo, não chegou?.

arrived fast Paulo, not arrived

c. Chegou o Paulo depressa.

arrived Paulo fast

Postverbal subjects: syntax and discourse – VSO and VOS orders

In case scrambling were to reduce to case-driven movement, this pattern
would remain very mysterious: if the central property of unaccusative verbs
is that they do not assign accusative case to their objects, scrambling of the
object of unaccusatives should always be bad, since there would be object-
movement to an inert case-licensing projection. As (107c) shows, that is not
the case: movement of the subject is possible, but case is not licensed, hence
the target of movement is not a case-licensing position. Note that it cannot be
argued that the object has moved to a potential landing site for Case-licensing,
since that yields ungrammatical results in other types of constructions in
Portuguese. For instance, in (108), a subject in a raising construction may
not be stranded in the embedded infinitival Spec,IP (a potential case-licensing
position):

(108) *Parecem os meninos comer o bolo.

Seem(3ps-pl) the kids eat the cake

Given the counter-arguments presented here, and also those presented in
Vikner (1994), I conclude that scrambling is not (case-driven) A-movement.

If scrambling is not A-movement, the other option is that it is A-bar move-

ment, not considering the existence of mixed positions. Diesing (1995) and
Diesing and Jelinek (1995) take the fact that scrambling seems to have
semantic effects, more in particular specificity effects, following de Hoop
(1992), as evidence for an approach taking semantics to be the motivation
for scrambling. According to Diesing and Jelinek (1995), scrambling is a
type-mismatch repair operation. Following Diesing’s (1992) Mapping hy-
pothesis, they propose that definite NPs and QPs have to scramble out of VP
in order to escape existential closure. That is, VP is a domain for existential
material. Since definite NPs and QPs are not existential, they may not be
within that domain by the time a derivation reaches LF. Since definites may
appear inside or outside VP at surface-structure, they suggest that this opera-
tion may apply overtly or covertly, provided that at LF the right mapping may
be established: at LF, there may be neither definite NPs nor QPs within VP.

A problem they are left with is the fact that indefinites do not fall under the

scope of this operation, since there is no problem for them to be within the
domain of existential closure. However, indefinites may scramble, as in (109).
In order to solve this problem, Diesing and Jelinek (1995) propose that
scrambling of indefinites follows from the assumption that the Scope Con-
dition applies at S-structure in German. According to this condition, for an
element to have scope over another element, the former must c-command the

Objects in VOS: Scrambling in European Portuguese

latter. This condition is semantically defined, and must hold at LF. According
to the authors, the specificity of German is that the effects of the Scope
Condition are surface-true. This assumption is based on the observation that
different semantic representations result, depending on whether the indefi-
nite follows or precedes an adverb.

(109) German (from Diesing 1995)

a. weil Elly immer Lieder singt.

since Elly always songs sings

a.’ ALWAYS

[time (t)]

∃

song(x) & sings (Elly,x,t)

b. weil Elly Lieder immer singt

since Elly songs always sings

b.’ ALWAYS

[song(x)] sings(Elly,x)

Diesing and Jelinek’s proposal may be falsified if we find a case of scram-
bling of indefinites across an adverb which does not affect the semantic/
temporal representation of the sentence. Such a case would be problematic
for their analysis, since we would have a case in which the indefinite does
not need to take scope over the adverb and yet it scrambles. Such examples
in which no ambiguity between adverb and indefinite arises, and scrambling
of indefinites is still possible, exist and are exemplified in (110). The differ-
ence between (109) and (110) lies in the choice of adverb. If we control for
an adverb that has no effect on the semantics of the sentence, the prediction
would be that scrambling of indefinites becomes impossible. However, that
is not true. In (110), there is no necessary scope relation to be established
between the indefinite and the manner adverb, and yet the scrambling order
is possible:

(110) a. O Paulo fala bem uma língua.

Paulo speaks well a language

b. O Paulo fala francês bem.

Paulo speaks a language well

An interesting fact about the adverbs that Diesing (1992,1995) and Diesing
and Jelinek (1995) use has been brought up to my attention by Danny Fox
(p.c.). He notes that their examples involve adverbs like always (cf. 109)
whose semantics is sensitive to focus in the sense of Rooth (1985). If these

Postverbal subjects: syntax and discourse – VSO and VOS orders

effects are controlled for, scrambling appears to be still possible. Moving a
bit ahead, and supposing that scrambling and focus are related, it is not sur-
prising that changing the location of focus will change the semantics of
these sentences. Depending on the location of focus, the adverb will associate
with the indefinite or not (captured in Diesing and Jelinek’s analysis in terms
of scopal relations). If the adverb is sensitive to focus and the indefinite is
focused, the adverb will associate with the indefinite. If the focus is on some
other constituents, the meaning of the sentence will be different.

Under a strictly semantic account, leaving aside discourse effects, this

correlation between adverbs, scope and focus would be hard to identify. I take
this observation as the first piece of evidence in favor of the idea that scram-
bling is basically associated with the discourse structure of the sentence, in
the sense of Reinhart (1995).

The last hypothesis concerning the motivation for scrambling and the

one I am going to adopt was outlined in Reinhart (1995). Adopting Cinque’s
(1993) sentence stress algorithm, which states that the most deeply embedded
constituent of a sentence is the one that will receive the most prominent
stress, Reinhart observes that in a normal SOV sentence in Dutch, the object
is the most embedded constituent; hence it is the element bearing the most
prominent stress and gets interpreted as the focus. Obviously, a distinction
must be established between default stress and marked stress: the default
stress is the one that follows from the stress algorithms proposed in Nespor
and Vogel (1986), Cinque (1993) and Nash (1995), and normally associates
with clause-final position and correlates with embedding. Heavy prosodic
stress is required when the syntactic configuration obviates the application
of the default rule (e.g. stress shifting and heavy stress on focused subjects
in English, cf. Zubizarreta 1995 among others). Now, where does scrambling
come in? According to Reinhart, the motivation for scrambling is to be found
at the interface between PF constraints and discourse-structure. The motiva-
tion for scrambling the object is to make it escape the default focus stress.
Actually, Reinhart proposes that scrambling allows the verb to be focused.
However, this description of the facts seems not to be entirely correct, for
several reasons: in both Dutch and German, the verb may be focused without
resorting to scrambling: it is sufficient to shift the stress; scrambling across
an adverb focuses the adverb independently of the stress that the verb receives
(see Baart 1987, Büring 1997 among others for the relevant data). The prob-
lem with Reinhart’s initial generalization is that the difference between de-
fault and marked stress, though acknowledged, is not completely spelled
out. Default stress appears as the normal stress in declarative sentences,

Objects in VOS: Scrambling in European Portuguese

marked stress requires a marked intonation contour on the focused constitu-
ent.

Once these two are differentiated, it is easy to see what the generaliza-

tion is like: the complement of the verb scrambles in order to leave the
adjunct in the most embedded position where the latter can receive the most
prominent stress by default. Of course, for the complement to be able to do
this, it must not require stress itself. Scrambling is thus a twofold operation:
on the one hand it involves defocusing of the object, on the other hand it is
an operation that allows another constituent to be stressed.

Scrambling is

possible with stress on the verb, but then marked stress is needed. Data
showing these differences are presented in Baart (1987). Independently of
this, Reinhart’s analysis seems correct: scrambling is used for creating
appropriate focus configurations, namely to make the element bearing the
sentence nuclear stress escape it.

Let us see how this hypothesis is corroborated by Portuguese (the data

are similar in Dutch and German): in (111), it is expected that the object be
the focus of the sentence given that it is the new information requested in the
question. Indeed in this case the object cannot be scrambled, as the infelici-
tous sentence in (111Ba) shows:

(111) A: O que é que o Paulo fala bem?

what does Paulo speak well?

B: a. #O Paulo fala francês bem.

Paulo speaks French well

b. O Paulo fala bem francês.

Paulo speaks well French

If, on the other hand, the adverb is questioned, so that the object is old infor-
mation, it must not be in the position where it gets default stress; hence it
must be scrambled, allowing the adverb to receive the default stress:

(112) A: Como é que o Paulo fala Francês?

how does Paulo speak French?

B: O Paulo fala francês bem.

Paulo speaks French well.

#O Paulo fala bem francês.
Paulo speaks well French

Postverbal subjects: syntax and discourse – VSO and VOS orders

An interesting piece of data brought to my attention by Martin Honcoop is
the case in which a scrambled element is associated with focus, which, in
principle, seems to disconfirm Reinhart’s theory:

(113) Jan zei dat ik DE KRANT gisteren las, (en het boek vandaag)

Jan said that I the newspaper yesterday read, and the book today

This example is interesting because the NP is scrambled while it is also part
of the focus of the sentence. However, there are two important differences
between this case and the cases Reinhart describes: first, the scrambled NP
bears heavy stress, and second, the NP is not the focus of the sentence by
itself. When this sentence is uttered, the speaker intends to focus both the
NP and the adverb gisteren/yesterday. According to Reinhart (1995), the
focus set of a sentence is the constituent bearing the most prominent stress
plus everything it dominates. That is, in order to build an appropriate focus
set for (113), the speaker has to make the object dominate the adverb. The
only way to do that is by scrambling it. Now, according to the motivation
for scrambling that explored by Reinhart, the scrambled position will not
allow the object to receive the default stress of the sentence, since it is not
the rightmost constituent anymore. The only way to make it the focus of the
sentence is to resort to prosodic stress (see also Nash 1995).

Therefore, example (113), rather than disconfirming Reinhart’s theory

presents evidence in favor of it: scrambling is related to the need to create
appropriate focus configurations: furthermore, it is related to defocusing,
since the NP in the scrambled position cannot be the focus of the sentence
unless it bears a heavy stress. The only way in which (113) would be prob-
lematic would be if the scrambled NP were be the focus and the adverb
were not included in the focus set of the sentence. Under that interpretation,
(113) is infelicitous.

Summing up, in this section, I have shown that European Portuguese has

scrambling of the German/Dutch type, by looking at the properties of scram-
bling in Dutch and German, and by comparing them with the order NP-Adv
in European Portuguese. I have argued that scrambling is an A-bar movement
operation that adjoins the complement of the verb to VP. Finally, I discussed
the motivation for scrambling, arguing in favor of Reinhart’s (1995) prosodic/
discourse explanation of scrambling. I showed that this was the hypothesis
facing the least number of empirical problems.

Crucial for the discussion in this chapter is the conclusion that subjects

in VOS orders may be analyzed as being stranded in Spec, VP.

Objects in VOS: Scrambling in European Portuguese

Two questions remain unanswered: in which contexts are in-situ subjects
and how is the licensing of in-situ subjects done? These will be the topics of
the next chapter.

Postverbal subjects: syntax and discourse – VSO and VOS orders

4. Inversion and information structure

4.1. Introduction

In this chapter, it is argued that, in a language like European Portuguese,
properties of the information structure partially determine the distribution of
arguments. It will be argued that subjects stay in Spec, VP, only if they are
focused, and that in-situ subjects are licensed under Agree (cf. Chomsky
2001). Likewise, they may only raise to Spec,IP if they are not focused, and
when the entire sentence is focused. The view on focus to be adopted is that
information focus is rightmost, for prosodic convergence. This analysis of
focus leads to a number of questions regarding my assumptions concerning
the exact formulation of this constraint. A very common view on focus is to
assume that focused constituents must be licensed at the specifier of a func-
tional projection (Focus Phrase), which in most analyses appears at the left
periphery of the sentence (Brody 1990; Kiss 1995; Rizzi 1997). In the analysis
to be developed here, the exact reverse effect is obtained: focused constituents
appear at the rightmost position of the sentence. As it will be shown later, the
two approaches are not incompatible, but they operate on two different kinds
of discourse objects. Nevertheless, it is important to be clear about what type
of focus I am talking about, and why I am defending an in-situ analysis of
focus rather than an analysis that involves raising to a functional projection.
Therefore, I will present some arguments against a focus-movement ap-
proach.

Let me emphasize, for the sake of clarity, that I will be looking at infor-

mation focus: that is, focus that conveys new information (Dik 1978; Büring
1997; de Hoop and Swart 2000; Kiss 1996 among many others). This
remark is important, since the term focus is often used with a very wide
variety of meanings, rendering discussion of phenomena and identification
of scope of research quite difficult (see Givón 1990 for classification and
examples of several constructions involving focus-related aspects). I am
interested in this study in the focus that is normally associated with a high
pitch accent (Selkirk 1984, among others) and that can be identified in ques-
tion-answer pairs and correction contexts. This type of focus is often called
information focus. That is, it is focus in the sense that it conveys new infor-

mation without altering the truth value of the sentence (see Vallduví 1990,
among others). This excludes from the discussion contrastive focus and
focus involving uniqueness or exhaustive listing in the sense of Szabolcsi
(1981). In the last section of this chapter, I will present some tests to distin-
guish the several types of focus, and show that they do not necessarily inter-
sect. The behavior of contrastive focus will be shown to be different both in
distribution and in meaning. Below I provide examples of each of the con-
structions involving foci, in order to specify which will be the topic of this
chapter:

(1)

Focus Constructions:

a. In situ:

Comeu a sopa o Paulo.
Ate the soup Paulo

b. Syntactically marked focus (cf. Raposo 1994, among others)

Muito vinho o João bebeu!
a-lot-of wine João drunk

c. Focus-preposing (ungrammatical in Portuguese, OK in other

languages).

*ESSE LIVRO, o João leu.
That book João read

(1a) will be the topic of this chapter. I will not talk about (1b), which is dis-
cussed in Uriagereka (1995), Raposo (1994), among others, since its discourse
characteristics are different from (1a), and I will argue that the ungrammati-
cality of (1c) in European Portuguese is an additional argument for not con-
sidering focus-preposing as an argument for movement of foci.

4.2. Focus-movement?

Focus is represented in different ways in different languages. While a language
like English displays focus in situ, as in (2), a language like Hungarian
seems to require movement for licensing focused constituents:

(2)

I saw JOHN.

Inversion and information structure

(3)

a. AZ ÚJSÁGOT dobtam el.

(from Horvath 1995)

the newspaper threw-I away

b. *Eldobtam AZ ÚJSÁGOT.

Chomsky (1976) has argued, on the basis of the contrast in (4), that even in
English focused constituents need to be moved. Chomsky showed that
focus as in (4b) induces weak-cross-over effects just like other operators do
e.g. in (5):

(4)

a. His

mother saw John

b. *His

mother saw JOHN

(5)

*Who

does his

mother like?

According to Chomsky, the most natural way to explain the parallelism
between (4b) and (5) is to assume that, at LF, focused constituents move
establishing an operator-variable relation with their traces, yielding a structure
like (6b) for a sentence like (6a):

(6)

a. Mary loves JOHN.

b. [

JOHN

[

Mary loves t

]]

This accounts for the weak cross over effects in a rather natural way. The
LF-representation of (6b) is a violation of Koopman and Sportiche’s (1983)
bijection principle in the same way the overt syntax of (5) is:

(5)

JOHN

[

his

mother loves t

]]

(6)

Bijection Principle:

There is a bijective correspondence between variables and A-bar
positions.

Given (6), the problem with (5) is that ‘John’ is binding two variables: the
pronoun and the trace, yielding a violation of the bijection principle.

The weak-cross-over argument together with the existence of overt

focus-movement in Hungarian led linguists to assume that languages with
focus in-situ like English need to move focused constituents at LF.

Focus-movement?

Movement theories of focus have been criticized in Anderson (1972), Rooth
(1985), von Stechow (1990), among others, on the basis of their failure to
account in an appropriate way for lack of ECP and subjacency effects, multi-
ple foci and crossing paths at LF. I will briefly discuss these three problems.

First of all, Focus displays a lack of ECP and subjacency effects. If focus

involves movement, focusing should be impossible within a barrier for ex-
traction. That this is not true can be seen from the following examples:

(7)

Focus on an embedded subject (that-trace effects are expected):
Mary thinks that JOHN will go to the movies.

(8)

Focus within an adjunct:
Mary arrived late because she was SICK.

(9)

Focus within shifted Heavy NPs:
I read yesterday all the books MY teacher recommended.

(10)

Focus inside Wh-islands:
I wonder what to write with THIS PEN.

(11)

Focus inside a complex NP:
John announced a plan to steal FIVE cars tonight.

(12)

Focus within a coordinate structure:
John saw Mary and ALL the other students.

Any movement approach to focus predicts that these sentences should be
ungrammatical, since ECP effects (and subjacency) are assumed to be opera-
tive at all levels of representation (May 1985; Huang 1982; Koster 1987; Bayer
1995). If alternatively, one would assume that ECP is not operative at LF,
one would miss the empirical generalization that may be drawn considering
LF-movement, namely the similarities with overt movement (cf. the studies
cited above).

Another problem with assuming that focused constituents move at LF (on a
par with wh-movement) is the existence of multiple foci:

(13)

a. JOHN saw MARY.

b. John gave the BOOK to BILL.

Inversion and information structure

The problem with these sentences is different depending on the theory of
focus-movement assumed. If it is assumed, as Brody (1990) does, that
focused constituents enter a Spec,Head agreement relation with a Focus
head and that Hungarian is the overt counterpart of English, this would
imply that a head has multiple specifiers.

This in turn predicts a lack of

adjacency between the foci and the element lexicalizing the head (V in
Hungarian, according to Brody (1990)). However, multiple focus movement
is not possible in Hungarian:

(14)

Évat János várta a mozi elótt.

(from Kiss 1995)

Eve-acc John waited the cinem in-front-of
Eve, JOHN waited for her in front of the cinema.

If multiple specifiers are permitted, sentence (14) might incorrectly yield
the English interpretation in (15). An uniformization with wh- in situ is not
possible, since the postverbal PP in (14) is not interpreted as focus:

(15)

JOHN waited for EVE in front of the cinema.

The third problem comes from the fact that focus movement may create
crossing paths at LF. This argument obviously does not go through in theories
allowing for crossing paths (e.g. Chomsky 1995). I will not take a position
with respect to whether the theory should or should not allow for crossing
paths. In case it should not (Pesetsky 1982, Kayne 1994), these cases remain
problematic. The creation of crossing paths at LF arises in a focus move-
ment analysis for a sentence like (16):

(16)

Who did JOHN wait for?

This sentence should be ungrammatical because moving JOHN at LF would
induce a crossing path, which is a ill-formed path according to Pesetsky
(1982):

(16’) JOHN who did’t wait for t?

These three types of evidence should be enough to discard an analysis of
focus in terms of movement. However, there is still the evidence from
WCO-effects and the distribution of focus in Hungarian. I will return to
these problems after finishing the discussion of the focus-movement analyses.

Focus-movement?

If focus-movement exists, it is also important to determine the exact nature
of this movement: focus movement establishes an operator-variable relation,
but why should there be such a relation?, and if there is movement, where
do focused constituents move to?

Two widely accepted theories of focus movement are the ones advocated

by Brody (1990) and Horvath (1986,1995). In this section, I will discuss the
major aspects of both theories showing that they are not empirically adequate.

Brody (1990) argues that there is a Focus Phrase, where focused con-

stituents move to in order to satisfy the focus-criterion, parallel to Rizzi’s
(1991) wh-criterion:

(17)

Focus-criterion:

a. At S-structure and LF, the Spec of an FP must contain a +f-phrase.

b. At LF, all +f-phrases must be in an FP.

The source of cross-linguistic variation is then whether (17a) is satisfied
overtly (Hungarian) or covertly (English). As noted by Horvath (1995), this
approach to focus is not satisfactory since it does not account for the fact
that languages other than English and Hungarian codify focus in positions
that are not either the base-position or the leftmost position of the sentence.

Horvath (1995) proposes that focus is either assigned like any other gram-

matical feature (e.g. Case) by a non-lexical head or freely assigned. The need
for V-adjacency in Hungarian comes from the need to lexicalize the func-
tional head that assigns the focus feature (which Horvath claims to be I° in
Hungarian). Her formulation of the focus parameter is given in (18). The
conditions in (133,1– 4) determine what type of manifestation of focus can
be found in different languages:

(18)

Focus parameter:

1. nature of the feature:

(i) freely occurring, i.e. not vs.

(ii) assigned by a specific X°

transferred from another

category (Hungarian)

category (English)

2. what X° functional category of the clause is the assigner, i.e. the

source of the feature

3. whether the feature-assigning category needs to be

lexicalized

Inversion and information structure

4. the mode/nature of the process of feature-assignment:

(i) feature transfer
(ii) Spec-head agreement

Horvath’s focus parameter has the advantage of accounting for crosslinguistic
differences on the representation of focus without crucially resorting to an
additional functional projection, making the set of primitive categories of
the theory simpler. However, relying on a specific functional projection for
assigning of focus creates problems for cases in which focus surfaces in iso-
lated constituents.

Let us consider the case of English. According to (18), focus in English

is not codified in the syntax. Therefore, focus is freely assigned. However,
as long as the English syntax becomes more flexible, alternations seem to
be able to codify different information structures:

(19)

A: What did you give to Mary?

B: a. I gave Mary a book.

b. #I gave a book to Mary.

(20)

A: Who did you give a book to?

B: a. #I gave Mary a book.

b. I gave a book to Mary.

If focus were freely assigned to any category, the contrast between (19) and
(20) should not obtain. Any of the alternations should be adequate inde-
pendently of the context.

One of the most widely accepted theories of focus-in-situ is Rooth’s

(1985). Rooth proposes a semantic theory for focus, according to which no
focus-movement is required: given a sentence where focus can be identified,
a set of alternatives is construed. The set of possible alternatives is con-
strained within a certain contextual domain reminiscent of Jackendoff’s
(1972) P(resupposition)-set.

Rooth (1985) accounts for Chomsky’s (1976) weak-cross-over independ-

ently of focus-movement. Rooth claims that the bound variable readings
require

λ-abstraction, so that the pronouns or noun phrases can be inter-

preted as bound variables and not as free variables. A mechanism enabling
this to happen is already available in the grammar: Quantifier Raising (May
1985). In other words, it is not necessary to have focus movement as an

Focus-movement?

independent rule of the grammar: like all NPs, focused NPs may be QR-ed
but need not. However, if a bound variable reading is intended, QR is obliga-
tory. In that case, LF-evaluation constraints are operative, and weak-cross-
over configurations are ruled out. Rooth’s approach does not yet explain
why deaccenting (that is, removing the stress from the focused constituent)
obviates the WCO effects, but it is empirically superior to Chomsky’s since
it excludes the obligatoriness of WCO with focus (see also Vallduví 1990
for discussion).

Adopting Rooth’s approach allows for dismissing the weak-cross-over

argument as evidence in favor of focus-movement. At best, we can keep it
as evidence for QR.

However, Rooth’s theory of interpretation of focus in situ does not say

anything concerning languages like Hungarian, in which focused constituents
move to a specific position. Also, it does not explain the behavior of lan-
guages like Portuguese, in which, as it will be shown, focused constituents
seem to stay very low in the structure.

Since Rooth’s analysis is not complete enough to take care of the word

order facts, I will keep it as a semantic approach to focus. Syntax together
with prosody will enable an identification of the focus set of constituents for
a given sentence. The identified focus-set will be operated on by semantics.
Provided that there is an algorithm permitting a correct identification of
focus, semantics may apply over the material identified as focus.

4.3. Word order and focus

Having established part of my assumptions concerning the representation of
focus, let us look at the relationship between word order and focus, as far as
the placement of subjects is concerned. The first important aspect to note is
that the attested word order variation is not free: each word order can be
used only in given contexts. This has been noted several times in literature
on Portuguese (Duarte 1987, 1996, 1997; Ambar 1992, 1994, 1996, 1997,
Martins 1994; Costa 1997, 1998). For each of the possible word orders listed
in (21), a different felicity context is associated. In (21), I also indicate the
position each constituent occupies in the clausal structure, according to the
conclusions reached in the previous chapter.

(21)

SVO: subject in Spec,IP, object in base-position
VSO: subject in Spec, VP, object in base-position

Inversion and information structure

VOS: subject in Spec, VP, object adjoined to VP (via scrambling)
OSV: subject in Spec,IP, object topicalized
OVS: subject in Spec, VP, object topicalized

The relation between each word order and context is demonstrated in the
following examples. I will proceed by providing an appropriate discourse
context and check which possible order is an appropriate continuation for
each case. For the proper characterization of these contexts, the notions topic
and focus will be relevant. I will assume the following tests to identify topics
and focus:

a) In a question-answer pair a focused constituent in the answer replaces

the wh-word in the question (cf. Dik 1978, Bresnan and Mchombo 1987,
Rochemont and Culicover 1990, among others)

b) A Topic is information already referred to in the discourse or a subpart of

a referent already mentioned (see Büring 1997 for discussion and rele-
vant examples).

(22)

Object focused:

A: O que é que o Paulo partiu?

what Paulo broke

B: O Paulo partiu a janela.

Paulo broke the window
#Partiu o Paulo a janela.
#Partiu a janela o Paulo.
#A janela o Paulo partiu.
#A janela partiu o Paulo.

(23)

Sentence-focus:

A: O que é que aconteceu?

what happened

B: O Paulo partiu a janela.

Paulo broke the window
#Partiu o Paulo a janela.
#Partiu a janela o Paulo.
#A janela o Paulo partiu.
#A janela partiu o Paulo.

Word order and focus

(24)

Subject and object focused:

A: Ninguém partiu nada.

noone broke anything

B: #O Paulo partiu a janela.

Paulo broke the window
Partiu o Paulo a janela.
#Partiu a janela o Paulo.
#A janela o Paulo partiu.
#A janela partiu o Paulo.

(25)

Subject is focused:
A: Quem é que partiu a janela?

who broke the window

B: #O Paulo partiu a janela.

Paulo broke the window
#Partiu o Paulo a janela.
Partiu (a janela) o Paulo.
A janela o Paulo partiu.
A janela partiu o Paulo.

I will not discuss in detail the cases in which the object appears in sentence-
initial position. For the moment, it is enough to say that for objects to appear
in sentence-initial position, they have to have been referred in previous dis-
course and/or have some contrastive force:

(26)

A: A Ana viu o Paulo?

Ana saw Paulo

B: O Paulo, ela viu.

Paulo she saw

(27)

A: Quem é que partiu as janelas?

who

broke the windows

B: Esta janela partiu o Paulo

this window broke Paulo.

Summarizing, in terms of information structure, all constituents that convey
new information appear to the right, and constituents that convey information

Inversion and information structure

previously referred in the discourse appear at the left periphery of the sen-
tence. The following question remains to be answered: why should these
relations between positions and discourse information exist? What does
each of the identified syntactic positions have to do with the correspondent
discourse function?

4.4. Mapping syntax-discourse

I will be assuming the following:

a) Old information has to be either topicalized or defocused, while new

information is the Focus of a sentence;

b) I will follow proposals concerning the correlation between syntax and

phonological sentence stress which suggest that sentential stress falls on
the rightmost constituent of a sentence. Thus, in a normal SVO sequence,
stress will fall on the object. Assuming with Jackendoff (1972) that the
focused element in a sentence is the one bearing the most prominent
stress, the proposals above capture the fact that in an SVO sentence with
unmarked intonation the object is interpreted as the focus (cf. Lambrecht
1994 who shows that objects tend to be focus).

Still concerning focus, I will follow Reinhart’s (1995) suggestion that

XPs may be marked as foci with a heavy stress. This happens to any XP
that does not get default stress by virtue of not being rightmost.

c) I will assume with Zubizarreta (1995) and Reinhart (1995) that some

syntactic operations are prosodically motivated.

d) I will assume that the following tendencies hold, as observed by Lambrecht

(1994): subjects tend to be topics; objects tend to be foci; definites tend to
be old information (topic); indefinites tend to be new information (focus).
It is important to emphasize that these are just tendencies and not absolute
statements. Optimally, it would be possible to derive these tendencies from
some structural property of subjects, objects, definites and indefinites.

Let us now come back to the issue under investigation and see how these
assumptions derive the facts described in the previous sections. From the
line of inquiry I am pursuing, it is obvious that some new facts have to be

Mapping syntax-discourse

added to the paradigms described. More specifically, since the notion of focus
is crucial, and I am assuming with Nespor and Vogel (1986), Cinque (1993),
Zubizarreta (1998) and Reinhart (1995) that focus is partly a prosodic phe-
nomenon,

it now becomes important to see how subjects behave with

respect to intonation in each of the positions identified above.

It is a well-known fact that focus is normally associated with high stress.

In order to get a clear idea of the correct way to interpret focus, it is necessary
to investigate where prominence appears in the sentences. By doing so, I will
also be able to derive the generalization of the preceding section.

Frota (1994,1995) has argued in favor of representing focus in Portuguese

as a phonological category that is freely assigned. Part of her arguments is
based on the behavior of clitics (see Frota and Vigário 1996 for complete
argumentation).

The domain of Frota’s argumentation is narrow (or contrastive) focus, and

not information focus (though of course, they may coincide). Actually, Frota
(1997) suggests that two types of prominence may be necessary to describe
the two types of focus marking. What is relevant for the discussion is that
independently of the type of focus looked at, Frota’s claim appears to be
correct: focus (contrastive and information) is marked phonologically in
Portuguese.

If the distribution of stress in each of the word orders discussed above is

considered, evidence may be found in favor of Frota’s claim: there is a one-
to-one correspondence between the first focused constituent of the sentence
and its most prominent stress.

Consider (28) below, where capital letters indicate high stress:

(28)

a. Partiu o PAULO a janela.

broke Paulo the window

b. *PARTIU o Paulo a janela.

c. *Partiu o Paulo A JANELA.

(28) is a VSO sentence in which subject and object are in focus. In that case,
both have to appear in the right periphery of the sentence. The subject is the
first focus that appears and it bears the most prominent stress of the sentence.
Conveying the same information by means of the same word order and with
other stress patterns is not possible (28 b,c).

In (29), only the subject is in focus (VOS). In that case, the subject bears

the heaviest stress, though it is not heavier than the neutral stress present in
unmarked SVO sentences.

Inversion and information structure

(29)

a. Comeu a sopa o Paulo.

ate the soup Paulo

b. *Comeu a sopa o PAULO.

c. *Comeu a SOPA o Paulo.

d. *COMEU a sopa o Paulo.

In SVO sentences, the most natural intonation is the rightmost prominence,
without a very heavy stress:

(30)

O Paulo partiu a janela.
Paulo broke the window

The literature on prosody often distinguishes between neutral stress and
heavy stress (see e.g. Chomsky and Halle 1968). This distinction has been
criticized by Selkirk (1984), among others, who claims that there is no em-
pirical advantage in proposing such a distinction. Selkirk’s argumentation is
based on focus projection. Focus projection is the term given to the fact that
neutral stress on e.g. a sentence-final object may yield an interpretation in
which only the object is focused, or the VP is focused or the whole sentence
is focused. Selkirk claims that not only neutral sentence-final stress projects
allowing an interpretation in which there is focus on sentence-final element
only, on the VP or on the whole sentence. Selkirk claims that heavy stress in
a non-final constituent also permits focus-projection in the sense just
explained. If Selkirk’s observations are applied to the Portuguese case, it is
expected that sentence (30) be an appropriate answer to any of the questions
in (31):

(31)

a. What happened? (Sentence-focus)

b. What did Paulo do? (VP-focus)

c. What did Paulo break? (Object-focus)

This is indeed correct. However, differently from Selkirk’s claim for English,
a distinction between neutral stress and marked stress is relevant for the
identification of focus in Portuguese. Crucially, every time there is a high
stress on a non-final constituent, there is no projection of focus in this lan-
guage. (30) is not a legitimate sentence for expressing VP-focus. Since high
stress is only necessary on constituents that are not in absolute sentence-

Mapping syntax-discourse

final position, this invalidates Selkirk’s claim: marked stress does not
project.

(32)

Partiu O PAULO a janela.
broke Paulo the window

Selkirk claims that stress on XP, which is a constituent of YP, will enable
percolation of stress to YP. If focus projection would apply in these terms,
(32) could be a felicitous answer to (31b), with focus on the whole VP,
since the subject is VP-internal. In other words, the subject corresponds to
XP and the VP to YP in Selkirk’s algorithm. Hence, stress on the subject
should project to VP. Now, this is not true, though it is predicted by Selkirk’s
theory.

Note that even neutral stress does not project if there is a change in the

unmarked word order of the language. Hence, a VOS sentence is not a legiti-
mate answer to a question that requires something else than the subject to
be in focus:

(33)

What did Paulo do?

a. O Paulo partiu a janela.

Paulo broke the window

b. #Partiu a janela o Paulo.

Although the sentence’s main stress fall on the rightmost constituent in both
cases, and in both cases the rightmost constituent is VP-internal, focus pro-
jection is not allowed.

Actually, even the English cases Selkirk presents as possible cases of

focus-projection without rightmost prominence are difficult to evaluate.
Selkirk claims that (34) may have VP-focus:

(34)

(from Selkirk 1984):
Did John give a BOOK to Bill?

Selkirk claims that (35) is an appropriate answer for this sentence:

(35)

No, he grew a pot of NARCISSUS for him.

In (35), the verb and the NP contrast with give a book in (36). Selkirk con-
cludes from this that prominence on the NP may give VP focus.

It seems,

Inversion and information structure

though, that this is not a very accurate conclusion, since if there were VP-
focus in (36), the sentences in (36) might as well be appropriate answers,
since (36a) and (36b) involve alternatives to the focused VP:

(36)

a. No, he grew a pot of NARCISSUS for Mary.

b. No, he killed Mary.

Now, (36a,b) are not appropriate answers to (34), presumably because in
(35) there is no VP-focus. Actually, Selkirk acknowledges that for (35) to
involve VP-focus, “John, Bill, and, say, Bill’s recent birthday are old infor-
mation in the discourse”(p.216, my emphasis). If Bill is old information, it
is difficult to understand how it can be maintained that the whole VP is in
focus. In other words, I am suggesting that the idea that VP-focus is in-
volved comes from the fact that grow X for Y and give X to Y are minimally
different in the relevant context. No new information is added by replacing
give with grow. In this sense, the only new information is the NP, and VP-
focus is only apparent. The use of a proper name in the question may also
determine the impression that there might be VP-focus even without Bill
being focused. Proper names never convey absolute new information. If
instead of a definite, someone is used, a VP-focus interpretation never arises
unless there is rightward prominence:

(37)

A: What did John do?

B: #John gave a BOOK to someone.

John gave a book to someone.

Speakers who accept (37B) report to me that the DP a book clearly must
have contrastive force. We have thus a case of overlap of two types of
focus: a contrastive focus on the DP and information focus on the VP. Since
these two types of focus may be distinguished in semantic terms, the behav-
ior of one should not be used as evidence for the other.

Summarizing, it is thus possible to interpret the stress pattern of (35) in

different terms: since the pronoun is old information (it refers to Bill), it
must not be heavily stressed. In the context given, grew a pot of narcissus
for him and give a pot of narcissus to him are equivalent. Hence, the only
new information which is contrasted is the DP object. The stress pattern
emerges as a consequence of shifting the stress from the sentence-final PP
to the object DP. The VP-focus effects arise only if the identity between

Mapping syntax-discourse

grow X for Y and give X to Y in the ‘birthday’-context are not acknowledged.
The fact that not any VP may replace the one in the question confirms that in
this case there is no VP-focus.

It seems thus that whenever there is a change either in the unmarked word

order (SVO for Portuguese) or in the normal intonational pattern (rightmost
prominence), there is no projection of focus.

The intonation of information focus in Portuguese can then be summa-

rized as follows: Focused constituents are prosodically prominent. If they
are rightmost they bear neutral stress, if they are not rightmost they are
assigned a high pitch accent. If there is more than one focus, the leftmost
bears the heavy stress: all constituents following the heavy stress are inter-
preted as focus. Information focus is not incompatible with other types of
contrast. Hence, any constituent may bear heavy stress, independently of its
being the focus of the sentence for contrast purposes. This will, however,
make projection of focus more difficult for the reasons pointed out above.

Under the review of Selkirk’s discussion of projection of focus made

above, I reach conclusions similar to hers regarding the status of projection as
not being exceptional, although in the exact opposite sense. She concluded
that focus-projection is not exceptional, since it may happen almost every-
where. Differently, I concluded that focus-projection is unexceptional, since
it does not need to be postulated. The reasoning goes as follows: all that is
needed is rightmost prominence. The effects of projection are a consequence
of coincidence of rightmost borders of constituents (NP,VP,IP). Any other
stress pattern will preclude projection, since projection does not exist as an
independent phenomenon. It is just the effect of the ambiguity of several
rightward constituent borders.

Combining the distribution of focus and the prosodic facts, the following

generalizations are obtained:

a) Focused constituents are rightmost in the sentence;

b) Focused elements bear high stress (neutral or marked);

c) If there are multiple foci, they appear all to the right of the non-foci ele-

ments.

d) If there are multiple foci, the first in a left-to-right fashion bears heavy

stress.

These observations may serve as the cues to the formulation of an algorithm
to identify information focus in European Portuguese, which is given in (38).

Inversion and information structure

Note that (38) is meant as an algorithm to identify information focus, not
the other types of focus, excluding thus the cleft constructions discussed
above, which were nevertheless useful for the study of focus projection.
This formulation is partially based on Reinhart’s (1995) observations con-
cerning the distribution of focus and on her own formulation of what a
focus-set of constituents is:

(38) The focus set of constituents of a sentence is the prosodically most

prominent constituent plus everything it c-commands.

Prosodic prominence is defined as in (39), adapted from Nespor and Vogel
(1986), Cinque (1993), Zubizarreta (1995) and Nash (1995):

(39) The prosodically unmarked most prominent constituent is the right-

most one, following the recursion pattern of a language.

(39) states that in VO languages, the most prominent constituent is the
rightmost one to the right of the verb, while in OV languages, the most
prominent constituent is the rightmost to the left of the verb (the XP in bold
in (40) below):

(40)

V O XP XP
O XP XP V

Let me now return to the algorithm presented above. When the conditions
for a constituent which has to be interpreted as focus to receive the sentence
(neutral) most prominent stress are not met, a heavy stress has to be assigned.
This happens e.g. in the case of multiple foci, in which two constituents can-
not occupy the rightmost position at the same time. Let us see how the algo-
rithm in (39) allows for identifying focus in the cases discussed above. In
the case of a VSO sentence with high stress on the subject, the set of focused
constituents will be the subject and the object. The subject is interpreted as
focus, because it is the most prominent constituent. The object is interpreted
as focus, because it is c-commanded by the subject. Since the verb is out of
the c-command domain of the subject, it is not interpreted as focus.

In the case of VOS sentences, the subject bears the main neutral stress and

does not c-command anything, hence it is the only constituent interpreted as
focus. The focus on the subject may not project, because the unmarked word
order is changed. The impossibility for rightmost focus to project when

Mapping syntax-discourse

there are changes at the unmarked word order will not follow from any con-
siderations in this chapter. A solution for this problem will be proposed later.

As for SVO, two situations are possible: the most natural is that the

object is the rightmost element, bearing sentence neutral stress. In that case,
since it does not c-command anything, only the object is interpreted as new
information. The other two possibilities are interpretation of these sentences
as VP-focus and everything in focus. These interpretations also follow from
the algorithm in (39), since the object is also the rightmost part of the con-
stituent IP or VP that one wants to focus in cases of sentence-focus and VP-
focus respectively. Focus projection is thus interpreted here as a natural
consequence of the general neutral stress rule: assign the most prominent
stress to the rightmost element. This rule is general to any constituent, inde-
pendently of its label, as observed by Cinque (1993) among others

(41)

a. [

the good man with the red shirt]

b. [

much more beautiful]

c. [

before midnight]

d. [

give something to someone]

e. [

John gave a book to Mary]

Now, why is focus projection impossible in a VSO sentence? The explanation
is simple: since the subject is the most prominent constituent, it cannot be the
case that the whole sentence is in focus, since the stress does not fall on its
rightmost element (see also Cinque 1993 and Reinhart 1995 for a similar
reasoning).

Should it be possible then that stress on a sentence-final subject would

yield sentence-focus? In principle, nothing precludes it, since the subject is
both the rightmost element and the most prominent one. There is however a
reason for excluding this possibility: if one compares a VOS sentence with
a SVO sentence, they are equal in terms of prosody. However, the former is
more marked, since the subject does not appear in its canonical position:
Spec,IP. Hence, since there is no difference in status in terms of possible
interpretations, the least marked SVO sentence is grammatical.

Let us now see why some sentences are impossible when associated to

some interpretations, and how that follows from the algorithm in (39). Why
can’t a VSO sentence be interpreted with focus on the subject only? That is,
why is (42B) not a felicitous answer to (42A):

Inversion and information structure

(42)

A: Who broke the window?

B: #Partiu o PAULO a janela.

broke Paulo the window

The problem with this answer is that it forces the object to be interpreted as
new information, since it is within the c-command domain of the subject
which bears the most prominent stress. Since window has been referred to in
the discourse, this interpretation is not felicitous.

Another case excluded by the algorithm in (38) is the interpretation for a

VSO sentence in which the subject is not stressed:

(43)

*Partiu o Paulo a JANELA.
broke Paulo the window

The problem with this sentence is that, according to the definitions in (39),
only the object can be interpreted as focus. Now, if that is the case, the sen-
tence is ruled out for either of the two following reasons: if the subject is to
be interpreted as new information, the sentence is out since the subject is
not included in the focus set of constituents. This is because the subject is
not c-commanded by the most prominent constituent. Alternatively, if the
interpretation required is one in which the subject is not to be interpreted as
focus, and then there is no reason for it to stay in a low position.

Also,

heavy stress should never appear in sentence-final position, unless for con-
trast purposes, since it is not necessary in order to assign prominence to a
constituent.

It seems thus that the principles given above make a large number of cor-

rect predictions, concerning the possible intonations and mappings between
word orders, intonations and discourse functions.

Summarizing the results of this section, I present the following table,

containing the several possible word orders with most prominent element
marked in bold. The second column of the table indicates the focus set of
constituents given by each word order, and the third column indicates the
reason why some pairs word order/intonation and word order/focus-set are
impossible:

Mapping syntax-discourse

Table 1.

Table 1 illustrates the relevance of two aspects for the identification of focus:
stress assignment and c-command by the most prominent element.

It is important to note at this stage that the algorithm in (39) just represents

the instructions hearers have to follow in order to identify the focus of a
sentence. As such, these principles are quite descriptive in nature. They are
explanatory only to the extent that they follow from independent principles:

Inversion and information structure

Word Order

Focus-set

Reason for ungrammaticality

SVO

O, VP or IP

VSO

S and O

VOS

*SVO

S is not c-commanded by the most
prominent element

*SVO

V and O are c-commanded by the
most prominent element; they
should be part of the focus-set

*SVO

The sequence with unmarked
stress blocks the more marked one

*VSO

Stress is not in the rightmost
constituent of IP

*VSO

SVO is preferable, since the word
order is less marked (cf. Costa
1998, and section 3.4)

*VSO

Stress falls on the subject

*VSO

Since the subject is not in focus,
there is no reason for it to stay low

*VOS

VP, IP

SVO is preferable, since the word
order is less marked (cf. 3.4)

*VOS

S and O

O has been scrambled for escaping
focus, so it may not be the focus.

in order to get to (39), it is necessary to have a theory of phrasal phonology
which relates sentence stress assignment to the directionality parameter
(Nash 1995; Nespor, Guasti and Cristophe 1995). In addition, a theory of
sentence structure making sure the c-command requirement is met in the
relevant configurations is required. Finally, there has to be an independent
explanation for the part of the principles that require focus to be prosodically
prominent (see Givón 1984 for an explanation in terms of attention span). It
is a combination of these three aspects that makes it possible to arrive at a
formulation of an algorithm for sentence stress identification.

4.5. No focus-movement in Portuguese

Concerning the debate focus-in-situ vs. focus-movement, from the Portu-
guese data discussed above, one may conclude that several rearrangements
of the sentence word order must be made for the sake of identifying focus,
although there is not one specific position where focused elements move to
on the left periphery of the sentence, as is the case in Hungarian. Instead,
focused elements seem to stay in the rightmost position, and in most cases
they do not undergo any movement in order to reach this position. Given
this behavior, the conclusion obtains that Portuguese is not of the Hungarian
type, and it can be treated as a language with focus in situ.

In Costa (1998,

2000b), empirical arguments are presented showing that a focus-movement
approach would make wrong predictions for dealing with these data.

For completeness, let me just point out that the syntactically marked

focus constructions that are found in Portuguese, described by Raposo
(1994), among others, involve quantified DPs, and is exemplified in (44)
(Raposo, p.c.):

(44)

Muito vinho o João bebeu!
Much wine João drank

This type of construction is much more natural if the sentence is exclamative,
which casts some doubt on their status as instances of focus-movement.

Comparing this construction with the Italian case, they appear to be alike.

Naturally, not all properties may be checked, since the construction is possi-
ble only with quantified DPs:

i) As far as combination with clitics is concerned, the preposed quantified

DP may not be doubled:

No focus-movement in Portuguese

(45)

a. Portuguese:

Muito vinho, o João (*o) bebeu(*-o)!
Much wine, João it drank it

b. Italian:

IL TUO LIBRO, (*lo) ho comprato
your book, it have-I bought

ii) There may only be one quantified DP in this preposing construction:

(46)

Portuguese:

a. Muitas vezes o João bebeu vinho!

Many times João drank wine

b. Muito vinho o João bebeu muitas vezes.

Much wine João drank many times

c. *Muitas vezes, muito vinho bebeu.

Many times much wine drank

Italian:

d. *IL LIBRO, A GIANNI, darò.

the book to Gianni, (I-will-)give

(46c) shows that two quantified DPs may not be preposed. This sentence is
only acceptable if the first preposed element is interpreted as a topic. Under
the exclamative interpretation that these preposed DPs receive (see below),
the sentence is ungrammatical.

iii) Preposed quantified DPs in the relevant interpretation are incompatible

with wh-elements:

(47)

Portuguese:

a. *Muito vinho, com quem é que o João bebeu?

Much wine, with whom João drank

b. *Com quem, muito vinho o João bebeu?

With whom much wine João drank

Italian:

c. *A GIANNI, che cosa hai detto?

to Gianni, what did you tell

Inversion and information structure

Again, (47a) is only possible under the topic interpretation for the DP muito
vinho. With the exclamative interpretation, the sentence is ungrammatical. This
is not surprising, since preposing is more natural in exclamative sentences,
and a sentence may not be simultaneously interrogative and exclamative.

Given the similarity of behavior with the Italian construction, when com-

paring in-situ with moved foci, I will use quantified DPs to be sure that we
are looking at syntactically marked foci.

The two types of constructions can be fruitfully compared contrasting the

function of preposing and focus-in-situ respectively (see also Raposo 1994;
Duarte 1987; Casielles 1996). It will be argued that for a constituent to be pre-
posed, it has to yield given information, even in the constructions involving
quantified DPs only.

I have been considering throughout this chapter that the focus that appears

in situ is used to introduce new information; hence it is felicitous as an answer
to a wh-question:

(48)

A: Quem é que o Paulo viu?

who did Paulo see

B: O Paulo viu muita gente.

Paulo saw many people

Now, if the difference between preposing and focus-in-situ is just a differ-
ence in terms of the locus of application of the operation focus-movement
in the derivational history of the sentence, both cases are expected to yield a
felicitous answer to a wh-question. However, that is not true, as the inappro-
priateness of (49) attests:

(49)

A: Quem é que o Paulo viu?

who did Paulo see

B: #Muita gente, o Paulo viu.

many people, Paulo saw

Note that the facts in (48) and (49) are true for any type of preposing. If we
prepose a PP, the same effects obtain:

(50)

A: Com quem é que o Paulo falou?

With whom

Paulo talked

B: O Paulo falou com {a Maria/muita gente}

Paulo talked with Maria / many people

No focus-movement in Portuguese

(51)

A: Com quem é que o Paulo falou?

With whom

Paulo talked

B: #Com {a Maria/muita gente} o Paulo falou

With Maria /many people Paulo talked

What is then a felicitous context for preposing? Consider the following
fragments of discourse, in which preposing is felicitous.

As mentioned before, preposing is possible if the preposed constituent is

given in the discourse (or in the context). These are cases in which preposing
is undistinguishable from topicalization. (52) and (53) exemplify such cases:

(52)

A: Quem é que comeu muita sopa?

who ate much soup

B: MUITA SOPA, ninguém comeu, (muito pão comeu o Paulo).

much soup noone ate (much bread Paulo ate)

(53)

A: Alguém leu o teu livro?

someone read your book

B: O MEU LIVRO, o Paulo leu (não o teu).

my book Paulo read (not yours)

Another context for preposing (now for the case considered in Raposo’s
work cited above) is exclamative sentences:

(54)

Muito vinho o Paulo bebeu!
Much wine Paulo drank

(55)

Muita gente tu encontraste!
Many people you met

In these cases there is no implicit contrast with anything else. Note however
that, even in an exclamative answer to a question, this word order is not licit if
the quantified DP is new information (the information focus of the sentence).
The relevant contrasts are given in (56) and (57):

(56)

A: O que é que o Paulo bebeu?

What Paulo

drank

B: a. O Paulo bebeu muito vinho!

Paulo drank much wine

b. #Muito vinho o Paulo bebeu!

Much wine

Paulo bebeu

Inversion and information structure

(57)

A: Quem chegou?

Who arrived

B: a. Chegou muita gente!

Arrived many people

b. #Muita gente chegou!

Many people arrived

A rather convincing confirmation of the exclamative force of the sentences

discussed in Raposo’s work was brought to my attention by Inês Duarte: it
is impossible to add a question tag to a sentence with a preposed quantified
DP, which shows that the sentence may not be interpreted as declarative:

(58) Muito vinho bebeu o João, (*não bebeu?)

much wine drank João, not drank

It is thus possible to conclude that the focus-preposing construction may

not serve to introduce new information. It is possible to obtain this word
order if the quantified DP may be deduced from the context.

(59) A: A festa foi óptima!

The party was great

B: Muito vinho o Paulo bebeu!

As the examples above show, in a felicitous context for preposing, the con-

stituent that is in the left periphery of the sentence has to have been pre-
viously referred to in the discourse or be deduced from the context. More-
over, it never introduces new information. Rather, it contrasts some piece of
old information with something else. If preposing would convey new infor-
mation, it might be used for answering questions. As shown above, this is
not true. As also mentioned in work by Inês Duarte (Duarte 1987, 1996),
anytime there is preposing of a constituent in Portuguese, the information
conveyed by that constituent is given (independently of whether it is further
contrastive or topical). Actually, the part that answers the question can
never be fronted as (60) and (61) illustrate, independently of whether the
constituent conveying new information is contrasted to another one or not:

(60) A: Quem é que viste ontem?

who did you see yesterday

B: ONTEM vi [muita gente]

, (hoje pouca)

yesterday I saw many people, (today few)

No focus-movement in Portuguese

An answer in which the focus is fronted and the contrasted temporal adverb
is left in situ is simply infelicitous:

(61)

A: Quem é que viste ontem?

who did you see yesterday

B: #[Muita gente]

vi ontem.

Many people I saw yesterday

This is true even if the focus of the sentence (i.e. the part that replaces the
wh-phrase) is itself contrasted:

(62)

A: O que é que puseste na prateleira?

What did you put on the shelf

B: # [MUITA COISA]

pus na prateleira, [POUCA COISA]

, na cadeira

Many things I put on the shelf, few things on the chair

A felicitous answer for this question would involve preposing the PP:

(63)

B: NA PRATELEIRA, pus [muita coisa]

, NA CADEIRA [pouca coisa]

on the shelf I put many things, on the chair few things

It seems thus, that in spite of the need to be contrasted, a constituent that
constitutes absolutely new information can never appear in fronted position.
On the other hand, topics may be contrasted. This conclusion suggests that
taking the constructions of preposing as the overt counterpart of focus in situ
is erroneous, and that the label focus-preposing is often used without looking
at the specific properties of the language involved and at the contexts in
which each construction may be used, as also pointed out by Büring (1997).

For the case of Portuguese, the so-called focus-preposing construction is

restricted to quantified DPs, and even in those cases, it is not an alternative
to focus-in-situ, since the discourse function is different. Hence the con-
struction involving preposing must not be taken as an argument for focus-
movement.

Summing up, it was shown in the previous sections that, for a subject to

occur postverbally in its base position, it must be focused. This observation
supports the view that interface conditions play a role in determining the
potential surface positions of subjects. The remaining question is how the
in-situ subjects are licensed. This is the topic of the next section.

Inversion and information structure

4.6. Phases, locality and subjects in Spec, VP

Following recent work by Chomsky (1998, 2000), it is legitimate to suppose
that the Case-features of the head attracting the subject may be licensed
under two mechanisms: Move or Agree. If the former is chosen, the subject
is attracted to Spec,IP. If the latter option is at stake, the subject stays in its
base-position, and Case-features are licensed under Agree. Note that, as pro-
posed above, if the subject is in Spec, VP, Spec,IP is empty:

(64)

a. [

[CASE]

[

t t

b. [

[CASE]

[

[CASE]

At first sight, this gives rise to optionality: there are two converging deriva-
tions for licensing subjects. Assuming with Adger (1994), that the choice
between two converging outputs is made post-syntactically, taking into
accout the information structure of the clause, this is a desirable result. Given
a pair of converging outputs SVO and VSO, the latter will only be picked if
the subject is focused.

If Agree is the relevant mechanism for licensing in-situ subjects, it is pre-

dicted that locality effects should arise.

This prediction is borne out. As the

following data show, an inverted subject may occur within a non-finite
clause:

(65)

a. Decidiram ler todos os alunos esse livro.

Decided-3pl read all the students that book

b. Querem ler todos os alunos esse livro.

Want-3pl read all the students that book

The surprising aspect of the examples in (65) is that the empty subject of
the matrix verb is correferential with the subject of the non-finite verb, and
yet no principle-C effects arise. This follows from the proposal that Spec,IP
is empty in inversion contexts.

It must be noted that this type of in-situ

subject within an infinitival clause is restricted: it only happens if the matrix
verbs do not select a CP. Let us consider clitic-climbing, as in (66), or weak-
adverb climbing

, as in (67), as a signal that the complement of the matrix

verb is defective. The contrasts in the examples below between finite and
non-finite complements show that these two phenomena are dependent on
the absence of a CP-node (cf. Gonçalves 1999, among many others):

Phases, locality and subjects in Spec, VP

(66)

a. Eu só o quero ler amanhã.

I only it-acc want read tomorrow
“I only want to read it tomorrow”

b. *Eu só o quero que leias amanhã.

I only it-acc want that (you) read tomorrow
“I only want that you read it tomorrow”

c. Eu só quero que o leias amanhã.

I only want that (you) it- acc read tomorrow
“I only want that you read it tomorrow”

(67)

a. Eu só lá decidi ir ontem.

I only there decided go yesterday
“I only decided to go there yesterday.”

b. *Eu só lá decidi que ia ontem.

I only there decided that (I) would-go yesterday
“I only decided that I would go there yesterday”

c. Eu só decidi que ia lá ontem.

I only decided that (I) would-go there yesterday
“I only decided that I would go there yesterday”

Accordingly, if there is no evidence for the absence of CP, either because
the matrix verb is not a restructuring verb, or if an embedded negation is
inserted blocking restructuring, or because there is no independent evidence
for any type of transparency relation between the matrix verb and the em-
bedded complement (independently of restructuring), the sentences become
ungrammatical, contrasting with their non-inverted counterparts:

(68)

Embedded negation:

a. ??Decidiram não ler todos os alunos esse livro.

Decided-3pl read all the students that book

b. Todos os alunos decidiram não ler esse livro.

All the students decided not to read that book

(69)

Non-restructuring verbs (projecting CP):

a. *Recusaram ler todos os alunos esse livro.

Refused-3pl read all the students that book

Inversion and information structure

b. Todos os alunos recusaram ler esse livro.

All the students refused to read that book

c. *Negaram ler todos os alunos esse livro.

Denied to read all the students that book

d. Todos alunos negaram ler esse livro.

All the students denied to read that book

These data provide the necessary evidence for assuming that Agree is at
stake. Assuming with Chomsky (1998, 2000) that the scope of Agree is the
strong phase, and accepting the standard view that not all non-finite com-
plements project CP (Boskovic 1997), these data follow. For instances, in the
restructuring contexts, CP is not projected, there is no strong phase boundary,
and Agree may operate.

A consequence of the present proposal is that in VSO contexts, there is

no strong phase boundary intervening between T and Spec,VP. In other
words, it must be assumed that either vP does not exist in European Portu-
guese, or it is a weak phase. In Costa (1998), following claims on the non-
universality of functional categories (e.g. Bobaljik and Thrainssón 1996), it
is suggested that AgrO does not project in languages in which scrambled
objects behave like adjuncts. If v is taken as the category providing the
landing site for objects in Object-shift languages, the assumption that AgrO
does not project may be adapted to v. If this is the case, there is indeed no
strong phase boundary between T and Spec,VP, and Agree may take place.

Alternatively, it is legitimate to assume that the subject is generated in

Spec,vP, and the results achieved for inverted subjects may be translated,
assuming that inverted subjects are stranded in the specifier of VP. In that
case, the relevant configurations are:

(70)

a. [

[CASE]

[

t t

b. [

[CASE]

[

[CASE]

Since the subject is at the edge of the strong phase (vP), Agree can probe
this position and the licensing of the relevant features takes place.

Phases, locality and subjects in Spec, VP

4.7. Related evidence: agreement in copular constructions

Before concluding, let us consider a set of data involving rightward agree-
ment, which may provide additional evidence for the proposal made above
that subjects in-situ are licensed under Agree. The data to be considered
comes from inverted copular constructions, in which the verb agrees to the
right. We will add data from Brazilian Portuguese, since this language differs
from European Portuguese in two interesting ways: first, it does not have
subject-verb inversion of the type discussed above (cf. Figueiredo Silva 1996,
among others);

second, it does not have clitic climbing (cf. Duarte and

Gonçalves 2000, among others). These two differences and the agreement
facts will permit checking whether there is any correlation between locality,
agreement and the possibility for subjects to occur in a postverbal position.

Let us then consider the facts. Both in Brazilian and in European Portu-

guese, in specificational sentences, the verb agrees to the right, if there is a
pronominal form on the right, as shown in (71):

(71)

O assassino sou eu.
The murderer am I

If the pronominal form is to the left of the verb, the verb agrees to the left, as
in (72):

(72)

Eu sou o assassino.
I am the murderer

The interesting difference between the two languages emerges when there is
a modal verb involved. If the specificational sentence occurs with a modal
verb, the same agreement pattern is reproduced in European Portuguese.
The modal verb agrees with the pronominal form to its right.

European Portuguese:

(73)

a. O assassino devo ser eu.

the murderer must-1sg be I

b. O assassino posso ser eu.

the murderer may-1sg be I

In Brazilian Portuguese, however, the agreement pattern with modal verbs
is different. If the specificational sentence contains a modal, the modal verb
agrees with the preverbal element:

100

Inversion and information structure

Brazilian Portuguese:

(74)

a. O assassino deve ser eu.

the murderer must-3sg be I

b. O assassino pode ser eu.

the murderer may-3sg be I

Note that, in (74), Brazilian Portuguese is behaving like English in normal
specificational sentences in not allowing agreement to the right:

(75)

a. The murderer is me.

b. *The murderer am I.

These agreement patterns raise the following questions:

i) What is the difference between Brazilian and European Portuguese modal

verbs underlying the differences between the agreement pattern in the
two languages?

ii) What underlies the difference between Brazilian Portuguese and English

simple cases, making agreement to the right possible only in the former
language?

We assume that the agreement pattern found in specificational sentences is
the result of there being inversion (cf. Tavares (in preparation)). Let us fur-
ther assume that the subject does not reach Spec,IP in this type of inversion,
as illustrated in (76), and remain agnostic as far as the target of the fronted
predicate is:

(76)

[

O assassino [

I’

sou [

In this sense, the structure in (76) is similar to other cases of predicate
inversion, as in (77):

(77)

Inteligente sou eu.
clever am I

The assumption that the case in which there is agreement to the right is a
(special) case of subject-verb inversion implies assuming an analysis similar

Related evidence: agreement in copular constructions

101

to the one proposed above for other cases of subject-verb inversion. In other
words, extending the analysis proposed in the previous section to these
structures, one can assume that the inverted subject eu in (76) is licensed
under Agree.

This assumption straightforwardly explains why there is agreement to

the right in Portuguese, German or Icelandic, but not in English (Tony Kroch,
p.c.). Only in the latter is there no subject-verb inversion, which blocks the
possibility of tracing rightward agreement as an effect of subject-verb inver-
sion. In our terms, English does not allow for any subject to be licensed
under Agree, which has reflexes on the lack of agreement to the right. The
option of licensing subjects to the right under Agree is restricted in Brazilian
Portuguese to some verb classes.

Let us now come back to the difference between the two varieties of Por-

tuguese. As illustrated above, in European Portuguese, a modal verb agrees
to the right with the subject embedded in the infinitival construction, which
is not true for Brazilian Portuguese. These agreement facts with modals
may be shown to follow from locality and the presence or absence of CP in
a way similar to the one outlined in the previous section.

As shown in (78), modal verbs in European Portuguese are restructuring

verbs (Gonçalves 1999), allowing clitic climbing:

(78)

Eu não lhe devo dar o livro.
I not him must give the book

“I must not give him the book”

Unlike in European Portuguese, restructuring is not productive in Brazilian
Portuguese. A well known fact about the latter is that it does not permit clitic
climbing (cf. Duarte and Gonçalves 2000):

(79)

Brazilian Portuguese:

a. Eu não devo lhe dar o livro.

I not must him give the book

b. *Eu não lhe devo dar o livro.

I not him must give the book

This difference in the behavior of modal verbs in the two languages gives us
the ingredients for understanding the agreement patterns. If in (73) there is
restructuring, the modal and the main verb work as a single phase domain,

102

Inversion and information structure

since CP is not projected. Because there is no phase-boundary between the
matrix Infl and the subject, the latter is licensed under Agree. The agreement
pattern is therefore expected: we find no difference between the construction
in (73) and the constructions in which there is subject-verb inversion within
a non-finite clause.

Now, let us consider the case of Brazilian Portuguese. If Brazilian Portu-

guese modal verbs do not restructure, as shown in (75), a locality constraint
makes it impossible to derive the configuration with agreement to the right.
The reason for a non-restructuring context to block agreement to the right
follows from the proposal made in the preceding section. CP is projected,
inducing a phase-boundary, and blocking the Agree relation between the
matrix Infl and the embedded subject.

An obvious difference in terms of results must be pointed out. A failure

to establish Agree induced ungrammaticality in European Portuguese, while
it induces a failure in morphological agreement in Brazilian Portuguese. I
do not have a solution to this problem. It may be the case that the embedded
Spec,TP plays a role in licensing the embedded subject. Negation is assumed
to be associated to TP in Romance languages (Zanuttini 1996). It may there-
fore be taken as a diagnostic to know whether Spec,TP is projected. As shown
in (80), for European Portuguese, the presence of an embedded negation
blocks restructuring, since it makes clitic climbing impossible. The data in
(80a,b) show that there is no problem with having a negation embedded
under the modal. The sentences in (80c,d) show that clitics can not climb in
the presence of the embedded negation:

(80)

a. Eu não devo ver o Pedro

I not must see Pedro

b. Eu devo não ver o Pedro.

I must not see Pedro

c. Eu devo não o ver.

I must not him see

d. *Eu devo-o não ver.

I must him not see

Combining these facts with the analysis of agreement to the right, it is pre-
dicted that the latter should not occur in Brazilian Portuguese. In this lan-
guage, there is no restructuring, thus, moving the predicate to Spec,IP above
the modal will violate locality, since there is another Spec,IP closer to it.

Related evidence: agreement in copular constructions

103

This is, however, not a satisfactory answer to the problem, since it predicts
that the embedded subject should be licensed by the embedded Spec,TP. I
will therefore leave the issue of how the subject in Brazilian Portuguese is
licensed for further research.

Interestingly, evidence for this analysis may be found internally to

European Portuguese. If a verb forces the projection of CP, agreement to the
right is never possible. This is the case with the modal verb necessitar ‘to
need’. The contrast in (81) shows that this verb does not allow clitic climbing;
the data in (82) shows the agreement pattern in the specificational context:

(81)

a. Eu só necessito de lhe dar-lhe um livro.

I just need to him give a book

b. *Eu só lhe necessito de dar um livro.

I just him need to give a book
“I just need to give him a book”.

(82)

a. *O problema necessito de ser eu.

The problem need-1sg to be I

b. ?O problema necessita de ser eu.

The problem needs to be I
“I must be the problem”.

The sentence in (82b) patterns like the Brazilian Portuguese cases discussed
above. This confirms that the projection of CP is the crucial factor to know
whether an embedded subject can be licensed to the right, under Agree.
Incidentally, note that (82b) is not ungrammatical, but just marginal, which
allows us to get back to the issue raised above for Brazilian Portuguese:
how is the embedded subject licensed in sentences like (82b)? The fact that
the question may be raised for European Portuguese as well, and the con-
trast between (82b) and the sentences discussed in the previous section in
which the subject occurs after a non-finite verb embedded under a control
predicate allows for conjecturing that the problem has to do with the pres-
ence of the copula rather than with the failure to Agree. In other words, the
presence of the copula be seems to enable an additional last resort way of
licensing the inverted subject. Obviously, this must be a last resort strategy,
since it would otherwise rescue all ungrammatical cases discussed above.

Summing up this section, although some issues remain open, its purpose

was to show that there is independent evidence for the claim that the presence

104

Inversion and information structure

of a CP-node induces a boundary for licensing an inverted subject under
Agree. The contrast between European and Brazilian Portuguese and its
relation with the availability of restructuring with modal verbs in the two
languages provides the necessary evidence for relating inversion with the
amount of projected structure. The special feature of the data discussed in
this section is that we see a correlation between morphological agreement and
the cases in which the inverted subject is licensed by the matrix Infl. In the
cases discussed, failure to Agree corresponds to a failure to morphologically
agree.

4.8. Conclusions

This chapter provided evidence for the possibility of subjects to surface in
Spec,VP. It was defended that this occurs when subjects are focused, in
compliance with requirements imposed by the mapping with prosody. It
was further shown that the possibility for subjects to emerge in Spec,VP is
primarily conditioned by syntax: it reflects the possibility of checking Case-
features under Move or Agree.

Conclusions

105

5. Optionality and left-dislocated subjects:

semantic and discourse properties

In chapters 2, 3 and 4, it was observed that there is a correlation between the
distribution of information focus and the placement of subjects in European
Portuguese. In particular, the following two generalizations arose: (i) infor-
mation focus surfaces at the clause’s rightmost position; (ii) for each context,
there is only one word order.

The discussion in chapter 2 led to the conclusion that preverbal subjects

in European Portuguese occupy an A-position. Yet, as mentioned, nothing
precludes left-dislocation of a subject, in the same way nothing precludes
left-dislocation of an object. However, nothing was said concerning contexts
favoring subject left-dislocation.

This chapter examines two contexts challenging the generalization that

for each context there is only one order: unaccusative contexts and answers
to multiple wh-questions. The latter will turn out to provide a context in
which subjects may be left-dislocated.

5.1. SV and VS in unaccusative contexts

As mentioned in chapter 3, one of the criteria used for determining the un-
marked word order of a language is to look at contexts of sentence-focus,
which may be tested in answers to what happened?. In this context, the only
felicitous word order in Portuguese is SVO. This is illustrated in (1) for
transitive verbs. All other word orders are unfelicitous in this context:

(1)

What happened?

a. O Paulo comeu o bolo.

Paulo ate the cake

b. #Comeu o Paulo o bolo.

c. #Comeu o bolo o Paulo.

d. #O Paulo o bolo comeu

e. #O bolo o Paulo comeu.

#O bolo comeu o Paulo.

If the verb is intransitive, the only felicitous answer to what happened? is the
order SV. This is illustrated in (2):

(2)

What happened?

a. O Paulo cantou.

Paulo sang

b. #Cantou o Paulo

The first problem raised by the behavior of subjects of unaccusatives has to
do with their behavior in sentence-focus contexts. The analysis presented in
the preceding chapters makes the prediction that subjects move to Spec,IP in
unmarked contexts independently of the verb they appear with. This predic-
tion was confirmed for transitives and intransitives. The same should be true
for unaccusatives. The complement of an unaccusative verb should always
raise to Spec,IP to receive nominative Case. However, in sentence-focus
contexts, movement of the subject to Spec,IP seems to be optional, as illus-
trated in (3)

(3)

What happened?

a. O Paulo chegou.

Paulo arrived

a.’ Chegou o Paulo.

arrived Paulo

b. A fábrica ardeu.

the factory burnt

b.’ Ardeu a fábrica.

This behavior raises the following question:

Why is inversion a felicitous word order in unmarked contexts with
unaccusatives only?

In other words, what is special about unaccusatives that enables their subjects
to remain low, differently from what happens with intransitives and transi-
tives?

b) Why is inversion optional with unaccusatives?

108

Optionality and left-dislocated subjects: semantic and discourse properties

In transitive and intransitive contexts, there is a one-to-one correspondence
between word orders and discourse contexts. In this sense, it may be argued
that there is no true optionality in the language, since each word order
serves different discourse functions. In this case, however, there seems to be
true optionality, since two different word orders emerge for one single dis-
course context. The answer to this question must involve showing that the
optionality is just apparent, or providing an analysis predicting two optional
outputs. A finer look at the data will reveal that the former approach is right.

5.1.1. Lack of agreement in postverbal position

Before examining the context favoring the alternation SV-VS, let us now
turn to an apparently unrelated problem: the patterns of verbal agreement
with pre- and postverbal subjects. As the following examples show, verbal
agreement does not vary with word order alternations. This is true for transi-
tives (4) and for intransitives (5). A plural subject triggers plural agreement
in preverbal position and in postverbal position. All sentences in which the
default 3rd person singular agreement appears are ungrammatical:

(4)

SVO:

a. Os meninos comeram o bolo.

the kids ate-3pl the cake

a.’ *Os meninos comeu o bolo.

the kids ate-3sg the cake

VSO:

b. Comeram os meninos o bolo.

ate-3pl the kids the cake

b.’ *Comeu os meninos o bolo.

ate-3sg the kids the cake

VOS:

c. Comeram o bolo os meninos.

ate-3pl the cake the kids

c.’ *Comeu o bolo os meninos.

ate-3sg the cake the kids

SV and VS in unaccusative contexts

109

(5)

SV:

a. Os meninos cantaram.

the kids sang-3pl

a.’ *Os meninos cantou.

the kids sang-3sg

VS:

b. Cantaram os meninos.

sang-3pl the kids

b.’ *Cantou os meninos.

sang-3sg the kids

At least for these verb classes, it seems to be possible to generalize that, in-
dependently of the word order, subject and verb obligatorily agree in person
and number.

A problem for this generalization appears when agreement with postver-

bal subjects of unaccusative verbs is considered.

In colloquial speech, it is

possible for plural subjects of unaccusative verbs to trigger 3rd person sin-
gular verbal agreement. As illustrated in (6), lack of agreement is optional.
It is also possible for the verb to fully agree with the postverbal subject:

(6)

a. Chegaram o Pedro e o Paulo.

arrived-3pl Pedro and Paulo

a.’ Chegou o Pedro e o Paulo.

arrived-3sg o Pedro e o Paulo

b. Fecharam muitas fábricas.

closed-3pl many factories

b.’ Fechou muitas fábricas.

closed-3sg many factories

c. Chegaram as cadeiras.

arrived-3pl the chairs

c.’ Chegou as cadeiras.

arrived-3sg the chairs

Note that this is not restricted to coordinated subjects, as (6b) illustrates, or
to weak quantifiers, as (6c) illustrates. Even a definite DP may occur in this

110

Optionality and left-dislocated subjects: semantic and discourse properties

construction. Partial agreement with coordinated subjects is not restricted to
a single verb class, as discussed in Colaço (1998):

(7)

a. Brincou o Paulo e o Pedro.

intransitive

played Paulo and Pedro

b. Chegou o Paulo e o Pedro.

unaccusative

arrived Paulo and Pedro

c. Comeu-o o Paulo e o Pedro.

transitive

ate it Paulo and Pedro

The possibility of obtaining partial agreement is restricted to the postverbal
position. In preverbal position, unaccusative and intransitive verbs behave
alike in disallowing partial agreement between the subject and the verb:

(8)

Intransitive verbs:

a. O Pedro e o Paulo cantaram.

Pedro and Paulo sang-3pl

a.’ *O Pedro e o Paulo cantou.

Pedro and Paulo sang-3sg

b. Muitos meninos brincaram.

many children played-3pl

b.’ *Muitos meninos brincou.

many children played-3sg

(9)

Unaccusative verbs:

a. O Pedro e o Paulo chegaram.

Pedro and Paulo arrived-3pl

a.’ *O Pedro e o Paulo chegou.

Pedro and Paulo arrived-3sg

b. Muitas fábricas fecharam.

many factories closed-3pl

b.’ *Muitas fábricas fechou.

many factories closed-3sg

SV and VS in unaccusative contexts

111

This behavior of subjects of unaccusative verbs with respect to agreement
raises the following questions:

a) What is the relation between the postverbal position and the lack of

agreement?

In other words, why is lack of agreement possible only in postverbal position?
What makes agreement obligatory when the subject is preverbal?

b) Why is this option restricted to unaccusatives?

Like the inverted word order is only felicitous in sentence-focus contexts
with unaccusative verbs, here, it may be observed that the lack of agreement
is only possible with unaccusative verbs. An explanation for this restriction
to one verb class is needed.

c) What is the relation between the unmarkedness of inversion and the

agreement patterns discussed?

Finally, one would like to know whether there is any relation between the
two problems discussed. Arguably, it is not a coincidence that the only verbs
that allow for a felicitous inverted word order are the only ones that allow
for lack of agreement. Ideally, one single explanation for the two phenomena
described could be found.

5.1.2. Hypothesis and arguments

Let us make the following assumptions:

a) Subject-verb agreement may be taken as a diagnostic to detect whether

nominative Case is assigned to the subject;

This assumption is a common one. For the specific case of European Portu-
guese, it is particularly straightforward to make this assumption, since agree-
ment is enough to license nominative case even in non-finite contexts (cf.
Raposo 1987):

(10)

O Paulo pensa irem eles à praia.
Paulo thinks go-3pl they-Nom to the beach

112

Optionality and left-dislocated subjects: semantic and discourse properties

b) Postverbal subjects of unaccusative verbs may be assigned some other

case.

In analyses of unaccusative verbs, such as Belletti’s (1988), it is proposed
that postverbal subjects of unaccusative verbs are assigned partitive Case.
Such proposals are crucial for making it possible to postulate that nominative
Case is not the only way to license the argument of the unaccusative verb.

c) Language-internal variation may be due to specific properties of lexical

items.

In other words, I will assume that the difference between the construction
with or without agreement has to do with the lexical properties of unac-
cusative verbs.

Accepting these assumptions, let us now formulate the hypothesis for

explaining the behavior of unaccusative verbs, which will partially follow
Belletti’s (1988) analysis of unaccusative verbs in Italian:

(11)

Arguments of unaccusative verbs in colloquial European Portuguese
are not obligatorily assigned nominative Case.

This hypothesis basically argues in favor of an ambiguous status for unac-
cusative verbs. Either their complements must receive nominative Case,
hence must move to the preverbal position Spec,IP, or they may be assigned
partitive Case in situ, as in Belletti (1988). In the latter case, there is no need
for them to move, and Spec,IP is occupied by an expletive, which brings the
unaccusative inversion close to locative inversion constructions. In Costa
and Figueiredo Silva (2003), evidence is presented in favour of this analysis.

In Coelho et alii (2001) and Costa et alii (2002), it is argued that inversion

with unaccusative verbs in Brazilian Portuguese is an instance of locative
inversion. This proposal is based on the following observations. First, there
is no true optionality. It is not the case that in sentence-focus contexts, SV
and VS alternate freely, as shown in the examples in (12)-(14):

(12)

O que aconteceu?
What happened

a. Caiu um avião.

Fell an airplane

b. ??Um avião caiu.

An airplane fell

SV and VS in unaccusative contexts

113

(13)

O que aconteceu?
What happened

a. Nasceram 93 bebés.

Were born 93 babies

b. ??93 bebés nasceram.

93 babies were born

(14)

O que aconteceu?
What happened

a. Um frigorífico descongelou.

A freezer unfroze

b. ??Descongelou um frigorífico.

In fact, both definiteness effects and subtypes of unaccusative predicates
seem to favour or disfavour the VS order in this context. The second argu-
ment for analyzing VS as locative inversion comes from the comparison
with English. Just like in English, VS in Brazilian Portuguese does not pro-
vide information-focus on the subject. The same typically holds for English
locative inversion and expletive constructions:

(15)

Brazilian Portuguese:

Quem morreu?
Who died

a. #Morreu o PM.

Died the PM.

b. O PM morreu.

The PM died.

(16)

English:

Who comes?

a. A man comes.
b. #There comes a man.

Based on these two observations, Costa and Figueiredo Silva (2003) hy-
pothesize that unaccusative VS sentences are locative inversions in Brazilian

114

Optionality and left-dislocated subjects: semantic and discourse properties

Portuguese (cf. Pinto 1997; Ambar 1998; Cornish 2002 for other languages).
Assuming with Levin and Rappaport Hovav (1995) that unaccusatives enter-
ing locative inversion constructions are associated to (potentially unrealized)
locative or temporal argument (cf. Pustejovsky (1995)), and that expletive
pro may correspond to the temporal / locative argument (Pinto 1997), the
Brazilian Portuguese facts follow straightforwardly. Since there is expletive
pro in Brazilian Portuguese, we expect this language to instantiate VS only
in the context in which locative inversion is adequate: the context of presen-
tational focus in which locative inversions are uttered.

This analysis makes some further predictions. First, it is expected that not

all unaccusatives invert, since according to Levin and Rappaport Hovav
(1995), some unaccusative predicates are not associated to locative or tem-
poral arguments. This would be the case for the predicate descongelar/un-
freeze, as illustrated in (17). Second, it is predicted that a few verbs other
than unaccusatives, selecting the same kind of argument, invert in Brazilian
Portuguese. This would be the case for the verb telefonar (to call), as also
argued in Pinto (1997), and Cornish (2002).

(17)

O que é que aconteceu?
What happened

Telefonou a Maria.
Called Maria

Finally, since information focus and presentational focus are not identical, it
is correctly predicted that they may be linked to different types of syntactic
constructions. In fact, English provides independent evidence for distin-
guishing the two types of VS order in Portuguese.

This hypothesis straightforwardly explains the agreement patterns

described. When the subject is preverbal, agreement is obligatory, as in
(18). This is because the subject is only moved to this position when it is
looking for nominative Case, and when expletive pro is not in Spec,IP:

(18)

Muitas fábricas fecharam
many factories closed-3pl

b. *Muitas fábricas fechou.

many factories closed-3sg

SV and VS in unaccusative contexts

115

In postverbal position, both full and partial agreement may be found, as in
(19):

(19)

a. Fecharam muitas fábricas.

closed-3pl many factories

b. Fechou muitas fábricas.

closed-3sg many factories

The two options are predicted under (11). (19a) corresponds to the construc-
tion in which the argument needs nominative Case, although it has not moved
to Spec,IP. In this sense, it is just like all other cases of subject-verb inversion,
independently of the verb class. (19b) corresponds to the locative inversion
construction in which the argument of the unaccusative verb is assigned
partitive Case. In that case, the verb does not agree with the subject.

Let us consider three arguments that favor the hypothesis formulated in

(11), and the analysis just developed.

Argument A:
The first argument in favor of the hypothesis formulated above comes from
the interaction between the patterns of verbal agreement and the distribution
of nominative pronouns. Pronouns are the only forms of the language that
are morphologically specified for Case. As expected, in preverbal position,
nominative pronouns obligatorily agree with the verb:

(20)

a. Eles chegaram.

they-NOM arrived-3pl

b. *Eles chegou.

they-NOM arrived-3sg

Interestingly, in postverbal position, full agreement is also obligatory, as
illustrated in (21):

(21)

a. Chegaram eles.

arrived-3pl they-NOM

b. *Chegou eles.

arrived-3sg they-NOM

116

Optionality and left-dislocated subjects: semantic and discourse properties

This result is expected under the hypothesis in (11). Since the pronoun is
morphologically specified for nominative Case, it must occur in the context
in which it is assigned case. It is therefore impossible to obtain the nomina-
tive pronoun with partial agreement, since in that Case there would be no
matching between the morphological Case and the structural Case assigned
(partitive). Since the pronominal paradigm of European Portuguese does not
include partitive subject pronouns, a nominative pronoun may not co-occur
with a non-agreeing verb.

Argument B:
According to Raposo and Uriagereka (1990), in dialectal European Portu-
guese, it is possible to find overt expletives in constructions involving
postverbal subjects. Crucially, this possibility is restricted to unaccusative
constructions. It is thus possible to find an overt expletive cooccurring with
a postverbal subject in a sentence like (22):

(22)

Ele morreu muitas pessoas naquele acidente.
he died-3sg many people in that accident

It is however not possible to obtain overt expletives in intransitive (23) or
transitive (24) contexts, independently of the agreement exhibited by the
verb:

(23)

a. *Ele cantaram os meninos.

he sang-3pl the kids

b. *Ele cantou os meninos

he sang-3sg the kids

(24)

a. *Ele comeram os meninos o bolo.

he ate-3pl the kids the cake

b. *Ele comeu os meninos o bolo

he ate-3sg the kids the cake

This behavior is expected under the hypothesis formulated above. Since the
Spec,IP position is not necessarily involved in nominative Case licensing
with unaccusatives, this position may host an expletive with nominative Case
morphology only in unaccusative contexts. With other verbs, nominative
Case associated with this position must be assigned to the postverbal subject.

SV and VS in unaccusative contexts

117

Before closing this section, let me point out that this behavior of unaccu-
satives provides evidence in favor of the traditional analysis of preverbal
subjects in Portuguese (Duarte 1987; Ambar 1992). According to these
analyses, preverbal subjects are in Spec,IP. As mentioned before, this
hypothesis is challenged in Barbosa’s (1995,1996) work, who suggests that
preverbal subjects in null subject languages are instances of left-dislocation.
As noted above, preverbal subjects of unaccusative verbs agree obligatorily
with the verb. This may be explained if preverbal subjects are in Spec,IP. In
this position, they are assigned nominative Case, which is reflected by ver-
bal agreement. If preverbal subjects were left-dislocated, obligatory agree-
ment might not be found. Note that left-dislocation of arguments does not
trigger changes in agreement:

(25)

a. Comeu os bolos.

ate-3sg the cakes

b. Os bolos, comeu-os.

the cakes ate-3sg them

Since, as we have seen, postverbal subjects of unaccusatives do not obliga-
torily agree with the verb, it could be predicted that they might be left-dislo-
cated, as any other subject, according to Barbosa’s proposal. In that case, no
change in agreement would be predicted. However, such prediction is not
confirmed by the data.

5.1.3. Summing up: is there optionality?

The analysis proposed above, based on the proposal of Costa (2001), basi-
cally denies the existence of true optionality. The data examined in Coelho
et alii (2001, 2002) and in Costa and Figueiredo Silva (2003) suggests that
there is no true optionality in the behavior of unaccusative contexts. Taking
the fact that under certain circumstances, as shown above, either VS or SV
are preferred, and given the comparison with Brazilian Portuguese, Coelho
et alii (2001, 2002) suggest that the SV-VS alternation is only apparent, and
that the cases of VS are instances of identificational focus.

Assuming the analysis of identificational focus put forward in Cornish

(2002), the authors claim that the inverted structures correspond to instances
of locative inversion, in which Spec,IP is occupied by an expletive:

118

Optionality and left-dislocated subjects: semantic and discourse properties

(26)

[

pro [ V [

t DP ]]]

As mentioned above, this analysis derives four independent facts: (i) this
word order is restricted to unaccusatives, since this verb class favors locative
inversion (Levin and Rappoport-Hovav 1995); (ii) agreement may be partial
if the inflectional head enters a Spec,head relation with the expletive; (iii)
this construction is possible in Brazilian Portuguese, a language in which
Spec,IP cannot be empty and only expletive pro is available; (iv) it is ex-
pected that not all unaccusative verbs behave alike, since lexical restrictions
determine which types of unaccusatives may enter locative inversion con-
structions.

It is crucial for the purposes of this chapter to note that the optionality

found in unaccusatives is just apparent, and that a closer look at morpholog-
ical and semantic properties of the inverted construction was important to
determine the differences between SV and VS.

5.2. VSO and SVO in answers to multiple wh-questions

Another case of apparent optionality may be found in answers to multiple wh-
questions. As shown in (27) and (28) both VSO and SVO constitute legiti-
mate answers to a multiple wh-question:

(27)

Quem leu o quê?
who read what

(28)

a. Leu o João o livro.

read João the book

b. O João leu o livro.

João read the book

These data, like the data involving unaccusative verbs, challenges the gen-
eralization that, for a single context, there is only one word order available.
Moreover, the data exemplified in (28) is problematic for the generalization
made in the previous chapter that all focus material is rightmost, since in the
answer in (28b) we observe that the focused subject surfaces to the left of
the verb, which is not focused.

In Costa (2002a), I argued that this optionality is just apparent. As it will

be shown, the semantic properties of the two answers are different. Interest-

VSO and SVO in answers to multiple wh-questions

119

ingly, we will conclude that the semantic aspects relevant for making the
SVO answer legitimate favor the left-dislocation of the subject.

5.2.1. Semantic properties of multiple-wh questions

According to Hornstein (1994) and Chierchia (1991), multiple wh-questions
have the following properties:

(29)

A: They can be answered with a pair-list;

B: The subject is D-linked (Pesetsky 1987);

C: The answer must be exhaustive.

Let us then check whether these properties differentiate the two word orders.
It may be observed that property (30A) does not distinguish the two word
orders:

(30)

A: Quem leu o quê?

who read what

B: Leu o João o livro, leu a Maria o jornal e leu o Pedro a revista.

read João the book, read Maria the newspaper and read Pedro the
magazine

(31)

A: Quem leu o quê?

who read what

B: O João leu o livro, a Maria leu o jornal e o Pedro leu a revista.

João read the book, Maria read the newspaper and Pedro read the
magazine

Both word orders may be used with lists of pairs. The other two properties
(exhaustivity and D-linking) differentiate the two types of answers. In this
section, arguments are presented in favor of the following claims:

A. SVO is not exhaustive.

When a multiple wh-question is answered with SVO, it is not necessar-
ily a complete answer.

120

Optionality and left-dislocated subjects: semantic and discourse properties

B. In VSO, the answer and the pair Subject-object are exhaustive.

When a multiple wh-question is answered with VSO, there are no other
possible answers, and there is a uniqueness relation between the subject
and the object.

The following arguments provide evidence in favor of the two claims formu-
lated above:

A – Continuation without assertion:
In spite of the fact that focused elements are rightmost, there is a type of
answer described in Ambar (1998), in which a focused subject appears pre-
verbally, as in (32):

(32)

A: Quem comeu o bolo?

who ate the cake

B: O João...comeu.

João...ate

This type of answer is possible with an intonational break between the subject
and the verb, and it means that the speaker does not provide an exhaustive
answer. In the case of (32), there may be other persons who ate the cake, but
the speaker is not sure about it.

Taking this type of non-assertive answer into consideration, it may be

tested whether something like this is behind the difference between VSO
and SVO in answers to multiple wh-questions. A way of testing this type of
answer is resorting to a continuation without assertion. As (33) and (34) illus-
trate, this type of continuation is possible with SVO but not with VSO:

(33)

A: Quem leu o quê?

who read what

B: O João leu o jornal e a Maria leu a revista…os outros não sei o

que leram.

João read the newspaper and Maria read the magazine…the others,
I don’t know what they read

(34) A: Quem leu o quê?

who read what

VSO and SVO in answers to multiple wh-questions

121

B: Leu o João o jornal e leu a Maria a revista...(*os outros não sei o

que leram.)

Read João the newspaper and read Maria the magazine...the oth-
ers, I don’t know what they read

B – VP-ellipsis:
A similar argument is provided VP-ellipsis. This will only be possible if the
preceding sentence is SVO:

(35)

A: Quem comeu o quê?

who ate what

B: O João comeu a sopa, e a Maria também comeu.

João ate the soup and Maria also ate

(36)

A: Quem comeu o quê?

who ate what

B: Comeu o João a sopa (*e a Maria também comeu)

ate João the soup and Maria also ate

This behavior is expected, considering the proposal formulated above. Since
a VSO answer is exhaustive, there cannot be a copy of the VP stating that
there is a relation between the same object and a different subject.

C – N-words:
Like other arguments, argumental N-words may surface either in preverbal
or postverbal position. This option is conditioned by their discourse function.
If they are focused, they are postverbal. If they are given, they are preverbal:

(37)

Quem chegou?
who arrived?

a. Não chegou ninguém.

not arrive no-one

b. *Ninguém chegou.

no-one arrived.

122

Optionality and left-dislocated subjects: semantic and discourse properties

(38)

O que é que ninguém fez?
what did no-one do

a. Ninguém chegou.

no-one arrived

b. *Não chegou ninguém.

not arrived no-one

According to this behavior, it would be expected that subject N-words would
pattern like other subject DPs in answers to multiple wh-questions. At first
sight, this appears to be the case:

(39)

A: Quem comeu o quê?

who ate what

B: Ninguém comeu a sopa.

Noone ate the soup

(40)

A: Quem comeu o quê?

who ate what

B: Não comeu ninguém a sopa.

not ate noone the soup

It is asserted that noone ate the soup, but the hypothesis that someone else
ate something else is not excluded. Thus, the two word orders are possible.
A different pattern emerges when both the subject and the object are N-words.
In this case, only SVO is possible:

(41)

A: Quem comeu o quê?

who ate what

B: Ninguém comeu nada.

noone ate anything

(42)

A: Quem comeu o quê?

who ate what

B: *Não comeu ninguém nada.

not ate noone anything

VSO and SVO in answers to multiple wh-questions

123

The answer in this case is necessarily exhaustive, since no subjects or objects
are left out of the answer. So far, exhaustivity has been linked to VSO word
orders, which might lead to the expectation that VSO should be the answer
in this case. Note, however, that there is only one possible answer in this
context, and that there is a difference in terms of syntactic markedness be-
tween SVO and VSO. Only the former involves movement of the subject to
Spec,IP, subjects in VSO being licensed under Agree. Since there are not two
options available, there is no reason to use the most marked option.

D – VSO is preferred in correction contexts.
Another context in which VSO can be used is in correction contexts, as in (43):

(43)

A: Ninguém comeu nada.

noone ate anything

B: Comeu o João a sopa.

ate João the soup

In this case, the SVO option is dispreferred. At first sight, there should be
no difference between the context illustrated in (43) and the multiple wh-
question, since in both cases the focus of the sentence will be constituted by
the subject and the object. There is however a crucial difference between
the two contexts. In the case of correction, a contrast is being established. In
the case of answer to a wh-question, the focus is merely informational. A
well-known property of contrastive focus is that it involves exhaustivity and
uniqueness (Szabolcsi 1981, Kiss 1996). Since, as mentioned above, VSO
involves exhaustivity, this word order is preferred in contrastive contexts.

E – Generic contexts:
The final piece of argumentation in favor of a semantic difference between
SVO and VSO answers comes from the emerging pattern in generic contexts,
as in (44):

(44)

Quem come o quê?
who eats what

A: As baleias comem peixes.

whales eat fish

B: ??*Comem as baleias peixes.

eat whales fish

124

Optionality and left-dislocated subjects: semantic and discourse properties

As the example above illustrated, in generic contexts, only SVO is available.
A VSO answer is not possible. This follows straightforwardly from the prop-
erties of generic contexts and from the properties claimed to be associated
with SVO answers to multiple-wh questions.

Genericity implies non-exhaustivity (see Krifka et alii 1995). As the con-

tinuation test and the VP-ellipsis test show, SVO answers are non-exhaustive,
hence it is expected that the word order involving exhaustivity may not be
used in generic contexts.

5.2.2. Syntactic consequences

This study of word order in the context of answer to multiple wh-questions
has some theoretical consequences for the syntactic analysis of the SVO-VSO
alternation:

a) The generalizations concerning apparent optionality and the distribution

of focus may be maintained.

First, the two generalizations concerning the non-existence of optionality and
the distribution of focused constituents may be maintained. SVO and VSO
are not free variants. Instead, they correspond to different semantic options.
Second, the generalization that new information is rightmost is not discon-
firmed by a focused subject appearing in the leftmost position, since these
subjects are always D-linked, which means that they share some properties
with topics.

b) Additional argument against FocP for contrastive focus:

It is traditionally considered that FocP is at the left periphery of the clause,
taken into consideration what happens in languages like Hungarian. Above,
some arguments were presented against the idea that foci occupy a discourse-
related functional category. Yet, one might suppose that a left-peripheral
Focus Phrase would be reserved for contrastive focus. However, the context
discussed in this paper shows that a contrastive and information focus occu-
pies a position to the right (VSO), not emerging in a left-peripheral position.

c) Left-dislocated subjects:

The context discussed in this section makes it possible for a subject to be left-
dislocated. As discussed in the previous chapter, while preverbal subjects in

VSO and SVO in answers to multiple wh-questions

125

sentence-focus contexts exhibit properties that lead to the conclusion that
they are in Spec,IP, preverbal subjects in answers to multiple wh-questions
provide evidence in favor of the analysis claiming that they are left-dislo-
cated. This reinforces the conclusion stating that an analysis of preverbal
subjects in Spec,IP does not obviate their left-dislocation.

The observation made above was that when there is no assertion, the sub-

ject is clause-initial. This behavior is expected, since the subject is D-linked,
and topical or given information tends to be sentence initial.

Let us now observe the syntactic properties of the initial subject:

A – It may be doubled by a pronoun

(45)

A: Quem comeu o quê?

who ate what

B: A Maria…ela comeu a sopa.

Maria…she ate the soup

Pronominal doubling is typical of left-dislocation constructions, which favors
the hypothesis that preverbal subjects in this context are left-dislocated.

This is unlike what happens in cases of sentence-focus:

(46)

A: O que é que aconteceu?

what happened

B: *A Maria…ela comeu a sopa.

Maria…she ate the soup

Since doubling in this context is ungrammatical, it may be defended that sub-
jects are not left-dislocated in sentence-focus contexts.

B – Indefinite subject are specific:

(47)

A: Quem comeu o quê?

who ate what

B: Um cão…comeu o osso.

(= one of the dogs)

a dog…ate the bone

126

Optionality and left-dislocated subjects: semantic and discourse properties

(48)

A: Quem comeu o quê?

who ate what

B: */??Um cão qualquer comeu o osso.

some dog ate the bone

As shown in (47) and (48), an indefinite preverbal subject in answer to a
multiple wh-question is necessarily specific. If a non-specific indefinite is
used, as in (48), the sentence is ungrammatical. This behavior is not repro-
duced in cases of sentence-focus:

(49)

A: O que é que aconteceu?

what happened

B: Um cão comeu o osso.

a dog ate the bone

B’: Um cão qualquer comeu o osso.

some dog ate the bone

Non-specific indefinites cannot be left-dislocated, which favors the analysis
proposed here for preverbal subjects in answers to multiple wh-questions.
As for Spec,IP, what triggers movement to this position is EPP and Case-
licensing, which is not related with specificity or definiteness. Hence, it is
expected that there is no semantic restriction regarding the type of subject
that can appear in preverbal position in sentence-focus contexts.

The observation that preverbal subjects in two different contexts leads to

the following consequence: the two analyses available for preverbal subjects
in European Portuguese do not contradict each other. They are both necessary
to account for different facts (see Raposo 2000 for a similar conclusion).
Again, as stated at the end of chapter 3, syntax may produce two converging
outputs: one with a left-dislocated subject, and another one with the subject
in Spec,IP. The choice between the two is made by the appropriateness of
each one of them, taken into consideration their semantic properties and the
discourse context.

VSO and SVO in answers to multiple wh-questions

127

6. Subjects in Spec,TP and the interface with

morphology

6.1. Spec,TP available in I-to-C contexts only

Among the several possible landing sites for subjects in the functional
domain, nothing has been said so far concerning the possibility for subjects
to be stranded in Spec,TP.

In the previous chapters, it was argued that preverbal subjects in European

Portuguese move to the specifier of AgrSP, while the verb undergoes short-
V-movement from V-to-T. This analysis derives the non-adjacency between
subject and verb in sentences like (1):

(1)

Ninguém provavelmente leu bem o livro.
no-one probably read well the book

The fact that the verb appears in between two adverbs in (1a,b) shows that it
is not the case that there is no V-movement at all in European Portuguese.
As mentioned in chapter 1, I assume that subject-oriented adverbs must be
TP-adjuncts, since they only appear in the position in between the subject and
the verb. All other positions for the adverb in (2) yield a manner reading:

(2)

a. O João estupidamente entornou o café.

Subj-Or.

João stupidly spilled the coffee

b. O João entornou estupidamente o café

Manner/*Subj-Or.

João stupidly spilled the coffee

c. O João entornou o café estupidamente.

Manner/*Subj-Or.

João stupidly spilled the coffee

In this kind of context, it is possible to show that Spec,TP is not an available
position for the subject, in spite of the fact that European Portuguese allows
for subject-verb inversion (Ambar 1992). In other words, subject-verb
inversion in declarative contexts is not to be analyzed as a case of subject in

Spec,TP and verb in AgrS. The unavailability of Spec,TP is attested in exam-
ple (3), in which the subject is doubled by a pronoun, blocking the topic
reading for the adverb, and the position for the pronoun in between the sub-
ject-oriented adverb and the verb is ungrammatical:

(3)

a. O João…ele estupidamente entornou o café.

João…he stupidly spilled the coffee

b. *O João…estupidamente ele entornou o café.

João…stupidly he spilled the coffee

So far, the evidence shows that the subject cannot stay in Spec,TP. However,
if one looks at wh-questions involving I-to-C movement, it is possible for
the subject to surface right after the subject-oriented adverb:

(4)

a. O que tinha estupidamente o João entornado?

what had stupidly João spilled

b. Quando tinha cautelosamente o João lido o livro?

when had carefully João read the book

There is thus an apparent contradiction: while the data in (3) show that
Spec,TP is not an available position for the subject, the data in (4) show that
Spec,TP is an available position for the subject. This puzzle becomes more
evident in the next example:

(5)

a. O João estupidamente tinha já entornado o café.

João stupidly had already spilled the coffee

b. O João…(ele) estupidamente (*ele) tinha (*ele) já (??ele) entor-

nado (ele) o café.

João…(he) stupidly (he) had (he) already (he) spilled the (he) the
coffee

c. O que tinha (ele) estupidamente (ele) já (??ele) entornado?

what had (he) stupidly (he) already (he) spilled

The crucial contrast is the one between the pronouns in bold in (5b) and (5c).
(5b) shows that, in the declarative sentence, the pronouns cannot occur in
any position between the subject-oriented adverb and the adverb já ‘already’.
In the interrogative context (5c), however, the position in between the two
adverbs is an available position.

130

Subjects in Spec,TP and the interface with morphology

The analysis outlined above for subject non-adjacency and the puzzle
regarding Spec,TP raise at least the following two questions:

(6)

a. If V does not raise to AgrS, how do Agr morphemes merge with V?
b. Why is Spec,TP an available position for subjects in I-to-C con-

texts only?

6.2. The interface with morphology

The suggestion I would like to make is that the availability of Spec,TP is a
consequence of morphological merger of AgrS to V.

Let me start by providing some background on how morphological merger

operates. According to some works in the framework of Distributed Mor-
phology (Halle & Marantz 1993, Bobaljik 1995), affixation takes place in
the Morphological component of the grammar. The fusion of heads is possi-
ble under syntactic adjacency, and lexical insertion is made in single slots.
Bobaljik (1995) provides two potential scenarios illustrating how affixation
may operate. Suppose there is cyclic head-movement, creating the syntactic
unit in (7):

Head-Movement A:

(7)

Agr

AgrO

As defended by Bobaljik, this type of object has consequences for morpho-
logical fusion, since there must be two independent morphemes for T and
Agr. This is because in a first moment V merges with AgrO, under syntactic
adjacency, while in a second step the unit V/AgrO would merge with T.
Now, this second step is impossible, since it would predict that the tense
morpheme and the verbal root would be competing for insertion in the same
slot. As a consequence, fusion does not take place, and both T and Agr mor-
phemes may cooccur.

The interface with morphology

131

The second type of scenario discussed by Bobaljik (1995) is the one in (8):

Head Movement B:

(8)

Agr
fo

AgrO

AgrS

AgrO T

AgrS

According to Bobaljik, this is the type of head created if T has weak N-fea-
tures, not being able to attract AgrO. Under such circumstances, AgrS attracts
T, and the complex V+AgrO. Unlike in (7), T and AgrS are syntactically
adjacent, therefore the two heads may undergo fusion. The consequence for
morphology is that T and AgrS morphemes will now compete for insertion
in the same slot. As mentioned, for the configuration in (8) to be obtained, T
must have weak-N features, hence Spec,TP is unavailable.

The big consequence from this type of analysis is that by looking at the

verbal morphology, one may know whether Spec,TP is projected. In other
words, if T and AgrS morphemes cooccur in a language, then Spec,TP is
projected. This analysis is the basis for Bobaljik and Jonas’ 1996 [Spec,TP
parameter]. They claim that this is evidence that morphology may act as a
filter on syntactic derivations, and that transitive expletive constructions
provide the syntactic evidence for knowing whether Spec,TP is used as a
landing site for the subject or not.

Let us consider some examples discussed by Bobaljik: Icelandic and

English contrast in that the latter only provides evidence for Agr or T mor-
phology but not for both, while in the former Agr and T morphemes cooccur:

Icelandic: kasta ‘throw’

English: tremble

Present

Past

Present

Past

kasta

kasta-

δi

tremble

tremble-d

kasta-r

kasta-

δi-r

tremble

tremble-d

kasta-r

kasta-

δi

tremble-s

tremble-d

köst-um

köstu-

δu-m

tremble

tremble-d

kast-i

köstu-

δu-δ

tremble

tremble-d

kasta

köstu-

δu

tremble

tremble-d

132

Subjects in Spec,TP and the interface with morphology

The translation of these facts into distributed morphology is the following:
T and AgrS are in complementary distribution in English, competing for
insertion in the same slot. The syntactic correlation is the expected one:
English lacks Icelandic-like transitive expletive constructions.

Note that, since there is no V-to-I in English, affixation must be made

under adjacency, a matter I will return to below.

In languages with V-to-I or V2, the same type of distinctions may be

found:

German (Dutch is similar): sagen ‘say’

Present

Past

sag-e

sag-te

sag-st

sag-te-st

sag-t

sag-te

sag-en

sag-te-n

sag-t

sag-te-t

sag-en

sag-te-n

Like in Icelandic, in German, T and Agr are not in complementary distribu-
tion, and there is evidence for the availability of two subject positions in the
IP-domain.

Bobaljik (1995) crucially presents the contrast between Swedish and

Afrikaans. In both languages, the present tense has no distinctive morphology,
and there is no evidence for competition between T and Agr. In Swedish, the
two types of morphology do not compete, since there is no Agr morphology,
and in Afrikaans, the past tense is formed with periphrastic constructions.

Swedish: smaka ‘to taste’

Present

Past

smaka-r

smaka-de

smaka-r

smaka-de

smaka-r

smaka-de

smaka-r

smaka-de

smaka-r

smaka-de

smaka-r

smaka-de

The interface with morphology

133

Afrikaans: werk ‘to work’

Present

Past

werk
werk

Past formed

werk

with auxiliary verbs

werk
werk
werk

Yet, there is syntactic evidence that, in Afrikaans, Spec,TP is available,
since there are transitive expletive constructions. Based on this contrast,
Bobaljik (1995) suggests that the crucial morphological evidence to know
whether T and Agr are in competition for insertion in the same slot in the
Germanic languages comes from the Past tense morphology.

Before getting back to the Portuguese data, let me just address the issue

of what counts for morphological adjacency, when there is no head-to-head
movement. Halle & Marantz (1993) and Bobaljik (1995) suggest that do-
insertion is used when V and I are not adjacent, blocking morphological
merger. The main idea, schematized in (9) is that any lexical material,
except for adverbs, blocks the adjacency requirement, and force do-inser-
tion.

(9)

a. [

Subj I [

I and V are adjacent, morphological merger is possible

b. [

Subj I [

Adv [

I and V are adjacent, morphological merger is possible

c. [

Subj I [

NegP

not [

I and V are not adjacent, morphological merger is not possible
(do-insertion)

d. [

wh I+C [

Subj t

[

I and V are not adjacent, morphological merger is not possible
(do-insertion)

With the background given above, we now have the necessary tools to
address the problems regarding the availability of Spec,TP in EP.

134

Subjects in Spec,TP and the interface with morphology

Let us start with the first question raised above, namely how does AgrS merge
with the verbal root, if there is no T-to-Agr movement? It is legitimate to
assume that the analysis proposed for verbal morphology in English, a con-
text in which there is no V-to-I movement, applies in EP: since there is no
movement from T to AgrS, there must be morphological merger under adja-
cency. If the subject is in Spec,AgrS, there is adjacency between AgrS and
T, independently of the presence of an adverb adjoined to TP.

This is illus-

trated in (10a). If the subject would stay in Spec,TP like in (10b), there
would be no adjacency and morphological merger would be impossible:

(10)

a. [

AgrSP

Subj AgrS [

(Adv) T+V [

I and V are adjacent, morphological merger is possible

b. [

AgrSP

AgrS [

Suj T+V [

I and V are not adjacent, morphological merger is not possible

So far, this straightforwardly explains how AgrS is merged with the verbal
root, and why Spec,TP is not an available position for subjects in declarative
contexts. Recall that, according to Bobaljik (1995), there is a correlation be-
tween the availability of Spec,TP and the existence of two slots for Agr and
T. His comparison between the several germanic languages also shows that
the past tense paradigms are the crucial ones. Crucially, the past tense in
European Portuguese only displays evidence for a single slot. It is not possi-
ble to distinguish independent T and Agr morphemes in the past tense:

(11)

Past tense falar ‘to speak’

fale-i
fala-ste
falo-u

falá-mos
fala-stes
fala-ram

It remains to be explained why Spec,TP is available for subjects in interrog-
ative contexts. As argued in Ambar (1992), wh-questions with bare wh-forms
involve I-to-C movement. European Portuguese does not have any strategy
like do-support. This implies that in order for the verb to go to I, T must raise
to AgrS, in compliance to the Head Movement Constraint:

The interface with morphology

135

(12)

a. [

C+T+AgrS [

AgrSP

t [

The head created in C has the shape in (13):

(13)

Agr C

Agr

Recall from Bobaljik (1995) and Bobaljik and Jonas (1996) that, in a head
like this, there may be no fusion of nuclei, otherwise the verbal root and T
would be competing for the same slot. If neither fusion nor morphological
merger can apply, adjacency between Agr and T is no longer relevant. The
syntactic consequence is that nothing prevents using Spec,TP as a position
for the subject. This explains why Spec,TP is only available when there is I-
to-C movement.

This analysis might make a different prediction, as pointed out to me by

J. Bobaljik (p.c.). It might be the case that when the verb moves to C, through
cyclic head movement, a different morphology would show up. This is how-
ever not the case. I will leave this issue unsolved here.

Two additional arguments show that the availability of Spec,TP depends

on the existence of I-to-C movement.

First, one may consider other adverbs that may only be adjoined to TP.

Such a case is the adverb sempre ‘always’, which occurring preverbally
means something like ‘after all’ (cf. Gonzaga 1997). This adverb provides
more robust evidence for the unavailability of Spec,TP in declarative con-
texts, since, unlike subject-oriented adverbs, it cannot be topicalized:

(14)

a. O João sempre tinha feito o trabalho.

João after all had done the work

b. *Sempre o João tinha feito o trabalho.

after all João had done the work

As predicted by the analysis, in questions involving I-to-C movement, the
subject may occur after this adverb:

136

Subjects in Spec,TP and the interface with morphology

(15)

a. ?Que trabalho tinha sempre o João feito?

which work had after all João done

b. Esses trabalho todos, tinha sempre o João feito?, não tinha?

all those works, had always João done?, hadn’t he?

Second, as shown in Raposo (1987), inflected infinitives in European Portu-
guese often involve I-to-C movement. Raposo (1987) shows that in factive
contexts I-to-C movement is not obligatory, as attested by the grammatical
word orders in (16):

(16)

a. Os alunos lamentam os deputados terem votado a proposta.

the students regret the deputies have-3pl voted the proposal

b. Os alunos lamentam terem os deputados votado a proposta.

the students regret have-3pl the deputies voted the proposal

Inflected infinitives provide a good testing ground for the proposal made in
this paper for two reasons. First, topicalization is impossible in this context
(Barbosa 2000, Costa and Gonçalves 1999):

(17)

a. * Eu lamento, esse livro, terem eles lido.

I regret that book have-3pl they read

b. * Eu lamento terem, esse livro, eles lido.

I regret have-3pl that book they read

Therefore, a pre-subject position for a subject-oriented adverb may not be
taken as an instance of topicalization of the adverb. Second, since I-to-C
movement is optional, the prediction is that the subject will be occurring in
the post-adverbial position (Spec,TP), if the verb is clause-initial, which
indicates that there is I-to-C movement. This prediction is borne out. (17a)
shows that the subject-oriented adverb may occur in between the subject
and the auxiliary verb. (17b) shows that if there is I-to-C movement, the
subject may remain in Spec,AgrSP, in the pre-adverbial position. The cru-
cial contrast is the one between (17c) and (17d). (17c) shows that, if there is
no I-to-C movement, the subject cannot occur in between the adverb and
the auxiliary verb. In other words, if there is no I-to-C movement the sub-
ject cannot stay in Spec,TP. In (17d), I-to-C movement occurred, and the
subject may surface after the adverb, in Spec,TP.

The interface with morphology

137

(17)

a. Os meninos lamentam os deputados cautelosamente terem votado

a proposta.

the children regret the deputies carefully have-3pl voted the pro-
posal

b. Os meninos lamentam terem os deputados cautelosamente votado

a proposta.

the children regret have-3pl the deputies carefully voted the pro-
posal

c. *Os meninos lamentam cautelosamente os deputados terem votado

a proposta.

the children regret carefully the deputies have-3pl voted the pro-
posal

d. Os meninos lamentam terem cautelosamente os deputados votado

a proposta.

the children regret have-3pl carefully the deputies voted the pro-
posal

6.3. The non-parametric availability of Spec,TP

The study developed in this chapter permits drawing the conclusion that the
availability of a specific A-position for subjects may derive from specific
constructions rather than from a global parameter.

This is theoretically interesting, since it raises a number of issues regard-

ing the format of the [Spec,TP parameter] and the null subject parameter.
The data from EP regarding the availability of Spec,TP suggest that there is
not really a parameter dividing languages as far as Spec,TP is concerned but
much more local constraints determining whether this syntactic position
may be used language-internally.

Another conclusion that may be drawn is that the comparison between

Romance and Germanic languages regarding structure of the clause may be
established through different paths. The unavailability of transitive expletive
constructions in Romance does not obviate taking into consideration these
languages for the discussion of the spec-TP parameter.

Finally, and getting back to the general question concerning the types of

constraints that affect the availability of A-positions as landing sites for sub-
jects, this discussion further confirms that there is not a single answer that

138

Subjects in Spec,TP and the interface with morphology

may be given to this question, since completely different factors seem to
determine the usage of different A-positions: while the availability of
Spec,AgrSP seems to be constrained by the syntax-discourse interface,
Spec,TP appears to be constrained by the syntax-morphology interface.

The non-parametric availability of Spec,TP

139

7. Syntactic outputs and the interfaces

The discussion carried out in the preceding chapters enables us to address
the issue of how the syntactic component interacts with interface issues. I
argued that syntax proper does not need to encode discourse notions. The
behaviour of subjects appears to indicate that, whenever two options can be
generated in the syntactic component, discourse constraints may choose the
one that best suits its purposes. However, in some circumstances, as discussed
in chapter 4, the two generations cannot be generated in the syntactic com-
ponent, for instance, for locality reasons. In that case, even if a subject is
focused, it will not be able to occur in the inverted position.

This type of data leads to the formulation of the following hypothesis:

(1)

When the computational system generates multiple convergent out-
puts, interface constraints may filter or select them.

The immediate consequence of this hypothesis is twofold: first, the interface
conditions do not act as syntactic triggers; second, the interface conditions act
on outputs either as selectors of one of multiple options, or as filters. Another
consequence of this proposal is that whenever only one output is generated
by the syntactic component, ambiguity and/or different types of interface
disambiguation strategies may take place.

Considering the interface with prosody and discourse, mentioned above,

it appears that if the syntax of a language allows multiple outputs in which
different constituents are placed at the position in which they bear sentence
nuclear stress, this will be the preferred option. Otherwise, stress shift is used
as a marked option. Under this alternative view, the syntactic rearrangement
is always preferred over stress-shift strategies. The goal of this chapter is to
provide further empirical argumentation for this view on the relation between
interface conditions and the syntactic component. The arguments to be pre-
sented will favour Chomsky’s (2001) proposal that the computational system
is ‘blind’ to most interface considerations.

In section 1, I will discusse the the behaviour of ditransitives in English

(Brandt 1999), and in European Portuguese; section 2 will address the behav-
iour of possessives in Portuguese and Italian (Cardinaletti 1998; Castro and

Costa 2003); in section 3, the differences between subject-verb inversion in
Brazilian and European Portuguese (Costa and Figueiredo Silva 2003).
These three types of evidence appear to lead to the following conclusion:
stress shift is used only when, for syntactic reasons, a constituent cannot
appear at the position in which it is assigned nuclear stress.

7.1. The behaviour of ditransitives in English and in European

Portuguese

The ditransitive alternation in English between V-DO-PP and V-IO-DO has
been argued to correlate with focus effects: if the direct object is the focus
of the sentence it appears rightmost, if the indirect object is the focus, the
order in which it is rightmost is used (Brandt 1999):

(1)

Who did you give the book to?
I gave the book to Mary.

(2)

What did you give to Mary?
I gave her a book.

Not all ditransitive verbs exhibit these two possibilities (e.g. Jackendoff 1990).
In those cases, if the non-final constituent is focalized, a marked stress must
be used:

(3)

What did you donate to the library?

a. I donated the BOOKS to the library.
b. *I donated the library the books.

These facts indicate that word order alternations are used for discourse pur-
poses, only if the syntax generates two options. When only one option is cre-
ated, stress-shift strategies are used. Note that, according to most authors, the
verb in (4) does not alternate due to lexical semantic reasons (e.g. Marantz
1984).

In European Portuguese, there is also a relationship between word order

and focus in ditransitive contexts, as illustrated in (4) and (5):

(4)

A: A quem é que deste o livro?

to whom did you give the book

142

Syntactic outputs and the interfaces

B: Dei o livro [

ao Paulo]

(I)gave the book to Paulo.
#Dei ao Paulo o livro.

(5)

A: O que é que deste ao Paulo?

what did you give to Paulo

B: Dei ao Paulo [

o livro].

(I) gave to Paulo the book
#Dei o livro ao Paulo.

Interestingly, if the focused argument must bind an anaphor contained within
the non-focused argument, the former cannot be rightmost. In that case, the
focused argument bears a heavy stress:

(6)

A: A quem é que deste os livros?

to whom did you give the books?

B: Dei [

A CADA AUTOR] o seu livro.

(I) gave to each author his book.
?*Dei o seu livro a cada autor.

(I) gave his book to each author

(7)

A: O que é que deste aos autores?

what did you give to the authors

B: Dei [

CADA LIVRO] ao seu autor.

(I) gave each book to its author
?*Dei ao seu autor cada livro.
(I) gave to its author each book

In Costa (1998a), it is argued that this behavior of focus binders provides
evidence for an Optimality-theoretical approach to focus, since the constraint
forcing foci to be rightmost can be violated when it conflicts with binding
requirements. However, this account failed to accommodate the stress-shift
effects. Under the view advocated here, the output in which the focus c-
commands the anaphor is the only legitimate ouput. Accordingly, syntax is
not generating two convergent ouputs, and stress-shif acts as a last resort
strategy.

Incidentally, this type of data not only sheds light on the issue of interfaces

but also on the structure of ditransitive contexts. So far, we have considered

The behaviour of ditransitives in English and in European Portuguese

143

cases in which syntactic transformations generate multiple converging
ouputs. The binding facts, and an asymmetry with VOS shows that multiple
converging outputs can also be the result of optional base-generations.

Let us first consider the debate considering the structure of VP in ditran-

sitive contexts (Kayne 1984; Larson 1988; Pesetsky 1995; Phillips 1996;
among others). As noted in Phillips (1996), part of the debated problems
derives from the fact that constituency tests yield contradictory results, when
applied to ditransitive VPs. For instances, NPI-licensing in (8) provides evi-
dence for a shell structure like in (10), while the fronting tests provide evi-
dence for a layered structure like in (11):

(8)

a. John gave nothing to any of the children on his birthday.

b. *John gave anything to none of the children on his birthday.

(9)

John intended to give candy to children on his birthday…

a. …and [give candy to children on his birthday] he did.

b. …and [give candy to children] he did on his birthday.

c. …and [give candy] he did to children on his birthday.

(10)

(11)

V’

DO V’

An asymmetry in binding may shed light on this debate. In Costa (1996,
1998), and in chapter 4, it was argued that in SVO orders the subject is in
Spec,IP. VSO and VOS orders instantiate subjects stranded in Spec,VP.
According to these results, the difference between VSO and VOS lies on the
status of the object. In VOS orders, the objects scrambles out of VP, creating
an adjunction configuration. Following this analysis, depicted in (12), scram-
bling is an instance of A-bar-movement:

144

Syntactic outputs and the interfaces

(12)

Viu a Maria o João.
[

V [

DO [

Subj t t ]]]

One of the arguments for the A-bar nature of object scrambling came from
binding facts. As illustrated in (13) scrambling of an object to the left of the
subject does not feed binding, unlike what happens in cases of A-movement
(13c).

(13)

a. *Viu o seu filho cada mãe.

VSO

saw her son each mother

b. *Viu cada mãe o seu filho.

VOS

c. Cada mãe foi vista pelo seu filho.

Passive

The binding facts in ditransitive contexts illustrated above pose a challenge to
this generalization. In particular, these facts raise the following two questions:

What is the nature of indirect object scrambling?

Why doesn’t indirect object scrambling across a direct object yield
the same results as direct object scrambling across a subject?

Rephrasing the problem, the binding facts suggest that an object moved
across a subject seems to be A-bar moved, while an indirect object moved
across the direct object seems to occupy an A-position.

The solution I will offer to this puzzle is based on the assumption that

syntactic structure is strictly binary, and generated from left-to-right, as pro-
posed in Phillips (1996).

I follow Belletti and Shlonsky (1995) in their claim that the order direct

object-indirect object in Romance is unmarked. Evidence in favor of this
comes from the emerging word order in sentence-focus contexts (14), and
from the order found in idioms formed with ditransitive verbs (15–16):

(14)

O que é que aconteceu?
What happened

a. O João deu uma prenda à Maria.

João gave a gift to Maria

b. #O João deu à Maria uma prenda.

João gave to Maria a gift

The behaviour of ditransitives in English and in European Portuguese

145

(15)

Dar pérolas a porcos.
To give pearls to pigs

*Dar a porcos pérolas.

To give to pigs pearls

(16)

Pôr mais lenha na fogueira.
To put more wood in the fire

*Pôr na fogueira mais lenha

to put in the fire more wood

However, I deviate from Belletti and Shlonsky (1995), since I assume that
the fact that this word order is unmarked does not necessarily mean that it
correspond to a base-generated word order.

Note that constituency tests also yield contradictory results in EP: on the

one hand, binding (18-19) and scopal facts (20) provide evidence for a shell-
structure:

(18)

a. Entreguei cada livro ao seu autor.

(I) gave each book to his author

b. Entreguei a cada autor o seu livro.

(I) gave to each author his book

(19)

a. *Apresentei a Maria

à filha da Maria

(I) introduced Maria to the daughter of Maria

b. Apresentei à filha da Maria

a Maria

(I) introduced to the daughter of Maria Maria

(20)

a. Apresentei uma mulher a todos os homens.

∃ > ∀ (preferred)

(I) introduced a woman to all men

b. Apresentei a todos os homens uma mulher.

∀ > ∃ (preferred)

(I) introduced to all men a woman

On the other hand, gapping (21) and fronting (22) provide evidence in favor
of a layered structure:

146

Syntactic outputs and the interfaces

(21)

a. Eu dei livros à Maria e o Pedro [ deu livros] à Ana.

I gave books to Maria and Pedro to Ana

b. Eu pus os livros na prateleira e tu [puseste os livros] na gaveta.

I put the books on the shelf and you in the drawer.

(22)

O Pedro queria dar os livros ao Rui ontem.
Pedro wanted to give the books to Rui yesterday

a. …e [dar os livros] ele deu ao Rui ontem.

and give the books he gave to Rui yesterday

b. …e [dar os livros ao Rui] ele deu ontem.

and give the books to Rui he gave yesterday

c. …e [dar os livros ao Rui ontem] ele deu.

and give the books to Rui yesterday he gave

It is important to note that the evidence for layered structures is not crystal-
clear, since with other types of ditransitives, stranding one of the arguments
is not good, a problem I will not address here:

(23)

O Pedro queria pôr os livros na prateleira ontem.
Pedro wanted to put the books on the shelf yesterday

a. ??*…e [pôr os livros] ele pôs na prateleira ontem.

and put the books he put on the shelf yesterday

b. …e [pôr os livros na prateleira ] ele pôs ontem.

and put the books on the shelf he put yesterday

c. …e [pôr os livros na prateleira ontem] ele pôs.

and put the books on the shelf yesterday he put

The contradictory tests suggest that an analysis along the lines of Phillips’
(1996) may be on the right track. The consequence of this type of analysis is
that there are two basic ways of generating the two word orders, even if only
one of them is unmarked. The conclusion from this type of approach is that
there is no direct relation between unmarkedness and base-generation.
Phillips’ analysis is based on the following two principles:

The behaviour of ditransitives in English and in European Portuguese

147

(24) MERGE RIGHT

New items must be introduced at the right edge of a structure.

(25) BRANCH RIGHT

Where a terminal can be attached to more than one position in the
existing structure with no effect on interpretation, the attachment that
results in the more right-branching structure must be chosen.

The interaction between these two principles operates in the following way
for the generation of V-DO-IO and V-IO-DO orders:

(26)

V-DO-IO:

Step 1: merge V with Direct Object, discharging theta-role

Step 2: Create a copy of V (reanalyzing the direct object as a specifier), and

merge a PP as its sister, discharging the goal theta-role.

V’

Phillips’ analysis predicts that constituency tests may target a step of the
derivation in which V and DO form a VP, or the final structure, in which there

148

Syntactic outputs and the interfaces

is c-command of the DO over the PP. For generating V-IO-DO order, (24) and
(25) operate in the following way:

(27)

V-IO-DO:

Step 1: merge V with Indirect Object, discharging goal theta-role

Step 2: Create a copy of V (reanalyzing the indirect object as a specifier),

and merge a DP as its sister, discharging the theme theta-role.

V’

Note that, as mentioned above, claiming that both orders are base-generated
does not entail that they should be optional. The structure built in (27) is only
generated if necessary for satisfying binding requirements (or any other con-
straint forcing the order IO-DO, such as heaviness).

Suggesting that V-IO-DO can be base-generated for binding purposes

makes two interesting predictions. First, as it is also pointed out in Phillips
(1996), it is expected that no case is found in which the verb and the indirect
object may be fronted stranding the direct object. This is confirmed by the
data in (28):

The behaviour of ditransitives in English and in European Portuguese

149

(28)

O Pedro queria entregar os prémios aos vencedores...
Pedro wanted to give the prizes to the winners

…*e entregar aos vencedores ele entregou os prémios.
and give to the winners he gave the prizes

The only circumstances under which V-IO fronting is possible with DO
stranding are cases in which binding is involved. Although the contrast is
subtle, (29) is better than (28):

(29)

O Pedro queria entregar a cada vencedor o seu prémio...
Pedro wanted to give to each winner his prize

…?e entregar a cada vencedor ele entregou o seu prémio.
and give to each winner he gave his prize

Second, this type of approach straightforwardly predicts well-known cases
of indirect object control. In (30), PRO contained in the direct object is con-
trolled by the indirect object, and it is not necessary to assume that the clause
is right-dislocated, which would be problematic. The analysis I am suggesting
predicts that both DO and IO occupy A-positions, and the control pattern
follows.

(30)

pedi aos meninos

para PRO

i/j

dançar.

I asked to the children to dance

Let us finally turn to the solution of the puzzle presented: why it is not pos-
sible for a scrambled direct object in VOS to bind into the subject. Recall that
the binding possibilities for V-IO-DO and V-IO-DO were derived from the
fact that both word orders are base-generated. Therefore, no A-bar-movement
takes place. The problem must now be stated in different terms: the issue now
is why VOS cannot be base generated.

In both cases discussed above, the theta-criterion is satisfied: either the

verb or its copy assigns the theta-roles to each one of the arguments. This
entails that the theta-criterion is satisfied in the course of the clause’s gener-
ation (and not pre-syntactically). This consequence strongly clashes with
Baker’s (1988) UTAH. In particular, there is no need to derive one of the
orders in ditransitive contexts from the other, since the theta-criterion may
be satisfied in both cases.

150

Syntactic outputs and the interfaces

Independent evidence for the claim that theta-roles may be assigned in the
course of the derivation, and that the realization of the theta-role may
depend on other factors comes from cases of complex predicate formation
(Wurmbrand 1998; Gonçalves 1999). As discussed in Gonçalves (1999), the
realization of an embedded agent as a DP or as a PP depends on the transi-
tivity of the embedded verb, on the type of main verb, and on the process of
complex predicate formation. This is illustrated in (31):

(31)

a. Eu mandei correr os meninos.

I made run the children

b. Eu mandei ler o livro aos meninos.

I made read the book to the children.

Crucially, if all agents were DPs generated in Spec,VP, the pattern in (31)
would be unexpected. The fact that the realization of the agent as a PP in
(31b) is forced by constraints on complex predicate formation provides
independent evidence for a theory of theta-role assignment that takes into
consideration the sentence as a whole rather than each of the theta-assigning
heads independently. The relevance of the type of main verb for determining
the shape of the embedded agent becomes even more obvious, in the cases
discussed in Gonçalves (1999), in which the matrix verb does not make
available a PP position, and, as a consequence, the agent theta-role may not
be discharged:

(32)

a. Eu vi correr os meninos.

I saw run the children

b. *Eu vi ler o livro aos meninos.

I saw read the book to the children

Returning to the binding difference between V-IO-DO and V-O-S, recall
that the object surfacing to the left of the subject reconstructs, while the IO
across DO is able to bind. As mentioned above, interpreting reconstruction
as a consequence of A-bar movement, and the binding relation between IO
and DO as a consequence of base-generation, the question to be answered is
why the object cannot be base-generated in VOS.

In what follows, I will show that it is not possible to satisfy left-to-right

merge and the theta-criterion, creating a base-generated VOS. Step 1 in (33)
represents the merging of the verb with the direct object, which like in the
case of ditransitive contexts is unproblematic.

The behaviour of ditransitives in English and in European Portuguese

151

(33)

V-O-S:

Step 1: merge V with Object, discharging theme theta-role

Step 2: Create a copy of V (reanalyzing the object as a specifier), and merge

a DP as its sister, discharging the agent theta-role.

Subj

V’

_=>

Subj

As shown above, the problem comes about when the subject is merged with
the copy of the verb, in compliance with MERGE RIGHT and BRANCH
RIGHT. The resulting configuration, illustrated in (33b) does not yield a
proper context for assignment of the external theta-role. In other words, the
problem is that the subject is not a sister to V’, and is therefore unable to
receive theta-role, according to Koopman and Sportiche’s (1991) proposal.

As a consequence, unlike V-IO-DO, VOS cannot be base-generated. Instead,
VOS must be an instance of movement. If it is A-bar movement, the structure
in (34) obtains and the binding possibilities follow.

(34)

[

V [

O [

S t t ]]]

152

Syntactic outputs and the interfaces

Crucially, for the discussion in this chapter, we observe a case in which two
syntactic outputs are base-generated, and are filtered out at the interface
with prosody. One of the two may be filtered out in the syntax, if binding
requirements are not met. In that case, stress shift applies.

7.2. The behaviour of possessives in Portuguese and Italian

Cardinalletti (1998) shows that pre-nominal and post-nominal possessives
are XPs in Italian, and that post-nominal possessives are used in definite
contexts, only if they are focussed:

(35)

a. *la SUA casa, non tua

Cardinaletti (1998: 19 –20)

the her house, not yours

b. la casa SUA, non tua

the house hers, not yours
“Her house, not yours”

There is evidence to argue that Portuguese pre-nominal possessives are
heads (cf. Castro and Costa 2003). This is confirmed by the following facts:

– they cannot be modified (36a);

– they can refer to entities with the feature [-human] (37a);

– they can be phonetically reduced in some dialects of Portuguese (38a).

(36)

a. *O só meu problema é que não percebo nada disto.

The only my problem

b. Um problema só meu é que não percebo nada disto.

a problem only my

(37)

a. A sua

tampa é muito prática.

i = Maria/frying pan

her lid is very practical
“Its/her lid is very practical”

b. Encontrei uma tampa sua

i = Maria/*frying pan

I found a lid her
“I found a lid of hers”

The behaviour of possessives in Portuguese and Italian

153

(38)

a. O m[e] livro

the my book
“My book”

b. Um livro *[me]/[mew]

a book my
“A book of mine”

In spite of there being an association between definiteness and the use of
pre- vs postnominal possessives, what is relevant for the purposes of this
chapter is the asymmetry between European Portuguese and Italian. While
in the latter, the strong postnominal form is used, even in definite contexts,
for focalization, in European Portuguese, this option is not available, hence
the prenominal possessive must be used even in such contexts, associated
with a heavy stress. In Castro and Costa (2003), it is proposed that prono-
minal possessives are weak (non-clitic) heads. This proposal makes the fol-
lowing predictions: first, it is predicted that the forms under analysis cannot
display clitic-like behavior, since they are not clitics;

second, the asymme-

try between the pre- and postnominal possessives is explained. Since the
prenominal weak forms are heads, they cannot be modified, they cannot be
coordinated without focalization, and they can be reduced.

This hypothesis also explains the focalization facts. Unlike in Italian, a

focalized possessive is not post-nominal:

(39)

a. Esse é o MEU problema, não o teu.

That is the MY problem, not the yours

b. *Esse é o problema meu, não o teu.

That is the problem my
“That is MY problem, not yours”

This asymmetry follows from the categorial status of these forms in the two
languages. Since pre-nominal possessives are heads in Portuguese, they do
not enter into configurations typical of XPs, which permit the post-nominal
order. In Italian, because of their XP-status, possessives may surface in two
different configurations. The conclusion that can be drawn from these facts is
that the position of the possessive and its relation with its discourse function
solely derives from its categorial status. If the DP-final position observed for
Italian were a consequence of focus-movement to the right, the asymmetry
between DP-internal behaviour and the availability of a final position for
foci at the clause level would become difficult to explain.

154

Syntactic outputs and the interfaces

In other words, the case of possessives when contrasted to the general avail-
ability of clause-final position for focused items indicates that the correla-
tion between domain-final position and focus is to be evaluated in each con-
struction, depending on its specific properties, rather than generalized for a
specific language. In the case of possessives, we observe that European
Portuguese behaves like a stress-shift language, just because there is no
legitimate syntactic ouput placing the possessive on the DP-final position in
definite contexts.

7.3. The differences between subject-verb inversion in Brazilian and

European Portuguese

As mentioned above, in Brazilian Portuguese, unlike in European Portuguese,
a subject which is focused does not invert, remaining in preverbal position
(Costa and Figueiredo Silva 2003). This contrast between the two languages
is illustrated in (40):

(40)

A: Quem comeu o bolo?

Who ate the cake

B: a. Comeu o João.

(EP/*BP)

Ate João

a.’ O JOÃO comeu.

(*EP/BP)

João ate

As indicated in (40a’), the non-inverted subject bears a heavy stress. This
contrast between the two languages is interesting for the present discussion,
since the two languages diverge only when the behaviour of subjects (of non
unaccusative verbs) is concerned. In other instances of focalization, Brazilian
Portuguese behaves like other languages in which focus occurs clause-finally,
as illustrated in the following examples for ditransitive contexts:

(41)

Brazilian Portuguese:

A: O que o João deu pra Maria?

What João gave to Maria

B: O João deu pra Maria um CD.

João gave to Maria a CD

The behaviour of possessives in Portuguese and Italian 155

(42)

Brazilian Portuguese:

A: Pra quem o João deu o CD?

To whom João gave the CD

B: O João deu o CD pra Maria.

João gave the CD to Maria

(43)

European Portuguese:

A: O que é que o João deu à Maria?

What João gave to Maria

B: O João deu à Maria um CD.

João gave to Maria a CD

(44)

European Portuguese:

A: A quem é que o João deu o CD?

To whom João gave the CD

B: O João deu o CD à Maria.

João gave the CD to Maria

As argued in Costa and Figueiredo Silva (2003), the data in (41)–(44) make
two different points. First, both languages allow for either complement of
the verb to surface clause-finally. Second, as shown by the question-answer
pairs, the choice between order V-DO-IO and V-IO-DO may be made in
compliance with discourse requirements. This case is of particular interest
for the present discussion, since we observe that both languages satisfy the
requirement that the focus of the sentence surfaces rightmost. Therefore,
positing that the subject in Brazilian Portuguese does not invert in (40) be-
cause Brazilian Portuguese does not codify focus in the syntax would not
account for the behaviour of complements of ditransitive verbs. What appears
to be at stake that differentiates the two languages is that, since Brazilian
Portuguese is loosing referential pro (Figueiredo Silva 1996, among others), it
is impossible to leave Spec,IP empty, which makes the inverted construction
ungrammatical. Therefore, we observe another case in which a syntactic out-
put is not generated for a purely syntactic reason, and therefore the best word
order for prosodic purposes does not arise. Again, under such circumstances,
stress shift is used as a last resort.

156

Syntactic outputs and the interfaces

7.4. Conclusions

The three cases presented instantiate mixed systems, in which languages
resort both to word order rearrangements and to stress shift in very similar
contexts. In all cases under consideration, it was shown that syntax proper
does not need to encode or refer to interface considerations. In other words,
the arguments put forward in this chapter provide evidence for a more auto-
nomous syntax, not making reference to interface conditions, and argue for
an articulated view of the grammar, in which the interfaces read the outputs
of syntax, but do not interfere in syntactic derivations.

Conclusions

157

8. Summary and conclusions

The primary goal of this book was to provide an analysis of the several
positions where the subject may surface in European Portuguese. Departing
from an architecture of the clause as sketched in early minimalist work,
containing two subject-related functional categories above VP (AgrP and
TP), it was shown that the subject may surface in all potential landing sites:
Spec,AgrP, Spec,TP and Spec,VP. Moreover, just like any other argument of
the clause, it was claimed that subjects also have the possibility of surfacing
in a left-dislocated position, arguably adjoining to the clause’s left periphery.

It was shown that there is no free variation. Each of these positions may

be occupied by the subject, only if two requirements are met:

i) The position is made available by syntax;

ii) The position does not violate any interface condition.

In other words, the following model was argued for: syntax generates legiti-
mate outputs. At the interface levels, each output may be selected or filtred
out, according to requirements of the interface. This interface licensing con-
ditions operate in the following way for each of the identified surface posi-
tions:

1. Spec, VP – The subject may surface in Spec,VP, because it is able to

check Case under Agree (Chomsky 2000). Likewise, Case may be
checked under Move. The consequence is that both SVO and VSO out-
puts are equally well-formed from a syntactic point of view. It is argued
that Information Structure and their interplay with prosody may choose a
VSO ouput over an SVO order, when the subject is the focus of the sen-
tence and must receive the sentence’s nuclear stress. The idea that Agree
is a legitimate way of licensing an in-situ subject was motivated by the
observation that in-situ subjects are constrained by locality conditions,
definable in terms of phase-boundaries.

2. Spec,TP – This position provided an interesting puzzle. Looking at

adverb positions, it appears that Spec,TP is an available surface position

for subjects in I-to-C contexts. This observation was explained taking
the interface with morphology into account. It was argued that the sub-
ject cannot be stranded in Spec,TP when the subject is in T, since it
blocks the merger of the heads Agr and T.

3. Left-dislocation and Spec,AgrP – Non-focused preverbal subjects

were shown to occupy the specifier of the topmost functional category of
the inflectional domain. This goes against recent claims in the literature
that preverbal subjects in null subject languages are left-dislocated.
Nevertheless, it was shown that the fact that preverbal subjects occupy an
A-position does not imply that they necessarily must occupy an A-posi-
tion. Looking at contexts of apparent optionality in answers to multiple
wh-questions, it was shown that in the appropriate context subjects may
be left-dislocated. For a subject to appear in adjunction to the clause, it
must meet semantic requirements such as non-exhaustivity.

The generalization obtained is that when the computational system gen-

erates multiple convergent outputs, interface constraints may filter or select
them. The immediate consequence of this generalization is twofold: first,
the interface conditions do not act as syntactic triggers; second, the interface
conditions act on outputs either as selectors of one of multiple options, or as
filters. Another consequence of this proposal is that whenever only one output
is generated by the syntactic component, ambiguity and/or different types of
interface disambiguation strategies, such as heavy stress marking, may take
place.

160

Summary and conclusions

As mentioned in chapter 7, the arguments put forward throughout the book
provide evidence for a more autonomous syntax, not making reference to
interface conditions, and argue for an articulated view of the grammar, in
which the interfaces read the outputs of syntax, but do not interfere in syn-
tactic derivations.

Summary and conclusions

161

Appendix:
On the nature of agreement in European Portuguese

Introduction

The discussion carried out in chapter 6 strongly bears on the need for Agr
inflection to merge with the verb in T. Since Agr typically encodes both per-
son and number inflection, it is important to clarify which type of agreement
is at stake here. I therefore dedicate this appendix to report on the results
achieved in Costa and Figueiredo Silva (2003a), who claim that number
morphology is, in most instances, a post-syntactic matter, based on a com-
parison between three register of Portuguese. If these results are on the right
track, the relevant feature for triggering V-movement or morphological
merger is person. Since we contend that the most substancial difference in
the inflectional systems of European and Brazilian Portuguese relate to the
presence of person features, two predictions are made:

(i) Number agreement morphology does not necessarily trace the presence

of a Spec,head configuration;

(ii) There are no crucial differences in terms of V-to-I movement between

the two languages, since in both, there is person morphology triggering
V-to-I movement (Costa and Galves 2002; Vikner 1997).

The main goal Costa and Figueiredo Silva’s (2003a) work is to discuss an
additional argument against the view that verbal morphology should be
explained in terms of verb movement alone. The following facts from three
registers of Portuguese are discussed, some of which were previously noted
in Galves (1993) and Figueiredo Silva (1996):

1. In European Portuguese, and in two registers of Brazilian Portuguese,

there is no evidence to posit different landing sites for the verb. According
to the traditional tests (adverb placement and floating quantifiers), in all
of them, the verb seems to have moved from V to T, without reaching the
topmost functional head (Costa 1996; Costa and Galves 2002).

2. Yet, the agreement patterns in these dialects are different. As shown

below, the following generalizations may be drawn:

– In European Portuguese (EP), all elements able to bear plural morphology

do so:

(1)

Os carros são lindos.
the-pl car-pl are beautiful-pl
‘The cars are beautiful’

– In one of the registers of Brazilian Portuguese (BP1), there is no DP-

internal agreement, although there is subject-verb agreement:

(2)

Os carro são lindo.
the-pl car-sg are beautiful-sg
‘The cars are beautiful’

– In the other register of Brazilian Portuguese (BP2), there is neither DP-

internal agreement, nor subject-verb agreement:

(3)

Os carro é lindo.
the-pl car-sg is beautiful-sg.
‘The cars are beautiful’

These facts raise at least the following questions:

a) If there is robust evidence that in all these language variants, the target of

verb movement is the same, can it be maintained that the morphological
differences follow from verb movement?

b) What type of micro-variation is at play in Portuguese that derives these

differences?

c) Is there any relation between the DP-internal facts and subject-verb

agreement facts?

The hypothesis developed in Costa and Figueiredo Silva (2003a) is that the
effects of visible agreement must be interpreted as a non-unitary phenome-
non. They derive from i) the type of morpheme (singleton or dissociated)
(Embick and Noyer 2001), and ii) whether Spec-head configurations trigger

164

Appendix: On the nature of agreement in European Portuguese

visible agreement. The possible combinations of these factors derive the
variation found across the three variants of Portuguese.

In section 1of this appendix, we will take a closer look at the data, in order

to provide a clearer understanding of the type of morphological variation at
stake; in section 2, the assumptions and proposal for explaining the crosslin-
guistic variation are presented; the predictions of the analysis are explored
in section 3.

1. The facts

The facts under scrutiny stem from two different domains: the number agree-
ment patterns within the DP (section 1.1), and the patterns of subject-verb
agreement (section 1.2). As mentioned, the data comes from three registers
of Portuguese: European Portuguese (henceforth EP), and two varieties of
Brazilian Portuguese (BP1 and BP2).

An important remark must be made concerning the Brazilian Portuguese

data: we are somehow idealizing the distinction between two dialects, since
the same speaker may use BP1 and BP2. This idealization is however con-
firmed by sociolinguistic research showing that in designated situations,
speakers opt for one of the two varieties.

If this observation is correct, we

are dealing with a case of competing grammars, each one used in different
situations, in the sense of Kroch (1994, 1997), and it is possible to tease the
two apart.

1.1. DP-internal number agreement

DP-internally, European Portuguese displays the pattern of number agreement
found in most Romance languages: plurality is expressed in all categories
able to bear this type of morphology (noun, determiner, quantifiers, adjec-
tives, possessives and demonstratives). This is shown in (4):

(4)

a. Os/estes/alguns/uns livros muito bonitos

The-pl / these-pl / some-pl /a-pl book-pl very pretty-pl
‘The/these/some/ books very pretty’

b. Os primeiros livros da biblioteca

the-pl first-pl books-pl of-the library
‘The first books of the library’

The facts

165

c. Os meus livros

the-pl my-pl books-pl
‘My books’

d. Todos os meus primeiros livros bonitos

all-pl the-pl my-pl first-pl book-pl pretty-pl
‘All my first pretty books’

Both dialects of Brazilian Portuguese behave alike and differ from European
Portuguese. In general, plurality is marked just on the determiner. Nouns and
post-nominal adjectives are not marked for plurality:

(5)

Os/estes/alguns/uns livro muito bonito
The-pl / these-pl / some-pl / a-pl book-sg very pretty-sg
‘The/these/some/ books very pretty’

The opposition between prenominal and postnominal positions within the
DP is crucial for establishing the agreement patterns. As noted by Menuzzi
(1994), plural markers may optionally surface on other prenominal elements,
but if the noun is not marked as plural, no post-nominal element may bear a
plural morpheme. This derives some variation in the prenominal domain. Pre-
nominal adjectives may or may not bear plural morphology. A pattern that is
not found is agreement on the adjective with a non-agreeing determiner:

(6)

a. Os primeiros livro da biblioteca

The-pl first-pl book-sg of-the library
‘The first books of the library’

b. Os primeiro livro da biblioteca

The-pl first-sg book-sg of the library
‘The first books of the library’

c. *O primeiros livro da biblioteca

The-sg first-pl book-sg of the library
‘The first books of the library’

Prenominal possessives differ from prenominal adjectives in an interesting
way: just like in the case of adjectives, both the determiner and the posses-
sive may bear the plural morpheme. However, if only one of them is to
agree, the possessive, and not the determiner will bear the plural marker,
which distinguishes this case from what was observed for adjectives:

166

Appendix: On the nature of agreement in European Portuguese

(7)

a. Os meus livro

The-pl my-pl book-sg
‘My books’

b. *Os meu livro

The-pl my-sg book-sg
‘My books’

c. O meus livro

The-sg my-pl book-sg
‘My books’

The emerging generalization from these data is the following: either the
determiner head or all the prenominal elements bear plural morphology in
Brazilian Portuguese. The interesting difference between adjectives and
possessives must be accounted for, since the latter is the only case in which
the plural-bearing element is not the determiner.

1.2. Subject-verb agreement

As shown in the preceding chapters, in European Portuguese, subjects and
verb agree independently of the position of the subject. The only exception
to this is the possibility for inverted subjects of unaccusative verbs not to
agree in colloquial speech (Costa 2000):

(8)

a. Os meninos comeram o doce.

The children ate-pl the candy
‘The children ate the candy’

b. *Os meninos comeu o doce.

The children ate-sg the candy
‘The children ate the candy’

c. Comeram os meninos o doce.

Ate-pl the children the candy
‘The children ate the candy’

d. *Comeu os meninos o doce.

Ate-sg the children the candy
‘The children ate the candy’

The facts

167

e. Comeram o doce os meninos.

Ate-pl the candy the children
‘The children ate the candy’

*Comeu o doce os meninos.
Ate-sg the candy the children
‘The children ate the candy’

(9)

a. Muitas florestas arderam.

Many forests burnt-pl
‘Many forests burnt’

b. *Muitas florestas ardeu.

Many forests burnt-sg
‘Many forests burnt’

c. Arderam muitas florestas.

Burnt-pl many forests
‘Many forests burnt’

d. Ardeu muitas florestas.

(colloquial)

Burnt-sg many forests
‘Many forests burnt’

Small clause predicates and passive participles also display obligatory num-
ber agreement:

(10)

a. As casas parecem bonitas.

The houses seem-pl pretty-pl
‘The houses seem pretty’

b. *As casas parecem bonita.

The houses seem-pl pretty-sg
‘The houses seem pretty’

c. As casas foram destruídas.

The houses were destroyed-pl
‘The houses were destroyed’

d. *As casas foram destruída.

The houses were destroyed-sg
‘The houses were destroyed’

168

Appendix: On the nature of agreement in European Portuguese

Before presenting the subject-verb agreement pattern in the two dialects of
Brazilian Portuguese, it is important to recall that the verbal paradigms are
different, as extensively discussed in the literature, in particular in Galves
(1993) and Figueiredo Silva (1996). In standard European Portuguese, there
are five different combinations of the person and number features:

(11)

Verb cantar ‘to sing’ – present tense

singular

plural

canto

cantamos

II cantas

cantam

III canta

cantam

In BP1, the verbal paradigm consists of four different combinations of the
same features, as shown in (12):

(12)

Verb cantar ‘to sing’ – present tense

singular

plural

canto

canta / cantamos

II canta

cantam

III canta

cantam

Analyzing these differences, Galves (1993) proposes that this dialect of
Brazilian Portuguese has lost a semantic distinction for person, but not its
syntactic feature. In fact, the author proposes that the pattern in (12) derives
from a combination of binary features for person and number:

(13)

[+person, -number]

-o

[+person, +number]

-mos

[-person, -number]

-a

[-person, +number]

-m

Obviously, for deriving the European Portuguese pattern, a binary feature
for person will not do, since it cannot account for the three-way distinction
found in the singular. Independently of the details of Galves’ proposal, the
crucial aspect to be kept in mind for this paper is that number is a distinctive
feature in the verbal morphology of BP1.

The facts

169

In BP2, the verbal paradigm is much more simplified. Taking the same case
presented above for the other two variants, it is possible to observe, that there
is just a distinction between first person and the rest:

(14)

Verb cantar ‘to sing’ – present tense

singular

plural

canto

canta

II canta

canta

III canta

canta

For the purposes of this appendix, the relevant aspect is that the difference
between BP1 and BP2 may be linked to the role played by number. This is a
relevant feature for distinguishing verbal forms in BP1, but not in BP2. In
the latter, only person plays a role.

Bearing this description in mind, let us now look at the subject-verb agree-

ment patterns in the two dialects of Brazilian Portuguese.

In BP1, there is number agreement between the subject and the verb:

(15)

a. Os menino comeram o doce.

The-pl child-sg ate-pl the candy
‘The children ate the candy’

b. *Os menino comeu o doce.

The-pl child-sg ate-sg the candy
‘The children ate the candy’

c. Os menino tossiram.

The-pl child-sg coughed-pl
‘The children coughed’

d. *Os menino tossiu.

The-pl child-sg coughed-sg
‘The children coughed’

Unlike what we did above for European Portuguese, for the sentences in
(15), it is not possible to test whether the position of the subject is relevant
for the pattern of agreement, since inversion is impossible with transitive and
inergative verbs. However, if we test unaccusative verbs, it is possible to
detect an interesting difference with respect to European Portuguese. Inverted

170

Appendix: On the nature of agreement in European Portuguese

subjects do not agree. Agreement with an inverted subject is felt by speakers
as posh, as if one is trying to mimic European Portuguese:

(16)

a. ??Queimaram muitas floresta.

Burnt-pl many-pl forest-sg
‘Many forests burnt’

b. Queimou muitas floresta.

Burnt-sg many-pl forest-sg
‘Many forests burnt’

Another interesting difference with respect to European Portuguese comes
from predicative and passive constructions: as shown in (17), unlike in EP,
there is subject-verb agreement, but the adjectival or participial form does
not display number agreement with the subject:

(17)

a. ??As casa parecem bonitas.

The-pl house-sg seem-pl pretty-pl
‘The houses seem pretty’

b. As casa parecem bonita.

The-pl house-sg seem-pl pretty-sg
‘The houses seem pretty’

c. ??As casa foram destruídas.

The-pl house-sg were-pl destroyed-pl
‘The houses were destroyed’

d. As casa foram destruída.

The-pl house-sg were-pl destroyed-sg
‘The houses were destroyed’

BP2 significantly differs from BP1. There is no subject-verb number agree-
ment in any context, neither with transitive and inergative verbs in SV order
,nor with unaccusative verbs in any order:

(18)

a. *Os menino comeram o doce.

The-pl child-sg ate-pl the candy
‘The children ate the candy’

b. Os menino comeu o doce.

The-pl child-sg ate-sg the candy
‘The children ate the candy’

The facts

171

c. *Os menino tossiram.

The-pl child-sg coughed-pl
‘The children coughed’

d. Os menino tossiu.

The-pl child-sg coughed-sg
‘The children coughed’

(19)

a. *Queimaram muitas floresta.

Burnt-pl many-pl forest-sg
‘Many forests burnt’

b. Queimou muitas floresta.

Burnt-sg many-pl forest-sg
‘Many forests burnt’

c. *Muitas floresta queimaram.

Many-pl forest-sg burnt-pl
‘Many forests burnt’

d. Muitas floresta queimou.

Many-pl forest-sg burnt-sg
‘Many forests burnt’

Predicative and passive constructions do not display any plural morphology
either. The plurality of the sentence is just marked on the subject’s deter-
miner.

(20)

a. *As casas parecem bonitas.

The-pl house-pl seem-pl pretty-pl
‘The houses seem pretty’

b. *As casas parecem bonita.

The-pl house-pl seem-pl pretty-sg
‘The houses seem pretty’

c. As casa parece bonita.

The-pl house-sg seems-sg pretty-sg
‘The houses seem pretty’

d. *As casas foram destruídas.

The-pl house-pl were-pl destroyed-pl
‘The houses were destroyed’

172

Appendix: On the nature of agreement in European Portuguese

e. *As casa foram destruída.

The-pl house-sg were-pl destroyed-sg
‘The houses were destroyed’

As casa foi destruída.
The-pl house-sg was-sg destroyed-sg
‘The houses were destroyed’

1.3. Summary

Before turning to the proposal made in Costa and Figueiredo Silva (2003a),
let me sum up the descriptive conclusions obtained, and the issues to be
addressed:

In European Portuguese, there is full DP-internal agreement;

ii)

In the two dialects of Brazilian Portuguese, number within the DP is
marked either on the D head or in all prenominal elements;

iii) Prenominal adjectives and possessives differ in Brazilian Portuguese,

in the sense that only the latter may carry number marking in the
absence of plurality on the definite article;

iv) In European Portuguese, there is full subject-verb agreement, independ-

ently of the position of the subject, except in the case of unaccusative
verbs in which agreement is optional with inverted subjects in colloquial
speech;

In European Portuguese, there is full number agreement with passive
participles and small clause predicates;

vi) In BP1, there is full subject-verb agreement, except for the case of in-

verted subjects;

vii) In BP1, there is no number agreement with passive participles and

small clause predicates;

viii) In BP2, there is no number subject-verb agreement;

ix) In BP2, there is no number agreement with passive participles and

small clause predicates.

The facts

173

These descriptive statements raise at least the following questions:

a) What is the difference between the plural marker in European and

Brazilian Portuguese, allowing it to spread over all elements able to bear
it only in the former?

b) Why is it possible to find multiple agreeing elements in BP in the pre-

nominal domain only?

c) What is conditioning the difference of behavior between prenominal

possessives and adjectives in BP?

d) What is the difference between subject-verb agreement and DP-internal

agreement, differentiating the two dialects of BP?

e) What is the difference between subject-verb agreement and agreement

with passive participles and small clause predicates, differentiating BP1
from EP?

f) Why does inversion have an effect on subject-verb agreement in BP1,

but not in EP?

2. Proposal

The hypothesis put forward in Costa and Figueiredo Silva (2003a) is that
the type of agreement morphology variation found is not a consequence of
different landing sites for verb movement. In other words, given the evidence
regarding verb movement, it is assumed that the number agreement facts are
independent from the locus of verb movement. In fact, taking into account
the traditional tests for tracing V-to-I movement, there are no significant
differences between European and Brazilian Portuguese, that might be pro-
vide evidence for postulating different landing sites for the verb. As shown
in Costa and Galves (2002), in both languages, the verb appears to move
from V-to-T, not targeting the highest functional head of the IP-domain.
This derives the fact that in both languages the verb may surface in between
adverbs, and precede or follow floating quantifiers:

(21)

Ninguém provavelmente leu bem o poema.

(EP/BP)

Nobody probably read well the poem
‘Nobody probably read well the poem’

174

Appendix: On the nature of agreement in European Portuguese

(22)

a. Os menino(s) todos beijaram a Maria.

(EP/BP)

The-pl child(ren) all kissed-pl Maria
‘The children all kissed Maria’

b. Os menino(s) beijaram todos a Maria.

The-pl child(ren) kissed-pl all Maria
‘The children all kissed Maria’

Contenders of the hypothesis that Brazilian Portuguese has less verb move-
ment than European Portuguese often base their argumentation on the mor-
phological facts (cf. e.g. Ambar, Gonzaga and Negrão 2002). The data in
(21) and (22) suggest that there is no positional evidence for this claim

The second assumption needed is to assume, following Halle and Marantz

(1993), Bobaljik (1995), and Embick and Noyer (2001), that morphemes
may attach to heads independently of movement. In other words, contra
Belletti (1990), among others, we contend that a specific inflectional mor-
pheme may surface on the verb as a consequence of syntactic head-move-
ment, or as the result of a morphological process. We will further follow the
tenets of Distributed Morphology in assuming that a morpheme may be real-
ized in two ways: as a singleton or as a dissociated morpheme. According to
Embick (1997) and Embick and Noyer (2001), a dissociated morpheme does
not figure in syntax proper. It is inserted after Spell-Out, only indirectly
reflecting syntactic structures. An application of this idea is proposed in
Embick and Noyer (2001) in order to explain the distribution of definiteness
markers in the Swedish DP. These markers attach to the nominal root, when
there is no other material, but, if there is a prenominal adjective, they also
obligatorily surface on the determiner:

(23)

a. Mus-en

b. Den gamla mus-en

mouse-df

the old mouse-df

‘the mouse’

‘the old mouse’

Embick and Noyer (2001) explain this behavior by postulating the following
PF-requirements on the realization of N and D[+df]:

(24)

a. N must be marked for definiteness when D is [+def].

b. D[+def] must have a host.

The compliance with these constraints may be established in the syntax. N-
to-D movement satisfies both and derives the pattern in (23a). However, if

Proposal

175

an adjective is present, blocking N-raising for some reason, a determiner
must be inserted in D for satisfying (24b). However, this is not sufficient for
satisfying (24a). Since [+def] is a dissociated morpheme in Swedish, it can be
inserted post-syntactically, attaching to the nominal head, and the constraint
in (24a) is satisfied. This analysis of the Swedish facts enables envisaging the
agreement between the determiner and the noun as not reflecting a specific
syntactic configuration.

Adopting for EP the assumption that the proliferation of a given mor-

pheme may be interpreted as a consequence of it being dissociated, we make
the following hypothesis:

(25)

Type of [plural] morphology in Portuguese:

a. [plural] is a singleton in Brazilian Portuguese.

b. [plural] is a dissociate morpheme in European Portuguese.

A corollary of the hypothesis that [plural] may be a dissociated morpheme
able to attach to roots independently of their position is that, for agreement
to obtain, a Spec-head configuration may be required, although that is not
obligatory. This observation conforms to Chomsky’s (2001) proposal that
agreement may be triggered under different types of operation, and is empiri-
cally founded on the observation that identical spec-head configurations do
not trigger overt agreement in all languages. A clear example is past-partici-
ple agreement with displaced objects in French, and the lack of it in other
Romance languages (Kayne 1989):

(26)

a. Je les avait faites. (French)

I them-acus-fem-pl had done-fem-pl
‘I have done them’

b. Eu tinha-as feito. (European Portuguese)

I had them-acus-fem-pl done-masc-sg
‘I have done them’

According to these assumptions, the observed crosslinguistic variation fol-
lows straightforwardly. The difference between European and Brazilian
Portuguese derives from the type of morpheme associated with plurality.
We contend that [plural] is a dissociated morpheme only in EP. This pro-
posal derives the fact that all elements able to bear (subject-agreeing) plural
morphology will actually carry such markers, independently of the existence

176

Appendix: On the nature of agreement in European Portuguese

of movement and spec-head configurations. In BP, on the other hand, [plural]
is not a dissociated morpheme. Therefore, it will attach to the element anchor-
ing the information concerning number. Following Enç (1991), D is the head
linking the DP to its LF-interpretation. Accordingly, the plural morpheme is
realized on this head, which is able to carry this marker. Since [plural] is not
a dissociated morpheme in BP, it will not surface in other categories. This
derives the behavior displayed in BP2: there is a single marking of plural.

As noticed above, under this type of approach, overt reflexes of Spec-

head configurations are possible, although not obligatory. This allows for
deriving the difference between BP1 and BP2: in the former, but not in the
latter, a Spec-head configuration between the subject and an inflectional
head yields visible agreement, just like in the case of past participle agree-
ment in French. The fact that the same effect does not obtain in BP2 is not
surprising, given the independent evidence for parametrizing the overt effects
of spec-head relations.

In short, the account for the variation in the morphological expression of

plurality may be summarized in the following schema:

(27)

dissociate morpheme?

Yes (EP)

No (BP)

Does Spec-head in I trigger overt agreement?

Yes (BP1, EP, French)

No (BP2)

Does Spec-head in AgrO trigger overt agreement?

Yes (French)

No (Portuguese)

With these ingredients, it is now possible to readdress the list of questions
listed at the end of the previous section:

a) What is the difference between the plural marker in European and

Brazilian Portuguese, allowing it to spread over all elements able to bear
it only in the former?

Since [plural] is a singleton morpheme in Brazilian Portuguese, it will only
surface in a single head. On the other hand, the fact that it is a dissociated
morpheme in European Portuguese enables its postsyntactic association
with all elements able to bear plural marking.

b) Why is it possible to find multiple agreeing elements in BP in the pre-

nominal domain only?

Proposal

177

The answer to this question does not entirely follow from the proposed analy-
sis, since it involves doing some additional research on the internal structure
of the DP. In any case, it is interesting to note that there is a correlation be-
tween the agreement facts in BP and the well-known, though poorly under-
stood, asymmetry between prenominal and postnominal material as far as
phrase structure is concerned. As discussed in DiSciullo and Williams
(1987), among many others, prenominal material displays some properties
typical of heads, resisting modification and not taking complements, while
postnominal elements display properties typical of phrases. Independently
of the precise implementation of these ideas, if morphology regards all
prenominal material as a complex array of heads, it is expected that the only
elements marked with plurality are the prenominal ones. By hypothesis, the
plural morpheme could then be spreading from D onto the other units of the
complex head.

c) What is conditioning the difference of behavior between possessives and

adjectives in BP?

As shown above, if only one of the prenominal elements is marked with plu-
rality, it must be the determiner in D-Adjective-N sequences, which follows
from the hypothesis that [plural] attaches to the D head, which serves as an
anchor to LF. However, if the sequence is D-Possessive-N, the element
bearing the plural morpheme is the determiner, as repeated in (28):

(28)

A minhas casa
the-sg my-pl house
‘My house’

This difference between possessives and adjectives follows from the assump-
tions concerning the locus of attachment of the plural morpheme. As just
mentioned, it must be attached to the head responsible for establishing the
link with semantic interpretation. As extensively argued in Castro (2001),
two facts must be taken into account in order to understand the behavior of
possessives in Portuguese: i) if they occur prenominally, the DP is definite, if
they occur postnominally, the DP is indefinite (cf. 29); ii) the definite article
in examples like (28) is expletive. Castro (2001) shows that the latter fact
becomes very obvious, when it is taken into account that Brazilian dialects
omitting the expletive determiner before proper names, also do so before
prenominal possessives, as shown in (30) and (31):

178

Appendix: On the nature of agreement in European Portuguese

(29)

a. O meu livro

the my book
‘My book’

b. *O livro meu

the book my
‘My book’

c. Um livro meu

a book my
‘A book of mine’

d. *Um meu livro

a my book
‘A book of mine’

(30)

European Portuguese:

a. *(O) João

(the) João
‘John’

b. *(o) meu livro é azul.

(the) my book is blue
‘My book is blue’

(31)

Dialectal Brazilian Portuguese:

a. (O) João

(the) João
‘John’

b. (O) meu livro é azul.

(the) my book is blue
‘My book is blue’

Since the definite article in definite possessive constructions is not the marker
of definiteness, and is just an expletive, and since it is assumed that the [plural]
morpheme anchors on the category codifying the information that is relevant
for the interface with the interpretational component, it is expected that the
number morphology surfaces on the possessive rather than on the expletive.

d) What is the difference between subject-verb agreement and DP-internal

agreement, differentiating the two dialects of BP?

Proposal

179

As spelled out above, DP-internal agreement and the visibility of subject-verb
agreement stem from two different conditions: no dialect of BP is expected to
exhibit D-N agreement, since the number morpheme is a singleton, and there
is no Spec,head relation between D and N. On the other hand, the parametric
specification schematized in (27) stipulates that the Spec,head relation be-
tween the subject and I triggers visible effects in BP1 only. These two sources
for the visibility of agreement combined derive the fact that the two dialects
of BP behave alike in the DP-internal domain, but differ in the subject-verb
agreement patterns.

e) What is the difference between subject-verb agreement and agreement

with passive participles and small clause predicates, differentiating BP1
from EP?

It was observed that, while in EP, number agreement surfaces everywhere,
including in passive participles and small clause predicates, in BP1 there is
only subject-verb agreement, but not agreement with the passive participles
and small clause predicates. This difference in behavior straightforwardly
follows from our analysis: since subject-verb agreement in BP1, unlike in EP,
is a consequence of Spec,head agreement, we expect to find it only when-
ever the subjects stands in a Spec,head configuration with the agreeing head.
Since within small clauses, there is no Spec,head relation between the subject
and the predicate, and there is a head-complement relation between passive
participles and the subject, the configurations needed for agreement to arise
do not exist in these two constructions. In EP, on the contrary, since agree-
ment is established via the postsyntactic insertion of the dissociated mor-
pheme, number agreement arises independently of the type of configura-
tional relation between the subject and the agreeing head.

f) Why does inversion have an effect on subject-verb agreement in BP1, but

not in EP?

The answer to this last question is related to the previous one. Assuming
with Belletti (1988) that inverted arguments of unaccusative verbs stay in
their base-generated position, there is no Spec,head relation between the
inflectional head and the subject, hence no visible agreement arises. In EP,
on the contrary, since number agreement is a dissociated morpheme, inde-
pendently of there being a Spec,head configuration, the verb and the subject
may agree.

180

Appendix: On the nature of agreement in European Portuguese

At this point, it is important to understand what goes on in colloquial Euro-
pean Portuguese, in which, as mentioned above, the inverted subject of an
unaccusative may not agree with the verb:

(32)

Ardeu muitas florestas.
Burnt-sg many-pl forests
‘Many forests burnt’

Costa et al. (2002) argue that (32) is an instance of locative inversion. In that
case, (32) exhibits Spec,head agreement with an expletive subject. In support
of their claim, they show that lack of agreement shows up in cases of identi-
ficational focus, but not in cases of information focus:

(33)

A: Quem é que chegou? (EP)

who is that arrived?
‘Who is arrived?’

B: a. ??Chegou os alunos.

Arrived-sg the students
‘The students arrived’

b. Chegaram os alunos.

Arrived-pl the students
‘The students arrived’

Looking at non-null-subject languages, like English, it is possible to know
that expletive constructions and locative inversions are not used in informa-
tion focus contexts:

(34)

A: Who is coming?

B: a. John is coming.

b. #There comes John.

The interpretation Costa et al. (2002) make of these facts is the following: in
information focus contexts, there is no expletive involved, which derives the
obligatory agreement with the argument. In locative inversion contexts, there
are two candidates for agreement, the expletive and the argument, and varia-
tion is found.

Proposal

181

3. Further predictions

Besides deriving the agreement facts presented in section 1, the present
analysis makes the following interesting predictions with consequences for
the treatment of other phenomena that we would like to emphasize:

a) In EP, independently of the word order found, there is always full number

agreement (Costa 1998), since agreement is not dependent on a specific
syntactic configuration. This prediction has consequences for analyses of
inversion. If this analysis is on the right track, the visibility of agreement
must not be taken as a sign that at some point of the derivation the inverted
subject stood in a Spec,head relation with the subject in the inflectional
domain.

b) Returning to the behavior of possessives in Brazilian Portuguese, an addi-

tional fact is correctly predicted by this analysis. Prenominal possessives
display number agreement in BP1, while postnominal possessives do not,
as shown in (35):

(35)

a. o meus livro

b. uns livro meu.

the-sg my-pl book-sg

some-pl book-sg my-sg

‘my books’

‘some books of mine’

Assuming with Schoorlemmer (1998), Castro (2001), among others, that
only the prenominal possessive is a head related to D, while the postnominal
possessive is an XP, it is expected that the singleton [plural] morpheme only
attaches to the former.

c) Scherre (1994) shows that in the variation found for the agreement pat-

terns in the Brazilian dialects there are very many variables that do not fit
well within a purely syntactic approach for deriving agreement patterns.
Such variables include aspects such as phonological salience of the num-
ber marker and linear position. It is interesting to note that locating the
analysis of agreement patterns at the interface between syntax and mor-
phology, at the PF wing of the grammar, opens up a way for accommo-
dating variables that are morphophonological in nature.

d) Finally, this analysis makes an interesting prediction concerning the

variation found within Romance languages as to what the target of V-to-I

182

Appendix: On the nature of agreement in European Portuguese

movement is and its correlation with differences in number agreement
inflection. One of the departure points leading this proposal was the
observation that the differences between the three variants of Portuguese
found in the verbal agreement pattern do not correlate with different
landing sites for the verb. It was further shown that the differences
between the three dialects were linked to the marking of number, since
they all encode person differences, even if in different degrees. Let us
consider Vikner’s (1997) generalization concerning the morphological
evidence for there being V-to-I movement:

(36)

Vikner’s generalization:

A language has V-to-I movement if there is Inflection for Person in
all tenses.

Since all three variants of Portuguese encode inflection for Person, Vikner’s
generalization and this analysis correctly predict that there is no difference
as far as the target of verb movement is concerned.

It is further predicted that a given language may move the verb higher

than the other and yet exhibit a weaker number morphology. Assuming with
Costa (1996) and Costa and Galves (2001) that the targets of V-to-I move-
ment are AgrS in French and T in European Portuguese, the prediction is
correctly borne out. French is a language moving the verb higher than
European Portuguese, and exhibiting less number agreement.

4. Conclusion

The proposed analysis not only accounts for the variation described between
the three varieties of Portuguese, but it also provides clear evidence for an
autonomous morphological component, deriving morphological aspects of
language in a way that is partly independent from syntax. Interestingly, it
was crucial for these conclusions to be reached to relate the DP-internal
agreement facts with subject-verb agreement.

Assuming these results for number agreement, it may be hypothesized

that the relevant part of inflection for the discussion carried out in chapter 6
concerns person inflection.

Conclusion

183

Notes

I am not adopting the radical position that all adverbs are adjoinable to different
projections. Rather, I will follow a mixed approach, accepting that some adverbs
have fixed positions and others may be adjoined to different projections. See
Costa (1998, 2002) for a discussion of the relevant data.

In Costa and Galves (2002), it is argued that the same holds for Brazilian Portu-
guese.

See Costa (2000a) and Costa and Duarte (2002) for a comparison between the
two analyses.

The non-adjacency between the subject and the verb is explained in terms of
partial-V-movement. See chapter 1.

Duarte (1996) mentions that those speakers who accept (9b) only accept it if the
preposed constituent is not doubled by a clitic, which favors an analysis of this
construction in terms of topicalization, rather than clitic-left-dislocation.

Note that this case is different from the type of example discussed by Barbosa
(2000), presented in (i):
(i) a. Cresceu uma flor em todos os vasos.

grew a flow in every pot

b. ???Uma flor cresceu em todos os vasos.

a flower grew in every pot

Barbosa (2000) presents this contrast as an argument in favor of her hypothesis,
since there is no a priori reason to consider that Spec,IP is incompatible with
wide scope reading of the subject. However, the contrast in (i) just leads to the
conclusion that, whenever there are two available positions for the subject, they
can be used to codify different readings (see also Adger 1994; Costa 1998; among
others), independently of the A- vs A-bar status of the positions.

As pointed out to me by P. Barbosa, for some reason, the verb telefonar ‘to
call’ behaves differently:
(i) O que é que aconteceu?

a. O João telefonou.
b. Telefonou o João.

This is the pattern found for unaccusative verbs (Costa 2001). Assuming with
Pinto (1997) that the verb telefonar behaves like unaccusatives in being able to
select a temporal operator, enabling locative inversion, this behavior follows
(see Costa and Figueiredo Silva 2003 for further arguments for this hypothesis).

In Costa (1997, 1998), a discussion of differences between definite and indefinite
preverbal subjects is presented. Since the data turn out to be not very clear, I will
not reproduce the discussion here.

See Lobo (2003) for an argument for not establishing a full parallel between
these I-to-C contexts and conditional clauses headed by a complementizer.

10. Note that the ungrammaticality of (17b) is problematic not only for Barbosa’s

analysis, but also to the claim put forward in this chapter that the two analyses
are not incompatible.

11. As noted by a reviewer of Costa (2003), the existence of mixed systems is not

enough to rule out the possibility that Barbosa’s analysis holds for languages in
which it is not possible to distinguish between different types of pro. The point
I am trying to make is that her analysis cannot be generalized to all types of
null subject languages.

12. For a difference between the judgements by Ribeiro (2002) and corpus data,

see Lobo (2003).

13. I am ignoring here the exact label of I. This may cause some confusion, when

this discussion is confronted with the assumptions of chapter 1, since I argued
for different landing sites for verbs in English, French and Portuguese. I decided
to do so to make the discussions clearer. For clarity: I is the functional category
where subjects land, that is, AgrSP in chapter 1. For the present discussion, the
crucial fact is whether we need or do not need a landing site for the verb above
the category where subjects move for case reasons.

14. VS order in embedded context is possible with all verbs and all types of main

verbs, which makes it different from the contextually dependent embedded V2
phenomena in languages such as Icelandic and Yiddish (see McCloskey 1992
for contexts of CP-recursion and Rizzi 1997 for a specific proposal on split-CP).

15. Actually (18a) is grammatical, but the adverb gets a different (aspectual) mean-

ing. For controlling for that factor, one may have another monosyllabic adverb,
in the ruled in position, with a contradictory meaning and the sentence would
remain grammatical:
(i) Bem comeu mal o Paulo maçãs.

16. I am not going to speculate here on the mechanism which renders binding possi-

ble in this context. I would like to note anyway that it is very likely that the
object QP is bound not by the right-dislocated subject but by the pro in subject
position.

17. Speakers of Dutch disagree with respect to the possibility of scrambling PPs.

There seems to be some variability with respect to the adverbs that intervene
between the PP and the verb. I keep these examples in the text, since they are
fully acceptable for some speakers, but I am aware that a more precise collec-
tion of data is in order. Thanks to Martin Honcoop, Ruben van de Vijver and
Marga Petter for pointing this problem out to me.

186

Notes

18. Monosyllabic adverbs seem to be a reliable test for the distribution of arguments

in Portuguese and in English, but not in Dutch. For this reason, I will contrast
the sentences in Portuguese with Dutch sentences involving other types of ad-
verbs. I have no explanation for why monosyllabic adverbs differ in behavior
cross-linguistically.

19. See also Costa and Duarte (2003) for an alternative proposal regarding the vari-

able status of the null object.

20. (55b) is possible if the adverb is modifying the adjective simpática ‘friendly’.
21. Deprez (1989) presents some data suggesting that scrambled DPs in German

license parasitic gaps. See Doetjes (1997) and Costa (1998) for discussion of
the status of FQs associated with objects.

22. Note the ungrammaticality of (i):

(i) *O Pedro leu os livros bem todos.

Pedro read the books well all

23. See example (71).
24. This argument only goes through under the assumption that remnant movement

is not a root phenomenon, which must be the authors’ assumption since VOS
interrogatives are possible in embedded contexts:
(i) Perguntei a quem deu o livro o Pedro.
(I) asked to whom gave the book Pedro

25. This argument only holds if the distribution of clitics is indeed syntactically

triggered. Barbosa (2000) has suggested that enclisis is triggered any time a clitic
is initial in a prosodic constituent, which would be the case in the sentences in
(99). If the prosodic analysis proves true, this argument against the remnant
movement analysis does not hold.

26. As mentioned above, I will call the languages in which the shift Adv-comple-

ment affects only NPs object-shift languages. These include Danish, Norwegian
and Swedish, which move only pronouns and Icelandic and Faroese, which move
NPs. When unnecessary, I will not distinguish the pronoun-shifting languages
from the NP-shifting languages.

27. I am using the alternation specific/non-specific here for ease of explanation. It

may be seen below that I will not consider specificity the trigger for scrambling,
but for now it is enough to note that similar effects arise with NPs and adverbs.

28. A further problem for Diesing’s approach, pointed out to me by Sjef Barbiers,

is that in a sentence like (i), the object must be analyzed as being outside VP,
and the subject inside VP, which is obviously not feasible:
(i) Er heeft iemand het pakje op tafel laten ligeen.

there has someone the pack on the table let lay.

29. See Ladd (1996) for the several types of marked stress.
30. As extensively discussed in Reinhart (1995), this analysis captures the so-called

specificity effects.

Notes

187

31. Note, however, that the combination of the two arguments leads to a surprising

contradiction, since Hungarian focus movement does not yield WCO-effects,
as noted by Brody (1995).

32. The relational nature of focus prevents an analysis in which the constituent

containing the focused part is pied-piped. In most analyses involving focus
movement (e.g. Ambar 1998), focused constituents move in order to form a
domain separate from the part of the sentence that yields given information. If
the two types of information occur in the same position because of pied-piping,
this type of relation is lost.

33. Such a hypothesis is entertained in Chomsky (1995). It has been pointed out to

me that the lack of multiple foci overtly moved in Hungarian may follow from
a feature-checking approach. If there is only one strong focus feature in the F
position in Hungarian, the other focused elements may stay in situ, on a pair
with multiple questions in English. I am not sure whether this analysis may
work, since as we will see later, Hungarian focus entails uniqueness and is not
recursive in the same way English cleft constructions are not recursive:
(i) *It was in front of the cinema that it was John that Eve waited for.

If, as it will be discussed below and is argued in É. Kiss (1996), clefts are indeed
the counterpart of Hungarian focus movement, it has to be explained why multi-
ple-clefting and multiple focused elements are impossible, if multiple in-situ
foci are possible. Furthermore, note that in Hungarian contrastive foci in situ are
ruled out, even if there is a focused constituent moved (K. Polgardi (personal
communication)). Thus, the parallel with wh-movement does not hold since: (i)
there does not seem to be a language with multiple focus movement or with co-
occuring moved contrastive focus and in-situ contrastive focus.

34. A related functional head has been proposed for Portuguese by Martins (1994),

Uriagereka (1995), Raposo (2000), among others.

35. Note that it is not my goal here to present a solution to Rooth’s problems. The

important point of Rooth’s theory which I will follow here is to eliminate Focus-
movement from the set of LF-operations. That is, when foci move, they move in-
dependently of their being foci: the creation of bound variable readings forces
them to be QR-ed. One may wonder what is the advantage of trading focus
movement for QR. The advantage is: this QR is independently necessary, and
applies to Quantifiers, not to foci without quantificational force.

One of my goals in this study is to try derive as many word orders as possible

without resorting to covert operations. The best evidence for covert operations
comes from the behavior of quantifiers (May 1985), hence, it should be possible
to reduce the evidence of some of the apparent focus movement constructions
to independently necessary QR operations.

36. It is important to note that this does not mean that topic is the complement of

focus. I will assume with Jäger (1996) and Büring (1995) among others a tri-

188

Notes

partite structure for information: topic, focus and background. These analyses
make the right prediction that a sentence may have only a focus, or no topic. See
Büring and references therein for discussion.

37. This implies that marked stress and rightmost placement are not competing

strategies for marking focus. Instead, stress appears as a last-resort mechanism.
See Costa and Figueiredo Silva (2003) for further arguments.

38. See arguments in favor of interpreting focus also as a prosodic phenomenon in

EP in Frota (1992).

39. Keep in mind that this type of prominence and the one described in Frota’s work

are distinct. In this section, no prosodic analysis of the prominence will be
made. I will just propose an algorithm enabling the hearer to identify the locus
of the main prominence. It is by no means my goal here to make any specific
claims about the nature of the stress assigned to focused constituents.

40. Her conclusions and my discussion are independent of the actual structure of VP,

since she claims that stress on any constituent of VP is sufficient for enabling
projection.

41. Cinque’s algorithm is defined in terms of embedding, while here I use linear

order. I will leave aside the discussion between embedding and linearity for the
moment, returning to it when it will be relevant. A reviewer to Costa (2002a)
points out to me that the algorithm in (49) does not apply to these cases, and that
it should be defined in terms of rightmost word rather than constituent. I will
stick to the definition in terms of constituent following the tradition in phrasal
phonology and in prosody, since the predictions are exactly the same. The same
reviewer argues that marked and unmarked stress assignment follow from dif-
ferent mapping conditions. This is not true, as discussed in this paragraph: in
case of VSO with main stress on the subject, the focus set is the subject and
everything it c-commands (the object), in case of SVO with main stress on the
object, the focus set is this element (which is ambiguously the rightmost part of
the NP, VP and IP as a consequence of iambic stress) and everything it c-com-
mands (nothing). The crucial difference between the two cases is whether or
not the constituent bearing main stress c-commands another constituent. When
that is not the case, the effects of focus-projection arise. Crucially, the algorithm
in (38) predicts both cases.

42. The same reasoning applies to exclude a sentence-focus interpretation for a VSO

sentence with unmarked stress. See the next two chapters for a formalization of
this analysis in terms of markedness, and for an explanation of cross-linguistic
variation in sentence-focus context.

43. It has been pointed out to me that subject movement in this case corresponds to

a case of defocusing yielding an unmarked word order, which is not common to
defocusing operations. Note however, that defocusing in these cases also has
the function of topic-promotion which very often is associated with subjects

Notes

189

being in their canonical Spec,IP position, given the natural tendency for subjects
to be the topic of the sentence (cf Lambrecht 1994, Li 1976). Under the assump-
tion that markedness arises as a consequence of a natural tendency not being
met, subject topic-promotion should not yield a marked word order.

44. Szendrói (2001) convincingly argues that, even for Hungarian, it is not necessary

to assume a focus position in the left periphery. According to this author, the
so-called focus movement is an instance of prosodic movement in the sense of
Reinhart (1995).

45. More subtle differences concerning the discourse function of preposed ele-

ments are given in Ambar (1998), who also shows that preposed elements may
not be absolute new information, but argues for the existence of what she calls
topic/focus in which there is fronting and some function overlap. Although such
cases do not disconfirm the main observation that the function of focus-in-situ
is different from the function of preposing, I think it is important to refer to
them for completeness.

46. See also Quer (2002) for arguments from Catalan to distinguish focus-move-

ment from QP-fronting.

47. It must be known why SVO is the only felicitous order when the whole sentence

is focused. In Costa (1998, 2000c), this fact is analyzed within Optimality Theory
as a case of Emergence of the Unmarked. In this work, it must be stipulated that,
under equal circumstances, Move takes precedence over Agree. See Wurmbrand
(2001) for a discussion of the choice between Move and Agree.

48. These facts are crucial for deciding between this analysis and the proposal

made in Costa (1998), in which inverted subjects were violating their licensing
requirements, for satisfying a constraint on focus.

49. Incidentally, these data provide an argument in favor of Hornstein’s (1999,

2000) hypothesis that control may be considered movement. I will not take a
strong position regarding this issue, since the sentences in (182b,d) must involve
PRO.

50. For the status of these adverbs, see Castro and Costa (2003).
51. As mentioned above, inversion n Brazilian Portuguese is restricted to unac-

cusative contexts. Crucially, it is important to note for the purposes of this sec-
tion that inversion with copula is possible in Brazilian Portuguese.

52. The lack of minimality effects in embedded contexts, as shown in (i) might lead

to the conclusion that inversion in this type of context is an instance of move-
ment of the predicate to Spec,IP rather than some type of topicalization:
(i) a. Eu pergunto-me em que ano o rei serei eu.

I wonder in which year the king will-be-1sg I

b. *Eu pergunto-me que ano o rei verei eu.

I wonder in which year the king will-see-1sg I

190

Notes

However, Tavares (in preparation) shows that a comparison between left-dislo-
cated elements and inverted predicates reveals that they share some properties,
permitting a treatment of inverted copular constructions as left-dislocation.

53. The following problem arises: Brazilian Portuguese has raising verbs of the

seem-type, allowing raising of the subject to the matrix IP, which provides evi-
dence for the lack of CP. However, contrary to the prediction made here, if a
copular verb is embedded under seem, the agreement still goes with the matrix
verb:
(i) O assassino parece ser eu.

The murderer seems to be I

Although the problem is obvious, it does not necessarily constitute a problem for
the analysis, since this type of verbs allow super-raising in Brazilian Portuguese,
even when CP is projected. Therefore, the raising of the subject in non-finite
contexts is not compelling evidence for saying that CP is not projected in this
language.

54. For an analysis of agreement in Portuguese, see Costa and Figueiredo Silva

(2003a).

55. The data involving lack of agreement with postverbal subjects of unaccusative

verbs come from colloquial speech. These constructions, although produced
quite often by the speakers, are not accepted as grammatical in prescriptive
grammars.

56. As discussed in Costa and Figueiredo Silva (2003a), this assumption is only

valid in a language or dialect in which nominative subjects trigger agreement
independently of their placement in transitive and intransitive contexts. In
Brazilian Portuguese, this assumption does not hold.

57. (3b) is only legitimate with a topic intonation for the adverb.
58. In most examples involving I-to-C movement, I will use auxiliary verbs so that

I am able to control the position of the inverted subject. Leaving the participle
behind allows for making sure that the subject surfaces to its left, and not in
Spec,vP.

59. The special status of adverbs remains unaccounted for. It is not clear why

adverbs should not count as interveners. As Bobaljik (1995) discusses, this is a
more general issue, however, since in other domains, adverbs seem to behave
in the same way, in not disrupting adjacency relations. For a possible solution
to this problem, see Bobaljik (2002).

60. In Costa and Figueiredo Silva (2003a), it is argued that the number agreement

morphology in Portuguese has the status of dissociated morphemes, in the
sense of Embick and Noyer (2001). This assumption does not compromise the
results of this chapter, since the scope of the discussion is both person and
number agreement. Cf. the appendix to this chapter for discussion on the status
of agreement in Portuguese.

Notes

191

61. The morpheme -va- of the imperfect in forms like fala-va-mos ‘we talked’ may

be an aspectual morpheme rather than a tense morpheme, since the past of an
imperfect may be expressed with an auxiliary construction, in which tense is
expressed by the auxiliary verb. According to some authors, aspectual heads
surface below T, and cyclic head-movement predicts that the aspectual mor-
phology surfaces as an independent morpheme. It is likely that the same holds
for future, an issue to be further explored.

62. The only cases in which other orders are found in idioms involve Heavy com-

plements:

(i) Dar a César o que é de César.

Give to Caesar what belongs to Caesar

63. Since this is not a restructuring context, isomorphism with a matrix verb will

not play any role.

64. The only cases in which there is apparent modification of a prenominal posses-

sive can be argued not to instantiate of modification of an XP by another XP, but
rather head adjunction (Castro and Costa 2003).

65. A question to be addressed is what the distinction is between Xº weak forms

and Xº clitics. Note that the same question arises for the distinction between
strong XPs and weak XPs. As already noticed by Cardinaletti and Starke
(1999), the distinctions do not follow from categorial status alone. In the case
at stake, we can hypothesize that the difference between clitics and weak Xº is
prosodic.

66. Pereira (1993), cited by Scherre (1994), reports the following rates of agreement

for one single literate speaker:

talking to his boss

talking to his family

talking to his employees

verbal agreement

98%

91%

24%

DP-internal agreement

90%

52%

42%

This case shows that the same speaker masters the three dialects under discussion
in this paper, using each one of them in different conversational situations, con-
firming our hypothesis that BP1 and BP2 are to be treated separately.

67. Costa and Pereira (2003) present evidence from agreement with the pronominal

form a gente (lit.the people), which is grammatically specified for 3

person

singular, and referentially specified as a 1

person plural, showing that the num-

ber marker that is at stake is semantic number marking.

68. Even in BP2, a dialect using tudo, a non-inflected form of the quantifier all, we

see the same possibilities of word order:

(i) Os menino tudo beijou a Maria

The-pl boy-sg all kissed-sg the Mary
‘The boys all kissed Mary’

192

Notes

(ii) Os menino beijou tudo a Maria

The-pl boy-sg kissed-sg all the Mary
‘The boys all kissed Mary’

69. Matters may turn out to be more complicated. Based on facts like (i), Tavares

(in preparation) claims that in the presence of two candidates for agreement, the
one entering in a Spec,head relation with I categorically agrees:

(i) Neste jogo, tu {és/*sou} eu e eu {*és/sou} tu.

In this game, you {are/am} I and I {are/am} you
‘In this game, you are me and I am you’

At this stage, it is however difficult to incorporate these results in the discussion
of the patterns of agreement in unaccusative contexts, since in the latter, there is
variation only in colloquial speech.

Notes

193

References

Adger, David

1994

Functional Heads and Interpretation. Doctoral dissertation, University
of Edinburgh.

Adragão, Maria do Mar

2001

Aquisição da inversão numa criança entre os dois e os três anos. Ms,
Universidade Nova de Lisboa.

Adragão, Maria do Mar and João Costa

2003

On the status of preverbal subjects in null-subject languages: evidence
from acquisition. To appear in GALA 2003 Proceedings.

Alexiadou, Artemis and Elena Anagnostoupoulou

1995

SVO and EPP in Null Subject Languages and Germanic. In: FAS
Papers in Linguistics 4: 1–21, Potsdam.

1998

Parametrizing Agr: word order, verb- movement and EPP-checking.
Natural Language and Linguistic Theory 16.3: 491–539.

Ambar, Manuela

1992

Para uma Sintaxe da Inversão Sujeito-Verbo em Português. Doctoral
dissertation, University of Lisbon.

1996

Aspects of Focus in Portuguese. paper presented at the International
Workshop on Focus, Univ. Paris X.

1998

The syntax of focus in Portuguese: a unified approach. Ms, University
of Lisbon.

Ambar, Manuela and Jean-Yves Pollock

1998

Topique et Commentaire dans quelques constructions à inversion du
sujet en français et portugais. Ms, University of Lisbon & CNRS,
Lyon.

Anderson, Steven

1972

How to get even. In Language, 48, 893–905.

Baart, Joan

1987

Focus, Syntax and Accent Placement. Doctoral dissertation, Leiden
University.

Barbosa, Pilar

1995

Null Subjects. Doctoral dissertation, MIT.

2000

Clitics: a window into the null subject property. In: João Costa (ed.)
Portuguese Syntax. New comparative studies. Oxford University
Press, 31–92.

Bayer, Josef

1995

Directionality and Logica Form. On the scope of focussing particles
and wh-in-situ. Kluwer Academic Publisher.

Belletti, Adriana

1988

The Case of Unaccusatives. Linguistic Inquiry, 19: 1–34.

1990

Generalized Verb Movement: Aspects of Verb Syntax. Turin:
Rosenberg and Sellier.

Belletti, Adriana and Ur Shlonsky

1995

The order of verbal complements: A comparative study. Natural
Language and Linguistic Theory 13: 489–526, Kluwer Academic
Publishers.

Bennis, Hans and Teun Hoekstra

1984

Gaps and Parasitic Gaps. The Linguistic Review, 29–87.

Bianchi, Valentina & Silva, Maria Cristina Figueiredo

1994

On some properties of agreement object in Italian and Brazilian
Portuguese. In: Michael Mazzola (ed.) Issues and Theory in Romance
Linguistics. Selected papers from the Linguistic Symposium on
Romance Languages XXIII, 181–197. Washington, D.C., Georgetown
University Press.

Bobaljik, Jonathan

1995

Morphosyntax: the syntax of verbal inflection. Ph.D. dissertation,
MIT, Cambridge, Massachusetts.

2002

A-Chains at the PF-Interface: Copies and Covert Movement, in

Natural Language and Linguistic Theory 20: 197–267.

Bobaljik, J. D. and D. Jonas

1996

Subject Positions and the Roles of TP. In Linguistic Inquiry, 27:
195–236.

Bobaljik, Jonathan David and Höskuldur Thrainsson

1998

Two heads aren’t always better than one. Syntax 1.1: 37–71.

Bok-Benema, Reineke

1998

Remnant VP movement in Spanish. Paper presented at Going
Romance, Utrecht.

Boskovic, Zeliko

1997

The syntax of non-finite complementation. Cambridge, Mass.: MIT
Press.

Brandt, Patrick

1999

Scope, topichood, and double objects. In: Todirascu, Amalia (ed.)
ESSLI Student Session 1999, Utrecht.

Bresnan, Joan and Sam Mchombo

1987

Topic, pronoun and agreement in Chichewa. In: Language 63.4.
741–82.

196

References

Brody, Michael

1990

Remarks on the Order of Elements in the Hungarian Focus Field. In:
I. Kenesei (ed.) Approaches to Hungarian, vol. 3, Szeged: JATE.

1995

Focus and Checking Theory. In: Approaches to Hungarian. Vol. 5,
JATE.

Büring, Daniel

1997

The 59th Bridge Accent: on the meaning of topic and focus. Routledge
Studies in German Linguistics 3, Routledge.

Cardinaletti, Anna

1998

On the Deficient/Strong Opposition in Possessive Systems. In:
Possessors, Predicates, and Movement in the Determiner Phrase.
Artemis Alexiadou & Chris Wider (eds.), 17–53. Amsterdam/Phila-
delphia: John Benjamins Publishing Company.

Casielles, Eugenia

1996

Focus Preposing (it is called). Ms, UMass at Amherst.

Castro, Ana

2001

Os possessivos em português europeu e português brasileiro: unidade
e diversidade, Actas do XVI Encontro Nacional da Associação Portu-
guesa de Lingüística, Associação Portuguesa de Lingüística, Colibri,
Portugal, 599–613.

Castro, Ana and João Costa

2003

Weak forms as Xº: prenominal possessives and preverbal adverbs in
Portuguese, in: A. T. Pérez-Leroux & Y. Roberge (eds.), Romance
Linguistics: theory and acquisition, Amsterdam & Philadelphia, John
Benjamins, 95–110.

Chierchia, Gennaro

1991

Functional WH and weak crossover. In Proceedings of the 10

West

Coast Conference on Formal Linguistics, Dawn Bates (ed.), 75–90.
Stanford, CA: Center for the Study of Language and Information.

Chomsky, Noam

1976

Conditions on Rules of Grammar. Linguistic Analysis, 2: 303–351.

1993

A Minimalist Program for Linguistic Theory. In: Ken Hale and
Samuel J. Keyser (eds.) The View from Building 20. Cambridge,
Massachusetts, MIT Press.

1995

The Minimalist Program. MIT Press, Cambridge, Mass.

1998

Minimalist Inquiries. MIT Occasional papers in Linguistics, MIT.

2000

Derivation by phase. MIT Occasional Papers in Linguistics, MIT.

2001

Beyond explanatory adequacy. Ms, MIT.

Chomsky, Noam and Morris Halle

1968

The Sound Pattern of English. MIT Press, Cambridge, Mass.

References

197

Cinque, Guglielmo

1993

A Null Theory of Phrase and Compound Stress. In: Linguistic Inquiry,
24: 239–298.

Colaço, Madalena

1998

Concordância parcial em estruturas coordenadas em Português europeu.
In Actas do XIV Encontro Nacional da Associação Portuguesa de
Linguística, Aveiro.

Coelho, Izete, João Costa, Maria Cristina Figueiredo Silva and Sérgio Menuzzi

2001

Ordem VS e sujeito nulo em PE e PB. paper presented at 2º Colóquio
do Projecto PE/PB, Universidade do Ceará, Fortaleza.

Cornish, F.

2002

Locative inversion in 8 languages: syntax, semantics, discourse-prag-
matics, and functional position. Paper presented at 10

International

conference on functional grammar, University of Amsterdam.

Corver, Norbert and Henk van Riemsdijk

1994

Studies on Scrambling. Mouton de Gruyter.

Costa, João

1995

Adverbs as an Argument for OV in Dutch. Ms, HIL\Leiden University
[presented at OV\VO meeting, Univ. of Amsterdam].

1996

Adverb Positioning and V-movement in English: some more evidence.
Studia Linguistica 50: 1, 22–34.

1997

Positions for Subjects in European Portuguese. In: Proceedings of
WCCFL XV, CSLI, Stanford.

1998

Word order variation. A constraint-based approach. Doctoral disser-
tation, Leiden University (published by Holland Academic Graphics).

2000a

Spec,IP vs. Deslocado: prós e contras das duas análises dos sujeitos

pré-verbais. In D.E.L.T.A. 17.1: 283–303.

2000b

Focus in situ: evidence from Portuguese. In Probus, 12–2, 187–228.

2000c

The Emergence of unmarked word order. In: Géraldine Legendre,

Jane Grimshaw and Sten Vikner (eds.), Optimality Theoretic Syntax,
171–203. Cambridge: MIT Press.

2001

Postverbal subjects and agreement in unaccusative contexts in Euro-
pean Portuguese, in: The Linguistic Review 18.1, 1–17.

2002

A multifactorial approach to adverb placement: assumptions, facts and
problems. In: A. Alexiadou (ed.), Lingua: special issue on adverbs.

2002a

Multiple focus in European Portuguese: apparent optionality and

subject positions. In: C. Beyssade et al. Romance Languages and
Linguistic Theory 2000, Amsterdam: Johns Benjamins, 93–108.

2002b

VOS in Portuguese: Arguments against an analysis in terms of remnant

movement. In: A. Alexiadou et al. (eds.). Dimensions of Movement.
From features to remnants. Amsterdam: John Benjamins: 69–89.

198

References

2003

Null vs overt Spec,TP in European Portuguese. In: J. Quer et al.

(eds.), Romance Languages and Linguistic Theory 2001, Amsterdam:
John Benjamins, 31–47.

Costa, João, Izete Coelho, Maria Cristina Figueiredo Silva, and Fátima de Oliveira

2002

Considerações sobre a Ordem VS e Sujeito Nulo em Português

Europeu e Português Brasileiro. Paper presented at the PEPB Third
Colloquium, University of Lisbon, Lisbon.

Costa, João and Sandra Pereira

2003

Phases and Autonomous Features: a Case of Mixed Agreement in

European Portuguese. Paper presented at the MIT Workshop on EPP
and Phases, MIT, Boston.

Costa, João and Charlotte Galves

2002

External subjects in two varieties of Portuguese: evidence for a non-
unified analysis. In: C. Beyssade et al. Romance Languages and Lin-
guistic Theory 2000, Amsterdam: Johns Benjamins, 109–125.

Costa, João and Inês Duarte

2002

Preverbal subjects in null-subject languages are not necessarily left-
dislocated. In Journal of Portuguese Linguistics, 159–176.

2003

Objectos nulos em debate. In: I. Castro and I. Duarte (eds.). Razões e
emoção. Miscelânea de estudos em homenagem a Maria Helena Mira
Mateus. Lisboa: Imprensa Nacional – Casa da Moeda.

Costa, João and Maria Cristina Figueiredo Silva

2003

On the (in)dependence relations between syntax and pragmatics.

Paper presented at GLOW Workshop on Information Structure, Lund.

2003a

Nominal and verbal agreement in Portuguese: an argument for Dis-

tributed Morphology. Ms, Universidade Nova de Lisboa.

Costa, J. & A. Gonçalves

1999

Minimal Projections: Evidence from Portuguese. Catalan Working
Papers in Linguistics, 7. 59–69.

Déprez, Viviane

1989

On the Typology of Syntactic Positions and the Nature of Chains.
Doctoral dissertation, MIT.

Diesing, Molly

1992

Indefinites. Cambridge, Massachusetts, MIT Press

1995

Yiddish VP Order and the Typology of Object Movement in Germanic.
Ms, Cornell University.

Diesing, Molly and Eloise Jelinek

1995

Distributing Arguments. In: Natural Language Semantics 3.2, 123–176.

Dik, Simon C.

1978

Functional Grammar. Amsterdam: North-Holland.

References

199

DiSciullo, Anna-Maria and Edwin Williams

1987

On the definition of word, MIT Press, Cambridge, Mass.

Den Dikken, Marcel

1995

Extraposition as intraposition, and the syntax of English. Tag ques-
tions. Ms, HIL\Vrije Universiteit Amsterdam.

Doetjes, Jenny

1997

Quantifiers and Selection. Doctoral diss., HIL/Leiden University.

Duarte, Inês

1987

A Construção de Topicalização na Gramática do Português. Doctoral
Dissertation, University of Lisbon.

1996

A Topicalização no Português Europeu: uma análise comparativa. In:
I. Duarte and I. Leiria (eds.), Actas do Congresso Internacional sobre
o Português, APL/Colibri, Lisbon.

1997

Ordem de palavras: sintaxe e estrutura discursiva. In: Ana Maria Brito
et alii (eds.), Sentido que a vida faz – estudos para Óscar Lopes.
Campo das Letras.

Duarte, Inês and Gabriela Matos

2000

Romance clitics and the Minimalist Program. In: João Costa (ed.).

Portuguese Syntax. New Comparative Studies. Oxford University
Press, 116–142.

Duarte, Inês and Anabela Gonçalves

2000

Construções Causativas em Português Europeu e Português Brasileiro.

Paper presented at PEPB – 2000, Universidade de Coimbra.

Embick, David

1997

Voice and the interfaces of syntax, unpublished Ph.D. dissertation,
University of Pennsylvania, Philadelphia.

Embick, David and Rolf Noyer

2001

Movement Operations after Syntax. Linguistic Inquiry 32, 555–595.

Enç, Muvet

1991

The Semantics of Specificity. Linguistic Inquiry 22, 1–25. Figueiredo
Silva, Maria Cristina. 1996. A posição do sujeito em português
brasileiro – frases finitas e infinitivas, Editora da Unicamp, Campinas,
Brazil.

Frank, R., Y-S. Lee and O. Rambow

1992

Scrambling as non-operator movement and the special status of
subjects. In: S. Barbiers et al (eds.) Proceedings of LCJL 3, Leiden
University.

Frota, Sónia

1992

A Prosódia do Advérbio na Frase: interacção e convergência. In:
Actas do VIII Encontro da Associação Portuguesa de Linguística –
Lisboa 1992. APL, Colibri, Lisbon.

200

References

1994

Is Focus a Phonological Category in Portuguese. In: M. Schoorlemmer
and P. Ackema (eds.), Console I Proceedings, Holland Academic
Graphics.

1995

Focus, Phrasing, Stress and Accent in European Portuguese. Paper
presented at the International Conference on Interfaces in Linguistics,
Oporto.

1997

Focus in the Prosodic Interface: evidence from European Portuguese.
talk presented at the University of Lisbon.

1998

Prosody and Focus in European Portuguese. Doctoral dissertation,
University of Lisbon.

Frota, Sónia and Marina Vigário

2002

Efeitos de peso no Português Europeu». In: Mateus, M. Helena;

Correia, Clara N. (eds.), Saberes no Tempo. Homenagem a Maria
Henriqueta Costa Campos. Lisboa: Colibri, 315–333.

Galves, Charlotte

1993

O enfraquecimento da concordância no português brasileiro, in: Ian
Roberts and Mary Kato (eds.), Português Brasileiro: uma viagem
diacrônica, Editora da Unicamp, Campinas, Brazil.

Givón, Talmy

1984

Syntax: a functional-typological introduction. Vol 1. John Benjamins.

1990

Syntax: a functional-typological introduction. Vol 2. John Benjamins.

Gonçalves, Anabela

1999

Predicados verbais complexos em contexto de infinitivo não preposi-
cionado em português europeu. Doctoral dissertation, University of
Lisbon.

Gonzaga, Manuela

1997

Aspectos da sintaxe do advérbio em português. Mass. dissertation,
Universidade de Lisboa.

Halle, Morris and Alec Marantz

1993

Distributed Morphology and the Pieces of Inflection. In: K. Hale and
S. J. Keiser (eds.), The View from Building 20: Essays in Linguistics
in Honor of Sylvain Bromberger, MIT Press, Cambridge, Mass.,
111–176.

Holmberg, Anders

1986

Word Order and Syntactic Features. Doctoral dissertation, University
of Stockholm.

de Hoop, Helen

1992

Case Configuration and Noun Phrase Interpretation. Doctoral disser-
tation, Groningen.

de Hoop, Helen and Henriette de Swart

2000

Topic and Focus, L. Cheng en R. Sybesma (eds.), The First Glot Inter-
national State-of-the-Article Book, Mouton de Gruyter, Berlin/New
York, 105–130.

References

201

Hornstein, Norbert

1994

Logical Form. From GB to Minimalism . Blackwell, Oxford.

1999

Movement and Control, Linguistic Inquiry, 69–96.

Horvath, Julia

1986

FOCUS in the Theory of Grammar and the Syntax of Hungarian.
Foris, Dordrecht.

1995

Structural Focus, Structural Case, and the Notion of Feature Assign-
ment. In: Kiss (1995).

Huang, James

1982

Logical Relations in Chinese and the Theory of Grammar. Doctoral
dissertation, MIT.

Iatridou, Sabine

1990

About Agr(P). Linguistic Inquiry, 21: 551–577.

Jackendoff, Ray

1972

Semantic Interpretation in Generative Grammar. Cambridge, Mass.,
MIT Press.

1990

Semantic Structures, MIT Press, Cambridge, Mass.

Jäger, Gerard

1996

Topic, Scrambling and Aktionsart. In: Laura Brugé et alii (eds.),
Proceedings of ConSOLE 3. Leiden: Leiden University.

Johnson, Kyle

1991

Object Positions. In Natural Language and Linguistic Theory 9:
577–636.

Kayne, Richard

1984

Connectedness and binary branching. Dordrecht: Foris.

1989

Facets of Past Participle Agreement in Romance, in: Paola Benincà
(ed.), Dialectal Variation and the Theory of Grammar, Foris,
Dordrecht, 85–103.

1994

The Antisymmetry of Syntax. Cambridge, Massachusetts, MIT Press.

1998

Overt vs. Covert Movement. Ms, New York University

Kayne, Richard and Jean-Yves Pollock

1998

New Thoughts on Stylistic Inversion. Paper presented at the Workshop
on Inversion in Romance, University of Amsterdam.

Kiss, Katalin

1995

Discourse Configurational Languages. New York: Oxford University
Press.

1996

Focus Operator and Information Focus. Working Papers in the Theory
of Grammar, Hungarian Academy of Sciences .

Koopman, Hilda and Dominique Sportiche

1983

Variables and the Bijection Principle, in TheLinguistic Review 2.3,
139–160.

1991

The position of subjects in Lingua, 85.1, 211–258.

202

References

Koster, Jan

1987

Domains and dynasties. Foris, Dordrecht.

Krifka, Manfred et alii

1995

Introduction to Genericity. In The Generic Book, 1–124.

Kroch, Anthony

1994

Morphosyntactic Variation. In: K. Beals (ed.), Proceedings of the
Thirtieth Annual Meeting of the Chicago Linguistics Society 2,
Chicago Linguistics Society, Chicago, 180–201.

2001

Syntactic Change. In: Mark Baltin and Chris Collins (eds.), The Hand-
book of Contemporary Syntactic Theory, Blackwell, UK, 699 –729.

Ladd, Robert

1996

Intonational Phonology. Cambridge University Press.

Lambrecht, Knud

1994

Information Structure and Sentence Form. Cambridge University Press.

Larson, R. K.

1988

On the double object construction. Linguistic Inquiry 1/9: 335–391.

Lee, Young-Suk and Beatrice Santorini

1994

Towards resolving Webelhuth’s paradox: evidence from German and
Korean. In: Norbert Corver and Henk van Riemsdijk (eds.), Studies on
scrambling. Movement and non-movement approaches to free word-
order phenomena (Studies in generative grammar 41), 257–300.
Berlin: Mouton de Gruyter.

Levin, B. e M. Rappaport Hovav

1995

Unaccusatives. Cambridge, Mass.: MIT Press.

Li, Charles

1976

Subject and Topic. Academic Press, New York.

Lobo, Maria

2003

Aspectos da Sintaxe das Orações Subordinadas Adverbiais do
Português, Ph.D. Dissertation, Universidade Nova de Lisboa.

Mahajan, Anoop

1990

The A/A-bar distinction and Movement theory. Ph.D. dissertation,
MIT.

Marantz, A. P.

1984

On the Nature of Grammatical Relations, MIT Press, Cambridge,
Mass.

Martins, Ana Maria

1994

Os clíticos na história do português. Doctoral dissertation, University
of Lisbon.

Mateus, Maria Helena et alii

1989

Gramática da Língua Portuguesa. Lisboa: Editorial Caminho.

May, Robert

1985

Logical Form. MIT Press, Cambridge, Mass.

References

203

McCloskey, James

1992

Adjunction, Selection and Embedded Verb Second. Ms, University of
Santa Cruz.

1992a

On the Scope of Verb Movement in Irish. Ms, UCSC.

Menuzzi, Sérgio

1994

Adjectival positions inside DPs. In: Crit Cremers and Reineke Bok-
Benema (eds.), Linguistics in the Netherlands, John Benjamins,
Amsterdam, 127–138.

1994

On the Role of F-features: empty categories, binding and the pro-
nominal system in Brazilian Portuguese. Ms, HIL\Leiden University.

Müller, Gereon and Wolfgang Sternefeld

1994

Scrambling as A-bar Movement. In: Corver & Riemsdjik 1994.

Nash, Lea

1995

Argument scope and Case Marking in SOV and in Ergative Lan-
guages: the case of Georgian. Doctoral dissertation, Univ. Paris 8.

Neeleman, Ad

1994

Complex Predicates. Ph.D. dissertation, Utrecht University.

Neeleman, Ad and Tanya Reinhart

1999

Scrambling and the PF-interface. In: W. Geuder and M. Butt (eds.).
The projection of arguments: lexical and compositional factors.
CSLI Publications, Stanford, 309–353.

Nespor, Marina and Irene Vogel

1986

Prosodic Phonology. Foris, Dordrecht.

Nespor, Marina, Teresa Guasti and Anne Cristophe

1995

What can Infants Learn from Prosodic Constituents? In GLOW News-
letter, 34.

Ordoñez, Francisco and Esthela Treviño

1995

Los sujetos y objetos preverbales en español. Paper presented at the
5th Colloquium on Generative Grammar, Coruña, Spain.

Ordoñez, Francisco

1997

Word Order and Clause Structure in Spanish and other Romance lan-
guages. Doctoral dissertation, CUNY.

Ouhalla, Jamal

1991

Functional Categories and Parametrization. Oxford: Routledge.

Pesetsky, David

1982

Paths and Categories. Doctoral dissertation, MIT.

1987

Wh-in-Situ: Movement and Unselective Binding. In: E. Reuland and
A. ter Meulen (eds.). The Linguistic Representation of (In)definite-
ness. 98–129, Cambridge, MIT Press.

1989

Language Particular Processes and the Earliness Principle. Ms, MIT,
Cambridge, Mass.

204

References

Pesetsky, David

1995

Zero syntax. MIT Press, Cambridge, Mass.

Phillips, C.

1996

Order and Structure. Doctoral dissertation, MIT.

Pinto, Manuela

1994

Subjects in Italian: Distribution and Interpretation. In: Reineke Bok-
Bennema and Crit Cremers (eds.), Linguistics in the Netherlands
1994, 175–187, Amsterdam, John Benjamins Publishing Company.

1997

Licensing and Interpretation of Inverted Subjects in Italian. Doctoral
dissertation, OTS/Utrecht University.

Pollock, Jean-Yves

1989

Verb Movement, Universal Grammar and the structure of IP. Linguistic
Inquiry, 20: 365– 424.

1994b

Notes on Checking Theory, pro-drop, Free Variation and Economy.
Paper presented at the Going Romance VIII, Utrecht.

Pratas, Fernanda

2002

O sistema pronominal do caboverdiano. Questões de gramática.
Mass. dissertation, Universidade Nova de Lisboa.

Pustejovsky J.

1995

The Generative Lexicon. Cambridge: MIT Press

Quer, Josep

2002

Edging Quantifiers: On QP-Fronting in Western Romance. In:

C. Beyssade et al., Romance Languages and Linguistic Theory 2000,
Amsterdam: Johns Benjamins, 253–270.

Raposo, Eduardo

1986

On the Null Object in European Portuguese. In: O. Jaegli and Silva-
Corvalan (eds.), Studies in Romance Linguistics, 373–390.

1987

Case Theory and Infl-to-Comp: the Inflected Infinitive in European
Portuguese. Linguistic Inquiry, 20: 85–110.

1994

Affective Operators and Clausal Structure in European Portuguese
and European Spanish. Paper presented at 24

Linguistic Symposium

on Romance Languages, UCLA/USC.

2000

Clitic Position and Verb Movement in European Portuguese. In: João
Costa (ed.), Portuguese Syntax. New Comparative Studies. Oxford
University Press, 266–298.

Reinhart, Tanya

1995

Interface Strategies. Ms, OTS/Utrecht University.

Ribeiro, Raquel

2002

As ocorrências da forma de gerúndio na variedade padrão e numa
variedade dialectal do português europeu. Mass. Dissertation, Uni-
versidade Nova de Lisboa.

References

205

Rizzi, Luigi

1982

Issues in Italian Syntax. Foris, Dordrecht.

1991

Residual Verb Secondand the Wh-criterion. Technical Reports in
Formal and Computational Linguistics, 2.

1997

The Fine Structure of the Left Periphery. In: Haegeman (ed.), Elements
of Grammar. Handbook in generative syntax. Kluwer, Dordrecht,
281–337.

Rochemont, Michael and Peter Culicover

1990

English Focus Construction and the Theory of Grammar. Cambridge,
England: Cambridge University Press.

Rohrbacher, Bernard

1994

The Germanic Languages and the Full Paradigm: A Theory of V to I
Raising, unpublished Ph.D. dissertation, University of Massachusetts,
Amsherst, Mass.

Rooth, Mats

1985

Association with Focus. Doctoral dissertation, University of Massa-
chusetts, Amsherst, Mass.

Ross, John

1967

Constraints on variables in syntax. Doctoral dissertation, MIT.

Santorini, B.

1990

ÎNFL and scrambling in German. Ms, University of Pennsylvania.

Scherre, Marta

1994

Aspectos da concordância de número no português do Brasil. Revista
Internacional de Língua Portuguesa – Norma e Variação do Português
12, 37– 49.

Schoorlemmer, Martin

1998

Possessors, articles, and definiteness’. In: Artemis Alexiadou and
Chris Wilder (eds.), Possessors, predicates and movement in the
Determiner Phrase. John Benjamins, Amsterdam, 55–86.

Selkirk, Lisa

1984

Phonology and Syntax: the relation between sound and structure.
MIT Press, Cambridge, Mass.

Sportiche, Dominique

1988

A Theory of Floating Quantifiers and Its Corollaries For Constituent
Structure. Linguistic Inquiry, 19: 425– 449.

von Stechow, Arnim

1990

Current Issues in the Theory of Focus. In: Semantics. International
Hanbook of Contemporary Research. Mouton de Gruyter, Berlin,
804–825.

Szabolcsi, Anna

1981

The Semantics of Topic-Focus Articulation. In Formal Methods in
the Study of Language.

206

References

Szendrói, Kriszta

2001

Focus and the Syntax-Phonology Interface. Ph.D. dissertation,
University College, London.

Uriagereka, Juan

1995

An F Position in Western Romance. In: Kiss (1995).

Vallduví, Enric

1990

The Informational Component. Doctoral dissertation, University of
Pennsylvania.

Valmala Elguea, Vidal

1994

Spanish Word-Order and Checking of Morphological Features. Talk
given at Going Romance 8, Utrecht.

Vanden Wyngaerd, Guido

1990

Object Shift as an A-movement Rule. In MIT Working papers in
linguistics.

Vikner, Sten

1994

Scandinavian Object Shift and West Germanic Scrambling. In:
Corver & Riemsdijk (1994).

1997

Vº-to-Iº movement and inflection for person in all tenses. In: Liliane
Haegeman (ed.): The New Comparative Syntax. Longman Linguistics
Library, Edinbourgh, UK, 189–213.

Webelhuth, Gert

1989

Syntactic Saturation Phenomena and the Germanic Languages.
Doctoral dissertation, UMass, Amherst

Williams, Edwin

1994

A reinterpretation of the evidence for verb movement in French. In:
D. Lightfoot and N. Hornstein (eds.), Verb Movement. Cambridge
University Press, 189–206.

Wurmbrand, Susi

2001

Infinitives. Mouton de Gruyter, Berlin.

Zubizarreta, Maria Luisa

1998

Word Order, Prosody, and Focus. MIT Press, Cambridge.

1998a

The Structure of the Higher Middle Field: the position of the verb and
the subject. Paper presented at the Workshop on Inversion in Romance,
University of Amsterdam.

Zwart, Jan-Wouter

1993

Dutch Syntax. A Minimalist Approach. Ph.D. dissertation, University
of Groningen.

References

207

A-bar, 8, 13–15, 32, 44–45, 56, 63, 65,

69, 73, 144–145, 150–152, 185,

adjacency, 4, 8, 14, 36, 40, 52, 53, 57,

75–76, 129, 131, 133–136, 185,
191

adjunction, 2, 14, 27, 29, 31, 36, 40,

45–46, 50, 68, 74, 99, 129, 144,
160, 192

adverbs, 2, 6–10, 14, 23, 25–29, 32, 36,

39–41, 43–44, 48, 50, 63–64,
66–69, 96–97, 129–130, 134–137,
159, 163, 174, 185, 186, 187, 190,
191

Agree, 1, 3, 71, 97–100, 102–105, 124,

159, 190

agreement, 3, 12, 75, 77, 100–105, 109–

119, 163–183, 191, 192, 193

AgrP, 1–5, 6, 20–22, 39, 45–46, 63, 99,

131–139, 159–160

anaphor, 33, 44, 143
A-position, 2–3, 6, 15–16, 37–38, 45,

107, 138–139, 145, 150, 160

auxiliary verbs, 7, 24, 26, 29, 48, 49,

134, 137, 191, 192

binding, 32–33, 37–38, 44, 73, 143–146,

149–153, 186

Brazilian Portuguese, 20, 100 –105,

113–115, 118–119, 142, 155–156,
163–183, 185, 190, 191

Capeverdean, 20 –21
Case, 1, 3, 21, 39, 45, 47, 62–65, 76, 97,

99, 105, 108, 112–118, 127, 159,
186

c-command, 58–59, 65, 87–91, 149, 189
Celtic, 24
cleft, 48–49, 52, 87, 188

climbing, 97, 100, 102–104
clitic left dislocation, 12, 21, 185
complementizer, 19, 26, 186
contrastive, 8, 55–56, 72–74, 80–82,

85–86, 89, 94–97, 124–125, 188

control, 22, 104, 150, 190

definite, 3, 8, 17–18, 34, 62, 65, 81, 85,

110, 114, 127, 153–155, 173, 175,
178 –179, 186

Distributed Morphology, 131–133, 175
ditransitive, 141–151, 155–156
D-linking, 120, 125–126
doubling, 18–21, 34, 48, 52–54, 126
Dutch, 8, 24, 35–45, 63, 67–69, 133,

186, 187

ellipsis, 122, 125
English, 7, 9, 14, 19, 24, 27–28, 39–41,

46, 67, 72–73, 75–77, 83–84,
101–102, 114 –115, 132–133, 135,
141–142, 181, 186, 187, 188

EPP, 12, 20–21, 127
exclamative, 91–95
exhaustivity, 2, 4, 16, 72, 120–125, 160
expletive, 20–21, 24, 29, 113–115, 117–

119, 132–134, 138, 178–179, 181

floating quantifiers, 9, 45, 48, 51–52,

163, 174, 187

focus, 1–5, 8, 18, 54–56, 58, 64–69, 74–

105, 107–108, 112–115, 119–121,
124–127, 141–145, 153–155, 159–
160, 181, 188, 189, 190

focus-movement, 9, 71–78, 91–96, 154,

190

French, 6–7, 9, 29, 47, 54, 176, 177, 183,

186

Index

fusion, 131–132, 136

generic, 124–125
German, 24, 35–45, 63, 65–69, 102,

133, 187

Germanic, 3, 4, 39–45, 62, 134–135, 138
Greek, 12, 22, 46

Head Movement Constraint, 135
head-movement, 29, 131–136, 175, 192
Heavy NP shift, 31, 74
Hungarian, 72–76, 78, 91, 125, 188, 190

Icelandic, 24, 28, 45, 63, 102,

132–133, 186, 187

if-clauses, 19
indefinite, 8, 17, 34, 56, 62, 65–67, 81,

126–127, 178, 186

indirect object, 52–53, 142, 145, 149–150
infinitives, 6 –7, 137
inflected gerunds, 22
Information Structure, 1, 3, 5, 71–105,

159

intransitive, 16, 64, 108–109, 111, 117,

191

inversion, 3, 4, 16, 20–27, 47, 54,

71–105, 108, 111–119, 129, 142,
155, 170, 174, 180–182, 185,
190

Italian, 8–9, 14, 30, 64, 91–93, 113,

153–155

I-to-C, 2–4, 14, 19, 23–25, 129–131,

135–137, 186, 191

language acquisition, 21
left-dislocation, 1–4, 6, 8,11–22, 35,107,

118, 120–127, 159–160, 185, 191

locality, 4, 97–103, 141, 159
locative inversion, 113–119, 181, 185

manner, 27, 66, 129,
merger, 2, 131–135, 160, 163
minimality, 14–15, 190

morphology, 2, 4–5, 117, 129–139, 160,

163–183, 191

Move, 1, 3, 97, 105, 159, 190

negation, 14, 98, 103, 122–123
nominative, 32, 47, 108, 112–113, 115–

118, 191

nuclear stress, 1, 58, 67–69, 81, 90, 141–

142, 159, 189

null object, 42– 43, 187
null subject, 2– 4, 11–22, 118, 138, 160,

181, 186

object shift, 45, 63, 99, 187
optionality, 2, 4, 6, 12, 97, 107–127, 160
OSV, 1, 11, 13, 79
OVS, 1, 11, 79

parasitic gap, 37–38, 42–43, 45, 56, 187
participle, 26, 29, 168, 171, 173–174,

176 –177, 180, 191

partitive, 113, 116 –117
passive, 21, 33, 37, 145, 168, 171–174,

180

phase, 4, 97–102, 159
possessive, 33, 44, 141, 153–155, 165–

167, 173–174, 178–179, 182, 192

postverbal, 3–4, 19–21, 23–70, 75, 96,

100, 109 –118, 122, 191

PP, 27, 36, 39, 43, 45, 75, 85, 88, 93,

96, 142, 148–149, 151, 186

pragmatic, 38
predicative, 21, 37, 43–44, 171–172
preposing, 13, 48–50, 72, 92–96, 190
preverbal, 2– 4, 11–22, 35, 53, 60, 100,

107–118, 121, 122, 125–127, 129,
136, 155, 160, 186

pro, 20 –21, 115, 119, 156, 186
PRO, 150, 190
pro-drop, 20
pronominal, 12, 18–22, 30–32, 34, 37, 45,

48, 52–53, 56, 73, 77, 85, 100,
116 –117, 126, 130, 154, 187, 192

210

Index

prosody, 1, 3, 5, 58, 78, 83, 88, 105, 141,

153, 159, 189

QP, 14, 33, 37, 44, 65, 186, 190

raising, 17, 22, 65, 191
reconstruction, 15, 59, 151
remnant movement, 3, 35, 47–62, 187
right-adjunction, 27

Scandinavian, 45, 63
scope, 15, 48, 58–59, 65–69, 146, 185
scrambling, 3, 8, 30, 33, 35–69, 79, 90,

99, 144–145, 150, 186, 187

small clause, 37, 43, 168, 173–174, 180
Spanish, 12, 22, 46–47, 55
Spec,IP, 3, 5, 9, 11–22, 23–27, 35, 46 –

47, 51–52, 60–61, 64–65, 71,
78–79, 88, 97, 101, 103, 108, 113,
116–119, 124, 126–127, 144, 156,
185, 190

Spec,VP, 1, 3, 5, 16, 20, 23–35, 38,

47–67, 71, 78–79, 97–99, 105,
144, 151, 159, 191

specificity, 38, 62–63, 65, 126–127, 187
Spell-Out, 5, 175
stress, 1, 3, 8, 14, 55, 58, 67– 69, 78,

81–91, 141–143, 153–157, 159 –
160, 187, 188, 189

subject-oriented, 26 –27, 129–130, 136 –

137

SVO, 1– 4, 11–22, 39, 59, 78, 81–83,

86–90, 97, 107, 109, 119–127,
144, 159–160, 189, 190

tag, 31, 33, 34, 48, 52–53, 95
topicalization, 12, 56–58, 61–62, 79, 81,

94–65, 126, 136 –137, 185, 190

TP, 1–5, 26–27, 47–48, 52, 59, 103–104,

129–139, 159–160

Transitive Expletive Constructions, 24,

132–134, 138

unaccusative, 20–22, 64–65, 107–119,

155, 167, 170–173, 180–181, 185,
190, 191, 193

unmarked, 11, 15–16, 21, 81–90, 107–

108, 112, 145–147, 189, 190

verb movement, 4, 7–9, 27, 39–40, 133,

135, 163–164, 174–175, 182–183

Verb-second, 24, 133, 186
VOS, 1, 3, 11, 23, 30 –70, 79, 82, 84,

87–88, 90, 109, 144–145, 150 –
152, 187

VSO, 1, 3, 4, 11, 21, 23–35, 78, 82, 87–

90, 97, 99, 109, 119–127, 144 –
145, 159, 189

wh-questions, 4, 18, 27, 47, 55–56, 58,

93, 107, 119–127, 130, 135, 160,
187

Index

211

Document Outline

Contents
Foreword
1. Introduction
2. Preverbal subjects: Spec,IP or left-dislocated?
3. Postverbal subjects: syntax and discourse – VSO and VOS orders
4. Inversion and information structure
5. Optionality and left-dislocated subjects: semantic and discourse properties
6. Subjects in Spec,TP and the interface with morphology
7. Syntactic outputs and the interfaces
8. Summary and conclusions
Appendix: On the nature of agreement in European Portuguese
Notes
References
Index

Subject Positions and Interfaces The Case of European Portuguese

Document Outline