(Current Issues in Linguistic Theory 245) Josep Quer (Ed.), Jan Schroten (Ed.), Mauro Scorretti (Ed.), Petra Sleeman (Ed.), Els Verheugd (Ed.) - Romance Languages and Linguistic Theory 2001

ROMANCE LANGUAGES AND LINGUISTIC THEORY 2001
AMSTERDAM STUDIES IN THE THEORY AND

HISTORY OF LINGUISTIC SCIENCE
General Editor
E. F. KONRAD KOERNER
(Zentrum für Allgemeine Sprachwissenschaft, Typologie
und Universalienforschung, Berlin)
Series IV – CURRENT ISSUES IN LINGUISTIC THEORY
Advisory Editorial Board
Raimo Anttila (Los Angeles); Lyle Campbell (Christchurch, N.Z.)

Sheila Embleton (Toronto); John E. Joseph (Edinburgh)
Manfred Krifka (Berlin); Hans-Heinrich Lieb (Berlin)
E. Wyn Roberts (Vancouver, B.C.); Hans-Jürgen Sasse (Köln)
Volume 245
Josep Quer, Jan Schroten, Mauro Scorretti, Petra Sleeman and

Els Verheugd (eds.)
Romance Languages and Linguistic Theory 2001: Selected papers from ‘Going
Romance’ 2001, Amsterdam, 6–8 December
ROMANCE LANGUAGES AND
LINGUISTIC THEORY 2001
SELECTED PAPERS FROM ‘GOING ROMANCE’
AMSTERDAM, 6–8 DECEMBER 2001
Edited by
JOSEP QUER
University of Barcelona
JAN SCHROTEN
Utrecht University
MAURO SCORRETTI
PETRA SLEEMAN
ELS VERHEUGD
University of Amsterdam
JOHN BENJAMINS PUBLISHING COMPANY

AMSTERDAM/PHILADELPHIA
TM The paper used in this publication meets the minimum requirements of American
8
National Standard for Information Sciences — Permanence of Paper for Printed

Library Materials, ANSI Z39.48-1984.
Library of Congress Cataloging-in-Publication Data

Romance languages and linguistic theory 2001: Selected papers from ‘Going Romance’, Amsterdam,
6–8 December 2001 / edited by Josep Quer ... [et al.]
p. cm. -- (Amsterdam studies in the theory and history of linguistic science. Series IV, Current
issues in linguistic theory, ISSN 0304-0763 ; v. 245)
Includes bibliographical references and index.
1. Romance languages--Congresses. I. Quer i Carbonell, Josep, 1962-. II. Series.
PC11 .R63 2003
440--dc22 2003045130
ISBN 90 272 4757 9 (Eur.) / 1 58811 445 7 (US) (Hb; alk. paper)
© 2003 – John Benjamins B.V.
No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other
means, without written permission from the publisher.
John Benjamins Publishing Co. • P.O.Box 36224 • 1020 ME Amsterdam • The Netherlands
John Benjamins North America • P.O.Box 27519 • Philadelphia PA 19118-0519 • USA
INTRODUCTION
Going Romance is a major European annual discussion forum for theoretically

relevant research on Romance languages; it is an international initiative of the
Dutch university community involved in research on Romance languages. The
proceedings volumes, titled Romance Languages and Linguistic Theory, contain
the selected papers of the Going Romance conferences, which have been
organized by and held at the various universities of the country. The first volume
contained the selected papers of the thirteenth conference, held in 1999.
This is the third proceedings volume. The articles form a selection of the papers
that have been presented at the occasion of Going Romance 2001 (XV) - which
was held at the University of Amsterdam on December 6 through December 8.
The three day program included a workshop on Determiners. The volume
contains articles on specifics of one or more Romance languages or varieties: the
architecture of the Determiner Phrase and properties of determiners, the left
periphery of the sentence and clause structure, null elements and their
interpretation, clitics, and other interesting phenomena.
The editors would like to thank everyone who contributed to the success of
Going Romance XV. Next to some of the editors, the organization committee
consisted of Ileana Comorovski (Université de Nancy 2), Denis Delfitto
(Utrecht, UiL OTS), Jenny Doetjes (Utrecht, UiL OTS), Frank Drijkoningen
(Utrecht, UiL OTS), Aafke Hulk (Amsterdam, HIL), Brigitte Kampers -Manhe
(Groningen, CLCG).
The selection committee for the more than fifty abstracts for the main session
and the workshop consisted of editors and organizers and was assisted by the
invited speakers João Costa (Lisbon), Richard Kayne (New York), Brenda Laca
(Paris 8), Giuseppe Longobardi (Trieste), Luigi Rizzi (Siena), Liliane Tasmowski
(Antwerp), Karen Zagona (Washington) and by the following independent
advisors: Leonie Bosveld (Groningen, CLGC), Francis Corblin (Paris 4
Sorbonne), Carmen Dobrovie-Sorin (Paris 7), Jenny Doetjes (Utrecht, UiL OTS),
Jean-Marie Marandin (Paris 7), Johan Rooryck (Leiden, HIL) and Lucia Tovena
(Lille 3).
VI INTRODUCTION
The organizers and the editors gratefully acknowledge financial support from the
Amsterdam Center for Language and Communication (ACLC), the Royal
Netherlands Academy of Arts and Sciences (KNAW) and the Netherlands
Organisation for Scientific Research (NWO), the Utrecht institute of Linguistics
OTS (UiL OTS), the Holland Institute of Generative Linguistics (HIL), and the
determiner group of the PICS project 'Formal semantics and French data' (co-
funded by CNRS and NWO).
Finally, we wish to thank Ans de Kok, who took care of Web support. We wish to
give special thanks to Jasper Roodenburg for his invaluable assistance in
organizing Going Romance 2001.
Josep Quer
Jan Schroten
Mauro Scorretti
Petra S leeman
Els Verheugd
CONTENTS
Spanish De-Clauses Are Not Always in the Right Mood 1

Luis Alonso-Ovalle
Mood and Focus 17

Claudia Borgonovo
Null vs Overt Spec,TP in European Portuguese 31

João Costa
Determiner Architecture and Phrasal Movement in French Lexifier Creoles 49

Viviane Déprez
On the Relation between Focus, Prosody and Word Order in Romanian 75

Edward Göbbel
Economy of Structure. The Case of Subject Clitics in Piedmontese 93

Cecilia Goria
Identificational Focus vs Contrastive Focus. A Syntactic Distinction 113

Daniela Isac
Null Objects and VP Ellipsis in European and Brazilian Portuguese 131

Mary Aizawa Kato
From Non-Identity to Plurality. French Différent as an Adjective and

as a Determiner 155
Brenda Laca & Liliane Tasmowski
On the Non-Unitariness of NP Subject Inversion. A Comparison of French

NP Subject Inversion in Interrogatives and Temporal Subordinates 177
Karen Lahousse
viii CONTENTS
Past Participle Agreement with Pronominal Clitics and the Auxiliary Verbs
in Italian and French 193
Paul Law
Deficient Pronouns and Linguistic Change in Portuguese and Spanish 213

Ana Maria Martins
Nominalizations of French Psychological Verbs. Syntactic Complements 231

and Semantic Participants
Judith Meinschaefer
Notes on Vocative Case. A Case Study in Clause Structure 247

Andrea Moro
Mapping out the Left Periphery of the Clause. Evidence from

North Western Italian Varieties 263
Sandra Paoli
The Left Periphery in Child French. Evidence for a Simply-Split CP 279

Dorian Roehrs & Marie Labelle
Plural Indefinite DPs as Plural-Polarity Items 295

Benjamin Spector
On the Status of the Partitive Determiner in Italian 315

Gianluca Storto
Determiners and Weakly Discretised Domains 331
Lucia Tovena
Index of languages and dialects 347
Subject Index 349

SPANISH DE-CLAUSES ARE NOT ALWAYS
IN THE RIGHT MOOD*
LUIS ALONSO-OVALLE
University of Massachusetts, Amherst
1. latridou's question
The benchmark theory of conditionals maintains that conditionals quantify
over a contextually restricted domain of worlds (Kratzer 1991). They are modal
statements. The antecedent contributes to the interpretation of the whole
conditional a proposition, a set of worlds. Conditionals quantify over a
contextually restricted domain of worlds in which the proposition that the
antecedent expresses is true. This is all antecedents do. In particular, the semantic
import of its tense and mood inflection is neglected: it is - at most - a merely
formal reflection of the type of modal in the consequent (Fintel 1998; Heim
1992; Kratzer 1991).
This last assumption has been recently challenged. The dissection of
counterfactual conditionals (latridou 2000; Ippolito 2001) has led to questioning
the semantic import of the antecedent's inflection and to wondering whether the
inflections of both the antecedent and the consequent are interpreted. This is, in
short, Iatridou's question.
(1) Iatridou's question: Is the tense/mood marking of both the antecedent

and the consequent interpreted?1
This paper reflects my views on the topic at the time when it was presented at Going Romance.
They have changed slightly since then (see Alonso-Ovalle, in preparation). For practical purposes. I
have limited myself here to the exposition of the original ideas and made no attempt to incorporate
my new views.
Thanks to Shai Cohen, Javier Gutiérrez-Rexach, Kevin Klement, Paula Menéndez-Benito, Josep
Quer, Mike Terry, two anonymous reviewers and the audience of Going Romance 2001. Special
thanks to Roger Higgins for his incisive comments on different parts of this work; to Barbara Partee
for her sharp insights, her enthusiasm and benevolence with too often too poorly developed
observations. 1 am indebted to Angelika Kratzer for more enouragement, help and advice - linguistic
and non-linguistic alike - than I could aknowledge here. My gratitude to Sandra Barriales for too
many hours unspent with and too many judgements asked for. Errors can only be mine.
2 LUIS ALONSO-OVALLE
Suppose the inflection of the antecedent is semantically vacuous. Then, if a

language allows antecedents to optionally lack any tense or mood marking
whatsoever, the interpretation of a conditional under both its finite and non-finite
versions should be equivalent.
Spanish conditionals, indicative and subjunctive alike, come in two
varieties depending on whether the antecedent is an inflected (si-conditionals) or
an infinitival clause (de-conditionals):
(2) a. Si Caligula está muerto, me tendrán que ascender

if Caligula is dead PRO me will-have-to promote.INF
b. De estar muerto Caligula, me tendrán que ascender.
DE be INF dead Caligula PRO me will-have-to promote.INF
"If Caligula is dead, I will have to get promoted."
The antecedent of a de-conditional (henceforth a de-clause) can host

aspectual heads and sentential negation, as illustrated in (3), but no overt mood
inflection.
(3) a. De haber matado a Caligula.

DE PRO have.INF kill.PART AC Caligula
b. De no haber matado a Caligula.
DEPRO NOT have.INF kill.PART AC Caligula
If de-clauses are truly uninflected (and if the preposition/complementizer is truly

semantically vacuous2), assuming that mood marking in the antecedent is
semantically vacuous, we expect - quite close to the intuitions - the si- and de
versions of a conditional to be interpreted in the same way. I will show that they
are not. Then, insofar as the si- and de-versions of a conditional are not
interpreted the same way, we are bound to consider the semantic import of the
mood inflection of antecedents.
1
"I did not address the question of whether ExclF ['Exclusion Feature': the interpretation of past
morphology in subjunctive conditionals, L.A.O.] plays an equal role in the antecedent and in the
consequent, leaving open the possibility that the appearance of ExclF in one is an agreement
phenomenon of sorts with the other." (Iatridou 2000:267).
2
Part of my work in progress addresses the issue of whether the preposition-complementizer is truly
semantically vacuous. The moment we consider the full range of prepositional conditionals in
Spanish, it becomes apparent that it is not. Nevins (2002) shows the existence of complementizers
that convey counterfactuality, sometimes even in the absence of overt inflection of the antecedent.
The semantic contribution of conditional complementizers must be taken seriously. Unfortunately, I
cannot do justice to the subtleties of the topic here.
SPANISH DE-CLAUSES 3
The paper is organized as follows: section 2 presents the data. Section 3

shows that si-conditionals and de-conditionals are not equally appropriate in the
same scenarios and suggests an explanation based on the assumption that the
mood inflection of the antecedent is interpreted. Section 4 makes the explanation
explicit. It deals with the interpretation of mood marking in the antecedent of
conditionals and its potential interactions with the modals in the consequent. It
also shows how the absence of mood inflection determines the interpretation of
de-conditionals. Finally, Section 5 presents some conclusions and open issues.
2. Marking-off th e territory
I start by borrowing a context from the literature. Kratzer (1979:133)
reports the following story from Ancient Rome:
Scenario 1
When Caligula left the arena one day, suddenly the doors shut behind him and
he was attacked by his own body-guard. The crowd in the arena heard him
screaming but they could only guess what had happened. Maybe Caligula was
dead, maybe he was still alive.
In this situation, if Marcus had spoken Spanish, he could have uttered the
sentence in (4a) or that in (4b).
(4) a. Si estámuerto, me tendrán que ascender.

if PRO is dead.IND PRO me will-have-to.3PL promote.INF
b. De estar muerto, me tendrán que ascender.
DE PRO be dead PRO me will-have-to.3PL promote.INF
"If he is dead, I will have to get promoted."
He could also have uttered the sentence in (5a), or that in (5b).
(5) a. Si estuviera muerto (ahora), me tendrían que

if PRO were.PSTSBJ dead now PRO me would-have-to.3PL
ascender.
promote. INF
"If he were dead (now), I would have to get promoted."
b. De estar muerto (ahora), me tendrían que

DE PRO be dead now PRO me would-have-to.3PL
ascender.
promote.INF
De-conditionals can be paired up with either indicative or subjunctive si-

conditionals. In fact, the two versions of the previous sentences look pretty
similar (Kany 1936, 1939; Söhrman 1991; Montolío 1999). Nevertheless, no
matter how close they look, si-conditionals and de-conditionals are associated
with different appropriateness conditions, which I illustrate next.
3. When is it appropriate to use a de-conditional?

The appropriateness conditions associated with de-conditionals can only be
understood when contrasted with the appropriateness conditions associated with
si-conditionals. To these, I now turn.
3.1 Presuppositions of indicative and subjunctive si-conditionals

I start by making the standard move within possible world semantics: I
assume that sentences express propositions, which are conceived of as sets of
worlds. A proposition is construed as the set of all those worlds in which it is
true. A proposition p expressed by a sentence a ([a]) is true in a world w if w is
a member of p.3
I then assume a modal analysis of conditionals. Modals are quantifiers.
They quantify over possible worlds. Like any quantificational statement, modal
statements are evaluated with respect to an implicitly understood domain of
quantification. One such possible domain is the 'context set', the set of worlds
that, for all the parties involved in the conversation know, could be the actual one
(Stalnaker 1998). In this paper I will focus on 'epistemic' conditionals,
conditionals whose domain of quantification is the context set and avoid the
complications that result when conditionals quantify over different domains.
I start by assuming the following appropriateness conditions of (si-)
conditionals. Indicative conditionals are appropriate in a context only if both the
antecedent and its negation are compatible with the context set (Stalnaker 1975;
Bigelow 1976; Kratzer 1979). Subjunctive conditionals are appropriate in a
context if the negation of the antecedent is compatible with the context set. A
frequent use of subjunctive conditionals (their 'counterfactual' use) requires that
3
When no confusion is likely to arise, I will use the terms 'antecedent' and 'consequent' to refer to
the propositions the antecedent and consequent of a conditional expresses.
the proposition expressed by the antecedent be inconsistent with the context set
(Kratzer 1979).
(6) Appropriateness conditions for indicative conditionals:

C ∩ [α] ≠ Ø and C∩[not-α] ≠ Ø
Appropriateness conditions for subjunctive conditionals:
(i) C ∩ [not-α] ≠Ø
(ii) Counterfactual uses: C ∩ [α] = Ø
3.2 Back to Caligula

I will now put the machinery to work. In the scenario with which I opened
section 2, it is an open possibility whether Caligula is dead or not. Consequently,
the utterances of the indicative conditionals in (4) are correctly predicted to be
felicitous. The scenario forces an epistemic interpretation of the conditionals. A
most plausible domain of quantification is the context set. The appropriateness
conditions associated with indicative conditionals require that the context set
contain worlds in which Caligula is dead and worlds in which he is not. Given
the way the scenario is set up, the requirement is easily met.
The non-counterfactual uses of the subjunctive conditionals in (5) are also
correctly predicted to be felicitous. It is required that some of the worlds in the
context set be worlds in which Caligula is not dead and the requirement is met.
Let us now change the context slightly.
Scenario 2
Some minutes later, the doors of the arena open and in comes Caligula, greeting the
crowd. (Kratzer 1979:134)
This event changes the common ground in a crucial way. Now we add to
the common ground the proposition that Caligula is alive. All worlds in the
context set are now worlds in which Caligula is alive.
Imagine, as in Kratzer (1979), that (4) and (5) are uttered again in this new
context. Then, the utterances of the indicative conditionals (4) are correctly
predicted to be totally inappropriate, since in no worlds in the context set is
Caligula dead.
Consider now the corresponding subjunctive conditionals:
(5) a. Si estuviera muerto (ahora), me tendrían que ascender.

A counterfactual use of (5a) would be totally appropriate, as predicted,

since the context set does not contain any world in which Caligula is dead.
Under the presumption that si and de-conditionals are totally parallel, the
corresponding de-conditional in (5b) should be equally felicitous:
(5) b. De estar muerto (ahora) , me tendrían que ascender
Contrary to the predictions, however, there is something odd about (5b) when
contrasted to (5 a) in this new context. The sentence in (5b) does not feel to be
appropriate. Some informants report that their intuitions about the felicity of an
utterance of (5b) in this new context are elusive. Bear this in mind.
If mood marking in the antecedent were semantically vacuous, the
interpretation of si- and de-conditionals should always be the same. Both types
of conditionals should be equally appropriate in Scenario 2. They are not. We
then hit upon a puzzle: how come the de-clause makes a difference in the
subjunctive conditional and not in the indicative conditional in this scenario?
And why do some speakers report their intuitions to be elusive?
Since (5a) differs from (5b) just in the type of antecedent, it is the type of
antecedent that must be blamed for the instability of judgments. The contrast
between (5 a) and (5b) shows that mood marking in the antecedent must be
interpreted after all.
My solution to this puzzle is the following: I take the mood inflection of the
antecedent to be interpreted. In the next section, I will adhere to the view that
indicative and subjunctive conditionals differ as to how they change the context
in which they are uttered. I will propose that the mood inflection in the
antecedent of epistemic conditionals signals how the domain of quantification is
modified. Indicative mood is by default associated with shrinking. An indicative
antecedent shrinks the domain of quantification by stripping away from it those
worlds in which the antecedent is false. Following Stalnaker (1975), I will take
this to be the default strategy. The marked strategy, associated with the
subjunctive, is the expansion of the domain. Subjunctive antecedents require that
the domain of quantification be (possibly) expanded so as to include worlds in
which the antecedent is true (Von Fintel 2001; Quer 2001).
Since de-clauses are moodless, they lack any overt instructions as to how
they should modify the domain of quantification. However, if the strategy
associated with the indicative is the default, they are expected to stick to it,
unless coerced by the modal to behave as subjunctive clauses. The contrast
between (5a) and (5b), what I will call the Caligula effect, illustrates what
happens when committing to the default strategy turns out to be a fatal move. If
the de-clause behaved as an indicative clause in Scenario 2, (5b) should be out. If
it behaved as a subjunctive clause, it should be felicitous. The tendency to stick
to the default strategy proves to be fatal in Scenario 2. However, the modal can
still coerce the de-clause to behave as a subjunctive clause, repairing the disaster.
This explains why intuitions might be elusive. Taking the de-clause to behave
(momentarily) as an indicative antecedent, makes it feel inappropriate. Realizing
that the modal can coerce it to behave as a subjunctive antecedent makes it feel
appropriate. This is, in a nutshell, my explanation of the Caligula effect.
In what follows, I will introduce some assumptions to make the reasoning a
little more explicit. Those readers that might not be interested in the particular
technical implementation of the previous reasoning can skip section 4 without
much harm.
4. Revisiting the Caligula effect

To be a little more explicit, I need to wax technical for a moment. I start by-
modeling a context as a pair <C,f>, where  is the context set, the set of all
worlds compatible with all the parties involved in the conversation know
(Stalnaker 1998) and f ('the modal horizon' in Von Fintel 2001) is an
accessibility function associated with C, a function from worlds to sets of
worlds. The modal horizon is responsible for determining the domains of
quantification of (counterfactual) subjunctive conditionals.
4.1 Subjunctive (counterfactual) antecedents expand the existing domains

Then, I stick to Von Fintel (2001) and Quer (2001) in assuming that
updating the context with a subjunctive conditional involves expanding the
domain of quantification by adding to it the set of worlds most similar to those
already in it in which the antecedent is true. We then need a notion of maximal
similarity between worlds. For the sake of simplicity, I will assume, as in Heim
(1992) and Von Fintel (2001) that the notion of similarity between worlds is
given. To define maximal similarity we need to simply import the order-theoretic
definition of greatest element. As in Heim (1992) and Von Fintel (2001), I will
rely on a definition of maximal similarity that compares worlds where a certain
proposition is true. For any world w and any proposition p, the function max≦w
(p) selects the closest worlds to p (according to the given ordering) in which p is
true (see Fintel 2001):4
4
The assumption being thatmax≦w(p) is defined for any w and p whatsoever, i.e. that for any world
w there exists a set of closest worlds in which p is true. In the context of developing a semantics for
counterfactuals, this assumption has been dubbed by Lewis 'the Limit Assumption'. In what follows,
I stick to it. For arguments against it, see Lewis (1973). For arguments in favor, see Stalnaker
(1984:140-142). For an overview of the role of the notion of similarity in the development of a
semantic theory of counterfactuals, see Nute (1984).
8 LUIS ALONSO-OALLE
(7) For any proposition p ε þ (W), any world w and any similarity
relation≤ w ,
max≤w (p) = {w': w' ε p &∀w": w" ε p → w " ≤ w ' }
I assume that, in an initial context, the parties involved in conversation

know nothing and hence every possible world could be the actual one:  is the
set of all possible worlds (see (8)). The modal horizon is minimal: it assigns to
each world w the singleton {w} (as in von Fintel 2001).
(8) The initial context: < W, λw.{w} >5
Updating the initial context with a subjunctive antecedent amounts to expanding

the available modal horizon by adding to it the closest worlds in which the
proposition that the antecedent expresses is true, as depicted in Figure 1. If the
proposition that the antecedent expresses is true in a world w, then f(w) does not
grow at all.
Fig. 1 : Updating the context with a subjunctive antecedent
In what follows, I will blame the antecedent for that context change.6
Uttering the antecedent does not change the common knowledge: it just modifies
the modal horizon. In symbols:
5
I follow the informal lambda notation used in Heim and Kratzer (1998). 'λw.{w}' is the name of
that function from worlds to sets of worlds that assigns to each world in its domain the singleton
consisting of that world.
6
The idea behind being that would needs a certain context to be licensed and that antecedents of
subjunctive conditionals are just one way to provide would with the required environment (see
Veltman 2002). Cf. the following examples, due to Veltman (2002):
(i) John didn't drink too much wine. He would have got sick.
(ii) (??) John drank too much wine. He would not have got sick.
(9) <C,f> + ⌈si α [+subjunctive]⌉ =<C, f*>

Where ∀w ε C: f*(w) = f(w)max≤w ([α])
What about the interpretation of subjunctive modals?7 Iatridou (2000)

surveys the morphological setup of verbs in subjunctive conditionals. Cross-
linguistically, past tense morphology appears in subjunctive conditionals
(Papago, Japanese and Korean, Hebrew, Turkish, Basque, English...). As for the
consequent, in English, the modal would arguably decomposes into a modal
component and a past component (Palmer 1986; Abusch 1988; Vlach 1993). In
Modern Greek, the consequent hosts a modal element Oa in combination with
past (Iatridou 2000). Modals in the consequent of subjunctive conditionals in
Romance host a verb in the so-called 'conditional mood'. Iatridou (2000) argues
that the conditional mood is sensibly analyzed synchronically as the combination
of a future modal plus past morphology:
(10) a. English: would= will (modal)+ 'past'

b. Modern Greek: Oa (modal) + 'past'
 Romance: -ia = rá (modal)+ 'past'
The feature whose phonetic realization we call 'past' provides, following

Iatridou, what she calls a 'skeletal meaning' of the following form:
(11) Where x ranges over times and worlds, 'T(x)' are 'the x that we are
talking about' and 'C(x)' are 'the x that for all we know are the x of
the speaker',
T(x)C(x) (Iatridou 2000:246)
A natural interpretation of Iatridou's system for the modal uses of 'past'

(when the variable in (11) ranges over worlds) equates T(x) with the domain of
quantification of the subjunctive modals, the modal horizon in our terms, and
C(x) with the context set. By assuming (11), we can provide the following
context change potential for the consequent of subjunctive conditionals.
(12)<C,f> + ⌈si α [+subjunctive]-ria ß⌉ =<C, f*>

Where
(i)∀w (p) = C:f*(w)Èmax≤w([α])
7
In what follows, I will talk of indicative and subjunctive modals to refer to the modals of indicative
and subjunctive conditionals, independently of whether they are in fact subjunctive or not.
(ii) C*= {w ε C: (f(w)Ç [α]) Ç [ß] = f(w) Ç [α]} 8

Defmedness conditions associated with 'past' (hence with ria-$ )
<C,f> + éria-αù is defined iff ∀w e  : f(w) Í 
A subjunctive conditional then performs two operations: the antecedent

changes the modal horizon in the way defined in (9) and then the subjunctive
modal modifies  by keeping all those worlds w in whose modal horizon all
worlds in which [a] is true are worlds where [ß] is also true. The process is
depicted in Figure 2.
The defmedness condition in (12) requires that the domain of quantification
provided by the modal horizon reach out of the context set. The reader must keep
this condition in mind, since it will play a major role in my analysis of the
Caligula effect. The condition is in the spirit of latridou's proposal. Although it
looks like a natural move to make, it is not as innocent as it seems. It will have to
be relaxed to account for non-counterfactual cases of subjunctive conditionals,
since, if the proposition expressed by the antecedent is true in a world w ε C,
then f(w) = {w} and {w} Í . I will leave it at that for my present purposes,
though.9
,s
If (f(w) Ç [α] Ç [ß] = f(w)Ç [α], then all worlds in f(w) n [a] are ß-worlds, since for any sets
A, B, A n B = A just in case all elements of A are elements of B.
9
I refer the reader to Von Fintel (1998) for a discussion of the presupposition of subjunctive
conditionals, where is proposed the weaker condition that the domain of quantification might be
outside C.
Assume [α]Ç [ß] = {w23,w43}
Fig. 2: Updating the context with a subjunctive conditional
4.2 Indicative restrictors shrink the domain

As for indicative conditionals, I will assume the context change potential in
Heim (1992), depicted in Figure 3. Indicative conditionals perform the following
operation on the context: for each world w, they take from within the context set
the closest worlds to w in which the antecedent is true. If w is a world where the
antecedent is true, the set of closest worlds to w where the antecedent is true is
the singleton containing w itself. If w is a world where the antecedent is not true,
the operation selects the closest worlds to w from within  where the antecedent
is true. The following step consists on checking whether in all those worlds
closest to w where the antecedent is true, the consequent is also true. If that is the
case, w survives, if not, it is thrown away. This is shown in Figure 3:
Assume V ε C:max≤w(C[α])[ß]={w)
Fig. 3: Updating the context with an indicative antecedent
In symbols:
(13)<C,f> + ⌈si α [+subjunctive],ß⌉ =<C*, f>

Where C*-{w ε : 77max≤w (C Ç[α] ) Ç [ß] = max≤w (C Ç [α] ) }
I have not provided a compositional context change potential for indicative

conditionals for the sake of simplicity. It will suffice to assume that the context
change potential of an indicative conditional involves intersecting the context set
with the proposition expressed by the antecedent.10
Notice that this context change potential requires  to contain worlds where
the antecedent is true. For consider again the definition of max≤w
(7) For any proposition p ε  (W), any world w and any similarity
relation ≤w ,
max≤w (p) = {w': w' ε p & ∀w": w" G p ® w" ≤w w'}
If C n [a] is the empty set, the reader can verify that for any w, max≤w (
Ç [α] ) = Ø, and hence that the condition that, for any w, max≤w (CÇ[α])Í
[ß] would be trivially satisfied. Unless there are worlds where the antecedent is
true, the context change potential of indicative conditionals would be useless. We
then impose the following defmedness condition:
10
See Heim (1992:196) for an illustration of the fact that the similarity relation must apply to a set of
worlds in the context, a proposition that retains all the information in the context set along with the
information contributed by the antecedent. See Alonso-Ovalle (in preparation) for a different setup
providing an independent context change potential for the antecedents.
(13') <C,f> + ési α [+subjunctive]ù =<C, f*>

Where C*-{w ε C:max≤w (C [α] )  [ß] = max≤w (C  [α] ) }
Definedness condition: C Ç [α]¹ Ø
This defmedness condition will play a major role in the Caligula effect. In
Scenario 2, the context set contains no world where Caligula is dead.
Consequently, any antecedent interpreted as an indicative antecedent of the form
If Caligula is dead... will make the whole conditional undefined.
4.3 Moodless clauses and the default strategy: revisiting the Caligula effect
What about de-clauses? Let us assume that, in fact, mood marking in the
antecedent signals the way it changes the context. I have assumed that
subjunctive antecedents affect the modal horizon. Although I have not provided a
context change potential for the antecedent of indicative conditionals, I have
assumed that the context change potential of indicative conditionals involves
intersecting the context set with the proposition expressed by the antecedent. De-
clauses are moodless. They will then lack any instructions on how to change the
context. They could change it either as indicative antecedents do or as
subjunctive antecedents. This property allows us to derive the Caligula effect as
follows.
Recall Scenario 2. The doors of the arena are finally opened and everybody
can see Caligula alive, greeting the crowd. Recall the problem: in this context
(5b) is slightly odd when compared to (5a), both repeated below. Intuitions are
reported to be elusive, at least for some speakers.
(5) a. Si estuviera muerto (ahora), me tendrían que

if PRO were.PSTSBJ dead now PRO me would-have-to.3PL
ascender.
promote.INF
b. De estar muerto (ahora), me tendrían que
DE PRO be dead now PRO me would-have-to.3PL
ascender.
promote. INF
The contrast shows that the mood of the antecedent must be interpreted, for
it were not, then (5a) and (5b) should be equally fine. That gives us a hint to
answer latridou's question. In fact, if we asssume that the mood marking of
antecedents is not semantically vacuous, we can explain why the de-clause
makes a difference in the subjunctive conditional and not in the indicative

conditional in Scenario 2. Assume mood signals the way antecedents change the
context. According to my assumptions, the way indicative conditionals change
the context involves intersecting the context set with the antecedent. This, in its
turn, requires that there be worlds in the context set where the antecedent is true.
Subjunctive in the antecedent signals overtly that the antecedent will modify the
modal horizon. If de-clauses are moodless, then they will lack any instructions
on how to update the context. In principle, they could change it as indicative
antecedents and then they would require that the context set contained worlds
where the antecedent is true. They could also change it as subjunctive
antecedents and they would then modify the modal horizon. Either way will do.
The behavior of both types of conditionals under Scenario 1 attests to this fact.11
Stalnaker (1975) proposed a uniform semantic analysis for both indicative
and subjunctive conditionals. According to him, the difference between
indicative and subjunctive conditionals is a pragmatic one: the use of subjunctive
marks that the default strategy of taking the domain of quantification to remain
within  is suspended. I will assume with Stalnaker that the context change
strategy associated with indicative conditionals is some sort of default.
The fact that de-clauses can behave as either indicative or subjunctive
antecedents can lead to trouble if any of the two strategies does. The indicative
conditionals corresponding to the examples in (5) are ruled out in Scenario 2,
because they require that there be some worlds in  where the antecedent, the
proposition that Caligula is dead, be true. The way the scenario is set up
precludes this. Everybody sees that Caligula is alive. No world in  is a world
11
Though it does not make a difference in the Caligula scenarios, in general, de-clauses require that
the context set contain some worlds where the proposition they express is true and some where it is
false. For consider the following scenario.
Scenario 3
Jigl has just landed here. Driving past that beautiful garden, he sees some nice hydrangeas. He does
not know what the weather is like, but he says to Jogl:
(i) Si crecen aquí las hortensias, entonces los inviernos tienen que
if grow.3PL here the hydrangeas, then the winters have-to.3PL
ser suaves.
be mild
Unlike (i), (ii) is not appropriate in this context.
(ii) V.De crecer aquí las hortensias, entonces los inviernos tienen que
DE grow.INF here the hydrangeas, then the winters have-to.3PL
ser suaves.
be mild
"If hydrangeas (can) grow here, then winters must be mild."
This type of contrast illustrates that even when they are moodless, de-clauses impose definedness
conditions on their own. In Alonso-Ovalle (in preparation) I blame the complementizer for that.
where he is dead. If mood is already interpreted in the antecedent by checking

whether the definedness condition holds, and indicative marking corresponds to
the default strategy, we expect the de-clause to clash with the context associated
with Scenario 2, exactly as if it were an indicative antecedent. I have it that it
does.
There is, however, a crucial difference between indicative antecedents and
moodless ones. Moodless antecedents can also be interpreted as if they were
subjunctives. In fact, under our scenario, the moment the verb in the conditional
mood in the consequent kicks in, things change. The context change of the
consequent of a subjunctive conditional requires that the modal horizon reach
beyond the context set. If we processed the de-clause as if it were a subjunctive
antecedent, the requirement would be met. In fact, since the de-clause is
moodless, it can be uploaded to the context as if it were a subjunctive clause,
expanding the modal horizon beyond the context set and providing the
consequent with a suitable context. This solves the problem.
We can now see why intuitions are reported to be elusive. De-clauses can in
principle change the context in two ways. The default strategy leads to disaster.
The marked strategy associated with counterfactuals, however, solves the
problem.
5. To conclude
In response to Iatridou's problem, I have shown that moodless antecedents
interact with subjunctive modals in a way that suggests that mood marking in the
antecedent of conditionals is interpreted. The Caligula effect is explained if
mood marking signals the way in which the antecedent affects the context.
Usually, inflected antecedents contribute to the domain of quantification of the
modal just in the way the modal requires. Indicative antecedents feed indicative
modals. Subjunctive antecedents feed subjunctive modals. Mood marking in the
antecedent of conditionals can be interpreted while still being a phenomenon of
sortal agreement. Its effects are then generally masked. In order to see them, we
need antecedents that could in principle feed both indicative and subjunctive
modals. The moodless antecedents of de-conditionals provide us with exactly
this kind environment.
References
Abusch, Dorit. 1988. "Sequence of Tense, Intensionality and Scope". Proceedings of the
Seventh WCCFL, ed. by Hagit Borer, 1-14. Stanford, Calif: CSLI.
Alonso-Ovalle, Luis, (in preparation). "Mood Inflection and the Context Change Potential
of the Antecedents of Conditionals". Ms., University of Massachusetts at Amherst.
Bigelow, John  1976. "If-Then Meets the Possible Worlds". Philosophia 2.215-235.
Fintel, Kai von. 1998. "The Presupposition of Subjunctive Conditionals". The Interpretive
Tract. MIT Working Papers in Linguistics, ed. by Uli Sauerland and Orin Percus,
29-45. Cambridge, Mass.: MIT.
Fintel, Kai von. 2001. "Counterfactuals in a Dynamic Context". Ken Hale. A Life in
Language, ed. by Michael Kenstowicz, 123-153. Cambridge, Mass.: The MIT Press.
Heim, Irene. 1992. "Presupposition Projection and the Semantics of Attitude Verbs".
Journal of Semantics 9.183-221.
Iatridou. Sabine. 2000. "The Grammatical Ingredients of Counterfactuality". Linguistic
Inquiry 31.231-270.
Ippolito. Michela. 2001. "Presuppositions and Implicatures in Couterfactuals". Ms., MIT.
any, Charles E. 1936. "Conditions Expressed by Spanish 'De Plus Infinitive'".
Híspanla, XIX.211-216.
Kany, Charles E. 1939. "More about Conditions Expressed by Spanish. 'De Plus
Infinitive'". Hispania XXII. 165-170.
Kratzer, Angelika. 1979. "Conditional Necessity and Possibility". Semantics from
Different Points of View, ed. by R. Bäuerle, U. Egli and A. von Stechow, 117-147.
Berlin: Springer-Verlag.
Kratzer, Angelika. 1991. "Conditionals". Semantik: Ein internationales Handbuch der
zeitgenössischen Forschung, ed. by Arnim von Stechow & Dieter Wunderlich, 651-
656. Barlin/New York: Mouton.
Lewis, David. 1973. Counterfactuals. Oxford: Blackwell.
Lewis, David. 1981. "Ordering Semantics and Premise Semantics for Counterfactuals".
Journal of Philosophical Logic 10.217-234.
Montolío, Estrella. 1999. "Las construcciones condicionales". Gramática descriptiva de
la lengua española, ed. by Ignacio Bosque and Violeta Demonte, 3643-3739.
Madrid: Espasa-Calpe.
Nevins, Andrew Ira. 2002. "Counterfactuality without Past Tense". NELS 32, ed. by
Masako Hirotani. 441-451. Amherst, Mass: GLSA.
Nute, Donald. 1984. "Conditional Logic". Handbook of Philosophical Logic, vol. II, ed.
by D. Gabbay and F. Guenthner, 387-439. Dordrecht: Reidel.
Palmer, Frank R. 1986. Mood and Modality. Cambridge: Cambridge University Press.
Quer, Josep. 2001. "Interpreting Mood". Probus, 13.81-111.
Stalnaker, Robert. 1975. "Indicative Conditionals". Philosophia 5.269-286 (Reprinted in
Stalnaker, R.C. 2000, 63-78)
Stalnaker, Robert. 1984. Inquiry. Cambridge, Mass.:MIT Press.
Stalnaker, Robert  1998. "On the Representation of Context". Journal of Logic,
Language and Information 7.3-19. (Reprinted in Stalnaker, R.C. (2000), 96-115)
Stalnaker, Robert  2000. Context and Content. Oxford: Oxford University Press.
Söhrman, Ingmar. 1991. Las construcciones condicionales en castellano contemporáneo.
Studia Romanica Upsaliensia. Uppsala: Universitatis Upsaliensis.
Veltman, Frank. 2002. "A Dynamic Approach to Mood and Modality". Lecture notes for
the 1st North American Summer School in Logic, Language and Information.
Stanford, California.
Vlach, Frank. 1993. "Temporal Adverbials, Tenses and the Perfect". Linguistics and
Philosophy 16.231-283.
MOOD AND FOCUS
CLAUDIA BORGONOVO
Université Laval
0. The problem
As is well known, subjunctive in the Romance languages must appear in
embedded clauses when selected lexically by some predicates ('intensional
subjunctive', in the vocabulary of Stowell 1993), ?nd may appear when
embedded under a declarative or epistemic verb which is itself under the scope
of an operator ('polarity subjunctive', again according to Stowell). The latter
case is the one of interest here, and some examples appear in (1):
(1) a. No sabían que Pedro se hubiera-SUBJ había-INDde viaje.

"They didn't know that Pedro had gone on a trip."
b. Has oído que Pedro se hubiera-SUBJ / se ha-lND ido ido de
viaje?
"Have you heard that Pedro went on a trip?"
Declaratives like decir "to say", perception verbs like oír "to hear" and
epistemics like saber "to know" normally select indicative but admit the
subjunctive when negated or interrogated. The choice of indicative or
subjunctive has semantic consequences; the most widely accepted view states
that the truth of the embedded proposition is presupposed by the speaker when
indicative is used but it is not with subjunctive.
Surprisingly, when Polarity Subjunctive triggers are embedded under a
strong intensional predicate1 like want "querer", the choice with regard to mood
disappears and indicative is the only grammatical option:
I thank P. Ackema, M.Espanol-Echevarria, O. Fernandez-Soriano, L.McNally, J. Quer and M.

Uribe-Etxebarria for useful comments. I also thank the audiences at the Instituto Universitario Ortega
y Gasset, the Pompeu Fabra University, the University of Girona, the LEHIA workshop in Vitoria-
Gasteiz and Going Romance, where versions of this paper were presented, for comments. My deepest
gratitude goes to L Bosque (Bosque 1990 suggests some of the ideas developed here) and C.Sánchez
López for sharing their ideas and discussing this material with me.
18 CLAUDIA BORGONOVO
(2) La prensa quiere que la gente no crea que el ministro es-IND/*sea-

SUBJ culpable.
"The media want people not to believe that the minister is guilty."
Purpose clauses, also creators of intensional contexts, cause the same modal
effect (i.e., suspension of mood choice or SMC) when they embed a polarity
subjunctive trigger:
(3) Te lo dije para que no pensaras que Pedro era-IND /*fuera-SUBJ

culpable.
"I told you that so that you would not think that Peter was guilty."
In section 2, I show the full empirical range of the phenomenon of SMC, which
has gone largely unnoticed in the literature.2 These data raise several questions:
1. Why is subjunctive blocked in these examples? What is the connection
between the impossibility of subjunctive and the syntax-semantics of intensional
predicates?
2. SMC data have an obvious impact on the explanation of the phenomenon of
polarity subjunctive as a whole; how? What exactly is the import of mood in
contexts such as SMC and in regular polarity subjunctive contexts?3
1. The hypothesis
The traditional hypothesis concerning polarity subjunctive (Rivero 1972
and following literature) states that indicative signals presupposition of the
embedded proposition by the speaker, whereas subjunctive indicates no such
commitment. Given the types of verbs used, it seems legitimate to ask why the
grammar would give the speaker the option of manifesting himself in such a way
only when the matrix verb is negated; it clearly does not when the matrix verb is
affirmed. This observation points to a fundamental problem with the classical
presuppositional view, which attributes no role to negation. Yet the construction
and its concomitant interpretation do not exist without negation. I develop an
1
The class of strong intensional predicates comprises volitional verbs like want, directives like order
and some modals. They contribute a set of worlds to which the propositional content of their
complement is anchored (see Farkas 1992; Quer 1998).
2
Kleiman, in a 1978 unpublished dissertation, discusses the impossibility of subjunctive in counter-
factuals, from a different perspective than mine; I thank I. Bosque for bringing Kleiman (1978) to
my attention.
3
In the rest of the paper 1 will concentrate on polarity subjunctive triggered by negation; the analysis
can be extended to interrogative contexts, but the data that will be discussed (particularly SMC) are
impossible with interrogatives.
MOOD AND FOCUS 19
approach that skirts this problem.4 Furthermore, it is not clear how the purely
presuppositional approach would account for SMC cases (see Borgonovo 2002
for discussion of this point, that I will not repeat here for reasons of space).
The hypothesis I will defend here is a simple one: mood in Spanish signals
how negation is to be interpreted. Indicative marks the matrix predicate as focus
of negation, whereas subjunctive signals that the focus of negation is the
embedded clause. Other linguists have pointed out that there is a co-relation
between indicative and external negation, subjunctive and internal negation
(Horn 1978; Bosque 1980, 1990; Sánchez-López 1999, and references quoted
there). Quer 1998 expresses this co-relation in terms of thetic and categorical
negation. I will do things a bit differently: 1. the co-relation between mood and
the interpretation of negation becomes the basis for the explanation of mood
choice and 2. what mood marks is two possible foci. I will show that many of the
good results of previous work can be derived from this minimally simple
hypothesis, and the latter will be supported with new data.
I will clarify what the relevant readings are that mood gives rise to in
Spanish. Let's take the following minimal pair:
(4) a. Juan no vio que Pedro s alier a-SUBJ.

b. Juan no vio que Pedro salía-IND.
"Juan didn't see that Pedro left."
Sentence (a) says that there is a relevant event of seeing (taking event to
comprise both events and states) which took place in the past and of which Juan
is the experiencer, and whose percept is not the event denoted by Pedro left.5
That is, Juan saw something (focal presupposition), but not the event denoted by
the embedded clause. Sentence (b), on the other hand, says that there was a non-
seeing event of which Juan is the experiencer and whose percept is the event
denoted by Pedro left. The focal presupposition of the latter sentence is the non-
focused embedded CP; as a result, the interpretation of (b) is that there was an
event of leaving, and of that event we are told that Juan did not see it.
There are pragmatic differences between the two utterances: neither can be
felicitously used in an out of the blue context (as tends to be the case with
negative statements in general; see Givon 1977, but also discussion in Horn
2001). After all, there are infinitely many events in the world that Juan has not
witnessed and speakers tend not to introduce them in conversation, given
4
Quer (1998, 2000) develops a presuppositional account that is free of this problem; I cannot discuss
it here for lack of space.
5
I ignore in this paper the so called external reading negation, paraphrasable with "it is not the case
that...", which also obtains with subjunctive.
Gricean maxims. But the contexts that make (4a) and (b) felicitous are different
in each case. In the case of (a), what might be discussed are events witnessed by
Juan; in the second case, there is a presupposed event, that of Pedro's leaving,
and of that event it is asserted that Juan did not see it.
About the organization of this paper: in section 2 I establish the empirical
range of SMC, the contexts in which a negated verb does not license subjunctive
in the lower clause. SMC is crucial because it illustrates the connection between
focus and mood in a particularly perspicuous way: if the embedding verbs in
CP2 (the intermediate CP between the strong intensional verb and the verb in the
indicative) include the typical Neg-raising verbs, e.g., epistemics, Neg-raising is
not possible. Neg-raising is extremely useful because it undisputedly shows how
negation should be interpreted: the negative operator, though it appears in the
matrix predicate, is interpreted and behaves syntactically as if it were in the
lower clause. Thus, Neg-raising provides the clearest evidence of what is the
focus of negation in the sentences where it obtains. It will be shown in section 3
that negation in SMC contexts must obligatorily be interpreted upstairs, and this
is exactly the reason why subjunctive is ungrammatical; the correlation between
indicative and higher negation is not a coincidence. Additional evidence in
support of the central claim is discussed in section 4, evidence that centers on 1.
the interaction of the particle sino, a contrastive focus marker, and mood; 2. the
connection between mood and focus in rationale clauses and 3. the impossibility
of simultaneously licensing a subjunctive clause and an NPI as arguments of the
same predicate. In section 5 I turn to non Neg-raising predicates and discuss the
corresponding focal evidence. Section 6 contains the conclusions.
2. Empirical range of SMC

In this section I will explore the contexts in which SMC obtains; to do so, I
will look at a variety of predicates and contexts to determine their effect on
mood.
2.1 Embedding under epistemic, declarative, perception andfiction predicates

There is no SMC effect.
(5) a. Marta piensa / dice / oye / sospecha / sueña que Pedro no cree
que se lo merecen-IND / se lo merezcan-SUBJ.
"Marta thinks, says, hears, suspects, dreams that Peter doesn't
believe that they deserve it."
MOOD AND FOCUS 21
2.2 Embedding under strong intensionalpredicates

The class of strong intensional predicates (Farkas 1992) comprises
volitional verbs like want (2 above), directives like order, some modals (6) and
purpose clauses (see 3 above, and also Quer 1997, ch. II, Fn 30). The effect is
robust with this type of predicates:
(6) Es necesario que no se crea que aquí hay-IND / *hay-SUBJ

problemas.
"It is necessary that it be not believed that there are problems here."
2.3 Embedding under Polarity Subjunctive trigger

There is no SMC effect:
(7) Marta no cree que él no piensa-IND / piense-SUBJ que Juan es -IND/

sea SUBJ culpable.
"Marta doesn't believe he doesn't think that Juan is guilty."
2.4 Embedding under factives

There is no SMC with factives:
(8) a. Lamento que no creas que él es-IND / sea-SUBJ culpable.
"I regret your not believing him to be guilty."
b. No me di cuenta de que no creías que él es-IND / fuera-SUBJ
culpable.
"I didn't realize that you didn't believe him to be guilty."
2.5 Embedding under adjunct clauses (concessive, temporal, because and

conditional clauses)
(9) a. Te ha enviado los documentos aunque no creas-SUBJ / crees-

IND que él es-IND / sea-SUBJ culpable.
"He sent you the documents even though you don't believe him
to be guilty."
b. Te lo dijo cuando ya no creías que él era-JND / fuera-SUBJ
culpable.
"He told it to you when you did no longer believed that he was
guilty."
 Te lo dijo porque no crees que él es-lND / sea-SUBJ culpable.
"He told it to you because you don't believe that he is guilty."
Polarity subjunctive triggers in concessive - (9a) -, temporal - (9b) - and because

- (9c) - clauses allow both moods. When the negated epistemic appears in the
protasis or the apodosis of a counterfactual, SMC obtains, but not with a realis
clause, as in (10a):
(10) a. Si no crees que Pedro es-IND / sea-SUBJ culpable, debes hablar

con el comité.
"If you don't believe Pedro to be guilty, you should talk to the
committee."
b. Si no creyeras ya que Pedro es-IND / *fuera-SUBJ culpable,
deberías hablar con el comité.
"If you didn't already believe Pedro to be guilty, you should talk
to the committee."
 No creería que Pedro es-IND / *fuera-SUBJ culpable si no
hubiera visto los documentos.
"I wouldn't believe Pedro to be guilty had I not seen the
documents."
2.6 Polarity subjunctive triggers as matrix predicates: imperatives
(11) No digas que Pedro es-IND / *sea-SUBJ culpable.y

"Don't say that Pedro is / is culpable."
The generalization is clear: the ungrammaticality of subjunctive embedded

under a Polarity subjunctive trigger is strong and without exceptions when the
matrix predicate is strong intensional, when it appears in the protasis or apodosis
of a counterfactual or when the trigger is in the imperative, and non-existent
everywhere else.
3. Neg-raising and SMC

Descriptively, there is Neg raising when the negative operator appears in an
embedding clause but is interpreted and behaves syntactically as if it were in the
embedded clause. The classical evidence for Neg-raising includes long distance
licensing of NPIs, pronominalization and sluicing (see Prince 1976; Bosque
1980; Horn 1978, 2001). I evaluate some of this evidence in connection with
SMC cases to show that Neg-raising does not obtain. The section ends with an
explanation of why this is necessarily the case. I first include the evidence with
6
Some traditional grammars of Spanish (see Gili Gaya 1966 and others) point out that subjunctive is
ungrammatical when embedded under a negated imperative.
MOOD AND FOCUS 23
regards to normal polarity subjunctive contexts, and afterwards I discuss it in

reference to SMC contexts/
a: NPI licensing :
(12) a. No creemos que haya-SUBJ movido un dedo en todo el tiempo

que estuvo empleado aquí.
"We do not think (s)he has moved a finger (=worked) in the time
she was employed here."
NPIs are licensed in CP2 (the first embedded clause), even though negation
appears on CP1, the matrix. Let's see how this evidence fares when the negated
epistemics appear in SMC contexts:
(13) * Marta quiere que no creamos que Pedro ha movido-IND un dedo en

todo el tiempo que estuvo empleado aquí.
With strong NPIs (such as NPI idioms) the sentences behave as expected when
negation is in the upstairs clause.
b: NPI Temporal adjuncts:

A temporal adjunct like hasta las 3 ("until three") is acceptable with durative
predicates, but not with telic ones (Karttunen 1977; Bosque 1980).
(14) a. Juan se quedó hasta las 3 / *Juan llegó hasta las 3.

"Juan stayed until 3 / *Juan arrived until 3."
When negated, telic predicates accept the until phrase:
b. Juan no llegó hasta las 3.

"Juan did not arrive until 3."
As expected, in Neg-raising contexts an embedded telic verb is OK with an until

phrase:
 Juan no cree que María llegue hasta las 3.
"Juan does not believe that Maria will arrive until 3."
7
Neg-raising verbs comprise verbs of opinion and expectation, intention and volition and perceptive
approximation (Horn 1978). In Spanish, all these classes lexically select subjunctive, except for the
verbs of opinion and perceptive approximation, which are used here.
Until phrases are not licensed in SMC contexts:
(15) *Si no creyeras-IND que Marta llego-SUBJ hasta las 3, ...

"if you did not believe that Maria arrived until 3..."
The evidence shows that negation in SMC cases is interpreted in the higher
clause: what is being negated here is not the content of the subordinate clause but
the predicate that embeds it. The question is why.
The particular configuration of predicates that trigger SMC has to be
partially or totally responsible for the effect; any answer as to why Neg-raising,
and subjunctive mood under the hypothesis developed here, are impossible in
these examples, has to take into account the predicates that embed the negated
epistemic verbs. Given that it is strong intensional predicates, counterfactuals
and imperatives that trigger SMC, it is necessary to find the trait that unites them,
i.e., we need to determine what sort of natural class they form.
Imperatives, counterfactuals and strong intensional predicates introduce a
set of possible worlds in which the truth of the embedded proposition is
evaluated. The worlds introduced do not comprise the world that models the
actual one; to put it differently, the worlds introduced are not compatible with the
world that models reality according to the subject. As a result, there is a
discrepancy between what obtains in the possible worlds and what obtains in the
actual one (or in the world that models it).8 Let's see how this applies to our
examples: if if x were to believe that p is uttered, it follows that, according to the
speaker's espistemic model, x does not believe that p in the actual world. When
the speaker says: If x were not believe that p, then he thinks that x actually
believes that p. In the latter case, e.g., as a result of this difference between the
world that models reality (in which x does believe) and the worlds introduced by
counterfactuals (in which x does not believe), the stative epistemics are
interpreted as inchoative: if x were not to believe that p means "if x were to stop
holding a certain belief'; if x were to believe that p means "if were to start
holding a certain belief'. This shift has semantic consequences: in Neg-raising
8
The same observation applies to imperatives: orders are given to change a present state of affairs.
With regards to strong intensional verbs, there is at least a pragmatic implicature involved about the
state of affairs that obtains in the actual world: if I say I want you not to believe that p, it is implied
that I think that you believe that p or that you hold no belief with regards to p, but see Giannakidou
(1998), where volitionals are analysed as non-veridical. I thank an anonymous reviewer for pointing
out Giannakidou's analysis of volitional verbs.
MOOD AND FOCUS 25
contexts, the proposition with lowered Neg entails the proposition with raised
Neg 9; e.g., b. entails a. in (16):
(16) a. Juan no cree que Marta sea-SUBJ culpable.

"Juan does not believe that Marta is guilty."
b. Juan cree que Marta no es culpable.
Entailment only holds when the epistemic is stative but not when it is inchoative:
if x were to stop holding a certain belief p does not entail if x were to start
holding the belief that - p. These two propositions are distinct, as lack of
entailment shows.
If there were Neg-raising in these examples, the difference between these
two propositions would be blurred. Horn (1978, 1989) claims that Neg-raising is
part of a general tendency in language to place negation as soon in the sentence
as is compatible with avoidance of communicative chaos: there is no Neg-raising
in our examples precisely because a crucial distinction between two propositions
would be canceled, thus crossing the border into communicative chaos. Negation
has to be interpreted where it appears.
Summarizing, there is no subjunctive in SMC because in these contexts
negation has to be interpreted upstairs, as the failure of the Neg-raising tests
shows.
4. Mood marks focus: more evidence

Lack of Neg-raising in SMC contexts has provided the first piece of
evidence for the hypothesis that mood marks how negation should be interpreted
in polar contexts. Let's now turn to other sources of evidence.
4.1 Evidence from sino

The particle sino in Spanish is used to correct the constituent in the
sentence that falls under the scope of negation; it thus requires negation and it is
used in the standard test for contrastive focus.
(17) a. No es inteligente*pero / sino loco.

"(S)he is not intelligent but crazy."
It can also be used to contrast sentences in indicative and in subjunctive:
9
Actually Partee (1973) claims that Neg-raising reduces to this entailment relation, and that there is
no Neg movement transformation involved. It is clear that there is no transformation, but the 'just
entailment' analysis cannot explain the syntactic evidence discussed in this section.
b. No quiere que te vayas-SUBJ sino que te quedesSUBJ

"(S)he does not want you to go but to stay."
When a subjunctive polarity clause is contrastively focused, subjunctive is

ungrammatical after sino:
(18) No dijo que se hubiera-SUBJ ido sino que se *hubiera-SUBJ escondido

/ había-IND escondido.
"(S)he did not say that (s)he had left but that (s)he was hiding-SUBJ,
was hiding-IND."
The data provided by sino support the hypothesis about the connection between
mood and scope: sino has to be followed by indicative even when the contrasted
sentence is inflected for subjunctive because the CP the conjunction introduces
must be outside the scope of negation.
4.2 Negation and because-clauses

It is well known that because- (and purpose) clauses can be either within
the scope of sentential negation or outside it. Purpose clauses in Spanish are
always inflected in the subjunctive, so they cannot be offered as evidence for the
focal claim made here, but rationale clauses are always in indicative except when
they are within the scope of negation:
(19) No te lo dijo porque quisiera-SUBJ / *quería-IND ofenderte, sino

porque pensó que no te importaría.
"He didn't tell you because he wanted to offend you but because he
thought you wouldn't mind."
(19) offers additional evidence for the claim that subjunctive marks focus.
4.3 NPIs, absorption, focus and a combination thereof

Bosque (1994) shows that a single negation cannot license a negative
polarity item (NPI) and a Neg-focused constituent at once. It is well known that
there is an absorption effect in Spanish that allows a single negative operator to
license more than one NPI with a (single) negative interpretation (a). Neg can
also license one focused constituent (b), but there seems to be a minimality
effect when the licensees are of different kinds, e.g., an NPI and a focused
constituent (c). I offer no analysis here for why this should be the case due to
lack of space; I merely capitalize on Bosque's observation:
MOOD AND FOCUS 27
(20) a. No te lo ha dado nadie nunca en ningún lugar.

"Nobody has ever given you anything anywhere."
b. No le he dado un libro a Juan, sino a Pedro.
"I did not give a book to Juan, but to Pedro."
c. *No le he dado nada a Juan, sino a Pedro.
"I did not give anything to Juan, but to Pedro."
The hypothesis that subjunctive marks the focus of negation makes a clear
prediction: a polarity subjunctive clause cannot co-occur with a co-argument
NPI. The prediction holds when the NPI is stressed (21a) or focalized (21b):
(21) a. *?No le dijimos a NADIE que estuviéramos-SUBJ aquí.

"We did not tell ANYONE that we were here."
b. A nadie (le) dijimos que estuviéramos-SUBJ aquí.
The same effect holds with focused rationale clauses in subjunctive, both with N-
words (b) and idiomatic NPIs:
(22) *No dije eso a nadie porque estuviera-SUBJ cansado (sino porque...)
"I did not tell that to anyone because I was tired (but because...)"
When the NPIs in (21) are pronounced with a neutral intonation, the sentence
improves considerably:
(23) (?)No le hemos dicho a nadie que estuviéramos aquí.
(23) suggests that a distinction needs to be made between simple negative

concord and focused NPIs. I will not develop an analysis of the difference here,
but I point out that Ladusaw (2000) makes a similar distinction between two
approaches to negative concord, one in which negative concord is essentially
quantificational and the terms represent the restriction of the operator, and a non-
quantificational account in which the terms are weakly construed descriptions
(Ladusaw 2000:240). I suggest that what we find in examples such as (23) is an
instance of the latter, but I will not develop the issue here.
4.4 Contrastive focus data

The data I will look at next include negated verbs of perception. They
provide the clearest evidence in the case of contrastive focus given a semantic
peculiarity of the conjunction sino: this conjunction is the contrastive focus
marker 'par excellence', but it also signals mutual exclusion of the two
contrasted constituents. I can say esto no es azul sino blanco ("this is not blue
SINO white") because blue and white exclude one another. In the case of
epistemic verbs, contrast with sino is harder to get because typically these verbs
do not exclude one another: e.g., think and suspect, guess and think are not
mutually incompatible. So a sentence like the following is odd, not because it is a
counterexample to the focal claims made here, but because the two verbs
contrasted with sino are mutually compatible:
(24) #Juan no cree que Pedro es-IND / sea-SUBJ culpable sino que lo
sospecha.
"Juan does not believe that P. is guilty, SINO he suspects it."
The following are clearer data, given that in principle there can be an
incompatibility between hearing and seeing an event; sino is far more acceptable
in the context of verbs of perception:
(25) a. ??No vio que Pedro saliera-SUBJ sino que lo oyó.

b. No vio que Pedro salía-INDsino que lo oyó.
"(S)he didn't see Pedro's leaving but (s)he heard it."
The oddness of (25a) is explained because negation takes as its focus the
embedded clause, but the corrective constituent targets the matrix. Indicative, on
the other hand, is completely acceptable.10
5. How the focal hypothesis derives the good results of other hypotheses
This section sketches how good results from previous work can be
recuperated under the present approach.
#1: the presuppositional effects obtain, as far as they do, because with indicative
the subordinate clause is outside the scope of negation. As a consequence, the
clause in question is interpreted as part of the presuppositional, not the focal,
component of the sentence. Presupposition, then, is not what indicative marks
per se; it follows from the interpretation mood forces on negation and its ensuing
division of the sentence between focus and presupposition.
#2: Indicative marks that the focus of negation is the matrix verb; it follows that
a sentence in which the matrix verb is a belief verb which is negated will
produce a strong assertion of dis-belief, hence the polemic flavour normally
attibuted to sentences such as (26):
10
An anonymous reviewer suggests that indicative is compatible with other matrix constituents being
under the focus of negation, and not only the matrix verb. I think this is true, but I cannot develop this
here due to lack of space; the hypothesis can be changed in the suggested way, while maintaining the
basis gist of the focal proposal.
MOOD AND FOCUS 29
(26) No creo que Marta va-IND a Londres.

"I don't believe that Marta is going to London."
#3: Verbs like think or believe express a mid-scalar degree of certainty. When the
negative operator acts on the clause these verbs embed, the semantics of the
subordinator plus the focalizing effect of the negative operator conspire to
produce a sentence which naturally lends itself to the assertion softening, or
hedging, such as the pragmatic approaches to Neg-raising propose (see Prince
1977, for example).
6. Conclusions
In this paper I have presented an analysis of modal choice in Spanish in
terms of negation and its focus. The analysis is maximally simple, supported by
ample evidence and it has the advantage of incorporating insights from previous
work. I have not given, evidently, a comprehensive analysis of mood; the
analysis as it stands does not carry to intensional, or lexical, subjunctive, and
does not carry unmodified onto other optional-subjunctive contexts such as
relatives. At first glance, this paper argues against such a unified analysis of
subjunctive. The analysis hinges crucially on the presence of a negative operator,
and it cannot be convincingly argued that all subjunctive contexts are negative.
Subjunctive morphology would not be the first instance in which a piece of
morphology does multiple work in a language, a glaring example of multiple
task morphology being the values -ed or -ing in English. The hypothesis and the
results presented here, then, offer an argument for a multi-valued analysis of the
subjunctive morphology. This view is in principle supported by the non
existence of polarity subjunctive in languages that use the subjunctive
morphology productively, such as the Balkan and Slavic languages. If the
subjunctive has different interpretive roles at the syntax-semantics interface, we
would expect that languages would differ in precisely the way they do: some
semantic distinctions would be uniformly marked by mood, whereas variation
would be evidenced in others.
References
Borgonovo,  2002. "Mood and focus". Cuadernos de linguistica IX. Instituto Ortega y
Gasset, Madrid. 42-63
Bosque, I. 1980. Sobre la negacion. Madrid: Cátedra.
Bosque, I. 1999. "Negación y el principio de las categorías vacías". Gramática del
español, ed. by V. Demonte, 167-199. México: Colegio de México.
Bosque, I. 1990. "Las bases gramaticales de la alternancia modal". Bosque 1990. 13-66.
Bosque, I., ed. 1990. Indicativo y subjuntivo, Madrid: Cátedra.
Bosque, I. & V. Demonte, eds. 1999. Gramática descriptiva de la lengua española.

Madrid: Gredos.
Cole, P., ed. 1978. Pragmatics. Syntax and Semantics 9. New York: Academic Press.
Farkas, D. 1992. "On the semantics of subjunctive complements". Romance languages
and modern linguistics theory, ed. by P. Hirschbuhler and K. Koerner, 69-104.
Amsterdam: John Benjamins..
Giannakidou, A. 1998. Polarity sensitivity in non-veridical contexts. Amsterdam: John
Benjamins.
Gili Gaya, S. 1961. Curso superior de sintaxis española. Barcelona : Spes.
Horn, L. 1978. "Remarks on Neg-raising". Cole 1978. 101-176.
Horn, L. 2001. A natural history of negation. Chicago: University of Chicago Press. CSLI
Publications.
Horn, L. & Y. Kato, eds. 2000. Negation and Polarity. Oxford: Oxford University Press.
Giannakidou, A. 1998. Polarity sensitivity as (non) veridical dependency. Amsterdam:
John Benjamins.
Givón, T. 1978. "Negation and presupposition". Cole 1978. 69-112.
Kleiman, A. 1974. A syntactic correlate of semantic and pragmatic relations of the
subjunctive mood in Spanish. Unpublished Ph.D. dissertation, University of Illinois.
Karttunen, L. (1977). "Until". Proceedings of the Chicago Linguistic Society10.284-297.
Ladusaw, W. (2000). "Thetic and categorical, stage and individual, weak and strong".
Horn & Kato 2000. 66-101.
Partee, . (1973). "The semantics of Belief-sentences. Approaches to natural languages,
ed. by J. Hintikka, E. Moravcsic and E. Suppes, 309-336. Dordrecht: Reidel.
Progovac, L. 1993. "The (mis)behaviour of anaphora and negative polarity". The
L ingu is tic Review 10.101-143.
Progovac, L. 2000. "Coordination, C-Command, and 'Logophoric' N-words". Negation
and Polarity, ed. by L. Horn and Y Kato, 102-145. Oxford: Oxford University
Press.
Quer, J. 1998. Mood at the interface. The Hague: Holland Academic Graphics.
Quer, J. 2001. "Interpreting mood". Probus 73.81-111.
Rivero, M.L. 1971. "Mood and presupposition in Spanish", Foundations of language
7.305-336.
NULL VS OVERT SPEC,TP IN EUROPEAN PORTUGUESE
JOÁO COSTA
Universidade Nova de Lisboa
1. Goals
The goal of this paper is to discuss the status of subjects in functional A-
positions in European Portuguese (EP), trying to answer the following question:
(i) What type of conditions make it possible for an A-position to be used

as a landing site for the subject?
Before answering this question, an issue must be addressed concerning recent

hypotheses about the null subject parameter (Barbosa 1995, Alexiadou &
Anagnostopoulou 1996):
(i)a Is it possible to claim that the only A-position for subjects in null
subject languages is Spec,VP?
It is important to provide a negative answer for question (i)a, since if it turned out
to be the case that the only A-position for subjects was their base-position, the
first question would never arise.
This paper also focuses on the availability of a [+/-Spec,TP parameter]
(Bobaljik & Jonas 1996), and its correlation with the functional structure of the
clause, trying to answer the following question:
(ii) What type of phenomenon may shed light on this issue (given the
unavailability of Transitive Expletive Constructions in Romance)?1
In previous work, I have discussed the status of the highest A-position

(Spec,AgrSP), arguing that it is a legitimate landing site for subjects for syntactic
1
For reasons not to consider VSO sentences in European Portuguese instances of Transitive
Expletive Constructions, see Costa (1998) and Coelho et alii (2001).
32 JOAO COSTA
reasons, and that the usage of this position may be constrained by the discourse-
syntax interface. In this paper, I will strengthen the argument that there are
preverbal A-positions in EP, focussing on Spec,TP. The points to be made are the
following:
a) The availability of Spec,TP as a landing site for the subject is

dependent on the existence of I-to-C movement."
b) The dependency between head movement and the availability of
Spec,TP may be understood as a consequence of the syntax-
morphology interface.
2. On the non-binarity of the null subject parameter

Recent proposals concerning the status of the null subject parameter
(Barbosa 1995, Alexiadou & Anagnostopoulou 1996, among others) have put
forward the claim that null subject languages are languages in which Spec,IP is
not projected as an A-position; the EPP-feature is checked by pronominal Agr;
the only A-position for subjects is Spec,VP; and, preverbal subjects are left-
dislocated.
This type of approach makes the prediction that the null subject parameter is
binary. Since Spec,IP is not projected and the burden is put on EPP, rather than
other properties of Spec,IP, the difference between different types of pro (Rizzi
1982, 1986) is lost. In other words, Rizzi's proposal that licensing conditions for
expletive pro are not contingent on the availability of a pronominal Agr, which
would be relevant for referential pro, is hard to recapture, since expletive pro is
related with EPP, and so is pronominal Agr, according to the proposals referred to
above.
In previous work (Costa 1998, 2001), I have presented several counter
arguments against adopting Barbosa's and Alexiadou & Anagnostopoulou's
analysis in European Portuguese. Those arguments are listed below:
SV(O) order in sentence-focus contexts;
Assuming that preverbal subjects are left-dislocated, there is no clear explanation

as for why focused subjects in sentence-focus contexts are good, since this is a
context in which left-dislocation is out.
2
I will make the distinction between AgrS and T when necessary. Elsewhere, I will refer to the set of
these two functional heads as I.
NULL VS OVERT SPEC,TP IN EUROPEAN PORTUGUESE 33
Obligatory agreement in preverbal position in unaccusative contexts;
In colloquial EP, the plural argument of an unaccusative verb may not agree with
the verb in postverbal position. In preverbal position, full agreement is
obligatory. Since left-dislocation does not trigger agreement, this behavior is
unexpected.
Lack of minimality effects in embedded wh-questions;
In some dialects of EP, embedded wh-phrases cannot cooccur with a left-

dislocated constituent, which can be understood in terms of relativized
minimality. In such dialects, a preverbal subject may follow a wh-phrase without
yielding ungrammatical results.
Lack of definiteness effects in Spec, IP;
Non-specific indefinites, which resist left-dislocation, may occur in preverbal

position. This is unexpected, if preverbal subjects were left-dislocated.
The contexts that legitimate clitic left-dislocation of non-subjects do

not match the contexts that legitimate preverbal subjects.
If it were the case that preverbal subjects are left-dislocated, there should be a
perfect match between the contexts that legitimate preverbal subjects and clitic
left-dislocation of non-subjects. This is not the case.
In this paper, I would like to add some additional evidence against the idea
that null subject languages are languages that check EPP by means of pronominal
Agr. The new evidence comes from three different domains.
First, there are languages which are only semi-pro-drop. This is the case of
Brazilian Portuguese (Coelho, Costa, Figueiredo Silva & Menuzzi 2001) and
-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:
(1) a. Chegou  Pedro. BP

arrived Pedro
"Pedro arrived."
b. *Viajou  Pedro. BP
traveled Pedro
34 JOAO COSTA
. *pro viajou. BP
traveled
(2) Está chovendo. BP
is raining
"It is raining."
(3) a. Txiga tres pesoa. CVC
arrived three persons
"Three persons arrived."
b. *Papia tres pesoa. VC
talk three persons
 *pro papia CVC
talks
(4) Txobi. CVC
rains
"It rains."
Inversion is possible in BP and CVC, in contexts in which Spec,IP is occupied 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
interaction between different constraints, as suggested by Rizzi (1982) and
Coelho et alii (2001), among others.3 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 referential and
expletive null subjects. A language in which Spec,IP is projected, but 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 availability of pro.
The second domain in which counter-evidence for the claim that preverbal
subjects are left-dislocated is found 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. A study of the acquisition of subjects in a child in his second year of
As noted by a reviewer, 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.
life (Adragäo 2001) reveals that inversion is highly marked and rare in the child's
early productions:
(5) % 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 important to note that, in her study, Adragäo
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 & Matos 2000).
These data provide additional evidence for the unmarked status of the
preverbal position for subjects. The correlation between inversion and
unaccusativity make it impossible to claim that children do not know whether
Spec,IP is projected.
Finally, the existence of inflected gerunds in Dialectal European Portuguese,
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 dialectal 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 (6):
(6) Dialectal EP:

a. Estandes ec cansado, tu, podes ir.
being tired, you may go.
"Since you are tired, you may go."
b. Estandes tu cansado, eu posso ir
being-2sg you tired, I may go
"Since you are tired, I may go."
 *Estandes ec, cansado, eu posso ir.
being-2sg tired, I may go
(6') Standard EP:

a. Estando ec, cansado, tu, podes ir
being-2sg tired, you may go
"Since you are tired, you may go."
36 JOAO COSTA
b. Estando tu cansado, eu posso ir

being you tired, I may go
"Since you are tired, I may go."
c. * Estando ec, cansado, eu, posso ir
being tired, I may go
The ungrammaticality of (6c) and (6'c), and the lack of difference between the
two dialects casts some doubt on the idea that Agr is pronominal in European
Portuguese.
The counterarguments presented in the previous work and this new evidence
allow for saying that Spec,IP is available in European Portuguese. This raises the
following question:
If Spec,IP is actually Spec,TP and Spec,AgrSP, can all these positions

be used as a landing site for the subject? Under what conditions?
In Costa (1998), I proposed that discourse may constrain the use of Spec,AgrSP
It was shown that focused subjects remain in VP, except for the case of sentence-
focus contexts.
In the next sections, I will address the issue of whether and when Spec,TP
may be used as a landing site for the subject.
3. The Puzzle: availability of Spec, TP as a landing site for the subject

In Costa (1996) and Costa & Galves (2000), it is argued that subjects in
European Portuguese move to Spec,AgrSP, while the verb undergoes short-V-
movement from V-to-T.4 This analysis derives the non-adjacency between subject
and verb in sentences like (7):
(7) a. Ninguém provavelmente leu b  livro.

no-one probably read well the book
The fact that the verb appears in between two adverbs in (7a,b) shows that it is
not the case that there is no V-movement at all in European Portuguese. In the
works cited above, it is claimed that subject-oriented adverbs must be TP-
4
As correctly pointed out by an anonymous reviewer, on its way to Spec,AgrSP. the subject goes
through Spec,TP either for locality reasons or for EPP-checking. The point to be made throughout
the paper is that the subject may not surface in Spec,TP in declarative sentences. This makes the
presence of a trace in this position irrelevant for the purposes of this paper.
adjuncts, since they only appear in the position in between the subject and the
verb. All other positions for the adverb in (8) yield a manner reading:
(8) .  João estupidamente entornou o café. Subj-Or.

Joäo stupidly spilled the coffee
b. O Joäo entornou estupidamente o café. Manner/* Subj-Or.
Joäo spilled stupidly the coffee.
c
c.  Joäo entornou o café estupidamente. Manner/* Subj-Or.
João spilled the coffee stupidly.
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.5 The unavailability of Spec,TP is attested in example (9), 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 subject-oriented adverb and the
verb is ungrammatical:6
(9) .  João ... ele estupidamente entornou o café.
João ....he stupidly spilled the coffee
b. *  Joäo ... estupidamente ele entornou o café.
Joäo.... 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:7
(10) .  que tinha estupidamente o Joäo entornado?

what had stupidly Joäo spilled
b. Quando tinha cautelosamente o Joäo lido o livro?
when had carefully Joäo read the book
5
Actually, Costa & Duarte (2002) present an analysis of inversion with presentational focus
involving this type of configuration. It is however compatible with the claims put forward in this
paper.
6
(9b) is only legitimate with a topic intonation for the adverb.
7
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.
38 JOAO COSTA
There is thus an apparent contradiction: while the data in (9) show that Spec,TP
is not an available position for the subject, the data in (10) show that Spec,TP is
an available position for the subject. This puzzle becomes more evident in (11):
(11) a.  João estupidamente tinha já entornado o café.

João stupidly had already spilled the coffee
b.  João ... (ele) estupidamente (*ele) tinha (*ele) já (??ele)
Joäo ...(he) stupidly (he) had (he) already (he)
entornado (ele) o café.
spilled (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 underlined pronouns in (llb-c). (1 lb)
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 (11c), however, the position in between the two adverbs is
an available position.
The analysis outlined above for subject non-adjacency and the puzzle
regarding Spec,TP raise at least the following two questions:
(12) 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
contexts only?
3.1 Background
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 Morphology
(Halle & Marantz 1993, Bobaljik 1995), affixation takes place in the
Morphological component of the grammar. The fusion of heads is possible 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 (13):
NULL VS OVERT SPEC, TP IN EUROPEAN PORTUGUESE 39
As defended by Bobaljik, this type of object has consequences for morphological

fusion, since there must be two independent morphemes for T and Agr. This is
because in a first step 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 morphemes may cooccur.
The second type of scenario discussed by Bobaljik (1995) is the one in (14):
According to Bobaljik, this is the type of head created if T has weak N-features,
not being able to attract AgrO. Under such circumstances, AgrS attracts T, and
the complex V+AgrO. Unlike in (13), 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 (14) 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 & 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.
40 JOAO COSTA
Let us consider some examples discussed by Bobaljik: Icelandic and

English contrast in that the latter only provides evidence for Agr or T
morphology but not for both, while in the former Agr and T morphemes cooccur:
Icelandic: kasta "throw" English: tremble

Present Past Present Past
kasta kasta-ði tremble tremble-d
kasta-r kasta-ði-r tremble tremble-d
kasta-r kasta-ði trembles tremble-d
köst-um köstu-ð- tremble tremble-d

kast-ið köstu-ðu-ð tremble tremble-d
kasta köstu-ðu tremble tremble-d
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-1 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 distribution,

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
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 (15) is that any lexical material, except for
adverbs, block the adjacency requirement, and force do-insertion.8
8
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.
42 JOAO COSTA
(15) a. [IPSubj I [ vp V I and V are adjacent,

morphological merger is possible
b. [IP Subj I [ vp Adv [VP V I and V are adjacent,
morphological merger is possible
 [IP Subj I [NegP not [ vp V I and V are not adjacent,
morphological merger is not possible (do-insertion)
d. [CP wh I+C [IP Subj t1 [ vp V I and V are not adjacent,
morphological merger is not possible (do-insertion)
32 Back to the puzzle on Spec, TP in EP
With the background given above, we now have the necessary tools to
address the problems regarding the availability of Spec,TP in EP.
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 context 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 adjacency. 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 illustrated in (16a). If the
subject would stay in Spec,TP like in (16b), there would be no adjacency and
morphological merger would be impossible:
(16) a. [AgrSP Subj AgrS [ T P (Adv)T+V [ VP t v

I and V are adjacent, morphological merger is possible
b. [AgrSP AgrS [TPSuj T+V [VPtv
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 between 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 possible to
distinguish independent T and Agr morphemes in the past tense: 9
9
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 morphology
NULL VS OVERT SPEC, TP IN EUROPEAN PORTUGUESE 43
(17) 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 interrogative

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:
(18) a. [CP C+T+AgrS [AgrSP [TP t
The head created in  has the shape in (19):
Recall from Bobaljik (1995) and Bobaljik & 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 however
not the case. A reviewer suggests that the feature composition of T in
surfaces as an independent morpheme. It is likely that the same holds for future, an issue to be
further explored.
44 JOAO COSTA
interrogatives and factive contexts may be different than in declarative contexts. 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". This adverb provides more robust evidence for the
unavailability of Spec,TP in declarative contexts, since, unlike subject-oriented
adverbs, it cannot be topicalized:
(20) .  João sempre tinha feito o trabalho.

João after all had done the work
b. *Sempre  João tinha feito o trabalho.
after all Joäo had done the work
As predicted by the analysis, in questions involving I-to-C movement, the subject

may occur after this adverb:
(21) a. ?Que trabalho tinha sempre o João feito?

which work had after all Joäo done
b. Esses trabalho todos, tinha sempre  Joäo feito?, não tinha?
all those works, had always Joäo done?, hadn't he?
"All those works, Joäo always had done them, hadn't he?"
Second, as shown in Raposo (1987), inflected infinitives in European

Portuguese 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 (22):
(22) a. Os alunos lamentam os deputados terem votado a proposta.

the students regret the deputies have-3pl voted the proposal
"The students regret that the deputies have voted the proposal."
b. Os alunos lamentam terem os deputados votado a proposta.
the students regret have-3p'l the deputies voted the proposal
"The students regret that the deputies have 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 & Gonçalves 2000):
(23) a. *Eu lamento, esse livro, terem eles lido.

I regret that book have-3pl they read
"That book, I regret that they have read it."
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. (24a) shows that the subject-oriented
adverb may occur in between the subject and the auxiliary verb. (24b) shows that
if there is I-to-C movement, the subject may remain in Spec,AgrSP, in the pre-
adverbial position. The crucial contrast is the one between (24c) and (24d). (24c)
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 subject cannot stay in Spec,TP. In (24d), I-to-C movement occurred, and the
subject may surface after the adverb, in Spec,TP.
(24) a. Os meninos lamentam os deputados cautelosamente terem

the children regret the deputies carefully have-3pl
votado a proposta.
voted the proposal
b. Os meninos lament am terem os deputados cautelosamente
the children regret have-3pl the deputies carefully
votado a proposta.
voted the proposal
 *Os meninos lamentam cautelosamente os deputados ter
the children regret carefully the deputies have-3pl
votado a proposta.
voted the proposal
d. Os meninos lamentam terem cautelosamente os deputados
the children regret have-3pl carefully the deputies
aprovado a proposta.
voted the proposal
4. Conclusions
The study developed in this paper permits drawing the following
conclusions:
46 JOAO COSTA
i) There are preverbal A-positions for subjects in null subject languages;

ii) The availability of a specific A-position for subjects may derive from
specific constructions rather than from a global parameter.
The second conclusion is theoretically interesting, since it raises a number of

issues regarding 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 the 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 initial question concerning the types of
constraints that affect the availability of A-positions as landing sites for subjects,
I hope to have shown that there is not a single answer that 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.
References
Adragão, M. M. 2001. "Aquisição da inversão numa criança entre os dois e os três anos".
ms, Universidade Nova de Lisboa.
Alexiadou, Artemis & Elena Anagnostopoulou 1996. "SVO and EPP in Null Subject
Languages and Germanic". FAS Papers in Linguistics, Potsdam.
Ambar, M. 1992. Para urna Sintaxe da Inversäo Sujeito-verbo em Português. Lisbon:
Colibri.
Barbosa, P. 1995. Null Subjects. Doctoral dissertation, MIT.
Bobaljik, J. D. 1995. Morphosyntax. The Syntax of Verbal Inflection. Doctoral
dissertation, MIT.
Bobaljik, J. D. & D. Jonas 1996. "Subject positions and the roles of TP. Linguistic Inquiry
27.195-236.
Coelho, I. et alii. 2001. "Ordern VS e sujeito nulo em PE e PB". Paper presented at 2°
Colóquio do Projecto /PB, Universidade do Ceará, Fortaleza.
Costa, J. 1996. 'Adverb positioning and V-movement in English: some more evidence".
Studia Linguistica 50.22-34.
Costa, J. 1998. Word Order Variation. A Constraint Based Approach. Doctoral
dissertation, Leiden University.
Costa, J. & C. Galves 2000. "External subjects in two varieties of Portuguese". Romance
Languages and Linguistic Theory 2000 ed. by Claire Beyssade, Reineke Bok-
Bennema, Frank Drijkoningen & Paola Monachesi 109-125. Amsterdam &
Philadelphia: John Benjamins.
Costa, J. & A. Gonçalves 2000. "Minimal projections: evidence from Portuguese".
Catalan Working Papers in Linguistics.
Costa, J. 2001. "Spec.IP vs Deslocado: prós e contras das duas análises dos sujeitos pré
verbais". D.E.L.T.A 17.20.283-303.
Costa, J. & 1. Duarte 2002. "Discourse configurationality and its (ir)relevance for subject
positions". Paper presented at the 12th Colloquium on Generative Grammar,
Lisbon.
Duarte, I. & G Matos 2000. "Romance Clitics and the Minimalist Program". Portuguese
Syntax. New comparative stúdies ed. by J. Costa. Oxford: Oxford University Press.
Halle, M. & A. Marantz 1993. "Distributed Morphology". The View from Building 20. ed.
by K. Hale & J. Keyser. Cambridge, Mass.: MIT Press.
Pratas, F. 2002. O Sistema Pronominal do Caboverdiano. Questões de Gramática. MA
dissertation. Universidade Nova de Lisboa.
Raposo, E. 1987. "Case Theory and Infl-to-Comp". Linguistic Inquiry 18:1.85-109.
Ribeiro, R. 2002. As Ocorrências da Forma de Gerúndio na Variedade Padräo e 
Variedade Dialectal do Português Europeu. MA dissertation, Universidade Nova de
Lisboa.
Rizzi, L. 1982. Issues in Italian Syntax. Dordrecht: Foris.
Rizzi, L. 1986. "Null subjects in Italian and the theory of pro". Linguistic Inquiry
17:3.501-558.
DETERMINER ARCHITECTURE AND PHRASAL
MOVEMENT IN FRENCH LEXIFIER CREOLES
VIVIANE DÉPREZ
Institut des Sciences Cognitives CNRS
1. Introduction
By comparing the determiner inventory and distribution of a variety of
French Lexifier Creoles (FLC), this paper aims at furthering the investigation of
the syntax of nominal projections within the framework of micro-parametric
syntax. FLC are here shown to feature both a remarkable uniformity in the
inventory of their overt determiners and a striking diversity in their syntactic
distribution, thus presenting a particularly fertile ground for a micro-parametric
comparison. Interesting empirical and theoretical questions arise as to whether a
common structure can be held to underlie the nominal projections of the distinct
FLC. Since many of these determiners are arguably heads, does variation occur
in the basic hierarchy of the functional heads of nominal projections, as
suggested for the sentential domain (Ouhalla 1991), or can the determiner
systems of FLC be analyzed as having a common single functional architecture?
If the latter, what should this architecture be? Furthermore, if movement is
involved in deriving the distinct surface orders, as is likely under a common
architecture approach, are both head to head movement and phrasal movement
necessary or does one of these two types of movement prevail?
This work presents a foray into the DP structure of FLC within the anti
symmetric perspective of Kayne (1994), providing a number of arguments for a
common basic architecture for all FLC nominal projections and exploring an
analysis of the observed variation that features no head movement but extensive
and highly constrained phrasal movement within the proposed DP architecture.
The first step in developing this analysis is to search for evidence in support of
an underlying architecture for the FLC nominal projections of the FLC here
examined. The structure proposed parallels the functional architecture developed
for nominal constituents within the recent generative literature. The comparative
Creole data provides empirical support for the existence of a number of distinct
functional projections hierarchically organized in a fixed order above the N
50 VIVIANE DEPREZ
projection (DefP/DP > DemP/AgrP > NumP > NP). Once the basic architecture
is determined, the distinct orders are shown to be systematically derivable
through a cascade of phrasal movements governed by a single general principle:
(1) Specifier of FLC nominal functional heads must always be filled:

they contain strong features overtly checked in the Spec of the
projection.
A motivation for (1) based on a modification of Thràinson's (1996) Real

Minimalist Principle ("Assume only those functional categories you have
evidence for") is explored in work in progress. It builds on the idea that strong
interpretable features can be made visible at PF by overt lexico-morphological
distinctions or by movement. On this view, either movement or
lexical/morphological specification may serve to motivate a semantically
content-full functional projection and it is the interplay of these two modes of
visibility along a scale of grammaticalization that is assumed to motivate the
variations observed. Given (1), many of the variations observed among the FLC
determiner systems can be predicted from the X' status of their particular
determiners as specifiers (XP) or as heads (Xo), a choice arguably attributable to
the variable degree of grammaticalization of the relevant items. The remaining
variations are here suggested to derive from two distinct notions of locality that
trigger distinct types of phrasal movement, Complement to Spec phrasal
movement or Spec to Spec phrasal movement.
2. The facts
This section introduces the inventory and distribution of the FLC
determiners. The description proceeds by classes of determiners and emphasizes
the similarities and differences encountered in the distinct Creoles.
2.1 Parallel distribution: indefinite determiners

The indefinite determiners of the FLC included in our survey can be
characterized by the following generalizations:
• All the FLC have a singular indefinite determiner derived from the French
numeral/indefinite marker un (with a variety of allomorphs)
• This indefinite determiner is consistently placed in a pre-nominal position
• It admits both specific and non-specific readings
• There are no overt plural counterparts to this indefinite determiner in the
FLC under consideration, yet the singular indefinite determiner is generally
in complementary distribution with the plural marker
DETERMINER ARCHITECTURE AND PHRASAL MOVEMENT 51
Characteristic examples of this indefinite determiner are provided below. The

spelling of these examples follows that of their sources (see references):
Seychelles Creole (SC): ~e

(2) a. ~e zom "a man" b. ~e lakaz "a house"
Mauritian Creole (MauC): en

(3) En lakaz "a house"
Antillean Creole:
St Lucie (SLC): õjõ
(4) a. õ koSõ "a pig" b. jo efô "an effort"
Martinique (MarC): an
(5) a. an tab "a table" b. an timanmay "a child"
Guadaloupean Creole (Gua): õ

(6) a. An vwè õ timoun. b. Mi ö bel wob wouj.
I see a child Here is a nice dress red
Louisiana Creole (LC): ein

(7) a. einfomme "a woman" b. ein couteau "a knife"
Haitian Creole (HC): yõ

(8) a. yõ fij "a girl" b. yõ liv "a book"
2.2 Diverse distribution: definite, demonstrative and plural markers

Most FLC, (except for Seychelles Creole (SC) and perhaps Old Louisiana
Creole (OLC) vs. New Louisiana Creole (NLC)), feature both a definite marker:
la and a distinct demonstrative marker: sa, although the latter commonly co-
occurs with the former. The relative order of these markers is greatly variable in
the distinct FLCs. So is the order of the plural marker that also commonly co-
occurs with the definite marker or has itself a definite reading (see below). The
range of possibilities is illustrated in Table 1 where the combined distribution of
definite, demonstrative and plural markers across the FLCs under study is
shown:
52 VIVIANE DEPREZ
Singular Plural
Pre-nominal Pl
SC DemN sa N Dem PL N sa ban N
MauC Dem N (Def) sa N-la Dem PLN(Def) sa ban N-la
Ant N Dem Def N-sa-(l)a, PL N Dem Def se N-sa-(1)a,
N-ta-(l)a se N-ta-(l)a
Gua N Def Dem N-la-sa PL N Def Dem se N-la-sa
NLC (Def) N Dem-(Def) (la) N sa-la (Def/PL) N Dem-(Def) le N sa-la
Post-nominal PI
OLC N Def/m N-la N (Dem= sila) PL N-(sila)ye
GyC Dem N Def sa N la Dem N PL Def sa N-y(e-l)a
HC N Dem Def N-sa-a N Dem PL N-sa-yo
Table 1: Distribution of definite, demonstrative and plural markers across FLCs
Before providing concrete examples illustrating these distributions, some

comments about Table 1 are in order.
The definiteness marker la: note that although called a definite determiner here,
la is not the exact counterpart of the French or English definite determiners in
any FLC. For instance, unlike French le, FLC la does not generally occur in
generic or kind denoting expressions, although this seems possible in some FLC
(HC). La is often said to have deictic force, although empirical tests are usually
not provided to support this claim that may relate to conjectures on likely sources
for this marker.1 For instance, the French post-nominal demonstrative reinforcer
là in expressions such as ce livre-là (this book there) is often considered a likely
source. Should this conjecture be correct, FLC la would indeed derive from a
former demonstrative (a common source for definite determiners) and plausibly
have preserved some demonstrative force. However, candidate substrate
languages also feature post-nominal definite determiners that are likely sources
for la. Ewe for instance has a lexical definite article la that occurs in NP final
position: ati lá "the tree" (Lyons 1999:77) and so does Wolof (Cérol 1991:85).
Most FLC also feature an adverbial locative marker la clearly derived from the
French locative adverb là, and also considered a likely source. There are,
however, in some FCLs (i.e. Antillean) rather clear arguments that the adverbial
marker and the definite marker are now distinct. In HC, for instance, only the
determiner la presents an allomorphic paradigm {la, a, an, lan, nan (Joseph
1988)) governed by the phonological form of the immediately preceding lexical
1
For an in depth discussion on the origin of the post-nominal marker LA in Haitian Creole, see
Fattier 2000.
item. The same is true in Martinique Creole and more generally Antillean 
(Baker, pc.) The marker la further seems to have clear properties of defmiteness.
For instance, it exhibits the characteristic uniqueness and familiarity feature of
definite determiners in so called anaphoric contexts i.e. in reference to a unique
entity previously introduced in the discourse. HC la can also designate a unique
object identifiable by situational reference, discourse reference or even common
knowledge reference (Fattier 2000:13).2 Since la is usually unstressed and shows
no proximate/distal contrast, it differs from common demonstrative markers.
Finally la can and sometimes must co-occur with other demonstrative markers
such as sa (cf HC (Fattier 2000) (or si(la)), a feature that is cross-linguistically
common for definites but uncharacteristic of demonstratives (Lyons 1999). Both
these co-occurrence factors and the familiarity/uniqueness properties rather
clearly suggest that la is a defmiteness marker (at least currently in some FLC).
Concerning its syntactic position, we note that the definite marker la occurs
in post-nominal position in all the FLC considered. Although this order clearly
reflects a strong tendency, it seems not absolute. There is a pre-nominal la in at
least one FLC, namely, Reunion Creole (RC) {la kaz "the house": Baker
2002:15). A pre-nominal definite determiner also occurs in present day Louisiana
Creole (NLC) (Neumann 1985). Both RC and NLC do seem to also feature a
post-nominal marker la that may sometimes even co-occur with the pre-nominal
one (Neumann 1985:134). Given the possible presence of a pre-nominal la in RC
and NLC, it seems that both pre- and post-nominal positions are possible for all
the markers in Table 1. That is, although the position of these elements is usually
fixed within a given Creole, the spectrum of variation across FLC does not
strictly preclude a pre- or post-nominal position for any of them. This contrasts
with the strictly pre-nominal position of the singular indefinite determiner, which
never varies across FLC or time.
Ordering generalizations: it is notable that when comparing the pre-nominal and

the post-nominal ordering of the functional elements in Table 1 neither a fully
parallel distribution (i.e. identical ordering on either side of N), nor a full mirror
image distribution (reverse ordering on either side of N) can be observed.
Determiner order on both sides of the nominal constituent seems to result from a
somewhat puzzling combination of both.
Consider first the pre-nominal ordering. We observe that pre-nominal
definite and demonstrative markers do not generally co-occur, thus appearing
unordered with respect to one another in this position. Both, however, clearly
systematically precede the pre-nominal plural markers leading to a Def/Dem > PI
2
Empirical work has led me to comparable conclusions for MarC.
54 VIVIANE DEPREZ
(N) pre-nominal ordering. The reverse ordering *PL > Dem/Def (N) is never
observed across FLC.
In post-nominal position, the order of demonstrative and plural markers is
also rigid, with the former always preceding the latter (N) Dem > Pl, paralleling
the order found in pre-nominal position. The reverse ordering seems possible
only when these two markers are distributed around the nominal projection PI >
(N) > Dem/Def, as in Antillean or Guadeloupe Creole, never when they both
precede or both follow the nominal projection. The post-nominal ordering of the
demonstrative and plural markers with respect to the definite shows a greater
amount of variability. In post-nominal position, the demonstrative and definite
markers clearly co-occur and are ordered There is a strong tendency across FLC
for post-nominal demonstratives to precede post-nominal definites (N) Dem >
Def, sa (l)a, in apparent mirror order with the pre-nominal position with,
however, one notable exception, namely Guadeloupe Creole (Gua), where the
opposite order, Def > Dem, is instantiated. Interestingly, however, in Gua,
variants showing the more general Dem > Def order also appear possible (Cérol
1991:85). Turning to plural and definite markers in post-nominal positions, the
most general tendency seems to be a complementary distribution. This is
observed in Haitian Creole and Old Louisiana Creole where plural and definite
markers do not co-occur and where the combination of NP + plural marker seems
to systematically receive a definite interpretation. It is also true to some extent in
Guyanese Creole, where the plural tends nowadays to be an agglutinated form ya
> ye + la. There are, however, both synchronic and diachronic variants of these
Creoles that manifest an apparent possible co-occurrence of the two markers. In
these variants, la appears to precede the plural marker Def > PI (la yo, la ye) in
HC and OLC. Guyanese, however, exhibits the reverse order, with the definite
following the plural marker, PI > Def (ye la →ya).
Interestingly, and it seems to me significantly, the three markers of Table 1
do not seem to be able to occur together in post-nominal positions. That is,
although definite demonstrative and plural markers are all found in post-nominal
positions, either individually or in pairs, the three of them are usually not found
together. Consider for instance the HC variants where la can co-occur with a
plural marker (la yo). In this variant, the demonstrative marker is also found to
commonly co-occur with the plural marker (sa yo) and independently, in the
singular, with the definite marker (sa (l)a). But the systematic co-occurrence of
the 3 markers together in the expected N Dem > Def > PI (sa (l)a yd) order has
not been observed (Fattier 1998, 2000). Interestingly, the only attested very rare
example of the three markers together (Fattier 1998) manifests the unexpected
Dem > Pl > Def order, also observed in Guyanese Creole.
There is a third generalization concerning the distribution of these three

markers, not apparent from Table 1 for want of an easy way to notate it in a two
dimensional table. The generalization is that a certain amount of more or less
obligatory co-distribution between the markers is observed. It is a general
tendency (although not an absolute one (cf. SC)) that demonstrative markers
(sometimes obligatorily) co-occur with a post-nominal definite determiner (Guy,
Ant, Gua, HC). Similarly, a certain amount of co-occurrence is observed between
pre-nominal plural markers and post-nominal definite markers. This co
occurrence seems obligatory in both Ant and Gua (Cyril 1997) but optional in
Mau or Guy. Note that when co-occurring, these markers distribute around the
nominal projection, PL being pre-nominal and Def post-nominal. In sum,
obligatory - between the above markers seems to obtain only when
they either distribute around the nominal constituent or are both post-nominal
(cf. Dem< Def). Following are some concrete data, illustrating Table 1 and the
tendencies discussed above.
• Seychelles Creole: Dem N Dem PL N

(9) a. sa kuto "this knife"
b. sa lasjet "this plate"
(10) a. ban 1er at "the rats"
b. sa ban zako "these/the monkeys"
• Mauritian Creole: Dem N (Def) Dem PL N (Def)

(11) a. liv la "the book"
b. sa loto-la "this car"
(12) a. ban politisyen "politician"
b. ban liv la "the books"
 sa ban lakaz dibwa "these wooden houses"
• Antillean Creole:
- St Lucie : N Dem Def PL N Dem Def
(13) a. tab la "this table"
b. kat mile sa la "those four mules"
 se koSõ an "the pigs"
- Martinique: N Dem Def PL N Dem Def
(14) a. timanmay la "the child"
b. bel kay tala "this beautiful house" (Def not optional)
(15) a. se tab la "the tables"
b. se boug tala "these guys"
c. *se tab
56 VIVIANE DEPREZ
• Guadaloupean Creole N Def Dem P1 N Def Dem

(16) a. tab la "the table"
b. boug lasa "this man" (Def not optional)
(17) a. se bong la "the men"
b. sé tab wouj lasa "these red tables"
c. *sé timoun "the children"
(18) Nonm pasé an chimen-sala. "Men were passing on this road."
• New Louisiana Creole

(19) a. La flj vini reste avec mwa isi.
Def girl came stay with me here
"The girl came stay with me here."
b. Je te pele vj(p mile-la Fiva
They Past call old mule-Def Five
"They called the old mule Fiva."
c. Eu twa to tõde par le pu dõs-sa-la?
Where you Past hear speak of dance Dem-Def
"Where did you hear about dance?"
(20) a. Le vje  par le e ta kreol
Def+P1 old people talk a lot Creole
"The old people talk Creole a lot."
b. le kokodri-sa-la "these crocodiles"
 le kokodri-la-la "these crocodiles there"
d. (le) kokodri sa ye "these crocodiles"
• Old Louisiana Creole

(21) Nom la et fom la rivé hiè.
"The man and the woman arrived yesterday."
(22) Nom yé et fommes yé rivé hiè.
"The men and the women arrived yesterday."
• Guyanese Creole (Baker 2002)3

(23) a. mé pagay-la, uben si upimigo, pran guvernay-la
(St. Quentin 1872:116)
b. agouti mo manjé kan mo té là bitachon yéla
(St Quentin 1872:116)
c. sa wom-la "cet homme"; sa fam-yéla "ces femmes"
(St Quentin 1872:122)
3
These examples are taken from Baker's paper. As no glosses or translations are given, this author
would rather not make them up for fear of erring in the wrong direction.
(24) a. Vf... sa PSEUDO-entélèktuèl-ya-ya

b. V[ sé PSEUDO-entélèktuèl-ya]
"these pseudo intellectuals"
(Data from Ludwig & Telchid 2001)
• Haitian Creole
(25) a. bel pye-bwa (l)a "the beautiful tree"
mange sa a "this food"
pitit sa a "this child" (Def not optional)
(26) a. monnyo "the people"
mouvman sa yo "these movements"
3. Structure and orders

This section presents the proposed common architecture of the nominal
projection in FLC and discusses the derivations of the possible variations. The
structure is discussed in section 3.1 and the derivations in section 3.2.
3.1 Common Nominal Structure

Creoles, as is well known, are quite generally head complement languages
(Bickerton 1981). This generalization, clearly verified in FLC in VP, AP and NP,
suggests that the pre-nominal order of determiners could well be the most basic
one, reflecting the common underlying architecture. The DP hypothesis,
motivated on much crosslinguistic work, further entails that determiners project
functional structures distinct from the lexical nominal ones. Based on an analysis
of verbal functional projections, it has been suggested that FLC may be said to
wear so to speak their functional structure on their sleeves (Rottet 1992). For
nominal constituents, such a view suggests that there may be at least three
distinct functional projections above NP, a definite projection, a demonstrative
projection and a number projection, each representing the distinct overt markers
commonly found in FLC. The discussion below explores some arguments for the
existence of these separate functional projections.
3.1.1 Although as observed above, no obvious pre-nominal order between the

definite and demonstrative projection emerges from our data, the fact that these
two elements can co-occur or occur separately on either side of the nominal
constituent suggests that they may have distinct projections in at least some FLC.
Crosslinguistic studies of definite and demonstrative co-occurrence further
suggest that a Def > Dem ordering may be a default one.4 Lyons (1999:64) notes
"...there is a strong tendency for the definite article, if a free form, to occur
4
The reverse order Dem > Def, however, is not excluded e.g. Greek. (Panagiotidis 2000)
58 VIVIANE DEPREZ
initially in the noun phrase, independently of a language's general constituent

order pattern". If this is correct, positing a left headed initial definite projection
has the advantage of being both compatible with the FLC data and motivated by
general crosslinguistic considerations.
Several linguists have argued that demonstratives are base-generated in a
constituent separate from DP (Giusti 1993; Bernstein 1997; Panagiotidis 2000).
For Giusti (1993), demonstrative markers are specifiers generated in a projection
below DP called AgrP. For Panagiotidis (2000) demonstratives can also head
their own projections and attract a deictic operator in their Spec. The latter
proposal is attractive for the FLC that manifest 'independent' definite and
demonstrative markers, the criteria for 'independency' being here 1) the presence
of distinct definite and demonstrative markers {la & sa) and 2) the possibility for
the definite marker to occur alone. These criteria seem achieved in most FLC
except in SC and OLC, which exhibit either the markers la or sa but not both and
do not have other clearly distinct definite and demonstrative markers. This
suggests that DP and DemP may be separate projections in most FLC but fused
ones in some. See below for more discussion of the distinct FLC.
The strict Dem > Pl pre-nominal order observed both in pre-nominal and in
post-nominal positions suggests a structure in which the demonstrative
projection dominates a number projection. Motivation for a separate NumP is
"discussed in 3.1.2. Summarizing the possibilities, the maximally expanded
functional nominal structure for FLC would be as in (27):
Demonstrative Phrase
Number Phrase (= NumP)
(27) presents the different Dem and PL markers as either potential heads or
potential specifiers of their containing projection. I suggest below that either
possibility can be instantiated in distinct FLC and that this may be an important
source of the ordering diversity. As for the singular indefinite determiner, I
propose that it is merged invariably in the Spec of NumP across FLC, just like
other number terms in a variety of languages (Ritter 1992). This will ultimately
account for its constant pre-nominal order.
3.1.2 Some motivation for a definite and a demonstrative projection has been
presented above. This section discusses empirical arguments in support of what
is perhaps the most controversial functional projection in FLC, namely the
projection of number. These empirical arguments concern chiefly the pre-
nominal number markers.
As the SC and MauC pre-nominal plural marker ban derives from the
French nominal bande, it may seem tempting to analyze ban + NP combinations
as binominal NN constructions rather than as the combination of a functional and
a lexical projection. But convincing empirical arguments can be given against a
NN analysis on the basis of a point-by-point comparison between the behavior of
common nouns in clear binominal constructions and that of the plural marker
ban (Rottet 1997). The arguments summarized here are based on MauC data but
comparable arguments could be made for SC. They all point to the conclusion
that ban cannot be treated as a noun and must rather be a functional projection.
In the regular binominal constructions of MauC, as (28) shows, both the
head noun and the complement noun can host independent modifiers:
(28) a. en gro sakpuasõ b. en sak gro puasõ

a big bag fish a bag big fish
"a big bag of fish" "a bag of big fish"
As (29) shows, however, ban itself can never be independently modified:
(29) a. *trua gro ban zanfan b. ban trua gro zanfan la

three big PL children PL three big children the
"the three big children"
Similarly, in binominal constructions, both the head noun and the complement
noun can be independently determined (29). For ban, this is impossible (30):
(30) a. sa karotsu-la sa loto-a b. *sa ban-la sa zanfan la

that tire of that car that PL the that children the
Comparing a [ban + NP] construction with a potentially synonymous measure

noun [grup + NP] construction further reinforces the above division, showing
that besides syntactic differences, there are also clear distinctions in
interpretation. While (31a) with grup necessarily leads to a partitive
60 VIVIANE DEPREZ
interpretation, the most natural interpretation of the ban+NP construction is that

of a (definite) plural. This interpretation is not available for a [grup+ NP]
construction. (31b) shows that, in such contexts, there is no partitive
interpretation for the ban construction. As for (31c), it shows that when ban
combines with an apparent indefinite determiner (or the numeral "one"), it has a
particular interpretation = "some", and lacks again the clearly partitive
interpretation of a [grup + NP] construction. In this respect, the combination [en
ban] recalls the French quantifier quelqu'un/quelques-uns ("someone")
historically derived from the combination of the adjectival quelque and the
numeral un.
(31) a. Mofintruv ban/*grup zanfan pe zue dan lakur

I have found the kids playing in the yard
b. Mo ti pe get trua grup/* ban zanfan pe zue dan lakur
I was watching three big groups of kids playing in the yard
 Mo fin truv en grup/en ban zanfan pe zue dan lakur
I found a group/some kids playing in the yard
To sum up, the preceding set of comparative examples shows clearly that the
construction [ban + NP] is neither syntactically nor semantically equivalent to
the construction [measure noun + NP]. Ban clearly has a more functional
behavior than a measure noun since it precludes both modification and
determination and is associated with a particular non-nominal interpretation. The
analysis of ban as a functional plural marker thus seems amply justified by the
data.
Further supporting evidence for the functional status of ban comes from
historical considerations. As noted in Baker (2002), ban was first used in MauC
in co-occurrence with an indefinite determiner {en ban) with the meaning "a
group of' {ein band'p'tits miletons "a group of small mules" (Lolliot 1855:67)).
Later (end of 19th century) ban started appearing alone and began to take on the
function of an optional plural marker. That ban began to function as a plural
marker when it began to enter in complementary distribution with the indefinite
en suggests that it came to occupy the same syntactic position of en i.e Spec
NumP, thus entering in competition with it. This would account for their present
day complementary distribution.
The question of the position of ban within the NumP projection is a matter
of some debate. For Rottet (1977) ban is the head of the NumP. However, both
the historical evolution of ban, and its distribution with respect to other
determiners raises problems for this assumption.5 Below is a quote from Bollé
(1977): "The plural marker ban can be preceded by the demonstrative, the
possessive and tu but not by a numeral or by a quantifier, apart from a few very
rare exceptions". Relevant examples, including one with the rare combination of
ban with a numeral are given below:
(32) a. ban liv la "the books"

b. sa ban lakaz-la "these/those houses" (MC)
c. tu sa ban gardjè "tous les gardiens" (SC)
d. sa ban trua gro zom "those three big men" (MC)
Particularly relevant for determining the position of ban within NumP are
considerations on the respective ordering of ban and numerals. If as some have
proposed (Zamparelli 1995), numerals are projected in NumP, and generally
merged in its Spec, it follows that ban cannot both occupy the head of NumP and
precede a numeral within NumP. As clearly shown by (32d), however, ban must
precede a numeral whenever they co-occur — however rare this co-occurence
may be. Examples like (32d) thus suggests that when ban and a numeral co-
occur, ban could occupy the Spec of NumP while the numeral itself would
occupy its head. This would satisfy the requirement that both ban and the
numeral be contained in NumP, even if it seems unusual for a numeral to head
NumP. Note, however, that the co-occurence of a numeral with a plural marker,
although clearly ordered PI > Numeral, is in fact not the common way of
expressing cardinality in FLC. In regular cases, numeral expressions are used
without plural markers, the number term alone being sufficient to indicate
plurality. That is, in contrast to languages like English, FLC in general appear
more 'economical' for number marking: number is expressed only once by the
presence of the numeral term. In regular cases, then, numerals are in fact in
complementary distribution with the plural marker. If as suggested above, the
plural marker occupies the Spec of NumP and not its head, this complementary
distribution could result from a competition between the number term and the
plurality marker for the same syntactic position, i.e. the specifier of NumP. In a
sense, number and numerals in languages like FLC would be assimilated to
numerals, a potentially attractive perspective that aligns singular and plural (cf.
the number "one" as a marker of singularity in Spec NumP) and is in line with
the proposal in Heycock & Zamparelli (2000) that plurals may have a cardinal
like interpretation in some languages. It is then only in particular cases, i.e.
apparently equivalent to definite numeral expressions {the three books) (further
5
The discussion here is based on Seychelles Creole data, but again there are comparable data in
Mauritian.
62 VIVIANE DEPREZ
research is required to better establish this fact) that numerals can co-occur with
plural markers. On our view then, it is only then that a numeral could come to
occupy the head of NumP. In the general case, the number term and the plural
marker both occupy Spec NumP, thus entering in complementary distribution.
3.1.3 Turning now to the pre-nominal plural marker se of Gua and MarC, we
observe that its distribution is strikingly similar to that of ban. First, se like ban
is usually in complementary distribution with numerals in the regular cases of
cardinal expression (33 a). Further, like ban, the plural marker se can follow the
universal quantifiers tout and when it co-occurs with a numeral, se must always
precede it (33c):
Guadeloupe
(33) a. twa timoun "three children"
b. *sé Uva timoun
c. tout se twa timoun la (sa)
all PL three children the these
"all (these) three children"
Martinique
(33')a. Man ni senk yich.
I have five children
b. Tout se moun ta la anfrans.
All PL people these the France
"All these people are in France."
 I vini épi sé dé misye a
He come with PL two men the
"He came with the two men."
The fact that expressions containing a numeral in Gua and MarC contain no
plural suggests again that numerals and the plural marker se are in
complementary distribution, both competing for the same syntactic position, i.e.
Spec NumP. Examples like (33c) and (33'c) also suggest that like numerals in
MauC and SC, the MarC and GuaC numerals may sometimes head the NumP
projection and thus co-occur within NumP, with the plural marker se in Spec
NumP. Cyrille (1997) also argued for the existence of a NumP projection in Gua
but proposed that se is the head of NumP. She was, however, unable to account
for the order exemplified in (33 c) or for the usual complementary distribution of
numerals and plural markers. Although quite similar, distributionally se differs
from ban in one important respect. In contrast to ban, the plural marker se in Mar
and GuaC never occurs alone with a noun. It must always obligatorily co-occur
with the definite marker la. That is, expressions such as *se liv "books", are
ungrammatical (Cyril 1997, Damoiseau 1999) as la must always be present
whenever se is. Moreover, while the pre-nominal plural marker ban seems
compatible with an indefinite reading, (ƒ ana ban move voler "there are bad
thieves" (VI: 158 Bollée 1977)) this seems impossible in MarC and Gua. Given
this property, it turns out that the co-presence of a plural marker with a numeral
in these FLC causes an obligatory change in the meaning of cardinal expressions
from indefinite as in (33 a) to definite as in (33 c) ("three boys" vs "the three
boys"). The association of a numeral with a plural marker in these FLC serves to
introduce defmiteness rather than plurality. I will assume that the pre-nominal
plural marker ban and se differ as follows: se is associated with a strong
+[defmite] feature that must be overtly checked. Ban in contrast, is unmarked for
defmiteness, and thus compatible with either a [+defmite] marker or with
indefmiteness.6 The consequences of this will be seen in the ordering derivations
discussed in the following section.
3.2 Deriving the different orders

Given our assumption that the basic architecture of nominal projections is
as in (27), the only predicted orders so far are those of pre-nominal determiners. I
will first assume that throughout FLC, the singular indefinite determiner
occupies the specifier of a singular NumP, thus blocking both head movement
and XP movement to this position and that it is negatively specified for a definite
feature [-def]. NP movement across the NumP projection, say to the specifier of
DP or DemP could in principle be possible. However, if as seems natural given
interpretative considerations, both DemP and DP are associated with a [+def]
feature, movement of NP above a [-def] projection would presumably be ruled
out on interpretative considerations since NPs themselves are not specified for
defmiteness. This then suffices to derive the general pre-nominal order of the
indefinite singular marker.
Concerning other attested pre-nominal orders, the la+NP order is found in
RC (and irregularly in NLC), sa+NP occurs in SC and MauCr, Pl+NP is realized
as ban+NP in SC and MauC and as se+NP (+la) in Antillean and Gua. All these
pre-nominal orders are derived straightforwardly given (27). The obvious
question now is: where do the post-nominal orders come from?
There are, of course, several possible answers to this question. First, the
architecture proposed in (27) could not be common across FLC with the distinct
FLC having different headedness for their functional projections. This has been
proposed by Lumsden and Lefebvre for HC whose markers all are post-nominal.
6
The indefmiteness marker en, on the other hand, can be assumed to be marked negatively for
defmiteness [-def]. As a result, it will be generally incompatible with definite markers.
64 VIVIANE DEPREZ
Such a proposal, however, seems more difficult for Creoles like MauC or Gua
whose markers distribute on both side of the nominal constituent. Questions arise
in these FLC as to why some determiner projections should be right headed but
not others. Even within HC, this proposal raises some unanswered questions,
since there is at least one pre-nominal determiner in HC, namely the indefinite
singular one, and all lexical projections are clearly left headed as in all FLC.
Apparent variations in the headedness of distinct constituent types are not
uncommon in the languages of the world. The tradition in generative syntax,
however, has been to consider these as resulting from movement. Such a
tradition has led to Kayne's interesting proposal that languages are uniformly left
headed. I will thus assume that post-nominal determiners result from movement,
not from distinctions in headedness. The question then is what type of movement
is involved and what drives it. Assuming that the architecture in (27) is common
to all FLC, movement may involve either a head or a maximal projection.
Consider a simple case of a post-nominal definite determiner.
The order liv la could a priori well derive from movement of the head noun to
the determiner head or from the movement of the entire NP over it. This
question, however, turns out to have a simple answer as soon as more complex
nominal constituents are considered. As is evidenced by (35) and (36), it is not
just the head noun that occurs before the determiner, but rather the noun with
various determination markers and all of its possible modifiers, adjective or
relative clauses, and all of its complements. Clearly then, it is the whole nominal
constituent, not the nominal head alone, that must move:
(35) bel kay che la ' (HC)

beautiful house expensive the
"the beautiful expensive house"
(36) sa ban zen zom katolik ki Mari kötä la (MauC)
Dem PL young men catholic that Mari loves Def
"those young Catholic men that Mary loves"
As comparable arguments can be given for the post-nominal position of the other
determiners, it is clear that phrasal movement is the key to the observed ordering
variation in FLC, not head movement. The question of why this should be
naturally arises. As numerous considerations in the generative literature have
found head movement to be generally linked to rich inflectional morphology, the
lack of head movement in FLC should come as no surprise, given the rather well
known paucity of inflectional morphology in Creoles.7 A further consideration

suggests itself. Given that heads are in effect the minimal part of XP projections,
there is a sense in which head movement can be considered a subcase of XP
movement but the reverse is not true. From this point of view then, XP
movement is a more basic type of displacement than head movement. If Creole
languages manifest a tendency to instantiate unmarked choices along available
parametric options (a tendency that may nonetheless be subject to stronger
constraints imposed by the influence of substrate or superstrate languages) the
choice of XP movement over head movement appears in line with this tendency.
Going a bit further, I would like to propose that in FLC nominal
projections, all markers that are heads must have their specifier occupied. That
is, all FLC determiner heads have strong features in the sense of Chomsky
(1995) and thus, systematically attract a phrasal projection in their Spec to check
them.
Proposal: XP movement
Nominal functional heads are strong probes in FLC
Again, questions arise as to why this should be, and perhaps more
particularly as to why such a systematic attraction pattern is manifest primarily
in nominal constituents but not in other functional domains, such as for instance
the verbal functional domain. In all FLCs, indeed, verbal functional projections,
and more generally Tense, Mood and Aspect markers quite generally precede the
verbal constituent. They do not follow it. Although an in depth comparison of the
verbal functional domain and the nominal functional domain in FLC is beyond
the scope of this paper, I would like to mention a few factors that may help
understanding this intriguing difference. First, it may be of relevance that
nominal functional projections manifest a certain amount of semantic or
pragmatic ambiguity. That is, markers in the nominal functional system seem to
have changed from one meaning to another across time (cf. in particular la sa),
without necessarily changing their morphological shape. This does not seem to
have occurred to the same extent in the verbal functional paradigm. Within the
general perspective of the Real Minimalist Principle which imposes a motivation
for the postulation of distinct functional projections in. any language, the
distinction observed between the verbal and the nominal system could perhaps
be interpreted as motivated by this distinction. The idea is that movement may
7
The case of verb movement in current Louisiana Creole convincingly argued for in Rottet (1992)
constitutes an interesting exception for head movement within FLC. As Déprez (1999) argues, this
case provides an important proof that the implicational relation between head movement and rich
morphology goes from head movement to rich morphology and not the other way around.
66 VIVIANE DÉPREZ
serve to disambiguate potentially ambiguous nominal functional markers in the

nominal domain but may not be needed in the verbal functional paradigm, if the
latter presents less interpretative ambiguity of its functional heads. Another
consideration seems also important with respect to the distinction in FLC
between the verbal functional domain and the nominal one, namely the
importance of substrate and superstrate influence. Both the substrate languages
and to some extent the superstrate language do seem to exhibit a certain amount
of phrasal movement in the nominal system but there is no comparable
displacement in the verbal system. More precisely, African languages like Ewe,
Fon Gbe and Wolof among others clearly manifest post-nominal determiners
rather similar, to those found in FLC. The positions of these determiners have
been argued to derive from phrasal movement (Aboh 2000) that constitute a clear
model for the FLC determiner order. Furthermore, as argued by Bernstein
(1997), French also seems to manifest phrasal movement in the derivation of its
demonstrative reinforcer pattern. Constructions such as ce livre-là, that are a
plausible source for part of the definite/demonstrative determiner system of FLC
are analyzed as involving phrasal movement of the nominal constituent. Such
considerations may have favored XP movement in FLC D?s.
It follows from the proposal made above and from a strict notion of locality
discussed below that if a determiner is a head in FLC, it will have a post-nominal
position. Since not all determiners are post-nominal, it must also be the case that
they are not all heads. We have discussed potential evidence based on
distributional facts that pre-nominal plural markers occupy the Spec of their
NumP projection. Although comparable evidence still need to be discovered to
characterize the status of demonstrative markers in FLC, it is clear that a survey
of the cross-linguistics data provides ample evidence that demonstrative markers
can be heads or specifiers. Such a distinction, for instance is explicitly argued for
in Schoorlemmer (1996) and Panagiotidis (2000) among others. Pending
supporting empirical evidence, I will assume that this distinction is also present
in FLC and may in fact be an important factor of variation. Within the realm of
diachronic data, a change of status from a specifier to a head (or from an XP to
an Xo) has been a frequently noted evolution. From the perspective of
creolization (as well as historical change), this change, I suggest, could also be
understood as a further step along the dimension of grammaticalization. As an
exploration of these factors remains to be done for the FLC under study, the
analysis developed here should be taken as preliminary. Its overall direction,
however, seems promising enough to warrant further exploration. Armed with
phrasal movement and variability in the specifier/head distinctions, I will now
proceed to offer a derivation for the determiner ordering of each of the FLC in
Table 1.
1. Seychelles : no XP movement. SC seems essentially to manifest the linear

order predicted by the structure (27). This is the case for the demonstrative and
the plural markers for which SC exemplifies the absence of movement. Things
are a bit less clear with respect to the definite marker la. In the data reviewed
from Bollée (1977) la never seems to occur as a determiner, sa in SC seems to
serve both as a demonstrative or as a weak deictic determiner, making reference
to an entity previously mentioned in the discourse. Possibly then in SC, the DefP
projection may be indistinguishable from the DemP projection and simply not
projected. If so there are only two functional projections in SC (DemP and
NumP) and their order results from the absence of movement. If there is no
distinct DefP projection in SC, then there cannot be a strong deictic or definite
feature in D° that needs to be checked. The plural projection moreover, being
apparently compatible with either a definite or an indefinite reading may be
assumed to be unmarked for defmiteness. In short no movement is required as
there are no strong features that need checking.
(37) [DemP SA [NumP BAN[NP N ]]]
2. Mauritian Creole: turning now to MauC, there is evidence here for the
existence of separate DP/DemP projections. Present day Mauritian Creole
distinguishes between the definite la and the demonstrative sa. As noted in
Baker (2002) for instance, there has been an interesting change in the syntax of
la in the history of MauC. The first mark of determination in MauC seems to
have been a bimorphemic sa N la where neither marker was independent of the
other, and both had essentially an indistinct definite/demonstrative value. Later
on, a distinction occurred between the la and the sa, and the former began to
appear alone with nominal constituents with a more distinct definite value. I
suggest that MauC exemplifies a case where the defmiteness marker is an
independent head while both the demonstrative and the plural markers are
specifiers of their own projections. The attested order is derived as follows: the
DemP projection is attracted to the spec of DP with everything that it contains. It
seems plausible to assume that with the projection of a separate DP comes the
presence of a strong [+Def] feature in D°. As this feature needs checking,
movement to Spec D° is required. Either the projection of NumP or the projection
of DemP can be attracted to Spec DP. As mentioned above, we may take the
projection of NumP to be underspecified for defmiteness whenever it is null or
contains a plural marker. When it contains the singular indefinite marker en, it
will be marked [-def] and movement will be impossible, presumably for
interpretability reasons. The result will be that in MauC whenever D° projects, it
must attract its immediate complement.
68 VIVIANE DEPREZ
3. Antillean Creole: AntC and in particular St L and MarC both involve one more
step in the derivation of their determiner order than MauC. I suggest that both the
Def and the Dem markers are heads that must each check a strong feature. First
Dem attracts its complement NumP into its specifier (presumably to check a
strong deictic [+deic] feature) (resulting in [ [(PI) NP] Dem]) and then Def also
attracts its complement to its specifier, presumably to check a strong [+defmite]
feature, resulting in a [[(PI) NP Dem] Def] order. The derivation is a pure case of
Comp to Spec movement. As the plural morpheme se in the Spec NumP has been
assumed above to also be positively specified for defmiteness, we may further
hypothesize that this feature is strong and needs checking. If so, a nominal
constituent that projects only a Plural Spec NumP will be ruled out, the strong
[+def] feature of the plural morpheme remaining unchecked. The projection of a
head able to check the strong [+def] feature of the plural marker is required,
accounting for the obligatory co-presence of the defmiteness marker with the
plural marker in this FLC.
4. Guadeloupean C: Recall that what distinguishes Gua from other Antillean  is

the fact that it manifests an unusual post-nominal Def > Dem order. Given the
proposed structure (27), this order can obtain easily if a Spec to Spec phrasal
movement is involved rather than a Comp to Spec one. Suppose that in GuaC,
NumP moves to Spec DemP : [se NP] sa and subsequently to Spec DP as in (40).
This will derive the Def > Dem order, i.e. a post-nominal order that is parallel to
the pre-nominal one.
Recall from above, that the order Def > Dem is apparently not the only one
found in GuaC: variants show the same order as Antillean (sa la). This second
order illustrates the same pure Comp to Spec derivation as in the other Antillean
creóles.
The question arises as to why only GuaC permits these two derivations?
After all, se clearly has a [+def] feature in MarC too, and so a Spec to Spec
derivation with [NumP + PI = se +NP] should also be possible. As a first point of
consideration, note that a Comp to Spec derivation seems simpler/more local
than a Spec to Spec one in some regards. For a Spec to Spec phrasal movement
to be possible, the complement of a head H must be transparent enough to allow
penetration inside its complement domain down to the Specifier of the
complement. A Comp to Spec phrasal movement in contrast requires no such

'penetration'. It is thus more 'local' in the sense that it can be said to make use of
more basic relations. Such considerations may explain why the Comp to Spec
phrasal movement type seems in general more common in languages like FLC,
which apparently favor more 'basic' syntactic relations.
5. Louisiana C: the fact that the OLC plural marker is post-nominal suggests that
it is a head with a strong [number] feature that attracts its NP complement.
Although there seem to be several dialects of Louisiana French Creole, the one
here called OLC (Broussard 1942) does not seem to clearly distinguish the Def
projection from the Dem one. In OLC, la is clearly used, but there is no mention
of sa anywhere and la is said to have a strong demonstrative force. This suggests
an analysis where la occupies the head of a single undistinguished DemP/DP
projection immediately above the NumP (as in SC) that attracts its complement
to check its strong feature. Given that Num° attracts its NP complement and that
D° does so as well, OLC instantiates a pure Comp to Spec derivation. Note,
interestingly, that in OLC, the plural marker seems to be in complementary
distribution with the Def/Dem one so that an NP ye la order is never attested. A
possible explanation for this fact could be that the definite marker la has come to
mark singularity in addition to defmiteness, hence becoming incompatible with
the plural head yo. Since plural NPs with an overt plural marker are usually
interpreted as definite NPs in OLC, it may be that a definite plural is represented
by a [+ definite] null head in D° with NumP sitting in its specifier and ye
occupying the head of NumP in Spec D°, or it may be that yo has come to be
absorbed phonologically by the Def head in the Spec of which the NumP
projection sits. If so, the plural marker would in fact end up in the D°head:
( 4 1 ) [Def/DemP [NumP [ N P N ] J YE tj]i 0 ti ]
OLC seems, however, to have had an earlier stage at which la occurred in co
occurrence with the plural marker ye. In this diachronic(?) variant the Def/Dem
marker la preceded the plural marker ye with the resulting order NP la ye:
(42) Pikan layé déchiré ma lapo.

These spikes have torn my skin
(Valdman from Neumann text in Veronique 1994:196)
Note, interestingly, that this variant can easily be derived with a simple change in
the derivation, i.e. using a Spec to Spec rather than a Comp to Spec one. That is,
NP would move first to Spec NumP and then on to Spec DP giving NP la ye.
70 VIVIANE DEPREZ
That is, in the NP la ye dialect NP also moves to Spec NumP but it then
continues to move to Spec Def/DemP leading to a post-nominal order of the
markers that parallels the one observed in pre-nominal position (Def/Dem > P1).
(42')[ Def/DemP [NP N] j LA [ ti YE ti]]
If this is correct we have here, as above for Gua, two (diachronic) variants that
are distinguished by the type of phrasal movement they involve (Spec to Spec vs
Comp to Spec ), rather than by the status of their markers (Spec / head). Since
the NP la ye stage of LC preceded the NP ye+def stage, where la and ye are in
complementary distribution, it might be hypothesized that the Spec to Spec
derivation in time moved to a Comp to Spec derivation, leading to a potential
[[NP ye] la] order, which through phonological absorption itself led to the
complementary distribution of la and ye. As we will see, Guyanese Creole below
offers a strikingly comparable scenario.
In contrast to OLC, NLC, as termed here in Table 1 (cf. the non-basilectal
dialect of Neumann 1985), presents both a pre-nominal definite determiner and
pre-nominal plural marking. Both of these characteristics are said to result from
the influence of Lousiana French, which is closer to French itself. As Louisiana
Creole seems at present to be rather irregular, so that a post-nominal
determination system co-exists with the newer pre-nominal one, a more detailed
study would be required to allow us to integrate this system within our proposal.
We thus leave this case for further study. NLC seems, however, to instantiate the
separation of DefP from DemP, with perhaps the Def marker in Spec DefP.
6. Guyanese C: Guy is also a post-nominal plural Creole, involving a PL head

that attracts its NP complement. Dem appears to be a Spec in Guy being always
pre-nominal. Def is a head that also attracts its complement as in other FLC. The
resulting order is derived as follows: the plural marker attracts its NP
complement ( NP > P1) and so does the Def marker ([Dem NP PL] Def) but
movement to Spec Dem does not occur, since sa is in Spec.
(43) [ DP [ Demp SA [NumP N] i YE ti ] ]j LA tj ]
This Creole is the only one with an attested post-nominal PL > Def order. The
two markers, however, seem to be distinguished only in older texts. In more
recent samples (Corpus Creole, Ludwig & Telchid 2001), there is a single
PL[+def] marker ya that results from the apparent fusion of the plural marker ye
with the definite marker (l)a. Rather than being an instance of head movement,
the only one that would occur across FLC, the fusion ye la > ya may result from
a phonological absorption due to the proximity of the two markers, assembled

through phrasal movement. If correct, this scenario would provide some
reinforcement for the above-suggested evolution of OLC in which a post-
nominal plural marker in NumP gets reanalyzed as a definite plural marker in
DefP under some absorption process.
7. Haitian C: HC has all of the makers of Table 1 in post-nominal positions.

Regarding the relations between the plural marker and the definite marker in this
FLC, however, there appear to be distinct dialects. Only the most common one,
in which the plural marker and the definite marker are in strict complementary
distribution, is considered here (the dialect in which both co-occur is considered
in work in progress). In this dialect, the demonstrative must co-occur with either
the definite marker la in the singular or with the marker yo in the plural and it
always precedes them. Note furthermore that in this FLC, perhaps more strongly
than in others, la seems strictly associated with singularity. We will thus
hypothesize that in addition to the Def feature, Def0 in HC contains a strong
Number feature that must be checked by the movement of a number marked
projection to Spec DefP.
Given (27), if we assume that the PL marke yo heads the NumP projection
in HC, the order NP Dem > P1 can be captured under a Spec to Spec derivation,
with NP moving first to Spec NumP and then subsequently to Spec DemP. For
the last step of the derivation, however, although in principle for plural nominal
expressions with yo a further Spec to Spec movement bringing NP to the Spec of
a null definite determiner would be possible, such a step appears in fact
incompatible with the Dem > Def ordering attested regularly in the singular (NP
sa (l)a). It thus seems that the last step (i.e. movement to Spec D° null) must
involve the movement of the whole DemP projection, thus involving a Comp to
Spec step topping an otherwise Spec to Spec derivation. In confirmation with
this derivation, the only attested (extremely rare) combination of the three
markers in HC manifests the order NP sa yo (I)a, an order that has to be derived
as described above, that is with NP moving Spec to Spec to Spec DemP and then
the whole DemP moving to Spec DefP. Such a derivation, however, seems oddly
unbalanced as it combines two distinct types of locality Spec to Spec, and then
Comp to Spec. We may think, however, that it is precisely this lack of balance in
the derivation that is at the source of the extreme rarity of the 3 det
combinations. The following is what I have in mind. The unbalanced derivation
suggests that this may have represented only a stage in the development of HC.
In similarity with what was suggested for OLC and GuyC above, the resulting
co-occurrence of a plural marker, present only in definite plural expressions, and
of a definite marker may have led to the phonological fusion of the two markers
72 VIVIANE DEPREZ
(as in Guyana Creole ye la→ya) subsequently followed by the reanalysis of the

plural marker as a definite plural determiner. After this reanalysis, the plural
marker  may come to be merged directly in the head of D°. Note that given
such a re-analysis, i.e. bringing yo eventually to be the head of D°, a fully
homogeneous Comp to Spec derivation then becomes possible for a plural
demonstrative expression NP sa yo: NumP moves to Spec DemP and DemP
subsequently moves to the Spec D°, containing yo. The suggestion then is that in
plural definite expressions in HC the presence of the final definite marker la has
been progressively eliminated through reanalysis of the plural marker as a
definite plural marker in the head of DefP. The result (and possibly the
motivating force for the reanalysis) is to allow a simpler homogeneous Comp to
Spec derivation, thereby eliminating the more complex mixed Spec to Spec +
Comp to Spec that was necessary to derive the demonstrative plural order NP sa
yo.
This completes our discussion on the structure and derivation of the distinct
FLC determiner ordering.
4. Conclusion
To recap, the essential features of the proposed analysis are as follows:
• The proposed nominal architecture matches rather closely the one proposed
in recent cross-linguistic works on nominal projections. The Creole data
provide interesting support for the existence of independent NumP and
DemP projections as proposed in Ritter (1994) and Bernstein (1997) or
Panagiotidis (2000) respectively.
• There is no head movement in FLC DP just as in verbal projections

(DeGraff 1998). Nominal projections, however, manifest intense and
systematic phrasal movement.
• Variations among the distinct Creoles involve two distinct factors:
1) The Spec vs. head status of functional elements: filled heads force local
movement to Spec, filled Spec prevents local movement to Spec.
2) XP movements can involve either a Comp to Spec or a Spec to Spec

derivation with the former seemingly favored somewhat over the latter. The
tendency to favor Comp to Spec derivations may be governed by Economy
considerations favoring the most stringent locality type whenever possible. What
remains to be further explored, however, is why this type of derivation, although
more common, is not always realized. I will leave this question for further
research.
References
Adone, Dany 1992. "For a DP Analysis in Mauritian Creole" Kognitionswissenschaft.
Bericht Nr 17. Oktober 1992.
Baker, Philip & Chris Corne 1982. Isle de France Creole. Ann Arbor: Karoma Publishers.
Baker, Philip 2002. ''Quelques Cas de Reanalyse et de Grammaticalisation dans
l'Evolution du Creole Mauritiern", ms.
Bernstein, Judy 1997. "Demonstratives and Reinforcers in Romance and Germanic
Languages". Lingua 102.87-113.
Bollée, Annegret 1977. Le Créole Français des Seychelles. Tübingen: Niemeyer.
Broussard. James. F. 1942. Louisiana Creole Dialect. Louisiana: Louisiana State
University Press. Baton Rouge.
Carrigton, Lawrence. D. 1984. Saint Lucían Creole: a Descriptive Analysis. Hamburg:
Helmut Buske.
Cérol, Marie-Josée 1991. Une Introduction au Créole Guadeloupéen. Pointe-à-Pitre,
Guadeloupe: Editions Jasor.
Cyrille, Odile Elzire 1997. Aspects of the Syntax of Guadeloupe Creole Nominal
Expressions. Master Thesis. Department of Modern Languages, University of
Salford.
D'Ans, André-Marcel 1968. Le Creole Français d'Haïti. The Hague-Paris: Mouton.
Damoiseau, Robert 1999. Eléments de Grammaire Comparée Français-Créole
Martiniquais. Cayenne: Ibis Rouge Editions.
DeGraff, Michel 1997. "Verb Syntax in Creolization (and Beyond)'". The New
Comparative Syntax, ed. by L. Haegeman, 64-94. London: Longman.
Déprez, Viviane & Marie-Therese Vinet 1997. "Predicative Constructions and Functional
Categories in Haitian Creole"'. Journal of Pidgin and Creole Languages 12:2.1-32.
Déprez, Viviane 1999. "De la Nature Sémantique des Nominaux sans Déterminant en Créole
Haïtien" in Langues 2:4.289-300. John Libbey Eurotext, Agence Universitaire de La
Francophonie.
2001. "On the Nature of Haitian Bare NPs". Current Issues in Romance Languages
ed. by D. Cresti., C. Tortora. & T. Satterfield, 48-64. Amsterdam & Philadelphia: John
Benjamins.
Fattier, Dominique, 1998. Contribution à lEtude de la Genèse d'un Créole: l'Atlas
Linguistique d'Haïti,Cartes et Commentaires, Thèse d'Etat, Aix-en-Provence.
2000. "Genèse de la Détermination Post-Nominale en Haïtien: l'Empreinte
Africaine". L'information Grammaticale 85:39-46.
,  & R. Zamparelli 2000. "Plurality and NP-Coordination". Proceedings of
NELS-30, Rutgers University.
Joseph, Frantz L. 1988. La Détermination Nominale en Créole Haïtien, Thèse de Doctorat
de 3 ème cycle, Université Paris VII.
Kayne, Richard 1994. The Antisymmetry of Syntax. Cambridge, Mass.: MIT Press.
Longobardi, G. "N-movement in Syntax and LF". Venice Working Papers in Linguistics
91.1-9.
74 VIVIANE DEPREZ
Ludwig, Ralph, Sylviane Telchid & Florence Bruneau-Ludwig 2001. Corpus Créole.
Hamburg: Helmut Buske Verlag.
Lyons, Christopher 1999. Definiteness. Cambridge: Cambridge University Press.
Neumann, Ingrid 1985. Le Créole de Beaux Bridge, Louisiane: Etude Morphosyntaxique,
Textes, Vocabulaire. Hamburg: Helmut Buske.
Panagiotidis, Phoevos 2000. "Demonstrative determiners and operators". Lingua 110.717-
742.
Ritter, Elizabeth 1992. "Cross-linguistic Evidence for the Number Phrase". Canadian
Journal of Linguistics 37.197-218.
Rottet, Kevin 1992. "Functional Categories and Verb Raising in Louisiana Creole. Probus
4.261-289.
1993. "The Internal Structure of DP in Mauritian Creole", ms. Indiana
University.
Schoorlemmer, Maaike 1998. "Possessors, Articles and Definiteness". Possessors,
Predicates and Movement in the Determiner Phrase, ed. by A. Alexiadou & Ch.
Wilder. 22.55-86
Syad, Anand 1996. "The Development of a Marker of Definiteness in Mauritian Creole.
Changing Meaning, Changing Functions ed. by P. Baker & A. Syea., 171-186.
London: University of Westminster Press.
Sylvain Suzanne, [1936] 1979. Le Créole Haïtien. Morphologie et Syntaxe. Geneva:
Slatkine Reprints (reprint of the edition of Wetteren-Port-au-Prince 1936)
Thràinsson, Hoskuldur 1996. "On the (Non)universality of Functional Categories".
Minimal Ideas: Syntactic Studies in the Minimalist Framework ed. by W. Abraham,
253-281. Amsterdam & Philadelphia: Benjamins.
ON THE RELATION BETWEEN FOCUS, PROSODY AND
WORD ORDER IN ROMANIAN*
EDWARD GÖBBEL
University of Tübingen
1. Introduction
Information structure, the partitioning of a proposition into focus and
background, is expressed by word order and/or prosodic means in a wide range
of languages. While certain discourse functions like topic and contrastive focus
may be encoded by movement into designated syntactic positions, typically at
the left-periphery of the clause (cf. Rizzi 1997; Drubig 2003), and can therefore
be considered core syntactic operations, there is a growing awareness that certain
movement types may have a prosodic trigger (Zubizarreta 1998) or occur in the
PF component (Holmberg 1999). In this paper we discuss the rules which govern
the focus-prosody relation in Romanian and we determine how prosody interacts
with word order. We will eventually provide an explanation for certain cases of
VP-internal scrambling operations which cannot be attributed to morphosyntac-
tic triggers. For example, the word order in (1) involves movement of a
defocused adverbial in front of the direct object. From a focus-theoretical point
of view, this construction is interesting because it is compatible with a broad
focus interpretation (i.e. it can answer a question like What did you do under
Mary's balcony?). Noncanonical word order is generally associated with narrow
focusing strategies in other Romance languages (e.g. Spanish, Italian,
Portuguese).
(1) Am cântat sub balconul ei o serenada.

have. 1 SG sung under balcony.the her a serenade
"I sang a SERENADE under her balcony."
This paper has benefited greatly from comments and suggestions by Bernhard Drubig, Remus
Gergel, Hanneke van Hoof, Wolfgang Sternefeld, Arnim von Stechow and two anonymous reviewers.
Gabi Frey and Romana Roman were of invaluable help with the Romanian data. .
76 EDWARD GOBBEL
However, this construction ceases to be puzzling if we consider scrambling as

the functional equivalent of deaccenting and allow broad foci to contain
defocused material. The fact that a broad focus may contain defocused material
has played a central role in accounts of the focus-prosody relation in Germanic
languages like English, German and Dutch, particularly within the framework of
the argument structural (AS) approach to focus structure, whose main proponents
are Schmerling (1976), Gussenhoven (1984, 1992), Selkirk (1984, 1995) and
Rochemont (1986). Although such an approach has generally been rejected for
Romance (cf. Ladd 1996 for Italian; Costa 1998, Frota 1998 for European
Portuguese; Zubizarreta 1998' for Romance generally), we argue in this paper
that the rules developed by the AS approach can also be applied to Romanian.
For Romance, it is generally claimed that focal stress is determined by
some version of the Nuclear Stress Rule (NSR), as can be found in work by
Cinque (1993) and Zubizarreta (1998). We argue that the NSR is also active in
Romanian and that it affects word order variation in the VP. In fact, the rules of
the AS approach and the NSR do not exclude each other; the former account for
the phenomenon known as F(ocus)-projection or focus-ambiguous readings,
whereas the NSR is a default accentuation rule, which applies mechanically and
which can be overridden under certain conditions.
The paper is structured as follows: section 2 introduces the basic tenets of
the AS approach and their formalization in Selkirk (1995). Several arguments
will be presented to show that Selkirk's approach faces difficulties if applied to
Romance. Zubizarreta's (1998) NSR account will be considered instead, which
accounts more naturally for the Romance data. Section 3 discusses the problem
of F-projection with respect to Romanian, whereas section 4 discusses the
application of the NSR in Romanian in order to show how it affects word order.
2. Approaches to the focus-prosody relation

Central to the AS approach is the claim that the focus-prosody relation is
mediated by argument structure. It relies on two major generalizations
concerning the distribution of accentual prominence.
The first generalization, which goes back to observations by Schmerling
(1976) and Gussenhoven (1984), can be formulated as follows: in broad focus
contexts, arguments and (at least VP) adjuncts are prominent, while predicates
may be deaccented if they have an internal argument. In (2), for example, the
verb is unaccusative and one accent on the internal argument is sufficient to
render the whole sentence focused. In (3 a), both arguments are accented,
whereas the verb may remain unaccented due to the presence of a prominent
internal argument. Adopting a term from Gussenhoven (1984), the verb and the
object form one 'focus domain'. In (3b), the verb is also prominent because there
FOCUS, PROSODY AND WORD ORDER 77
is no internal argument with which it could form one focus domain; here the
subject, verb and adverbial constitute their own focus domains.
(2) (What's happened?) [F The KEYS have disappeared].

(3) A: What's that terrible noise?
a. [F ALAN's mowing the LAWN].
b. [F ALAN's MOWING in the GARDEN].
The second generalization goes back to observations by Ladd (1980), who has
shown that prominence is retracted from the internal argument to the selecting
head if the argument cannot be accented due to its anaphoric status. In (4), the
accent is retracted from the object onto the verb, whereas in (5) from the PP to
the selecting noun. Note that both examples are cases of broad focus which
contain defocused material. The scope of the focus sensitive particle in (5) is the
VP, not the noun book.
(4) A: I've just cleared the sandwich and the other remnants of the meal
from the table.
B: [F I was EATING that sandwich]!
(5) A: Does Mary know anything about bats?
B: She even [F wrote a BOOK about bats].
These two generalizations are embodied in Selkirk's (1995) focus rules in (6-7).1
(6) Basic Focus Rule

An accented word is F-marked.
(7) Focus Projection
a. F-marking of the head of a phrase licenses the F-marking of the
phrase.
1
Note that these rules refer only to the minimal marking of focus by pitch accents. It is a well-known
fact that prefocal constituents may also be accented. For example, a peak accent can be inserted in the
pre-nuclear stretch, particularly if the focal accent comes late, as in (i). Other factors, like
phonological weight of constituents, speed and expressiveness of utterance, also have an effect on
phrasing and concomitant insertion of pitch accents. As Ladd (1996) has pointed out, narrow focus
can be distinguished from broad focus by employing an emphatic stress (i.e. increased pitch range,
intensity, etc.). This strategy, however, is not obligatory and certain patterns of prominence may be
ambiguous, particularly head-complement sequences (a cup of COFFEE) or nouns modified by
adjectives (five FRANCS). What is important for our discussion are systematically produced
prominence patterns in broad focus contexts.
(i) A: Who does Mary read the letters from her lover to?
H* H* L-L%
B: She reads the letters [F to MeLINda].
78 EDWARD GÖBBEL
b. F-marking of an internal argument of a phrase licenses the F-

marking of the head.
c. F-marking of the antecedent of a trace left by NP- or wh-
movement licenses the F-marking of the trace.
The Basic Focus rule states that a prominent word is assigned a F(ocus) feature.
The Focus Projection rules account for focus-ambiguous readings. (7a), for
example, accounts for the fact that the F feature on the prominent noun book in
(5B) projects to the NP. The focused NP itself contains the defocused
complement about bats. The verb wrote may inherit the focus feature from its
object a book about bats by rule (7b) and project it further to the VP. Rule (7c)
comes into play when the internal argument has moved from its base position, as
in the unaccusative example (2) above.
Several arguments can be adduced against applying these rules to Romance.
Firstly, adjuncts in a DP need not be accented in English, but are necessarily
accented in Romance. In the examples (8), originally due to Newman (1946),
main prominence is assigned to the verb or the noun depending on whether the
infinitival clause functions as a complement (8a) or as a relative clause (8b). In
the corresponding Italian examples (9), main prominence is invariantly assigned
to the verb.
(8) a. I have instructions to LEAVE.

(= I was instructed to leave.)
b. I have INSTRUCTIONS to leave.
(= there are instructions which I have to leave.)
(9) a. Ho istruzioni di PARTIRE.
b. Ho istruzioni da LASCIARE. (Cinque 1993:262)
Secondly, Romance lacks subject-prominent sentences. For example, thetic

sentences in European Portuguese have a prominent verb, as in (10). In English,
the verb is prominent only if it is unergative, and the subject is an external
argument.
(10) A minha carteira DESAPARECEU.

the my wallet disappeared
"My WALLET is missing." (Frota 1998:211)
Thirdly, Ladd's deaccenting rule in broad focus contexts does not apply (at least)
in Spanish and Italian (cf. Ladd 1996, Zubizarreta 1998). In the Italian example
(11) the same constituent is accented twice, which is not possible in English.
(11) Le inchieste servono a metiere a POSTO cose

the investigations serve to put to place things
andate fuori POSTO.
gone out-of place
(12) The investigations are helping to put back in ORDER things that have
got OUT of order. (Ladd 1996:176)
It has often been observed that Romance languages may have recourse to word
order variation where English resorts to contextual deaccenting. In the Spanish
example (13), the defocused PP is scrambled in front of the focused object.
(13) A: What did Ana hide under the bed?

B: Ana escondió debajo de la cama [F la MUÑECA].
Ana hid under the bed the doll
"Ana hid the DOLL under the bed." (Zubizarreta 1998:130)
In a recent investigation of the relation between focus, prosody and word order
in Romance, Zubizarreta (1998) has argued that sentence-final prominence
should be attributed to a phrasal stress rule. Specifically, the removal of
defocused material from the position of neutral stress is the result of the
interaction between the NSR and the Focus Prominence Rule (FPR) in (14) and
(15). These rules are defined over 'metrical sisters'. Simplifying somewhat, only
content words are metrical sisters regardless of whether they are syntactic
sisters.2
(14) NSR
Given two sister categories Ci, and Cj, the one lower in the asymmetric
c-command ordering is more prominent.3
2
Zubizarreta actually develops a modularized NSR which also includes argument-structural notions
in order to account for subject-prominent and a subset of V-final sentences in German (cf. (i)).
However, for examples with sentence-final prominence like (ii), the NSR in (14) is assumed to assign
main prominence. Her approach is highly complicated and involves several pages of definitions,
which we cannot discuss here due to limitations of space. We only note that (ii) does not allow a
broad focus reading unless the adverbial is also accented. In fact, German is similar to English
regarding nonintegrativity of adjuncts with verbs into one focus domain (cf. Winkler & Göbbel 2002
for a detailed critique of her account of German accentuation).
(i) Peter hat an einem AUFSATZ gearbeitet.
Peter has on a paper worked
(ii) Peter hat an einem Tisch GEARBEITET.
Peter has at a table worked
3
Asymmetric c-command is defined as in (i). The nodes relevant for this definition are constituents
which are visible for the syntactic computation (i.e. heads and maximal projections); therefore, it can
be said that a specifier of a head X asymmetrically c-commands X and also an intermediate
80 EDWARD GÖBBEL
(15) FPR
Given two sister nodes Cj (marked [+F]) and Cj (marked [-F]), Ci is
more prominent than Cj.
Under this approach, the NSR determines prominence on the most deeply
embedded constituent in examples with broad focus like (10). In examples with
narrow focus, the NSR and the FPR may give rise to conflicting prominence
assignments whenever a sentence-final constituent is defocused. This conflict is
resolved either by movement of the defocused constituent in front of the focused
constituent, as in (13), or by analysing defocused constituents as 'metrically
invisible'. Zubizaretta claims that French examples like (16), but also their
English and German counterparts, reflect the latter process. After movement of
the defocused constituent or its reanalysis as metrically invisible, the NSR can
reapply and assign main prominence to the focused constituent. In other words,
the NSR does not 'see' anaphoric constituents in languages which allow
contextual deaccenting.
(16) A: What did you put on the table?

B: Nous avons mis TROIS LIVRES sur la table.
we have put three books on the table
The main consequence of Zubizarreta's approach is that focal accent (even if not
sentence final) is eventually subsumed under the NSR, despite the fact that this
rule does not refer to focus (only the FPR does). She further argues that only
contrastive and emphatic focus are instances of free stress assignment. It is
claimed that emphatic focus has a purely metagrammatical function, whereas
contrastive focus is partly metagrammatical and partly focus-related. This
distinction between information focus and contrastive focus is reminiscent of the
normal and contrastive stress patterns of the early generative tradition. Like
Chomsky (1972), she assumes that contrastive focus and presumably all
sentences containing operators which interact with focus (e.g., negation),
constitute a completely different phenomenon from information focus, subject to
different interpretation and stress-assignment rules.
However, we disagree with several aspects of her approach. Firstly, the
notion of 'metrical invisibility' is not independently defined or embedded in
some current theory of metrical phonology and remains an artefact (cf. Mörnsjö
1999 and Winkler & Göbbel 2002). Secondly, an NSR approach does not seem
projection of X. The latter is not visible for the syntactic computation and therefore does not c-
command the specifier.
(i) a asymmetrically c-commands ß =dcf a c-commands ß and ß does not c-command a. (p. 35)
to be adequate for all Romance languages. French, for example, differs from
other Romance languages like Portuguese, Italian and Romanian, and also from
Germanic languages, in that focus is not marked by pitch accents, but by
boundary tones.4 Furthermore, tonal events are often observed in the postfocal
stretch (Le Gac & Yoo 2002) which, among other factors, may be conditioned by
the metrical organization of the postfocal material (Delais-Roussarie et al. 2002).
Therefore, both the application of a phrasal stress rule and a notion like 'metrical
invisibility' must be seriously called into question for this language.
Thirdly, given the fact that the notion 'metrical invisibility' has to be
rejected, noncontrastive focal accent in sentence-nonfinal position cannot be
subsumed under the NSR either, and must be a case of free stress assignment.
This conclusion will be reinforced in section 4 below where we show that the
NSR in Romanian is only operative in broad focus contexts. We do, however,
accept Zubizarreta's argument that word order variation may be triggered by a
phrasal stress rule. Before addressing the application of a phrasal stress rule in
Romanian, we turn to focus projection.
3. The issu e of focus projection

Focus in Romanian is identified at PF in terms of pitch accents. In fact, the
asymmetry in prominence assignment to predicate + internal argument vs.
predicate + external argument/adjunct sequences, which is embodied in Selkirk's
rules, is observed in Romanian in terms of type of pitch accent, not necessarily in
terms of presence vs. absence of accentual prominence. In order to show this, we
have to introduce some basic aspects of intonation in declarative sentences. For
the phonological representation of intonational contours we adopt the auto
segmental theory as developed by Pierrehumbert (1980). The notation will
follow the ToBI Guidelines (Beckman & Ayers 1994, Beckman & Hirschberg
1994), which incorporate certain revisions of the original Pierrehumbert system.
Romanian employs two simple accents, a peak accent H* and a low accent
L*, and four bitonal accents, of which one is rising (L+H*) and three are falling
(H+!H*, H+L* and H*+L). Furthermore, two phrasal tones, which mark the end
of an intonational phrase (L% and H%), and one initial boundary tone (%H) are
distinguished. We will only illustrate what we need for the discussion here (cf.
Göbbel 2001 for further details). Narrow focus, for example, has two different
instantiations. If it is final in the intonational phrase (IP), it is associated with a
4
We are indebted to an anonymous reviewer for pointing out this problem. French being a language
without lexical stress, focus is mainly signalled by phrasing, as argued, for example, by Féry (2001).
The rough equivalent of deaccenting in French is dephrasing (i.e. lack of any tonal events). Féry
concludes that "prefocal constituents are always phrased, but postfocal ones are mostly (but crucially
not always) dephrased" (p. 171).
82 EDWARD GÖBBEL
falling tone (H+!H*, step down to mid from a high peak). If it is nonfinal in IP, it
is associated with a high fall (H*+L). The two answers in (17) and the associated
contours in Figure 1 illustrate these two accent types (accented syllables are
capitalized).
(17) A: Cine vine? ("Who is coming?")

H+!H* L%
a. Vine [j- MaRIAna].
come.PRES.3SG Mariana
"MARIANA is coming."
H*+L L%
b. [F MaRIAna] vine.
Figure 1: F()-contours of examples (17a) and (17b)
In longer, broad focus utterances, the nuclear accent is generally realized as an

H+!H* tone and pre-nuclear accents are either peak accents (H*) or rising peak
accents (L+H*). As can be seen in Figure 2, a sample pitch extraction of example
(18), the contour is declining with compression of pitch range. What is
noteworthy here is that the last two accents form a 'hat pattern'.
(18) A: Ce se întâmplă în bucătărie?

"What's going on in the kitchen?"
L+H* L+H* H+!H* L%
B: MaRIA face  prăjiTUră pentru GEORge.
Maria makes a cake for George
Figure 2: F() contour of example (18)
Because the nuclear accent in declarative sentences is generally a falling one, the
contour has to go up first. This explains why examples like (19), which would
hardly conform to the rules of the AS approach, generally have a pre-nuclear
accent (typically a peak accent), regardless of whether the verb is part of the
focus or not. This sentence with the indicated accentual pattern is focally
ambiguous; it may answer either Who did he give the pot to? or What did he do
with the pot?. Evidence that H* is the default accent in Romanian, whose
insertion is determined by the general rise-fall pattern in the nuclear stretch,
comes from examples with more than one defocused constituent preceding the
focused one. In (20) the peak accent is associated with the adverb, not the verb.
Sample pitch extractions of these two examples can be seen in Figure 3. Pre-
nuclear accents which are not predicted by Selkirk's rules are not necessarily
counterexamples to her approach, but the conditions under which they occur
have to be stated (cf. also fn. 1).
H* H+!H* L%
(19) [FI-a DAt-o [F OAnei]].
CL.DAT-has given-CL.ACC Oana.DAT
"He gave it to OANA."
H* H+!H* L%
(20) {Ce limbă vorbeşte bine?) Vorbeşte BIne [F daNEza].
("Which language does he speak well?") speaks well Danish.the
"He speaks DANISH well."
84 EDWARD GÖBBEL
400T"""------.-----., 500T"""---,...----,----.,
400
300 :
300
~M.-..fl\-
i::i: '
t:J 200
;!;.
~
;,:
100
-..f""
100+------+-----1
io.DATo O.'.".i vo r b. s t. BI". d 0. r; E z 0.
W H !W H 'w
0.92712 1.47551
TI",. (s) TI",. (s)
Figure 3: Fo-contours of examples (J 9) and (20)
Once we abstract away from simple peak accents, focus intonation in Romanian
conforms to the rules of the AS approach. Arguments are generally associated
with perceptually more prominent bitonal ac cents (L+H* and H+!H*), whereas
verbs may be deaccented (cf. Figure 2 above) or, more commonly, associated
with the default H* accent, as in (21).
(21) A: Jmi dai mie statuia din bronz?

"Will you give me that bronze statue?"
L+H* H* H+!H* L%
B: Nu. MaRlA i-a proMIs-o MaRlAnei.
no Maria CL.DAT-has promised-cL.ACC Mariana.DAT
"No. Maria promised it to Mariana."
Verbs followed by internal arguments are associated with an H* accent or (less

commonly) deaccented, as in (22). Abitonal L+H* accent on the verb sounds
decidedly exaggerated or emphatic in this case. However, bitonal accents may be
associated with the verb if it is followed by an adverbial or an 'external
argument', as can be seen in (23) and (24). In these cases the verb cannot be
deaccented. Two pitch tracks are provided in Figure 4. The sequence H*+L
H+!H* in (24) gives rise to a terraced contour.
H* H+!H* L%
(22) (Ai auzit vestea?) VIne MAma.
("Have you heard the news?") come.PRES.3SG mother.the
"MOTHER's coming."
L+H* H+!H* L%
(23) {Ce face Ion?) Se PLIMbă în grăDInă.
("What is John doing?") CL.REFL walks in garden
"He's taking a WALK in the GARDEN."
H*+L H+!H* L%
(24) (Ce-i zgomotul acela?) PLÂNge coPIlul.
("What's that noise?") cry.PRES.3SG child.the
"The baby's crying."
Figure 4: F() contours of examples (23) and (24)
Further evidence that Romanian focus intonation conforms to the rules of the AS
approach is the fact that the nuclear accent may be retracted from an internal
argument to the selecting head if the argument has an antecedent in the
discourse. In other words, broad focus may contain defocused material like in
English. In the following examples, an internal argument (direct object in (25),
PP complement in (26), and sentential complement in (27)) cannot be accented,
but we have maximal 'projection' of focus.
(25) A: Am strâns sandvişul şi celelalte resturi de mancare de pe masa.

"I've cleared the sandwich and the other remnants of the meal
from the table."
H*+L L%
B: [FAm vrut sămăNÂNC sandvişul acela]!
have. 1SG wanted to eat.SUBJ. 1SG sandwich.the that
"I wanted to EAT that sandwich!"
86 EDWARD GÖBBEL
(26) A: Sepricepe Luminita la lilieci?

"Is Luminita well-informed about bats?"
: Cum să nu. [F A scris şi  ARte despre lilieci].
of course has written even a book about bats
"Of course. She even wrote a BOOK about bats."
(27) (A vertit George la petrecere?) [F A VRUT ă vină].
("Did George come to the party?") has wanted to come.SUBJ.3SG
"He WANTED to come."
Concluding this section, we have argued that only bitonal accents signal focus,
while simple peak accents are generally inserted due to the rise-fall pattern of the
intonational contour. Furthermore, we have shown that broad foci may contain
defocused material; deaccenting of an internal argument without giving rise to
narrow focus on the selecting head is actually one of the main motivations for
Selkirk's focus projection rules. In the next section, we discuss apparent
counterexamples to the AS rules and we argue that a phrasal stress rule is also
operative in Romanian.
4. The NSR as a trigger for scrambling

Apparent counterexamples to an AS approach in Romanian are examples
like (28) and (29), in which focus can only be narrow (cf. the context questions
in square brackets). Selkirk's rules predict that the F feature on the prominent
internal argument can be inherited by the verb and project up to VP and higher.
This would, for example, integrate the defocused adverbial into the VP-focus in
(28).
(28) Am cântat o SERENADĂ sub balconul ei.

have. lSG sung a serenade under balcony.the her
"I sang a SERANADE under her balcony."
[What did you sing under her balcony?]
[*What did you do under her balcony?]
(29) Ar fi fost ION arestat.
would be been John arrested
"JOHN would have been arrested."
[Who would have been arrested?]
[*What would have happened?]
We believe that such examples are not counterexamples to an AS approach,

because broad focus generally requires sentence-final prominence in Romanian.
In other words, a phrasal stress rule is operative in this language. If the nuclear
accent is assigned to the participle in (29), the sentence may have a broad focus
interpretation. Furthermore, defocused constituents are not readily tolerated in

sentence-final position if they can be 'removed', for example by cliticization, as
in (30). The phrasal stress rule, however, is a violable intonational constraint. It
cannot override independent grammatical restrictions: in examples (25) through
(27) focus is broad and one would expect sentence-final prominence. However,
the defocused constituents cannot scramble or undergo deletion without violating
conditions on movement or recoverability. Therefore, prominence is retracted to
the selecting head without giving rise to a narrow focus interpretation.
(30) A: Nu găsesc ziarul ("I can't find the newspaper.")

a. #[FA aruncat MARIANA ziarul].
has thrown-away Mariana newspaper.the
b. [FL-a aruncat MARIANA].
CL.ACC-has thrown-away Mariana
"Mariana threw it away."
It seems to be the case that the phrasal stress rule is not operative in narrow focus
contexts. Scrambling effects, which have been attributed to the NSR in other
Romance languages are not observed in Romanian. For example, adverbials
scramble in Spanish (31), but not in Romanian (32). Objects shift across manner
adverbs in European Portuguese (33), but resist movement in Romanian (34).
(31) A: What did John plant in the garden? (Zubizarreta 1998:134)

B: Juan plantó en el jardín [F un ROSAL].
Juan planted in the garden a rosebush
(32) A: Ce ai plantat în grădină? ("What did you plant in the garden?")
a. Am plantat [F un TRANDAFIR] în grădină.
have.1SG planted a rosebush in garden
"I planted a ROSEBUSH in the garden."
b. #Am plantat în grădină un TRANDAFIR.
(33) A: How does Paolo speak French? (Costa 1998: 178)
a.  Paolo fala francês bem.
Paolo speaks French well
b. #0 Paolo fala b em francês.
(34) A: Cum vorbeşte daneza? ("How does he speak Danish?")
a. Vorbeşte fF BINE] daneza.
speak.PRES.3SG well Danish.the
"He speaks Danish WELL."
b. ?* Vorbeşte daneza [F BINE].
88 EDWARD GÖBBEL
However, the NSR applies with full force in broad focus contexts. The adverbials
in (35) and (36) have to scramble in front of the direct object. If they do not
scramble, only a narrow focus interpretation on the prominent argument is
available.
(35) A: Ce ai făcut în grădină? ("What did you do in the garden?")

a. [F.-Am plantat în gradina un TRANDAFIR].
have. 1SG planted in garden a rosebush
b. #Am plantat un TRANDAFIR în gradina.
(36) A: De ce e supărată Maria? Ce ai făcut sub balconul ei?
"Why is Mary angry? What did you do under her balcony?"
a. [F Am căntat sub balconul ei o SERENADA].
have. 1SG sung under balcony.the her a serenade
b. #Am cântat o SERENADA sub balconul ei.
These examples clearly show that noncanonical word order is compatible with
broad focus. The rules of the AS approach require the internal argument to be
prominent. The NSR, on the other hand, blindly assigns prominence to the last
lexical item in the intonational phrase. Therefore, sentence-final defocused
constituents will be removed from the position to which nuclear stress is
assigned. In Zubizarreta's (1998) approach, outlined in section 2, deaccented
constituents are analysed as 'metrically invisible'. This notion is only needed in
order to attribute prominence assignment in both narrow and broad focus
contexts to the NSR. In other words, the NSR does not 'see' defocused material
if the language allows contextual deaccenting. Such an approach is clearly
problematic for Romanian since one would have to assume that defocused
constituents are analysed as metrically invisible only in narrow focus contexts,
but not in broad focus contexts. Contextual deaccenting in clause final position
actually means that the NSR does not apply in this case.
Following suggestions by Frascarelli (1999), we formulate the NSR as a PF
rule. The first condition in (37), which replaces Zubizarreta's Focus Prominence
Rule in (15), states that a word bearing NS (i.e. right peripheral prominence in
the intonational phrase) is a focus or part of the focus. This condition is needed
in order to account for scrambling in (35) and (36) above. The second condition
states that the domain of the NSR is the intonational phrase.5
3
Note that examples like (iB) are split up into two intonational phrases, each with its own nuclear
accent. Formulating the NSR in terms of depth of embedding would wrongly predict main
prominence on probleme because the noun does not c-command the adjunct clause (thanks to Richard
Kayne for pointing out this problem for a syntactic formulation of the NSR).
(i) A: De ce nu vi la prelegere7 ("Why aren't you coming to the lecture?")
(37) NSR
a. A word bearing NS is a focus or part of the focus.
b. NS is assigned to the rightmost lexical item in the intonational
phrase.
4.1 Where does phonologically-driven movement occur?
Both the syntactic component and PF have been proposed as levels at which
defocused constituents are moved (cf. Zubizarreta 1998 and Holmberg 1999,
respectively). In this section we briefly consider some evidence that
phonologically-driven movement (p-movement) applies in the syntax. The idea
is that, if there is evidence that the NSR can affect core syntactic operations, then
p-movement occurs in the syntax. At least one syntactic operation is affected by
the NSR, namely, dative shift.
In double object constructions in which both internal arguments are F-
marked either order is possible. This can be seen in (38). Following proposals by
Larson (1988), Baker (1997) and many others, we assume that the indirect object
(IO) is the lower argument in the underlying representation.
(38) A: Ce face Maria? ("What is  doing?")

a. Maria [¡Ai face  prănuraălui George."
Maria CL.DAT makes a cake DAT George
"Maria is baking George a cake."
b. Maria [F îi face lui George o prăjitură].
However, the NSR can either force or block movement of the IO. In broad focus
contexts in which the IO is construable from the context, it has to move in front
of the direct object (DO). As shown in (39), the IO targets a position between the
VP-internal subject and the DO. In (40), the IO cannot shift because it would
give rise to a configuration in which the defocused DO receives nuclear stress.
(39) A: De ce este Anuta fericită? ("Why is Anuta happy?")

a. [FI-a dat Ion Anutei o VERIGHETÄ].
CL.DAT-has given John Anuta.DAT a wedding-ring
"John has given Anuta a wedding-ring."
b. #I-a dat Ion o VERIGHETĂ Anutei.
c. ??I-a dat Anutei Ion o VERIGHETÄ.
H+!H* L% H+!H*L%
 : [IP Am nişte proBLEme] [IP de rezolVAT].
have. l SG some problems of solve.SUPINE
'T have some PROBLEMS to solve."
90 EDWARD GÖBBEL
(40) A: Ce afàcut Maria cujurnalul ei?

"What did Mary do with her diary?"
a. I-a dat jum alui ei lui ION.
CL.DAT-has given diary.the her DAT John
"She gave her diary to John."
b. #I-a dat lui ION jurnalul ei.
Dative shift in Romanian must be considered a case of A-movement. As can be

seen in the examples (41) from Gierling (1996), dative shift feeds binding of
variable pronouns. We assume that the 10 moves to the specifier of a light verb
projection, sandwiched between VoiceP, which hosts the external argument (cf.
Kratzer 1994), and the projection of the lexical verb, in which both arguments
are theta-marked (cf. 42).
(41) a. I-am dat [fiecärui prieten], cartea lui,.

CL.DAT-have.lSG given each.DAT friend book.the his
"I've given each friend his book."
b. *I-am dat cartea lui, [fiecärui prieten],. (Gierling 1996:116)
(42) [TP i-a datį [VoiceP 1(^ ti [VP Anuteij t, [VP  verighetä [ v -t¡ tj]]]]
CL.DAT-has given John Anuta.DAT a wedding-ring
Since the NSR can affect core syntactic operations, p-movement must occur in
the syntax. The displaced PP-adverbials in (35) and (36) are probably adjoined to
VP.
5. Concluding remarks
This paper has explored the relation between focus, prosody and word order
in Romanian. By concentrating on information focus, we argued that Romanian
focus intonation can be submitted to an argument structural approach. In
particular, we have shown that the asymmetry in prominence assignment to
predicate + internal argument vs. predicate + external argument/adjunct
sequences, which is embodied in Selkirk's (1995) focus rules, is observed in
Romanian in terms of type of pitch accent. These rules are also needed to
account for the fact that broad focus may contain defocused material. In the
second part of this paper we have argued that the NSR is active in broad focus
contexts and affects the position of adverbials and indirect objects.
An important question not addressed so far is the relation between the two
rule systems. We believe that Selkirk's focus rules do not apply only in Germanic
languages and Romanian, but are more universal if prominence for the purpose
of F-projection is understood to include syntactic and/or morphological marking.
In Hungarian, for example, focused constituents are moved to a structural focus

position, yet only an object in this position, but not a subject or adjunct, can
signal sentential focus (Kenesei 1998). In Somali, a language with
morphological focus marking, a focus marker on an object can signal VP and S-
focus; focus marking on the verb can signal VP-focus just in case the object is
defocused (Svolacchia et al. 1995). The NSR, on the other hand, is an
independent accentuation rule applying mechanically in most Romance
languages (and also Swedish, cf. Mörnsjö 1999). The application of the F-
projection rules in Romance languages has not been noticed because the NSR
assigns main prominence to the position where internal arguments are typically
found.
References
Baker, Mark 1997. "Thematic Roles and Syntactic Structure". Elements of Grammar ed.
by Liliane Haegeman, 73-137. Dordrecht: Kluwer.
Beckman, Mary E. & Gayle M. Avers 1994. "Guidelines for ToBI Labelling", version 2.0.
Unpublished manuscript and accompanying speech materials, Ohio State University.
& Julia Hirschberg 1994. "The ToBI Annotation Conventions". Unpublished
manuscript and accompanying speech materials, Ohio State University.
Chomsky, Noam 1972. "Deep Structure. Surface Structure and Semantic Interpretation".
Studies on Semantics in Generative Grammar ed. by Noam Chomsky, 62-119.
Cambridge: Cambridge University Press.
Cinque, Guglielmo 1993. "A Null Theory of Phrase and Sentence Stress". Linguistic
Inquiry 24.391-444.
Costa, João 1998. Word Order Variation: A constraint based approach. Doctoral
dissertation, HIL/Leiden University. The Hague: Holland Academic Graphics.
Delais-Roussarie, Elisabeth, Annie Rialland, Jenny Doetjes & Jean-Marie Marandin 2002.
"The Prosody of Post-focus Sequences in French". Proceedings of Speech Prosody
2002, Aix-en-Provence, France, ed. by Bernard Bel & Isabelle Marlien.
Drubig, Bernhard H. 2003. "Towards a Typology of Focus and Focus Constructions".
Linguistics 41.1 -40.
Féry, Caroline 2001. "Focus and Phrasing in French". Audiatur Vox Sapientiae: A
Festschrift for Arnim von Stechow ed. by Caroline Féry & Wolfgang Sternefeld,
153-81. Berlin: Akademie Verlag.
Frascarelli, Mara 1999. "The Prosody of Focus in Italian (and the Syntax-Phonology
Interface)". Probus 11.209-238.
Frota, Sónia 1998. Prosody and Focus in European Portuguese. Doctoral dissertation,
University of Lisbon.
Gierling, Diana 1996. "Further Parallels between Clitic Doubling and Scrambling".
ESCOL '96.113-123.
Göbbel, Edward 2001. Syntactic and Focus-structural Aspects of Triadic Constructions.
Doctoral dissertation, University of Tübingen.
Gussenhoven, Carlos 1984. On the Grammar and Semantics of Sentence Accents.
Dordrecht: Foris.
92 EDWARD GÖBBEL
1992. "Sentence Accents and Argument Structure". Thematic Structure: Its Role
in Grammar ed. by Iggy M. Roca, 79-106. Berlin: Foris.
Holmberg, Anders 1999. "Remarks on Holmberg's Generalization". Studia Linguistica
53.1-39.
Kenesei. István 1998. "Adjuncts and Arguments in VP-Focus in Hungarian". Acta
Linguistica Hungarica 45.61-88.
Kratzer. Angelika 1994. "On External Arguments". Functional Projections ed. by Elena
Benedicto & Jeffrey T. Runner. 103-130. Amherst: University of Massachusetts
Occasional Papers 17.
Ladd. Robert D. 1980. The Structure of Intonational Meaning: Evidence from English.
Bloomington: Indiana University Press.
1996. Intonational Phonology. Cambridge: Cambridge University Press.
Larson. Richard 1988. "On the Double Object Construction". Linguistic Inquiry 19.335-
391.
Le Gac. David & Hi-Yon Yoo 2002. "Intonative Structure in French and Greek". Romance
Languages and Linguistic Theory 2000 ed. by Claire Beyssade, Reineke Bok-
Bennema, Frank Drijkoningen & Paola Monachesi. 213-231. Amsterdam &
Mörnsjö. Maria 1999. "Theories on the Assignment of Focal Accent as Applied to
Swedish". Working Papers in Scandinavian Syntax 64.37-78.
Newman. Stanley 1946. "On the Stress System of English". Word 2.171-187.
Pierrehumbert. Janet B. 1980. The Phonology and the Phonetics of English Intonation.
Ph.D. dissertation, MIT.
Rizzi, Luigi 1997. "The Fine Structure of the Left Periphery". Elements of Grammar ed.
by Liliane Haegeman, 281-337. Dordrecht: Kluwer.
Rochemont, Michael S. 1986. Focus in Generative Grammar. Amsterdam & Philadelphia:
John Benjamins.
Schmerling. Susan F. 1976. Aspects of English Sentence Stress. Austin: Texas University
Press.
Selkirk. Elisabeth O. 1984. Phonology and Syntax. Cambridge, Mass.: MIT Press.
1995. "Sentence Prosody: Intonation, Stress, and Phrasing". The Handbook of
Phonological Theory ed. by John A. Goldsmith. 550-569. Cambridge, Mass.:
Blackwell.
Svolacchia, Marco, Lunella Mereu & Annarita Puglielli 1995. "Aspects of Discourse
Configurationality in Somali". Discourse Configurational Languages ed. by Katalin
È Kiss, 65-99. Oxford: Oxford University Press.
Winkler, Susanne & Edward Göbbel 2002. "Focus, P-movement, and the Nuclear-stress
Rule: A view from Germanic and Romance." Linguistics 40.1185-1242.
Zubizarreta, Maria L. 1998. Prosody, Focus, and Word Order. Cambridge, Mass.: MIT
Press.
ECONOMY OF STRUCTURE
THE CASE OF SUBJECT CLITICS IN PIEDMONTESE
CECILIA GORIA
University of Nottingham
1. Introduction
Piedmontese as other Northern Italian Dialects (NIDs) has Subject Clitics
(henceforth SCLs), i.e. clitic elements which encode subject features and show
up between preverbal subjects and finite verbs. Since Rizzi (1986), SCLs have
been analysed as elements of agreement. This together with the variation typical
of the morphological and distributional properties of these elements across the
NIDs has justified the claim that SCLs occupy multiple agreement projections.
Most recently, Poletto (2000) posits an Agreement Field which involves four
different projections designated to host different morphological types of SCLs.
These projections extend from the IP to the CP domain. In this framework the
distributional properties of these elements are also derived from structural
differences. According to Poletto, Piedmontese SCLs are merged in the CP
section of the Agreement Field, with the exception of SCL -t of 2sg it/at (see
Table 1 below) which is merged in the IP section. This claim is motivated by the
morphological content of these clitics ('deictic clitics' and 'person clitics'
respectively) and by properties such as optionality, interaction with elements in
CP, omission in coordination, which are all claimed to be related to specific
positions inside the structure.1
In this paper, arguing against Poletto's (2000) claim that Piedmontese has
two morphologically and structurally different SCL types, I propose that at least
some of the properties that motivate the Agreement Field can be accounted for
The reader is referred to Poletto's (2000) own work for more details on the Agreement Field and to
Goria (forthcoming) for a detailed discussion against its motivations.
94 CECILIA GORIA
without resorting to structural complexity.2 Adopting Chomsky's (1995, 2000,

2001) Agr-less structure, I establish the position and function of Piedmontese
SCLs in T and account for the morphological and distributional properties of
these elements in two varieties of Piedmontese: Turinėse and Astigiano. I
propose an analysis which relates the distributional properties of SCLs to their
morphological make-up without relying on multiple agreement projections. More
precisely, I claim that the distribution of Piedmontese SCLs is the result of the
interaction of morphological constraints concerned with the overt realization of
agreement features. This is done in accordance to the general principles of
Optimality Theory.
The assumption underlying my proposal is the view that the morphological
realisation of agreement features pertains to a component of grammar distinct from
narrow syntax. This allows me to suggest that different types of rules apply
inside the two components. Thus, this paper juxtaposes the Minimalist Program
and Optimality Theory and provides an account for the morphosyntactic
properties of subject clitics in Piedmontese by taking the mechanisms of these
two formalisms to work in a complementary fashion. The result is the
minimisation of structural complexity.
2. The proposal: position and function of Piedmontese SCLs

In the wake of Chomsky's T-model, I propose that Piedmontese SCLs are
merged as separate heads D adjoined to T, where they overtly express the
features added to T as it is drawn from the lexicon (section 2.1), i.e. T's EPP
feature and its  features.3 Furthermore, I maintain that SCLs do not function as
checkers of the EPP (section 2.2). In support of the former claim, I argue that
unmarked preverbal subjects are in Spec TP. That is, neither are they left
dislocated (Alexiadou & Anagnostopoulou 1998; Barbosa 1995, 2000, among
others), nor are they in a topic-like position inside the left periphery of the clause
(Poletto 2000) Hence, the order S SCL V reveals that SCLs are inside the TP.
2.1 Unmarked Preverbal Subjects in Italian and the NIDs: arguments against
the left dislocation analysis
Beginning with the dislocation analysis, I refer to Barbosa's (1995, 2000)
claim that in Null Subject Languages (NSLs) preverbal subjects, with the
exception of contrastively focussed DPs and bare QPs, are base-generated in a
2
I must point out that this work on Piedmontese intends to be the first step towards a larger project in
which the theoretical approach presented here is tested against different morphological types of SCLs
and different morphosyntactic phenomena.
3
See Ledgeway (2000) and Torrego (1998a) for a similar analysis of Object Clitics in v.
ECONOMY OF STRUCTURE 95
dislocated position and linked to a resumptive pro contained inside the clause
(Cinque 1990, 1997). The literature contains a number of arguments against this
idea. Below those that can be directly related to the NIDs are listed (1-3) and
some new ones are added (4-5).4
1. Poletto (2000) argues that the fact that in Venetian SCLs are required with left
dislocated subjects but are ungrammatical otherwise reveals in general the need
of distinguishing left dislocated from non-dislocated subjects.
2. Cardinaletti (1997) referring to (1) and (2) below from Conegliano argues that
the choice of the SCL la is determined by the presence of a preverbal subject in
the canonical subject position: pro in (1), la Maria in (2).
(1) La a telefona, la Maria.

SCL has telephoned, the Mary
(2) La Maria la a telefona.
the Mary SCL has telephoned
3. Cardinaletti (1997) observes a pragmatic/semantic contrast between left

dislocated and non-left dislocated subjects in standard Italian.5 The same
interpretational differences are found in Piedmontese. The following are the
direct Piedmontese equivalents of Cardinaletti's examples:
(3) I Thai vist ier a la Biennale 'n film su Wim ¡Venders. 'L regista, Pier
a l'ha peui ancontralo al bar n'ora pi tarde.
"I saw yesterday at the Biennale a film on Wim Wenders. The film
director, Piero saw him in the bar an hour later."
(4) I l'hai vist ier a la Biennale 'n film su Wim Wenders. Dop la
proiesion, 'l regista  l'ha avu 'n premi a la carera.
"I saw yesterday at the Biennale a film on Wim Wenders. After the
show, the film director received a prize for the career."
(5)  l'han premia 'n film su Wim Wenders. 'L regista, al premi, a l'ha
avulo dal ministro.
"They have awarded a prize to a film on Wim Wenders. The film
director, the prize, has received it from the Minister."
In (3) the direct object 7 regista is left dislocated and it can only refer to the
director of the film shown at the Biennale. Similarly, in (5) where the subject
4
See also Costa (1998) for the same claim with respect to European Portuguese.
5
Cardinaletti (1997) provides additional arguments against the dislocation analysis with reference to
standard Italian. These have been left out, here, as they are not directly concerned with the NIDs.
96 CECILIA GORIA
7 regista is left dislocated since it precedes a left dislocated object, it has the
same interpretation as in (3). On the other hand, in (4) the subject 7 regista may
refer to the director of the film shown at the Biennale as well as to Wim
Wenders. This, as suggested by Cardinaletti (1997) for standard Italian, marks
the difference between dislocated and non-dislocated subjects.
4. Goria (forthcoming) takes the following grammaticality contrast as further
argument in favour of distinguishing dislocated from non-dislocated subjects:
(6) Non capisco perché tutti i bambini, la pasta, la mangiano. solo

guando è fr edda.
"I do not understand why all children, pasta, they eat it only when it is
cold."
(7) Non capisco perché i bambini, la pasta, la mangiano tutti solo quando
è fredda.
"I do not understand why all children, pasta, they eat it only when it is
cold."
(8) *Non capisco perché i bambini, la pasta, tutti la mangiano solo
quando è fredda.
In (6) the quantified subject tutti i bambini is dislocated as it precedes the

dislocated object la pasta. According to Cinque (1997), a dislocated XP as tutti i
bambini is connected to a resumptive pro, here in subject position. In (7), i
bambini is dislocated and the quantifier tutti modifies the trace in Spec Light-vp
of the resumptive null phrase, which has abandoned the quantifier (Sportiche
1988) and moved into Spec TP for licensing reasons (Cardinaletti 1997). As for
(8), its ungrammaticality reveals the inadequacy of the connection between the
dislocated subject i bambini and a resumptive pro modified by tutti if the QP
[tutti pro] has risen to Spec TP. This is not surprising given Cinque's (1997)
claim that dislocation is subject to island conditions. In contrast, sentences (9)
and (10) below indicate that the extraction of the DPs i bambini and gli studenti
from a dislocated [tutti-XP] is grammatical:
(9) Non capisco perché i bambini, la pasta, tutti, all'età di tre anni, la
mangiano solo quando è fredda.
"I don't understand why all children, pasta, at the age of three, they
eat it only when it is cold."
(10) Sono sicura che gli studenti, questo libro, tutti, per l 'esame di fine
anno, lo compreranno e lo legger anno attentamente.
"I am sure that all the students for the end of the year examination will
buy and read this book carefully."
The relevant point here is that while extraction of the subjects i bambini and gli
studenti is allowed from a dislocated position, it is not from the [tutti-XP] in
Spec TP. Clearly this contrast is not predicted if the subject position is regarded
as being always dislocated.
5. Finally, if all preverbal subjects are left dislocated, it is difficult to distinguish
marked subjects from subjects not marked as topic. In other words. how is the
contrast between (11) and (12) accounted for?
(11) Gianni, Maria conosce il motivo per cui è scappato.

''Gianni, Maria knows the reason why he has left."
(12) Maria conosce il motivo per cui Gianni è scappato.
"Maria knows the reason why Gianni has left."
Under the dislocation analysis Gianni in (11) as well as in (12) is marked as

[+topic] and connected to a resumptive pro inside its clause, predicting that these
two distinct structures have identical numerations. Under Minimalist
assumptions such a situation never arises, since Economy would ban (11). where
the long movement of Gianni is less economical.
Having presented some arguments against the left dislocation analysis of
unmarked preverbal subjects, I now look at the subject in CP analysis (Poletto
2000).
2.2 Subject in CP
Poletto (2000) suggests that lexical preverbal subjects in Italian and the
NIDs are not left dislocated but raise cyclically to a topic-like position in CP: a
CP subiect. In this section, I provide arguments against this idea.
Firstly, if the raising of the subject to CPsubject is due to a feature [+topic],
Poletto's (2000) claim about the unmarked character of preverbal subjects in
Spec CP is weakened. In fact, any feature responsible for movement into the Left
Periphery would mark the subject.
Secondly, the subject-in-CP analysis takes as compelling evidence
sentences like the Piedmontese suppletive imperatives illustrated in (13) below:
(13) Mario ch' a parta subit.

Mario that SCL leaves soon
"Mario is to leave soon."
The position of the subject Mario in (13) on the left of the complementizer che is
taken to leave no doubt that the unmarked position of the subject is inside the CP
98 CECILIA GORIA
layer.6 However, this analysis is weakened by the fact that (13) above is marked,
and so is the placement of the subject to the left of che. More precisely, (13) is to
be compared with (14) below, which I take to be its unmarked counterpart, still
retaining its imperative value:
(14) Che Mario a parta subit.

that Mario SCL leaves soon
"Mario is to leave soon."
M. Parry (p.c.), suggests that in suppletive imperatives like (13) the subject
is focussed and therefore has moved to a focus position to the left of the
complementizer. This view clearly highlights the marked nature of (13).
However, departing from M. Parry, I suggest that (13) carries special emphasis
on the verb, rather than the subject, indicating that the position of the subject on
the left of che is outside the scope of focus. This idea rises from the following
grammaticality contrast:
(15) *Mario ch'a parta, nen Gianni.

"Mario must go, not Gianni."
(16) Mario ch 'a studia, nen ch 'a lesa l'giornal.
"Mario must study, not read the newspaper."
If this is correct, the position of Mario in the subject+ che structure above cannot
be taken to be the unmarked position of the subject. Goria (2000) offers an
explanation along the lines of Zubizarreta's (1998) p-(rosodic) movement,
suggesting that the subject in (13) has moved to the left periphery in order for
prosodic prominence to be assigned correctly to the verb. Here it suffices to note
that subject+ che and che+subject imperatives have different readings, so that the
position of the subject in the former cannot be given as evidence in favour of the
subject-in-CP analysis.
Finally, even within Poletto's analysis, the position of Mario in (13) turns
out not to be the unmarked one. Making the distinction between deictic and non-
deictic suppletive imperatives, Poletto posits two projections for the realization
I am aware of the claim that a conservative variety of Piedmontese allows suppletive imperative
structures introduced by two complementizers.
(i) che gnum ch ' a bogia
that nobody that SCL moves
These structures need further studying before claims are made with respect to the position of the
subject. It appears that these sentences are highly marked and restricted to certain discourse contexts.
Hence one should be cautious in taking these as evidence in support of the subject-in-CP analysis.
of these different types of imperativity: -deictic CP and +deictic CP. The former
is assumed to be higher than the latter. In (13), a -deictic imperative, che is in
-deictic  However, the -deictic position must be higher than the unmarked
position of the subject, witness the che+subject -deictic imperative in (14). It
follows that the order subject+che in (13) must be the result of moving Mario to
a different position. This supports the view that the position of the subject in
subject+ che imperatives cannot be taken to be the unmarked position of
preverbal subjects.
To summarise, I have presented evidence against the left dislocation and
the subject-in-CP analyses of preverbal subjects. If the latter are no longer
claimed to be in the Left Periphery, SCLs which always follow preverbal
subjects can only be placed inside the TP domain, assuming Chomsky's Agr-less
structure.7 The next section continues with discussing the function of
Piedmontese SCLs and the fact that they do not check the EPP.
3. SCLs and the EPP

Chomsky (1995) proposes that the EPP feature is a categorial feature
assigned to T responsible for subject raising to Spec TP in SVO. Thus,  features
and Case are no longer the trigger for subject raising, as they may be checked via
long-distant agreement (Chomsky 2000). It has been suggested, however, that in
NSLs lexically realised preverbal subjects do not satisfy the EPP (Alexiadou &
Anagnostopoulou 1998; among others). The latter may be checked by verbal
inflection in NSLs without SCLs, thanks to the [+N] feature of verbal inflection,
and by the SCLs in the NIDs. In this section, I argue against this interpretation of
SCLs.
Firstly, following Chomsky's view, if SCLs are EPP checkers, Spec TP is
no longer available for subject raising and preverbal subjects must be differently
motivated. This is essentially the claim grounding the dislocation analysis which
has been evaluated and excluded with respect to Piedmontese.
Secondly, the claim that SCLs check the EPP predicts incorrectly that
sentences involving SCLs are structurally different from those that do not. In
fact, in Piedmontese there is no evidence that omission of SCLs affects sentence
structure.
Thirdly, if SCLs check the EPP, the interpretative properties of inverted
locative constructions described in Benincà (1988) and Pinto (1997) find no
explanation. In brief, VS in Italian, and Piedmontese, is generally licensed by
In this paper, the claim that Piedmontese SCLs are in T is defended only with respect to the position
of unmarked preverbal subjects. The reader is referred to Goria (forthcoming) for more arguments
supporting this claim.
100 CECILIA GORIA
Focalisation. That is, postverbal subjects are focalised, hence derivational

processes moving the subject to the relevant Focus Phrase are required for the
satisfaction of [+focus]. Nonetheless, VS can bear wide focus if the verb takes a
LOC(ative) argument. The function of LOC is to check the EPP feature in T
allowing the subject to stay in situ (Pinto 1997). If the verb does not take LOC,
subject raising is required by the EPP.8 Clearly this analysis can no longer be
maintained if SCLs are invariably the checkers of the EPP. This is an unwanted
result given that attributing to LOC the ability to check the EPP in locative
structures neatly captures the focus reading of postverbal subjects in
Piedmontese.
Finally, presentative constructions, as those illustrated below, provide more
evidence against the idea that SCLs are the checkers of the EPP.
(17) Tuti ij ani, an sla statai per Ast, a j meuir vàire

every the year on the road to Asti SCL LOC die many
giovo.
young people
"Every year, on the road to Asti, many young people die."
In presentative constructions, the subjects are postverbal and the sentences carry
wide focus. As mentioned earlier, Pinto (1997) argues that a LOC argument
checks the EPP, hence the subject does not need to raise to Spec TP. In
Piedmontese, the presence of LOC is signalled by the locative particle je/j '. The
compatibility between the SCL a and je/j' reveals that the former is not involved
in satisfying the EPP, but that the latter is, witness the ungrammaticality of a
preverbal subject:
(18) *La nona 'd Domenico a l' è mort-je.

the grandmother of Domenico SCL C1 7 is died-LOC
"Domenico's grandmother has died."
Note that the co-occurrence between preverbal subjects and je/j' becomes
more acceptable, although marked, if the former is contrastively focused, (19)
below. My explanation is that the subject in (19) carries feature [+focus], hence
it raises to FocusP (Rizzi 1997) regardless of the fact that the EPP is satisfied by
LOC.
See Goria (forthcoming) for a detailed discussion about inverted locative constructions in
Piedmontese. The reader is also referred to Belletti (2001) and Zubizarreta (1998) for the relation
between inverted subjects in Romance and Focalisation.
(19) So ,  l è mort-je, nen so mama.

his grandmother, SCL CL 7 is died-LOC, not his mother
"His grandmother has died, not his mother."
Nonetheless, if the subject is dislocated, the compatibility between

preverbal subjects and je/j' is again weakened, as the EPP is checked by LOC
preventing pro from raising to Spec TP.
(20) * La cesa, la sman-a passá, a l' è brusa-je.

the church, the week gone, SCL CL 7 is burnt-LOC
"The church, last week, burnt down."
Finally, if -je is the referential clitic with the meaning "to him/her"(cf. al
preive "to the priest"), homophonous with the locative particle, (20) gains
acceptability. Under this interpretation the EPP is no longer checked by LOC and
pro is licensed by raising to Spec TP.
To conclude, I have argued that Piedmontese offers no evidence for placing
SCLs outside the TP domain and for the claim that they function as EPP
checkers. Hence, I maintain that SCLs are Ds adjoined to T and are not involved
in EPP checking. SCLs encode the Agr(eement) properties of T. This analysis
presents the following core advantages:
1. no subject positions in the left periphery of the clause and, most importantly,
no extra SCL projections need be postulated;
2. the interpretative differences that characterise postverbal subjects in locative
constructions are explained straightforwardly;
3. it is correctly predicted that the presence vs. absence of SCLs has no
syntactic consequences;
4. it establishes a parallel between the domains of T and v (see fn. 3).
I now go on to deal with the morphology and distribution of Piedmontese SCLs.
4. Optimal Agreement and Piedmontese SCLs

The morphology and distribution of SCLs is characterised by a high degree
of variation. In Piedmontese, morphologically different SCL systems are
attested. Furthermore, SCLs may always be optionally omitted with no syntactic
implication. I label this pattern of optional omission Full Optionality. At the
same time, the frequency of omitting SCLs depends on their person and number
102 CECILIA GORIA
features. That is, SCLs encoding 2sg, 3sg and 3pl are dropped less often than the
other SCLs in the paradigm. I call this pattern Person Optionality.9
The view, presented earlier, that SCLs are adjoined to T and encode T's 
features and its EPP feature hardly allows for the variation just described. In fact,
it does not offer the tools for explaining why some features and not others are
overtly expressed, giving rise to the distinct clitic-paradigms attested in
Piedmontese. Additionally, it predicts that optionality results from distinct
numerations and subsequently structural differences. As mentioned earlier, this
prediction is incorrect with respect to Piedmontese, given that there is no
evidence that a sentence with SCLs is syntactically different from the same
sentence without SCLs. Thus, the optionality of SCLs in the varieties examined
here (see below) is a case of free variation.
For these reasons, without departing from the structural account of SCLs
suggested in the previous section, I rely on the general principles of Optimality
Theory and propose that the moiphological and distributional properties of
Piedmontese SCLs derive from the competition of constraints concerned with the
overt realization of agreement features. These operate in a component of
grammar separate from narrow syntax. It will become clear that the advantage of
my analysis is that the properties of SCLs described above do not require a
structural explanation. In this way, a full account of Piedmontese SCLs can be
given within the T-model of sentence structure (Chomsky 1995).
I will look at SCLs in two varieties of Piedmontese: Turinese and
Astigiano. The SCLs in question are given in Table 1:
Turinese Astigiano
lsg i mangio a mangio
2sg it mange at mange
3sg a mangia a mangia
lpl i mangioma a mangioma
2pl i mange a mange
3pl a mangio a mangio
Table 1
The starting point of the analysis proposed here is Samek-Lodovici's

(1996) Optimal Agreement, according to which different agreement patterns are
the outcome of different rankings of agreement constraints. A subgroup of the
In a similar fashion, SCLs are optionally omitted in the second conjunct of coordinated structures.
Because of lack of space this issue cannot be dealt with in this paper. The reader is referred to Goria
(forthcoming) for more details.
latter promotes agreement features, while a constraint labelled No -FEATURES

penalizes them. Samek-Lodovici makes the agreement constraints relative to
features and to the heads they are associated to, explaining variation within one
and the same variety as well as across languages. In the present analysis, the
agreement constraints are constraints promoting or penalizing the overt
morphological realization of agreement features; they are relative to T, being
concerned with the realization of the  features in T; and are families of
constraints as they are relative to each single feature of the major agreement
categories of person, number and gender.
4.1 The proposal

The goal in this section is to account for the morphological properties of
SCLs in Astigiano and Turinese and for Full and Person Optionality. Given that
the agreement constraints operate on the morphological realization of features, it
is necessary to examine the featural content of the SCLs under examination.
Beginning with Astigiano (Table 2):
1sg a [-(add,sg)]
2sg at [+(add,sg)]
3sg a [-(add,sg)]
1p1 a [-(add,sg)]
2pl a [-(add,sg)]
3pl a [-(add,sg)]
Table 2: Astigiano
2sg is morphologically distinct from all other persons. As 2sg corresponds to the
feature [+(addressee +singular)] (henceforth [+(add,sg)]; see Table 3 below),
SCLs in Astigiano mark the distinction between [+(add,sg)] vs. [-(add,sg)].10 I
label this as the Basic System, as it exemplifies the basic morphological
distinction that can be expressed by a system of SCLs.11
Under the proposal that SCL systems derive from constraint interaction, the
Basic System results from the competition between constraints that promote the
realisation of features [+(add,sg)] and [-(add,sg)], i.e. [+(ADD,SG)] and
In all varieties 2sg has unique morphology (Renzi & Vanelli 1983). Hence, it seems legitimate to
postulate a morphological feature [+(add,sg)], that contrasts with all other persons, i.e. [-(add,sg)].
This is supported by the pragmatic prominence of 2sg.
11
See Goria (forthcoming) for the compatibility of such an idea with the claim that in some varieties
SCLs express no subject agreement.
104 CECILIA GORIA
[-(ADD,SG)] respectively, and No -FEATURES which penalizes agreement

features:
i) [+(ADD,SG)] » [-(ADD,SG)] » No -FEATURES gives rise to the Basic System,
see Tableau 1:
[+(ADD,SG)] [-(ADD,SG)] No

FEATURE
☞ lsg a [-(add,sg)] * * 1
0 * *!
☞ 2sg at [+(add,sg)] * *
0 *! *
☞ 3sg a [-(add,sg)] * *
0 * *!
☞ 1pl a [-(add,sg)] * *
0 * *!
☞ 2pl a [-(add,sg)] * *
0 * *!
☞ 3pl a [-(add,sg)] * *
0 * *!
Tableau 1: Basic Systeni
By re-ranking these constraints a number of other SCL systems are derived:

ii) NO -FEATURES » [+ (ADD,SG)] » [-(ADD,SG)] gives rise to a system with no
SCLs ( 0 SCLs).12 This system is represented in Tableau 2:
" Further research is required for a better understanding of the nature of 0 SCLs. The question is
whether a 0 SCL is to be interpreted as absence of the head hosting the clitic or as a null SCL, i.e.
one which does not express agreement features. As  features are assigned to T regardless of a
language's ability to express such features overtly, it seems reasonable to think that if a language has
SCLs than a 0 SCL is the null equivalent of its overt counterpart. Note that this view goes along with
Torrego's (1998b) proposal that the difference between NSLs and non-NSLs lies in the way the EPP
feature and the agreement features of T are encoded. Torrego suggests that in the former, but not in
the latter, a separate head D adjoined to T encodes (covertly in her proposal) the features of T. At the
same time, Cardinaletti (2002) argues against covert clitics arguing that the absence vs. the presence
of a clitic (object clitic in Right Dislocation and Marginalization in her analysis) equates to syntactic
differences. Without challenging Cardinaletti's proposal, here it suffices to say that there is no
evidence in Piedmontese that a sentence with SCLs is structurally distinct from the same sentence
without SCLs (see also Goria forthcoming for a detailed discussion on this topic).
[+(ADD,SG)] [-(ADD,SG)]
FEATURES
[-(add,sg)] *!
☞ * *
1sg 0
[+(add,sg)] *! *
☞ 2sg * *
0
[-(add,sg)] *!
☞ 3sg 0 * *
[-(add,sg)] *!
☞
1pl 0 * *
[-(add,sg)] *!
☞ * *
2pl 0
[-(add,sg)] *!
☞
3pl 0 * *
Tableau 2: ØSCLs
iii) [+(ADD,SG)] » No -FEATURES » [-(ADD,SG)] gives rise to a system with

only 2nd sg SCL (cf. Renzi & Vanelli's (1983:128), System 6). See Tableau 3:
No
[+(ADD,SG)] FEATURES [-(ADD,SG)] |
[-(add,sg)] * *!
☞
1sg Ø * *
☞ 2sg [+(add,sg)] * *
Ø *! *
[-(add,sg)] * *!
☞ 3sg Ø * *
[-(add,sg)] * *!
☞
1pl Ø * *
add,sg)] * *!
☞
2pl Ø * *
[-(add,sg)] * *!
☞
3pl Ø * *
Tableau 3: Renzi & Vanelli's System 6
106 CECILIA GORIA
iv) [-(ADD,SG)] » No -FEATURES » [+(ADD,SG)] gives rise to a system with

all persons other then 2nd sg. This system, however, has never been attested. The
pragmatic prominence of 2sg may be the factor responsible for the dominance of
[+(ADD,SG)] over [-(ADD,SG)].
Turning to Turinese, its SCL system is morphologically richer. Not only is
2sg distinct from all other persons, as in Astigiano, but also are 1sg and 1pl, and
2pl distinct from 2sg and 3sg and 3pl. In order to give such a system the correct
interpretation, I look at the componential analysis of the category of person
(Table 3) and illustrate the morphological distinctions marked by the system of
SCLs in Turinese (Table 4), which I call the Deictic System because it marks the
role of person in the speech act.
Speaker; ([speak])
Participant in the Singular; ([sg])
speech act; ([part])
lsg + + +
2sg - + +
3sg - - +
lpl + + -
2pl - + -
3pl - - -
Table 3: Componential analysis of the category of person
(adapted from Calabrese 1998)
lsg i [-(add,sg)] [+part]

2sg it [+(add,sg)]
([+part] [-speak] [+sg])
3sg a [-(add,sg)] [-part]
([-speak])
lpl i [-(add,sg)] [+part]
2pl i [-(add,sg)] [+part]
3pl a [-(add,sg)] [-part]
([-speak])
Table 4: Turinese13,14
13
[+(add,sg)] is also inevitably [+part] [-speaker] [+sg]. Hence, [+(add,sg)] also satisfies [+PART],
[-SPEAK], [+SG] violating No  FEATURES four times. Analogously, [-(add,sg)] [-part] is also
([-speak]), and therefore it violates No  FEATURES three times.
Although the reduction of the set of features involved is attractive, it would not suffice to capture
the morphological patterns attested in Piedmontese as well as in several SCL systems across the
NIDs (Goria forthcoming). In particular, the simplification of the complex feature (±add,sg) into
In order to derive the Deictic System, a ranking that favours feature [+part]
in addition to the basic [±(add, sg)] distinction is necessary. All three systems
seen so far must be taken to be in competition: i) the 0 SCL System; ii) the
Basic System, with [±(add,sg)] SCLs; iii) the Deictic System with [±part] SCLs.
This is illustrated in Tableau 4:
SCLs Features [+(ADD.SG)] [+PART] NO [-(ADD,SG)] [-PART] I

FEAT.
☞ 1sg i [-(add,sg)] ** *
r+part]
a [-(add,sg)] *! * *
Ø *! * * *
☞ 2sg it/at [+(add,sg)] **** * *
Ø *! * * *
3sg a [-(add,sg)] ***
[-part]
a [-(add,sg)] *! *
Ø *! * * **
1pl i [-(add,sg)] ** *
[+part]
a [-(add,sg)] *! * *
0 *! * * *
2pl i [-(add,sg)] ** *
[+part]
a [-add,sg)] *! * *
Ø *! * * *
3pl a [-(add,sg)] ***
[-part]
a f-(add,sg)] *! * *
0 *! * * *
Tableau 4: Deictic System
To sum up, the interaction of three agreement constraints determines four SCL
systems. These have been illustrated in Tableau 1 for the Basic System, Tableau 2
with no SCLs, Tableau 3 for Renzi & Vanelli's System 6, and Tableau 4 for the
Deictic System.
(±addressee) (¿number), although a desirable option, would not adequately describe the
morphological distinction that characterises the Basic System in Astigiano.
108 CECILIA GORIA
4.2 Full and Person optionality

The core ideas in this section are i) optionality is obtained from one and the
same input. That is, the set of  features assigned to a given T; ii) Full and
Person Optionality are the result of the co-existence of several rankings within a
given variety (Anderson 1996, 2000). This provides the speakers with the choice
of which system to adopt. It is in these terms that the influence of Italian on the
use of SCLs is to be interpreted. The close contact with the standard language
has an effect on the speakers' choice favouring the Ø SCL System to the
detriment of the other systems.
Person Optionality gives rise to the following paradigms, shown in Table 5:
Turinese Astigiano
1sg Ø mangio Ø mangio
2sg it mange at mange
3sg a mangia a mangia
1pl Ø mangioma Ø mangioma
2pl Ø mange Ø mange
3pl a mangio a mangio
Table 5: Person Optionality
This is a deictic system, i.e. Deictic System 2, in which [-(add,sg)] [+part] are
covert, but still distinct from [-part] and from [+(add,sg)]. Crucially, the systems of
Astigiano and Turinese end up expressing the same features (cf. Table 1 and Table
5). This is captured, here, by claiming that the Deictic System 2 is obtained in both
varieties from one and the same constraint ranking. By re-ranking the hierarchy
established for the Deictic System 1 in Tableau 4, the Deictic System 2 is obtained
(Tableau 5, below).
Having established in this and the previous sections the rankings necessary to
obtain the Basic System, the Deictic Systems 1 and 2 it is now possible to clarify
optionality with respect to the distribution of these SCLs. Assuming that optionality
is the result of multiple hierarchies freely alternating within the same linguistic
system, the optionality patterns attested in Turinese and Astigiano are
straightforwardly accounted for. The rankings available in each variety are listed
below:
D Rankings available in Turinese:

a) Deictic System 1 : [+(ADD,SG)] » [+PART] » No  FEATURES » [-(ADD,SG)]
» [-PART]
b) Deictic System 2: [+(ADD,SG)] » [-PART] » NO  FEATURES » [-ADD,SG)]
» [+PART]
c) Ø SCL System: No  FEATURES » [+(ADD,SG)] » [-(ADD,SG)] » [+PART] »

[-PART]
SCLs Features [+(ADD,SG)] [-PART] № [-(ADD,SG)] [+PART]

FEAT.
i [-(add,sg)] * * *!*
[+part]
a [-(add,sg)l * * *!
☞
1sg Ø * * * **
☞ 2sg it/at [+(add,sg)] * ** ** *
Ø *! * * *
☞ 3sg a [■(add,sg)] * *** *
[-part]
i [-(add,sg)] * *! *
Ø * *! *
**
*!*
i [-(add,sg)] * *
[+part]
a [-(add,sg)] * * *! *
☞ 1pl Ø * * * *
*!*
i [-(add,sg)] * *
[+part]
a [-(add,sg)] * * *! *
☞ 2pl Ø * * * *
☞
3pl a [-(add,sg)] * *** *
[-part]
a [-(add,sg)l * *! * *
Ø * *! * *
Tableau 5: Deictic System 2
ii) Rankings available in Astigiano:15

a) Basic System: [+(ADD,SG)] » [-(ADD,SG)] » No  FEATURES » [+PART] »
[- PART]
b) Deictic System 2: [+(ADD,SG)] » [-PART] » No  FEATURES » [-(ADD,SG)]
» [+PART]
c) Ø SCL System: No  FEATURES » [+(ADD,SG)] » [-(ADD,SG)] » [+PART] »
[- PART]
15
The unavailability in Astigiano of the ranking i) a) simply means that in this dialect there is no
means of expressing overtly [-(add,sg)] [+part].
CECILIA GORIA
110
The consistent use of SCLs is due to the rankings (i-a) and (ii-a) available
in Turinese and Astigiano respectively. Full Optionality is due to the free
alternation between the ranking for Ø SCL System and the other rankings in both
varieties. Person Optionality is due to the alternation between the ranking for the
Deictic System 2 and the other rankings.
The adoption of the Optimal Agreement framework offers a number of
advantages:
1. it provides a formal account for the availability of SCLs in a given variety
and for their morphology. Crucially all systems are obtained from the
interaction of the same agreement constraints;
2. it captures the similarities between Turinese and Astigiano with respect to
the Deictic System 2. In fact, the  perspective offers the tools to relate these
systems by virtue of their featural content;
3. it provides a unique systematic explanation for the causes and the effects of
Full and Person Optionality in Piedmontese;
4. it clarifies the tendency of Piedmontese speakers to leave out their SCLs.
Here, it is suggested that the close contact of Turinese and Astigiano with
standard Italian is responsible for such a change. Analogously, it is
legitimate to conjecture that the use of the Deictic System 2 in Astigiano
may be due to the influence of Turinese;
5. it makes sure that the optionality typical of Piedmontese SCLs is treated as
authentic free variation, by deriving it from one and the same input;
6. last but not least, it does not involve complex structures.
5. Summary and conclusion

In this paper, I have provided an encompassing account of Piedmontese
SCLs by taking the general principles of the Minimalist Program and Optimality
Theory to be complementary to one another. I have first dealt with the structural
position and function of these elements, proposing that they are Ds adjoined to T
realizing T's EPP and  features, and that they are unrelated to the checking of
the EPP. The sentence structure adopted as well as the inviolable nature of the
EPP are those promoted by the Minimalist Program. The main advantage of this
proposal is that SCLs are treated as elements of agreement without abandoning
the task to reduce structural and computational complexity. This analysis
however does not offer the tools to deal with language internal variation. In the
second part of the paper, I have overcome such a drawback and dealt with
variation by looking at the relation between the morphological make-up of SCLs
and their distribution. The core idea of my proposal is that the latter is unrelated
to the position in the structure occupied by SCLs and is the result of the
interaction of morphological constraints. These are separated from narrow syntax
mechanisms, are concerned with the morphological realization of agreement and

compete in the fashion of Optimality Theory. Once again, in my explanation, I
avoid resorting to structural complexity. In this respect, the present proposal
contrasts sharply with existing accounts of Piedmontese SCLs.
As a final observation, the analysis presented in this paper needs testing
against other SCL systems and morpho syntactic phenomena. In Goria
(forthcoming) this task is undertaken by deriving the morphological properties of
several SCL systems described in the seminal work of Renzi & Vanelli (1987) as
the result of the same mechanism of agreement constraints ranking and re-
ranking. It is the task of future work to extend the field of research and evaluate
further the validity of the present proposal in a comparative perspective.
References
Alexiadou, A. & E. Anagnostopoulou. 1998. "Parametrizing Agr: Word Order, V-
Movement and EPP-Checking". Natural Language and Linguistic Theory 16:3.491-
539.
Anderson, S. 1996. "How to put your clitics in their place or why the best account of
second-position phenomena may be a nearly optimal one". The Linguistic Review
13.165-191.
2000. "Towards an Optimal Account of Second Position Phenomena". Optimality
Theory: Phonology, Syntax and Acquisition ed. by J. Dekkers et al., 302-332.
Oxford: Oxford University Press.
Barbosa, P. 1995. Null Subjects. Doctoral Dissertation. Cambridge, Mass: MIT Press.
2000. "Clitics: A Window into the Null Subject Property". Portuguese Syntax:
New Comparative Studies, ed. by J. Costa, 31-93. Oxford & New York: Oxford
University Press.
Benincà, P. 1988. "L'ordine degli elementi della frase e deile construzioni marcate".
Grande grammatica italiana di consultazione ed. by L. Renzi, 115-125. Bologna: Il
Mulino.
Belletti, A. 2001. " 'Inversion' as Focalization". Subject Inversion in Romance and the
Theory of Universal Grammar, ed. by A. Hulk, & J-Y Pollock, 60-90. Oxford
University Press: New York.
Calabrese, A. 1998. "I sincretismi fra pronominali clitici nei dialetti italiani e sardi e la
teoría della morphologia distribuita" . Atti del XII Congress  Internazionale di
Linguistica e Filología Romanza. Centro di studi filologici e linguistici siciliani,
Università di Palermo ed. by G. Ruffino, 107-122. Tuebingen: Max Niemeyer
Verlag.
Cardinaletti, A. 1997. "Subjects and clause structure". The New Comparative Syntax ed.
by L. Haegeman, 33-63. London: Longman.
2002. "Against optional and null clitics. Right Dislocation vs. Marginalization".
Studia Linguistica 56.29-57.
Chomsky, N. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press.
2000. "Minimalist Inquiry: The Framework". Step by step ed. by R. Martin, D.
Michaels & J. Uriagereka, 89-155. Cambridge, Mass: MIT Press.
112 CECILIA GORIA
2001. "Derivation by Phase". Ken Hale: A life in language, ed. by M. Kenstowicz,

1-52. Cambridge, Mass: MIT Press.
Cinque, G. 1990. Types ofA'-Dependencies. Cambridge, Mass: MIT Press.
1997. "'Topic' Constructions in Some European Languages and Connectedness'".
Materials on Left Dislocation ed. by E. Anagnostopoulou, et al, 93-118. Linguistik
Aktuell, 14. Amsterdam: John Benjamin.
Costa, J. 1998. Word Order Variation. A constraint-based approach. The Hague, The
Netherlands: Holland Academic Graphics.
Goria, C. 2000. The Syntax and Morphology of Piedmontese SCLs: an analysis based on
the Minimalist Program and Optimality Theory. Doctoral Dissertation. University
of Manchester.
forthcoming. Subject Clitics in the Northern Italian Dialects: a comparative
analysis based on the Minimalist Program and Optimality Theory. Natural
Language and Linguistic Theory Series. Dordrecht, the Netherlands: Kluwer
Academic Publishers.
Johannessen, J.B. 1998. Coordination. Oxford & New York: OUP.
Kayne, R. 1975. French Syntax: The Transformational Cycle. Cambridge Mass: MIT
Press
Ledgeway, A. 2000. Comparative Syntax of the Dialects of Southern Italy: A Minimalist
Approach. Oxford: Blackwell.
Pinto, M. 1997. Licensing and Interpretation of Inverted Subjects in Italian. Utrecht, The
Netherlands: Utrecht Institute of Linguistics OTS.
Poletto, C. 1995. "Complementizer Deletion and Verb Movement in Italian". University
of Venice Working Papers in Linguistics 5:2.49-79.
2000. The Higher Functional Fields. Evidence from the Northern Italian Dialects.
Oxford & New York: Oxford University Press.
Renzi, L. & L. Vanelli. 1983. "I pronomi soggetto in alcune varietà romanze". Scritti
Linguistics in Onore di Giovan Battista Pellegrini ed. by P. Benincà et al. 120-145.
Pisa: Pacini.
Rizzi, L. 1986. "On the Status of Subject Clitics in Romance". Studies in Romance
Linguistics ed. by O. Jaeggli, & C. Silva-Corvalan,, 391-419. Dordrecht: Foris.
1997. "The Fine Structure of the Left Periphery". Elements of Grammar ed. by L.
Haegeman, 112-158. Dordrecht: Kluwer.
Samek-Lodovici, V. 1996. Constraints on Subjects. An Optimality Theoretic Analysis.
Ph.D. Dissertation. Rutgers University: New Brunswick, New Jersey.
Sportiche, D. 1988. "A theory of floating quantifirs and its corollaries for Constituent
Structure". Linguistic Inquiry 19:3.425-449.
Torrego, E. 1998a. The Dependencies of the Objects. Linguistic Inquiry Monographs 34.
Cambridge, Mass: MIT Press.
1998b. "Nominative subjects and pro-drop infl". Syntax 1,2206-2X9. Oxford:
Blackwells.
Zubizarreta, M.L. 1998. Prosody, Focus, and Word Order. Linguistic Inquiry Monograph.
Cambridge Mass: MIT Press.
IDENTIFICATIONAL FOCUS VS CONTRASTIVE FOCUS
A SYNTACTIC DISTINCTION
DANIELA ISAC
UQAM/Concordia University
1. The Proposal
In this paper, we propose that the properties of clitic doubled direct object
constructions in Romanian are best understood if we take clitics to be Focus
operators of an anaphoric nature (Rizzi 1997). More specifically, we propose that
the clitic anaphorically connects the doubled object to a set of alternatives, i.e.
the set of contextually or situationally given elements for which the predicate
phrase can potentially hold. This proposal has two desirable consequences. First,
the anaphoric nature of the clitic operator explains why bare quantifiers cannot be
doubled by clitics. A bare quantifier doubled by a clitic leads to a conflicting LF
configuration in which the (empty) object position must be interpreted at the
same time as a null constant and as a variable. Second, under this view, one can
account for the constraints on the interpretation of a clitic doubled object. In
particular, one can account for the absence of variable readings and of kind level
readings for the clitic doubled object and for why only individual constants are
retained as possible referents of clitic doubled object DPs in Romanian.
The paper is organized as follows: in section 2 we argue for an analysis of
accusative clitics in Romanian as Focus operators; in section 3 we show that
Romanian accusative clitics are non-quantificational and that they are anaphoric
operators instead; in section 4 we discuss the consequences of our analysis for the
interpretation of doubled objects; and in section 5 we draw the conclusions.
2. Accusative clitics as Focus operators

2.1. Clitic doubling in Romanian
Romanian DOs must be doubled by an accusative clitic if the former is
[+specific] and [+human]. Moreover, clitics can only double objects preceded by
114 DANIELA ISAC
the preposition pe. Pe must be used with person denoting proper names and
personal pronouns and is optionally used with all person-denoting nouns.
(1) a. *(Li)-am vãzut pe Ioni.

CL-have-I seen pe Ion
"I have seen Ion."
b. (Li)-am ascultat pe studenti/ studentui.
(CL)-have-I interviewed pe student / student-the
"I have interviewed the student."
 *Li-am vãzut pe âii.
CL-have-I seen pe dog
"I saw the dog."
d. *Li-am ascultat pe disi.
CL-have-I listened pe disc
"I've listened to the disc."
2.2. Clitics and focus

The focus nature of a constituent can be tested by resorting to a diagnostic
proposed by Rochemont and Culicover (1990).
(2) In a well formed question/answer sequence, all and only the

information requested in the question is focused in the response.
The application of this test to clitic doubled constructions shows that clitic
doubled objects are focused: a suitable answer to (3a) is (3b), which contains a
clitic doubled object, but not (3c).
(3) a. Pe cine (*l)-ai vãzut?

pe who CL-have-you seen ?
"Who have you seen?"
b. L-am vãzut pe student.
CL-have-I seen pe student
"I have seen the student."
. #Am vãzut studentul.
have-I seen student-the
"I have seen the student."
Notice that (3 a) contains no clitic, and that in fact clititic doubling yields an
ungrammatical result in this case. So the presence of the clitic in (3b) cannot be
IDENTIFICATIONAL FOCUS VS CONTRASTIVE FOCUS 115
due to some requirement on the symmetry of question-answer pairs. Rather, an

object DP which is not doubled by a clitic is not in focus, but one that is clitic
doubled is interpreted as focused.
2.3 Clitics and identificational focus

The following types of focus are distinguished in the literature (Rochemont
1986, Erte schik-Shir 1997, Kiss 1998): (i) information focus, which simply
expresses non-presupposed or new material; (ii) identifie ational focus, which is
the exhaustive subset of the set of contextually or situationally given elements for
which the predicate phrase can potentially hold; and (iii) contrastive focus, a
particular case of identificational focus, which operates on a closed set of entities
whose members are known to the participants of the discourse (Kiss 1998). In
this case, the identification of a subset of a given set also identifies the
contrasting complementary set. Syntactically, the assumption is that both
identificational and contrastive focus involve movement to a scope position, i.e.
to the specifier position of a functional projection in the left periphery of the
clause.
In 2.2. above we have established that Romanian clitic doubled DOs
express focus. In view of the distinction between several types of focus, the
question is what kind of focus is relevant for clitic doubled DOs. Our claim is
that clitic doubled (in situ) DOs in Romanian express identificational focus. This
indicates that one can distinguish between identificational and contrastive focus
not only on semantic and pragmatic grounds, but also based on syntactic
considerations. One way to argue in favor of this claim is to test whether clitic
doubled objects have the properties associated with identificational focus. The
[+exhaustive] feature of identificational focus can be tested as follows (Szabolcsi
1981): if in a pair of sentences the first sentence contains a focus consisting of
two coordinated DPs, as in (4a), and the second sentence differs from the first
one only in that one of the coordinated DPs has been dropped (as in (4b)), and if
the second sentence is not among the logical consequences of the first one, then
the focus expresses exhaustive identification.
(4) a. It was a hat and a coat that Mary picked for herself.
b. It was a hat that Mary picked for herself. (Kiss, 1998)
(5) a. Mary picked a hat and a coat for herself.
b. Mary picked a hat for herself.
116 DANIELA ISAC
(4b) is not a logical consequence of (4a); rather, it contradicts it. Consequently, a

hat in (4b) expresses identificational focus. In contrast, (5b) is a logical
consequence of (5a), and as such it does not express identificational focus.
Now, turning to clitic constructions, (6b), which contains a clitic doubled
object, is not a logical consequence of (6a), but (7b), where the object is not clitic
doubled, is a logical consequence of (7a).
(6) a. I-am invitat pe [Ion si Maria] ia noi.

CL have-I invited pe Ion and Maria at us
"I invited John and Mary to our house."
b. L-am invitat pe Ion la noi.
CL have-I invited pe Ion at us
"I invited John to our house."
(7) a. Am invitat studentu si profesora la noi.
Have-I invited students-the and teachers-the at us
"I invited the students and the teachers to our house."
b. Am invitat studentu la noi
Have-I invited students-the at us
"I invited the students to our house."
This shows that the set defined by a non-clitic doubled conjoined DO is not
exhaustive, and that the [+exhaustive] feature is actually a contribution of the
clitic.
Furthermore, there are certain restrictions on the type of constituents that
can be possible identificational foci; in particular, universal and existential
quantifiers seem to be incompatible with identificational focus, as argued by Kiss
(1998). The same restrictions seem to operate on the clitic doubled object.
(8) (*Îl) acceptãm pe oricine.

(him) accept-we pe anybody1
"We accept anybody."
(9) (*L)-am vãzut pe cineva.
Him-have-I seen pe somebody
"I have seen somebody."
4
As noticed by one of the reviewers, the gloss to example (8) indicates "anybody", and not
"everybody". Even though the so-called 'free choice' "anybody" is analysable as a universal
quantifier, we do not think that the distinction between "anybody" and "everybody" is relevant to our
discussion. For further discussion, see section 3.3.
Kiss (1998) also notices that identificational focus is incompatible with "also"
phrases and "even" phrases. Romanian, however, allows such phrases in clitic
doubled constructions, which seems to pose a problem to our proposal that clitic
doubled object express identificational focus.
(10) a. L-am invitat si pe Ion

CL-have-I invited also pe Ion
"I also invited Ion."
b. L-am invitat chiar si pe director.
CL-have-I invited even also pe director
"I even invited the director."
 L-am invitat chiar pe director.
CL-have-I invited even pe director
"I invited the director himself"
However, this problem is only apparent. First of all, the "also" phrase in (10a) is
acceptable precisely in a context where it can be understood to identify a member
of a relevant set in addition to one or more members identified previously, with
the rest of the set still excluded. Moreover, an "also" phrase can also be clefted
under such an interpretation, as shown by (11), even though clefts in English
typically express identificational focus.
(11) A: Bill danced with Mary. (Kiss 1998)

B: No, it was Sam that danced with Mary.
C: It was also John that danced with Mary.
With regard to the "even" phrase, notice that it can have two slightly different
interpretations in Romanian, as the glosses to (10b,c) indicate. Under the
interpretation in (10b), it is basically similar to an "also" phrase, as it identifies a
member of a relevant set, i.e. the director, in addition to one or more members
identified previously, with the rest of the set still excluded. Under the
interpretation in (10c), it is simply an emphatic element, as the gloss indicates,
and as such it is perfectly compatible with a [+exhaustive] interpretation.
To conclude this section, clitic doubled DOs have the same properties as
constituents that express identificational focus: they are [+exhaustive] and they
are subject to the same type of restrictions as identificational foci.
118 DANIELA ISAC
2.4 Clitics and contrastive focus

Consider (12b) and (12c) below, which are both possible answers to (12a).
(12c) shows that clitic doubled direct objects can be contrastively focused.
(12) a. Ai întrebat profesorul la ce ora sãvii?

Have-you asked teacher-the at what hour to come?
"Did you ask the teacher what time you should be there?"
b. N-am întrebat PROFESORUL, ci STUDENTU.
Not have-I asked teacher-the but students-the
"No. I didn't ask the teacher, I asked the students."
c. Nu l-am întrebat pe PROFESOR ci pe STUDENT
Not CL-have-I asked pe teacher, but pe student
"No. I didn't ask the teacher, I asked the students."
This is not very surprising, given that contrastive focus is semantically a subtype
of identificational focus. However, the question arises whether clitic doubling
plays any role in distinguishing between identificationally focused and
contrastively focused objects. To answer this, let us first examine the syntactic
properties of contrastive focus in Romanian in more detail. As shown by Alboiu
(2000), Romanian contrastive focus is correlated with two syntactic positions: the
preverbal position, as in (13), and a post-verbal position, as in (12) above.
(13) Pe MIHAI l-am strigat, nu pe Ion.

Pe Mihai CL-have-I called, not pe Ion
"It is Mihai I called, not Ion."
When fronting is involved, there is a clear syntactic distinction between the way
in which contrastive focus is expressed and the way in which identificational
focus is expressed. Under the assumption that the clitic is associated with the DO
position, the properties of the clitic chain in (13) clearly differ from its properties
in (12c), given the different syntactic location of the DO. However, when an in-
situ DO is contrastively focused, the distinction between a contrastively focused
DO and an identificationally focused DO is blurred, as in both cases the object
can be clitic doubled. In order to solve this puzzle, let us first notice that in (12c)
above, contrastive focus is signalled prosodically, by emphasis (higher pitch). In
fact, prosodical marking is obligatory in both (12c) and in (13), i.e. whether the
contrastively focused constituent moves or remains in situ. The generalization
seems to be that what is crucial for a contrastive interpretation to obtain is not
clitic doubling, but that the contrasted objects be under stress. Both a clitic
doubled object, as in (12c), and an object that is not clitic doubled, as in (12b),
can be interpreted contrastively, as long as they are under stress.
2.5. Clitics and information focus

If it is true that a clitic doubled object expresses identificational focus, i.e.
that it identifies a subset of a set of contextually ox situationally given elements,
and given that information focus introduces new, non-presupposed material, then
the expectation is that a clitic doubled object should not be able to express
information focus, since the object cannot at the same time relate to a
contextually given set and introduce new material. However, as brought to my
attention by one of the reviewers, (14) seems to deny this expectation.
(14) a. L-am cãutat pe Petru.

CL-have-I looked for pe Petru.
"I looked for Petru."
b. Who did you look for?
 What happened next?
The problem seems to be that (14a) can be an answer to any of the questions in
(14b,c), and as such it seems to allow a focus reading which is wider than the
object. In particular, when (14a) is an answer to (14c), it allows for a wide, IP-
focus reading. Since only information focus can project in this way, it looks like
the clitic doubled object in (14a) can simply express information focus.
However, (14a) does not necessarily show that clitic doubled objects do not
consistently express identificational focus. Instead, it may show that the two
types of focus - identificational and information focus- can coexist. As noted by
Szabolcsi (1983), Szabolcsi and Zwarts (1993), and Kiss (1998), identificational
focus is semantically restricted to a particular type of constituents, namely to
constituents which denote unordered sets of distinct individuals. This is because
complement formation, which is crucial for exaustive identification to obtain, can
take place only in the case of unordered sets of individuals. On the other hand,
while identificational focus is limited in this way, information focus is not. Any
constituent can qualify as information focus. This allows then for the possibility
of having both information focus and identificational focus expressed in the same
sentence. Kiss (1998) discusses cases in which information focus is contained
within identificational focus. We propose that (14a) illustrates the other possible
coexistence relation, i.e. information focus is wider than identificational focus.
When (14a) is an answer to (14c), it allows indeed for wide, IP-focus reading.
But the IP is an instance of information focus, and this does not exclude the
possibility of having an identificational focus embedded within. The new
120 DANIELA ISAC
information is that a 'looking for x' event happened next. This new, non-
presupposed event, however, contains an identificational focus which is linked to
presupposed information, i.e. to a set of several individuals that are contextually
relevant for the event 'looking for x'. In other words, even when (14a) is an
answer to (14c), it still states that out of a set of relevant entities, it was Petru that
I looked for, to the exclusion of the other relevant individuals; Petru is identified
as the individual of which the predicate exclusively holds.
To conclude so far, we have shown that object clitics are identificational
focus operators in Romanian, i.e. that in-situ, non-stressed clitic doubled objects
consistently and exclusively express identificational focus in Romanian. In what
follows, we explore the nature of the operator expressed by the clitic, as well as
the nature of the variable bound by the clitic operator.
3. Romanian accusative clitics are anaphoric operators

Rizzi (1997) splits A-bar dependencies into those involving a quantifier
which binds a variable and those that involve non-quantificational A-bar binding.
The latter case involves binding of a null constant by an anaphoric operator, i.e.
an element inherently characterized as an operator but different from
quantificational operators in that it does not assign a range to its bindee, but seeks
an antecedent to which it connects the bindee.
(15) Johni, whoi his mother really loves ti , is in big trouble.
(15) shows that the wh-chain in appositives is not quantificational, since no Weak
Cross Over (WCO) effects arise." Consequently the trace cannot be analyzed as a
variable. The chain that results is anaphoric in the sense that who, which is
inherently an operator, does not assign a range to its bindee, but seeks for an
antecedent, i.e. John, to which it connects the bindee.
We propose that object clitics are anaphoric operators. The clitic
anaphorically relates the object to a p-set, in the sense of Rooth (1992). Rooth
proposes that focusing always creates a set of alternatives which are under
consideration in the discourse. This set of alternatives is defined by replacing the
focused element with a variable. The set of alternatives under consideration in
(16a) is (16b), where  ranges over a set of individuals under consideration in the
discourse.
As proposed by Lasnik and Saito (1984), Weak Cross Over configurations, i.e. configurations in
which a pronoun is coindexed with a variable to its right, can be used as a test for the
quantificational nature of a chain.
(16) a. John phoned (only) Mary,

b. John phoned y
Under our view, the correlation between an identificationally focused constituent

and a set of alternatives is expressed by the clitic operator, which anaphorically
connects the object (a constant) to a set of alternatives, i.e. the set of contextually
given elements for which the predicate phrase can potentially hold.
Let us now turn to arguments for the analysis of clitics as anaphoric
operators.
3.1 Specificity effects

It is a commonly observed fact that clitic doubled constructions always have
a specific interpretation, even when the object is indefinite or bare. We assume
that an NP can be characterized as specific when the speaker has an individual in
mind as its referent.
(17) a. ÎI vãd pe student.

CL see-I pe student
"I can see the student."
b. L-am întrebat pe un student.
CL-have-I asked pe a student
"I asked a (certain) student."
Under the view that clitics are anaphoric operators, the specificity of the doubled
object (in the sense defined above) is obtained for free: anaphoric relations
always send to antecedents that are well established in the discourse.
3.2Weak Cross Over (WCO)

The non-quantificational nature of the object clitic chain can be tested in
WCO configurations. Consider (18), from Dobrovie-Sorin (1990).
(18) a. ??Mama eii va ajuta [una din student ele tale]¡.

mother her will help one of your students
"??Her mother will help one of your students."
b. Maria va ajuta [una din student ele tale].
Maria will help one of your students
"Maria will help one of your students."
 Mama eii  va ajuta pe [una din studentei e tale],.
mother her CL will help pe one of your students
"??Her mother will help one of your students."
122 DANIELA ISAC
(18a) shows that Romanian partitives normally give rise to WCO. Notice that the
unacceptability of (18a) is not due to the absence of the object clitic, since (18b),
which includes the same partitive object, is perfectly fine without any clitic.
However, when the partitive object in a context like (18a) is doubled by a clitic,
as in (18b), no WCO effects are detectable anymore. This clearly shows that
clitic doubled objects do not involve genuine quantification.
3.3 Quantifiers
Our analysis can also account for the ungrammaticality of (19), where the
object is a bare universal or existential quantifier.
(19) *ÎIi iubeste pe oricinei pe cinevai.

CL loves-he pe anybody/ pe somebody
"He loves anybody/ somebody."
A bare quantifier doubled by a clitic leads to a conflicting LF configuration, in

which the empty object position must be interpreted at the same time as a null
constant (by virtue of being part of an anaphoric chain) and as a variable (since it
is part of a quantificational chain formed by Quantifier Raising).
(20) QPi CLi pe [t]i
However, the empty category left behind by the movement of a quantifier is not
always a variable, and the resulting chain is not always quantificational. In (21a)
the quantifier induces WCO, whereas in (21b) the quantifier does not induce
WCO effects, and does not create a quantificational chain at LR
(21) a. *Mama lui, iubeste pe oricinei/ pe cinevai.

Mother his loves pe anybody/ pe somebody
"*His mother loves anybody/ somebody."
b. Mama lor I -îii iubeste pe unii copii.
mother their-CL loves pe some children
"??Their mother loves some of the children."
Clitic doubling is, as expected, sensitive to this contrast. Objects like oricine
"anybody" or cineva "somebody" cannot be clitic doubled, whereas unii copii
"some children" can, and in fact must, be clitic doubled.
(22) a. *Îl iubeste pe oricine / cineva.

CL loves-(s)he pe anybody / somebody
"(S)he loves anybody / somebody."
b. Ii iubeste pe cativa / unii (studenti ) / toti (studentii).
CL loves-(s)he pe a few /some students /all students-the
"(S)he loves several/some/ all (students)."
The quantifiers in (22a) create a quantificational chain at LF, which conflicts with
the anaphoric nature of the clitic, as shown in (20). In contrast, the quantifiers in
(22b) do not form a quantificational chain, and therefore clitic doubling is
possible.
3.3.1 Two types of Quantifiers. The contrast illustrated in (22) is supported by the
fact that the two types of quantifiers involved have distinct semantic, pragmatic
and syntactic properties. First, the contrast in (22) correlates with the distinction
between D-linked quantifiers and non D-linked quantifiers in Romanian. As
argued by Alboiu (2000), D-linked quantifiers behave like topics and they create
an operator - null constant chain, in which the resumptive clitic acts as an
anaphoric operator, while non D-linked quantifiers need to bind variables within
the IP over which they have scope, and they allow no clitic doubling. The two
types of quantifiers also have distinct syntactic properties. First, quantifiers that
cannot be clitic doubled, i.e. non D-linked ones, require verb adjacency when
moved in the preverbal field, as shown in (23), while D-linked quantifiers are not
subject to such restrictions, as shown in (24).
(23) Pe oricine, (*dacã va fi nev oie), va sprijini (dacã va fi nevoie).

Pe anywho if will be need will-he support if will be need.
"He will support anyone if necessary."
(24) Pe toti elevii, daca va fi nevoie, îi va sprijini.
Pe all students-the if will be need CL will support
"He will support all the students if necessary."
A second syntactic property that distinguishes between the two types of

quantifiers is whether the respective quantifier is exhaustively dominated by the
XP node that undergoes movement. D-linked quantifiers saturate their
quantificational features within the XP they occur in (Erteschik-Shir 1997,
124 DANIELA ISAC
Pesetsky 1987, Dobrovie-Sorin 1990, 1994). In other words, they do not project
their quantificational features to the respective XP and do not bind variables
outside of XP. In contrast, non D-linked quantifiers do project their
quantificational features to the XP they occur in, and they can bind variables
outside the respective XP. In Romanian, bare quantifiers are all non D-linked and
cannot be clitic doubled, because they are exhaustively dominated by the object
XP node. This allows for the percolation of the quantificational features to the
respective XP node, and for the binding of a variable outside the XP.
It seems, therefore, that the restrictions regarding the type of quantifiers that
can be clitic doubled should be expressed in terms of D-linking and in terms of
the ability to project quantificational features, and not in terms of their universal
or existential nature. In Romanian, there are several quantifiers with a universal
interpretation: toti/toate "all", fiecare "each", and oricine (the free choice
"anybody"). Even though they are all universal quantifiers, only the first two can
be clitic doubled, since only the first two are D-linked and cannot project their
quantificational features to the XP they occur in.
(25) I-am salutat pe [XP toti [(studentu)]].

CL-have-I greeted pe all (students-the)
"I greeted them all/ I greeted all the students."
L-am salutat pe [XP flecare [(student)].
CL-have-I greeted pe each (student)
"I greeted each of them/ I greeted each student."
(*L)-am salutat pe [XP oricine].
CL-have-I greeted pe anywho
"I greeted anybody."
3.4 Constraints on identificational focus

The anaphoric nature of the clitic operator can also account for the ban on
that-clauses, infinitival clauses, VPs, and predicative NPs/APs as identificational
foci.
Kiss (1998) suggests that the reason for this restriction is semantic: such
phrases cannot function as identificational foci because they do not denote
individuals, which serve as the primary domain of quantification. These
restrictions easily follow from our aproach: the clitic is an anaphoric operator that
can only bind a variable whose nature must be pronominal.
3.5 Contrastive focus

Another argument in favor of the anaphoric nature of the clitic operator is
the way in which clitic doubling correlates with the distinction between two types
of contrastive focus. As shown by Alboiu 2000, there is evidence for challenging

the exclusive quantificational nature of contrastive focus in Romanian:
contrastive focus involves either a quantificational operator or a non-
quantificational, anaphoric operator. Moreover, the quantificational or non-
quantificational nature of contrastive focus in Romanian is not sensitive to the
syntactic position of the latter. (26) and (27) show that non D-linked contrastively
focused constituents create WCO effects, whereas D-linked ones do not,
irrespective of whether the contrastive object moves overtly or not.
(26) a. Mama lor, a salvat ÃTEIIi nu pisicile.

Mother their has saved puppies-the, not kittens-the
"It's the puppies that their mother saved, not the kittens."
b. *Mama lori a salvat ÃTEIi, nu pisici.
Mother their has saved puppies, not kittens
"It's puppies that their mother saved, not kittens."
(27) a. Mama lor, ÃTEIIi i-a salvat ti nu pisicile
Mother their puppies-the CL-has saved not kittens-the
"It's the puppies that their mother saved, not the kittens."
b. * Mama lori CÃTEIi a salvat ti nu pisici.
Mother their puppies has saved, not cats
"It's puppies that their mother saved, not kittens."
The examples above show that contrastively focused elements on a definite

reading do not observe WCO and they do not involve genuine quantification.
Under our analysis of clitics as anaphoric operators, the expectation is that only
non-quantificational contrasted objects can be clitic doubled. As the examples
below show, the expectation is borne out.
(28) a. CÃTEI, (*ii)a salvat ti nu pisici.

puppies CL-has-he saved t, not kittens
"It is puppies that he saved, not kittens."
b. CÃTEIII *(ii-)a salvat ti' nu pisicile.
puppies-the CL-has-he saved t, not kittens-the
"It is the puppies that he saved, rather than the kittens."
3.5.1 Preverbal vs. postverbal clitic doubled objects. Notice that the object in
(28b) is not subject to the restrictions mentioned in section 2.1. In particular, the
object in (28b) is not preceded by the preposition pe and it is not [+human], but it
is nevertheless clitic doubled. This suggests that the clitic might perform a
different role in (28b), as compared, for instance, to (1). What distinguishes
126 DANIELA ISAC
between these two instances of clitic doubling is the overt position of the object.
The clitic can double both an object in situ and a dislocated one, but the clitic has
different properties in each case. What brings together (29a) and (29b) is the
anaphoric nature of the clitic operator.
(29) a. CL V Object
b. Objecti CLi V ti
When the object is moved, the moved object must be D-linked, or non-
quantificational and it is interpreted either as contrastive focus, or as a topic.
As shown by various authors (Dobrovie-Sorin 1990, 1994, Cornilescu 1997,
Alboiu 2000), if the object is D-linked, both dislocation to topic and dislocation
to focus involve clitic doubling in Romanian. Moreover, the preverbal focus
position in Romanian is always interpreted as contrastive focus. This contrasts
with the situation in Spanish and Italian, but for lack of space, we will not get
into spelling out the differences here. What is important is that the rationale of
the clitic in configurations involving object dislocation (either to topic, or to
focus) is to act as a resumptive pronoun that binds the trace of the moved object.
This contrasts with the role of the clitic in configurations like (29a), where
the clitic anaphorically relates the object to a set of alternatives. This contrast is
paralleled by the different nature of the element to which the clitic relates in these
two instances: in (29a) the clitic mediates the relation between a constant (the
entity or set of entities expressed by the object) and a set of alternatives. In (29b),
the clitic acts as a resumptive pronoun and it mediates the relation between an
empty category -t - and the moved object.
4. Interpretation of the clitic doubled object

A clitic doubled object is subject to certain interpretive retrictions. In
particular, the set of potential readings for the object never includes variable
readings, and kind level ones. We show that this restriction follows from our
analysis of clitics as Focus operators that anaphorically relate the DO to a set of
contextually given alternatives.
4.1. Exclusion of variable readings

In section 3.3 above we accounted for the exclusion of quantificational
readings by showing that there is a mismatch between the nature of the empty
category left behind by QR, i.e. a variable, and the nature of the empty category
that a clitic can anaphorically bind, i.e. a constant. The same analysis can be
extended to account for the exclusion of non-specific indefinites as clitic doubled
objects. Non-specific indefinites introduce variables, and as such cannot be

bound by a clitic, which can only bind null constants. Even though variable
readings are 'tolerated' by pe, they are incompatible with clitic doubling;
indefinites doubled by a clitic are always interpreted as specific.
4.2. Exclus ion of kind readings

There are three pieces of evidence that kind readings are excluded for clitic
doubled objects. First, with verbs like iubi "love", uri "hate", admira "admire",
respecta "respect", which allow the kind reading in object position, plural
definites without pe are felicitous and allow the kind interpretation, but pe DPs in
the plural are infelicitous if the generic reading is intended.
(30) a. Ion iubeste femeile. (generic) (Cornilescu 2000)

Ion loves women-the
"Ion loves women."
b. ??Ion le iubeste pe femei. (generic)
Ion them loves pe women
"Ion loves women."
c. Ion o iubeste  femeie si n-o va pãrãsi (non-generic)
Ion her loves pe woman and not her will abandon
"Ion loves the woman and will not abandon her."
As (30c) shows, pe+DP in the singular is possible, but the interpretation is not
generic, but referential. The interpretation is that Ion loves a particular woman.
Second, kind denoting NPs like tip "type" and fel "kind" cannot appear with
pe, as shown in (31a). In contrast, such NPs can appear without a preposition, as
in (31b), or with a preposition but in the absence of clitic doubling, as in (31c).
(31) a. *Ion o iubeste pe acest tip de femeie.

Ion CL loves pe this type of woman
"Ion loves this type of woman."
b. Ion iubeste acest tip de femeie.
Ion loves this type of woman
"Ion loves this type of woman."
 Ion e obsedat de acest tip de femeie.
Ion is obsessed by this type of woman
"Ion is obsessed with this type of woman."
128 DANIELA ISAC
Thirdly, one can check the D-linked, non modal nature of clitic doubled pe by
testing the compatibility of pe DPs with subjunctive relatives DP.3 As (32) shows,
clitic doubled objects are incompatible with subjunctive relatives.
(32) *Ion o cauta pe fatãcare sä stie fizicã.

Ion CL is seeking pe a girl who sä know (SUBJ) physics
"Ion is looking for a girl who would know physics."
This is because the clitic signals the existence of an epistemically salient referent
of the DO in the context world, while the subjunctive signals that there may not
be any referent for the pe DP in the context world, and that the referent should be
placed in an alternative possible world.
5. Conclusions
We proposed an analysis of Romanian clitic doubled direct objects as
identificational focus. This analysis allows for a syntactic distinction between
identificational and contrastive focus : identificational focus is correlated with
clitic doubling and involves a non-quantificational, anaphoric A-bar chain,
whereas contrastive focus is not necessarily associated with clitic doubling, and it
involves a quantificational A bar chain created by movement of the (object) XP to
a scope position. Also, this view is able to account for why bare quantifiers
cannot be doubled by clitics even if they can show up as objects of the
preposition pe, which must precede [+human] objects in Romanian. Having a
bare quantifier doubled by a clitic leads to a conflicting LF configuration in
which the empty object position must be interpreted at the same time as a null
constant and as a variable. At the same time, one can account for why clitic
doubled objects are interpreted as constants, and never have a
quantificational/variable interpretation, or a kind interpretation. This restriction
follows from our analysis of clitics as Focus operators that anaphorically relate
the direct object to a set of contextually given alternatives.
3
Carlson 1980 stresses that kinds are intensional individuals, presupposing reference to several
possible worlds, so that the genuine kind generic reading is modal and normative. Since clitic
doubled objects range only over the entities of the discourse, or context world, a true modal
dimension, conferred by reference to possible worlds, is missing.
References
Alboiu, Gabriela 2000. The Features of Movement in Romanian. Doctoral dissertation.
University of Manitoba.
Carlson, Greg 1980. Reference to Kinds in English. New York: Garland Publishing.
Cornilescu, Alexandra 1997. "Some notes on the syntax of the subject". Revue roumaine
de linguistique XLII. 101 -147.
2000. "Notes on the interpretation of the prepositional Accusative in Romanian".
Bucharest WPL. 91-107.
Dobrovie-Sorin, Carmen 1990. "Clitic doubling, wh-movement, and quantification in
Romanian". Linguistic Inquiry 21.351-397.
1994. The Syntax of Romanian. Berlin: Mouton de Gruyter.
Erteschik-Shir, Nomi 1997. The Dynamics of Focus Structure. Cambridge: Cambridge
University Press.
Farkas, Donka 1992. "On the semantics of subjunctive complements". Romance
Languages and Modern Linguistic Theory ed. by P. Hirschbühler, 69-105,
Amsterdam & Philadelphia: John Benjamins.
Kiss, Katalin E. 1995. "NP-movement, operator movement and scrambling in
Hungarian". Discourse Configurational Languages ed. by Katalin E. Kiss, 207-243.
Oxford: OUR
1998. "Identificational Focus versus Informational Focus". Language 74:2.245-
273.
Lasnik, Howard & Mamoru Saito 1984. "On the nature of proper government". Linguistic
Inquiry 14.235-289.
1992. Move alpha. Cambridge, Mass.: MIT Press.
Pesetsky, David 1987. "Wh-in-situ: movement and unselective binding". The
Representation of (In)definiteness ed. by Eric Reuland & Alice ter Meulen, 98-130.
Cambridge: MIT Press.
Reinhart, Tanya 1983. Anaphora and Semantic Interpretation. Chicago: University of
Chicago Press.
Rizzi, Luigi 1997. "The fine structure of the left periphery" Elements of Grammar:
Handbook of Generative Syntax ed. by Liliane Haegeman, 281-339. Dordrecht:
Kluwer.
Rochemont, Michael 1986. Focus in Generative Grammar. Amsterdam & Philadelphia:
John Benjamins.
& Peter Culicover 1990. English Focus Constructions and the Theory of
Grammar. Cambridge: Cambridge University Press.
Rooth, Mats 1992 "A theory of focus interpretation". Natural Language Semantics 1.75-
116.
Safir, Ken 1999. "Vehicle change and reconstruction in A'-chains". Linguistic Inquiry
30.587-621.
Szabolcsi, Anna 1981. "The semantics of Topic-Focus articulation". Formal Methods in
the Study of Language ed. by Jeroen Groenendijk, Theo Janssen & Martin Stokhof,
513-541.
130 DANIELA ISAC
1983. "Focusing properties, or the trap of first order". Theoretical Linguistics

10.125-145.
& Frans Zwarts 1993. "Weak islands and an algebraic semantics for scope
taking". Natural Language Semantics 1.235-285.
Vergnaud, Jean-Roger & Maria Luisa Zubizaretta 1992. "The definite determiner and the
inalienable constructions in French and in English". Linguistic Inquiry 23:4.595-
652.
NULL OBJECTS AND VP ELLIPSIS IN
EUROPEAN AND BRAZILIAN PORTUGUESE*
MARY AIZAWA KATO

Universidade Estadual de Campinas (UNICAMP)
1. Introduction
The aim of this paper is to analyze null objects in European and Brazilian
Portuguese. We will show that, while in European Portuguese all types of null
objects can be treated as a unitary variable category, in Brazilian Portuguese null
objects involve two distinct categories: a weak demonstrative and an empty
category that results from the remnant movement of the higher VP.
In the Principles and Parameters theory considerable attention has been
devoted to the possibility of an empty category in object position, a phenomenon
first perceived in Chinese by Huang (1984). The author analyzes this empty
category as a variable bound by a null topic, showing that, in Chinese, it cannot
be a pronominal, since it cannot occur as the object of a complement clause co-
indexed with the matrix subject. His theory predicted that null objects in any
language would also be variables, and thus be banned from such position.1
However, Cole (1987) shows that, contrary to Huang's prediction, languages like
Imbabura Quechua allow null objects in subordinate clauses co-indexed with a
matrix subject, revealing its pronominal nature.2 In later work, Huang (1991)
* I wish to thank the Going Romance 2001 audience for their questions and comments on the version
presented at the Conference. Special thanks are due to Jairo Nunes and João Costa, who discussed
the main ideas presented here. I am also grateful to the anonymous reviewer of the article submitted
for publication for the relevant contributions to improve form and content. Thanks also to Marcello
Marcelino for the editing work. Needless to say, all the remaining mistakes and shortcomings are of
my own responsibility.
1
He adopts the classic Binding theory, which prevents a pronoun to be bound inside its governing
category, and his Generalized Control theory, according to which the same locality conditions hold
for both PRO and pro. These principles would impose contradictory requirements for a null
pronominal in the referred object position.
2
The contrast between Chinese and Quechua can be seen in the following examples:
(i) *Zhangsami shuo Lisi bu renshi ei
Zhangsam says Lisi not knows ("Zhangsam says Lisi doesn't know him.")
132 MARY KATO
suggests that many cases of null objects in Chinese are disguised cases of VP
ellipsis, a proposal corroborated by Otani & Whitman (1991), who worked with
Japanese null objects.3
European Portuguese (EP) and Brazilian Portuguese (BP) are both
languages that allow null referential object constructions with a) no overt
antecedent, as in (la), b) with an expressed antecedent in A'-position, as in (lb),
and c) with an antecedent in the first conjunct of a coordination, as in (lc), the
last case known as VP-ellipsis (VPE).
(1) .  Manuel trouxe agora mesmo. √ E √ BP

the Manuel brought just now.
"Manuel has brought it just now."
b. Esse livro, eu só encontrei na FNAC. √ EP√BP
this book I only found at the FNAC
"This book, I have only found it at the FNAC."
 Pedro comprou  novo livro de Chomsky e Maria também
Pedro bought the new book of Chomsky and Maria also
vai comprar √ EP√BP
goes buy+inf
"Pedro has bought Chomsky's new book, and Maria is also going
to buy it."
Despite the surface similarities, we will show that different categories are
involved in the empty position of the two varieties of Portuguese.
In section 2 the main analyses of the null complement in the Portuguese
literature will be reviewed and in section 3 the problems of adopting such
analyses for BP will be presented. In section 4 I will provide an analysis for BP
gap constructions with and without an antecedent in A'-position, and in section 5
the proposed analysis for the A'-bound null complement will be extended to
account for the phenomenon of VP-ellipsis.
I will be assuming a split VP analysis - VP1, or vP, and VP2, or VP-shell -
as well as a mid-field FocusP, in line with Belletti (1999).
(ii) Juzij nin Marya ei juyanata

Juzi said Marya will love ( " Juzi said that Marya will love him.")
3
According to them a) Null Objects allow the sloppy reading in the second conjunct just like in
VPE, and b) Null Objects exhibit "locality effects" on the sloppy reading, just as English VPE does.
See a review of these studies as well as a different position in Hoji (1998).
NULL OBJECTS AND VP ELLIPSIS 133
2.  unified variable analysis in EP

Like Huang for Chinese, Raposo (1986) analyzes the empty category in (1),
for European Portuguese, as a variable bound by a null operator (cf. (2a)). In his
(1998) paper, on the other hand, he accounts for both the null object in (la) and
the empty category in (lb) as a unitary phenomenon. In the two cases, these
empty category constructions would involve a null definite determiner in object
position, which takes a null NP (pro), as its complement, as represented in (2b).
Pro is then claimed to move to a functional projection F to have its features
identified.4 The two representations show that Raposo has always in mind an
antecedent in A'-position, whether overt or null.
(2) a. OPi [ O Manuel trouxe eci agora mesmo.] (Raposo 1986)

b. (DP)[FPproi+F[TP... [VP [DP D ti]]] (D=null article)(Raposo 1998)
Raposo's movement, or variable analysis accounts for the island effects found in
this variety of Portuguese as can be seen below:
(3) a. (Esse bolo), [o rapaz que trouxe agora mesmo

this cake the boy who brought just now
da pastelaria] era teu afilhado. *EP
from the bakery was your godson
"This cake, the boy who brought it just now from the bakery was
your godson"
b. (Os dólares), eu informel à policía da possibilidade do
the dollars I informed the police about the possibility of
Manel ter guardado no cofre. *EP
Manei have+inf kept in the safe
"The dollars, I informed the police about the possibility that
Manel has kept them in the safe."
Costa & Duarte's (2001) work expands Raposo's (1986) analysis, restricted to
DP variables, to other constituents: PP, AdvP and the VP-shell. Their examples
are also possible in BP:
4
His analysis accounts for an interesting correlation: languages that have null generic articles license
null referential objects and null resumptives; languages that require overt definite articles in generic
NPs do not allow null referential objects and null resumptives. Since our analysis shows that BP has
a different sort of Null Object than EP, the question remains if such correlation holds for the different
types of Null Object in BP.
134 MARY KATO
(4) (Aos teus pais), tu nunca obedeces quando deves. √ EP√BP

to your parents you never obey+2p when should+2p
"Your parents, you never obey them when you should."
(5) (Bern), aquela criatura nunca se comporta em casa. √ EP√BP
well that creature never refl behaves at home
"Well, that creature never behaves at home."
(6) (O Pedro à Maria), apresentei ontem. aqui. √ EP√BP
the Pedro to Maria (I) introduced yesterday here
"I introduced Pedro to Maria yesterday."
As in Raposo's analysis, Costa & Duarte justify their variable analyses using the
fact that the empty complement can never occur inside islands in EP.
(7) (Sobre política), conheço gente que só conversa em bares. *EP

About politics, (I) know people who only talk in bars
"Politics, I know of people who only talk about it in bars."
(8) (Os parafusos na estante), sei de alguém que pode
the screws in the shelves, (I) know+lp of someone who can
colocar *EP
put
"The screws in the shelves, I know someone who can put them in."
3. The impossibility of a unified variable analysis in BP

In this section, we will look at two facts that made researchers of Brazilian
Portuguese reject the variable analysis.
As illustrated in (3 a and b), the prediction of island effects of a variable
analysis holds true for EP. However, as was shown by Galves (1989) and Farrell
(1990), BP null objects can freely occur in islands. This is also true for other
types of empty complements. Both Galves and Farrell propose a pro for the
empty object in BP, but their proposal cannot be extended to the gaps in the other
cases, which do not constitute a DP.5
3
Cinque's (1991) proposal of Left Dislocation (considered important by the reviewer), with a
referential resumptive pro could be considered a possible solution for BP, since it is shown to be
insensitive to islands. However, LD and its gap in BP are not restricted to DPs and, moreover, the
dislocated element can be non-referential.
(i) Dinheiro, eu não conheço ninguém que guarde.
Money I not know anyone who saves
"Money, I don't know of anyone who saves it."
(ii) Sem dinheiro, eu não sei quando ela ficou .
without money I not know when she rested
The variable analysis also fails to account for the possibility of an A-bound
null object in BP, discussed by Bianchi & Figueiredo Silva (1994) and Ferreira
(2000), which is non-existent in EP. The curious fact about this A-bound null
object is that it cannot have a [+human] antecedent. The authors propose that the
A-bound null object is a null pronoun with a [-human] feature.
(9) a. O Joséi sabe que a Maria gostaria de conhecer i.*EP *BP

the José knows that the Maria would like of meet
"José knows that Maria would like to meet him."
b. Esse prato, exige que o cozinheiro acabe de preparar
this dish requires that the cook finishes to prepare
na mesa. *EP√BP
at the table
"This dish requires that the cook finishes preparing it at the
table."
On the other hand, the A'-bound null object, with or without an overt Topic, does
not exhibit the [-human] requirement and could be analyzed as involving a
variable.
(10) a. Esse ator, eu acho que (eu) não conheci i √EP √ BP

this actor I think that (I) not met
"This actor, I think I have not met him."
b. Esse livro, eu acho que (eu) não li i √ EP√BP
this book I think that (I) not read
"This book, I think I have not read it."
But the movement analysis does not account for sentences like (11) in BP, where
there is an apparent DP-extraction:6
"Without money, I don't know when she has been."

6 The reviewer wonders whether Topic constructions and wh-movement constructions would not
share a similar derivation. Though wh-questions, especially the D-linked ones, can sometimes violate
islands (cf. Negräo, 2000), the two derivations do not seem to be related. Thus, if we replace the
topic in (3), which is grammatical, by a wh-expression, the resulting sentence is clearly
ungrammatical.
(i) *Que bolo o rapaz que trouxe agora mesmo da pastelaria era teu afilhado.
136 MARY KATO
(11) a. Esse autor, eu nao conheço [ninguém que tenha

this author I not know [nobody who has
elogiado ] *EP √ BP
praised ]
"This author, I don't know of anyone who has praised him."
b. Esse livro, a Maria conhece [todo mundo que leu ].*EP √ BP
this book the Maria knows [everyone that read ]
"This book, Maria knows everyone that has read it."
 Esse autor,, a Maria lamenta [ofato de que ninguém tenha
this author, the Maria laments [the fact of that nobody has
conhecido / *EP √ BP
met ]
"This author, Maria laments the fact that nobody has met him."
Summary of BP facts:
in BP A'-bound null objects can be [± human] as in EP, but are different
from the Portuguese ones, as they are not sensitive to islands;
BP also differs from EP in licensing A-bound [-human] null objects.
4. Analysis of BP
4.1 Null objects without a linguistic antecedent
The null object may appear without any linguistic antecedent in the sentence
or in discourse, provided that its reference is understood as a deictic category, a
deep anaphor (☞). It is not restricted to null object languages and it appears
sometimes restricted to imperatives in many languages (Kato 1993):
(12) a. Pull ☞ ! English (instruction on doors)

b. Send ☞ by mail. English (instruction on telephone bills)
(13) a. Empurre ☞! √ EP√BP
" Pull!"
b. Envie ☞ pelo correio, √ EP√BP
"Send by mail"
The null deictic is the default null object in emerging grammars, and often
appears with a locative, which is later completed with a demonstrative pronoun
(cf. Kato 1994):
7
In line with Raposo (1999), sentences with a null element in Topic position will be analyzed as an
ordinary case of A'-bound null object in section 4.3.
(14) a. Abi ☞ aqui. (Raquel 1;9,8)

"Open here"
b. Liga ☞ aqui. (Raquel 1;9,8)
"Switch on here".
(15) a. Tila eche aqui (Raquel 2:00)
put away this here
" Remove this"
b. Eu quelo tamhém eche aqui (Raquel 2;01, 16)
I want also this here
4.2 Null object as an "it"-like weak demonstrative in BP

The weak form of demonstratives can take the form of a neuter third person
as in English, a masculine/feminine clitic as in EP or a null form as in BP. The
[-human] interpretation of the Brazilian A-bound null object comes from the lack
of person features of the demonstrative.
(16) Demonstratives Strong (deictic) Weak (anaphoric)

English this/that / ☞ it
EP isto/isso/aquilo/ ☞ o/lo; a/la
BP isso/aquilo/ ☞ Ø
Like EP, BP used to have the third person clitic forms for the anaphoric function
and these were used both for [+human] and [-human] antecedents. But with the
loss of the third person clitics since the middle of the 19th century (Cyrino 1994,
Pagotto 1993, Nunes 1993), BP started exhibiting an apparently defective clitic
paradigm. But Kato (1993) claims that BP still has the whole paradigm, with a Ø-
clitic filling the gap. However, this clitic has only the third person feature, which
is sufficient when the antecedent is [-human], but not when the referent is
[+human], in which case we have the non-oblique forms ele/ela.8
(17) EP BP
me me-
te te-
o/lo; a/la Ø- "it"; ele "him"; ela "her"
8
Also the deictic null object exhibits the same distribution:
(i) [Pegue Ø] ! "Catch it!"
(ii) [Pegue ele] ! "Catch him!"
The non-oblique forms started appearing as strong forms doubling clitics, but today, in BP, they have
grammaticalized as weak forms, replacing the clitics (cf. Kato, forthcoming). Young generations
already accept ele/ela for [-human] entities.
138 MARY KATO
As a pronoun, the Ø-proclitic can appear with an antecedent in A-position.
(18) a. Esse prato, nao permite que você Ø-cozinhe

this dish not allows that you cook
em fogo alto. *EP√BP
in fire high
"This dish does not admit that you cook it in strong fire."
b. Esse prato, nao permite que você o cozinhe em
this dish not allows that you it cook in
fogo alto. √ EP √ written BP
strong fire
"This dish does not admit that you cook it in strong fire."
(19) a. *Essejuiz inao permite que a gente Ø- contradiga quando
this judge not allows that the people contradict when
ele fala. *EP *BP
he speaks
"This judge does not allow us to contradict him when he speaks."
b. Esse juiz i nao permite que a gente contradiga- o quando
this judge not allows that the people contradict-him when
ele fala. √ EP √ written BP
he speaks
c. Esse juizi nao permite que a gente contradigaliquando
this judge not allows that the people contradict he when
ele fala *EP √ BP
he speaks
4.3 A '-bound null objects in BP as the result of VP remnant movement

We saw above in examples (10a-b), repeated below as (20a-b), that contrary
to the A-bound null object (A-NO), which has to be [-human] in BP, the A'-bound
null object (A '-NO) is free to have a [+human] or [-human] antecedent.
(20) a. (Esse atori) eu acho que (eu) nao conheci i √ EP √

b. (Esse livro,) eu acho que (eu) não li i √ EP√BP
My analysis is that, in BP, while the A-NO is a neuter pronoun (Ø-clitic),

the gap in the A-NO results from movement. But before proposing the categorial
nature of the moved element, we will look into the similarities between
VPTopicalization and the A'-NO.
4.4 VPTopicalization
Matos (1992) has shown that VPTopicalization is not sensitive to islands in
EP, and we can see that that also holds true for BP.9 Matos proposes that such
constructions, therefore, do not involve movement, the topicalized VP being
merged in-situ.10
(21) a. Visitar os amigos, a Maria nao conhece [ ninguém

visit the friends the Maria not knows [nobody
que visite /. √ EP √ BP
that visists ]
"Visit her friends, Maria does not know anybody who does."
b. Visitar os amigos, a Maria lamenta [o fato de ninguém
visit the friends, the Maria laments [the fact that nobody
visitar /. √ EP √ BP
visits ]
"Visit her friends, Maria laments the fact that nobody does."
c. Visitar os amigos, [[que a Maria visite /
visit the friends [[that the Maria visits ]
é evidente]. √ EP √ BP
is evident].
"Visit her friends, that Maria does is evident."
Instead of assuming Matos' hypothesis that the VP is merged in-situ in Topic

position, we could follow a different line of reasoning, namely that VP moves,
and that the lack of island effects has to do with the nature of the constituent
moved rather than with structural barriers.11
4.5 VP remnant movement

Let us assume that VPTopicalization is derived by movement. Since both
EP and BP have V-to-I, VP in the two varieties has a copy of the moved verb.12 If
this copy is retained, the result of the movement is what we traditionally call VP
Topicalization, like the sentences in (22). If the copy is erased, constituting a
trace in GB terms, we will have a " remnant movement" of VP (Kayne 1998).
9
However, see Bastos (2001), who shows that island effects in VPTopicalization are sensitive to the
definiteness of the object.
10
Analyzing BP, Bastos (2001) proposes that VPTopicalization in BP can sometimes be derived
through movement, an analysis that we will endorse in section 4.5.
11
Bastos (2001) shows that the more specific the DP, the less acceptable the extraction becomes.
12
I am not assuming the fully inflected V hypothesis in Chomsky (1995), but the classical bare verb
in V, which manifests the infinitive marker if no inflection is added.
140 MARY KATO
We can see now that there are two ways to obtain a A'-NO: a) through
movement of DP or b) through remnant movement of VP, as illustrated in (24a)
and (24b):
(22) Os amigos, a Mari visita todos os anos. √ EP √ BP

the friends the Maria visits every year.
"Her friends, Maria visits them every year."
(23) a. [DP Os amigos] J a Maria i visita v [VP ti tv tj] todos os anos.
b. [VP ti tv os amigos]j a Maria i visitav [VP ]j todos os anos.
Let us suppose that, just like ordinary VPTopicalization, with the verb overt as in
(21a-c), VP-Remnant Movement is not sensitive to islands. Let us further
conjecture that EP chooses movement of DP for (22), while BP chooses VP
remnant movement. This would explain the lack of island effects in BP A'-NO,
and in both EP and BP VPTopicalization.
(24) a. [VP tsu tv os amigos], a Maria nao conhece [ninguem que

visite [vp ]]
b. [VPtsutvos amigos], a Maria lamenta [o fato de ninguém
visitar [vp ]].
c. [VP tsu tv os amigos], [que a Maria visita [VP ] ]
¿evidente. *EP √ BP
Let us now consider VPTopicalization, with the VP containing more than

one complement (cf. 25) and the corresponding VP Remnant Movement
structures (cf. 26):
(25) a. Depositar  dinheiro no banco ,  governo depositou

deposit the money in the bank, the government deposited
[VP / ontem. √ EP √ BP
yesterday
"Deposit the money in the bank, the government did yesterday."
b. Dar urn presente para/a os seus pais, a Maria pretende dar
give-inf a gift to her parents, the Maria intends to give
[vp ] sem demora. √ EP √ BP
without delay
"Give a gift to her parents, Maria intends to do it without delay."
 Partilhar segredos com  Pedro, a Lulu partilha [VP ]

share secrets with the Pedro, the Lulu shares
desde o ano passado. √ EP√BP
since the year last
"Share her secrets with Peter, Lulu does since last year."
(26) a. Dinheiro no banco, o governo depositou [VP ]
money in the bank the government deposited
ontem. √ EP√BP
yesterday
"Money in the bank, the government deposited it yesterday."
b. Um presente a/para os seus pais, a Maria pretende
a gift to her parents, the Maria intends
dar [VP ] sem demora . √ EP√BP
to give without delay.
"A gift to her parents, Maria intends to give it without delay."
 Segredos com  Pedro , a Lulu partilha[VP______]desde.
secrets with the Pedro, the Lulu shares since
o ano passado √ EP√BP
the year last
"Secrets with Peter, Lulu has shared since last year."
But if EP allows topicalization of a VP-shell, then we have to assume that it has

remnant movement of the lower VP, as it contains the trace of the verb:
If VP-movement is what is behind the lack of sensitivity to islands, we

should expect that topicalization of the VP-shell in EP would be possible from an
island, but we saw that EP cannot move any type of constituent from such
contexts (cf. (8)).
142 MARY KATO
Differently from EP, VP-shell constructions can appear topicalized from any
kind of islands in BP:13
(28) a. Dinheiro no banco, [que  governo deposite ] é o que 

money in the bank [that the government deposits ] is what the
povo espera. *EP √ BP
people expect,
b. Urn presente para os pais , eu conheço alguém que vai dar
a gift to her parents, I know (of) someone who goes give
no Natal. *EP √ BP
for Xmas
 Segredos com  Pedro, a Maria lamenta  fat
secrets with the Pedro, the Maria laments the fact
dela ter partilhado quando colegas. *EP √ BP
of her hav(ing) shared when (they were) classmates
4.6 Split VP
The question that still remains to be answered is: why do we have
differences regarding islands between EP and BP if the former can also have VP
remnant movement? Assuming a split VP hypothesis (Hale and Keyser, 1993;
Chomsky 1995), we propose that what can be moved without island restrictions
is not the lower VP, but the higher VP (or vP).
Recall that the BP topicalized forms were proposed to have the same nature
as VP Topicalization, except that in the latter the verb is retained.
VPTopicalization in both EP and BP can be analyzed as movement if we assume
that sensitivity to islands is not conditioned solely in terms of structural barriers,
but by the type of constituent involved in the movement.14
Assuming a movement analysis for VPTopicalization, now reanalyzed as
vPTopicalization, in both EP and BP, let us show the sort of constituent that we
have to move in order to obtain the order V DP PP in the topic position. Using
(25a) to illustrate, we have the following constituent:
13
There is no corresponding translation into English in these cases.
14
Other factors can also be involved such as the definiteness of the complement, as shown by Bastos
(2001).
Observe that VPTopicalization in both EP and BP has to move vP and not VP. If
the lower VP were moved, the order would be DO V PP.
While in vPTopicalization the copy of the verb in v is retained as in (25a-c),

what we have in vP Remnant movement is the trace of V, after it has raised to T,
as in the Brazilian sentences (26a-c).15 Let us illustrate the difference between EP
and BP using sentences (25a) and (26a), repeated here as (30a and b),
superficially identical in the two varieties of Portuguese:
(30) a. [ Depositar dinheiro no banco],  governo depositou ontem.
15
Bastos (2001) shows that the Copy Theory of Movement (Chomsky, 1993) and Distributed
Morphology (Halle & Marantz, 1993) solve the problem of the two verbal instances and their
difference in verbal inflexion.
144 MARY KATO
b. [Dinheiro no banco],  governo depositou ontem.
The representation of the moved constituent in (30a) in both EP and BP is (31a),

a bracketed version of (29a). The representation of the topicalized constituent in
(30b) is different in the two varieties: (3 lb) for BP and (31c) for EP.
(31) a. [vP ti depositarv [VP o dinheiro tv no banco]] EP BP

b. [vp ti tv [Vp dinheiro tv no banco ]] BP
c. [VP dinheiro tv no banco ] EP
Thus, when the verb appears overtly as in (31a), EP and BP both have moved vP;
when the verb does not appear overtly, what is moved is the lower VP-shell in
EP, and the full vP in BP. This explains why only EP is sensitive to islands. It
moves arguments and the VP-shell.
Summarizing our proposal for A'-NO: what is ordinarily called
VPTopicalization in both EP and BP is the result of vP movement to an A'-
position, with the copy of the verb in v not erased; BP allows vP Remnant
movement, which results in A '-NO; EP moves arguments and the VP-shell; vP
movement is not subject to islands.16
4.7 vP Topicalization as p art of VPEllips is (VPE)

In the previous sections, we showed that A-NO and A'-NO in BP cannot be
analyzed as a unitary phenomenon as in EP, because the former has necessarily a
[-human] antecedent and the latter is free of such restriction. In this section, we
will show that VPE shares with vPTopicalization and vP Remnant movement the
property of licensing both [+human] and [-human] antecedents, a fact that we will
draw on to propose a movement analysis for VPE.
Matos and Cyrino (2001) provide an easy clue to identify null objects and
VPE in BP, using the identity of the verb: in VPE the verbs are identical as in
(32):
(32)  Joäo comeu a banana, e a Maria

the Joäo ate the banana and the Maria
também comeu √ EP √ BP
also ate
"John ate the banana, and Maria also did."
16
A more radical proposal would be to consider that EP always moves the VP-shell in its remnant
form.
When the empty category is a null object (A-NO), the governing verbs are
different, as in (33):
(33) .  João descasc a banana, e Maria comen . ??EP √ BP

the Joäo peeled the banana and Maria ate
"João peeled the banana and Maria ate it."
b. O João trouxe o jornal, mas a Maria.
the Joäo brought the newspaper but the Maria
nao leu _ *EP √ BP
not read
"John brought the newspaper, but Maria didn't read it."
Examples such as (33) are generally unacceptable in EP which requires an overt

third person clitic a/o, while in BP the sentences are good because the empty
category is interpreted as a weak neuter clitic, an A-NO, as was seen in section
4.2.17
For the same reason, this kind of coordination with a gap in object position
is ruled out in BP when the antecedent is [+human]:
(34) *João trouxe a Maria, mas Pedro nao beijou .

Joäo brought Maria but Pedro not kissed
"Joäo has brought Maria, but Pedro hasn't kissed her."
In contrast, when we have real cases of VPE, namely with identical verbs, the
[+human] restriction is not operative. (35) is well-formed both in BP and EP.
(35) Joäo cumprimentou a Mariai, mas Pedro

Joäo greeted the Maria but Pedro
nao cumprimentou i. √ EP √ BP
not greeted
"Joäo has greeted Maria, but Pedro hasn't."
A solution to this contrast can be advanced if we assume that VPE involves

vPTopicalization. In other words, the second conjunct of the coordination in VPE
17
Matos & Cyrino consider such sentences good for both EP and BP, but, according to many
Portuguese speakers (Joäo Costa, p.c.), EP allows Null Objects only with verbs that can be used
intransitively.
146 MARY KATO
is a vP Topicaiization (vPTop) structure as in (36a), and the whole coordinate

structure has a structure like (36b):
(36) a. Cumprimentar a Maria,  Pedro nã cumprimentou. vPTop

greet the Maria, the Pedro not greeted
"Greet Maria, Pedro didn't."
b.  Joäo [cumprimentou a Maria], mas [cumprimentar a Maria]
the João greeted the Maria but greet the Maria
 Pedro nao cumprimentou [vP ]. VPE √ EP √ BP
the Pedro not greeted
"Joäo greeted Maria, but Pedro didn't."
On the other hand, when we have an A-NO and not VPE in coordination,
the gap is a weak demonstrative, and no underlying VP Topicalization has to be
assumed:
(37) .  João descas  bi e a Maria comeu-Ø i

b. *OJoão descascou a banana e [[comer a banana] a Maria
the João peeled the banana and eat the banana the Maria
oeu]
ate
What I defend here is that when there is no pronoun to recover the antecedent,
vP moves to be identified. The vP in the first conjunct acts as a discourse
antecedent for the topicalized vP, which, though erased at PF, acts at LF as the
topic that recovers it.
5. Potential problems for the vP Remnant movement analysis

5.1 Cases of a single complement of a ditransitive verb as Topic18
One possible theoretical argument against vP Remnant movement in BP
comes from constructions where only one complement of a ditransitive verb
appears fronted. We proposed that EP moved smaller constituents (DPs, PPs, and
the VP-shell), and that BP moved only the larger vP with its traces. If the analysis
is in the right direction, one would expect topicaiization of the object alone or of
the indirect object alone to be impossible in BP. However, this is not accurate.
Both (38a) and (38b) are possible in BP.
I thank Jairo Nunes for pointing out this problem.

(38) .  noivo, a Maria ja apresentou aos pais. √ EP √ BP

the fiancé, the Maria already introduced to the parents
" Her fiancé, Maria has already introduced him to her parents."
b. Aos pais, a Maria já apresentou  noivo. √ EP √ BP
to the parents the Maria already introduced the fiance
"To her parents Maria has already introduced her fiancé."
Though, on the surface, both varieties of Portuguese may exhibit identical forms
in (38), I maintain that EP moves the smallest constituent that constitutes the
topic (DP, PP, VP-shell), while BP always moves the largest constituent, namely,
the whole vP, and whatever remains inside it. This movement can take place after
one of the complements has scrambled out of VP to a mid-field FP position,
proposed by Belletti (1999) to be higher than vP.19
(39) a. [PP aos paiSj ],  Maria, já apres entou [vP ti [VP  noivo tv tJ] EP
b. [vP ti tv [tk tv aos pais]], a Maria, já apres entou
[FPO noivok [vP ] BP
(40) a, [DP  noivok], a Maria, já apresentou [vP ti [VP tk tv aos pais] EP
b. [vP ti tv  noivo t 10]j a Maria já apresentou [FP para os pais
[VP_] BP
In other words, in order to move everything except the scrambled-out element,

what is moved is the vP.
5.2 Prepositionless Topics20

The second potential problem for my analysis has to do with the possibility
of preposition drop in A'-position (41d), which is not possible before extraction
(41b) and neither in vPTopicalization (41f):
(41 ) .  Pedro gosta desse livro.

the Pedro likes of this book
"Pedro likes this book."
b. *Ø Pedro gosta esse livro.
the Pedro likes this book
19
The unmarked word order is SVDO IO. If the indirect object is a topic, it cannot stay in sentence-
final position, where it gets heavy stress. The effect of scrambling, or of the further movement to
topic position is precisely the type of P-movement proposed by Zubizarreta (1998).
20
1 am grateful to the reviewer to have pointed out this problem.
148 MARY KATO
c. Desse livro, o Pedro gosta .

of this book the Pedro likes
d. Esse livro, o Pedro gosta .
this book the Pedro likes
e. G ostar desse livro, o Pedro gosta .
like of this book, the Pedro likes
f. * G ostar esse livro, o Pedro gosta .
Like this book the Pedro likes
If the Topic derives from a vP Remnant movement, one would expect that the
preposition would always appear as in (41c).
Before solving the problem, we should stress the fact that preposition-drop
is not a general process. It is possible with verbs that cannot assign case to their
complement, like gostar and precisar and need a preposition as a case marker.
The following contrast should be observed between the sentences in (41) and
those in (42), where we have the verb depender, which selects a PP
complement:21
(42) a. Pedro depende da Maria desde criança.

Pedro depends on the Maria since child
"Pedro depends on Maria since he was a child."
b. * Pedro depende a Maria desde criança.
c. Da Maria, Pedro depende desde criança.
on the Maria Pedro depends since child
d. *A Maria, Pedro depende desde criança.
e. Depender da Maria, Pedro depende desde criança.
Depend on the Maria, Pedro depends since child
f * Depender a Maria, Pedro depende desde criança.
The solution that I propose is based on the assumption that, contrary to verbs like
depender, verbs like gostar select a DP, but have no case features (cf. Raposo
1992). This DP can have oblique Case added before merging or it may have
"default" case. If it has Case added, it will require case checking, which can only
be done if the numeration contains a preposition, which, adjoined to the DP, will
provide the condition for the V+prep[+oblique] and the DP[+oblique] to be checked,
eliminating the uninterpretable Case feature.
21
Notice that the caseless verbs do not require a preposition in English:
(i) I like these books.
(ii) I need these books.
(43) a.
The "default" case, unlike added Case, appears with DPs in positions that do not
require Case-checking.22
Sentences (41)a,  and e would start with a numeration containing the
preposition and a DP with oblique case features. As the preposition has the same
oblique Case feature as the DP, the derivation succeeds. Sentences (41)b and f
are ungrammatical because the DP argument appears in a checking position, but
the "default" case is not a case that satisfies checking requirements, as no verb or
preposition has a "default" case feature to check.
Let us now suppose that the numeration does not contain the preposition
and the DP has "default" case, as in configuration (43a). This has to be
considered a possible beginning for a convergent derivation in Brazilian
Portuguese since a sentence like (41a) is well-formed. My claim is that, if the DP
ends up in a position that allows it to manifest the "default" case, namely the
nominative in BP, then the derivation converges.
Let us illustrate the derivation of sentences with gostar, starting the
computation with and without the preposition.23 I will use a pronominal
complement to make things clearer.
22
Kato (1999) shows that each language chooses a specific morphological case for its "default"
case, which appears in topic and predicate position.
(i) a. It's ME (ii) a. ME, I drink beer.
b. C'est MOI b. MOI, je bois de la bière.
. Soy YO. c. YO, pro bebo cerveza.
d. Sou EU d. EU, eu bebo cerveja.
23
I am assuming the traditional view that the verb may be inserted without inflection, and that
inflectional morphology appears in INFL.
150 MARY KATO
(44) a. De mim, Joäo gosta.

of me Joäo likes
"John likes me."
b. [vp gost ar de[+oblique case] mim[oblique case]]
c. [vP João gostarv [vp tv de mim]]
d. [IP Joãoi gostar+I [vPti tv[VPtv de mim]]]
e. [vptitv [Vp tv de mim]]j [IP Joãoi gosta [vP_ tj]]
(45) a. EU, João não gosta.
b. [VP gostar EU [default c a s e ] ]
c. [vP Joãoi gostarv [VP tv EU[default case]]]]
d. [IP Joãoi gostarv+I [vP ti tv [VP tv EU [ d e f a u l t case]]]]
e. [vp ti tv [VP tv EU[default case]]] [IP João, gostarv+I [vP ti]]
The DP that remains in A'-position can manifest the 'default' case.
6. One empirical phenomenon in favor of the vP remnant movement

analysis
Empirical evidence for vP remnant movement in BP is hard to find, but we
can attest it with topicalization of predicates with light verbs, whose natural place
is the v-head. The examples below show that while vPTopicalization is possible
in both EP and BP, vP Remnant movement is only possible in BP.
(46) a. Dar o golpe do bau na Maria a gente acha que

give the strike of the trunk on Maria the people think that
o José está querendo dar [vP ] √ EP √ BP
the José is trying (to)give
(dar o golpe do baú = "marry a rich wife")
b. Por a boca no trombone  Pedro pôs [vP ]
put the mouth in the trumpet the Pedro put
durante a reunido. √ EP√BP
during the meeting
(por a boca no trombone = "denounce scandals, set up a great
fuss")
(47) a. O golpe do baú na Maria, a gente acha que o José está
the strike of the trunk on Maria, the people think that the José is
querendo dar [vP ] *EP √ BP
trying (to) give
b. Aboca no trombone, ο Pedro pôs [vP J durante

the mouth in the trumpet, the Pedro put during
a reunido *EP  BP
the meeting
Even with idioms containing light verbs, topicalization is possible from islands in
BR
(48) A boca no trombone, nao conheço ninguém que se atreva

the mouth in the trumpet (I) not know (of) anybody who dares
apor [vP J durante a reunido. *EP  BP
to put during the meeting
The possibility of A'-contructions with such common idioms may well be the
trigger for the Brazilian child to learn that the object gap in his/her language
involves vP Remnant movement. The trigger can be much simpler than the
examples given:
(49) A boca no trombone,  Lulu pôs durante a festa.

The mouth in the trumpet the Lulu put during the party
7. Final remarks
We showed, in this paper, that, though EP and BP exhibit similar
phenomena on the surface, a closer scrutiny reveals that their gaps in complement
position behave differently in island constructions.
The major theoretical contribution of this paper was to propose that absence
of islands effects in topicalized structures in BP cannot be explained using only
"merge in-situ" of the topic. In the analysis proposed here lack of sensitivity to
islands was assumed to be not only due to merge-in-situ of the Topic (the case of
Left Dislocation), but also conditioned by the type of constituent that undergoes
movement. The category vP was proposed to freely move out of islands both in
its full vP Topicalization form and in its remnant form.
References
Bastos, A.C. 2001. Fazer, Eu Faço! Topicalização de Constitiiintes Verbais em Portuguâs
Brasileiro, MA Thesis, Universidade Estadual de Campinas.
Belletti, Adriana 1999. "Inversion as Focalization". ms., Università di Siena.
Bianchi,Valentina & Maria Cristina Figueiredo Silva 1994. "On Some Properties of
Agreement-Object in Italian and in Brazilian Portuguese". Issues and Theory in
152 MARY KATO
Romance Linguistics; Selected Papers from LSRL XXIII, ed. by Michael Mazzola,
181-97. Washington, DC: Georgetown University Press.
Chomsky, Noam 1993. "A Minimalist Program for Linguistic Theory". Hale & Keyser
1993. 1-52.
-1995. The Minimalist Program. Cambridge, Mass.: The MIT Press.
Cinque,Guglielmo 1991. Types ofA'-dependencies. Cambridge, Mass: The MIT Press.
Cole, Peter 1987. "Null objects in Universal Grammar". Linguistic Inquiry 18:597-612.
Costa, Joäo & Inez Duarte 2001. "Objectos Nulos em Debate" ms, Universidade Nova &
Universidade de Lisboa.
Cyrino, Sonia 1993. "Observando a Mudança Diacrônica no Português do Brasil: Objeto
Nulo e Clíticos". Roberts & Kato 1993. 163-184.
Farrell, Patrick 1990. "Null Objects in Brazilian Portuguese". Natural Language and
Linguistic Theory 8.325-346.
Ferreira, M. B. 2000. Sujeitos Nulos e Objetos Nidos no Português Brasileiro. MA
Thesis, Universidade Estadual de Campinas.
Galves, Charlotte 1989. "O Objeto Nulo no Português Brasileiro; Percurso de urna
Pesquisa". Caderno de Estudos Lingüísticos 17.65-90.
Hale, Kenneth & Samuel Jay Keyser 1993. "On Argument Structure and the Lexical
Expression of Syntactic Relations. Hale & Keyser 1993. 53-110.
— eds. 1993. The View from Building 20: Essays in Linguistics in Honor of Silvain
Bromberg. Cambridge, Mass: The MIT Press.
Halle, Morris & Alec Marantz 1993. "Distributed Morphology and the Pieces of
Inflection". Hale & Keyser 1993. 111-176.
Hoji, Hajime 1998. "Null Object and Sloppy Identity in Japanese". Linguistic Inquiry
29:1.127-152.
Huang, C-T James 1984. "On the Distribution and Reference of Empty Pronouns."
Linguistic Inquiry 15.531-574.
1991. "Remarks on the Status of the Null Object". Principles and Parameters in
Comparative Grammar, ed. by Robert Freidin, 56-76. Cambridge, Mass: MIT
Press.
Kato, Mary Aizawa 1993. "The Distribution of Null and Pronominal Objects in Brazilian
Portuguese." Linguistic Perspectives on the Romance Languages: Selected Papers
from the XXI Linguistic Symposium on Romance Languages, ed. by William Ashby,
Marianne Mithun, Giorgio Perissinoto & Eduardo Raposo, 225-235. Amsterdam &
1994. "A Theory of Null Objects and the Development of a Brazilian Child
Grammar". How Tolerant is Universal Grammar? ed. by Rosemarie Tracy & Elsa
Lattey, 125-153. Tübingen: Niemeyer.
— 1999. "Strong Pronouns, weak Pronominals and the Null Subject Parameter".
Probus 11:1.1-37.
(forthcoming) "Pronomes Fortes e Fracos na Gramática do Portugués Brasileiro".
Revista Portuguesa de Filología.
Kayne, Richard 1998. "Overt vs. Covert Movement". Syntax 1.128-191.
Matos, Gabriela 1992. Construções de Elipse de Predicado em Português: SV Nulo e
Despojamento. Doctoral dissertation. Universidade de Lisboa.
& Sonia Cyrino 2001. "Elipse de VP em Português Europeu e Português
Brasileiro". presented at the Colóquio sobre o Portugués Europeu e Portugués
Brasileiro, Fortaleza.
Negrõo, Esmeralda Vailati 2000. "Wh-extractions and Relative Clauses in Brazilian
Portuguese". D.E.L.T.A 16 (Special Issue): 141-164.
Otani, Kazuyo & John Whitman 1991. "V-raising and VP-ellipsis". Linguistic Inquiry
22:345-358.
Nunes, Jairo (1993). "Direçào de Cliticização, Objeto Nulo e Pronome Tônico na Posiçäo
de Objeto em Portugués Brasileiro". Roberts & Kato 1993. 207-222.
Paggotto, E. 1993. "Clíticos, Mudança e Seleçao Natural". Roberts & Kato 1993. 185-
206.
Raposo, Eduardo 1986. "On the Null Object in European Portuguese". Studies in
Romance Linguistics ed. by Osvaldo Jaeggli & Carmen Silva-Corvalán, 373-390.
Dordrecht: Foris.
1991. Teoría da Gramática: a Faculdade da Linguagem. Lisboa: Caminhos.
1998. "Definite/zero Alternations in Portuguese". Romance Lingusitics:
Theoretical Perspectives ed. by Armin Schwegler, Bernard Tranel & Myriam Uribe-
Etxebarria, 197-212. Amsterdam & Philadelphia: John Benjamins.
Roberts, Ian & Mary Aizawa Kato eds. 1993. Português Brasileiro: Uma Viagem
Diacrônica. Campinas: Editora da Unicamp.
Zubizarreta, Maria Luisa 1998. Prosody, Focus, and Word Order. Cambridge, Mass.: The
MIT Press.
FROM NON-IDENTITY TO PLURALITY
FRENCH DIFFÉRENT AS AN ADJECTIVE
AND AS A DETERMINER
BRENDA LACA & LILIANE TASMOWSKI

Université de Paris 8 Universiteit Antwerpen-UlA
UMR 7023 FWO 7010
0. Introduction
One of the recurrent questions in contemporary research on the syntax and
semantics of determiners concerns the status of the elements introducing
indefinite or existential NPs. In DRT and related approaches that treat indefinites
as e-type expressions, elements like three or some are analysed as cardinality
predicates, in GQ-approaches they are analysed as expressions of the relation
between two sets (or, equivalently, as functions from sets to generalized
quantifiers). A host of mixed approaches exist, some of them associating each
analysis with a particular position in the syntactic-semantic partition of the
sentence (so for instance Löbner 1987; Diesing 1992), others deriving one sort of
denotation from the other via type-shifting operations (Partee 1987; Landman
2002), others still distinguishing between denotation and representation
(Szabolcsi 1997). Against this background, the study of items that may introduce
indefinite NPs and have distinct semantic and syntactic properties when they
function as adjectives or as determiners may prove rewarding. It may help us
understand which properties belong to determiners because they are determiners
(i.e., because they occupy a given syntactic position in the N projection). Such
items are relatively rare. French différents (as well as divers, certains) is
undoubtedly one of them.
We are grateful to  Beyssade, I. Comorovsky, F. Corblin, L. De Ryck, D. Godard, J.

Lowenstamm, J.-M. Marandin, and B. Spector for judgements and discussion. None of them
necessarily agrees with the views presented here.
156 BRENDA LACA & LILIANE TASMOWSKI
French différents can function as a symmetrical relational adjective and as

an indefinite plural determiner, and is peculiar both as a relational adjective and
as an indefinite plural. In this paper, we will first introduce the distinction
between adjectival and determiner différent, and we will then successively
examine the semantics of A-différent and that of D-différents. Our hypothesis is
(a) that one reading of A-différent, which we will call the NP-internal reading,
provides the link between adjectival and determiner-like uses of différents, and
(b) that some semantic properties of Ό-différents stem directly from the
semantics of A-différent, whereas others (in particular distributivity) stem from
the fact that D-différents cannot be an adjective.
1. French différent as an adjective and as a determiner

1.1 Syntax
In contemporary French, différent has two syntactically distinct uses, as a
relational adjective (A-différent) and as a determiner (Ό-différents).
A-différent (a) can appear as a predicate or in adnominal position, (b). is
gradable, since it can be modified by intensifying adverbs and appear in the
comparative, and (c) has a second argument position to be filled by a de-NP:
(1) a. Je vous imaginais différente.

"I thought you were different."
b. ll est devenu un homme différent.
"He became a different man."
(2) a. Ce texte existe en deux versions légèrement/entièrement
différentes.
"This text exists in two slightly/entirely different versions."
b. Cette version de votre article est moins différente de la
précédente qu 'annoncé.
"This version of your article is less different from the precedent
one than announced."
(3) a. Je vous imaginais différente de votre soeur.
"I thought you were different from your sister."
b. ll souhaitait un avenir différent de celui de son frère.
"He wanted a different future than his brother's."
FROM NON-IDENTITY TO PLURALITY 157
D-différents appears in prenominal position in the plural, and it is in

complementary distribution with indefinite determiners, in particular with de(s).
Since French admits no determinerless NP in argument position, this counts as
evidence for its syntactic status as a determiner:
(4) a. Différents organismes ont demandé sa collaboration.

"Various organisations requested his collaboration."
b. (* De/* Plusieurs) différents organismes...
"(De /Several) different organisations.."
Ό-différents is not gradable and has no second argument position.

Correspondingly, prenominal différent modified by an intensifying adverb or
with an explicit second argument exhibits adjectival syntax:
(5) a. *Si différentes personnalités ne pourront j amais s'accorder.

b. De si différentes personnalités ne pourront jamais s'accorder.
"So different personalities will never get to agree."
(6) a. *ƒ/ a différentes aspirations de ce que tu t'imagines.
b. ll a de bien différentes aspirations de ce que tu t'imagines.
"He has very different ambitions than what you imagine."
D-différents patterns like a plural indefinite determiner. Like the plural

indefinite des, and unlike the other indefinite determiners, it is not possible in
NPs lacking a lexical nominal head:
(7) Les étudiants étaient très fâchés. (Certains/ Plusieurs/ *Différents/

*Des) sont venus se plaindre.
"The students were very angry. (Some / Several / *Different/ *Des)
came to complain."
Prenominal différents, on a par with some plural indefinites, can be

preceded by a definite determiner:
(8) a. (les/ces) (trois/quelques) objections qu'il a soulevées

"(the/these) (three/few) objections that he put forward"
b. (les/ces) différentes objections qu'il a soulevées
"(the/these) different objections that he put forward"
As is well known, there are at least two competing analyses for

constructions involving plural indefinites preceded by a definite determiner, one
of them treating {the three) as a complex determiner, the other treating the
indefinite as a numerical adjective in an NP-construction {three objections) (see
Landmann 2002 among many others). As far as différents is concerned, the fact
that the possibility of adverbial modification is preserved in this context provides
evidence for its adjectival status in the case at hand:
(9) {les/ces) {si/bien) différentes objections qui ont été soulevées

"(the/these) (so/very) different objections that have been put forward"
1.2 The readings ofA-différ ent and D-differents

The distributional properties distinguishing D-différents from A-différent
correlate with important semantic differences.
When A-différent lacks an explicit second argument, it has several distinct
readings, only one of which is paralleled by D-différents. Following Carlson
(1987) and Moltmann (1992), in Laca/Tasmowski (2001a), we distinguished
among (i) external, (ii) quantified antecedent, (iii) dependent plural, and (iv) NP-
internal readings of K-différent. These readings differ in the way in which the
arguments of the relation denoted by différent are obtained.
In external readings (i), the second argument of the relation is obtained
deictically or anaphorically from the context:
(10) a. Jean a proposé une solution différente.

"John proposed a different solution."
b. Cette photo est légèrement différente.
"This picture is slightly different."
Quantified-antecedent (ii) and dependent plural readings (iii) can be

subsumed under the label of sentence-internal readings. In both cases, the
arguments of the relation are obtained via a referential dependency upon some
other element in the sentence, which acts as a licensor for the sentence-internal
reading. In (11a-b), the licensor is the quantifier chaque enfant, in (12a-b), the
plurality denoted by Jean et Marie'.
(11) a. Chaque enfant a reçu un cadeau différent/des cadeaux

différents.
"Every child got a different present/ different presents."
b. Chaque enfant est différent.
"Every child is different."
(12) a. Jean et Marie ont reçu des cadeaux différents.

"John and Mary got different presents."
b. Jean et Marie sont très différents.
"John and Mary are very different."
In NP-internal readings (iv), the arguments of the relation are obtained

inside the NP. If the NP is introduced by a cardinal determiner, the arguments
are the atomic elements of the plurality denoted by the N. If the NP is introduced
by des, there might be some latitude as to the possibilities of having
subpluralities of the plurality denoted by the N as arguments. Thus, (13a)
requires that each of the four letters be different from the others, whereas (13b)
can possibly describe a situation in which the same letter occurs more than once:
(13) a. Ce carré contient quatre lettres différentes.

"This square contains four different letters."
b. Ce carré contient des lettres différentes.
"This square contains different letters."
This parallels the behavior of NP-internal readings of relational adjectives:

again, (14a) requires each line to be parallel to the others, while (14b) can
describe a situation with different groups of parallels:
(14) a. Ce carré contient quatre lignes parallèles.

"This square contains four parallel lines."
b. Ce carré contient des lignes parallèles.
"This square contains parallel lines."
D-différents exhibits neither external nor sentence-internal readings. In fact,

it only has one interpretation, which is close to the NP-internal reading of des N
différents. Thus, (15) has three possible interpretations, an external reading
paraphrased in (a), a dependent plural reading paraphrased in (b) and an NP-
internal reading paraphrased in (c).
(15) Lui et moi, nous avons obtenu ce résultat par des procédés différents.
a. We got this result by different procedures than those used by
somebody else.
b. The procedures by which he got this result are different from the
procedures by which I got it (and vice versa).
 We got this result by various procedures.
In contrast, (16) has only one reading, roughly corresponding to (c):

(16) Lui et moi, nous avons obtenu ce résultat par différents procédés.
Note that the unavailability of external and sentence internal readings is not
linked to prenominal position as such, but to the syntactic status of différents as a
determiner. The prenominal différents with adjectival syntax in (17) exhibits the
same range of readings as (15):
(17) Lui et moi, nous avons obtenu ce résultat par de bien différents
procédés.
1.3 From A-différent to D-différents

It can be assumed that D-différents evolved diachronically from A-différent
through a process of gradual specialization clustering prenominal position,
determiner-like syntax and NP-internal readings. 17th and 18th Century French
differs from contemporary usage on several accounts, showing that these three
properties did not cluster in the same way as today.
Prenominal différents alternated between a determiner-like and an adjectival
syntax (18a-b), and it could be preceded by cardinal indefinites and other
quantity expressions (19a-c):
(18) a. Ces humeurs ont différentes qualités. (BOSSUET, J-B / De la

connaissance de Dieu / 1704)
"These humours have different qualities."
b. Les membres extérieurs reçoivent aussi de différentes
dispositions. (BOSSUET, J-B / De la connaissance de Dieu / 1704)
"External members get also de different dispositions."
(19) . l' église fut troublée par plusieurs différentes contestations
(ABBADIE J / Traité vérité relig. chrét. 2 /1684)
"the Church was troubled by several different challenges"
b. Après ces accidents, j'ai eu beaucoup de différentes
indispositions. (GUILLERAGUES / Lettres portugaises / 1669)
"After those accidents, I had a lot of different indispositions."
c. Dans un grand nombre de différents poèmes (AUBIGNAC, ABBÉ
D' / Conjectures académiques /1676)
"In a large number of different poems"
Furthermore, prenominal différents could exhibit sentence-internal readings.

Thus, in (20a) the arguments of the relation are obtained via a referential
dependency upon tous les deux, in (20b) they are obtained via a referential
dependency upon ils, and in (20c) via a referential dependency upon tout ce que
vous venez d'appeler:
(20) a. Ce qui fait que les uns vont à la guerre et que les autres n 'y vont
pas est ce même désir qui est dans tous les deux accompagné de
différentes vues. (PASCAL, BL. / Pensées sur la religion / 1662)
"What makes some of them go to war and the others not to go is
this very same desire, which is in the two cases accompanied by
different ways of seeing things."
b. Ils étoient presque tous habillés de différentes manieres.
(AuLNOY.M.C/ La chatte blanche / 1698)
"They were almost all dressed in different fashion."
 mais, dit la reine, tout ce que vous venez d'appeler vient en
différentes saisons (AULNOY.M.C/ La chatte blanche / 1698)
"but, said the queen, all the things you have just mentioned come
in different seasons"
Note, however, that in the above contexts it is impossible to decide whether

we have adjectival or determiner-like syntax, since the preposition introducing
the différents-Phrase either obligatorily cancels determiner DE (*de de) or
independently favors determinerless NPs (en hiver, en bonnes mains). Since we
have not found examples of clear determiner-like syntax with other than NP-
internal readings, we assume that, in the examples given above, we have to do
with a prenommai A-différent.
The above mentioned facts can thus be interpreted in a uniform way.
Contemporary French differs from 17th and 18th Century French in as far as the
prenominal position of indefinite nominal projections has been specialized for D-
different, hence the preference for the alternative with postposed différent in
cases like :
(21) a. Les membres extérieurs reçoivent aussi des dispositions

différentes. (cf. 18b)
b. l' église fut troublée par plusieurs contestations différentes
(cf.19a)
 Ils étaient presque tous habillés de manières différentes, (cf. 20b)
Even an exceptional contemporary example such as (22a) is only

exceptional as far as prenominal position is concerned. In all other respects,
adjectival syntax and adjectival semantics coincide in the expected way: in fact,
the context shows unmistakably that an external reading is required, roughly
equivalent to d'autres autels, a reading that would be lost in (22b):
(22) a. Aucun autre dieu n'a inspiré à ses adorateurs le mépris et la

haine de ceux qui prient à de différents autels. (Yourcenar)
"No other god has inspired his followers to despise and hate
those who pray at different/other altars (than themselves)."
b. "No other god has inspired his followers to despise and hate
those who pray at different/various altars. "
2. The semantics o f A-différent

2.1 Two different A -différent?
Relying on the lexical contrast between ander and verschieden in German,
Beck (2000) has recently argued for the existence of two different different, a
relational adjective (different-verschieden) and a comparison operator (differ ent
anders). Correspondingly, she proposes two distinct analyses for the sentence
internal readings of different. Whereas the quantified-antecedent readings in (11)
above would involve the comparison operator, the dependent plural readings in
(12) would involve the relational adjective plus a hidden reciprocal (from each
other). As for NP-internal readings, Beck assumes that they instantiate the
relational adjective plus a hidden reciprocal, whereas (if we are to follow the lead
of German), external readings would instantiate the comparison operator.
However, the meaning of the relational adjective and that of the comparison
operator are not actually disjoint in Beck's treatment. Although she does not
elaborate on this point, the fact is that the semantics of the relational adjective
different appears in the definition of the truth conditions of the comparison
operator. The meaning of differ ent-verschieden is sketched by Beck as in (23),
that of differ ent-anders as in (24): |
(23) [[different']] (a, b) = 1 iff (i) or (ii):

(i) a ≠ b
(ii) a and b belong to kinds a' and b' and a' ≠ b'
(24) [[different']] (χ, y) (R<e, <e, t») iff u [R(x) (u) & different'
(u,max(λv[R(y)(v)])]
The second clause (23ii) in the semantics of differ ent-v erschieden is

probably meant to capture the fact that different is apt to suggest qualitative
differences over and above non-identity of individuals. Thus, (25) is most
1
(24), which is meant to apply to examples such as Luise owns a different car than Otto could be
paraphrased as follows: "Luise and Otto may be said different when compared as to their ownership
of a car iff there is a car that Luise owns and that is not identical to the car that Otto owns." (Beck
2000: 112 (31a)).
naturally interpreted as asserting that the books Pierre reads are of a different
kind than those read by Paul (van Peteghem 2001):
(25) Pierre et Paul lisent des livres différents.

"Pierre and Paul read different books."
Now, whatever the identity conditions for kinds might be, neither identity of
kinds nor its negation can possibly be gradable relations. But as we have seen,
A-différent is a gradable predicate. This leads to the conclusion that its definition
should include a non-similarity clause, so that in (23) a clause of the form a ≈/≈ b
should be added.2 The stronger clauses on non-identity of kinds or non-similarity
are crucial to the understanding of why différents is excluded from some
contexts, and what the particular meaning effects it produces in other contexts
are. Indeed, in NP internal readings, différents does not make sense when these
clauses are excluded for semantic or for pragmatic reasons, as in (26a-b), where
non-identity of kinds or non-similarity are either impossible or irrelevant:
(26) a. # ll a parcouru cent kilomètres différents.

"He travelled a hundred different kilometres."
b. # Des milliers de sauterelles différentes ont envahi mon jardin.
"Thousands of different locusts have invaded my garden."
Furthermore, the presence of différents suggests non-identity of kinds or

non-similarity with regard to some property in contrasts such as:
(27) a. Dans l 'aquarium nagent des poissons.

"In the aquarium swim some fish."
b. Dans l'aquarium nagent des poissons différents.
"In the aquarium swim different fish."
(28) a. Pour la toiture, on a fait venir trois ouvriers.
"We called in three workers for the roof."
b. Pour la toiture, on a fait venir trois ouvriers différents.
"We called in three different workers for the roof."
2
We cannot go into the details of the definition of this relation here, but we assume that it should
eventually replace (23ii). Like ≈, it should take into account the number of properties shared by two
individuals. Similarity holds among individuals with regard to properties, and can thus obtain to
different degrees, according to the number of shared properties. Of course, non-similarity entails non-
identity of the individual objects, but so does Beck's condition on the non-identity of the kinds the
objects belong to. Note that the contrast between différent and autre in French is, among other
things, characterized by the fact that autre is not fully gradable and does not express non-similarity
(see also van Peteghem 2001).
Finally, contrasts such as those in (29a-b) show that différent as a predicate

is not simply a way of expressing non-identity of individual objects:
(29) a. L'Etoile du Matin est différente de l'Etoile du Soir [par

l'intensité de sa lumière].
"The Morning Star differs from the Evening Star in brightness."
b. L'Etoile du Matin n'est pas l'Etoile du Soir [*par l'intensité de
sa lumière].
"The Morning Star is not the Evening Star *in brightness."
2.2 A-différent as a symmetrical relational adjective

The distinction proposed by Beck (2000) between a relational adjective and
a comparison operator is certainly well anchored in a lexical contrast in German.
But the main motivation for the analysis of ander and of one of the uses of
different in English as a comparison operator stems from the fact that in contexts
with an explicit second argument, both exhibit properties that are similar to those
of comparatives: a second argument introduced by a complementizer and the
possibility of establishing an indirect comparison:
(30) a. Luise saw a taller man than Otto.

b. Luise sah einen größeren Mann als Otto.
(31) a. Luise saw a different man than Otto.
b. Luise sah einen anderen Mann als Otto.
However, the lexical item in French that unmistakably shows the syntax and
the semantics of a comparative is not différent, but autre (see van Peteghem
1997,2001):
(32) a. Jean a lu un autre livre que Pierre/ que hier.

John read a different book than Peter/ than yesterday.
b. *Jean a lu un livre différent de/que Pierre.
c. *Jean a lu un livre différent de/que hier.
It seems thus much less justified for French to derive the external and the
quantified antecedent readings from a comparative. French différent is a
relational adjective in all its uses, and the problem is that of understanding how
these are linked together, and how and why they differ from those of other
relational adjectives.
2.2.1 Singular différent. The most difficult cases involve singular différent. Both
différent and other symmetrical relational adjectives exhibit the external reading,
although autre seems to be clearly preferred to différent as far as external

readings are concerned:
(33) a. Jean a proposé une solution différente (=ander).

"John proposed a different solution."
b. [Pierre travaille sur le Swahili]. Jean connaît une langue
apparentée.
"[Peter works on Swahili]. John knows a related language."
But quantified antecedent readings pose a real problem. Un N différent is

fully acceptable in contexts such as (34-35), whereas other relational adjectives
are not.
(34) a. Chacun des assistants a travaillé sur une langue différente

(= ander). Ils pourront donc comparer leurs résultats.
"Every one of the collaborators has worked on a different
language. They will thus be able to compare their results."
b. 11 Chacun des assistants a travaillé sur une langue apparentée.
Ils pourront donc comparer leurs résultats.
"Every one of them has worked on a related language. They will
thus be able to compare their results."
(35) a. Mes étudiants ont travaillé chacun sur une langue différente
(= ander).
"My students have each worked on a different language."
b. ?? Mes étudiants ont travaillé chacun sur une langue apparentée.
"My students have each worked on a related language."
In fact, most speakers reject (34b, 35b) while accepting (34a, 35a). The
difficulty for constructing the quantified antecedent-interpretation is the same in
both cases: in order to obtain the arguments for the relational adjective, it is
necessary first to 'multiply' the singular un N via a referential dependency from
the quantifier, which involves processing the host NP and the adjective on two
distinct steps. For some reason, this seems easier to do with différent. As far as
we can see at present, the possibility of giving rise to quantified-antecedent
readings in the singular is a lexical property of différent. However, contrarily to
what is assumed by Beck, it is not simply a property it shares with comparatives,
since the latter do not easily give rise to internal interpretations:
(36) .a. * Chacun travaille sur une langue plus/ moins/ aussi difficile.
"Every one of them works on a more/ less difficult language/ as
difficult a language."
b. *Mes étudiants travaillent chacun sur une langue plus/ moins/

aussi difficile.
"My students work each on a more/ less difficult language/ as
difficult a language."
2.2.2 Plural différents. As for plural différents , it is necessary to distinguish des

N différents from NPs introduced by some other indefinite determiner. Genuinely
sentence-internal readings arise only with des. In Laca &Tasmowski (2001a), we
had labelled sentence-internal readings of des N différents 'dependent plural'
readings, relying on the fact that both the licensing environments and the
interpretations involved are closely parallel to those holding for the phenomenon
of dependent plurality as described by Kamp & Reyle (1993), Roberts (1987)
and particularly Bosveld-de Smet (1997). As is the case with the dependent
plural in (37), sentence-internal readings of des N différents require the presence
of a plurality in the sentential context (38a). This property distinguishes des N
différents from le même N (38b vs 38c):
(37) a. Monocycles have wheels.

b. #A monocycle has wheels.
(38) a. Les membres de la commission ont voté pour des candidats
différents.
"The members of the committee voted for different candidates."
b. #Toute la commission a voté pour des candidats différents.
"The whole committee voted for different candidates."
 Toute la commission a voté pour le même candidat.
"The whole committee voted for the same candidate."
The intervention of a singular between a dependent plural and the plurality

it depends upon blocks the dependent plural reading, and the same holds for the
sentence-internal reading of des N différents :
(39) a. The women bought cars with automatic transmissions.

b. #The women bought a car with automatic transmissions.
(40) a. Les deux témoins ont aperçu des voitures de couleurs différentes.
"The two witnesses saw cars of different colours."
b . #Les deux témoins ont aperçu une voiture de couleurs
différentes.
"The two witnesses saw a car of different colours."
c. Les deux témoins ont aperçu une voiture de la même couleur.
"The two witnesses saw a car of the same colour."
Both dependent plurals and sentence-internal readings of des N différents

can be licensed by a plural event, such as is the case with adverbials which
trigger event pluralities, like those in (41):
(41) a. ll a publié des livres chez Fayard et chez Elsevier.

"He has published books at Fayard and at Elsevier."
b. Des candidats différents ont été sélectionnés dans le premier et
dans le second tour.
"Different candidates were selected in the first and in the second
vote."
And, for some speakers at least, sentence-internal readings of des N

différents can, like dependent plurals, be licensed by adverbs of quantification:
(42) a. Jean porte souvent des cravates en soie.

"John often wears silk neckties."
b. Jean porte souvent des cravates différentes.
"John often wears different neckties."
The interpretive processes involved are analogous, if we assume, following

Bosveld-de Smet (1997), that dependent plural interpretations arise when we
map a partition of the licensing plurality onto some partition (or cover) of the
dependent plural expression. In the most often discussed cases, we have atomic
partitions on both sides and obtain thus a bijection at the level of atomic
individuals, but cases such as (43a-b) can be analysed in the same way:
(43) a. Doctors wear gloves

b. The demonstrators carried banners
In the case of the sentence-internal readings of des N différents, the

arguments of the relation denoted by différents are obtained via a cover of the
set denoted by the host NP. The cells of this cover are determined by a partition
of the licensing plurality.
(44) a. Jean et Marie ont reçu des cadeaux différents.

"John and Mary got different presents."
b. Les trois enfants ont reçu des cadeaux différents.
"The three children got different presents."
c. Les enfants ont tous reçu des cadeaux différents.
"The children all got different presents."
Beck (2000) labels such sentence-internal readings 'plural NP-dependent

readings', and analyses them as containing a hidden reciprocal (different from,
each other). She adopts a theory of reciprocity based on Heim, Lasnik & May
(1991), combined with an analysis of distribution that makes use of covers, as
proposed by Schwarzschild (1996). Thus, (44a) is analysed as follows:3
(44) a' X [presents (X) & **got (John & Mary, X) & ∀x [x < X & χ e
Cov → ∀y [y < X & y e Cov & χ ≠ y → different (χ, y)]]]
This has exactly the desired effects: the arguments of différents are obtained
via a cover for the set denoted by the NP hosting différents, and this cover can, in
turn, be obtained from a partition of another plurality in the sentence. Beck only
discusses cases in which coordinated NPs are responsible for the contextually
salient cover. The way in which we partition the licensor expression is in this
case fully determined by the way the plurality is mentioned, and we obtain the
following cover determining the arguments of différents:
(45) a. {the present or presents that John got, the present or presents that
Mary got}
In cases such as (44b) and (44c), we get a fully distributive (atomic)

partition of the licensor expression as basis for the cover, because les trois N and
the floating quantifier tous impose such an interpretation, and the corresponding
covers are:
(45) b.{the present or presents that child A got, the present or presents
that child  got, the present or presents that child  got}
 {the present or presents that child A got, the present or presents
that child  got, the present or presents that child  got, }
In other cases, the partition of the licensor expression upon which the cover
for the set denoted by the NP hosting différents depends, remains undetermined:4
3
Roughly "There is a set X of presents cumulatively (= **) received by John and Mary and they can
be said to be different if no element of a salient cover of the set is identical to the other elements of
the cover" (see Beck 2000: 122 (61b)).
4
The fact that plural definite generics are not good licensors for sentence-internal readings of des N
différents, as in ??.Les Indiens meurent de maladies différentes. "The Indians die of different
diseases.", can be related to the full indeterminacy of the possible partitions of a generic plural.
(46) a. Ces étudiants suivent des cours différents.

"These students take different courses."
b. {the course or courses followed by the relevant subplurality A,
the course or courses followed by the relevant subplurality B,....}
We take Beck's analysis to be essentially correct, though, as she

acknowledges herself, a solution has to be found for cases in which the licensor
expression is not an NP but involves pluralities of events. In such cases, the
cover for the set denoted by the host NP is dependent upon the partition of a
plural event.
Labeling sentence-internal readings of des N différents 'plural NP-
dependent readings', as proposed by Beck, or 'dependent-plural readings', as
proposed in Laca &Tasmowski (2001a) might seem a moot point. However, it is
difficult to see how to deal with the intuitions of some speakers who accept
examples like (47a) and even (47b) without appealing to the notion of dependent
plural:
(47) a. Jean porte souvent des cravates différentes. [a single necktie for
each situation possible]
b. Les enfants de Marie ont chacun des pères différents. [a single
father for each child possible]
Note, furthermore, that for speakers who accept such examples, différents
is not just a staple symmetrical relational adjective. Most dependent plural and
NP-internal readings of différents have parallels in the reciprocal interpretation
of other symmetrical relational adjectives, to wit:
(48) a. Jean et Marie ont travaillé sur des langues apparentées. Ils
pourront donc comparer leurs résultats. cf. 44a)
b. Ces trois étudiants ont travaillé sur des langues apparentées, (cf.
44b)
c. Mes étudiants ont tous travaillé sur des langues apparentées, (cf.
44c)
d. Mes étudiants ont travaillé sur des langues apparentées. Ils
pourront donc comparer leurs résultats, (cf. 46a)
But for the speakers who reject (34b/35b), the interpretations corresponding
to (47a-b) are excluded:
(49) a. Jean travaille souvent sur des langues apparentées, [at least two
languages for each situation are required]
b. Mes étudiants ont chacun travaillé sur des langues apparentées.

#Ils pourront donc comparer leurs résultats [at least two
languages for each student are required]
In other words, whereas 'normal' symmetrical relational adjectives seem to

require a pairing with the two relata for each and every event they are involved
in, différents is able to provide a relatum by itself.
2.3 NP- internal readings

Plural NP-dependent readings can only arise with des N différents, because
des does not of itself determine any sort of partition of the set/plurality denoted
by the host NP (see Bosveld-de Smet 1997 and examples (13) and (14) above). It
is as it were entirely transparent for any sort of cover constructed from the
outside. With other indefinite determiners, the only possible reading is the NP-
internal reading, in which the arguments of the relation result from an atomic
partition of the set/plurality denoted by the host NP:
(50) a. Jean et Marie ont reçu trois cadeaux différents.

"John and Mary got three different presents."
b. Les trois enfants ont reçu plusieurs cadeaux différents.
'The three children got several different presents."
Of course, if the cardinality of the NP hosting différents coincides with the

cardinality of a co-argument, this can give rise to bijective readings (as in 51a),
which are a well known effect of co-arguments of equal cardinality (51b) (Link
1991):
(51) a. Jean et Marie habitent dans deux maisons différentes.

"John and Mary live in two different houses."
b. J'ai présenté deux filles à deux garcons,
"I introduced two girls to two boys."
As remarked by Carlson (1987), Beck (2000) and Laca &Tasmowski

(2001a), the contribution of différents in such NP-internal readings is truth-
conditionally vacuous in its non-identity sense (non-identity of individuals being
entailed by the cardinal determiner). Hence the tendency to interpret it in terms
of qualitative differences (applying the stronger clause on non-identity of kinds
or non-similarity in the definition of différent in (23) above), or as a means of
emphasizing cardinality.
Non-identity of kinds or non-similarity is involved in examples (27b) and
(28b), repeated here for convenience:
(52) a. Dans l'aquarium nagent des poissons différents.

b. Pour la toiture, on a fait venir trois ouvriers différents.
We surmise that effects of event pluralization, such as those illustrated in

the following contrast, arise from the need to find some property with regard to
which the individuals of the set/plurality denoted by the host NP differ:
(53) a. Le suspect a été interrogé par deux policiers.

"The suspect was cross-examined by two policemen."
b. Le suspect a été interrogé par deux policiers différents.
"The suspect was cross-examined by two different policemen."
[two cross-examination situations as preferred reading]
The emphasis on cardinality is very clear in contexts like:
(54) Ces douze millions de partisans sont répartis sur plus de vingt pays
différents.
"Those twelve million partisans are scattered in more than twenty
different lands."
Emphasis on cardinality also accounts for a peculiar effect of différents in

cardinalized host-NPs, namely that of favoring narrow scope even in inverse
scope configurations, such as in:
(55) a. Deux policiers sont intervenus dans la plupart des

interrogatoires.
"Two policemen took part in most cross-examinations."
b. Deux policiers différents sont intervenus dans la plupart des
interrogatoires.
"Two different policemen took part in most cross-examinations."
Emphasizing either cardinality or qualitative differences would contribute to

dissociate topic and subject status, thus allowing for the subject-verb
composition required in inverse scope configurations (see Laca & Tasmowski
2001b and references therein). This can be seen once more in (56). (56a), where
an internal object has been frontalized without any apparent reason, is odd,
whereas (56b) is fully acceptable, precisely because différents implies there
exists a reason for topicalizing:
(56) a. #Des arguments ont été invoqués contre le projet'.

"Arguments have been put forward against the project:"
b. Des arguments fort différents ont été invoqués contre le projet'.
"Different arguments have been put forward against the project:"
Whenever a cover cannot be constructed from the partition of a plurality in

the sentence, des N différents cannot but have an NP-internal reading (or, of
course, an external reading). But in this case, emphasis on cardinality is
excluded, since des does not express any cardinality at all. Des N différents in the
NP-internal reading can only be interpreted in terms of qualitative differences,
i.e. as non-identity of kinds or non-similarity. Only if the context suggests such
differences will it be readily acceptable:
(57) a. La reconstruction scientifique peut se faire avec des matériaux

différents: formules mathématiques, images, modèles mécaniques
de laboratoire.
"Scientific reconstruction can be carried out with different
materials: mathematical formulae, images, mechanical lab
models."
b. Pierre a ? acheté des pommes différentes / 0K choisi des pommes
différentes.
"Peter has bought different apples / chosen different apples."
As the lexeme choisir, differently from the lexeme acheter, suggests by

itself some qualitative basis for the choice to be done, out of the blue, choisir des
χ différents is more expected than acheter des x différents.
3. The semantics of D-différents

Although closely parallel to NP-internal readings of des N différents, D-
différents cannot be conceived of as a relational adjective with a reciprocal
interpretation. If this were the case, D-différents would be compatible with
different partitions of the plurality denoted by the NP and thus give rise to
dependent plural readings, contrarily to what was shown in 1.2. above. In losing
its status as a predicate, D-différents also loses its relational status. But it
preserves some peculiarities of the adjective it is related to. Most importantly, it
requires the possibility of some qualitative difference among the instantiations of
the noun it introduces, which explains why it is impossible with measure nouns:
(58) a. *ll a parcouru différents kilomètres.

"He travelled different kilometres."
b. * ll a bu différents litres de bière.

"He drank different liters of beer."
Since it stems from a predicate of pluralities (a symmetrical relational

adjective in a reciprocal interpretation), D-différents requires, like plusieurs and
unlike certains, to be verified by a plurality (Corblin 2001):
(59) a. - Craignez-vous que certaines personnes fassent obstruction?

- Oui, Dupont.
"- Do you fear that some people will object? - Yes, Dupont."
b. - Craignez-vous que différentes /plusieurs personnes fassent
obstruction? - #Oui, Dupont.
"- Do you fear that different/several people will object? - #Yes,
Dupont."
D-différents is further characterized by distributivity. It resists collective

interpretations with mixed predicates, strongly favoring distributive readings.
Thus (60a) necessarily involves more than one gift, which is not the case for
(60b):
(60) a. Différents invités m'ont apporté un bouquet de fleurs.

"Different guests brought me a bunch of flowers."
b. Quelques invités m'ont apporté un bouquet de fleurs.
"Some guests brought me a bunch of flowers."
However, this property holds for some speakers also of plusieurs. More
crucial is the fact that D-différents does not allow for collective generic
interpretations, which are currently assumed to rely on generic or adverbial
quantification over group variables:
(61) a. #Différentes personnes décidées peuvent entraîner une foule.

"Different determined people can lead a crowd."
b. Plusieurs personnes décidées peuvent entraîner une foule.
"(A group of) several determined people can lead a crowd."
This is a property D-différents shares with certains (see Corblin 2001) and
with non-intersective (strong) determiners:
(62) a. #C'ertaines personnes décidées peuvent entraîner une foule, [only

distributive reading]
b. #Toutes les personnes décidées peuvent entraîner une foule.

[only distributive reading]
c. #La plupart des personnes décidées peuvent entraîner une foule.
[only distributive reading]
Note, furthermore, that des N différents allows for the collective generic
interpretations that are excluded for D-différents:
(63) a. Des propriétés différentes ont souvent une origine commune.

"Different properties often have a common origin."
b. Des commissions différentes peuvent faire échouer un plan bien
construit.
"Different committees may ruin a well conceived plan."
The question thus arises as to the origin of. the inherent distributivity that
apparently characterizes D-différents and prevents DPs of the type différents N
from providing a group variable for generic or adverbial quantification. In the
case of non-intersective determiners, it is normally assumed that non-
intersectivity goes hand in hand with quantifier status, which would exclude
group-variable representations. On the other hand, most intersective determiners,
in particular numerical adjectives, can function as cardinality predicates applying
to group variables. In their uses as cardinality predicates, they have the 'exactly
n' interpretation (Landman 2002), an interpretation that also arises in collective
generic contexts, and in collective interpretations in general:
(64) a. These three lions are dangerous,

b. Three apples cost $ 1.
 Three men carried the piano upstairs, [falsified in the collective
effort interpretation if there were actually four men].
Thus, D-différents is as exceptional as certains among indefinite

determiners. In the case of certains, the fact that expressions of the form certains
N do not provide a group variable has been linked by Corblin (1997, 2001) to the
assumption that certains is not actually intersective. But such an explanation
cannot hold for D-différents, which - unlike certains - does not carry any
implicature as to the existence of a non-null complement set (N-NnP). This
leads us to hypothesize that group variable interpretations are excluded for
expressions introduced by determiners that necessarily occupy a DET position
and cannot function as predicates inside an NP. Given the distribution of
numerical adjectives, which are never postnominal, we just cannot see the
position they occupy in an N projection. With différents, the position it occupies

(as a DET or as a predicate) is made visible by linear position and/or by the
presence or absence of a determiner. And we only get 'group variable' readings
with A-différents.
4. Final remarks
We have shown that French (unlike English or German) makes it possible to
distinguish two distinct items, A-différent and D-différents. A-différent is a
symmetrical relational predicate which differs from other symmetrical relational
predicates mainly in the possibility of obtaining its arguments via a dependency
from a distributive quantifier. This fact, however, does not substantiate Beck's
analysis in terms of a comparison operator, since (a) A-différent, unlike autre,
does not exhibit the properties of a comparative and (b) comparatives do not
behave in the same way as A-differ ent with quantified antecedents. This remains
the most puzzling fact about this item. On the other hand, Beck's proposal for
the analysis of NP-plural dependent readings via reciprocity effects related to
covers provides a neat way of accounting for sentence-internal readings of des N
différents. As for NP-internal readings, some of the meaning effects associated
with différents in cardinalized NPs can be linked to the truth-conditionally
vacuous character of différents in its non-identity-of-individuals sense in such
constellations. NP-internal readings of des N différents provide the link with D-
différents. Some properties of the latter (qualitative differences and cardinality
>1) are inherited from A-différent, but its distributive nature seems to depend
upon its status as a DET. A deeper analysis of the latter hypothesis must await
further clarification of the relation between syntax and semantics (Landman
2002).
References
Amiot, Dany, Walter De Mulder & Nelly Flaux, eds. 2001. Le syntagme nominal: syntaxe
et sémantique. Arras: Artois Presses Universitaires.
Beck, Sigrid. 2000. "The Semantics of Different: Comparison Operator and Relational
Adjective". Linguistics & Philosophy 23.101-139.
Bosveld-de Smet, Leonie. 1997. On Mass and Plural Quantification: The Case of French
des/du-NPs. Doctoral dissertation. Groningen.
Carlson, Greg N. 1987. "Same and Different: Some Consequences for Syntax and
Semantics". Linguistics & Philosophy 10.531-565.
Coene, Martine & Yves D'Hulst, eds. 2003. From NP to DP. Amsterdam: Benjamins.
Corblin, Francis. 1997. "Les indéfinis: variables et quantificateurs". Langue Française
116.8-32.
Corblin, Francis. 2001. "Où situer certains dans une typologie sémantique des groupes
nominaux?". Kleiber, Laca & Tasmowski 2001. 99-11.
Diesing, Molly. 1992. Indefinites. Cambridge, Mass.: MIT Press
Gondret, P. 1976. "Quelques, plusieurs, certains, divers: Etude sémantique". Le Français

Moderne 44.143-152.
Heim, Irene, Howard Lasnik & Robert May. 1991. "Reciprocity and Plurality". Linguistic
Inquiry 22.63-101
Kamp, Hans & Uwe Reyle. 1993. From Discourse to Logic. Dordrecht: Kluwer
Kleiber, Georges, Brenda Laca &Liliane Tasmowski, eds. 2001. Typologie des groupes
nominaux. Rennes: Presses Universitaires de Rennes.
Laca, Brenda & Liliane Tasmowski. 2001a. "Distributivité et interprétations dépendantes
des expressions d'identité". Kleiber, Laca & Tasmowski 2001. 143-166.
Laca, Brenda & Liliane Tasmowski. 2001b. "Faits de distributivité et classification des
déterminants en français". Amiot, De Mulder & Flaux 2001. 47-64.
Landman, Fred. 2002. "Predicate-Arguments Mismatches and the Adjectival Theory of
Indefinites". Coene & D'Hulst 2003. 212-237.
Link, Godehart. 1991. "Plural". Semantik/Semantics ed. by Arnim von Stechow & Dieter
Wunderlich, 418-440. Berlin: Mouton-de Gruyter..
Löbner, Sebastian. 1987. "Natural Language and Generalized Quantifier Theory".
Generalized Quantifiers, ed. by P. Gärdenfors, 181-201. Dordrecht: Reidel..
Moltmann, Frederieke. 1992. "Reciprocals and same/differ ent. Towards a Semantic
Analysis". Linguistics & Philosophy 15.411-462.
Partee, Barbara H. 1987. "Noun Phrase Interpretation and Type-Shifting Principles".
Studies in Discourse Representation. Theories and the Theory of Generalized
Quantifiers, ed. by J. Groenendijk et al., 115-143. Dordrecht: Foris..
Roberts, Craige. 1987. Modal Subordination, Anaphora and Distributivity.
PhD.Dissertation. Amherst, Mass.
Schwarzschild, Roger. 1996. Pluralities. Dordrecht: Kluwer.
Szabolcsi, Anna. 1997. "Strategies for Scope Taking". Ways of Scope Taking ed. by A,
Szabolcsi, 109-154. Dordrecht: Kluwer.
Van Peteghem, Marleen. 1997. "Sur un indéfini marginal: même exprimant l'identité".
Langue Française 116.61-80
Van Peteghem, Marleen. 2001. "Autre vs différent: du pareil au même?". Amiot, De
Mulder & Flaux 2001. 141-160.
ON THE NON-UNITARINESS OF
NP SUBJECT INVERSION
A COMPARISON OF FRENCH NP SUBJECT INVERSION IN
INTERROGATIVES AND TEMPORAL SUBORDINATES*
KAREN LAHOUSSE
F.W.O.-Vlaanderen/.U.Leuven/Paris 8
1. Introduction
The subject of this article is NP subject inversion in French (henceforth:
VS), a phenomenon which has received a lot of attention in the literature.1
However, although the analyses differ (cf. 4.1.), (almost) all authors focus on NP
subject inversion in wh-contexts like (1) and (tacitly) assume that their analysis
can be extended to the other syntactic contexts NP subject inversion occurs in,
i.e. (indirect) wh-interrogatives, exclamatives, relatives, clefts, PP
topicalizations, all kinds of adverbials, subjunctive complements, sentences with
a topicalized adverb and scene-setting sentences like Arrive Pierre "Peter
arrives".
(1) Quand partira ce garçon? (Kayne 1972)

when will leave this boy
"When will this boy leave?"
(2) a. Quand vint le tour d'Arnaud,... (Fr)2
when came the turn of Arnold,...
"When it was Arnold's turn, ..."
* Many thanks to B. Lamiroy, J. Rooryck, D. Vermandere, A. Zribi-Hertz and the audience of Going
Romance 2001 in Amsterdam, as well as two anonymous reviewers, for their remarks and
encouragement.
1
I use VS to indicate the verb-subject word order.
2
Almost all my examples are simplifications of attested examples. Literary examples from Frantext
are marked with Fr, journalistic examples from Le Monde 1998 with LM, and examples from the
Internet (mostly journalistic sites) with Ya. I am grateful to the Institute for Modern Languages in
Leuven (ILT) for allowing me to use their Le Monde 1998 corpus.
178 KAREN LAHOUSSE
b. Quand retombera la poussière du bombardement, ... (LM)

when will fall down the dust of the bombardment
"When the dust of the bombing will settle,..."
In this article, I will take a closer look at NP subject inversion in temporal

subordinates (2), a kind of inversion that has not been studied very often. I will
show that a number of syntactic phenomena indicate that the inversion
mechanism found in temporal clauses (VST) differs from the one that is
operative in interrogatives (VSI), and, hence, that an analysis based on inversion
in interrogatives cannot be applied to inversion in temporals. Consequently,
contrary to what has been implicitly assumed by many generativists, I will
conclude that NP subject inversion cannot be considered a unitary phenomenon/
In section 2, I will provide some descriptive evidence in support of my
claim that NP subject inversion is not a unitary phenomenon. More precisely,
inversion in temporal subordinates (VST) and in interrogatives (VSI) will be
shown to differ with respect to their flexibility, their complexity and their
pragmatic function. In sections 3 and 4, I will show that different mechanisms
are at work in SI. Specifically, on the basis of the position of the verb with
respect to aspectual adverbs and on the basis of extraction of quantitative en out
of postverbal subjects, I will argue that the verb in VST is in the head of a
functional projection lower than T°, while this is not the case for the verb in VSI.
This result immediately implies that Kayne and Pollock's (2001) analysis of VSI,
which is based on phrasal movement, cannot be applied to VST. Since my data
show that the differences between VST and VSI cannot be explained if VS is
considered to be a unitary mechanism, I will conclude that such is indeed not the
case.
2. NP subject inversion in temporal subordinates and in interrogatives:

some descriptive differences
In this section, I show that there are clear descriptive differences with
respect to the 'rule' of inversion in interrogatives (VSI) and in temporal
3
This proposal is not entirely new. Especially in non-generative frameworks (but see Kampers-
Manhe 1998 for a generative proposal), attention has been paid to the different properties of inversion
in different syntactic environments. For example, in an HPSG framework, Bonami, Godard and
Marandin (1998) clearly distinguish inversion in extraction contexts from inversion in "spatio-
temporally dependent clauses, instantiated in three contexts: time adverbials, subjunctive
complements, and sentences with a thetic interpretation in a narrative". Marandin (2001) calls the
latter two contexts "unaccusative inversion", but does not mention inversion in temporal
subordinates.
ON THE NON-UNITARINESS OF NP SUBJECT INVERSION 179
subordinates (VST).4 A first clear indication in favour of my claim that a

distinction has to be made between VSI and VST is the fact that the former
inversion type is obligatory (3a-b), while the latter type is optional (cf. 4a-b),
since inversion (VST) can alternate with the canonical subject-verb word order
(CW):
(3) a. VSI: Quand arrivera ton père?

when will arrive your father
"When will your father arrive?"
b. CW: * Quand ton père arrivera?
(4) a. VST: Quand entra Butler, le vieillard était assis près du feu
when entered Butler the old man was sitting by the fire
"When Butler entered the room, the old man was sitting by the
fireplace"
b. CW: Quand Butler entra, le vieillard était assis près du feu (LM)
Moreover, in temporal subordinates, NP subject inversion is the only possible

type of inversion (5), while in interrogatives (6), it alternates with 'complex'
inversion (CI):
(5) a. VST: Quand Butler entra,...

b. CI: * Quand Butler entra-t-il, ...
when Butler entered-he
"When Butler entered the room, ..."
(6) a. VSI: Quand arrivera ton père?
b. CI: Quand ton père arrivera-t-il? 5
when your father will arrive-he
"When will your father arrive?"
Another related argument concerns the pragmatic function of inversion. On the

one hand, VSI correlates with the shift from affirmative to interrogative, and
mainly has a "grammatical function" (Le Bidois 1952:1). VST, on the other
hand, does not affect the modality of the clause, given that both (4a) and (4b) are
affirmative, and it has an "expressive value" (Le Bidois 1952:1).
4
However, I do not claim that these descriptive differences mark the structural properties of VSI and
VST. In section 3 and 4, I give data and arguments relating to the different structural properties of
VSI and VST.
5
Note that, in interrogatives, native speakers often prefer 'complex' inversion (6b) or the
interrogative marker est-ce que to NP subject inversion (6a). The alternations seem to be due to the
register on the one hand and to the informational status of the elements of the sentence on the other
hand (cf. Le Goffic 1997 for more details).
180 KAREN LAHOUSSE
These descriptive differences between VSI and VST suggest that inversion
in interrogatives and inversion in temporal subordinates are distinct. In the
following sections, I will demonstrate, on the basis of arguments with respect to
the structural properties of NP subject inversion, that this is correct and that there
are (at least) two different inversion mechanisms: one that is operative in
temporals and another one that is at work in interrogatives.
3. Adverbs as indicators of the position of the verb

Since antisymmetry (Kayne 1994), postverbal subjects are generally
assumed not to be in a right-adjoined position, nor in a right-hand specifier.
Consequently, the verb-subject word order in NP subject inversion must be
analysed as involving (at least) some kind of leftward movement of the verb
across the subject.
3.1 The aspectual adverbs 'tout à coup, soudain, brusquement, peu à peu' as
tools to determine the verbal position
In Emonds (1978), and especially in Cinque (1999), adverbials are used as a
tool to analyse the target positions of verbal movement. More specifically,
Cinque (1999) claims that the relative position of different types of adverbs
indicates a hierarchy of functional projections: each class of adverbs is in the
specifier of a functional projection with which it shares semantic features and the
head position is the position where the verb moves through on its way to T:
(7) T-verb v [Asp1P adverb1 [Aspl° t"v [Asp2p adverb2 [Asp2° t'v [Vp tv ]]]]]
Cinque (1999:11), in his relative ordering of 'lower' French adverbs (8)
(8) généralement "generally" > (ne) pas "not" > déjà "already" > (ne)
plus "no longer" / encore "still" > jamais "never" / toujours "always"
> complètement "completely" I partiellement "partially" > tout
"everything" / rien "nothing"
does not include the adverbs soudain, tout à coup, brusquement ("suddenly") and
peu à peu ("gradually"). However, it can easily be shown that these adverbs
constitute a class and also fit in the relative ordering (8).6 First, as is generally
recognized in the literature (Cinque 1999, Steinitz 1969, Jackendoff 1972, Quirk
6
For reasons of brevity, I will not give examples of all the adverbs, but the reader can easily verify
that the four adverbs display exactly the same behaviour.
et al. 1985), two adverbs of one class cannot co-occur in a clause. This also
appears to be true for the four adverbs that are at stake:7
(9) a. * Tout à coup, les gens arrivent peu à peuadv

suddenly the people arrive gradually
"Suddenly the people gradually arrive"
b. * Jean surgit tout à coupadv soudainadv
John appears all of a sudden suddenly
"John all of a sudden suddenly appears"
 * Les enfants sont soudainadv peu àpeuadv apparus
the children are all of a sudden gradually appeared
"The kids have all of a sudden gradually appeared"
Second, these adverbs also share a semantic feature: they all indicate the
boundedness of the verb as they do not combine with states (10a) or activities
(10b) but are allowed with achievements (10c) and accomplishments (10d):
(10) a. ??? Les télévisions existent soudain/tout à coup/peu à peu

the televisions exist suddenly/suddenly/gradually
"Televisions exist suddenly/suddenly/gradually"
b. ??? Jean dort soudain/tout à coup/peu à peu 8
John sleeps suddenly/suddenly/gradually
"John sleeps suddenly/suddenly/gradually"
c. Les enfants arrivent soudain/tout à coup/peu à peu
the children arrive suddenly/suddenly/gradually
"The children suddenly/suddenly/gradually arrive"
d. Jean mange peu à peu les bonbons
John eats gradually the chocolates
"John eats gradually the chocolates"
Third, they occupy the same linear position with respect to other adverbs.
Indeed, these adverbs occur to the right of the negative adverb pas "not" (11-12):
7
Besides their aspectual use, these four adverbs can also be used as adverbs of manner. As predicted,
when used as adverbs of manner, they also have another position in the sentence. Twill not elaborate
on this.
8
Note that this example is correct if it has the meaning of "fall asleep". However, in this case, the
verb does no longer denote an activity.
182 KAREN LAHOUSSE
(11) a. Jene me suis pas tout à coup senti plus libre ... (Ya)
I NEG myself have not suddenly felt more free
"I did not suddenly feel freer."
b. * Je ne me suis tout à coup pas senti plus libre
(12) a. le légal ne devient paspeu à peu le critère du moral(Ya)
the legal NEG becomes not gradually the criterion of the moral
"legality does not gradually become the criterion of morality"
b. * le légal ne devient peu à peu pas...
and to the left of the negative adverb plus "no longer" (13-14):
(13) a. comme s'il ne comprenait soudain plus ce qui se passe (Ya)

as if he not understood suddenly more what happens
"as though he didn't no longer understand what is happening"
b. * comme s'il ne comprenait plus soudain ce qui se passe
(14) a. Les hirondelles ne sont peu à peu plus venues (Ya)
the swallos not have gradually more come
"The swallows gradually no longer came"
b. * Les hirondelles ne sont plus peu à peu venues
Furthermore, with respect to déjà "already", these four adverbs occur to its right,
as the contrasts (15) and (16) show:
(15) a. il a déjà soudain abandonné ses études, et...

he has already suddenly abandoned his study and
"he has already suddenly dropped out of college, and..."
b. ?* il a soudain déjà abandonné ses études
(16) a. elle s'était déjà peu à peu accoutumée à des doses fortes (Ya)
she got already gradually accustomed to ART doses strong
"she had already gradually gotten used to strong doses"
b. ?* elle s'était peu à peu déjà accoutumée à des doses fortes
The examples (11-16) show that the adverbs soudain, tout à coup, brusquement
and peu à peu occupy the same linear position located between déjà "already" to
the left and plus "no longer" to the right. Consequently, (the relevant part of)
Cinque's hierarchy (8) can be extended in the following way:
(17) pas "not" > déjà "already" > soudain, tout à coup, brusquement
"suddenly", peu à peu "gradually" > plus "no longer"
Note however that, since Cinque (1999:95) has established that encore "still"
follows immediately plus "no longer", this also predicts, by transitivity, that my
adverbs should precede encore "still". Hence, the grammaticality of the order
encore > peu à peu in (18) is unexpected:
(18) certains d'entre eux revenaient encore peu à peu à Budeya (Ya)
some of them came back still gradually to Budeya
"some of them gradually returned again to Budeya"
First, it is rather surprising that these adverbs co-occur at all, since the
'continuative' aspect (Cinque 1999:95) of encore "still" is expected to be
incompatible with the bounded aspect (cf. supra) of peu à peu "gradually".
However, on taking a closer look at the example, it becomes clear that encore in
this sentence is not the continuative adverb "still", but rather the repetitive
adverb "again", as the gloss indicates.9 Since repetitive adverbs can occur,
following Cinque (1999:106), in a position higher than that of déjà "already",
which is also higher than the four adverbs that are at stake here, the unexpected
order in (18) is accounted for.10 In conclusion, I have demonstrated that the
aspectual adverbs soudain, tout à coup, peu à peu and brusquement constitute a
class of adverbs and that their precise position is between déjà "already" and
plus "no longer". Along the lines proposed by Cinque (1999), these adverbs can
thus be used as a tool to identify the position of the verb in temporal subordinates
and in interrogatives (3.2).
3.2 The position of the νerb

3.2.1 The position of the verb in canonical word order (CW). In Cinque's
hierarchy, lower aspectual adverbs are in a position between T° and VP (see (7)).
This predicts that, in canonical word order (CW), the aspectual adverbs soudain,
9
This is particularly clear in examples like (i): given that the continuative adverb encore "still"
cannot be combined with perfective tenses, this adverb, when it occurs in sentences with perfective
tenses like (i), has the meaning of "again":
(i) ... par crainte d'être rappelé encore tout à coup par quelque ordre inattendu (Ya)
for fear of being called back still suddenly by some order unexpected
"for fear of being suddenly called back again by some unexpected order".
10
According to Cinque (1999:95) repetitive adverbs can also occur in a very low position (lower than
"no longer"), which also predicts that they are able to occur in a position lower than soudain, tout à
coup, brusquement and peu à peu. This is borne out, given the grammaticality of examples like (i)
and (ii) in French. As the glosses indicate, the meaning of encore in these examples is "again". Note
that ancora in Italian has the same behaviour (cf. Cinque 1999:32).
( i) ll a tout à coup encore tapé à la porte. "He has suddenly knocked at the door again"
(ii) Les patients ont peu à peu encore appris à parler. "The patients have gradually learned to
speak again"
184 KAREN LAHOUSSE
tout à coup, brusquement and peu à peu are also situated, with simple tenses,
between the main verb and its complements, and, with complex tenses, between
the auxiliary and the participle.11 This prediction is borne out, as the examples
(19-22)respectively show:12
(19) a. *? Jean tout à coupadv répondverh à la question.

John suddenly answers to the question
"John suddenly answers the question"
b. Jean répondverb tout à coupadv à la question.
(20) a. *? Les gens peu àpeuadv se rendentverh compte des horreurs...
the people suddenly become aware of the horrors
"The people become gradually aware of the atrocities"
b. Les gens se rendentverbpeu àpeuadv compte des horreurs...
(21 ) a. l 'arrière m 'aaux soudainadv échappépart (Ya)
the back me has suddenly escaped
"I suddenly lost control over the back"
b. *? I 'arrière soudainadv m 'aaux échappépart
(22) a. le AC 40 aaux peu àpeuadv réduitpart ses gains (Ya)
the CAC 40 has gradually reduced its gains
"the CAC 40 has gradually reduced its gains"
b. *? le CAC 40peu àpeuadv aaux réduitpart ses gains
In conclusion, with canonical word order (CW), the position of the verb with
respect to soudain, tout à coup, brusquement and  à peu is the following:
(23) inflected VCw > soudain, tout à coup, brusquement, peu à peu
> participle
3.2.2 The position of the verb in inversion in temporals (VST). Let us now try to
determine the exact position of the main verb in VS in temporal subordinates. In
my corpus research, I found that the main verb in VST stays in a position lower
than soudain, tout à coup, brusquement and peu à peu:
11
In French clauses with canonical word order, both main verbs and auxiliaries are generally taken to
be in T°.
12
Note that I do not consider "parenthetical uses" of adverbs, which (cf. Cinque 1999:30) can make
unacceptable sequences grammatical. Cf. the contrast between the grammatical (i) and (ii) and the
ungrammatical (19a) and (20a) respectively:
(i) Jean, tout à coup, répond à la question.
(ii) Les gens, peu à peu, se rendent compte des horreurs...
(24) a. VST: Lorsque tout à coupadv surgitverh l'image d'une

when suddenly emerges the picture of a
représentation théâtrale (Fr)
representation theatrical
"When suddenly the picture of a theatrical representation
emerges"
b. VST: Lorsque soudainadv survientverh le drame, (Fr)
when suddenly occurs the tragedy
"When suddenly tragedy occurs"
c. VST: quand peu à peuadv vinrentverh s'en ajouter d'autres (Fr)
when suddenly came of them add others
"When others gradually added up to the rest"
The contrast between (23) and (24) clearly indicates that the verb in VST (24)
surfaces in a position lower than that of the verb in W (23).
However, in order to conclude that the verb in VST also stays in the head
position marked by soudain, tout à coup, brusquement and peu à peu, the verb
should also follow déjà "already" (since it precedes the four adverbs under
consideration, cf. (17)) and precede encore "still" (since encore follows soudain,
tout à coup, brusquement and peu à peu). This prediction is borne out, as the
examples (25-26) respectively show:13
13
Given that the negative adverbs (ne) pas "not" and (ne) plus "no longer" are used by Cinque (1999)
to determine the relative position of adverbs with respect to each other (see (8)), it is tempting to use
these adverbs also to determine the verbal position. And at first sight, the results seem to be
promising since verbs in VST cannot combine with (ne) pas (i) whereas they can combine with (ne)
plus (ii):
(i) * Quand n 'arrivaient pas les linguistes
when NEG arrived not the linguists
"When the linguists did not arrive"
(ii) Quand ne retentirent plus les cliquetis de vaisselle, un silence se fit.
when NEG resounded no longer the clashing of the crockery a silence started
"When the clashing of the crockery no longer resounded, there was a silence."
This is exactly what would be predicted if the verb in VST was in the head position following the
adverbs I am considering (since these adverbs are in a position between (ne) pas and (ne) plus, cf.
(17)). However, even if (some) examples with (ne) plus are grammatical, they cannot be used to test
the position of the inflected verb in VST, since plus always has to follow the inflected verb.
Moreover, it can be shown that there are semantic effects that could explain the constraints on
negation in temporal subordinates in general (cf. Le Draoulec 1995), and in VST more specifically.
Unfortunately, I cannot go into this, since this is beyond the scope of this article.
186 KAREN LAHOUSSE
(25) a. CW: ?? quand ;l'èclosion de l'art roman déjàadv s'apprêteverb14

"when the expansion of Romanesque art is already in the
making"
b. VST: quand déjàadv s 'apprêteverb l'eclosión de l 'art roman (Fr)
(26) a. VST: Quand lui ρarvenaitverb encoreadv l'écho de son cri (Fr)
when him reached still the echo of her cry
"When the echo of her cry still reached him"
b. VST: * Quand encoreadv lui parvenaitverh l'écho de son cri
I conclude from these observations that the main verb in inversion in temporal
subordinates is in the head of a functional projection with the adverbs soudain,
tout à coup, peu à peu, and brusquement in its specifier:
(27) T°- inflected verbCw UspP soudain,tout à coup,brusquement,peu à peu

[Asp° inflected verbVST [VP tverb ]]]
Now, since it is generally assumed (Pollock 1989) that French finite verbs and
auxiliaries are in the same position, auxiliaries in VST should also be in the head
of a ftmctional projection with the adverbs soudain, tout à coup, peu à peu, and
brusquement in its specifier. This turns out to be true:
(28) a. VST: Quand soudainadv aaux explosé cette bombe

when suddenly has exploded this bomb
"When suddenly this bomb exploded"
b. VST: * Quand aaux soudainadv explosé cette bombe
c. CW: Quand cette bombe aaux soudainac/v explosé
(29) a. VST: Quand peu à peuadv sontaux arrivés les clients
b. VST: * Quand sontaux peu à peuadv arrivés les clients
d. CW: Quand les clients sontaux peu apeuat¡v arrivés
when the clients have gradually arrived
"When the clients gradually arrived"
Native speakers perceive a sharp contrast between the (a) examples (with the
auxiliary in a position lower than soudain, tout à coup, brusquement and peu à
peu) and the (b) examples (with the auxiliary in a position higher than these
adverbs). This shows that the auxiliary in VST, just as the finite verb in VST,
does not move to the left of soudain, tout à coup, brusquementamdpeu à peu, as
14
Note that this example is grammatical with the 'parenthetical use' of déjà "already".
the auxiliary does in CW (cf. 21-22). Hence, the auxiliary in VST stays in a
head position that is lower than the position (T°) of the auxiliary in CW.
My conclusion is that both main and auxiliary verbs in VST stay in an
aspectual head position lower than T° with soudain, tout à coup, brusquement
and peu à peu in the specifier position:
(30) inflected V Cw > soudain, tout à coup, brusquement, peu à peu

> inflected Vvs
Remember that the goal of this article is to show that inversion in temporal
clauses (VST) and inversion in interrogatives (VSI) are fundamentally different.
If this is true, verbs in VSI are predicted not to occur in the same position as
verbs in VST. As the contrast (31) shows, the auxiliary in VSI has to be in a
position to the left of the adverb soudain:
(31) (Two soldiers are talking about an explosion)

a. VSI: ... mais quand aux soudainadv explosé cette bombe
but when has suddenly exploded this bomb
dont tu me parles?
of which you me speak
"But when has this bomb you talk about suddenly exploded?"
b. VSI: *... mais quand soudainadvaaux explosé cette bombe dont tu
me parles?
The ungrammaticality of (31b) sharply contrasts with the grammaticality of

(28a) and (29a) and shows that the auxiliary in French VSI is not in the same
position as the auxiliary in VST, and must surface in a position higher than the
adverbs soudain, tout à coup, brusquement and peu à peu.16
15
Note that the position of the auxiliary with respect to the adverb encore "still" cannot be tested,
since this adverb only occurs with non-perfective verbs.
16
An anonymous reviewer mentions ungrammatical examples such as *Quand cette terrible bombe
a-t-eUe soudain explosé? "When has this terrible bomb suddenly exploded?" *Cette terrible bombe
a-t-elle soudain explosé? "Has this terrible bomb suddenly exploded?"*Quand le CAC a-t-il peu à
peu diminué? "When has the CAC gradually diminished?" *Le CAC a-t-il peu à peu diminué? "Has
the CAC gradually diminished?" which allegedly show that adverbs such as soudain and peu à peu
are not appropriate in any interrogative context, and, hence, that there is a semantic constraint
preventing the use of these adverbs in interrogative contexts. However, it should be noted on the one
hand that these examples all instantiate 'complex inversion', a type of inversion I am not dealing with
in this paper. On the other hand, even if such a semantic constraint exists, it clearly does not hold for
the kind of inversion I am dealing with, given the grammaticality of my example (31a), which shows
that soudain can occur in VS in an interrogative context.
188 KAREN LAHOUSSE
3.2.3 Interim conclusion. In keeping with the line of reasoning of Cinque (1999),
I have shown that the verb in inversion in interrogatives (VSI) surfaces in a
position higher than the one where the auxiliary in inversion in temporal
subordinates (VST) surfaces, since the verbs in VSI and VST are to the left and
right respectively of the position of the aspectual adverbs soudain, tout à coup,
brusquement, md peu à peu. This immediately implies that different mechanisms
are at work in the two types of constructions, and that these data cannot be
accounted for if VS is taken to be a uniform mechanism.
Moreover, the verb in VST has been shown to occupy the head position
immediately to the right of the specifier position in which soudain, tout à coup,
brusquement, and peu à peu are hosted (see (27)). This entails that the verb must
have undergone head movement, rather than phrasal movement. Consequently,
an account of VS based on phrasal movement cannot easily be extended to VST.
I also demonstrated that the verb in inversion in interrogatives is in a position
higher than soudain, tout à coup, brusquement, and peu à peu. However, this
does not automatically entail that the verb in VSI is also in T°: it could be that it
has undergone head-movement to T°, but it could also have been moved
subsequently (remnant movement of the whole IP) to a still higher position. As a
consequence, while the data for VSI are compatible with an analysis based on
phrasal movement, it does not explain the word order distribution found in VST.
In the following section, I will show that extraction of quantitative en
indeed suggests that the verb in VST has undergone head movement, while the
verb in VSI has undergone phrasal movement. This is further evidence in favour
of my central claim that VST and VSI should be distinguished.
4. Extraction of quantitative en
4.1 X movement versus XP movement
In the generative literature, there are two analyses for NP subject inversion
in French (cf. Hulk and Pollock 2001 for more details): on the one hand,
according to De Wind (1995), Déprez (1998, 1999) and Valois and Dupuis
(1992), the DP subject stays in a low position (the specifier of vP or VP) and the
verb undergoes head movement to the head of one of the functional projections
of the split-Infl layer. On the other hand, Kayne and Pollock (2001) argue that
the DP subject surfaces in the left periphery and that the verbal phrase (IP with
the trace of the previously moved subject) undergoes subsequent phrasal
movement to a higher functional projection in the left periphery.
According to Kayne and Pollock (2001), in French NP subject inversion,
the DP subject starts out as the specifier of a silent subject clitic that heads a
larger DP (32a). The DP subject moves out of its canonical SpecIP position to the
specifier of a higher functional projection FP, leaving the silent subject clitic
(32b). Then, the whole IP is moved leftward past the subject (remnant
movement) to the specifier position of a higher functional projection GP (32c):
(32) a. [ipDPsubject SCL [rverb]]

→ movement of the DP subject to SpecFP
b. [FP DPsubject [IP SCL tsubj [I°verb]]]
→ remnant IP movement to SpecGP
 [GP[IP SCL tsubj [I° verb]] [FP DPsubject [IP tIP]]]
For Kayne and Pollock (2001), one of the major arguments in favour of IP
movement in inversion in wh-contexts is based on quantitative en-extraction.
4.2 Extraction of quantitative en: VSI versus VST

Kayne and Pollock (2001:112) observe that extraction of quantitative en out
of the postverbal subject in interrogatives is not possible, as shown by the
ungrammaticality of (3 3):
(33) VSI: * A quiwh encl ont téléphoné trois t?

to whomwh of-them have telephoned three ?
"To whom have telephoned three of them?"
Since quantitative en must be extracted to a position c-commanding its original

position the ungrammaticality of (33), following Kayne and Pollock, suggests
that en does not c-command the postverbal subject, and, since en is attached to
the verb, neither does the verb c-command the postverbal subject in VSI.17 This
can only be explained if the verb and the clitic en are contained in a phrasal
constituent that moves leftward past the subject.18 Thus, the extraction of
quantitative en shows that verbal movement in VSI is phrasal movement.
NP subject inversion in temporal subordinates (VST) does not behave in the
same way with respect to en-extraction: there is a stark contrast between
examples like (33-34a) with en extracted out of the postverbal subject in VSI,
and examples like (34b), with quantitative en extracted out of the postverbal
subject in VST:
17
According to Pollock (1986), this explains the subject-object contrast between (i) and (ii):
(i) Deux hommes sont arrivés. * Deux t encl sont arrivés.
two men have arrived two of-them have arrived
(ii) J'ai mangé deux pommes. J'encl ai mangé deux t.
I have eaten two apples I of-them have eaten two
18
This phrasal constituent has to be IP, since movement of intermediate projections like  is
generally not allowed. This also entails the consequence that the subject must be moved out of IP
prior to the IP movement (31b), because otherwise, the inverted word order could not be obtained.
190 KAREN LAHOUSSE
(34) (I haven't received any more letters since several weeks)

a. VSI: *Quand en arrivera une?
when EN will-arrive one
"When will there arrive one (letter)?"
b. VST: Quand en arrivera une, je serai folle de joie.19
when EN will-arrive one, I will be overjoyed
"When there will arrive one (letter), I will be overjoyed."
The grammaticality of (34b) indicates that in inversion in temporal subordinates,

en c-commands the subject, and, consequently, cannot be contained in a moved
phrasal constituent.20
In conclusion, evidence based on the cliticization of quantitative en shows
that the VS mechanism of Kayne and Pollock (2001) cannot account for both
VST and VSI, and that inversion in temporal subordinates and inversion in
interrogatives are to be analysed as involving different mechanisms: VST seems
to involve head movement, whereas VSI can be explained by phrasal movement.
5. General conclusion
In this article, I have argued that different mechanisms account for
inversion in interrogatives on the one hand and in temporal subordinates on the
other hand, since they differ with respect to (i) their flexibility, complexity and
pragmatic function, (ii) the position of the verb with respect to the aspectual
adverbs soudain, tout à coup, brusquement and peu à peu, and (iii) extraction of
quantitative en out of the postverbal subject. These data have also been shown to
indicate that the verbal movement in inversion in temporal subordinates is head
movement rather than phrasal movement and that, consequently, the analysis of
Kayne and Pollock (2001), based on inversion in interrogatives, cannot be
extended to inversion in temporal subordinates.
19
Whereas all my informants reject sentences like (33-34a), some of them accept sentences like
(34b). Moreover, the following example is attested in Frantext: - Mais enfin, dit Jacquemort, à bien y
réfléchir, il ne passe jamais une voiture sur cette route. Ou si peu. - Justement, dit Clémentine. Il en
passe si peu qu'on ne se méfie plus et quand par hasard en arrive une, c'est d'autant plus dangereux.
"But, said Jacquemort, now that I think of it, there is never a car on this road. Or very few - Exactly,
said Clémentine. There are so few of them that you do not pay attention anymore and when one
happens to pass, it is even more dangerous." (Vian 1953:169). The fact that en-extraction out of the
postverbal subject of a temporal subordinate, even though it is possible, is not more frequent (in
Frantext, I found just one example of this type on a total of approximately 280 examples with
inversion in temporal subordinates) could be attributed to the fact that postverbal subjects constitute a
weak island. Thanks to L. Rizzi for pointing this out.
20
The possibility of extracting quantitative en can also be linked to another difference between
inversion in temporals and in interrogatives. In fact, extraction often out of postverbal subjects yields
a strong focalization of the subject, and only the postverbal subject in temporals can be focalized, the
interrogative element in VSI being the focus of the question. This question is left for further research.
ON THE NON-UNITARTNESS OF NP SUBJECT INVERSION 191
My overall conclusion is, then, that, since the differences between NP

subject inversion in interrogatives and in temporal clauses cannot be accounted
for by one single mechanism, NP subject inversion in French is not a unitary
phenomenon.
References
Bonami, Olivier, Daniéle Godard & Jean-Marie Marandin 1999. "Constituency and word
order in French subject inversion". Constraints and Resources in Natural Language
Syntax and Semantics ed. by Gosse Bouma, Erhard W. Hinrichs, Geert-Jan W.
Kruijff& Richard T. Oehrle, 21-40. Stanford: CSLI.
Cinque, Guglielmo 1999. Adverbs and Functional Heads. A Cross-linguistic Perspective.
Oxford: Oxford University Press.
Déprez, Viviane 1988. "Stylistic inversion and verb movement". Proceedings of the Fifth
Eastern States Conference on Linguistics, 71-82. Columbus: Ohio State University.
1990. "Two ways of moving the verb in French". MIT Working Papers in
Linguistics, vol. 13: Papers on wh-movement ed. by Lisa Cheng & Hamida
Demirdache. Cambridge, Massachusetts: MIT Press.
De Wind, Maarten. 1995 Inversion in French. Groningen: Groningen Dissertations in
Linguistics.
Emonds, Joseph 1978. "The verbal complex V'-V in French". Linguistic Inquiry 9:2.151-
174.
Hulk, Aafke & Jean-Yves Pollock 2001. "Subject positions in Romance and the theory of
Universal Grammar". Hulk & Pollock 2001. 3-19.
Hulk, Aafke & Jean-Yves Pollock 2001. Subject Inversion in Romance and the Theory of
Universal Grammar. Oxford: Oxford University Press.
Jackendoff, Ray 1972. Semantic Interpretation in Generative Grammar. Cambridge: MIT
Press.
Kampers-Manhe, Brigitte 1998. "Je veux que parte Paul: a neglected construction".
Romance Linguistics, Theoretical Perspectives, ed. by Armin Schwegler, Bernard
Tranel & Myriam Uribe-Extebarria, 129-141. Amsterdam & Philadelphia:
Benjamins.
Kayne, Richard 1972. "Subject inversion in French interrogatives". Generative Studies in
Romance Languages ed. by Jean Casagrande & Bohda Saciuk, 70-126. Rowley:
Newbury House.
Kayne, Richard 1994. The Antisymmetry of Syntax. Cambridge: MIT Press.
Kayne, Richard & Jean-Yves Pollock 2001. "New thoughts on stylistic inversion. Hulk &
Pollock 2001. 107-161.
Le Bidois, Roger 1952. L'inversion du Sujet dans la Prose Contemporaine (1900-1950).
Paris: Artrey.
Le Draoulec, Anne 1995. "La négation dans les subordonnées temporelles". Cahiers
Chronos 2/3.257-275.
Le Goffic, Pierre 1997. "Forme et place du sujet dans l'interrogation partielle". La Place
du Sujet en Français Contemporain ed. by Catherine Fuchs, 15-42. Louvain-la-
Neuve: Duculot.
192 KAREN LAHOUSSE
Marandin, Jean-Marie 2001. "Unaccusative inversion in French". Romance Languages

and Linguistic Theory 1999 ed. by Yves D'Hulst, Johan Rooryck & Jan Schroten.
Amsterdam & Philadelphia: John Benjamins.
Pollock, Jean-Yves 1986. "Sur la syntaxe de EN et le paramètre du sujet nul". La
Grammaire Modulaire ed. by Mitsou Ronat & Daniel Couquaux. Paris: Editions de
Minuit.
Pollock, Jean-Yves 1989. "Verb movement, Universal Grammar and the structure of IP".
Linguistic Inquiry 20.365-424.
Quirk, Randolf, Sidney Greenbaum, Geoffrey Leech & Jan Svartvik 1985. A
Comprehensive Grammar of the English Language. London: Longman.
Steinitz, Renate 1969. Adverbial-Syntax. Berlin: Academie Verlag.
Valois, Daniel & Fernande Dupuis 1992. "On the status of (verbal) traces in French, the
case of stylistic inversion". Romance Languages and Modern Linguistic Theory ed.
by Paul Hirschbuhler & Konrad Koerner, 325-338. Amsterdam & Philadelphia:
John Benjamins.
Vian, Boris 1953. L'arrache-coeur. Paris: Hachette.
PAST PARTICIPLE AGREEMENT WITH
PRONOMINAL CLITICS AND THE AUXILIARY VERBS
IN ITALIAN AND FRENCH*
PAUL LAW
Freie Universität, Berlin
1. Introduction
In contrast with Spanish, Portuguese and Romanian, (Standard) Italian and
French have two remarkable properties in compound tenses: (i) past participle
agreement (PPA) with pronominal clitics and (ii) auxiliary alternation. PPA is
possible with (mostly direct object) pronominal clitics, but not with full DPs in
argument position.
(1) a. Paul les a repeintes. (French)

Paul them has repainted.FEM.PL
"Paul repainted them."
b. Paul a repeint/*repeintes les chaises.
Paul has repainted.UNMARKED/FEM.PL the chairs.FEM.PL
"Paul repainted the chairs."
The auxililary "be" (French être and Italian essere) occurs in passive and with
unaccusative verbs, where the participle agrees with the surface subject.
(2) a. Due camicie sono state lav ate/* lavata. (Italian)

two shirts.FEM.PL be.PL been.FEM.PL washed.FEM.PL/FEM.SG
"Two shirts have been washed."
b. Maria è/*ha arrivata/* arrivate.
Maria be/have.SG arrived.FEM.SG/FEM.PL
"Maria arrived."
* I am indebted to Nicoletta Puddu and Livio Gaeta for their unfailing help with the Italian data. I
also would like to thank two anonymous reviewers and Petra Sleeman for very helpful comments and
suggestions for improvement of the paper. All errors of fact and interpretation are my responsibility.
194 PAUL LAW
The auxiliary "have" (French avoir and Italian avere) appears with other verbs,
where PPA is possible with neither the unergative subject nor with the direct
object in argument position (in standard Italian and French) as in (3).1
(3) a. Maria ha/*è dormito/*dormita. (Italian)

Maria has/is slept.UNMARKED/FEM
"Maria slept."
b. Maria ha/*èletto/*letta il libro.
Maria has/is read.UNMARKED/FEM the book. MASC.SG
"Maria read the book."
The participle may agree with indirect object pronominal clitics, but only if they
are reflexive/reciprocal; in this case, the auxiliary be instead of have shows up.
(4) a. i ragazzi si sono/*hanno telefonati/*telefonate. (Italian)

the boys self are.PL/have.PL telephoned.MASC.PL/FEM.PL
"The boys called themselves/each other."
b. i ragazzi gli hanno/*sono telefonato/* telefonati.
the boys 3 PL have.PL/are.PL telephoned.UNMARKED/MASC.PL
"The boys called him/them."
The grammatical patterns in (l)-(4) obviously raise the question of whether

they are related to other facts in Italian and French. I argue that PPA is part of a
more general pattern of agreement including adnominal adjective agreement, and
is subject to a condition on the positioning of the argument with respect to the
syntactic projection of the agreeing predicate (section 2). I suggest to relate the
different patterns of PPA with clitic pronouns to the different morphological
expressions of formal features, and to bring the binding-theoretic difference
between reflexive and non-reflexive clitics to bear on their different patterns of
PPA (section 3). I claim that the lack of PPA in constructions with an overt
expletive is reduced to predicate agreement, and that the choice of the auxiliary
1
A particular form of a past participle may underlyingly have different formal features. The participle
in (3b) is formally invariant regardless of the number and gender features of the direct object, while
that in (i) varies according to the number and gender features of the pronominal clitic (lo ha and la ha
are phonologically [la], a phonological fact that I will ignore throughout):
(i) a. Maria lo ha lettol* letta. (Italian)
Maria it.MASC has read.MASC/FEM
"Maria read it."
b. Maria la ha letta/*letto.
Maria it.FEM has read.FEM/MASC
"Maria read it."
The same past participle letto is thus annotated as unmarked in (3b) and as masculine in (ia).
PAST PARTICIPLE AGREEMENT IN ITALIAN AND FRENCH 195
verb for the reflexive clitic is related to the same restriction on clitic pronouns
co-occurring with the auxiliary in its main verb use. I show how immediate
elimination of the checked fomial features bearing on the relative positioning of
the co-occurring clitic pronouns may explain why PPA is possible with
Accusative but not Dative reflexive clitics (section 4). In conclusion, I argue that
my analysis is fully compatible with the specificity property of PPA (section 5),
2. The structural condition on agreement

Kayne (1989:86) suggests to assimilate PPA to subject-finite verb agreement
in that the same structural relation holding for the subject and the finite verb
holds for the participle and the argument it agrees with as well. Thus, if the
structural relation between the subject and the agreeing finite verb is as in (5b),
(5) a. Jean est/*sont intelligent. (French)

Jean is/are intelligent
"Jean is intelligent."
b. DP AGR[ v p V... ]
then the structural relation between the argument and the participle agreeing with
it should similarly be sufficiently local. In (la), even though the pronominal
clitic apparently is not structurally local to the agreeing participle, it has in fact at
some point in the derivation moved through a position where it is structurally
local to the agreeing participle, as in (6).
(6) Paul lesja [ e J, AGR repeintes [ e ]¡ (French)
In (6) the participle does not directly agree with the pronominal clitic in its
surface position, but via the first [e] position governed by AGR, which Kayne
(1989:89) takes to be adjoined to an AGR projection (=IP).2 The structural
relation between the first [e] position and AGR in (6) is therefore very much like
that between DP and AGR in (5b), insofar as AGR governs them.
Kayne (1989) argues that PPA with pronominal clitics of the sort in (1) is
part of a general pattern of agreement, including that in constructions with wh-
movement (optional in colloquial French, and impossible in Italian). As shown in
(7) and (8), PPA is also possible when a direct object wh-phrase is extracted, but
crucially not when the overt expletive il is present.
2
The assumption that AGR governs the subject position does not seem to square well with earlier
work on the subject/object extraction asymmetry (Ross 1967, Chomsky 1981).
196 PAUL LAW
(7) a. Les chaises que Paul a repeintes. (French)

the chairs that Paul has repainted.FEM.PL
b. Je me demande combien de chaises Paul a repeintes.
I me ask how many of chairs Paul has repainted.FEM.PL
"I wonder how many chairs Paul repainted."
(8) a. Je me demande combien de chaises
I myself ask how many of chairs
il sera repeint(*es) cette année.
it will-be repainted this year
"I wonder how many chairs there will be repainted this year."
b. ?Je me demande combien de couleurs
I myself ask how many of colors
il a déteint(* es) (sur ce vêtement).
it has run on that clothing
"I wonder how much color it/there has run on that clothing."
 ... wh-phrase, il Vaux [ e ],- AGRi Vpp [ e ]l,-
If the wh-phrase in (8) were to agree with the participle, then it would have to
move through the first [e] position preceding AGR in (8c), an Α-bar position
(Kayne 1989:89). As uninterpretable elements like expletives cannot be present
at LF (Chomsky 1986a), the expletive il in (8) must be eliminated, just like the
expletive there in the English existential construction.
(9) a. S-structure: There is a man in the garden,

b. LF: A mani, is ti in the garden.
Now, if the relationship between the first [e] and the expletive il in (8) is on a par
with that between the expletive there and the associate DP a man in (9), then il
would be replaced by the first [e], just as there would be replaced by a man 3 But
such replacement would result in improper movement (Chomsky 1986b) as it
involves movement from an Α-bar position to an A-position.
The lack of PPA in the construction with an overt expletive thus seems to be
an empirical argument for the first [e] position in (6).
2.1. Some problems with the AGR-based analysis of past participle agreement
Kayne's improper movement analysis of the examples in (8) assimilating it
to the account of the existential construction in (9) is conceptually and
3
It is immaterial whether the associate literally replaces the expletive, i.e. appears in its place, or
simply adjoins to it (Chomsky 1986a). In any event, the same analysis should apply to both the
examples in (8) and (9).
empirically problematic, however. First, there is no independent motivation for

movement of traces. Second, if the position from which the extracted phrase
agrees with the participle is an Α-bar position, then it would vitiate the idea that
PPA is the same as subject-finite verb agreement, the subject position being
uncontroversially an A-position. A more serious problem is that PPA with an
unaccusative subject and the surface subject of a passive verb would be
incorrectly excluded.
(10) a. Maria, è t, AGR arrivata ti (Italian)

Maria is arrived.FEM.SG
"Maria arrived."
b. Molti esp erti, furono t, AGR invitati ti
many experts.MASC.PL were invited.MASC.PL
"Many experts were invited."
Movement to subject position in (10) should be excluded as a case of improper

movement, just as the movement of the first empty category to replace the
expletive in subject position in (8c).4 PPA with unaccusative verbs and in passive
thus constitutes clear evidence against the idea that the position from which the
argument agrees with the participle is an Α-bar position.
Third, the ungrammaticality of the example in (11) shows that there is not
obviously a position preceding AGR for the extracted argument to move through.
(11) *Paul a ces tables, AGR repeintes ti,. (French)

Paul has these tables.FEM.PL repainted.FEM.PL
"Paul has repainted these tables."
Kayne (1989:89) considers this example, arguing that since nothing would allow
one to say that the argument ces tables "these tables" in (11) is in an A-bar
position, it is ruled out by the condition in (12).
(12) If a Case-marked chain is headed by an A-position, then that A-

position must be assigned Case.
In (11), the argument ces tables is in an A-position, but is not assigned Case by
the auxiliary avoir "have" (cf. section 4.3), its Case being assigned by. the
participle (Kayne 1989:87-88).
In Kayne's analysis, then, there are two positions preceding AGR, one A-
¿
This problem does not arise in Kayne's (1993:19) analysis in which PPA with unaccusative verbs is
not the same as PPA with an extracted object, but is reduced to adjective agreement.
198 PAUL LAW
position, where the object ces tables "these tables" is in (11), and one A-bar
position, the first [e] position in (6), where an extracted argument agrees with the
participle, schematically as in (13), the first [e] being adjoined to the projection
of AGR, while the second being the Spec of AGR.
(13) ... DP ... [e] [[e] AGR...
However, the problem that arises is why the example in (11) cannot have the
structure in which the object ces tables "these tables" is in the first [e] position in
(13), an Α-bar position where a wh-phrase can move through and agree with the
participle, so that the condition in (12) does not come into play. There may be
independent reasons for why no object can occur between the auxiliary and the
participle in (11), e.g. a wh-phrase must end up in SpecCP, a clitic pronoun must
precede the auxiliary verb, but the ungrammaticality of the example in (11)
suggests that independent evidence for two positions preceding AGR is lacking.
Fourth, PPA with an unaccusative or passive verb in (10) is also a problem
for the view that the agreement relation holds of AGR and the second [e] in (13),
which later work takes to be where the Accusative argument checks its Case
(Chomsky 1991). It is precisely because of the lack of such a Case-theoretic
position that the unaccusative subject and the logical object of a passive verb
must move to SpecIP in order to be assigned (Chomsky 1981) or check its Case
(Chomsky 1995). It is thus doubtful that there is an AGR head in these cases.
Fifth, the AGR-based account for PPA neither covers adnominal adjective
agreement nor relates PPA to predicate agreement. I argue that all these types of
agreement are subject to the same structural condition (sections 2.2, 2.3 and 4.2).
2.2. Agreement in noun phrases

Across Romance, adnominal adjectives agree in number and gender with
the co-occurring nouns.
(14) a. les bonnes/* bons églises (French)

the nice.FEM.PL/nice.MASC.PL churches.FEM.PL
"the nice churches"
b. i grandi/*gran pensatori. (Italian)
the great.MASC.PL/great.SG thinkers.MASC.PL
"the great thinkers"
The issue here is whether the structural relation between the adjective and the
noun in (14) is amendable to that between the first [e] and AGR or to that
between the second [e] and AGR in (13).
Traditionally, (extensional) adnominal adjectives are assumed to be

adjoined syntactically to the NPs they modify, reflecting their semantic function
of restricting the property denoted by the nouns. The structural relation between
the adnominal adjective and the noun in (14) may therefore seem to resemble
that between the first [e] and AGR in (13). This is not the case, however. On the
one hand, the adnominal adjective, a predicate, is adjoined to (a projection of)
the noun in (14), while in (13), the argument is adjoined to the projection of the
predicate it agrees with. On the other hand, if AGR governs the first [e] position
in (13) (cf. footnote 2), then the relation between them is not the same as that
between the adnominal adjective and the NP. There is no reason to assume that
the NP is governed by the adjective, for it cannot be extracted out of the DP.
Various problems arise if the Spec-head relation is to hold of the adnominal
adjective and the noun the same way it holds of the second [e] and AGR in (13).
As the noun appears to the right of the adjective, it cannot be in the Spec of the
adjective, the Spec position being to the left of the head. One way to avoid this
problem is to posit some abstract structure in which the NP is in the Spec of the
AP, as in (15a), and the surface word-order is derived by moving the NP to the
right, as in (15b). If rightward movement is in principle not available (Kayne
1994), then the order A-NP can be derived by first moving the NP out of the AP,
and the AP is subsequently moved to the left of the NP, as in (15c) (cf. den
Besten and Webelhuth 1990).
(15) a. [DPi ... [AP[NP pensatori ] [ grandi ]]]

b. [DP i ... [AP ti [ grandi ] ... [NP pensatori ]i, ]]
 [DP i ... [AP ti [ grandi ]]j... [NP pensatori ] i ,... t, ]
In order to justify the derivations in (15), we not only need independent evidence
for the structure in (15a), but also must show that the various movements are not
just to derive the correct word-order, but have independent motivation.
The assumption that the NP is base-generated in the Spec of the AP as in
(15a) is not entirely implausible, for APs headed by adjectives like gran "great"
do appear in predicate position, with the subject originating in the Spec position
of the predicate (Koopman and Sportiche 1985, Kuroda 1988, Koopman and
Sportiche 1991) and moving to SpecIP for Case reasons. However, an example
like that in (16a) cannot have a structure like that in (15a), since adjectives like
altro "other" do not occur in predicate position.
(16) a. la altra/* altro ragazza. (Italian)

the other.FEM.SG/MASC.SG girl.FEM.SG
b. *La ragazza è altra.
The girl.FEM.SG is other.FEM.SG
200 PAUL LAW
It thus seems that the example in (16a) is base-generated as such, with the Spec-
head relation failing to hold for the adnominal adjective and the right-occurring
noun. The examples in (14) can then have the same structure, with the Spec-head
relation similarly failing to hold for the adnominal adjective and the noun.
The conceptual problem with the movement of the NP in (15) is that it is
not clear where the NP moves to, and why. If the NP in (15a) and (15b) is to
appear in a derived position, then it must be an A-position, for extraction of the
complement of the NP is possible (Cinque 1980).
(17) a. [una persona [di cui, (Italian)

a person of who
[apprezziamo [DPla grande ... [NP.generosità t¡]...]]]
appreciate the great generosity ]
"a person of whom we appreciate the great generosity."
b. Di chiį apprezziamo[DP[la grande ...[NPgenerosità tt ] ... ]]]
of whom appreciate the great generosity
"Who do we appreciate the great generosity of?"
But independent evidence for such an A-position is hard to come by.

The derivational problem in (15) can be avoided, one might argue, if the
example in (14b), for instance, in fact has a structure as that in (18a) where the
adjective is in the Spec of NP, just like that in (18b) for the example in (16a).
(18) a. [DP i [NP [APgrandi] [pensatori]]] (Italian)

the great.MASC.PL thinkers.MASC.PL
"the great thinkers"
b. [DP la [NP [AP altra ] [ ragazza ] ] ]
the other.FEM.SG girl.FEM.SG
"the other girl"
The problem with the structures in (18) is that the agreement relation is the
reverse of that for subject-finite verb agreement. The predicate in (18) is in the
Spec of the NP with which it agrees, while the subject is in the Spec position of
the agreeing finite verb. The most troublesome aspect of the structures in (18) is
that the predicates sit in the Spec position of their arguments. As far as I can tell,
such subject-predicate relation does not hold elsewhere.
A logically possible analysis for the example in (14b) would be as in (19a)-
(19b) where an AP-internal null subject PRO is controlled by a projection of N.
(19) a. [DPi [NP. [APPROi [ grandi ]][N'p ensat ori ]']]]

b. [DPi[NP. [APPROi [ grandi ]] [NMpensatori ]']]
c. [DPll pens atore]' sper ava di PRO' r¡solvere il problema

the thinker hoped to solve the problem
As PRO is controlled by a DP elsewhere, as in (19c), there is therefore no reason

to suppose that the structures in (19a)-(19b) are possible.
The examples in (14) then would have no plausible structure in which the
adjective stands in the Spec-head relation with the agreeing noun on a par with
the subject and the finite verb, and from which the correct word-order can be
derived. These facts thus show that agreement is possible, at least sometimes,
between elements that do not stand in the Spec-head relation. In light of these
various problems, we need to look for an alternative account for PPA.
2.3. Syntactic projection and agreement

An obvious difference between the examples in (1) is that in (lb) the object
is overtly realized in argument position while that in (la) is not. In the Italian
examples (2) and (4) the internal argument of the verb is not in object position
either. Particularly noteworthy is the fact that the same also holds for other cases
of agreement where no pronominal clitic occurs. PPA with wh-movement in (7),
PPA with unaccusative and passive subject in (10), and adnominal adjective
agreement in DPs in (14) are cases where the argument is not in the syntactic
projection of the predicate with which it agrees. Agreement is thus subject to the
general structural condition in (20) allowing a predicate to agree with an
argument just when the argument is not syntactically realized in its projection.
(20) Y and its argument XP may agree in the configuration

... XP ... [yp ... Y . . . ] . . . XP ... (order irrelevant)
PPA is therefore simply a special case of this general pattern of agreement that
also includes adnominal adjective agreement.
The example in (21a) may give the impression that the embedded subject is
in the projection of the agreeing adjective, unexpected if the condition in (20)
holds.
(21) a. Je considere [Mary compétente]
I consider Mary competent.FEM.SG
b. Je considère [(*toujours) Mary (toujours) compétente ]
I consider always Mary always competent.FEM.SG
"I consider Mary always competent."
This not the case, however. As shown in (21b), the embedded subject necessarily
202 PAUL LAW
occurs to left of an adverb modifying the adjective, hence outside its projection.5
Subject-finite verb agreement in (5) turns out to fall under the condition in
(20) for thematic verbs, for the subject is not in the VP-projection. If auxiliary
verbs head VPs (Zagona 1988) and move out of their projections, to Io (or AgrS),
then subject-finite auxiliary verb agreement too falls under the condition in (20),
even though the subject is not an argument of the auxiliary verb. This is not a
problem specific to subject-finite auxiliary verb agreement, but is arguably part
of a more general pattern. Pronominal clitics related to the arguments of the
thematic verb too occur on the auxiliary verb, not on the thematic verb.
In sum, apart from determiner-noun agreement (footnote 5), all other kinds
of agreement share the property that the argument lies outside the projection of
the agreeing predicate, as captured by the condition in (20). Intuitively, we can
take agreement on the predicate to be an indication that one of its arguments is
not realized in its projection, although it remains unclear why that should be.
3. Formal features ofpronominal clitics and past participle agreement

As we can see from the examples in (l)-(2), the past participial form of a
verb in Standard Italian, though less obvious in Standard French, formally
terminates in a vowel which often varies according to the number and gender
features of an (internal) argument of the verb. It is thus natural to take the
terminating vowel on the participle as the formal expression of the number and
gender features of the internal argument. Being uninterpretable on verbs, these
must be eliminated before the derivation proceeds to LF (Chomsky 1995) by
being checked off by an element with the matching number and gender features:
PPA is possible with Accusative clitics and Dative reflexive clitics in Italian
(cf. (la) and (4a)), but not with Dative non-reflexive clitics (cf. (4b)). I suggest to
relate the different patterns of PPA to the morphological properties of the clitics.
3
Number and gender agreement between determiners and nouns falls neither under the structural
condition in (20) nor under the AGR-based account. If Ds head noun phrases and take NP-
complements (Brame 1982, Szabolsci 1983, Abney 1987), then the NP is in the projection of D:
(i) a. [Dp [les/*le [NP chevaux]]] (French)
the. PL/the. SG horses. PL
b. [Dp [i/*il [wpensatori ]]] (Italian)
the.PL/the.SG thinkers.PL
But if N moves to D (Giorgi and Longobardi 1991) at LF when D is overt, then it is quite natural to
assume that the number and gender features of N and D must match.
Sleeman (2002) suggests that adnominal adjective agreement differs from predicate adjective
agreement in that it is not mediated by the Spec-head relation, but is the overt manifestation of the
theta-identification relation between the noun and the adjective, resulting in an extended chain
comprising the AP and the NP it modifies. Her analysis is entirely compatible with the structural
condition for agreement in (20), for in her syntactic representations for DPs with adnominal
adjectives the NPs lie outside the AP-projections of the adjectives, the APs being in the Spec of
functional projections of NP (Cinque 1994).
3.1. Pronominal clitics as morphological expressions of formal features

The forms of the Italian Accusative and Dative third person non-reflexive
pronominal clitics are given in (22).
(22) a. Accusative: lo [SG, MASC], la [SG, FEM]

li [PL, MASC], le [PL, FEM]
b. Dative: le [FEM], gli
The Accusative clitic clearly has morphological distinctions for the different
combinations of the number and gender features. As the Dative clitic gli may
have an antecedent with any combination of the number and gender features,
except that of feminine singular, it seems« natural to assume that it is inherently
unspecified for these features. The partial similarity of the Dative clitic le to gli
and to the Accusative clitic le can be captured if the Dative clitic le is taken to be
unspecified for the number feature (like gli), but specified as having the feminine
feature (like the Accusative le). Despite their unspecified features, I suggest that
le is interpreted in the semantics as singular and gli as any combination of
number and gender features other than that of feminine singular.
3.2. Feature-checking and participle agreement with non-reflexive clitics

Suppose pronominal clitics are morphological expressions of the number,
gender and Case features of arguments (cf. Borer 1984). These formal features,
like other uninterpretable features, must be checked and eliminated from the
syntactic representation before the derivation proceeds to LF (Chomsky 1995).
For pronominal clitics on thematic verbs, the checking of uninterpretable
formal features is fairly straightforward. Formal features on the participle can be
checked by the matching features of the clitic pronoun on the verb. If the
auxiliary verb forms some sort of complex predicate with the thematic verb it
takes as its complement (cf. Rizzi 1978, Grimshaw 1982, Burzio 1986, Zagona
1988), then formal features of the participle too can be checked by those of the
pronominal clitics occurring on the auxiliary verb.
PPA is possible with the Accusative clitic as the number and gender features
of the participle can be checked by those of the clitic. The Dative clitic lacks the
number feature, the uninterpretable number feature on the participle, if present,
can therefore not be checked. Lack of PPA with the Dative clitic now follows, as
6
It is crucial that abstract Case be distinguished from morphological case. A specific form realizing
different abstract Cases has no morphological case. Thus, the formally invariant clitic le "them/to
her" may be the morphological realization of the Accusative or Dative Case, but has no
morphological case. The clitic la "her" realizes the Accusative Case and also has morphological
accusative case, it being formally distinct from the clitics le and gli with the dative case.
204 PAUL LAW
the representation with the uninterpretable number feature on the participle

violates the Principle of Full Interpretation (FI) at LF (Chomsky 1986a).
3.3. Past participle agreement with Dative pronouns

It is no surprise that the past participle shows agreement with Accusative
reflexive clitics, since it generally does so with Accusative non-reflexive clitics.
(23) a. Gianni la ha lav ata/* lavato. (Italian)

Gianni it.FEM.SG.ACC has washed.FEM.SG/UNMARKED
"Gianni washed it."
b. Maria si è lavata/* lavato.
Maria self is washed.FEM.SG/UNMARKED
"Maria washed herself."
But it is quite intriguing that in Italian PPA is only possible with Dative reflexive
clitics, but not with Dative non-reflexive clitics or Dative non-clitic reflexives.
(24) a. I ragazzi si sono telefonati/*telefonato. (Italian)

the boys self are telephoned.PL/UNMARKED
"The boys telephoned each other/themselves."
b. i ragazzi gli hanno telefonato/*telefonati.
The boys them.DAT have telephoned.UNMARKED/PL
"The boys telephoned them."
c. i ragazzi hanno telefonato/*telefonan a se stessi.
the boys have telephoned.UNMARKED/MASC.PL to selves
"The boys telephoned each other/themselves."
If PPA is impossible with Dative non-reflexive clitics because (at least) the
number feature of the participle cannot be checked, as the Dative non-reflexive
clitics are unspecified for this feature, then the same should be true of the
formally invariant Dative reflexive clitic. In fact, by the same reasoning, we
should also expect PPA to be impossible with the (formally identical) Accusative
reflexive clitic. The examples in (23) and (24) show that this is not the case.
The difference between reflexive and non-reflexive clitics with respect to
PPA can be brought to bear on their binding-theoretic difference. I claim that the
formally invariant reflexive clitic is not lexically specified for the number and
gender features, i.e. these features are unvalued, but the unvalued features are
valued by virtue of the reflexive being bound by a (local) binding-theoretic
antecedent. That is, the inherently unvalued number and gender features of the
reflexive clitic si in (23b) and (24a) are valued as being the same as those of its
antecedent. Consequently, the agreement morphology on the participle can be

checked off by the matching number and gender features of the reflexive clitic.
By contrast, the Dative non-reflexive clitic in (24b) cannot be bound (in a
local domain), in accord with binding theory; therefore, its inherently unvalued
number feature (and the gender feature in the case of gli) cannot be valued. If
agreement morphology were to occur on the participle in (24b), then at least its
number feature cannot be checked off, the Dative non-reflexive clitic being
unvalued for the number feature. The unchecked number feature of the participle
will then remain in the derivation as it proceeds to LF, inducing a violation of FI.
PPA in (24c) is excluded as the structural condition in (20) is not satisfied, the
non-clitic reflexive being in the projection of the participle.
Essentially the same account can be given to explain the lack of PPA with
Dative non-reflexive clitics in French. The Dative non-reflexive pronouns lui
"him/her" and leur "them" have no morphological distinction for gender.
(25) Le garçon lui/leur a écrit/*écrite/* écrites. (French)

the boy 3SG/3PL has written.UNMARKED/FEM.SG/FEM.PL
"The boys wrote to him/her/them"
If the participle were to carry agreement morphology, then its gender feature
cannot be checked off, violating FI.
The pattern of PPA in Italian in (24) does not hold in French, most obvious
in (26) where the agreement morphology on the participle is audible/overt.
(26) Les filles se sont écrit/* écrites. (French)

the girls self are written.UNMARKED/FEM.PL
"The girls wrote each other/themselves."
The different patterns of PPA in Italian and French may be due to some
additional condition, however.7 One possibility is that PPA is only possible with
(morphological) non-dative in Italian, but only with (abstract) non-Dative in
French. Thus, even though the unspecified gender and number features of se are
valued as being the same as those of their antecedents, just like si, PPA is
nonetheless impossible since se is related to a Dative argument. A reason why
PPA is further conditioned in this way is conceivably related to an independent
morphological difference between Italian and French. While PPA morphology is
7
An anonymous reviewer points out that the different patterns of PPA may be due to Italian si having
fewer features than French se, which is in turn related to Italian si having a wider distribution than
French se. For instance, Italian si occurs in the impersonal construction where the logical object of
the verb appears in subject position, agreeing with the main verb. Nevertheless, it remains unclear
what feature present in se, but not in si, is responsible for their different distributions and for PPA.
206 PAUL LAW
mostly overt/audible in Italian, it is mostly covert/inaudible in French. That is,

PPA is conditioned by a morphological feature in Italian where PPA morphology
is overt, while it is conditioned by an abstract feature in French where PPA
morphology is (largely) covert.8
4. Past participle agreement, predicate agreement and reflexive clitics

The lack of PPA in the construction with the expletive il (8) is unexpected
from the perspective of (20), as the wh-phrase is clearly not in the VP-projection
of the verb taking it as its argument, just like that in (7). Yet, PPA is possible in
(7), but not in (8). I argue that PPA in (8) is impossible for independent reasons.
4.1. Lack of agreement in the construction with an overt expletive

Quite generally, an adjective or nominal in predicative position occurring
with the auxiliary be agrees with the surface subject.
(27) a. La porte est blanche/* blanc. (French)

the door is white.FEM.SG/MASC.SG
b. Marie est étudiante/* étudiant.
Marie is student.FEM.SG/MASC.SG
"Marie is a student."
We can now bring the agreement facts in (27) to bear on the lack of PPA in (8a).
In (8a), the surface subject is the expletive il, homophonous with the third
person masculine singular pronoun il. Thus, just like in (27), the predicate
complement of the auxiliary verb be in (8a) must agree with the surface subject,
whence the lack of PPA with the extracted object. The example in (8b) is related
to that in (28a) where the non-wh-counterpart of the phrase combien de couleurs
"how many colors" occurs after the verb.
(28) a. lla déteint(*es) beaucoup de couleurs sur ce vêtement.
8
The first and second person clitics, e.g. Italian mi "me", ti "you", etc, are not formally distinguished
for gender (or Case), i.e. unspecified for the gender feature, even though they generally can refer to
either a male or female person, depending on the natural gender of the speaker/hearer.
PPA with first/second person Accusative reflexive clitics is very much like PPA with the third
person clitic si; the gender feature on the participle is checked by the gender feature of the reflexives,
after being valued as the same as that of their binding-theoretic antecedents. The lack of PPA with
first/second person Dative non-reflexive clitics is the same as that for third person Dative non-
reflexive clitics. The problem, however, is PPA with first/second person Accusative non-reflexive
clitics. It is not obvious how the gender feature on the participle is checked. One may assume that
they are in fact specified for the gender feature without overt morphology, while the Dative non-
reflexive clitics are not, accounting for their difference with respect to PPA. But such an assumption
seems difficult to justify, since they are formally the same. I do not see a more satisfying solution, but
like to point out that nothing specific to my analysis of PPA leads to this very general problem.
it has run off a lot of colors on this cloth

"There ran off a lot of colors on this cloth."
b. Beaucoup de couleursi ont déteint(*es) t, sur ce vêtement.
a lot of colors have run off on this cloth
"A lot of colors ran off on this cloth." (French)
The structure for the example in (8b), after the expletive il has been replaced by
its associate combien de couleurs "how much color" (cf. footnote 3), is identical
to that of (28b), where PPA is impossible, for a predicate generally cannot agree
with the surface subject of the auxiliary have (cf. the examples in (2) and (3)).9
The lack of PPA in the construction with an overt expletive in (8) can thus
be reduced to predicate agreement. There is then no longer reason to take it as
evidence for an Α-bar position preceding AGR (cf. the first [e] in (13)).
4.2. Predicate agreement and reflexive clitics

Kayne (1993) suggests that the different surface forms of the auxiliary verb
are in fact manifestations of a unique underlying abstract auxiliary BE. BE takes
a DP complement headed by an abstract determiner/preposition, and is spelled
out phonetically as have if the head of its complement incorporates into it.
(29) SUi,- BE+D/P, [DP ti [ tj [vp V OB ]]]] (BE+D/P=have)
Otherwise, the abstract auxiliary BE would surface as be.

As we can see in (4a), the auxiliary verb is be when a reflexive clitic
appears. Kayne claims that the reflexive clitic adjoins to AgrS, activating it in
such a way as to allow it to move to D/P and turn SpecDP into an A-position.
(30) SU,- BE [DP ti [ reflexive+AgrSj+D/P [tj ... ti ...]]]
The argument DP in the VP now may move via SpecDP on the way to Spec BE,
rendering D/P-to-BE incorporation unnecessary; BE is then spelled out as be.
I cannot go into the details of Kayne's (1993) analysis here, but like to point
out two problems. Conceptually, there is apparently no independent motivation
for the assumption that AgrS may be activated in a way allowing AgrS to move
to D/P and turn SpecDP into an A-position. Empirically, and more
problematically, it incorrectly predicts that PPA is always possible with the
reflexive clitic. In fact, when a reflexive and a non-reflexive clitic co-occur, PPA
is possible with the reflexive clitic, but only if it is related to an Accusative, not
9
The lack of PPA in (28b) can be explained in the same way as that for (8b) in terms of improper
movement. PPA with unaccusative and passive verbs in (10) remains problematic, as discussed.
208 PAUL LAW
Dative, argument (the reflexive preceding another clitic is phonologically se).
(31) a. le si sono presentati/* presentata. (Italian)

her.DAT self ACC.MASC.PL are introduced.MASC.PL/FEM.SG
"They introduced themselves/each other to her."
b. se la sono presentata/* presentati.
self DAT.PL her.ACC are introduced.FEM.SG/MASC.PL
"They introduced her to themselves/each other."
I argue presently that the appearance of the auxiliary verb be rather than
have with the reflexive clitic is related to the same restriction on the occurrence
of the clitics on these same verbs in their main verb use, and that the patterns of
PPA in (31) are due to the checking and immediate elimination of formal features
from the syntactic representation and to the position of the reflexive clitic.
4.3. Pronominal clitics and the verbs have and be

Kayne (1989:88) takes the grammatical contrast in (32) and (33) to be
evidence that French être "be" assigns Case but avoir "have" does not.10
(32) a. Paul sera photographié par Marie. (French)

Paul will be photographed by Marie
"Paul will be photographed by Marie."
b. Paul le sera par Marie.
Paul it will be by Marie
(33) a. Paul a telephoné (à Marie).
Paul has telephoned to Marie
"Paul has telephoned Marie."
b. * Ρ aul l' a (à Marie).
Paul it-has to Marie
There is reason to believe that the grammatical contrast between (32b) and (33b)
has more to do with the predicate proform le than with Case, however.
The clitic pronoun le in (32b) is formally identical to the personal pronoun
le related to an Accusative argument, but the two differ in some respects. While
the personal pronoun formally agrees with its antecedent in number and gender,
the predicate proform le is formally invariant.
10
This view evidently does not relate the auxiliary have to the main verb have as in Kayne (1993).
(34) a. Marie sera étudiante. (French)

Marie will be student.FEM.SG
"Marie will be a student."
b. Marie le/* la sera.
Marie it.MASC.SG/it.FEM.SG will be
"Marie will."
(35) a. Marie et Sophie sont étudiantes.
Marie and Sophie are students
b. Marie et Sophie le/* les seront.
Marie and Sophie it.MASC.SG/it.PL will be
"Marie and Sophie will be."
(36) a. Marie sera grande et mince.
Marie will be tall and slim
b. Marie le/* les sera.
Marie it.MASC.SG/them will be
"Marie will be."
If DPs must be Case-marked (cf. Rouveret and Vergnaud's 1980 Case Filter) in
order to be Theta-marked (Aoun 1979 and Chomsky 1981), then non-DPs like
the predicate proform le do not need Case.11
It is noteworthy that the verb have in its main verb use in French and Italian
can occur with a non-reflexive, but not with a (non-argument) reflexive clitic.
(37) a. Pier o e Maria li hanno. (Italian)

Piero and Maria them.ACC have
"Maria have them."
b Maria/* si) ha un panino.
Maria self has a bread
"Maria has some bread."
 Maria (si) mangia un panino.
Maria self eat a bread
"Maria is eating some bread."
Although it remains unclear why that should be, it comes as no surprise that the
same verb used as an auxiliary shows the same co-occurrence restriction.12
11
The grammatical contrast between (32b) and (33b) may also be due to the predicate proform le
being property-denoting, which the auxiliary be may take as its complement, but the auxiliary have
may not (cf. * Marie a étudiante litt. "Marie has student").
12
It may be that the verb avere "to have" (in its main verb use) has the Case-related property that the
co-occurring clitic pronoun must have distinctive morphology for abstract Case. The reflexive clitic
therefore cannot co-occur with avere in (37b) and (38a) since it is formally invariant regardless of the
210 PAUL LAW
(38) a. (loro) si sono/*hanno telefonati. (Italian)

they self are/have telephoned.MASC.PL
"They telephoned each other/themselves."
b. (loro) gli hanno telefonato.
they them.DAT have telephoned.UNMARKED
"They telephoned them."
4.4. Immediate elimination o f l features

From the perspective of predicate agreement in passive, predicative
adjectives and predicative n o i n a l s (cf. (27)), it is understandable that the past
participle agrees with the syntactic subject when the auxiliary be appears. PPA in
(31b) is thus of particular interest, for the past participle does not agree with the
syntactic subject (or with the reflexive), even though the auxiliary here is also be.
Suppose formal features not only must be checked as early as possible (cf.
Pesetsky 1989), but must also be immediately eliminated from the syntactic
representation (Chomsky 1995). Then in (31a) where the reflexive is attached
first to the auxiliary, its number and gender features, valued the same as those of
its binding-theoretic antecedent, can check the matching features of the
participle. The Dative non-reflexive clitic poses no particular problem, as it
generally does not agree with the participle anyway. But in (31b), the Accusative
clitic is attached first to the auxiliary, checking the number and gender features
of the participle. As its checked formal features are immediately eliminated from
the syntactic representation, the participle would no longer have formal features
to check those of the reflexive clitic; PPA with the Dative reflexive thus fails.
5. Conclusion
In this paper, I claim that PPA is subject to a structural condition requiring
that the argument be outside the syntactic projection of the predicate with which
it agrees. The structural condition for agreement is necessary but not sufficient,
however, as PPA also bears on the co-occurring auxiliary verb, and the presence
of the formal features on the argument. The advantages of this view are that
agreement has a unified account encompassing PPA, subject-finite verb
agreement, adnominal adjective agreement and predicate agreement, and that no
functional category AGR is needed, in line with recent work (Chomsky 1995).
Obenauer (1992) argues that PPA correlates with the specificity of the
argument. As the condition in (20) is only necessary but not sufficient for
agreement, it is therefore not too surprising that PPA is not always possible, in
underlying abstract Case, while the accusative clitics and the dative case clitic gli "to him/them" can,
these being morphologically distinctive. The lack of such a Case-related property in the auxiliary
essere "to be", contra Belletti 1988 (cf. Law 1996), explains why the reflexive is possible in (38a).
particular, when the argument is not specific. Why that should be then requires
an independent explanation. If the specificity of the argument correlates with its
being outside of the projection of the predicate (Diesing 1992) (cf. Déprez 1998),
then the facts showing the specificity property of PPA fall under the structural
condition in (20) as much as under the AGR-based account.
The account I offer here is far from being complete. Many complex patterns
of agreement and auxiliary selection in other varieties of Romance discussed by
Kayne (1989, 1993) clearly lie beyond the confines of my analysis. Lack of
access to detailed descriptions and speakers of these varieties prevents me from
undertaking a systematic study comparing them with the standard varieties. I
hope to rectify this empirical limitation in a near future.
References
Abney, Steven 1987. The English Noun Phrase in Its Sentential Aspects. PhD dissertation.
MIT.
Aoun, Joseph 1979. "On Government, Case-Marking and Clitic Placement." Ms, MIT.
Belletti, Adriana 1988. "The Case of Unaccusatives." Linguistic Inquiry 19.1-34.
Besten, Hans den & Gerd Webelhuth 1990. "Stranding". Scrambling and Barriers ed. by
Günther Grewendorf & Wolfgang Sternefeld, 77-92. Amsterdam: John Benjamins.
Borer, Hagit 1984. Parameter Syntax. Dordrecht: Foris.
Brame, Michael 1982. "The Head-Selector Theory of Lexical Specification and the Non-
Existence of Coarse Category". Linguistic Analysis 8.321-25.
Burzio, Luigi 1986. Italian Syntax. Dordrecht: Reidel.
Chomsky, Noam 1981. Lectures on Government and Binding. Dordrecht: Foris.
Chomsky, Noam 1986a. Knowledge of Language: Its Nature, Origin and Use. New York:
Praeger.
Chomsky, Noam 1986b. Barriers. Cambridge, Mass.: MIT Press.
Chomsky, Noam 1991. "Some Notes on the Economy of Derivation and Representation".
Principles and Parameters in Comparative Grammar ed. by Robert Freidin. 417-
454. Cambridge, Mass.: MIT Press.
Chomsky, Noam 1995. The Minimalist Program. Cambridge, Mass.: MIT Press.
Cinque, Guglielmo 1980. "On Extraction from NP in Italian". Journal of Italian
Linguistics 5.47-99.
Cinque, Guglielmo 1994. "On the Evidence for Partial N-movement in the Romance DP".
Paths towards Universal Grammar. Studies in Honor of Richard S. Kayne ed. by
Gulielmo Cinque et al., 85-110. Washington D.C.: Georgetown University Press.
Déprez, Viviane 1998. "Semantic Effects of Agreement: The Case of French Participle
Agreement". Probus 10.1-66.
Diesing, Molly 1992. Indefinites. Cambridge, Mass.: MIT Press.
Giorgi, Alexandra & Giuseppe Longobardi 1991. The Syntax of Noun Phrases.
Grimshaw, Jane 1982. "On the Lexical Representation of Romance Reflexive Clitics".
The Mental Representation of Grammatical Relation ed. by Joan Bresnan, 87-147.
Cambridge, Mass.: MIT Press.
212 PAUL LAW
Kayne, Richard 1989. "Facets of Past Participle Agreement in Romance". Dialect

Variation and the Theory of Grammar, ed. by Paola Benincà, 85-103. Dordrecht:
Foris.
Kayne, Richard 1993. "Toward a Modular Theory of Auxiliary Selection". Studia
Linguistica 47.3-31.
Kayne, Richard 1994. The Antisymmetry of Syntax. Cambridge, Mass.: MIT Press.
Koopman, Hilda & Dominique Sportiche 1985. "Θ-Theory and Extraction". Abstract in
GLOW Newsletter, number 14, Department of Language and Literature. Tilburg
University.
Koopman, Hilda & Dominique Sportiche 1991. "The Position of Subjects". Lingua
85.211-258.
Kuroda, Shigeru. 1988 "Whether We Agree or Not: A Comparative Syntax of English and
Japanese". Lingvisticae Investigationes 21.1-46.
Law, Paul 1996. "Remarks on the Verb be and the Expletive there in English".
Linguistische Berichte 166.492-529.
Obenauer, Hans-Georg 1992. "L'interprétation des Structures wh et l'Accord du Participe
Passé". Structures de la phrase et théorie du liage ed. by Hans-Georg Obenauer &
Anne Zribi-Hertz, 169-193. Paris: Presses Universitaires de Vincennes.
Pesetsky, David 1989. "Language-particular Processes and the Earliness Principle". Ms,
MIT
Rizzi, Luigi 1978. "A Restructuring Rule in Italian Syntax". Recent Transformational
Studies in European Languages, ed. by Jay Keyser, 113-158. Cambridge, Mass.:
MIT Press.
Ross, John Robert 1967. Constraints on Variables in Syntax. PhD dissertation, MIT.
Rouveret, Alain & Jean-Roger Vergnaud 1980. "Specifying Reference to the Subject:
French Causative and Conditions on Representations". Linguistic Inquiry 11.97-
102.
Sleeman, Petra 2002. "Adjectival agreement within DP without feature movement".
Current Issues in Romance Languages, ed. by Teresa Satterfield, Christina Tortora,
Diana Cresti, 301-317. Amsterdam & Philadelphia: John Benjamins.
Szabolcsi, Anna. 1983 "The Possessor that Ran Away from Home". Linguistic Review
3.89-102.
Zagona, Karen. 1988. Verb Phrase Syntax: A Parametric Study of English and Spanish.
Dordrecht: Kluwer Academic Publishers.
DEFICIENT PRONOUNS AND LINGUISTIC CHANGE IN
PORTUGUESE AND SPANISH
ANA MARIA MARTINS

University of Lisbon
1. Introduction
Different authors addressing linguistic change in Romance have proposed
that clitic pronouns evolved from second position clitics (P2) to verbal clitics,
correspondingly changing their categorial status (from XP to Xo) and their
syntactic distribution - see Rivero (1986, 1991, 1997), Barbosa (1993, 1996),
Fontana (1993, 1997), Halpern & Fontana (1994), Halpern (1995), among others.
Old Romance clitics, however, differ from typical P2 clitics in two ways: a) Old
Romance clitics are not strictly second but may occur in third, fourth, or more
rightward positions; b) the possibility of breaking up the linear adjacency
between clitic and verb is severely restricted in Old Romance. With respect to the
former trait of Old Romance, it has been claimed that leftward sentential
adjuncts (or left dislocated phrases in the Specifier of a recursive Top position)
are freely allowed in certain Romance languages, such as Old Portuguese and
Old Spanish; these constituents would be irrelevant for the computation of the
'second position', being treated for this matter as sentence external.1 As for the
strongly preferred linear adjacency between clitic and verb (even in languages
like Old Portuguese and Old Spanish which do not strictly adhere to this
positional restriction), it has been proposed that Old Romance manifested verb
second (V2) properties; because both the clitic and the verb competed for the
second position in the sentence, they would tend to occur 'clustered'. This
situation would have favoured the ulterior reanalysis of the Old Romance P2
clitics as verbal clitics, (cf. Rivero 1986, 1991; Salvi 1990, 1991, 1993, 1997;
Kaiser 1992; Barbosa 1993, 1996; Fontana 1993, 1997; Benincà 1995; Ribeiro
1995a, 1995b; Moraes 1995; Galves 1997, 2001).
1
Cf. Benincà (1995:336): "an adverbial clause can either be generated as a constituent of the main
clause (or a SpecCP occupier) - in which case enclisis is impossible - or as an extra-sentential
complement [in SpecTopP] - in which case enclisis is obligatory".
214 ANA MARIA MARTINS
The type of analysis sketched above predicts that in Clitic Left Dislocation
constructions, if only the left dislocated constituent precedes the verb, like in (1)-
(2) below, clitics would obligatorily be enclitic on the verb in Old Romance (see
Salvi 1991, Benincà 1995) - the verb counting as the first sentence internal
constituent. This prediction however is contradicted by the empirical data, as
sentences (3)-(4) below show. Although exemplifying the same type of syntactic
configuration as sentences (l)-(2), sentences (3)-(4) display proclisis. All the
sentences - (1) to (4) - belong to the letters written from India to the king of
Portugal by Afonso de Albuquerque in the early sixteenth century.2
(1) à pessoa que vos tall dise ou espreveo,

to-the person that you-DAT such-thing told or wrote,
pergumtelhe vosalteza omde estava syman afomso
ask-him-DAT Your-Highness where was Simam Afomso
"Your majesty should ask the person that told you or wrote you such
thing where Syman Afomso was"
(Cartas de Afonso de Albuquerque. Pato 1884:99)
(2) As joyas que a vossalteza manda elrrey de siam,
the jewels that to Your-Highness sends the-king of Siam,
levaqs_ nuno vaz.
carries-them-ACC Nuno Vaz
"Nuno Vaz is taking you the jewels that the king of Siam offers you."
2
Afonso de Albuquerque, who was born in the second half of the fifteenth century, was the first
Portuguese viceroy of India. The pattern of clitic placement represented in his letters from India is the
typical pattern of clitic placement found in late Old Portuguese (that is, in the fifteenth and early
sixteenth centuries, a period also referred to as Middle Portuguese). During the thirteenth century and
most of the fourteenth century the constraint against placing clitics in sentence initial position known
by the name of Tobler-Mussafia excludes clitics not only from the absolute sentential left-periphery
but also from the immediate first position after a Left Dislocated constituent, a clausal adjunct
(namely an adverbial clause), or a coordinate conjunction. In the fifteenth century, however, the
interdiction against placing clitics sentence initially appears to be restricted to the absolute sentence-
peripheral position. From the second half of the fourteenth century, clitics can be attested following a
coordinate conjunction. In the next century clitics are also found following a left adjoined adverbial
clause or a Left Dislocated phrase. Whatever the nature of the Tobler-Mussafia constraint is (be it
syntactically or prosodically motivated), I take these facts to show that there was a weakening of this
constraint in the fifteenth century, instead of three near simultaneous changes affecting the syntax of
coordinate structures, the syntax of complex sentences integrating adverbial clauses, and the syntax
of Clitic Left Dislocation structures - but see Benincà (1995) for a different view.
DEFICIENT PRONOUNS AND LINGUISTIC CHANGE 215
(3) aos que imda lá sam, lhe_ tenho dado seguros

to-those that still there are, them-DAT I-have given safe-conducts
e lhe mando agora noteficar ho voso perdam.
and them I-send now notify-ofthe your pardon
"I have been giving protection to the people that are still there and
now I will notify them of your pardon."
(4) Algüas cousas mevdas de quaa da Imdia, que sera
some things small from here from-the India, that will-be
necessareas sabelas vossalteza, as_ esprevo
necessary to-know-them Your-Highness, them-ACC I-write
aquy nesta carta gramde.
here in-this letter long
"In this long letter I inform Your Majesty about some facts arising in
India which Your Majesty should be aware of."
Moreover, Rivero (1993, 1997) and Kaiser (1999) have brought into
consideration serious objections to the characterization of Old Portuguese and
Old Spanish as V2 languages (of the Germanic type).3 Their view further
challenges the P2 hypothesis for Old Romance clitics since it leaves unexplained
why clitics tend to cluster together with the verb.
In this paper I construct an empirical argument against the hypothesized
phrasal status of Old Portuguese and Old Spanish clitic pronouns by comparing
the syntax of the accusative/dative/se pronouns with the syntax of the oblique
pronouns i (locative) and en(de) (separative, locative, partitive, genitive). Space
considerations preclude me from offering in this paper an account of the changes
observed in clitic placement in Portuguese and Spanish which is compatible
with the view that clitics are heads throughout the history of the Romance
languages. Such an account is developed in Martins (in press; forthcoming).4
3
Rivero believes that through the history of the Romance languages clitic pronouns evolved from P2
clitics to verbal clitics. However, in her 1997 work Rivero advocates that Old Spanish was from the
earlier documented times undergoing a typological shift, therefore displaying a mixed system with
respect to the typology of clitics:
"OSp has the mixed characteristics which derive from the combination of its I- and C-Systems in
embedded clauses but seldom in main clauses. In main clauses, OSp shows an overwhelming
preference for the I-system [i. ., verbal clitics] that survives in later stages, and exhibits restrictions
in its use of the C-system [i. ., 2 clitics], which provides the basis for the diachronic evolution that
eliminates it in later periods". (Rivero 1997:170)
4
In forthcoming work, I derive synchronic and diachronic variation in clitic placement in Romance
from the variable featural make-up of the functional heads Σ and AgrS, namely from the interplay
between the 'strenghten' property of Σ (with respect to a V-feature - see Martins 1994a) and the EPP
2. An argument against the 'XP to X° hypothesis': comparing the syntax of

the oblique pronouns "i" and "ende" with the syntax of the accusative,
dative and "se"pronouns
The pronominal system of Old Portuguese and Old Spanish included the
oblique pronouns i and ende, the cognates of Catalan hi and en, French i and en,
Italian vi and ne. In the fifteenth century the partitive/locative ende vanishes
whereas i is reanalysed as a locative adverb (cf. Badia Margarit 1947, Teyssier
1981, Wanner 1991, Muidine 2000). In this paper I will not be concerned with
explaining why the oblique pronouns i and ende were not preserved as part of the
pronominal system of Portuguese and Spanish. My aim will be to compare the
syntax of i and ende with the syntax of the accusative, dative and se pronouns,
having as background the proposal of a tripartite typology of pronouns built up
by Cardinaletti and Starke (1996, 1999). Let me start with enunciating the form
of the argument that I will make to support the view that accusative, dative and
se pronouns did not change their categorial nature (from XP to Xo) during the
history of Romance. Working on the assumption that the tripartite typology of
pronouns put forth by Cardinaletti and Starke is correct, a comparative approach
to the syntax of the deficient accusative, dative, and se pronouns, on the one
hand, and of the deficient oblique pronouns i and ende, on the other, shows that
the latter classify as weak pronouns whereas the former classify as clitic
pronouns. Under the tripartite typology, weak pronouns are phrasal constituents
(XP); clitics are syntactic heads (Xo). The distinct syntactic distribution of the
obliques and of the accusative/dative/se pronouns in Old Portuguese and Old
Spanish can be straightforwardly derived from their different phrase structure
status. This line of reasoning leads us to the conclusion that the accusative,
dative and se pronouns are Xo items from Old Romance through Modern
Romance.
In a series of articles published in the nineties, Anna Cardinaletti and
Michael Starke demonstrated that the classical bipartite typologies of pronouns
properties of AgrS. Within this kind of account the burden of accounting for diachronic variation in
clitic placement in Romance is not put in the syntax of left dislocation. Thus variation between
enclisis and proclisis is not seen, at any stage of the evolutionary path of the Romance languages, as a
product of the availability/unavailability of left dislocation. This seems a welcome result in view of
the fact that empirical evidence undermines the claim that 'pure' configurations of Clitic Left
Dislocation would necessarily display an enclitic pattern in Old Romance. Moreover, with regard to
Portuguese, the 'left-dislocation-based' account of change in clitic placement depends on the
assumption that Subjects went from being optionally left dislocated (up to the sixteenth century,
when both enclisis and proclisis appear in SV sentences), to being not allowed to be left dislocated
(in the seventeenth-eighteenth centuries, when proclisis is the usual pattern of clitic placement in SV
sentences), to being obligatorily left dislocated (giving rise to the invariable enclitic pattern of
contemporary Portuguese). This seems a very unlikely path in a language that throughout the
historical period under consideration was consistently pro-drop (cf. Barbosa 2000).
('stressed' vs. 'unstressed', 'strong' vs. 'weak', 'full' vs. 'clitic', etc.) are
irredeemably contradicted by cross-linguistic empirical evidence and inadequate
on theoretical grounds. Starting with a thorough examination of Germanic and
Romance, subsequently extended to other language families, Cardinaletti and
Starke came to identify three types of pronominal forms: 'clitic' pronouns,
'weak' pronouns, and 'strong' pronouns. This tripartite typology was the key to
solve enduring issues in the linguistics literature such as the (typological) status
of the personal pronouns of German or of the Italian dative pronoun loro. Under
the tripartite typology of pronouns, Cardinaletti and Starke derive the particular
semantic, syntactic and prosodic properties of each class of pronouns from the
following features: a) clitics and weak pronouns, in contrast to strong pronouns,
are structurally 'deficient' - at the sub-lexical level of syntactic structure; b)
clitics are syntactic heads (Xo) while weak pronouns and strong pronouns are
maximal projections (XP); c) clitics lack word stress, thus being prosodically
dependent words; weak pronouns like strong pronouns carry word stress (but
differently from strong pronouns can easily undergo prosodic restructuring with
an adjacent stressed word). Three kinds of constraint affect clitics and weak
pronouns as a result of their 'deficient' nature: they cannot introduce new
referents in discourse; they cannot be coordinated; they cannot be modified.5
5
An anonymous reviewer pointed out the three Old Spanish sentences given below, which were
intended as counter-examples with respect to the generalization that deficient pronouns, namely
accusative/dative/se pronouns, cannot be modified. As the translations I give below show, I do not
take those sentences to involve (relative) modification of an accusative/dative pronoun. In my
interpretation, (i) below displays a complement clause selected by found; (ii) displays an (object-
related) adverbial predication clause; (iii) below displays a sentential adverbial clause, namely a
reason clause. With respect to the non accusative clitic lo in (i), see Menéndez-Pidal (1946:§ 130) and
Fernández-Ordoñez (1999). As for the predication clause in (ii), see Brucart (1999:442).
(i) E el Candilero Zifarparo mientes en aquel cauallo que ama ganado del cauallero
and the knight Zifar set his-mind in that horse that he-had won from-the knight
que auía muerto a la porta de la villa, e fallolo que era bueno e mui
that he-had killed at the door of the town and found-it that it-was good and very well
enfrenado e mui valiente.
trained and very brave
"And the knight Zifar paid attention to the horse that he had won from the knight that he had
killed at the entrance of the town and found (about it) that it was a good horse, very well
trained and brave." (cf. Wagner 1929:55)
(ii) Et el día que llegó a Toledo ader eçò luego a casa de dom Yllán et
and the day that he-arrived to Toledo he-went straight to the-house of Sir Yllán and
fallólo que estava lleyendo en una cámara muy apartada.
found-him that was reading in a room very retired
"And when he arrived to Toledo he went straight to Sir Yllán's house and he found him (while
he was) reading in a retired room." (cf. Blecua 1969:96)
(iii) e llamotę_ que solías estar muy arredrado de Su voluntad
and I-call-you as you-used to-be very distant from His will (cf. Ishikawa 1990:115)
In Old Portuguese and Old Spanish both the oblique pronouns i and ende
and the accusative/dative/se pronouns display the incapacity for introducing new
referents and for being coordinated, therefore classifying as 'deficient' pronouns.
In an extensive study of the Old Portuguese oblique pronouns Paul Teyssier
(1981) showed beyond doubt that i and ende cannot dispense with a discourse
antecedent; besides, Muidine (2000) brought up evidence showing that the
oblique pronouns i and ende do not allow coordination. Sentences (5) to (7)
below, extracted from Old Portuguese legal documents, illustrate the
'replacement' of i and ende by strong forms (i.e., Prep + strong pronoun) when
coordination comes into play. Example (7) displays coordinated strong forms
occurring in exactly the same kind of discourse/textual context as the
(uncoordinated) weak pronouns i and ende in examples (5)-(6).
(5) no dito cassall e terras e cassas (...) que daquj

in-the mentioned farm and lands and houses (...) that from-now
en.diantefaçom hy como de coussa ssua propria forra
on they-can-do there like of thing his own free
"that from now on they are free to use the lands, the farm and the
houses [which they have bought] in any manner that they may wish"
(year 1483. Martins 2001:526)
(6) das ditas herdades e quintãa e Casal dela (...)
of-the above-mentioned lands and farm and house of-it (...)
que daqui adeãte o ouuesse o dito Monsteiro
that from-now on it should-own the above-mentioned monastery
liureměte e ě paz e fezesse ende .que lhj aprouuesse
freely and in peace and should-do of-it whatever itself wished
sem ebargo seu nëhüu
without constraint of-him none
"that from now on the monastery owns the lands and the farm with its
house without any constraints, using them freely in any manner that
the monastery may wish" (year 1339. Martins 2001:198)
(7) Aíam (...) as sobreditas Casas (...) e façã délias
own the above-mentioned houses (...) and do of-they-FEM
e ë ellas todo.Aquello que lhys Aprouguer
and in they-FEM everything that themselves they-may-wish
"they will own the houses and use them in any manner that they may
wish." (year 1383. Martins 2001:462)
The facts considered above show that the oblique pronouns i and ende
qualify as deficient pronouns like the accusative/dative/se pronouns. However
the two groups of deficient pronouns diverge from each other in the following
aspects:
A. In (finite) subordinate clauses, the accusative/dative/se pronouns are

normally preverbal, differing in this way from strong pronouns and DPs in
general. In this respect Old Portuguese and Old Spanish are like contemporary
Portuguese and Spanish. Oblique pronouns, on the other hand, do not display a
'special' placement in subordinate clauses. Thus i and ende are often post-verbal
in subordinate clauses.6 This fact is in harmony with the hypothesis that Old
Romance accusative/dative/se pronouns are 'special clitics' (in the terms of
Zwicky 1977) in contrast with the oblique pronouns.7 Since both the oblique
pronouns and the accusative/dative/se pronouns are deficient pronouns, their
different placement indicates that while the latter are clitics the former are weak
pronouns. The observed distributional contrast between the two sets of pronouns
under consideration is highlighted by sentences (8)-(9) below.
(8) das quaes cousas o dito Martim dominguiz queixada

of-the which things the mentioned Martim Dominguiz Queixada
pedyu a m que Ihy desse ende hun testemoyo
asked to me that him I-would-give of-them a written-statement
"Martim Dominguiz Queixada asked me to give him a written
statement certifying the facts"
(Portuguese, year 1294. Martins 2001:365)
6
Like full DPs, oblique pronouns may also occur in preverbal position in subordinate clauses. Old
Portuguese and Old Spanish allowed Object IP-scrambling, deriving SOV order in subordinates (as
well as in a certain kind of main clauses). See Martins (2002).
7
Some apparent exceptions to the regular proclitic pattern of subordinate clauses can be found in
particular contexts. Complement clauses selected by declarative verbs, for example, may allow
enclisis (see Martins 1994b:98-102). Torrego & Uriagereka (1993), however, claim on independent
grounds that these are instances of parataxis, not 'true' subordination (i. e. hypotaxis). On the other
hand, enclisis appears to be also attested in coordinate structures within embedded domains -
relevant examples are found not only in Old Portuguese and Old Spanish but in other Old Romance
languages, as pointed out for Old French by Labelle & Hirschbühler (2002). Since in the medieval
texts punctuation does not signal syntactic boundaries in the way it does in contemporary texts, the
level at which coordination applies is not always easily identified. But even if we are left with a
handful of clear examples of enclisis in clauses involving coordination below CP, the argument in A
above still stands. The relevant point is that oblique pronouns occur widespreadly in post-verbal
position in subordinate clauses, being not restricted to the particular environments where
accusative/dative/se pronouns can also be found.
(9) et moraban cerca de una villa, et un dia que se_

and they-lived close to a village and one day that se(-PASSIVE)
facía y mercado
made there a-market
"and they lived close to a village and a certain time when there was a
market there"
(Spanish. Cf. Badia Margarit 1947:120)
B. It is a well established generalization about the Romance languages with

pre-verbal negation that only an Xo element, namely a clitic, is allowed to break
the linear adjacency between the sentential negative marker and the verb. In Old
Portuguese and Old Spanish the accusative/dative/se pronouns not only may
occur between negation and the verb but actually do occur in that position quite
often (see Fontana 1993, Martins 1994b). In opposition to the
accusative/dative/se pronouns, the oblique pronouns (with a few exceptions for
Old Spanish i ) cannot intervene between the sentential negative marker and the
verb; therefore, they do not qualify as clitics whereas the accusative/dative/se
pronouns do.8 In negative sentences, either the oblique pronouns are post-verbal
or, if preverbal, they precede the string 'negation-verb'.9 This scenario is
expected under the hypothesis that the latter are weak pronouns (i.e., XPs) while
the former are true clitics (i. ., X°s). Sentences (10) to (12) exemplify the
exclusion of i and ende from the 'inter-negation-and-verb' position where clitics
- like the dative pronoun in (11) and the reflexive in (12) - are permitted.
(10) E  Monesteiro oú Ejgreia que nõ oúuer des L" Casáaes a

and the monastery or church that not owns from fifty farms to
Jusu no pouse hj Ricomě
more not stay there man-of-noble-birth
"And if the monastery or church does not own fifty farms or more, no
aristocrat is to be hosted there."
(Portuguese. 13th/14th century. Martins 2001:166)
8
The Old Spanish oblique i is attested in the string 'neg-i'-V', especially when V equals haber 'have'.
The fact that Old Spanish i may occur in such position shows that it could be a clitic at least in some
dialect of Castillian. This is not that surprising in view of the fact that in the Eastern Iberian Romance
languages, namely in Old Aragonese and Old Catalan, the oblique pronouns displayed clitic behavior
in tandem with the accusative/dative/se pronouns (cf. Badia Margarit 1947). Nevertheless, in
Castillian the clitic variant of the oblique i did not displace the weak non clitic variant. In fact, more
often than not Old Spanish i behaves as an XP: it occurs post-verbally in finite negative clauses and
in all kinds of subordinate clauses; it may be clause initial; it does not undergo mesoclisis; it is not
part of clitic clusters.
9
In this respect as well oblique pronouns contrast with accusative/dative/se pronouns, since the latter
are always preverbal in finite negative clauses.
(11) que lo emiende (...) e nol fagan end ningún amor

that it he-corrects (...) and not-him show for-it any fondness
"that he corrects it and nobody should show him any sympathy for
that"
(12) Aves torpes nin roncas hi non se acostavan.
birds injured nor snorty there not themselves would-lie
"Injured birds or snorty birds would not be able to sleep there."
 Similarly to strong pronouns, the oblique pronouns i and ende are always
post-verbal in configurations where accusative/dative/se pronouns display
variation between proclisis and enclisis (that is, in affirmative main clauses
without proclisis triggers).10 In independent work (Martins 1994b, 1995, in press,
forthcoming), I tackle the issue of the variable placement of the
accusative/dative/se pronouns in Old Portuguese and Old Spanish. The account
proposed ties such variation to the Xo character of the accusative/dative/se
pronouns. Thus the weak oblique pronouns, being XPs, do not participate in it.
Sentences (13)-(14) exemplify the constant post-verbal position of the oblique
pronouns in affirmative main clauses (without proclisis triggers nor focusing of
the oblique) in contrast to the variable (preverbal or post-verbal) placement of
the accusative/dative/se pronouns.11
10
The phrasal constituents which induce proclisis in affirmative main clauses are identified in
Martins (1994b, 1995).
11
Old Portuguese and Old Spanish are SVO languages, but the order OV can be derived in main
clauses by left-dislocating or focusing the object - the former displacement operation marks a
constituent as a topic, the latter as an identificational focus (see Martins 2002). Although rare,
sentences with focusing of the oblique pronouns i, en(de) are found in the medieval texts - see (22)
and (23) below. With repect to the accusative/dative/se pronouns, variation between preverbal and
post-verbal position is independent of focusing (which accusative/dative/se pronouns cannot
undergo). Diachronically viewed, the frequency of preverbal placement makes clear the distinction
between obliques on the one hand and accusative/dative/se pronouns on the other: the preverbal
placement of the oblique pronouns (resulting from focusing) shows stable infrequency throughout the
medieval period; as for the accusative/dative/se pronouns, there is a gradual and steady increase of
the preverbal placement from the 12th/13th century to the end of the medieval period. In late Old
Portuguese and Old Spanish proclisis arises more often than not in affirmative main clauses (without
proclisis triggers). See Martins (1994b, 1995).
222 ANA MAMA MARTINS
(13) E  Monesteiro où Ejgreia que nõ oúuer des La Casáaes a

and the monastery or church that not owns from fifty farms to
Jusu no pouse hj Ricomě mais façãlhj ende seruiço
more not stay there man-of noble-birth but do-him of-it service
hüa vez no ãno.
one time in-the year
"And if the monastery or church does not own fifty farms or more, no
aristocrats are to be hosted there but the monastery has the obligation
to provide them with its services once a year."
(Portuguese. 13th/14th century. Martins 2001:166)
(14) E porque ele esto dezia muito ameudi, os seus amigos que
and because he that would-say very often the his friends who
estovan derredor dele preguntaron-lhi a quem dezia e
were around him asked-him to whom he-was-saying and
ele se. maravilhou ende muito
he himself marveled of-it much
"and because he would say that very often, his friends who were with
him asked him whom he was talking to and he was very surprised
with that question" (Portuguese. Cf. Teyssier 1981:173)
D. As Cardinaletti (1999:40) points out "no Romance language allows for

lexical material to intervene between the verb and an enclitic pronoun". This is a
very robust generalization about clitics in Romance. It describes the behavior of
the accusative/dative/se pronouns throughout the history of Portuguese and
Spanish. The oblique pronouns i and ende, on the other hand, may occur non
adjacent to the verb in post-verbal position. Once more, this distributional
contrast may be derived from the different character of the two relevant sets of
pronouns. The accusative/dative/se pronouns are clitics (Xo) from earlier times
requiring an Xo host; the oblique pronouns i and ende are weak pronouns (XP)
which may stay on their own - although they may as well undergo prosodic
restructuring, thus changing a neighbour into a host (see Cardinaletti & Starke
1999:172-173). Sentences (15)-(17) exemplify the permitted discontinuity
between the verb and a post-verbal oblique pronoun. Moreover, (15) highlights
the contrast between the placement of the oblique pronoun and the placement of
a post-verbal dative pronoun, which as expected surfaces adjacent to the verb.
(15) E filarü li_ illos inde VI casales cữ torto.

And stole him they of-it six farms against the-law
"And they illegally deprived him of six of those farms."
(Portuguese. Cf. Castro 1991:232)
(16) E o santo homen pois esto ouvio ficou logo ende mui
and the holy man as this he-heard was immediately of-it very
triste.
sad
"And as soon as he heard that, the holy man was very sad."
(Portuguese. Cf. Teyssier 1981:173)
(17) & faze a los om[n]es la malazon tan fuerte que muere[n]
and he-makes to the people the illness so strong that die
los omnes ende
the people of-it
"and he causes people to be so ill, that people die because of it"
(Spanish. Cf. Wanner 1991:354)
E. 'Mesoclisis' terms the situation in which a clitic is placed inside a

morphological word, namely a verb, surfacing in between the verbal root and the
inflectional suffixes (T+Agr). Since morphological words, namely verbs, are Xo
items, only another Xo can become a segment of the former. The weak oblique
pronouns being XPs do not undergo mesoclisis whereas the clitic
accusative/dative/se pronouns do.12 Sentences (18)-(21) below illustrate the
relevant distinction.
(18) E enton dar-lh '-ia Deus lume de seus olhos

and then give-him-CONDITIONAL God light of his eyes
"and then God would give him (some of) the light of his eyes"
(Portuguese. Cf. Silva 1989:850)
(19) e gram nojo receberias ende
and great suffering you-get-CONDITIONAL from-it
"and that would cause you a great deal of pain"
(Portuguese. Cf. Silva 1989:237)
(20) & venir-te-α ende periglo & crebanto
and come-you-FUTUR from-there danger and ruin
"and you will receive from it danger and damage"
(Spanish. Cf. Wanner 1991:355)
12
With futur and conditional forms of the verb, accusative/dative/se pronouns can be either
mesoclitic or proclitic/enclitic in Old Portuguese and Old Spanish. Crucially the obliques are never
attested in mesoclitic position (independently on whether the oblique cooccurs with an
accusative/dative/se pronoun in the same clause).
Roberts (1993) sees the future and conditional markers in Old Spanish as verbal clitics, not as
verbal affixes. Even if we were to adopt Robert's view on this matter, the argument in E above would
still stand.
(21) dar-vos-he y de las nueses

I-give-you-FUTUR there of the nuts
"there I will give you some nuts"
F. The oblique pronouns escape the Tobler-Mussafia restriction (see Tobler

(1875, 1889; Mussafia 1886) which states that clitics are excluded from the
absolute left peripheral position in a sentence. The ban against being 'first',
which obligatorily affects the accusative/dative/se pronouns, does not necessarily
affect the oblique pronouns i and ende, as sentences (22)-(23) below show. It is
not clear what motivates the Tobler-Mussafia restriction (see Benincà 1995,
Halpern 1995, Rivero 1997, Fontana 1997 for references and discussion) and I
will not deal with the issue in this paper. But I take the fact that i and ende may
escape the Tobler-Mussafia constraint as a further indicator of their (non-clitic)
weak character and, conversely, of the clitic character of the accusative/dative/se
pronouns. Moreover, following Kayne (1991) and Cardinaletti and Starke (1996,
1999), I take 'being a clitic' to mean 'being an X o '.
(22) E eu Pedro dominguiz publico tabeliõ de Lixbõa rogado das

and I Pedro Dominguiz public notary of Lisbon requested by-the
partes de.suso ditas å éesto presente ffoy / êde duas
parts above mentioned to this(-act) present was / of-it two
cartas partidas per a. b. . õ mhamão propria escreuy.
charters broken by ABC with my hand own I-wrote
"And I, Pedro Dominguiz, notary public of Lisbon, was present at this
public act at the request of the parts involved in the agreement. I wrote
myself two documents certifying it."
(Portuguese. year 1297. Martins 2001:397)
(23) Buscó algún lugar de gran religión:
he-looked-for some place of great mysticism:
-I sóvo escondido faciendo oración.
- There he-stayed hidden doing prayers
"He looked for some mystical place. He hid there doing his prayers."
G. A characteristic feature of clitics is their clustering together within the same

clause. The accusative, dative and se pronouns obligatorily form clitic clusters
within the clause, but the oblique pronouns may occur separated from the
accusative/dative/se pronouns. This distinct behavior is one more piece of
evidence strenghtening the case for the different typological status of the oblique
pronouns i and ende, on the one hand, and of the accusative/dative/se pronouns
on the other.13 The latter but not the former behave as clitics, i.e., as X o syntactic
entities. The intraclausal positional dissociation between the oblique pronouns
and the accusative/dative/se pronouns arises in configurations where the two
types of pronouns would be allowed to be contiguous. In sentence (24) below
both the accusative la and the oblique hy could be placed before the string
'negation-verb'; in sentence (25) below both the accusative as and the oblique hj
could be either left adjacent to the verb or right adjacent to the complementizer.
See also sentences (8), (14) and (15) above.
(24) Pora levar el agua a los de iuso, que no la sabíen

to carry the water to the-ones of downwards who not it knew
hy nunquas puesta.
there never placed
"To carry the water to those who were downwards, who did not know
it to be placed there ever." (Spanish. Cf. Rivero 1991:276)
(25) e pollo dito escambho todallas vjnhas que o
and for-the mentioned exchange all-the vineyards that the
dito Moesteiro ha ë AlfforneL (...) Assj como as
mentioned monastery owns in Alffornel in-the-way that them
o dito Mosteiro hj a
the mentioned monastery there owns
"you will irrestrictedly own all the vineyards that the monastery owns
in Alffornel in exchange for the properties which you agreed on
giving to the monastery"
(Portuguese. year 1372. Martins 2001:451)
13
The fact that ι and en(de) may be contiguous with the accusative/dative/rø pronouns (or with each
other) is by itself irrelevant. Contiguity per se does not show whether there is clitic cluster formation.
What signals that this is the case is the obligatoriness of contiguity. Moreover, it should be noted that
while contracted forms involving two clitics can be found in Old Spanish and Old Portuguese, there
are no such contracted forms involving i or en(de). In this respect, Old Spanish and Old Portuguese
contrast sharply with Old Catalan where en and i contract together (giving nie,nhi) as well as with
the accusative/dative/se pronouns (see Badia Margarit 1947:213-219).
An anonymous reviewer suggested that ''oblique pronouns probably do not cluster within the same
clause because the locative meaning of both i and ende rules out this possibility". Old Catalan
contracted forms such as ni undermine this claim. The presence of ι and en(de) within the same
clause is also attested in Old Spanish and Old Portuguese, as sentence (i) below exemplifies,
(i) Ε todos aqueles beschos que na horta andavan e as verças comiam partiron-se
and all those animals that in-the garden were and the cabbages ate left-themselves
do horto e nunca ende hi hữữ f icou
from-the garden and never of-them there one stayed
"And (all) the animals that used to be in the garden and eat the cabbages left, not staying there
any of them." (Portuguese. Cf. Teyssier 1981:174)
The comparative inquiry undertaken in this paper offered for consideration

a set of facts which have not gotten enough attention in the literature on Old
Romance clitics. Under the interpretation of the data that I have argued for,
Rivero's claim (in 1986, 1991) that the oblique pronouns and the
accusative/dative/se pronouns of Old Romance are syntactic entities of the same
nature, that is, XPs, cannot be maintained. Assuming the tripartite typology of
pronouns proposed by Cardinaletti and Starke (1996, 1999), I take the facts
described above to show that the set of deficient personal pronouns of Old
Portuguese and Old Spanish splits into two subsets: a group of weak pronouns
that do not survive beyond the 15th century and a group of clitic pronouns which
(as expected under the assumption that clitics are heads) keep their Xo nature
unchanged across all the documented lifetime of the Romance languages.
The conclusion that Old Portuguese and Old Spanish accusative/dative/se
pronouns are not weak pronouns, but true clitics, might be reinforced by the
observation that a weak third person accusative pronoun (phonologically similar
to the third person nominative, i. ., ele/ela) is attested in OP (see sentence (26)
below). This weak accusative pronoun is to be distinguished from a strong third
person accusative pronoun also attested in Old Portuguese when coordination,
modification or focusing are at play (see sentences (27)-(29) below). The weak
third person accusative is in variation with the accusative clitic in the same type
of syntactic contexts; the strong third person accusative occurs in configurations
where a clitic is not allowed. The weak accusative pronoun (ele/ela), like the
weak oblique pronouns (i and ende), does not survive beyond the medieval
period.
(26) Eu dõ pedro soarez dauãdito que esta carta encomedej

I Dom Pedro Soarez above-mentioned that this chart sent
a fazer ela· date homéés cõ mas maos roboro.
to do it (the chart) before men with my hands I-validate
"I Dom Pedro Soarez validate this document that I sent to write in the
presence of witnesses."
(year 1277. Martins 2001:129)
(27) Contando como cativarom elle e os outros oito.
Telling how they-caught him and the other eight
"Telling how they caught him and the other eight of them."
(cf, Said Ali 1931:94)
(28) Per di ela, que foy a rrě milhor.
I-lost her who was the thing best
"I lost her, who was the best thing." (cf Dias 1918:71)
(29) El amiga, achei eu andar tal / come morto.

him friend found I to-be such like dead
"I saw HIM, my friend, looking like a dead person."
(cf. Brea et alii 1996:209)
3. Conclusion
In this paper I argue that the deficient accusative, dative and se pronouns of
Portuguese and Spanish did not change their categorial nature throughout the
history of these languages, being continuously X o entities. Since, for this matter,
Old Portuguese and Old Spanish have been widespreadly taken in the literature
as representative of an archaic stage in the history of Romance, I hence conclude
that Romance clitics are just true clitics (i. e. heads) from Old Romance to
Modern Romance. This fact reveals itself sharply when we compare the syntax
of the accusative, dative and se pronouns with the syntax of the oblique pronouns
i and en(de) in Old Portuguese and Old Spanish. The latter display properties of
weak pronouns (XPs) in contrast to the clitic properties of the former.
References
Ali, Manuel Said 1931. Gramática Histórica da Lingua Portugueza. S. Paulo:
Melhoramentos.
Badia Margarit, Antonio M. 1947. Los Complementos Pronominalo-Adverbiales
Derivados de IBI e INDE en la Peninsula Iberica. (= Anejos de la Revista de
Filología Española, 38). Madrid: Consejo Superior de Investigaciones Científicas.
Barbosa, Pilar 1993. "Clitic Placement in Old Romance and European Portuguese".
Papers from the Twenty-Ninth Regional Meeting of the Chicago Linguistic Society.
Chicago: Chicago University Press.
1996. "Clitic Placement in European Portuguese and the position of Subjects'·.
Second Position Clitics and Related Phenomena, ed. by Aaron L. Halpern & Arnold
M. Zwicky, 1-40. Stanford, Calif: CSLI Publications.
2000. "Clitics: A Window into the Null Subject Property". Portuguese Syntax:
New Comparative Studies, ed. by João Costa, 31-93. Oxford & New York: Oxford
University Press.
Benincà, Paola. 1995. "Complement clitics in Medieval Romance: the Tobler-Mussafia
Law". Clause Structure and Language Change, ed. by Adrian Battye & Ian Roberts,
325-344. New York & Oxford: Oxford Universiry Press.
Blecua, José Manuel, ed. 1969. Don Juan Manuel, El Conde Lucanor o libro de los
enxiemplos del Conde Lucanor et de Patronio. Madrid: Castalia.
Brea, Mercedes et alii, eds. Lírica Profana Galego-Portuguesa: Corpus completo das
cantigas medievais, con estudio biográfico, análise retórica e bibliografa
específica. Santiago de Compostela: Ramón Piñeiro.
Brucart, José Ma. 1999. "La estructura del sintagma nominal: Las oraciones de relativo".
Gramática Descritiva de la Lengua Española, ed. by Ignacio Bosque & Violeta
Demonte, 395-522. Madrid: Espasa Calpe.
Cardinaletti, Anna 1999. "Pronouns in Germanic and Romance Languages: An

Overview". Clitics in the Languages of Europe, ed. by Henk van Riemsdijk, 33-82.
Berlin: Mouton the Gruyter.
& Michael Starke 1996. "Deficient Pronouns: A view from Germanic. A Study in
the Unified Description of Germanic and Romance". Germanic Syntax, ed. by
Höskuldur Thráinsson, Samuel David Epstein & Steve Peter, 21-65. Dordrecht,
Boston & London: Kluwer.
1999. "The typology of structural deficiency: A case study of the three classes of
pronouns". Clitics in the Languages of Europe, ed. by Henk van Riemsdijk, 145-
233. Berlin: Mouton the Gruyter.
Castro, Ivo 1991. Curso de História da Lingua Portuguesa. Lisboa: Universidade Aberta.
Dias, A. E. da Silva 1918. Sintaxe Histórica Portuguesa. Lisboa: Clássica Editora. 1970
(5th edition).
Fernández-Ordoñez, Inés 1999. "Leísmo, laísmo, loísmo". Gramática Descritiva de la
Lengua Española, ed. by Ignacio Bosque & Violeta Demonte, 1317-1397. Madrid:
Espasa Calpe.
Fontana, Josep M. 1993. Phrase Structure and the Syntax of Clitics in the History of
Spanish. PhD dissertation. Philadelphia: University of Pennsylvania.
1997. "On the integration of second position phenomena". Parameters of
Morphosyntactic Change, ed. by Ans van Kemenade & Nigel Vincent, 207-249.
Galves, Charlotte 1997. "Do português clássico ao portugués europeu moderno: Urna
análise minimalista". Estudos Lingüísticos e Literários, 19, ed. by R. V. Mattos e
Silva, 105-128. Salvador: Universidade Federal da Bahia. [Included in Galves 2001]
2001. Ensaios sobre as Gramáticas do Português. Campinas, Sao Paulo: Editora
da Unicamp.
Gifford, D. J. & F. W. Hodcroft. 1959. Textos Lingüísticos del Medioevo Español. Oxford:
The Dolphin Book.
Halpern, Aaron 1995. On the Placement and Morphology of Clitics. Stanford, Calif:
CSLI Publications.
& Joseph M. Fontana. 1994. "Xo Clitics and Xmax Clitics". Proceedings of the
West Coast Conference on Formal Linguistics (WCCFL XII). Stanford, Calif: CSLI
Publications.
Ishikawa, Masataka 1990. Factorization in the Grammar and Syntactic Change: A study
in the categorial evolution of clitics in Spanish. Ann Arbor, Michigan: Ann Arbor
Microfilms International.
Kaiser, George 1992. Zur Grammatik der klitischen Personalpronomina im Französischen
und Portugiesischen: Eine synchronische und diachronische Analyse im Rahmen
der Prinzipien- und Parameter- theorie. Ph. D. Dissertation. Universität Hamburg.
1999. "A ordern das palavras e a posição do verbo finito no portuguĉs antigo".
Actas do Congress o Internacional Organizado por Motivo dos Vinte Anos do
Português no Ensino Superior, 248-259. Budapeste: Departamento de Lingua e
Literatura Portuguesas da Faculdade de Letras da Universidade Etövös Loránd.
Kayne, Richard 1991. "Romance Clitics, Verb Movement, and PRO". Linguistic Inquiry
22.647-686.
Labelle, Marie & Paul Hirschbühler 2002. "Changes in clausal organization and the
position of clitics in Old French". Paper delivered at VII Diachronic Generative
Syntax Conference, Girona.
Martins, Ana Maria 1994a. "Enclisis, VP-deletion and the nature of Sigma". Probus
6.173-205.
1994b. Clíticos na Histόria do Português. Ph.D. Dissertation. Lisboa: Faculdade
de Letras da Universidade de Lisboa.
1995. "Clitic Placement from Old to Modern European Portuguese". Historical
Linguistics 1993, ed. by Henning Andersen, 295-307. Amsterdam & Philadelphia:
John Benjamins.
2000. "A minimalist approach to clitic climbing". Portuguese Syntax: New
Comparative Studies, ed. by João Costa, 169-190. Oxford & New York: Oxford
University Press. [Reprinted, with minor revisions, from: Ana Maria Martins 1995,
same title. Papers from the 31st Regional Meeting of the Chicago Linguistic Society,
215-239].
2001. Documentos Portugueses do Noroeste e da Região de Lisboa: Da
Produção primitiva ao Século XVI. Lisboa: Imprensa Nacional - Casa da Moeda.
2002. "The loss of IP-scrambling in Portuguese: Clause structure, word order
variation and change". Syntactic Effects of Morphological Change, ed. by David
Lightfoot, 232-248. Oxford & New York: Oxford University Press.
in press. "Tipología e mudança linguísticas: os pronomes pessoais do português e
do espanhol". Santa Barbara Portuguese Studies VII, ed. by Eduardo Raposo &
Harvey Sharrer.
forthcoming. "VP-ellipsis, clitic placement, and scrambling in Romance".
[Selected papers from VII Diachronic Generative Syntax Conference, Girona].
Menéndez Pidal, R., ed. 1946. Cantar de Mio Cid: Texto, Gramática y Vocabulario.
Madrid: Espasa-Calpe. 1980 (5th edition).
Moraes, M. Aparecida Torres 1995. Do Português Clássico ao Portugués Moderno: Dm
Estudo da Cliticizaçao e do Movimento do Verbo. PhD dissertation. São Paulo:
Uni camp.
Muidine, Soraia Aboo 2000. Os pronomes i e en(de) no portugués dos séculos XIII a XVI.
M. A. Dissertation. Lisboa: Faculdade de Letras da Universidade de Lisboa.
Mussafia, Adolfo 1886. "Una particolarità sintattica deila lingua italiana dei primi secoli".
Miscellanea di Filología e Linguistica in Memoria di Napoleone Caix e Ugo Angelo
Canello, ed. by G. I. Ascoli et alii, 255-261. Firenze: LeMonnier.
Pato, R. A. Bulhäo, ed. 1884. Cartas de Afonso de Albuquerque Seguidas de Documentos
que as Elucidam. Vol. I. Lisboa: Academia Real das Sciencias de Lisboa.
Ribeiro, liza 1995a. "Evidence for a verb-second phase in Old Portuguese". Language
Change and Verbal Systems, ed. by Adrian Battye & Ian Roberts, 110-139. New
York & Oxford: Oxford University Press.
1995b. A Sintaxe da Ordern no Portugués Arcaico: O Efeito V2. PhD
dissertation. São Paulo: Unicamp.
Rivero, M. Luisa. 1986. "La tipología de los pronombres átonos en el español medieval y
en el español actual". Anuario de Lingüística Hispánica, II, 197-220.
—.._____ 1991. "Clitic and NP Climbing in Old Spanish". Current Studies in Spanish
Linguistics, ed. by Héctor Campos & Fernando Martínez-Gil, 241-282. Washington,
D. C: Georgetown University Press.
- 1993. "Long Head Movement vs. V2, and null subjects in Old Romance".
Lingua 89.217-245.
- 1997. "On two locations for complement clitic pronouns: Serbo-Croatian,
Bulgarian and Old Spanish". Parameters of Morphosyntactic Change, ed. by Ans
van Kemenade & Nigel Vincent, 170-206. Cambridge: Cambridge University Press.
Roberts, Ian 1993. "A Formal Account of Grammaticalisation in the History of Romance
Futures". Folia Linguistica Historica XIII/1-2.219-258.
Salvi, Giampolo. 1990. "La sopravvivenza deila legge di Wackernagel nei dialetti
occidentali deila Peninsola Iberica". Medioevo Romanzo 15.177-210.
1991. "Difesa e illustrazione deila legge di Wackernagel applicata alle lingue
romanze antiche: La pozicione delle forme pronominali clitiche". Per Giovan
Battista Pellegrini: Scritti degli allievi padovani, 439-462. Padova: Unipress.
1993. "Ordine deile parole e struttura deila frase nelle lingue romanze antiche".
Alfa 37.187-203.
1997. "From Latin weak pronouns to Romance clitics". Estudos Lingüísticos e
Literários, 19, ed. by R. V. Mattos e Silva, 85-104. Salvador: Universidade Federal
da Bahia.
Silva, Rosa Virginia Mattos e. 1989. Estruturas Trecentistas: Elementos para urna
Gramática do Português Arcaico. Lisboa: Imprensa Nacional - Casa da Moeda.
Teyssier, Paul 1981. "Le système des deíctiques spatiaux en portugais aux XIVe, XVe et
XVIe siècles". Cahiers de Linguistique Hispanique Médiévale 6.5-39. Reprinted in:
Études de Littérature et de Linguistique, 161-198. Paris: Fundação Calouste
Gulbenkian & Centro Cultural Português. 1990.
Tobler, Adolf 1875. Review of "J. Le Coultre, De L'ordre des mots dans Chrétien de
Troyes". Göttingische Gelehrte Anzeigen, Stück. 34.1057-1082. Reprinted in:
Vermischte Beiträge zur Französischen Grammatik V. 1902 & 1971 [Amsterdam:
Rodopi].
—- 1889. "Pronominales Objekt zu Infinitiv oder Partizipium" Zeitschrift für
romanische Philologie 13.186-191. Reprinted in: Vermischte Beiträge zur
Französischen Grammatikll. 1902 & 1971 [Amsterdam: Rodopi].
Torrego, Esther & Juan Uriagereka 1993. "Indicative Dependents". Ms. University of
Massachusetts at Boston & University of Maryland at College Park.
Wagner, Ch. Ph., ed. 1929. El libro del cavallero Zifar. Ann Arbor, Michigan: University
of Michigan Press.
Wanner, Dieter 1991. "The Tobler-Mussafia law in Old Spanish". Current Studies in
Spanish Linguistics, ed. by Héctor Campos & Fernando Martinez-Gil, 313-378.
Washington. D.C.: Georgetown University Press.
Zwicky, Arnold 1977. On Clitics. Bloomington: Indiana University Linguistics Club.
NOMINALIZATIONS OF FRENCH PSYCHOLOGICAL
VERBS
SYNTACTIC COMPLEMENTS AND SEMANTIC PARTICIPANTS*
JUDITH MEINSCHAEFER
University of Konstanz, Germany
1. Introduction
This paper addresses the question of how derived nominalizations realize
their arguments. If argument realization in derived nominalizations is subject to
the same rules as argument realization in verbs, differences in argument structure
between verbs and nominalizations must relate to the morphological process of
nominalization, which has been claimed by some to affect the argument structure
of a lexeme. If, on the contrary, argument realization in verbs and nouns is
governed by different rules, differences in argument structure might arise from
these different rules. In this paper, this question is addressed on the basis of the
grammaticality contrast exemplified in (1) and (2). The base verbs in (la) and
(2a) are both transitive verbs with a similar meaning. Still, the derived
nominalizations in (lb) and (2b) present a striking difference with regard to how
their arguments can be realized by syntactic complements.
(1) a. Le livre l'a fasciné.

"The book fascinated him."
b. La fascination du livre m'a surpris.
"The fascination of the book surprised me."
(2) a. La lettre l'a déçue.
"The letter deceived her."
b. #La déception de la lettre m'a surpris.
"The deception of the letter surprised me."
* This work was supported by the DFG through its SFB 471 in Konstanz. I would like to thank
Carmen Keilling, Bruce Mayo, Christoph Schwarze, and two anonymous reviewers for their
comments on earlier versions of this paper, and to Martine Lorenz-Bourjot for helping me with the
data. All remaining errors are mine.
232 JUDITH MEINSCHAEFFER
It is argued that the grammaticality contrast in (1) and (2) is correlated with
a difference in the semantic structure of the underlying verbs. It is further argued
that the contrast can be explained by assuming that the realization of arguments,
or better, of semantic participants, of derived nominalizations can be projected
from the underlying semantic form of the verbs involved, drawing on three
realization rules for semantic participants of nominalizations. The discussion is
restricted to psychological verbs and their derived nominalizations, because only
within this semantic class we find verbs showing the same surface syntax and
similar meaning, but presenting derived nouns contrasting crucially in how their
semantic participants can be realized.
The paper is structured as follows. Section 2 introduces some
characteristics of psychological verbs and nominalizations in French. In section
3, a number of descriptive generalizations about the realization of semantic
participants in psychological nominalizations are presented. Section 4 explores
the difference between classes of psychological verbs. In section 5, semantic
representations for three classes of psychological verbs are introduced. In section
6, I propose three participant-realization rules by means of which the descriptive
generalizations stated in section 3 can be derived from the semantic
representations given in section 5.
2. Psychological verbs as a semantic class

By psychological verbs, we mean verbs denoting an attitudinal or emotional
relation between two entities, where the precise nature of this relation is part of
the lexical meaning of the verb. Examples are given in (3).
(3) chagriner "grieve", mépriser "scorn", regretter "regret", angoisser

"afflict", amuser "amuse", étonner "astonish",fasciner "fascinate"
One constant property of psychological verbs appears to be that their semantic

representation (on some level) includes an entity experiencing the attitudinal or
emotional relation denoted by the verb; hence, these verbs are sometimes
subsumed under the class of 'experiencer verbs'. In the examples given in (4),
the experiencing entity is realized as subject in (a), as direct object in (b), and as
indirect object in (c).
(4) a. Lili adore sa petite sœur.

"Lili adores her little sister."
b. Ce tableau l'a toujours fasciné.
"This picture has always fascinated him."
NOMINALIZATIONS OF FRENCH PSYCHOLOGICAL VERBS 233
 Cette chanson ne plaît pas à Natalie.

"This song does not please Natalie."
The thematic roles assigned to the arguments of psychological verbs are

sometimes said to include the roles EXPERIENCER (of the attitudinal or emotional
relation), THEME (what is experienced), or CAUSE or STIMULUS (of the
experience). But it remains unclear whether the arguments of different classes of
psychological verbs are assigned the same thematic roles (as assumed, e.g., by
Belletti & Rizzi 1988; Grimshaw 1990) or different thematic roles (e.g., Pesetsky
1995; Iwata 1995; Wanner 2001), given that their syntactic realizations differ
considerably, cf. (4a-c).
Here, we assume that psychological verbs constitute a class of verbs sharing
certain lexical and conceptual features, but not necessarily sharing the same
thematic grid. Therefore, in the following we do not speak of arguments as
bearing certain thematic roles. Instead, we assume that the semantic participants
of psychological predicates are assigned two kinds of semantic or conceptual
roles, that of 'experiencing entity' and that of 'target or cause of the experience'.
At this point in the discussion, we distinguish only two classes of
psychological verbs, on the basis of how the two semantic participants are
realized syntactically: psychological verbs realizing the experiencing entity as
subject are termed 'experiencer-subject verbs' or ES-verbs; psychological verbs
realizing the experiencing entity as direct object are termed 'experiencer-object
verbs' or EO-verbs. Examples for both verb classes are given in (5a) and (5b),
respectively.
(5) a. ES-verbs: mépriser "scorn", adorer "adore", souhaiter "wish"

b. EO-verbs: réconforter "comfort", étonner "astonish"
The next section addresses the question of how nominalizations derived from
verbs of the two classes realize the semantic participants of the events they
denote.
3. Psychological nominais and their syntactic complements

In French, nominalizations can be derived from verbs by means of
affixation of different nominalizing affixes, such as -ation or -ement, as
exemplified in (6a) and (6b). In addition, nominalizations can be derived without
affixation, as in (6c). Finally, we find non-derived psychological nominals, as in
(6d).
(6) a. admiration "admiration", from admirer "admire"

b. étonnement "astonishment", from étonner "astonish"
c. regret "regret", from regretter "regret"

d. intérêt "interest", from Latin interest "it is important"
A crucial difference between verbs and derived nominalizations is that for

verbs, syntactic realization of semantic participants is obligatory, while for
nominalizations (as well as for non-derived nominais) it is optional. Therefore, it
is controversial whether the syntactic complements of nominalizations
corresponding to arguments of the base verbs should be termed arguments too, or
whether they should rather be classified as adjuncts (e.g., Dowty 1989;
Grimshaw 1990; Zucchi 1993:135; Alexiadou 2001:66). This question is not
addressed here; when referring to phrases like de Max ('de-phrase') or pour sa
faute ('pour-phrase'), the term 'complement' is used. Likewise, we do not speak
of arguments of lexemes as bearing thematic roles, but we do speak of semantic
participants of events denoted by lexemes as bearing the semantic or conceptual
roles 'experiencing entity' and 'target or cause of the experience'. We then say
that the complements of derived nominalizations realize the semantic
participants of the event denoted by the nominalization, or that they refer to the
semantic participants.
In French, nominalizations derived from psychological verbs basically
dispose of three types of syntactic constructions to realize semantic participants
in the event denoted: prepositional phrases headed by de "of', prepositional
phrases headed by prepositions other than de, such as par "by" and pour "for",
and possessive determiners like son, sa "his, her, its". The distribution of
possessive determiners is not discussed here (cf. Zubizarreta 1987:68).
First, de-phrases can realize both the experiencing entity and the target or
cause of the experience, (7b) and (7c). The sentence given in (7a) serves to
disambiguate the conceptual roles of the two semantic participants.
(7) a. Max regrette sa faute.

"Max regrets his mistake."
b. le regret de Max
"the regret of Max"
 le regret de la faute
"the regret of the mistake"
As can be seen in (8), the same holds for non-derived psychological nominais
like intérêt "interest". Again, the sentence given in (8a) is intended to
disambiguate the conceptual roles.
(8) a. Nicolas s'intéresse aux détails.

"Nicolas is interested in the details."
b. l 'intérêt de Nicolas
"the interest of Nicolas"
 l'intérêt des détails
"the interest of the details"
Second, par-phrases can occur only with EO-verbs, as in (9a, b), but not
with ES-verbs, (9c). Where a par-phrase is grammatical, it refers to the target or
cause of the experience; it cannot refer to the experiencing entity.
(9) a. le recours à des techniques de fascination par l'image (Le

Monde)
"the recourse to techniques of fascination by images"
b. la déception par le Gauìt Millau (WWW)
"the deception by the Gault Millau"
c. #l'admiration de cet auteur par Natalie
"the admiration of this author by Natalie"
With a number of nominalizations derived from EO-verbs, a par-phrase is

considered ungrammatical by some speakers. Some derived psychological
nominalizations, however, do allow a par-phrase. Among those nominalizations
for which I have found corpus1 examples with a par-phrase are affliction
"affliction", consolation "comfort", déception "deception", désappointement
"disappointment", fascination "fascination", satisfaction "satisfaction",
stupéfaction "stupefaction". Nominalizations derived without affixation, like
réconfort "comfort", as well as non-derived psychological nominais, like chagrin
"grieve", and deadjectival psychological nominalizations, like inquiétude
"anxiety", do not occur with par-phrases.
Third, prepositional phrases headed by pour occur primarily with
nominalizations derived from ES-verbs. Pour-phrases refer to the target or cause
of the experience, as in (10a), but not to the experiencing entity, (10b). With
(most) EO-verbs, they can realize neither participant, (10c) and (10d).2
(10) a. l'admiration du lecteur pour ce livre

"the admiration of the reader for this book"
b. # l'M'admiration de ce livre pour le lecteur
"the admiration of this book for the reader"
1
Using the database Frantext (http://zeus.inalf.cnrs.fr/frantext.htm), sequences of the type
'nominalization' + par were searched in texts dating from 1900 to 2000. In addition, the relevant
sequences were searched in the WWW, using the search engine Google (http://www.google.fr).
2
Still, one does find phrases like la fascination de ce livre pour les lecteurs contemporains "the
fascination of this book for contemporary readers". Examples of this type are discussed in section 6.
c. # amusement des enfants pour ce jeu

"the amusement of the children for this play"
d. 'amusement de ce jeu pour les enfants
"the amusement of this play for the children"
Pow-phrases are also found with non-deverbal psychological nomináis; they

likewise refer to the target or cause of the experience, but not to the experiencing
entity, see (11).
(11) a. Nicolas s'intéresse aux détails.

"Nicolas is interested in the details."
b. #/ 'intérêt pour Nicolas
"the interest for Nicolas"
 l'intérêt pour les détails
"the interest for the details"
Interestingly, ¿/e-phrases do not show the same distribution with respect to

all classes of psychological nominalizations, as exemplified by (1) and (2) above.
De-phrase complements of ES-nominalizations can realize either the experiencing
entity, as in (7b) above, or the target or cause of the experience, (7c) above. The
same is true for ¿fe-phrase complements of some EO-nominaiizations, like
fascination. In (12b), the ¿/e-phrase refers to the experiencing entity; in (12c), it
refers to the target or cause of the experience.
(12) a. L'orateur a fasciné l'auditoire.

"The speaker fascinated the audience."
b. La fascination de l'auditoire était facile.
"The fascination of the audience was easy."
 La fascination de l'orateur m'a surpris.
"The fascination of the speaker surprised me."
Other EO-verbs, however, like déception, do not allow ¿fe-phrases to refer to the
target or cause of the experience, (13c); rather, they permit only ¿fe-phrases that
refer to the experiencing entity, as in (13b).°
3
Corpus analyses have shown that one does find examples like (27a, b) below, where the ¿fe-phrase
refers to a participant which might at first glance appear to be like a target or cause of the experience.
In such expressions, however, déception de χ "deception of χ" has rather the meaning of χ est une
déception "χ is a deception" or χ est ce qui déçoit "χ is what deceives" than the meaning of χ a déçu
"χ has deceived". While these cases certainly merit a closer analysis, they are clearly distinct from
the examples studied in this paper.
(13) a. La réalité a déçu la jeune fille.

"Reality deceived the young girl."
b. La déception de la jeune fille m'a surpris.
"The deception of the young girl surprised me."
c. #La déception de la réalité m'a surpris.
"The deception of reality surprised me."
I have not found any underived psychological nouns sharing the properties oí
derived nominalizations like déception. Still, some deadjectival nominalizations
do present the same type of argument realization. Thus, the deadjectival noun
inquiétude "anxiety" occurs with a ¿fe-phrase realizing the experiencing entity,
(14b), but it cannot occur with a ^-phrase referring to the target or cause of the
experience, (14c).
(14) a. Ses paroles ont inquiété I 'enfant.

"His words troubled the child."
b. L'inquiétude de l'enfant (à propos de ses paroles) était évidente.
"The anxiety of the child (about his words) was evident."
c. #L 'inquietude de ses paroles m'a surpris.
"The anxiety of his words surprised me."
It is interesting to note that deception-nomimlizations behave like

nominalizations derived from canonical transitive verbs like destruction, in that
they show an effect of 'suppression' of the argument which is realized as subject
of the underlying verb (Grimshaw 1990:107), as exemplified in (15).
(15) a. the destruction of the city (by the enemy)

a', la destruction de la ville (par l'armée)
b. the enemy's destruction of the city
 #the destruction of the enemy [if enemy ~ subject of destroy]
c' #la destruction de l'armée [if armée ~ subject of détruire]
This effect of 'argument suppression' has led some researchers to describe

nominalization as an operation similar to passive formation (Grimshaw 1990), or
to conceive of nominalizations as ergative constructions (Williams 1987;
Alexiadou 2001). For the psychological nominalizations investigated here, we
(27) a. ce roman m 'exposait [...] à toutes les déceptions du réel (Yourcenar, 1931, Frantext)
"this novel exposed me [...] to all the deceptions of the real"
b. La déception de mon corps, j'en prenais mon parti. (Beauvoir, 1954, Frantext)
"The deception of my body, I came to terms with it."
find a minimal contrast in grammaticality, with one class of verbs allowing

realization of both participants in a ¿/e-phrase, and with the other class showing
'argument suppression' (Grimshaw 1990). At the same time, the underlying
verbs present no differences in surface syntax, with both classes realizing the
experiencing entity as the direct object. In the following, we propose to show
that this contrast in grammaticality can be related to semantic differences of the
underlying verbs. To this aim, in the next section we propose that two classes of
EO-verbs can be distinguished in French on the basis of aspectual criteria.
4. Aspectual features of psychological verbs

The two classes of EO-verbs distinguished in the previous section present
differences with respect to their compatibility with aspectual adverbials (cf.
Voorst 1992 for a more comprehensive account). Fasciner-verbs are compatible
with adverbs of duration like pendant longtemps "for a long time", as in (16a),
but they are less felicitous with adverbial expressions implying the existence of a
target state, like en peu de temps "in little time", see' (16b). Décevoir-vQvhs shov/
the opposite distribution; they are not felicitous with adverbs of duration, (16c),
but they are compatible with target state oriented expressions, (16d).
(16) a. Le livre Γα fasciné pendant longtemps.

"The book fascinated him for a long time."
b. #Le livre I ''a fasciné en peu de temps.
"The book fascinated him in little time."
 #Le livre  déçu pendant longtemps.
"The book deceived him for a long time."
d. Le livre a déçu en peu de temps.
"The book deceived him in little time."
As can be seen from their distribution relative to adverbial expressions, fasciner-

verbs are not telic, i.e., they do not imply a target state, while décevoir-verbs are
telic, i.e. they do imply a target state (cf. Vendler 1967; Dowty 1979; Smith
1997). Hence, the syntactic difference between the two classes of EO-verbs is
correlated with an aspectual difference.
In fact, there are even aspectual differences between the ES and EO classes:
ES-verbs are stative, while EO-verbs are non-stative. That ES-verbs are Stative can
be inferred from their incompatibility with progressive constructions, e.g., with
the French paraphrase être en train de (Borillo 1988; Smith 1997), cf. (17).
(17) #Max est en train de regretter sa faute.

"Max is regretting his mistake."
EO-verbs, in contrast, are non-stative, since they are compatible with progressive
operators (unless they are telic and punctual, like décevoir), as in (18).
(18) a. L'acteur est en train de fasciner le public.

"The actor is fascinating the audience."
b. Le clown est en train d'amuser les enfants.
"The clown is amusing the children."
Hence, we distinguish three aspectual classes of psychological verbs in

French. First, ES-verbs are stative. Besides regretter, verbs like admirer
"admire", estimer "estimate" belong to this class. Second, some EO-verbs, i.e.,
fasciner-wQvbs, are non-stative and atelic. With derived nominalizations, both
semantic participants can be realized in a ¿/e-phrase. Besides fasciner, verbs like
dégoûter "disgust", humilier "humiliate", tracasser "worry" belong to this class.
Verbs of this class denote psychological processes where the attitude or emotion
can be ascribed to the experiencing entity only in the presence of the target or
cause of the experience. Third, other EO-verbs, i.e., décevoir-verbs, are non-
stative and telic. With derived nominalizations, only one of the semantic
participants can be realized in a ¿/e-phrase. Besides décevoir, many deadjectival
psychological verbs, e.g. attrister "sadden", contenter "content", as well as many
punctual psychological verbs, e.g. étonner "astonish", émerveiller "amaze",
surprendre "surprise" belong to this class. Verbs of this class denote
psychological events where the attitude or emotion can be ascribed to the
experiencing entity not only in the presence, but also in the absence of the target
or cause of the experience. Finally, note that we find in French non-derived
psychological nouns whose participant-realizing properties correspond to
nominalizations derived from regretter-\Qrbs and to fasciner-vQvbs, while we
find only deadjectival psychological nominalizations sharing the participant-
realizing properties of nominalizations derived from décevoir-verbs. The next
section proposes semantic representations for the three classes of psychological
verbs. In section 6, it is shown that, drawing on three additional realization rules,
the correct distribution of the different types of complements to psychological
nominalizations can be derived from the semantic representations proposed for
the verbs.
5. Semantic representations for psychological verbs

It is a commonly held assumption in lexical semantics that the meaning of a
lexeme can be decomposed into atomic predicates. One of the basic claims of
such an approach is that verbs sharing certain atomic predicates also share some
of their grammatical properties. Hence, the decompositional semantic represen-
tation of a lexeme is claimed to encode not only the meaning, but also some of
the grammatical properties of this lexeme. Within a decompositional framework,
the analysis presented here aims at deriving the argument-realizing properties of
three different classes of psychological nominalizations from the meaning of the
underlying verbs. To this end, I propose that the meanings of the three verb
classes from which the nominalizations are derived can be represented as in (19).
(19) a. regretter PSYCH_RELATlON (χ, y)

b. fasciner CAUSE (ACT (y), PSYCH_RELATION (x, y))
 étonner CAUSE (ACT (y), CHANGE (PSYCH_STATE (x))) 4
The decompositional representations are intended to be interpreted in terms

of the paraphrases given in (20).
(20) a. "If χ regrets y, then χ stands in a psychological relation to y."

b. "If y fascinates x, then some action of y causes χ to stand in a
psychological relation to y."
 "If  astonishes x, then some action of  causes χ to come to be
in a psychological state."
Note that, since we are interested here in expressing generalizations concerning

the relation between semantic participants and syntactic complements, but not in
properties of individual lexical items, the representations in (19) consist entirely
of highly general atomic predicates like CAUSE, ACT, PSYCH_STATE, termed 'base
predicates', abstracting away from semantic differences which do not appear to
be reflected in the syntactic realization of semantic participants.
The base predicates in (19) are intended as representations of how the
arguments of a lexeme participate in the event denoted by the lexeme. In other
words, the conceptual roles assigned to the semantic participants of the event
denoted by the lexeme can be inferred from the base predicates (cf. Jackendoff
1987). More precisely, it is assumed here that the first participant of a
psychological relation (PSYCH_RELATION) is an experiencing entity, and the
second participant of a psychological relation is the target of the experience; cf.
(21a). Likewise, the (single) participant of a psychological state (PSYCH_STATE)
is an experiencing entity; cf. (21b). The (single) participant of an action (ACT) is
an agent; cf. (21c).
4
For this class of psychological verbs, similar representations have been proposed by Wanner (2001)
and Iwata (1995). The assumption that EO-verbs are causative is also made in some syntactic
accounts of psychological verbs (e.g., Pesetsky 1995).
Note that at this point, we can dissociate the roles 'target of the experience'
and 'cause of the experience', which have been conflated throughout the
previous paragraphs as 'target or cause of the experience'. By 'cause of the
experience', we denote the participant which is introduced as an argument of the
predicate standing in a causal relation to a a psychological state or relation, cf.
(2 ld).
(21) a. PSYCH_RELATiON (experiencing entity, target of the experience)

b. PSYCH_STATE (experiencing entity)
 ACT (agent)
d. 'cause of the experience': participant causing a psychological
state or relation
Finally, we assume by convention that psychological relations (represented

by PSYCH_RELATlON) and psychological states (represented by PSYCH_STATE) are
states, i.e., they are stative; actions (represented by ACT) are processes, i.e., they
are non-stative and atelic; changes of state (represented by CHANGE
(PSYCH_STATE)) are events, i.e., they are non-stative and telic (cf. Mourelatos
1978; Smith 1997). Let us further stipulate that in the case of complex
predications, only the aspectual type of the aspectually highest predicate is
projected onto the complex predication, where the hierarchy is
'event > process > state' (cf. Ehrich & Rapp 2000), with '>' meaning "is
aspectually higher than". Under this assumption, verbs with a semantic structure
as in (19a), i.e., ES-verbs, denote states. Verbs with a semantic structure as in
(19b), i.e., EO-fasciner-verhs, denote processes, and verbs with a semantic
structure as in (19c), i.e., EO-décevoir-verbs, denote events.
To conclude, the representations given in (19) for the three different classes
of verbs encode the aspectual type of the denoted event as well as the conceptual
roles assigned to the participants of the denoted event. In the next section, we
will see that, drawing on three additional participant-realization rules, we can
derive the syntactic realization of the semantic participants for the derived
nominalizations from the representations in (19).
6. Participant-realization rules f or psychological nominalizations

Section 3 above described which types of prepositional phrases (¿fe-phrase,
pour-phrasQ, /-phrase) occur with which class of psychological
nominalizations (derived from ES-verbs and two types of EO-verbs), and to
which semantic participants (experiencing entity, target or cause of the
experience) these can refer. The present section aims at formulating three
participant-realization rules by means of which the descriptive generalizations
stated in section 3 can be derived from the semantic representations given in
section 5. In particular, we want to derive the differing behavior of EO-verbs of

the fasciner-type and of EO-verbs of the décevoir-type with regard to which
semantic participant can be realized in a ¿/e-phrase. For convenience, the
semantic representations of the verbs are repeated in (22).
(22) a. ES regretter PSYCH_RELATION (χ, y)

b. EO fasciner CAUSE (ACT (y), PSYCH_RELATION (x, y))
 EO décevoir CAUSE (ACT (y), CHANGE (PSYCH_STATE (x)))
As to the ¿/e-phrase, regreí-nominalizations, as well as fascination-

nominalizations, can realize both participants; experiencing entity and target of
the experience. Dece/rt/cw-nominalizations, in contrast, can realize only the
experiencing entity, i.e., the participant over which the resulting psychological
state is predicated.
According to the semantic representations proposed here, the semantic
forms of fasciner-WQvhs and décevoir-verbs have identical first elements; the
structure of the first element should therefore play no role. Furthermore,
regretter-verbs and fasciner-verbs (which behave alike) have identical second
elements, while fasciner-verbs and décevoir-verbs (which show the contrast)
differ with regard to the second element of the predication. Hence, the relevant
difference should reside in the second element of the complex predications. More
precisely, for fasciner-verbs (as well as for regretter-verb's), the second element
contains two arguments, while for décevoir-verbs, the second element contains
but a single argument. In other words, for verbs of the fasciner-type, the state
which is caused is transitive, i.e., it is a psychological relation, while for verbs of
the décevoir-type, it is intransitive, i.e., it is a psychological state. The
generalization which can be drawn is then that the semantic participants which
can be realized by the de-phrase are those which appear in the second, or
rightmost, element of the complex predication. If this element is transitive, both
arguments can be realized in a ¿fe-phrase; if it is intransitive, only one argument
can be realized in a de-phrase. Hence, if the interpretation of ¿fe-phrase
complements of déverbal nominalizations is sensitive to which participants
appear as arguments of the second, or rightmost, element, the difference between
EO-nominalizations of the fascination-type and those of the déception-type
receives a straightforward interpretation. We therefore propose that the rule in
(23) regulates how semantic participants are realized by ¿/e-phrase complements
of nominalizations.
(23) Rule 1 : De-phrases can refer to the semantic participants introduced

by the rightmost predicate in the semantic form.
Note that this is a 'positive' formulation (cf. Markantonatou 1995 for a similar
proposal) of a previously made 'negative' generalization about nominalizations
of the destruction-type, cf. (15) above. This generalization has previously been
referred to as 'suppression' of the external argument (i.e., of the agent or causer),
(Grimshaw 1990; Alexiadou 2001), and it has also been termed 'causer deletion'
(Iwata 1995). Hence, previous approaches have argued that the agent or causer,
i.e., the semantic participant of the leftmost predicate, cannot be realized by of-
phrases in English, corresponding to French ¿fe-phrases, because it is suppressed
or deleted from the argument structure. The present proposal, in contrast, argues
that a semantic participant which has the conceptual role of agent or causer can
be realized in a ¿fe-phrase in French, but only if it has, in addition, a role to play
in the rightmost predicate. Thus, the argument 'suppression' effect found for
destruction-nomimlizations and déception-nommalizaXions, but not for
fascination-nomimlizations, is claimed to arise from the participant-realization
rule in (23).
The second generalization drawn in section 3 stated that prepositional
phrases headed by par can be interpreted only as referring to semantic
participants of events denoted by EO-verbs. In addition, it was said that a par-
phrase can refer only to what was termed the target or cause of the experience,
but not to the experiencing entity. Looking at the semantic representations in
(22), we see that EO-verbs, i.e., fasciner-verbs and décevoir-vevbs, share their
leftmost element, i.e., ACT, the presence or absence of which furthermore
distinguishes EO-verbs from ES-verbs. Therefore, we assume that it is the
semantic participant introduced by ACT, i.e., the agent, which can be realized in a
par-phrase. This generalization is formulated in (24).
(24) Rule 2 : Par-phrases can refer to the semantic participant introduced

by ACT, i.e., to the agent.
Finally, prepositional phrases headed by pour were said to occur only with
ES-nominalizations, but not with EO-nominalizations; they can refer to the target
of the experience, but not to the experiencing entity. This observation, however,
immediately raises the question why for EO-nominalizations the target of the
experience cannot be realized in a pour-phrase. What is then the relevant
difference between ES-verbs and EO-verbs of the fas einer-type"] One difference
consists in the aspectual type of the verbs: ES-verbs are stative, while EO-
fasciner-verbs are non-stative.
Evidence that stativity is indeed a factor governing the distribution of pour-
phrases can be gained from the examples in (25), showing that/?owr-phrases are
grammatical with nominalizations of the fascination-type, but only under certain
circumstances.
(25) a. la fascination progressive de ce livre /par ce livre

"the growing fascination of this book / by this book"
b. Γ énorme fascination de ce livre /'#par ce livre
"the enormous fascination of this book / by this book"
c. une énorme fascination pour ce livre
"an enormous fascination for this book"
As shown by (25a), fascination is a process nominalization (as demonstrated by

the adjective progressif "growing"); it is compatible with a ¿/e-phrase or par-
phrase. (25b) shows that fascination can have a second interpretation: it can
occur with the degree-modifying adjective énorme "enormous", which we
assume to modify (stative) qualities, but not (non-stative) processes. Thus, when
fascination has a stative interpretation, as in (25b), the č/e-phrase is still possible,
but the par-phrase appears to be less felicitous. Importantly, with a stative
interpretation, fascination is compatible also with a /r-phrase, as in (25c). We
therefore propose that the rule in (26) regulates the interpretation of pour-
phrases.
(26) Rule 3: Pour-phrases can refer to the second semantic participant

introduced by PSYCH_RELATlON, i.e., to the target of experience, of
stative predicates.
To conclude, this section has shown that the distribution and interpretation
of syntactic complements of psychological nominalizations can be derived from
the semantic representations introduced in section 5, by drawing on three
additional rules governing the realization of semantic participants by three types
of syntactic complements. The lexical representations and the participant-
realization rules proposed can thus not only account for the grammaticality
contrast exemplified in (1) and (2), but also for the distribution of /?ar-phrases
and-phrases with nominalizations derived from psychological verbs.
7. Conclusion
The paper started with the question how derived nominalizations realize
their semantic participants by means of syntactic complements. The discussion
has shown that, given appropriate semantic representations of the base verbs
from which information about conceptual roles of semantic participants can be
gained, the syntactic realization can be derived with three realization rules. A
central concern was to explain the grammaticality contrast found for two
different classes of derived nominalizations with regard to č/e-phrase
complements. It was shown that the contrast can be related to the semantic
structure of the underlying verbs, assuming that only certain substructures of a
complex predication are accessible for participant realization in ¿fe-phrases.

Here, it would be worthwhile to ask whether this proposal can in some way be
related to existing models of participant realization in verbs. As to the other two
types of syntactic complements discussed, par-phrases and /?oz/r-phrases, it was
argued that they are subject to realization rules sensitive to particular conceptual
roles of participants and aspectual features of the underlying verbs. It would be
interesting to see whether these rules can be related to a more general account of
the semantics of the prepositions p a r mdpour. In sum, the discussion has shown
that participant realization in nominalizations is governed by rules different from
those regulating participant realization in verbs. This assumption receives further
support from the observation that non-derived psychological nomináis as well as
deadjectival psychological nominalizations realize participants in a similar way
as déverbal nominalizations.
References
Alexiadou, Artemis. 2001. Functional Structure in Nomináis. Nominalization and
Ergativity. Amsterdam & Philadelphia: John Benjamins.
Belletti, Adriana & Luigi Rizzi. 1988. "Psych-Verbs and Θ-Theory". Natural Language
and Linguistic Theory 6.297-352.
Borillo, Andrée. 1988. "L'expression de la durée : Construction des noms et des verbes de
mesure temporelle''. Lingvisticae Investigationes 12.363-396.
Dowty, David R. 1979. Word Meaning and Montague Grammar. Dordrecht: Reidel.
- 1989. "On the Semantic Content of the Notion 'Thematic Role'". Properties,
Types and Meanings ed. by Gennaro Chierchia, Barbara H. Partee & Raymond
Turner, 69-129. Dordrecht: Kluwer.
Ehrich, Veronika & Irene Rapp. 2000. "Sortale Bedeutung und Argumentstruktur: ung-
Nominalisierungen im Deutschen". Zeitschrift für Sprachwissenschaft 19:2.245-
303.
Frantext. http ://zeus. inalf. cnrs. fr/frantext. htm
Grimshaw, Jane. 1990. Argument Structure. Cambridge, Mass.: MIT Press.
Iwata, Seizi. 1995. "The Distinctive Character of Psych-Verbs as Causatives". Linguistic
Analysis 129.95-120.
Jackendoff, Ray. 1987. "The Status of Thematic Relations in Linguistic Theory".
L ingu is tic Inquiry 18.369-411.
Markantonatou, Stella. 1995. "Modern Greek Déverbal Nomináis: An LMT Approach".
Journal of Linguistics 31.267-299.
Mourelatos, Alexander P. 1978. "Events, Processes and States". Linguistics and
Philosophy 2.415-434.
Pesetsky, David. 1995. Zero Syntax. Experiencers and Cascades. Cambridge, Mass. &
London: MIT Press.
Smith, Carlota S. 1997. The Parameter ofAspect. Second edition. Dordrecht: Kluwer.
Vendler, Zeno. 1967. Linguistics in Philosophy. Ithaca, N.Y.: Cornell University Press.
Voorst, Jan van. 1992. "The Aspectual Semantics of Psychological Verbs". Linguistics &
Philosophy 15:1.65-92.
Wanner, Anja. 2001. "The Optimal Linking of Arguments: The Case of English
Psychological Verbs". Competition in Syntax ed. by Gereon Müller & Wolfgang
Sternefeld, 377- 399. Berlin: Mouton de Gruyter.
Williams, Edwin. 1987. "English as an Ergative Language: The Theta Structure of
Derived Nouns". Papers from the 23rd Annual Regional Meeting of Chicago
Linguistics Society 23:1.366-375.
Zubizarreta, Maria Luisa. 1987. Levels of Representations in the Lexicon and in the
Syntax. Dordrecht: Foris.
Zucchi, Alessandro. 1993. The Language of Propositions andEvents. Dordrecht: Kluwer.
NOTES ON VOCATIVE CASE
A CASE STUDY IN CLAUSE STRUCTURE*
ANDREA MORO
Università "Vita - Salute " San Raffaele, Milano
1. Vocative Case in modern syntax: two questions

One of the major achievements of modern syntax is that all the different
relations that are standardly referred to as 'Case assignments' have been unified
under a single syntactic relation, namely a specifier - head relation holding
between a noun phrase and a proper functional head (see Chomsky 1995 and
references cited there). In this paper I would like to address some questions
concerning the nature of the so called 'Vocative Case' which naturally arise
under such perspective. As a premise, let me notice that the literature concerning
Vocative Case is unusually scarce, especially if compared with the literature
concerning other Cases. I think that this is not accidental, for the very existence
of a specific morphological Vocative Case has been often denied as will be
shown on the base of some citations from classical works in the field.
In this paper I will focus on two conceptually distinct questions:
(1) a. What is the internal structure of a phrase assigned a Vocative

Case?
b. What kind of structural environment does Vocative Case
assignment require?
This paper is organized as follows: in section 2. diagnostics for Vocative Case

will be discussed attempting to approach question (la), including references to
classical works; in section 3 we will focus on the distribution of Vocative Case in
the left periphery, addressing the issue posed in question (lb). It goes without
saying that the present work can only be considered as a preliminary attempt to
approach the relatively poorly understood phenomena related to Vocative Case
and does not by any means aim at reaching definitive conclusions on the issue;
* Progressive versions of this paper have been presented at the "Incontro di Grammatica Generativa"
in Siena in 1999 and at "Going Romance 2001" at the University of Amsterdam. I am indebted to the
audience of these conferences for many helpful comments. Special thanks to Guglielmo Cinque,
Giorgio Graffi, Giuseppe Longobardi, Luigi Rizzi, Franc Floricic and two anonymous reviewers for
their comments.
248 ANDREA MORO
rather, the goal of this paper is to address some interesting questions and possibly
refute some unadequate answers.
2. Diagnostics f or Vocative Case

Is there a Vocative Case? The question is not trivial. Let us preliminary
clarify the situation from a terminological point of view. We can distinguish
between a 'Vocative Case' and a 'Vocative Phrase'. Vocative Case is a(n overt)
morphological mark assigned to a Vocative Phrase;1 a Vocative Phrase, instead,
is a noun phrase which does not belong to the thematic grid of a predicate and is
used to attract someone's attention, in a broad sense. More precisely, from a
purely denotational point of view, one can distinguish at least two types of
Vocative Phrases depending on whether they refer to an entity which is not
referred to in the thematic grid of the predicate (call it 'Extradeictic Vocative') or
to an entity which is referred to in the thematic grid of the predicate by means of
a pronoun (call it 'Infradeictic Vocative'), being it the subject, the direct or the
indirect object:2
1
As usual, I will assume that morphological Case must not necessarily be overt and that in
languanges like Italian, morphological Case is overtly manifested only in the pronominal system. The
pronominal system is partially sensitive to Vocative Case: for example, when first person pronouns
are used as Vocative Phrases they cannot be assigned Nominative Case and they show the Accusative
or Default Case:
(i)  povero me /*io, dove posso andaré?
 poor I-ACC-DEF/I-NOM, where can I go
With second person pronouns the situation is different, though, since in that case Nominative can
show up as in:
(ii)  tu /te che muovi la ruota...
 you-NOM/you-Acc. who move the wing..
2
I will not discuss here the important issue concerning the type of relation between the noun phrase
and the pronoun which corefers with it (whether it is Binding or independent coreference). Notice
that non trivial coreference phenomena can also be detected by exemples like:
(i) .  mio/* suo/* tuo re, Gianni vuole un cavallo.
 my/ his/ your king, John wants a horse
b.  testimoni del sno arr ivo, Gianni è il nostro re.
 witnesses of his coming Gianni is our king
First person pronoun is selected in (ia) as if the clause containing a Vocative Case had an implicit
second person argument which the speaker obligatory refers to (leaving courtesy expressions like
vostra altezza ( "your highness") aside).
As for the nature of the coreference relation, it interesting to notice that Vocative Phrases don't give
rise to Binding Opacity or Relativized Minimality Phenomena, witness examples like the following:
(ii) a. O pover a Maria į, i ragazzi non la¡/j /ti į/*j aiutano.
o poor Maria the boys not her / you help
b. [Quale infermiera] į credi, [Maria]į, che i ragazzi denuncino t¡ ?
which nurse you think Mary that the boys impeach
 [In che modo] credi, Mariay. che i ragazzi denuncino que sta infermiera t¡ ?
in what way you think o Mary that the boys impeach this nurse
NOTES ON VOCATIVE CASE 249
(2) a.  Gianni, Maria sta abbracciando Pietro! Extradeictic Voc.

 Gianni Maria is hugging Peter
b.  Gianni,, pro, colpisci Pietro! Infradeictic (Subj.)Voc.
 Gianni, pro hit Peter
c.  Gianni j, Maria vuole abbracciare teh ,/luihhl
 Gianni Maria wants to hug you/him
Infradeictic (Dir. Obj.) Voc.
d. O Giannij, Maria vuole dare un libro a teh, /luih ,!
 Gianni Maria wants to give a book to you/him
Infradeictic (Ind. Obj.) Voc.
This, of course, is not sufficient to answer the question whether there is a

Vocative Case. In fact, it could well be that the noun phrase involved here is
assigned another Case (or, perhaps, no Case at all). Of course, one important
piece of evidence that there is indeed a Vocative Case, in fact the very motivation
for this discussion, comes from those languages where such a noun phrase which
does not belong to the thematic grid of the predicate is morphologically marked
by a distinct inflectional ending. A classic example is Latin:
(3) a. Tityre tu patulae recubans sub tegmine

Tityr-Voc. you-NOM. leafy-GEN. laying under cover-ABL
f agi-
beech-GEN
"Tityrus, you laying under the cover of a leafy beech..."
The proper name Tityrus is put in its Vocative Case {Tityre) at the very beginning
of the sentence, to attract the attention of Tytirus. Interestingly, such an overt
piece of evidence has often been considered as insufficient to conclude that there
be a Vocative Case, although it can hardly be denied that there is a Vocative
Phrase. In fact, many influential scholars have proposed that Vocative Case is not
a 'real' Case. The following citations illustrate such a rather murky situation:
(4) a. "de même qu'un appel échappe au contexte de la phrase, de

même le vocatif se trouvait en dehors de la flexion. Sa forme
était celle du thème nu (gr. koure, lat. domine) comme
l'impératif, ou de la racine au degré réduit (gr. pater). Isolé dans
la déclinaison, il devait disparaître, remplacé par le nominatif."
(Ernout-Thomas 19532: 14)
250 ANDREA MORO
b. "un cinquième cas, le vocatif, que les Stoïciens [...] ne

comptaient pas plus parmi les cas que les grammairiens indiens
qui, n'étant pas dupes d'une forme qui le rapproche si
fréquemment du nominatif, le laissaient en dehors de la flexion.
[...] le vocatif n'existe pas dans des catégories aussi importantes
que les pronoms personnels, les adjectifs possessifs, les adjectifs
et pronoms démonstratifs." (Humbert 19542: 247; 294)
c. "L'histoire du vocatif, d'Homère à la Koinè, est liée à
l'extraordinaire développement qu'a pris l'interjection  [...]. Elle
a fini par s'associer étroitement au vocatif et par perdre son
autonomie." (Humbert 19542: 295)
d. "En nostre Langue, et dans les autres vulgaires, ce cas s'exprime
dans les noms communs qui ont un article au nominatif, par la
suppression de cet article. Le Seigneur est mon esperance.
Seigneur vous estes mon esperance." (La grammaire de Port-
Royal; p. 44, in Brekle (éd.) 1966)
All in all, despite the difficulties to agree on the existence of a separate

morphological Case for Vocative, three purely syntactic factors emerge here that
correlate with a Vocative Phrase: the Vocative Phrase does not belong to the
thematic grid of the main predicate of the clause, the Vocative Phrase may not
co-occur with an article, the Vocative Phrase may be preceded by an emphatic
interjection. Clearly, these are quite weak diagnostics to conclude straightfor
wardly that there be a Vocative Case. Nevertheless, they clearly show that
Vocative Phrases behave differently from argumentai noun phrases and can be
fruitfully used to explore and refine our understanding of the syntax of Vocative
Phrases. Remember that the underlying issue we are concentrating on here is that
if there is a Vocative Case one must find a head which assigns it to the Vocative
Phrase it is associated with, if one wants to maintain the unified theory of Case
assignment that is currently adopted.
One important issue concerning the identification of a Vocative Case from a
syntactic point of view is that the role of the absence/presence of the article is not
uniform across and within languages nor diachronically stable. Consider Italian,
for example (the presence of the interjection ais irrelevant here):
(5) a .  (*il /un) ragazzo/ (*il) Gianni, la Maria è partita.

 the /a boy / Gianni, the Maria has left
b·  [DP donna / mia t]], Maria è partita.
 woman my Maria has left
c. Ό quel giovine [...] per i vostripoveri morti, fate la carita

 that young man, for the your poor dead...
d'andaré... ' (A. Manzoni, I promessi sposi, XXXIV)
d. 'Ditemi, quel giovine, al vostro paese che regola si usa?"
tell me, young man, in the your country what rule governs
(C. Goldoni, Lefemminepuntigliose; in Mazzoleni 1995)
e. ?0 il mi o caro ragazzo, es ci di qui!
o the my dear boy, get out of here
f. O caro il mió ragazzo, es ci di qui.
g. O (povero) me/*io, Maria è partita.
 poor me/ I, Maria has left
h. O tu/te, che non sai nulla...-
o you, who not know anything..
Prima facie, the example in (5a) shows that the definite and indefinite article
must be omitted in the Vocative phrase in Italian, even in those varieties of
Northern Italian which allow the article to precede the proper name. A simple
minded solution could be that the syntactic position where articles occur in a
noun phrase, namely D°, must be omitted. This conclusion, however, is not
correct for several reasons. First, adopting Longobardi's 1994 theory of №-to-D°
raising, (5b) shows that such a position can be realized, witness raising of donna
"woman" over the possessive adjective mia "my"; second, in XIX century
Italian examples such as those in (5c) and (5d) (the latter reproduced from
Mazzoleni 1995) D° can be clearly realized and occupied by a demonstrative;
third, if an emphatic adjective like caro "dear" precedes the noun phrase, the
definite article can easily occur with the Vocative Phrase as shown in (5f); fourth,
pronouns, which arguably occupy a D° position as standardly assumed, can
occur as Vocative Phrases as in (5g-h), whether or not an emphatic adjective co-
occurs with it. All in all, one cannot conclude that the absence of a D° projection
be a diagnostics for Vocative case. Rather, it must be the case that some different
reason explains the absence of the article with Vocative Phrase, possibly related
to the referential capacities of the noun phrase involved. In fact, notice that the
only other case where the definite article is impossible with proper names, of
course in those varieties which allow proper names to occur with articles such as
in many Northern Italian varieties, is the case where the proper name plays the
role of a predicate such as in:3
3
1 will not consider here the predicative use of proper names in copular sentences; see Moro 1997 for
a detailed discussion.
252 ANDREA MORO
(6) a. Il Gianni ¡o chiama (*il) Mario.

the Gianni him calls the Mario
b. Mi chiamo (*il) Mario.
me call the Mario
"My name is Mario."
These examples suggest that the reason why the article is absent most arguably
depends on the referential capacities of the noun phrase involved and, crucially,
is not specific to Vocative Case. Thus, the absence of the article cannot be used
as a diagnostics.4
Interesting facts also correlate with Vocative Phrases on phonological
grounds. This has been noticed in different domains of analysis. Floricic (2000)
pointed out that in Southern Italian varieties proper nouns can occur in a
truncated form, such as Ante vs. Antonio. These truncated forms can only be
used as Vocative Phrases. Interestingly, Floricic noticed that truncation can also
occur with verbs but only in the imperative form and that the two (Vocative
Phrase and imperatives) can co-occur. The paradigm is as follows:
(7) a. Ante/Antonio, Lucl/Lucia, Robbè/Roberto

b. Lucl/Lucia, vieni qua!
Luci/Lucia come here
 Lucia/*Lucí viene qua!
Lucia/ Luci comes here
d. Tie VTieni ("Keep";, Gua '/Guarda ("See"), To 77bg//("Take this")
e. Lucia guarda/*gua' il panorama.
Lucia looks at the panorama
This paradigm not only shows that Vocative Phrases in fact behave differently
than other noun phrases in that they allow truncation, it also draws a parallelism
between Vocative Case and imperative whose co-occurrence is often attested.
Although Floricic's discussion of the various facets of this parallelism cannot be
reproduced here, it is at least worth emphasizing that the absence of the article in
Vocative Phrases is paralleled by the absence of negation in second person
imperatives, indirectly reinforcing our conjecture that the absence of the article
in Vocative Phrases is not specific to Case assignment but is rather the
4
French, for example, is interesting. Most grammars indicate that the definite article must be present
with plural noun phrases and absent with singular:
(i) (*Le) garçon/* (Les) garçons, Jean est arrivé.
 the boy / the boys, Jean is arrived
Nevertheless, some speakers do accept definite articles with singular. I am indebted to Richard Kayne
for having pointed this out to me.
consequence of a more general condition on interpretation : "Ce n'est pas un

hasard si la forme non canta ("Il ne chante pas") n'est possible que dans le plan
délocutoire, de même que le syntagme nominal i ragazzi dont nous avons vu
qu'il ne pouvait être interpreté comme vocatif [fh. omitted]" Floricic 2000: 261.
Similar considerations have been addressed by Lazzeroni 1995. Studying the
well-known phenomenon of stress retraction in Vocative Case morphology in
languages like Ancient Greek and Sanskrit, Lazzeroni concludes that "nelle
coppie diatoniche del greco e del sánscrito la baritonesi segnala il termine
caratterizzato dal tratto che occupa il posto più alto nella gerarchia
dell'individuazione: il nome proprio rispetto al nome comune, il sostantivo
rispetto all'aggettivo, il nome d'azione rispetto al nome d'agente (assimilato
all'aggettivo), l'agente individuato rispetto ail'agente genérico" (Lazzeroni 1995:
4-5). Typical examples could be the following:
(8) a. adelphos / adelphe

brother Nom. / brother Voc.
b. leukos /'Mukös
white - adjective / white thing
c. karpos / Kar p os
fruit / fruit proper name
d. tomos / tomos
cut / cutter
The conclusion Lazzeroni suggests is that retraction of the stress in Vocative

Case is not special: 'Tassegnazione dell'accento non è vincolata a categorie
semantiche  morfologiche specifiche, ma alia gerarchia di individuazione che
oppone i termini di una coppia" (Lazzeroni 1995:5). In both case studies
illustrated here the phonological phenomena affecting Vocative Phrases cannot
be related to the necessity to distinguish a special Case from the others; rather,
these phenomena (truncation and retraction of stress) are due to independent
factors, most arguably related to 'semantic' reasons, in the broad sense involving
denotation, predication, etc.
We can conclude this section exploring the possible diagnostics for
Vocative Case and its internal structure. We have seen that the situation is rather
unclear and can be summarized as follows. A Vocative Phrase is a full noun
phrase containing both № and D° projections which does not belong to the
thematic grid of the predicate, although it might be related to it by means of a
pronoun; a Vocative Phrase may display anomalous behavior both syntactically
(absence of the article, presence of an interjection which immediately precedes
the noun phrase, selective referential capacities) and phonologically (truncation
and stress retraction); a Vocative Phrase appears to be marked by a special
254 ANDREA MORO
inflection in some languages, although many scholars suggested that Vocative is

just a bare root form rather than a specific Case ending. All in all, even if the
very existence of a separate Case inflection cannot be undisputedly proved, still
the question as to whether there is a specific syntactic configuration where a
Vocative Phrase can occur can be properly raised. In the next section we will
approach such an independent issue.
3. On the distribution of Vocative Phrase in the left periphery

Where can a Vocative Phrase occur? To answer this question we will focus
on Italian data. Prototypically, Vocative Phrases occur at the very beginning or at
the very end of the sentence:
(9) a. (O) Maria, pro vieni qui!

 Maria pro come here
b. pro vieni qui, (o) Maria!
pro come here,  Maria
The obvious preliminary question then is what portion of the left periphery they
occupy, assuming the non-neutral assumption that (9b) is a derived form from
(9a) by remnant movement.5 To answer this, I will assume the split Comp field
analysis stemming from Rizzi's 1997 work (see also Cinque 1979) which can be
synthesized as follows:
3
To support this assumption consider the following case. In Central Italian varieties like Marchigiano
spoken in Fano, for example, it is possible to have a wh-word like perché "why" to precede or
follow the IP it refers to as in:
(i) a. Perché [sei venutoj?
why did you come
b. [Se i venutoj perché ?
That this process is syntactically governed, and not just 'stylistic', can be proved by showing that the
same type of 'inversion' cannot take place with come "how":
(ii) a. Come [sei venutoj?
how did you come
b. * [Se i venutoj come ?
Now, assuming that perché is generated in the left periphery (cf. Rizzi 1990), one can conclude that
(ib) is derived from (ia) with remnant movement of IP over the position where perché is generated.
Interestingly, a Vocative Phrase can occur at the end of the sentence only if perché precedes the
sentence:
(iii) a. Perché [sei venutoj Mario?
why did you come Mario
b. *[Sei venutoj perché Mario? / * ?[Sei venutoj Mario perché?
c. Mario [se i venutoj perché ?
This can be explained by assuming that remnant movement is already involved in generating 'perché
inversion', suggesting that the impossibility of a Vocative Phrase at the end of the sentence is due to
the fact that the same strategy cannot be exploited twice.
(10) C° = ... Force0 > (Top° > Foc0 > Top° >) Fin° ... (Rizzi 1997)
Different types of elements fill different positions in (41) [(10) here,
a.m.]. Straightforward distributional evidence suggests that relative
pronouns are in the spec of Force0, while interrogative pronouns in
main questions compete with focused phrases for the spec of Focus0.
Complementizers such as that, que, etc. are in Force0 (when the
topic-focus field is activated), while prepositional complementizers
in Romance are in Fin° (Rizzi 1997: 325).
Thus, the obvious step to take is to see if a Vocative Phrase can occupy any
of the split Comp field specifier positions. Let us start by considering Fin°, i.e.
the lowest head. Since Fin° can be activated in different contexts in Italian, we
can provide different tests, namely infinitival declaratives, absolute small clauses
and Aux-to-Comp constructions:
(11) a. Gianni dice, (o) Maria/(o) ragazza, di andaré a Roma.

Gianni says,  Maria /o girl, to go to Rome
b. * Gianni dice di, o Maria, andaré a Roma.
Gianni says to o Mary go to Rome
 *Con,  Maria, Gianni malato, Pietro non puô partiré.
with  Maria Gianni sick Pietro not can go
d.  Maria, con Gianni malato, Pietro non puô partiré.
 Maria with Gianni sick Pietro not can go
e. *Avendo, (o) Maria, Pietro letto il libro...
having  Maria Pietro read the book...
f. O Maria, avendo Pietro letto il libro...
 Maria, having Pietro read the book...
As each pair indicates, the Vocative Phrase must precede the material contained
in the Fin° head (i.e. di, con and avendo, respectively) indicating that it must
occupy a higher specifier position.6
6
This conclusion fits in with the following contrast in English where did occupies the Fin0 position,
as suggested in Rizzi 1997:
(i) a. Did, (*o Mary), John read the book?
b. O Mary, did John read the book?
This also correlates with the fact that a Vocative Phrase cannot occur between a wh-word and do but
in the case the wh-word is why, which we independently know is higher than the other wh-words (see
Rizzi 1990 and references cited there)
(ii) a. Why (John) did you do that?
b. What (*John) did you do today?
256 ANDREA MORO
Consider now a Top° head. In Italian, such a head is exploited in the so

called Clitic Left Dislocation constructions (CLLD) in the sense of Cinque 1990.
In such constructions involving a phrase in the specifier of Top°, a clitic is in fact
involved in Italian. Consider then the following cases:
(12) .  Maria, I ragazzi, li aiuta Gianni.

 Maria the boys them helps Gianni
b. ?I ragazzi,  Maria, li aiuta Gianni.
the boys  Maria them helps Gianni
c.  Maria, I ragazzi, (* o Pietro), li aiuta Gianni.
 Maria the boys  Pietro them helps Gianni
d. ?I libri,  Maria, in questo scaffale, ce li mette Gianni.
the books, o Mary, on this shelf, there-them puts Gianni
A Vocative Case is clearly compatible with a CLLD construction proving that it

does not compete with the same specifier position as Topic Phrases. Moreover,
there is clear preference for the Vocative Phrase to precede the Topic Phrase.
Interestingly, the paradigm in (12) allows a further refinement: since Top° heads
can occur recursively, one may ask whether Vocative Phrases can occur higher
than the lower or the higher Top° or both. The case in (12c) clearly indicates that
there cannot be two Vocative Phrases: a Vocative Phrase can be licensed only
once and only higher than the higher Top° head.
What about the Foc° head? Tests show that Focus constructions in Italian
are compatible with a Vocative Phrase and that the Vocative Phrase must occur
higher than Foc°:
(13) .  Maria, I RAGAZZI, Gianni aiuta, non i conigli.

0 Maria, the boys Gianni helps not the rabbits
b. */ RAGAZZI,  Maria, Gianni aiuta, non I conigli.
the boys,  Maria, Gianni will help not the rabbits
Moreover, a further test can be provided with respect to Foc°, since we know that
wh-phrases compete for the same position as Focus Phrases. Consider the
following cases:
(14) a. Mi chiedo,  Maria, quale libro leggi?

1 wonder  Maria which book read
Similarly, notice also that a Vocative Case cannot separate existential there from the copula: if so, the
sentence is interpreted as a locative sentence (cf. precopular here):
(iii) a. There (o Mary) is a solution to the theorem,
b. Here (o Mary) is a solution to the theorem.
b. ?*Mi chiedo quale libro, o Maria, leggi?

I wonder which book  Maria read
Vocative Phrases must precede the Foe0 head even in the case the specifier
position hosts wh-phrases such as quale libro "which book". Notice that there is
no intrinsic prohibition for a wh-word to precede a proper noun (or more
generally a noun phrase), witness the following examples involving CLLD and
wh-movement:
(15) a. Mi chiedo Maria quale ragazzo la sopporti.

I wonder Maria which boy her stands
b. Mi chiedo quale ragazzo Maria la sopporti.
I wonder which boy Maria her stands
 Mi chiedo quale negoziante il vino lo venda di Domenica.
I wonder which dealer the wine it sells on Sundays
Leaving Top° and Foc° heads, let us now make one step further up in the left
periphery, exploring the co-occurrence of Force0 with a Vocative Phrase.7 In fact,
the phrase stemming from Force0 can involve different elements. First, the
specifier of Force0 can host relative operators such as i quali "the who":
(16) a. * Coloro, o Maria, i quali sonó arrivati ieri partir anno

those  Maria the who are arrived yesterday will leave
domani.
tomorrow
b.  Maria, coloro i quali sono arrivati ieri partir anno
 Maria those the who are arrived yesterday will leave
domani.
tomorrow
7
The tests involving Top° and Foe0 heads suggest a further test involving Vocative Phrases. Consider
die following case:
(i) a. O povero/caro ragazzo, Gianni è troppo intelligente.
 poor /dear boy, Gianni is too intelligent
b. *0 nessun/ogni ragazzo, Gianni è troppo intelligente.
 no /every boy, Gianni is too intelligent
 Nessun/ogni ragazzo, Gianni (*lo) odia.
no /every boy Gianni him hates
Vocative Phrases appear to behave like Topic Phrases in that they cannot be quantificational, unlike
Focus Phrases.
258 ANDREA MORO
This contrast indicates that the Vocative Phrase must be higher than the relative
operator occupying the specifier position of Force0.8 Second, Force0 can be
realized as an overt complementizer such as declarative che "that":
(17) a. Gianni pensa, () Maria/() ragazza, che Pietro abbia letto un
Gianni thinks  Maria/  girl that Pietro has read a
libro.
book
b.. * Gianni pensa che, () Maria/() ragazza, Pietro abbia letto un
Gianni thinks that  Maria/  girl Pietro has read a
libro.
book
Also in such a case, the Vocative Phrase must precede Force0. Notice that there is
no independent prohibition for a noun phrase to occur lower than declarative che,
witness the possibility for a proper noun to be in such a position in a CLLD
construction like the following:
(18) a. * Gianni pensa Mario che lo vedrà domani.

Gianni thinks Mario that him will see tomorrow
b. Gianni pensa che Mario lo vedrà domani.
Gianni thinks that Mario him will see tomorrow
We have thus reached the leftmost head of the split Comp Field. There is but one
option left, namely that a Vocative Phrase occurs as a specifier of an independent
head assigning Vocative Case to it. In other words, we must extend the split
Comp field including (at least) one more head/feature: the Vocative Phrase is
hosted in the spec of the head projected by a Voc° feature governing Force0.
Formally, the split Comp Field must accordingly be increased as follows:
(19) C° = ... Voc° > Force0 > (Top° > Foc° > Top° >) Fin0 ...
Although this is prima facie not a very satisfactory solution, as all solutions
stipulating an ad hoc entity are, it seems to me that this proposal can be regarded
8
Notice that a Vocative Phrase cannot be expoited to yield a Verb Second construction such as:
(i) a. *  lieber Andreas habe ich endlich das Buch gelesen.
 dear Andreas have I eventually the book read
b. Gestern habe ich endlich das Buch gelesen.
yesterday have I eventually the book read
as less expensive on theoretical grounds once one explores the distribution of the
interjection o, which many classical authors considered a diagnostics for
Vocative Case. In the remaining of this section I will briefly explore the syntax
of such an element.
A priori there are two options for the syntax of o: either  is a functional
projection" belonging to the noun phrase (i.e. to the D°-system) or  is a
functional projection belonging to the clausal structure (i.e. to the C°-system).
How can we choose between the two hypotheses? We have already seen that D°
can be exploited in Vocative Phrases, even if the article for independent reasons
might not be compatible with a Vocative Phrase. In fact, there are good reasons
to assume that  does not compete for D°. A first piece of evidence comes from
the (rather trivial) fact that  cannot occur with other noun phrases, even if they
occur in the left periphery, such as in the CLLD construction in (2Od) or in the
Focus construction in (20e):
(20) .  Pietro, Gianni baciava Maria in giardino.

 Pietro, Gianni kissed Maria in the garden
b. * Gianni baciava  Maria in giardino.
Gianni kissed  Maria in the garden
c. *0 Pietro baciava Maria in giardino.
 Pietro kissed Maria in the garden
d. *  Pietro, Maria lo ama.
 Pietro Mary him loves
e.  PIETRO, Maria ama.
 Pietro, Maria loves
The crucial piece of evidence, however, comes from testing coordination of two
distinct Vocative Phrases:
(21) .  Maria, Gianni è arrivato.

 Maria, Gianni is arrived
b.  Maria e Pietro, Gianni è arrivato.
 Maria and Pietro, Gianni is arrived
 *o Maria e o Pietro, Gianni è arrivato.
 Maria and  Pietro, Gianni is arrived
If  belonged to the DP system it would be hard to understand why it cannot be

iterated by coordination, as in a simple coordination of two DPs such as il
ragazzo e la ragazza "the boy and the girl". Thus, we are forced to conclude that
 rather belongs to the split Comp Field on empirical grounds. If this conclusion
proves tenable, then, it diminishes the ad hoc flavor of the solution proposed
260 ANDREA MORO
here, i.e. that Vocative Phrases are generated in a special extra head not
previously included in the split Comp Field. In fact,  provides overt evidence
that such an inventory must be independently increased to include more slots. Of
course, it remains for us to understand what kind of information such a higher
portion of the left periphery contains. A full understanding of such a role can
only be the topic of future research, but it is not unreasonable to conjecture that
such a higher head conveys deictic and propositional information pertaining to
the root clause. Notice also that the idea that  is a feature/head belonging to the
left periphery of the root clause can be independently attested by examining
cases like the following from a dialect of Tuscany (Pratese):
(22) a. (O) i che tu f ai costi?

o what that you do there
b. La mamma, la vole sapere (*o) i che tu fai costi?
the mother she wants know  what that you do there
The contrast in (22) shows that in Pratese the interjection  can only occur in the
higher left periphery, yielding independent evidence that this element is part of
the informational endowment of the root clause.
We can conclude our preliminary approach to the field of Vocative Case. If
on the one hand we have seen that there is no straightforward diagnostics to
identify a Vocative Case, on the other we have been able to isolate some
defmitory aspects of Vocative Phrases. Vocative Phrases are noun phrases which
do not belong to the thematic grid of a predicate, although they can corefer with
a pronoun playing the role of an argument of a predicate; Vocative Phrases
behave anomalously with respect to the distribution of articles, Binding
Theoretical phenomena and certain phonological facts (such as truncation and
retraction of the stress), although these characteristics may well be independently
motivated. Moreover, we have explored the position of Vocative Phrases in the
left periphery suggesting that they occupy the spec position of a dedicated Voc°
head which is higher than Force0 in the split Comp field and arguably contains
propositional and deictic information specific to the root clause.
References
Brekle, H.E. (ed.) 1966. Grammaire générale et raisonnée on la Grammaire de Port-
Royal, facsimile of the 1676 edition. Stuttgart: Friedrich Fromann Verlag.
Cinque, G. 1979. "Left Dislocation: A syntactic and pragmatic analysis", Studi di sintassi
e pragmática. Padova: Clesp.
Cinque, G. 1990. Types of Α-bar Dependencies, Linguistic Inquiry Monograph Series 17.
Cambridge, Mass.: The MIT Press.
Chomsky, N. 1995. The Minimalist Program. Cambridge, Mass.: The MIT Press
2
Ernout, Α. & F. Thomas 1953 . Syntaxe Latine. Paris: Editions Klincksieck.
Floricic, F. 2000. "De l'impératif italien sii (soisl) et de l'impératif en général". Bulletin
de la Société de linguistique de Paris, XCV: 1.227-266.
2
Humbert, J. 1954 . Syntaxe Grecque. Paris: Editions Klincksieck.
Lazzeroni, R. 1995. "La baritonesi come segno dell'individuazione: il caso del vocativo
indeuropeo". Vitalia dialettale, vol. LVIII (N.S. XXXV).33- 44.
Longobardi, G. 1994. "Reference and Proper Names: a Theory of N-movement in syntax
and Logical Form". Linguistic Inquiry 25.609-665.
Mazzoleni, M. 1995. " vocativo". Grande Grammatica italiana di Consultazione ed. by
L. Renzi, G. Salvi & Α. Cardinaletti, III. Bologna: Mulino.
Moro, A. 1997. The Raising of Predicates: Predicative Noun Phrases and the Theory of
Clause Structure. Cambridge, England: Cambridge University Press.
Rizzi, L. 1990. Relativized Minimality, Linguistic Inquiry Monograph series. Cambridge
Massachusetts: The MIT Press.
Rizzi, L. 1997. "The fine structure of the left periphery". Elements of Grammar:
Handbook of Generative Syntax, ed. by L. Haegeman, 281-337. Dordrecht: Kluwer.
MAPPING OUT THE LEFT PERIPHERY OF THE CLAUSE
EVIDENCE FROM NORTH WESTERN ITALIAN VARIETIES*
SANDRA PAOLI
University of Manchester
1. Introduction
Pollock (1989) initiated a paradigm of research where functional
projections played an increasingly important role'. The tendency is to refine the
structure by breaking down each functional projection into a series of heads
semantically and syntactically distinct. The most comprehensive work with
respect to this tendency is Cinque (1999), where the IP is fragmented into a
myriad of projections.
The CP, as a major functional projection, could not escape its destiny: Rizzi
(1997) split the traditionally labelled CP into four different heads: Force, Topic,
Focus and Finiteness. Force and Finiteness are the positions where the standard
Italian complementisers - the finite che and the non-fínite di respectively - are
located. While Force expresses the illocutionary force of the sentence, Finiteness
is seen as containing some modal information (Rizzi 1997:284). Topic and Focus
host left dislocated (LD) and focalised sentence initial phrases respectively.
Two dialects - a conservative variety of Turinėse (Tur) and a variety of
Ligurian (Lig) - spoken in North western Italy allow for two che - a higher one,
chel and a lower one, che2 - to co-occur in some subordinate clauses. Che2 is
always optional; nevertheless, its presence is totally excluded from some
subordinate clauses, and it is on this negative evidence that I have based my
analysis. The examples in (1) show the relevant data.1
I am deeply indebted to Cecilia Poletto and Paola Benincà for their constant help and
encouragement. This paper owes a great deal to their ideas. I am also grateful to the anonymous
reviewers for their helpful comments and insightful observations. The interpretation of their
suggestions and any mistakes that may stem from them are, of course, my own responsibility.
1
The glosses contain the following abbreviations: SCL = Subject Clitic; rf = reflexive clitic; part =
partitive clitic; L = invariable clitic used together with other SCLs before 'to be' and 'to have' verb
forms beginning with a vowel. The Turinėse examples are marked by a 'T' and the Ligurian ones by
an 'L'. All subjunctive verb forms are marked by 'S'.
264 SANDRA PAOLI
(1) a. Gioanin a sper a che Ghitin ch' as ne Τ

John SCL hopes that Margaret that SCL+rf part
vada tôst.
go.S soon
"John hopes that Margaret leaves as soon as possible."
i. A Teeja a credda che a Maria ch' a L
the Teresa SCL believes that the Mary that SCL
parta dum an.
leave.S tomorrow
"Teresa believes that Mary leaves tomorrow."
The aim of this paper is to make a small contribution to the understanding

of the structure of the left periphery through the investigation of the status,
function and position of che2 elaborating Rizzi's idea of a lower head with
modal content and the subject positions identified in the higher portion of the
clause.
Since the area under investigation lies at the border between the IP and the
CP domains, this paper focuses on the relative order of the elements with respect
to each other, although an effort is made to establish their absolute positions, too.
In order to achieve this, section 2 investigates the elements that form a cluster
with che2, Subject Clitics (SCLs). Determining their position is of vital
importance to identify the position filled by che2. Section 3 turns to the syntactic
status and function of che2. Section 4 focuses on the status of subjects that
appear to the left of LD phrases and claims that they are LD themselves, be they
quantificational or non quantificational. The conclusions are summarised in
section 5.
2. The identity and position of SCLs

This section aims to establish the nature of SCLs in Tur and Lig adopting
Poletto's (2000) system, in order to identify the position filled by che2 exploiting
its linear adjacency with SCLs.
In her investigation of Northern Italian Dialects, Poletto (2000) recognises
four morphological classes of SCLs, each of which is assigned a precise position
within what she labels as agreement field, an area comprising the CP and the IP.
This agreement field is split into two parts, pre- and post- preverbal negation:
preceding the preverbal negation are Invariable and Deictic SCLs, following it,
Person and Number SCLs. While the former types are placed in the CP layer, the
latter belong to the IP domain. Each type realises different morphological
features and has particular properties that distinguish it from the others. There
are a number of syntactic tests that can be used to reveal these characteristics and
MAPPING OUT THE LEFT PERIPHERY 265
thus identify the type to which a SCL belongs. These are: position with respect to
preverbal negation, various types of verbal coordination, clustering with the
complementiser and compatibility with LD items. I analyse Tur and Lig SCLs in
turn.
2.1. Tur SCLs

Tur has a complete set of SCLs, shown in (2):
(2) . ,
person singular plural
1st i i
2nd it i
3rd a a
The first test cannot be carried out, since Tur has post verbal negation.
Turning to the second test, Poletto considers a particular type of
coordination, involving two verbs that share most semantic features and differ
only in tense or aspect. Following Benincà and Cinque (1993), these are
analysed as a complex verb rather than two separate ones. Under this type of
coordination, Person SCLs necessarily need to be repeated in the second
conjunct. Tur SCLs do not need to be repeated - cf. (3) - suggesting that they do
not belong to this type:
(3) a. A les e arles Γ istess líber.

SCL reads and re-reads the same book
"S/He reads the same book again and again."
b. It fase e arfase sempe l' istess tr av aj.
SCL do and redo always the same job
"You're always doing the same job again and again."
In Polerto's system, only the two higher types of SCLs interact with the
complementiser: they both necessarily cluster with it, while this operation is only
optional for Number and Person SCLs. All the examples above show how che2
and the SCL of the embedded clause form a cluster: lack of cliticisation yields
ungrammaticality.
2
This is not due to generalised elision - cf. the following example, where three vowels are adjacent
to each other; yet, no elision occurs:
(i) Mi i ambreiijo mai.
I SCL cheat never
"I never cheat."
266 SANDRA PAOLI
(4) a. *Giôrs a sper a che Majo che as ne vada.

George SCL hopes that Mario that SCL+rf part goes.S
"George hopes that Mario goes away soon."
b. *Luch a pensa che ij tô che a sio
Luke SCL thinks that the your that SCL are.S
ancorzuss-ne.
realised-part
"Luke thinks that your parents have realised it."
In view of (4) we can conclude that Tur SCLs belong to one of the higher types,
either Invariable or Deictic SCLs.
In order to choose between them, we turn to the interaction with LD
phrases: only Deictic SCLs are compatible with them, while Invariable SCLs are
not:
(5) a. Ti, cola ca', it la cate?

you, that house SCL it buy
"And you, that house, are you buying it?"
b. Colafietina, cole robe afrose,  ha dile nen/
that little girl those things horrible SCL L has said-them not
"That little girl, those horrible things, she hasn't said them."
The syntactic tests have revealed that Tur SCLs belong to the Deictic type and
are therefore placed in the left periphery of the clause.
Let us now tum to Lig SCLs.
2.2. Lig SCLs

Lig, unlike Tur, has an incomplete set of SCLs. These are shown in (6).
(6)
person singular plural
1st — —
2nd ti —
rd
3 fern a i
ord
u i
3 mase
The first test - i.e. position with respect to preverbal negation - immediately
reveals a clear cut division within Lig SCLs: while 3rd person singular and plural
SCLs appear before negation, 2nd person singular must follow it.
(7) i. A Teeja a nu mangia de carne.

the Teresa SCL not eats of meat
"Teresa doesn't eat meat."
ii. U Mar    mangia de carne.
the Mark SCL not eats of meat
"Mark doesn't eat meat."
iii. Gianni e a Teėjai nu mangian de carne.
the John and the Teresa SCL not eat of meat
"John and Teresa don't eat meat."
iv. 77 nu ti mangi de carne.
you not SCL eat of meat
"You don't eat meat."
This sharp division suggests that 3 rd person singular and plural belong to one of
the higher types while 2nd person singular belongs to one of the lower ones. In
order to help identifying to which of the higher and lower types Lig SCLs
belong, we can use two types of verbal coordination. The first one, which
coordinates two distinct verbs with two distinct objects, allows for the omission
of Invariable SCLs in the second conjunct. This is applied to 3 rd person SCLs -
cf. (8) i-iii. The second one, the coordination of a 'complex verb' - cf. above
discussion - requires the obligatory repetition of Person SCLs in the second
conjunct. This is applied to 2nd person SCL - cf. (8) iv:
(8) i. U Giorgiuu mangia da menestra e *(u) beiva du vin

the George SCL eats of soup and SCL drinks of wine
"George eats soup and drinks wine."
ii. A Maria a canta  ti e *(a) bala  mi.
the Mary SCL sings with you and SCL dances with me
"Mary sings with you and dances with me."
iii. U Marcu e u Gianni i mangian da menestra
the Mark and the John SCLeat of soup
e *(i) beivan du vin.
and SCLdrink of wine
"Mark and John eat soup and drink wine."
iv. 77 dij a e *(ti) dijeà sempre a mejima cosa.
SCL say and SCL will say always the same thing
"You say and always will say the same thing."
The examples in (8) show how:

• 3 rd person singular and plural SCLs do not belong to the Invariable type;
• 2nd person singular SCL seems to belong to the Person type.
268 SANDRA PAOLI
In order to confirm that 3rd person SCLs belong to the Deictic type we turn
to their interaction with LD phrases: since the two are compatible, 3 rd person
SCLs must belong to the Deictic type:
(9) i. Luatri, in regain, i gh'e Γ an ζ à cat au.

they a present SCL to him it have already bought
"And they, a present, they have already bought it for him."
ii. Le, quell a c a, u a-a catta?
he that house SCL it buys
"And he, that house, is he buying it?"
iii. A Teeja,  libru a nu gh'e I'ha anca dattu.
theTeresa the book SCLnot to him it has yet given
"And Teresa, the book, she hasn't given it to him yet."
Summing up, Lig 3rd person SCLs are of the Deictic type, 2nd person
singular of the Person type, while Tur SCLs all belong to the Deictic type. Those
SCLs that form a cluster with che2 - all the Tur SCLs and the 3rd person SCLs in
Lig - all belong to the Deictic type, and are placed in the CP. Since che2
precedes them, the conclusion we reach is that che2 itself must fill a position in
the left periphery of the clause.
The next section investigates the nature and function of che2 and makes a
more precise suggestion as to its position.
3. Che2: its syntactic status and function

3.1. Syntactic status
In all the above examples we have seen that chel is always found in its full
form, i.e. che, while che2 forms a cluster with the linearly adjacent clitics, be
they subject, reflexive, dative or partitive clitics, and never bears tonic stress.
The order within the cluster is rigid: it cannot be altered - cf. (10) a-ii - nor can
the sequence be interrupted by a phrasal element such as an adverb - cf. (10) c-
iii:
(10) a. Maria a sper a che, die fior ch' as ne' T

Mary SCL hopes that of the flowers that SCL+rf part
desmentía gnun.
forgets.S nobody
"Mary hopes that nobody forgets about the flowers."
b. * Maria a sper a che, die fwr che n' as Τ

Mary SCL hopes that of the flowers that part SCL+rf
desmentía gnun.
forgets. S nobody
i. A Maria a credda che, a-a Teeja, ch'u L
the Mary SCL believes that to the Teresa that SCL
gh'e n' aggia zà parlau.
to her part has.S already spoken
"Mary thinks that he has already spoken to Teresa about it."
ii. *A Maria a credda che, a-a Teeja, che gh 'e L
the Mary SCL believes that to the Teresa that to her
n' u aggia zà parlan.
part SCL has.S already spoken
 *Giôrs a pensa che Gioanin che mir acó as T
George SCL thinks that John that perhaps SCL+rf
ne desmentía.
part forget. S
"George thinks that perhaps John has forgotten about it."
iii. *A Maria a credda che, a-a Teeja, che fos ci L
the Mary SCL believes that to the Teresa that perhaps
 gh'e n' aggia zà parlau.
SCL to her part has .S already spoken
"Mary thinks that perhaps he has already spoken to Teresa about
it."
These facts are reminiscent of the restrictions on clitic clusters, suggesting that
che2 itself is a clitic.
3.2. Function
Although, as already mentioned, the realisation of che2 is optional, its
presence is subject to two restrictions. Each of these in isolation is necessary to
capture its realisation but not sufficient: it is their combination that accounts for
its usage. The first restriction is that there must be some phonetically realised
syntactic material between chel and che2\ the sequence chel-che2 is not
allowed, either in Tur (11) a, or in Lig (11) i:
(11) a. * Gioanin a sper a che ch' as ne vada tôst.

John SCL hopes that that SCL+rf part go.S soon
"John hopes that s/he goes away soon."
270 SANDRA PAOLI
i. *A Maria α ρ ens av a che ch'  avesciazà vistu.

the Mary SCL thought that that SCL it had.S already seen
"Mary thought that he had already seen it."
This resistance against the repetition of morphemes is reminiscent of the

Obligatory Contour Principle in phonology (Leben 1973 and much subsequent
work), which requires adjacent phonemes to be contrastive. Neeleman & van de
Koot (2001) show how this phonological restriction is also active at the syntactic
level. The ungrammaticality of the examples in (11) can thus be accounted for in
terms of syntactic haplology.
The second restriction is that the verb in the embedded clause must be in
the subjunctive mood: che2 is disallowed with any other tense. Present and
future indicative and conditional, for examples, do not trigger it, in either variety,
as shown respectively in (12) a, b and  for Tur and i, ii and iii for Lig:
(12) a. *A dis che luatri ch a m angi  nen 'd rane.

SCL says that they that SCL eat not of frogs
"S/He says that they do not eat frogs."
b. *Giors a sper a che Majo ch' as n' andará tôst.
George SCL hopes that Mario that SCL+rf part will go soon
"George hopes that Mario goes away soon."
c. *Majo a pensa che Luch ch' as n' ancorzeria.
Mario SCL thinks that Luke that SCL+rf part would realise
"Mario thinks that Luke would realise it."
i. * U Gianni  disa che a Maria ch' a nu
the John SCL says that the Mary that SCL not
mangia de rainocce.
eats of frogs
"John says that Mary does not eat frogs."
ii. * U Francu u sper a che u Gianni ch' 
the Frank SCL hopes that the John that SCL
telefuneà a-u ciü fitu.
will phone at the more soon
"Frank hopes that John phones as soon as possible."
iii. * U Gianni  creada che Margaitin ch ' 
the John SCL believes that Margaret that SCL
mangées cia guentea sta turta.
would eat willingly this cake
"John thinks that Margaret would eat this cake willingly."
This restriction is not a general restriction on the complementiser che: chel is

MAPPING OUT THE LEFT PERIPHERY Z/l
triggered whenever the higher verb selects a finite embedded clause, and its
realisation is obligatory/
(13) a. *Ghitin a dis loràutri a s er iv-/s er ivr an/s er ivr lo T

Margaret SCL says they SCL write/will write/would write
na litra.
a letter
"Margaret says they write/will write/would write a letter."
i. * U Giorgiu  disa a lalla a L
the George SCL says the aunt SCL
telefua/telefunea/telefuneescia stas ers a.
phones/will phone/ would phone tonight
"George says the aunt phones/will phone/would phone tonight."
The connection between subjunctive mood and realisation of che2 is thus clear,
and I would like to claim that che2 is a subjunctive particle. Tur and Lig, as well
as marking subjunctive mood on the verb morphology - although in a reduced
way, cf. Table 1 -, also make optional use of an additional element, che2.
It has been argued in the literature (cf., for example, Giorgi and Pianesi
1997, von Stechow 1995) that the subjunctive is deficient. Morphologically,
because of the lack of morphological differentiation between some forms of the
present indicative and the present subjunctive; semantically, because on its own
the subjunctive does not give raise to any real temporal interpretation. This
deficiency is also witnessed in Tur and Lig: the morphological differentiation
between the present indicative and subjunctive is minimal, as Table 1 shows:
TPres Ind T Pres Subj L Pres Ind L Pres Subj

Is mi i parlo che mi i parla mi p ar lu che mi parla
lis ti't parle che ti't parle ti ti parli che ti ti parli
Ills chiel a parla che chiel a parla lé u parla che lé u parla
Ip noi i par loma che noi i parlo nuatri parlemmu che nuatri parlemmu
Up voi i parle che voi i parle vuatri parlei che vuatri parlei
\  lor a parlo che lor a parlo luatri i parlan che luatri i parlan
Table 1: Indicative and Subjunctive in Turinėse and Ligurian
The table gives the paradigm for a verb of the first conjugation, Tur parlé and
Lig parla "to speak", and shows in italics the forms that are distinct in the two
3
This fact reinforces the claim that chel and che2 are different.
272 SANDRA PAOLI
moods. Thus che2 would disambiguate between those forms that are identical in
the indicative and subjunctive. This analysis finds further support in the fact that
the optional use of che2 becomes even more so with first person singular and
plural in Tur and first person singular in Lig, precisely those forms that already
mark the mood distinction on the verb morphology.
Chel then is what may be labelled as the 'canonical' complementiser, i.e. in
Rizzi's (1997) structure it fills Force0 - but cf. Benincà (2001) for an alternative
analysis. Che2, on the other hand, is a subjunctive marker, and following Rizzi's
(1997) intuition of a lower  head encoding modal information and the facts
seen in Section 2,1 would like to propose that che2 is an expletive that fills Fin°.
Having identified more precisely the position filled by che2 as belonging to
the left periphery of the clause we have interesting consequences for the subject
that appears between chel and che2, and these are investigated in the next
section. But before moving on I would like to point out that the type of modality
encoded by che2 is not related to the realis / Irrealis distinction often quoted in
the literature. Che2 is triggered in embedded clauses selected by factive verbs -
which cannot be [-realis] by definition - as shown in (14) a and i:
(14) a. María a regret a che Giôrs ch' a sia dësmentiass-ne. T

Mary SCL regrets that George that SCL is.S forgotten.rf-part
"Mary regrets the fact that George forgot about it."
i. A-α Mariagh'e desplaza che u Mario ch' u L
to-the Mary to-her hurts thatthe Mario that SCL
sbragia sempre.
scream. S always
"Maria is sorry that Mario is always screaming."
Thus this lower head cannot be the same as the one identified by Poletto
(2000:118 ff) linked to the 'complementiser deletion' phenomenon, which seems
to encode a [-realis] feature.
4. Che2 and subject positions in the left periphery

This section aims to establish the status of the subjects that fill the position
between chel and che2 and whether there are any restrictions on the types of
subjects allowed there.
4.1. Preverbal subject positions

In the examples in (1) the position occupied by the subject can be analysed
in two ways: it either is a canonical subject position or a position where the
subject is discourse prominent, i.e. either LD or focalised. In order to choose
between these two possible interpretations we need to consider more data.

The presence of a LD object immediately following a subject forces a
reading where the subject is LD itself and fills [Spec, Top].4
(15) a. Gioanin a sper a che Teresin, cost robe, ch' a T

John SCL hopes that Teresa these things that SCL
Γ abi a nen dile.
L has. S not said-them
"John hopes that Teresa hasn't said these things."
i. A Maria a credda che u Gianni,  piimmu, ch' u L
the Mary SCL believes that the John the apple that SCL
aggi a z à mangiau.
it has.S already eaten
"Mary thinks that John has already eaten the apple."
The same sequence is also grammatical with a quantified subject:
(16) a. Maria a sper a che gnun, die fior, ch' Τ

Mary SCL hopes that nobody of the flowers that
as ne desmentía.
SCL+rf part forget.S
i. A Maria a spera che nisciun, a-n Gianni, L
The Mary SCL hopes that nobody to-the John
ch'  gh'e aggia zà dattu.
that SCL to-him it has. S already given
"Mary hopes that nobody has already given it to John."
Quantificational elements, because of their indefinite nature, are usually unable

to undergo left dislocation. So, what is the status of the subjects in (16)?
Cinque (1990), in his investigation of clitic left dislocation constructions
(CLLD), classifies them as an instance of A' dependencies, together with
successive cyclic Wh-movement, long Wh-movement and apparent Wh-
movement of NPs.5 He also defines as the prerequisite for undergoing long Wh-
4
Benincà (2001) has argued, contra Rizzi (1997), that TopP is not a recursive projection. Benincà and
Poletto (2001) further claim that TopP and FocusP are not to be analysed as single projections but as
fields, hosting a number of projections for LD and focalised phrases respectively: the Top field hosts
LD elements and elements with a list interpretation; the Foe field is a landing site for contrastive and
informational focus elements. I will therefore assume that each LD projection fills a different
Specifier of a projection within the Top field.
5
The LD element is resumed through a clitic and the two are in a binding relationship.
274 SANDRA PAOLI
Movement "the intrinsic referential character of a phrase", a condition which is

to be considered as a condition on A' chains in general. CLLD structures being
an instance of such a chain, means that referentiality - in the sense of Pesetsky's
(1987) D-linking - is a requirement for left dislocation. Thus, an element can be
left dislocated only if it is referential, which in turn bears on its ability to enter
into a binding relation with its resumptive clitic. Therefore, if a quantified
element is referential, i.e. if it can be interpreted as specific, it can indeed be left
dislocated. In (16) both gnun and nisciun are interpreted as referring to a specific
individual, perhaps "nobody of our friends", or "nobody of the people we know",
etc. If, on the other hand, a specific interpretation is not available or difficult to
attain, then the sentences are deviant, as shown in (17).
(17) a. *?Giôrs a cher d che gnun, d' un bon consèj, T

George SCL believes that nobody of a good advice
ch' as n' ambrigna prôpi.
that SCL+rf part not cares.S really
"George believes that nobody would ignore a good piece of
advice"
i. *?ATeeja a credda che nisciun, ste cose, L
The Teresa SCL believes that nobody these things
ch ' u-e dij a.
that SCL-them say.S
"Teresa believes that nobody would say these things."
Concluding, the subjects to the left of LD phrases are to be considered LD

themselves, be they quantified or not, and the position they fill as belonging to
the LD field.
Such an interpretation is supported by some interesting facts in Paduan.
Paduan does not exhibit obligatory SCLs co-occurring with subjects: they are
only present when the subject is left dislocated. Interestingly, a quantified subject
is compatible with a SCL only when it appears followed by a LD phrase - Paola
Benincà (p.c.):
(18) a. Credo che nissuni, na idea simile,*?(el) possaver-la vua.

Ibelievethat nobody an idea similar SCL can.S have-it had
"I think that a similar idea, nobody can have had it."
b. Credo che nissuni (*el) poss a ver-Ία vua.
I believe that nobody SCL can have-it had
"I think that nobody can have had it."
If we then turn to investigate the preverbal subject position to the right of LD

phrases and to the immediate left of che2, we notice an interesting difference:

while in Tur only non quantificational subjects are allowed, in Lig no such
restriction applies:
(19) a. Ghitin a sper a che, 'd son, Luch ch' as T

Margaret SCL hopes that of this Luke that SCL+rf
ne' desmentía nen.
part forget. S not
"Margaret hopes that Luke doesn't forget about this."
b. *?'María a sper a che, die fior, gnun ch ' T
Mary SCL hopes that of the flowers nobody that
as ne desmentía.
SCL+rf part forget.S
i. U Giorgiu  pensa che, a-a lalla, i fanti L
The George SCL thinks that to the aunt the children
ch' i gh'e aggian zà catau in regain.
that SCLto-her L have.S already bought a present
"George thinks that, to their aunt, the children have already
bought a present."
ii. A Maria a credda che, a-a Teeja, nisciun ch ' L
The Mary SCL believes that to-the Teresa nobody that
 gh'e n' aggia z à parlan.
SCL to-her part has.S already spoken
"Mary believes that nobody has already spoken about it to
Teresa."
Let us briefly summarise the evidence gathered so far on subject positions. (20)
shows where quantified (Q) and non quantified (DP) subjects can or cannot
appear in the left periphery:
(20) a. ...c/?e7/QSubj//DPSubj LD *Q Subj//DP Subj che2 T

i. ...chel /QSubj/ZDPSubj LD / Q Subj//DP Subj che2 L
The interesting variation concerns the position to the right of LD phrases and to
the left of che2. Why would the two dialects differ with respect to the types of
subject allowed to fill what appears to be the same position? A possible answer
derives from the fact that while Lig allows for sentence initial focalised phrases,
Tur does not, suggesting that Tur has less structure available in the left periphery;
more specifically, Tur has a TopP but not a FocP projection.
We could therefore assume that the subject to the left of LD phrases still
276 SANDRA PAOLI
lies within the Top field - cf. Benincà and Poletto (2001) - in Tur, but is already
part of the Focus field in Lig. Thus, while the subjects in (19) a and b fill [Spec,
Top] in Tur, they are in [Spec, Foe] in Lig. The restriction on the types of subject
that can appear in Tur could not be due to gnun receiving a non-specific
interpretation, since the context is exactly the same as (16) a. Perhaps it is due to
a hierarchical constraint on the way phrases can be left dislocated. More
specifically, quantificational elements, if LD, can only fill a higher position than
non quantificational ones.6
Turning now to Lig, if the subject in (19) ii did indeed fill the Specifier of a
contrastive focused position, it would not be able to co-occur with another
contrastively focussed element, given that only one focalised phrase is permitted
in any one sentence. (21) i shows how such a combination is deviant:
(21) i. ?*A Teeja a pensa che, a torta, nisciun ch' L

the Teresa SCL thinks that the cake nobody that
u-α mangia STASEIA, nu duman.
SCL-it eat.S tonight not tomorrow
"Teresa thinks that nobody would eat the cake TONIGHT, not
tomorrow."
In (21) staseia "tonight" is contrastively focussed in a postverbal position and its

co-occurrence with nisciun to the right of a LD phrase is not allowed. The
sentence is fine if nisciun appears postverbally and bears no contrastive stress.
This suggests that the position occupied by nisciun is a focalised position:
ii. A Teeja a pensa che, a torta, STASEIA che L

the Teresa SCL thinks that the cake tonight that
-a mangia nisciun, nu duman.
not-it eat nobody neg tomorrow
Thus (20) can be reformulated as (22).7
(22) a. chel [ / Q Subj//DP Subj ]LDLD [*QSubj//DPSubj]LD che2 T

i. chel [/QSubj//DPSubj] LD LD [ / Q Subj//DP Subj ]Focche2 L
6
Some support for this idea comes from Paduan where a quantified subject following a LD phrase
does not trigger the SCL, suggesting it is not LD.
7
1 am not claiming that in Tur the Top field is bigger than in Lig.
MAPPING OUT THE LEFT PERIPHERY
277
5. Conclusions
This paper has presented some interesting data that supports Rizzi's (1997)
intuition of a lower  head with modal content and further elaborates it making a
connection between this head and the modality encoded by the subjunctive. An
investigation of the nature of the subject positions identified in the left periphery
has suggested that a quantifie ational subject appearing in a position to the left of
LD phrases has a specific interpretation and is itself LD. Finally, the paper has
suggested that there may be a hierarchy at work in the left dislocation process
that only allows LD quantificational elements with a specific interpretation to
appear to the left of non quantificational LD phrases.
References
Benincà, Paola 2001. "The position of Topic and Focus in the left periphery". Current
studies in Italian Syntax: Essays offered to Lorenzo Renzi ed. by Guglielmo Cinque
& Giampaolo Salvi, 39-64. Amsterdam: Elsevier.
& Guglielmo Cinque 1993. "Su al cune differenze fra enclisi e proclisi". Omaggio
a Gianfranco Folena, 2313-2326. Padova: Editoriale Programma.
—- & Cecilia Poletto 2001. "Topic, Focus and V2: defining the CP sublayers".
Manuscript. University of Padova.
Cinque, Guglielmo 1990. Types ofA'- Dependencies. Cambridge, Mass.: The MIT Press.
- 1999. Adverbs and Functional Heads. Oxford: Oxford Uiniversity Press.
Giorgi, Alessandra & Fabio Pianesi 1997. Tense and Aspect: from Semantics to
Morphosyntax. Oxford: OUP.
Leben, William 1973. Suprasegmental Phonology. PhD Dissertation, Cambridge, Mass.:
MIT.
Pesetsky, David 1987. "Wh-in-situ: Movement and Unselective Binding". The
Representation of (In)definiteness ed. by Eric J Reuland & Alice G.B. ter Meulen,
98-129. Cambridge, Mass.: MIT
Neeleman, Ad & Hans van de Koot 2001. Syntactic Haplology. Manuscript, UCL.
Poletto, Cecilia 2000. The Higher Functional Field. Oxford: Oxford University Press.
Pollock, Jean-Yves 1989. "Verb Movement, Universal Grammar and the Structure of IP".
Linguistic Inquiry 20:3.365-424.
Rizzi, Luigi 1997. "The fine structure of the left periphery". Elements of Grammar ed. by
Liliane Haegeman, 281-337. Berkeley: Kluwer.
Stechow, Armin von 1995. On the proper treatment of tense. Manuscript. University of
Tübingen.
THE LEFT PERIPHERY IN CHILD FRENCH
EVIDENCE FOR A SIMPLY-SPLIT CP
DORIAN ROEHRS & MARIE LABELLE

Indiana University and Université du Québec à Montréal
1. Introduction
In this paper, we investigate a non-adult-like usage of the complementizer
QUE. Labelle (1993, 2000) reported a number of examples from various corpora
of child French in which the complementizer is not produced where it is
expected (called 'Misplaced QUE') or where it is unnecessarily repeated after a
left dislocated DP (called 'Intrusive QUE'). We show that the data provide
evidence in favor of Rizzi's split CP system (Rizzi 1997), as well as in favor of a
view of movement as a combination of Copy, Merge, Chain Formation, and
Chain Reduction (= Delete) (Nunes 1999, 2001). We follow Mayer, Erreich &
Vahan (1978) in assuming the Basic-operations Hypothesis according to which
in language acquisition one (or more) basic operation may fail to apply. In
particular, we claim that 'Misplaced' and 'Intrusive' QUE constructions are
simply-split CPs arising from a not fully matured control over Merge, Chain
Reduction and agreement. In that sense we treat them as performance errors: the
child knows the underlying processes but has an as yet imperfect mastery over
the control procedures that would allow her to execute them without error in all
instances. We therefore view these performance errors here as misapplications of
competence-driven operations: the child's competence dictates what operations
she should perform, but cognitive overload or insufficient short-term memory
occasionally impedes her from carrying out the operations correctly. Intrusive
and Misplaced QUE constructions therefore provide a window on the underlying
operations of the computational system.
In section 2 we survey the facts to be accounted for; in section 3, we present
the fundamental ideas of Rizzi (1997); in section 4, we discuss complex CPs in
child French; section 5 details our account of Intrusive and Misplaced QUE's.
We then show that Misplaced and Intrusive QUE's can be considered as one
example of a more general difficulty that children have in the acquisition of the
CP system.
2. Misplaced and Intrusive Q UE

Labelle (1993) lists 23 examples of constructions of the type illustrated in
(1) to (6), which she labels 'Misplaced and Intrusive QUE'.
280 DORIAN ROEHRS & MARIE LABELLE
2.1 'Misplaced Q UE '

In the Misplaced QUE construction, the complementizer QUE follows a
dislocated subject DP instead of preceding it. The construction is illustrated in
(1) to (3) with a complement clause, an adverbial clause and a relative clause; the
expected position of the complementizer is indicated by an underline.
(1) Complement clause:

II s 'est aperçu _ la porte QUE elle était ouverte. (GL, MG 5)1
he noticed the door THAT it was open
"He noticed that the door was open."
= (expected) : Il s'est aperçu que la porte elle était ouverte.
(2) Adverbial clause:
C'est pour Christian OU' ι(l) vient. (MP, child of 6)
it's for Christian THAT he comes
"It's in order that Christian come."
= (expected) : C'est pour que Christian il vienne.
(3) Relative clause:
Où elle est la boule _ Gabriel OU' il a cassée ? (ML, C1.4;6)
where she is, the ball Gabriel THAT he broke ?
"Where is the ball that Gabriel broke?"
= (expected) : Où elle est la boule que Gabriel il a cassée ?
2.2 'Intrusive QUE'

In the Intrusive QUE construction, a complementizer is followed by a
dislocated DP, which in turn is followed by a second instance of a
complementizer. This second complementizer is the 'Intrusive' QUE. The
construction is illustrated in (4) to (6) with a complement clause, an adverbial
clause and a relative clause.
1
Examples from GL are from an unpublished corpus by Guy Labelle (UQAM) where the age of
children is given in years only; MP stands for Méresse-Polaert (1969) who does not give the exact
age of the six-year-old children she studied; BP stands for Bouvier & Platone (1976); ML stands for
Labelle (1989). The examples from PHI (Philippe) and GRE (Grégoire) are from the CHILDES
database (Macwhinney 1991). The child's identification may follow these initials.
THE LEFT PERIPHERY IN CHILD FRENCH 2 81
(4) Complement clause:

Ils savaient pas QUE leur maman OU' elle était rentrée.
they knew not THAT their mother THAT she was back
"They didn't know that their mother had come back." (MP, child of 6)
= (expected) : Ils savaient pas que leur maman elle était rentrée.
(5) Adverbial clause:
Quand QUE les indiens OU' i(ls) veulent Γ attaquer?
when THAT the indians THAT they want-to him attack ?
"When the indians want to attack him?." (MP, child of 6)
= (expected) : Quand (que) les indiens ils veulent l'attaquer?
(6) Relative clause:
un trésor QUE le bandit OU' i(l) avait camuché (MP, child of 6)
a treasure THAT the thief THAT he had hidden
"a treasure that the thief had hidden"
= (expected) : Un trésor que le bandit il avait ca(mu)ché
Observe in (5) that the Intrusive QUE is distinct from the QUE found in doubly-
filled COMP constructions: the adverbial clause is introduced by a doubly-filled
COMP (quand + que), followed by a DP, followed by an Intrusive QUE.
Although infrequent, these constructions are produced by children of
various linguistic and socio-economic backgrounds; the examples in Labelle
(1993) come from five distinct studies ranging from spontaneous production, to
elicitation of relative clauses, some conducted in France, others in Canada. The
ages of the children producing the examples vary from 3;5 to 6. To give an idea
of the frequency of the constructions, Méresse-Polaert (1969) cites ten such
examples out of a corpus of about 40,000 words of narrative speech elicited from
72 six-year-old children. The corpus of relative clauses gathered by Labelle
(1989) contains seven such examples (produced by six different children) out of
a total of 1348 relative clauses, that is 0.5%. Interestingly, a similar proportion
(0.4%) has been estimated by Stromswold (1990: 60) for double tensing errors in
English children's question structures (see section 6). The fact that misplaced and
intrusive QUE's are reported by different authors and are attested in different
settings suggests that they reflect a feature of developing French.
The constructions produced by the children have the general form
illustrated in (7):2
(7) a. Misplaced QUE: ... 0 DP que [IP pronoun]

b. Intrusive QUE: ...que DP que [IP pronoun]
2
There is a second type of these forms involving qui as the lower complementizer. For clarity, we
postpone its discussion to section 5.2.
In all the examples reported, the DP preceding the erroneous complementizer

corresponds to the subject and is interpreted as a left-dislocated (= topicalized)
subject DP in an embedded clause. Importantly, neither type of construction is
produced with a comma intonation between the dislocated DP and QUE. Labelle
(1993) proposed an analysis in terms of CP-adjunction, as schematized in (8):
(8) a. ... [cpDP[ C p£uę[IP]]] (Misplaced QUE)

b. [Cp (WH) [ (que) [CP DP [CP que [ IP ]]]]] (Intrusive QUE)
However, the assumption that grammars of child French generate sentences

with the form in (8a) raises formal learnability problems since adjunction to an
embedded CP is not allowed in adult French, as shown in the complement and
relative clauses in (9b) and (9d).
(9) a. Je pense que Pierre, il est parti.

I think that Pierre he is-gone
"I think that Pierre is gone."
b. * Je pense Pierre qu 'il est parti.
 l'endroit où Pierre il est parti
the place where Pierre he is-gone
"the place where Pierre is gone."
d. * en droit Pierre ou il est parti
Furthermore, CP-recursion of the type schematized in (8b) is not a feature of

adult French either. In both cases, it remains unclear how the language acquiring
child could retreat from its non-adult-like grammar without negative evidence.
We show that the data find a better explanation within Rizzi's Split CP proposal.
3. Brief Summary ofRizzi (1997)

Rizzi (1997) proposes that the traditional CP may be viewed as a system of
four distinct heads, as illustrated in (10). The FORCE head faces outside: it is the
interface between a proposition and a higher clause or discourse, marking a
proposition as a question, declarative, relative, etc. TOPIC expresses old
information. Left-dislocated elements appear in Spec,TopP. Topics are recursive.
FOCUS expresses new information. Focused elements appear in Spec,FocP.
FINITENESS is the interface of the complementizer system facing inside. It
expresses the fact that complementizers agree with the fmiteness of the
embedded clause, and that they may, in some languages, agree in person or
number with the verb.
(10) [ForceP [TopP [FocP [TopP [FinP [IP]]]]]]

THE LEFT PERIPHERY IN CHILD FRENCH 283
The Topic or Focus phrases are not projected if the clause contains no overt topic
or focus element. If no constituent is topicalized or focused, two possibilities for
the complementizer system are discussed by Rizzi. Either Force and Fin are
realized as a syncretic head, that is, a single head as in (11), or two heads are
projected into a simply-split CP (12) (Rizzi, fh. 28).
(11) Force/FinP
Force/Fin IP
(12) ForceP
Force FinP
/ \
Fin0 IP
In the second case (12), the complementizer is generated in Fin and moves to
Force to check the Force features. Assuming with Nunes (1999, 2001) that
movement consists of Copy, Merge, Chain Formation, and Chain Reduction (see
section 5.1. for details), this movement from Fin to Force is represented as in
(13). It is excluded by minimality when a Focus or a Topic head intervenes.
(13) ... [ForceP que1 [FinP ... qua [ίΡ ...]]]
In what follows, we propose that Misplaced and Intrusive QUE

constructions result from the children's projection of a simply-split CP. Non-
adult-like instances of 'early' Merge as well as improper Chain Reduction result
in Misplaced or Intrusive QUE:
(14) a. Misplaced QUE: [ForceP que, [FinP DP qu^ ...]]

b. Intrusive QUE: [ForceP 4 ue i [πηΡ DP quej ...]]
4. Complex CP Structures in French

Before entering the core of the analysis, it is appropriate to ask what aspects
of the grammar of a French-speaking child might provide a source for the
constructions under discussion. Do children project a complex CP system? Do
they have evidence for QUE heading Force or Fin? As it turns out, French two-
year olds already produce complex CP structures.
4.1 Clitic Left Dislocation (CLLD)

Clitic left dislocation is a normal feature of French. In fact, subject CLLD is
so frequent that some researchers (e.g. Hulk 1995 for child language) suggest
that the left dislocated element is in reality a subject in Spec,AgrSP (the clitic
being a part of the verbal inflexion)/ However, Labelle & Valois (1996) show
that two-year-old children already left and right dislocate both subjects and
objects, (15a) being an example of a left-dislocated object. Also, Labelle (2000)
shows that at least some left dislocated subjects are topicalized elements; in
(15b), the marker là delimits the Topic (furthermore, the DP is dislocated twice).
Clitic left dislocations are found even in embedded clauses, as shown by the
early example from Philippe in (16).
(15) a. L'argent on le garde. (PHI 2;7,11, Labelle&Valois 1996:69)

"The money, we keep it."
b. Pi lui, le g(r)os bonhomme là, (il) s'appelle quoi?
"And him, the fat man là, (he) is called what?"
(Max 2;2.9,de Cat 2002)
(16) Elle a vu [QUE [le lapin il était parti ]]]
she saw THAT the rabbit it was gone
"She saw that the rabbit was gone." (PHI 2;11.7, file 6:832)
We conclude that French-speaking children produce CLLD constructions,

and take Intrusive and Misplaced QUE construction as providing further
evidence in favor of dislocated subject DPs in child language. Assuming Rizzi's
system, (16) has the dislocated DP in Spec,TopP and the Force head filled by the
complementizer QUE (the exact status of subject clitic pronouns is not crucial
for the point we want to make, and we won't discuss it further):
(17) [porceP Que [lopP D P i [ Fm P [iP Ui- · · ] ] ] ]
4.2 Pseudo-relative constructions

Pseudo-relative constructions are found in presentational constructions
(18a) and as complements of verbs of perception (18b). They are among the
earliest complex constructions produced by French-speaking children (Labelle
1989).
3
G Labelle (1976) gives the following figures for five-year-old children. Children from Montreal,
Canada: 89% of lexical subjects are dislocated. Children from Paris, France: 79%. For early child
language, Labelle & Valois (1996) calculated, for example, 31 preverbal lexical subjects vs 25 left
dislocated subjects (and 32 right dislocated subjects) in the corpus of Grégoire (10 files) (CHILDES).
(18) a. Ça c'est une tortue qui mange des herbes.

that it's a turtle who is-eating some grass
'That's a turtle eating grass." (GRE 2;5.13 (09:1429))
b. On regarde le grand-père qui est dans le tracteur.
one look-at the grandfather who is in the tractor
"We are looking at the grandfather in the tractor."
(GRE 5.27(10:529))
c. ( 'est une petite fille...) ...qu 'a va en promenade.
(that it's a little girl...) ... that she is taking a walk
(That's a little girl)... "talking a walk." (ML; 3;4 07A)
The (pseudo-)relativized DP always corresponds to the embedded subject. In

(18c), the clause is introduced by QUE followed by a subject pronoun (a = elle
'she'), instead of being introduced by QUI followed by a null subject. Marie-
Hélène Côté (1998:154) proposes to analyze pseudo-relatives as in (19a), which
we reformulate as in (19b), with the DP in Spec,FinP:
(19) a. ... [Cp DPj [- QUI, [IP ec¡... ]]] (Côté 1998: 154 )
b. ...[FinpDPitFin'QUiitipeCi ...]]]
While in CLLD a complementizer precedes the dislocated subject (17), in

pseudo-relatives, a complementizer follows the DP. We hypothesize that pseudo-
relatives, providing evidence for a lower complementizer, are the source of
Misplaced and Intrusive QUE constructions.
5. The Proposal Proper

5.1 Source of the Child Forms and Grammatical Constraints
The leading idea of our proposal is that Intrusive and Misplaced QUE
constructions are performance errors with complex CPs. Under taxing
conditions, a limited number of computational operations are either applied too
soon or not at all. As a matter of execution, we follow the general framework of
the Minimalist Program (Chomsky, 1995) and Nunes' (e.g. 1999; 2001)
'Copy+Merge' theory of movement.
Nunes argues that instances of apparent movement are the result of the
interplay of four independent primitive operations, Copy, Merge, Chain
Formation, and Chain Reduction (= Delete): an element is merged, copied, and
this copy is merged in a higher position checking features in the process. Nunes
assumes that only the upper copy checks its formal features, not the lower one.
Chain Formation then applies under asymmetical c-command. Chain Reduction
may delete either one of the copies: if the lower copy is deleted, the derivation
converges without additional operation. If the upper copy is deleted, the formal
features of the lower copy must be deleted by a subsequent operation, FF-
Elimination, for Full Interpretation at PR Consequently, Chain Reduction is
optimal if FF-Elimination does not apply. This ensures that the derivation with
the lower copy deleted is more economical. FF-Elimination is thus independent
of feature checking operations and basically governs the phonetic realization of
copies after Chain Formation. In this theory, while Merge is a single operation,
movement consists of four operations, one of them being Merge. Now, Mayer,
Erreich & Valian's (1978: 1) Basic-Operations Hypothesis predicts that "for any
transformation which is composed of more than one basic operation, there exists
a class of errors in child speech correctly analyzed as failure to apply one (or
more) of the operations specified in the adult formulation of the rule." This
allows for the possibility that in child language Copy, Merge, Chain Formation,
and Chain Reduction may not be properly employed. We suggest that this is what
happens in Intrusive and Misplaced QUE constructions.
In CLLD, the correct construction would require (among other operations)
merging a null head under Fin, then a Top head with the topic DP in Spec,TopP,
and finally the complementizer under Force, giving a fully complex CP:
(20) [ForceP que [TopP DP Top [FinP 0 [IP ... ]]]]
We propose that the child, attempting to produce a CLLD construction,

erroneously merges the overt complementizer and the topicalized subject DP
'early' in the structure. She merges the overt complementizer in Fin and the
topicalized DP in Spec,FinP, basically applying the operations deriving pseudo-
relatives instead of those required for topicalization. With QUE and the DP
merged in the earliest possible (relevant) positions, the child then projects a
simply-split CP in order to check the Force feature of QUE. The simply-split CP
allows the complementizer to move to Force and the derivation to converge.
(21) ForceP
With movement of QUE to Force in a simply-split CP, the most economical

derivation results in the Intrusive QUE construction in which both Chain
Reduction and FF-Elimination have not applied :
(22) [ForceP que [FmP DP que ...]]
Misplaced QUE constructions follow from improper Chain Reduction and failure
in the application of FF-Elimination:
An anonymous reviewer suggests that Misplaced QUE could also be due to

covert movement of the complementizer to Force. In this scenario, the child
follows the economy principle Procrastinate (or any other interpretation of covert
movement). Either case involves movement of the complementizer.
Deletion of the lower copy yields (24), which is superficially like the adult
forms (cf. (16)), except that the Topic is erroneously merged in Spec,FinP.
This implies that the superficially correct constructions produced by the children
may, in a few cases, be structurally non-adult.
The early Merge proposed here may follow from economy considerations:
the child attempts to take out lexical elements from the numeration as soon as
possible with the result of building up a more economical representation. The
numeration is 'emptied' earlier, less structure is projected, and movement and
construal relations hold across fewer nodes. With an early Merge, the
'derivational horizon' (Uriagereka, 1998) is narrowed down sooner and the
computational burden is reduced. We assume that, given their limited capacities,
children are particularly constrained by this condition on the economy of
representation. The early merge of QUE in Fin would then be triggered by a
combination of 1) the pressure to reduce the computational burden on short-term
memory and 2) the frequency of the operation, used to construct pseudo-relatives
(19b). Unlike Mayer, Erreich & Valian (1978) then, we view errors such as these
as performance phenomena which result from the computational system
erroneously failing to apply or typically applying too soon, probably because of
computational overload.
5.2 The Case of 'qui'

For the sake of clarity, we only discussed child forms with QUE as the
lower copy of the complementizer. However, there is a second subtype,
involving QUI as the lower copy. Compare (25) with (26), with a feminine DP to
the left of Fin. While in (25) the complementizer QUE (elided to QU') is
followed by the feminine personal pronoun eile, in (26) the complementizer is
realized as QUI, followed by a null subject. (With a masculine DP, the phonetic
realization [ki] is ambiguous between QUI and QU'/ (/ being a reduced subject
pronoun U ("he") or Us ("they")), so that QU'ifl/s) in (1) to (6) may also be
interpreted as QUI.)
(25) pendant [la soupe [OU' elle refroidit]] (MP, child of 6)

while the soup THAT it cools down
"while the soup is cooling down"
= (expected) : pendant que la soupe elle refroidit
(26) le camion  QUE [ la dame [QUI donne de V essence]]]
the truck where THAT the lady THAT gives some gas
"the truck that the lady is filling up with gas" (BP, 5 ;6)
= (expected) : Le camion où (que) la dame elle donne de l'essence
These examples have the following structures (ec = empty catetory):
(27) a. [
ForceP
(que) [FmP DP que [IP pron ]]]
u
- LForceP
(que) [FmP DP qui [IP ec ]]]
Our simply-split CP analysis involving movement accounts straight-forwardly
for the QUE cases. We now have to account for the fact that in (26) the upper
complementizer surfaces as QUE and the lower one as QUI. Assuming that QUI
is an agreeing ailomorph of QUE (Kayne, 1976), we follow the spirit of Rizzi
(1990) who proposes that C° is realized as QUI in the context illustrated in (28),
in which a Spec-head agreement relation holds in both CP and IP and the Spec
position of the lower phrase contains an empty category:
(28) [ X P i C V e C i l V - ] ]
Translating this idea into the Split-CP framework and employing Distributed
Morphology (Halle & Marantz, 1993), we assume that at PF QUE under Fin is
spelled-out as QUI, if (29) holds, that is, if Fin 0 formally agrees with Io:
(29) [XPiFin^teCilV..]]
(26) is then analyzed as (30) where the configuration of (29) is given in FinP but
crucially not in ForceP. Consequently, the complementizer in Fin is spelled out as
QUI and the one in Force as QUE.
(30) le camion [ForcepOÙ QUE [FinP la dame¿ QUI, [IP ec¡ donneį de
l'essence]]]
The QUI examples provide an argument in favor of our early Merge

analysis. While CLLD constructions involve a clitic subject, pseudo-relatives are
introduced by QUI with an empty category in subject position. Given that
children often produce QUE+pronoun as well as QUI+ec in relative clauses and
in pseudo-relatives (Labelle 1990 provides the following figures: with a feminine
antecedent, qu'elle: 103 (26.9%), qui 280 (73.1%)), the occurrence of both QUE
and QUI as the lower complementizer is an indication that they are constructing
a child-like pseudo-relative type of structure (cf. (19b)) and not a CLLD topic
construction.
Nothing so far has been said about the correlation between the realization of
the complementizer and that of the subject. QUI involves movement of the DP
leaving a trace (an empty category), whereas QUE occurs with a base-generated
DP and a resumptive pronoun (Haegeman 1994: 409). We tentatively assume that
the correlation can be derived by double Spec positions in FinP coupled with
some version of the Shortest Link Condition. In (31a) the DP in the lower Spec
position is 'close enough' to have moved from Spec,IP, leaving a trace, whereas
in (3 lb) the DP in the upper Spec is not close enough and the subject position is
spelled out as a resumptive pronoun:
(31) a. [... [FinPDPįqui [ÏP ec¡ ]]]

b. [Finp OVX [FmP 0 que [IP pronį ]]]
The constellation of (29) holds in (31a), yielding QUI, but not in (31b). A full
discussion of the nature of the intervening null element (0) as well as of the
nature of the violation leading to the resumptive-pronoun strategy would go
beyond the scope of this paper (see Aoun, Choueiri & Hornstein 2001 for some
recent discussion of resumptions). This analysis accounts for (32), in which two
topic phrases are in the C-system:
(32) [quand[mapetite sœur [le soir [QU' [elle dort]]]]]

when my little sister at night THAT she sleeps
"when my little sister sleeps at night" (MP, child of 6)
= (expected) : quand (que) ma petite sœur le soir elle dort
Assuming that both Spec positions are filled by a dislocated DP, the adverbial DP
intervening between the left-dislocated subject and the subject position results in
the subject position being spelled out with a resumptive pronoun:
(33) [ForceP quand [FinP ma petite sœurj [FmP le soir [Fin. QU' [elle, dort]]]]]
In the absence of intervening head positions, the complementizer is allowed to

move to Force. (32) is then analyzed as Misplaced QUE with the upper copy
either inappropriately deleted, as discussed above, or independently deleted by
the Doubly-Filled Comp Filter. If no copy of the complementizer is deleted, we
derive example (5) as an instance of Intrusive QUE.4,5
In our data, no logical options other than those in (27) have been observed
(children ofthat age never produce a construction of type [DP que ec]). We take
these restrictions on the realization of the complementizers to show that Intrusive
and Misplaced QUE's are occasional performance errors, that is, errors in
applying the computational operations required to derive a construction acquired
by the child. This explains the apparent gap between the first emergence of
CLLD (2;0-3;0) and the occurrence of the errors discussed here (3;5-6). The type
of performance error we have in mind is only expected when the competence
system has acquired the construction but where the operations are not
streamlined yet so that they may tax the computational system. In the present
case the embedding of CLLD in a subordinate clause may require all the
computational resources of the child, so that in cases of stress or other cognitive
load, early Merge errors are produced.
5.3 The case of Topic

We proposed that Intrusive and Misplaced QUE constructions result from
the eiToneous merge of a complementizer in Fin and a Topic in its Spec,
producing a structure constructed on the model of a pseudo-relative. The error in
4
With two topic DPs, the construction in (33) is structurally similar to (i), observed in some North
western Italian dialects, here exemplified by Ligurian (Paoli 2001: ex. (17)):
(i) A Maria a creada che ti a-u Gianni che ti ghe l'agi za dato.
The Maria SCL believe THAT you to-the John THAT SCL to-him L.have already given
"Mary thinks that you have already given it to John."
Note, however, that these dialects allow two overt instances of the complementizer che only when the
matrix verb selects the subjunctive mood. Furthermore, the lower complementizer is also allowed
between the two topics (i.e. between ti and α-u Gianni). For these reasons, we think that this is a
distinct phenomenon involving, perhaps, Merge of distinct (homophonous) heads. Despite these
differences between examples like (i) and child French, both types of structures involve split CPs,
and we take them to show that the application of Rizzi's system is on the right track here.
3
Alternatively, (32) could involve two topics in Spec,Top, with Top heads blocking movement of
QUE from Fin to Force. The Force feature would then be checked by quand merged in Spec,ForceP.
Topic positioning may be accounted for by Jakubowicz' (1999) Hypothesis on

Computational Complexity (HCC). According to the HCC, syntactically
necessary functional phrases which are part of the obligatory functional skeleton
are computed more easily than semantico-conceptual ones, present only in some
sentences. Given that Topic and Focus are 'optional' parts of the embedded CP
while Fin and Force are part of the functional skeleton, the HCC implies that
Topic and Focus phrases are more complex to compute. This increased
computational burden may lead to errors in merging these elements. The HCC
also accounts for the fact that Fin and Force are projected in Intrusive and
Misplaced QUE constructions, while Topic is not.
Note now that the positioning of a topic in Spec,FinP results in a mismatch
of features in FinP, as a topic element is in the Spec position of a non-topic
phrase:
(34) ... [ Fm pDP[+Top]que[ IP ...]]
Interestingly, even at five and six, children continue to experience difficulties

with agreement, be it subject-verb or adjective-noun (Méresse-Polaert 1969),
showing that they do not have a perfect control over all aspects of feature
checking. We propose then that Intrusive and Misplace QUE's disappear when
the child gets control over Agree. As the child gains full control over feature
checking, feature mismatch is avoided. Whereas the performance preference for
early Merge is 'violable', feature checking is a Condition on Convergence of
derivations. This forces the child to project a TopP in order to merge the
dislocated DP in Spec,TopP As a fully complex CP does not allow movement of
the complementizer from Fin to Force, the (overt) complementizer is then
merged in Force and only one copy of it is in CP. It follows that only one copy
can be phonetically realized.
6. Independent Evidence for the Proposal

Considering complex CPs in child French, we have shown that children
experience difficulties when a Topic intervenes between the two interfaces in the
CP system. The facts add to the evidence from other languages and from other
kinds of clauses that the acquisition of structural aspects (e.g. CP) as well as the
application of some primitive operations (e.g. Merge and Delete) may result in
problems. For example, Schonenberger (1996) employs a simply-split CP to
account for systematic Verb Second phenomena in embedded clauses, produced
by Swiss German children. Bearing in mind that the finite verb in embedded
clauses appears in final position in adult Swiss German, the complementizer
followed by the finite verb is non-adult-like (her ex. 28a):
(35) Chas ch mer achli Gift geh, [ForccP dass [FwP w erdet¡ [¡P mini Ohre
can me some poison give THAT turn my ears
au bru t; ]]].
also brown
"You can give me some poison so that my ears turn brown too."
In root questions, two- to four-year-old English-speaking children

occasionally produce double tensing errors of the type illustrated in (36) (data
from 'Grady 1997: 161). We view these Intrusive and Misplaced Tenses as
failure or misapplication of Chain Reduction after movement of {PAST} and will
(this is followed by ¿/o-support and affix-hopping in (37a)):
(36) a. What did you brought?

b. Where 0 the other Joe will drive?
(37) a. [ForceP what {PAST} [IP you {PAST} bring ]]
b. [ForceP where will [IP the other Joe will drive ]]
We take these examples to provide support for the type of analysis we have
proposed here.
7. Conclusion
We have assumed that the internalized grammar of the children is adult-like
with respect to the relevant structural aspects of complex CPs and claimed that it
does not allow them to freely produce Intrusive and Misplaced QUE
constructions. We suggested that these constructions result from processing
errors due to performance limitations and immature control of feature checking.
Under conditions taxing to the computational system, children occasionally
produce simply-split CP constructions by means of 'early' Merge. This may
result in improper Chain Reduction. We have linked these errors to other errors
in the production of complex CP constructions in child language.
More generally, Intrusive and Misplaced QUE constructions provide
evidence in favor of Rizzi's (1997) split-CP hypothesis by showing that the
lower interface of the complementizer system surfaces under certain conditions.
They also provide psycholinguistic evidence in favor of the Basic-operation
Hypothesis, in general, and the analysis of movement in terms of operations such
as Copy, Merge, Chain Formation, and Chain Reduction, in particular. In this
respect, the study of language acquisition provides a window to some generally
invisible operations of the computational system CHL and is thus well worth our
attention for the construction of a theory of grammar.
References
Aoun, Joseph, Lina Choueiri & Norbert Hornstein. 2001. "Resumption, Movement, and
Derivational Economy". Linguistic Inquiry 32.371-403.
Bouvier, N. & F. Platone. 1976. "L'étude génétique de la construction d'une détermination
■linguistique complexe: l'expression d'un même contenu par des enfants d'âges
différents." Cahiers du CRESAS 16A. Paris: Institut National de Recherche
Pédagogique.
Cat, Cécile de. 2002. Dislocations in adult and child French. Doctoral dissertation.
University of York.
Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press.
Côté, Marie-Hélène. 1998. "Quantification over individuals and events". Proceedings of
the West Coast Conference on Formal Linguistics 17.147-161.
Haegeman, Liliane. 1994. Introduction to Government & Binding Theory. (2nd ed.)
Oxford: Blackwell.
Halle, Morris & Alec Marantz. 1993. "Distributed Morphology and the Pieces of
Inflection". The View from Building 20. Essays in Honor of Sylvain Bromberger, ed.
by Kenneth Hale & Samuel K. Keyser, 111-176. Cambridge, Mass.: MIT Press.
Hulk, Aafke. 1995. "L'acquisition du sujet en français". Recherches linguistiques de
Vincennes 24.33-53.
Jakubowicz, Celia. 1999. "Functional categories in (ab)normal language acquisition".
Paper presented at the GALA conference at Potsdam.
Kayne, Richard. 1976. "French Relative que". Current Studies in Romance Linguistics,
ed. by F Hensey & M. Lujan, 255-299. Washington, D.C.: Georgetown University
Press.
Labelle, Guy. 1976. "La langue des enfants de Montréal et de Paris." Langue Française
31.55-73.
Labelle, Marie. 1989. Prédication et mouvement: l'acquisition des relatives chez les
enfants francophones. Ph.D. diss., Ottawa University.
Labelle, Marie. 1990. "Predication, WH-Movement, and the Development of Relative
Clauses". Language Acquisition 1.95-119.
Labelle, Marie. 1993. "Intrusive or Misplaced que in the language of French-speaking
children". Proceedings of the Twenty-fifth annual child language research forum, ed.
by Eve V. Clark, 252-264. Standford: CSLI Publications.
Labelle, Marie. 2000. "Explorations on the Acquisition of the Left Periphery". Paper
presented at Indiana University.
Labelle, Marie & Daniel Valois. 1996. "The status of post-verbal subjects in French child
language". Probus 8.53-80.
MacWhinney, Brian. 1991. The CHILDES project: Tools for Analyzing Talk. Hillsdale,
N.J.: Lawrence Erlbaum Ass.
Mayer, Judith W., Anne Erreich & Virginia Vahan. 1978. "Transformations, basic
operations and language acquisition". Cognition 6.1-13.
Méresse-Polaert, Janine. 1969. Etude sur le langage des enfants de 6 ans. Neuchatel:
Delachaux et Niestlé.
Nuneš, Jairo. 1999. "Linearization of Chains and Phonetic Realization of Chain Links".
Working Minimalism, ed. by Samuel David Epstein & Norbert Hornstein, 217-249.
Cambridge, Mass.: MIT Press.
Nunes, Jairo. 2001. "Sideward Movement". Linguistic Inquiry 32.303-344.
'Grady, William. 1997. Syntactic Development. Chicago: The University of Chicago
Press.
Paoli, Sandra. 2001. "Mapping out the sentence: Evidence from North-western Italian
Varieties". Paper presented at Going Romance 2001, Amsterdam.
Rizzi, Luigi. 1990. Relativized Minimality. Cambridge, Mass.: MIT Press.
Rizzi, Luigi. 1997. "The fine structure of the left periphery". Elements of grammar, ed. by
Liliane Haegeman, 281-337. Dordrecht: Kluwer Academic Publishers.
Schonenberger, Manuela. 1996. "Why do Swiss-German Children like Verb Movement
so much?". BUCLD 20 Proceedings, ed. by Andy Stringfellow, Dalia Cahana-
Amitay, Elizabeth Hughes and Andrea Zukowski, 658-669. Somerville, Mass.:
Cascadilla Press.
Stromswold, Karin. 1990. Learnability and the acquisition of auxiliaries. Ph.D. diss.,
MIT.
Uriagereka, Juan: 1998, Rhyme and Reason: An Introduction to Minimalist Syntax.
Cambrigde, Mass.: MIT Press.
PLURAL INDEFINITE DPS
AS PLURAL-POLARITY ITEMS
BENJAMIN SPECTOR
Université de Paris VII/Laboratoire de Linguistique Formelle
1. Introduction
In what follows, I try to account in a principled way for the interpretation of
French DPs introduced by the so-called 'partitive article' des (hereafter, 'des-
DPsV·1 Unlike other Romance languages, French does not generally license bare
plurals, except in predicative structures, coordinated NPs (NPpl et NPpl), and
enumerations. Des-OVs are often said to be the French counterparts of Romance
bare plurals, because they must generally take narrow-scope with respect to all
types of operators. It turns out, however, that des-DYs are sometimes able to take
wider scope than would be expected if they were exactly equivalent to, say,
Spanish bare plurals:
(1) Je veux acheter des chemises qui sont en vente dans ce magasin.
I want to buy des shirts that are-lND sold in this shop
"I want to buy some shirts which are sold in this shop."
According to prescriptive grammar, the indicative mood indeed forces des

chemises to take wide-scope, and des chemises must be interpreted 'de re', e.g
takes scope over the intensional verb. The Spanish counterpart of (1) requires the
introduction of unas and some Italian speakers report they need to use delle/dei
in order to get the same reading (though this does not seem to be true in standard
Italian). In spoken French, narrow-scope readings are allowed in contexts like
(1), but it remains true that wide-scope readings require the indicative mood.
It also turns out that des-OVs are allowed to take scope as far as singular
indefinites can in certain contexts:
1
In this paper, plural indefinites' is meant to refer only to bare plurals and ¿fes-DPs, not to other
plural indefinites. I depart from the usual terminology.
296 BENJAMIN SPECTOR
(2) Si des cousins à moi meurent, je serai riche.

if des relatives of mine die, I will be rich
"If some relatives of mine die, I will be rich."
(2) has a reading according to which there are relatives of mine such that, if
they die, I will be rich. Des-DPs are therefore able to take scope over an /^clause
to which they belong, e.g to escape a scopal island, just like singular indefinites
are (see Reinhart 1997).
But des-DPs are nevertheless forced to take narrow-scope in a lot of
contexts:
(3) Tous les garçons ont lu des livres.

all the boys have read des books
"All the boys have read books."
(3) can never be interpreted as meaning that there are certain specific books
that all the boys have read.
Given the fact that wide-scope readings seem quite constrained for des-
DPs, the question arises how to account for this variability (e.g. des-DPs are in
most contexts equivalent to, say, Spanish bare plurals, but are not always).
In what follows, I will connect the scopal behaviour of des-DVs to another
aspect of their interpretation: although morphologically plural, des-DVs are very
often interpreted as number-neutral:
(4) - Avez-vous des enfants ? - Oui, j'en ai un.

"Do you have children? Yes, I have one child."
(4) shows that a des-DV can receive an 'at-least one' interpretation, even
though it is morphologically plural. And for (3) to be true, it is sufficient that
each of the boys has read one book. Des livres is therefore interpreted as a
dependent plural, just like wheels in (5):
(5) Monocycles have wheels
There are contexts, however, in which des-OVs must receive a genuine

plural interpretation:
(6) Il y a une heure, Pierre a vu des filles.

"One hour ago, Pierre saw some girls."
PLURAL INDEFINITE DPS AS PLURAL-POLARITY ITEMS 297
(6) would be a false sentence if Pierre saw only one girl one hour ago.
The hypothesis I am going to present in this paper is basically the
following: des-DPs are forced to be interpreted as dependent plurals whenever
they can, e.g. whenever there is an item (the 'licenser') on which they can
depend; the class of licensers will include not only plural DPs, but also
intensional verbs (plural quantifiers over possible worlds) and some abstract
aspectual operators. Only when no dependent reading is possible will des-DPs
get a genuine plural interpretation, and be free to take wide-scope.
The paper is organised as follows: in section 2, I present and criticize a
pragmatic account of the plural/number-neutral contrast. In section 3, I present
my proposal and show how it can predict some basic facts about the
interpretation of des-DPs. In section 4, I will show that my proposal sheds light
on the interpretation of des-DPs in habitual and iterative sentences. In section 5,1
will tentatively extend my hypothesis in order to deal with the differences
between French des-DPs and Spanish Bare Plurals. Section 6 is the conclusion.
2. A pragmatic account of the number-neutral/plural contrast

A very intuitive account of the contrast between (4) (repeated as (7)) and
(8) would be to argue that the 'at-least two' interpretation of des enfants in (8)
arises as an implicature of what is in fact an 'at least one' interpretation:
(7) - Avez-vous des enfants ? - Oui, j'en ai un.

"Do you have children? Yes, I have one child."
(8) Pierre a des enfants.
"Pierre has children."
(8) is felt as true only if Pierre has at least two children. Suppose, however,
that des-DPs always convey an 'at least one' interpretation. Then what has to be
explained is not the interpretation of (7), but that of (8): (8) could then be argued
to implicate, rather than to entail, that Pierre has more than one child, because of
the availability of (9):
(9) Pierre a un enfant.

"Pierre has a child."
The pragmatic reasoning giving rise to the implicature would be the

following (roughly): if there had been exactly one child, (9) would have been
uttered instead of (8); hence, (9) is false.
As is well known, questions tend to cancel implicatures. This is why the
number-neutral interpretation of des enfants in (7) is so obvious.
Such a pragmatic account could be extended to dependent plurals in the

following way: plural operators should also be able to cancel the 'at-least two'
implicature which is associated with des-DPs, given the following reasoning:

What kind of interpretation is the availability of (4') below able to block?
(4') Tous les garçons ont lu un livre.

"All the boys have read a book."
In a situation in which each boy read exactly one book, it is reasonable to assume
that (4') would be chosen instead of (4). Hence, a possible implicature of (4) is
that (4') is false:
(4") It is not the case that all the boys read exactly one book each.
In other words, (4) should have the implicature that at least one of the boys read
at least two books, but should not implicate that all the boys did so. Therefore,
the 'at-least two' implicature associated with des should indeed disappear in
sentences like (4), and be replaced with a much weaker implicature, namely,
(4").
There are reasons, however, to doubt that such a pragmatic treatment is
really able to explain the number-neutral reading of des-DPs as arising from the
cancellation of the usual 'at least two' implicature.
First, the reasoning which has just been developed regarding (4) should also
be valid for (10):
(10) Chacun des garçons a lu des livres.

"Each of the boys has read some books."
Yet it turns out that (10) entails that each of the boys read more than one book; I
conclude that in (10), the 'at-least two' interpretation of des livres cannot be
itself an implicature, since such an implicature should be cancelled in this very
context. One has therefore to assume that, at least in some cases, the 'at-least
two' interpretation is really part of the meaning of des-DPs, and is not simply
implicated.
Second, assuming that des has a number-neutral interpretation, it is not that
clear that the singular indefinite determiner un/une should count as logically
stronger than des, given the fact that un/une itself does not have an exactly-one
interpretation. Rather, the exactly-one interpretation associated with un/une

arises itself as an implicature, and un/une is generally compatible with an 'at
least one' reading:
(11) - Avez-vous une cigarette ? - Oui, j'en ai encore trois.

"- Do you have a cigarette ? -Yes, I do. I still have three of them/'
The pragmatic account is therefore not straightforward, and would require a

better understanding of the way implicatures triggered by determiners are
computed.2
3. Des-DPs as PLURAL-polarity items

In order to capture the fact that bare plurals and des-DPs generally force
narrow-scope readings, most theoreticians have assumed that they are not
standard indefinites (which, on the contrary, can take 'long distance' scope). For
instance, it has been suggested that they are names of kinds (Carlson 1978;
Chierchia 1998) or that they denote properties (Dobrovie-Sorin & Laca 2000). In
both cases, their 'semantic type' is different from that of standard quantified DPs:
<e> if they name kinds, <e,t> if they denote properties. But we know at least
another class of expressions which are also restricted to narrow-scope readings,
namely, (negative) polarity items. It is tempting, then, to see whether bare plurals
and des-DPs could be treated as some kind of polarity items. Their distribution
should then be sensitive to the presence of certain licensing elements. And just as
negative polarity items can be thought of as 'semantically deficient', so that they
need to be licensed under the scope of some operator, they would also be
'deficient' in a sense: I will suggest that the discrepancy between plural
morphology and non-plural readings is a reflex of such a deficiency.
When we look at contexts in which des-DPs can be interpreted as number-
neutral, we find, among others, certain w/ze^-clauses.
(12) Quand des Italiens vont à Paris, ils vont visiter le Louvre.
"When Italian people go to Paris, they visit the Louvre."
(12) has a reading according to which when one or more Italian people go
to Paris, he or they visit(s) the Louvre. This is exactly what a number-neutral
reading for des Italiens amounts to.
2
Some recent works show that implicatures are actually not computed simply on the basis of the
global meaning of a given sentence, and that an adequate theory of implicatures is much more
complex than one could have thought. See Chierchia (2001).
A closer look at conditional contexts makes clear that des-NPs in if-or-

w/ze/7-clauses can be number-neutral only if the conditional clause somehow
involves a plurality of events or situations:
(13) Quand des gens sont venus l'an dernier, je les ai rencontrés.
"When people came last year, I met them."
In (13), the when-clmse can be understood as pointing to a presupposed single

event of SEVERAL people coming. It can also be read as referring to several
events of people coming, in which case there may be only one person for each
event. But it is not possible to understand the w/ze/7-clause as denoting a single
event of only one person coming: the number-neutral reading of the des-DP
requires there to be several events.
In view of those facts, it seems quite intuitive to think that number-neutral
readings of des-NPs are licensed, in particular, by contexts which involve some
kind of plural quantification on which des-DU* s can be dependent I now
formulate the proposal itself: des-DP s are morphologically plural but
semantically number-neutral, and they must be interpreted under the scope of
some suitable operator, such as plural DPs, intensional verbs and some 'plural'
aspectuo-temporal operators/ Only if there is no licenser will they be interpreted
as genuine plurals, and then may be specific. More formally, des-DPs are marked
as [+pl] (morphological plural), and [+pl] is licensed within the scope of any
element marked as [+PL] (typically, plural quantified NPs or 'plural' aspectual
markers). [+PL] expresses semantic plural; it can, and even must, be introduced
into the des-DP itself only when there is no other [+PL] which could license the
[+pl] feature.4 In the latter case, a des-DP is not interpreted as number-neutral,
but as a genuine plural.
My proposal makes the following prediction, among others: if a des-DP
takes maximal scope, then it cannot be number-neutral. As far as I know, this
prediction is correct. Consider (1) again:
3
I haven't included negation among the licensers, even though, when a des-DP is licensed below
negation, it must take narrow-scope and be interpreted as number-neutral. The point is that des-DPs
cannot generally appear as objects of negated verbs: des must be replaced with de, except in some
contexts which need to be carefully identified. I won't say anything more about negation in this
paper, because the des/de alternation must first be studied in itself.
4
It is natural to think that the [+pl]-feature marks any noun displaying plural morphology, and that
plural DPs other than des-OPs satisfy the licensing condition internally, thanks to some sort of
agreement between the determiner and the noun (maybe via a Number projection). Plural determiners
except des would all bear the [+PL] feature, which would license the [+pl] feature of the noun. Only
des-DPs, then, would preferably resort to 'external' licensing, internal licensing being a last-resort
option in that case.
PLURAL INDEFINITE DPS AS PLURAL-POLARITY ITEMS 3 01
(1) Je veux acheter des chemises qui sont en vente dans ce magasin.
"I want to buy des shirts that are-IND sold in this shop."
Indicative mood in the relative clause forces [des chemises qui ....] to take scope
over the intensional verb, which therefore cannot count as a licenser for [+pl].
Consequently, [+PL] must occur within the des-DP itself, which prevents it from
being interpreted as number-neutral. This yields the following interpretation:
'There are at least two shirts which are sold in this shop and which I want to
buy". Consider now (6):
(6) ll y a une heure, Pierre a vu des filles.

"One hour ago, Pierre saw some girls."
Des filles is interpreted as a genuine plural, e.g. (6) is false if Pierre saw exactly
one girl. This is again predicted: since there is no licenser in the sentence, des
filles must contain a [+PL] feature. In (4) (repeated below), the fact that the des-
DP can be licensed by the subject rules out the possibility of interpreting it as a
genuine plural ([+PL] can be introduced only 'at last resort'), which in turn
forbids a wide-scope reading:

Note that chaque and chacun do not allow des-DP s to be number-neutral, as

shown by (10) :
(10) Chacun des garçons a lu des livres.

"Each of the boys has read some books."
Therefore, DPs headed by chaque or chacun must not be marked as [+PL],

which is consistent with the fact that chaque-DPs are morphologically singular
and that, contrary to DPs headed by tous les, chaque-DPs are not compatible
with collective predicates, as (14) illustrates:
(14) * Chaque membre de l'équipe s'est réuni.

"Each member of the team gathered."
In this case, the necessary narrow-scope reading of the object cannot be

explained in terms of plural polarity. I will assume that it is simply a
consequence of the distributive nature of each.
The fact that, contrary to tous les, chaque and chacun are not marked as
[+PL], is also able to account for the following contrast:
(15) a. Tous les garçons étaient en train d'embrasser des filles lorsque
je suis entré.
"All the boys were kissing girls when I came in."
b. 11 Chaque garçon était en train d'embrasser des filles lorsque je
suis entré.
"Each boy was kissing des girls when I came in."
In (15b), des filles cannot be interpreted as number-neutral. Therefore, each boy

must have been kissing several girls at the same time for the sentence to be true.
Assuming that it is physically close to impossible to kiss several people at the
same time, (15b) is expected to be odd for pragmatic reasons. On the other hand,
(15a) is fine because des filles can and must be interpreted as number-neutral,
since its [+pl] feature is c-commanded by the [+PL] feature of the universally
quantified plural subject.
It has to be noticed, however, that such 'dependent plurals' could of course
be analysed differently. Consider (4) again :

(4) could be argued to receive a cumulative interpretation : (4) would assert

the existence of a set Y of several books such that the set of all the boys - call it
X - and Y stand in the following relation: each member of X has read at least one
member of Y, and each member of Y has been read by at least one member of X.5
It is easy to see that such an interpretation is very close, in terms of truth
conditions to a narrow-scope number-neutral interpretation for des livres. In fact,
if it is added that the boys cannot have read exactly the same book each, then the
number-neutral narrow-scope interpretation is exactly equivalent to the
cumulative reading: assume that each boy has read at least one book, and that
they have not read the same book each. Take the set Y of all the books read by at
least one boy; this set includes more than one member, and every boy is
connected to a member of Y, while, by definition, each member of Y has been
read by a boy. Therefore the number-neutral narrow-scope reading entails the
5
I am using the definition of cumulativity that can be found in Szabolcsi 1997: 64: Two quantifiers
stand in the cumulative relation if they introduce two sets (witness sets of the quantifiers) X and Y
such that "every element of X (is) connected to some element of Y, and ... every element of Y (is)
connected to some element of X".
cumulative reading. And it is easy to see that the entailment also holds in the
other direction.
If the cumulative analysis were correct, there would then be no scopal
asymmetry between the subject and the object. But one could expect, then, that
the passivization of (4) would preserve its truth-conditions:
(4"') Des livres ont été lus par tous les garçons.
"Books have been read by all the boys."
As a matter of fact, this prediction is not borne out: for (4'") to be true,
there must be several books such that each of them has been read by all the boys
(distributive reading). I conclude that the subject and the object, in (4), cannot be
scopally independent, and that the subject takes scope over the object.
4. Event-dependent plurals
In the previous section, it has been shown that plural DPs count as licensers
for the [+pl] feature. I will now argue that it is also possible to license [+pl] in a
des-DP by interpreting it under the scope of some aspectuo-temporal 'plural'
operator. Des-DPs can indeed be dependent not only on other plural DPs, but
also on expressions which denote pluralities of events (an idea which was
already illustrated by (12)).
4.1 Differentiated scope
(16) he chasseur a tué des lapins à plusieurs reprises.

"The hunter has killed rabbits repeatedly."
(16) means that what the hunter did was kill a different rabbit at different times.
Des lapins is therefore interpreted as a number-neutral narrow-scope indefinite.
Consider now (17):
(17) ??Le chasseur a tué un lapin à plusieurs reprises.

"The hunter killed a rabbit repeatedly."
What is odd in (17) is that it suggests that the hunter repeatedly killed the same
rabbit. Un lapin must take scope over à plusieurs reprises. The peculiar scopal
behaviour of plural indefinites is the reason why such constructions are said to
exhibit a phenomenon of 'differentiated scope'.
It is quite natural to treat the indefinite plural as 'dependent' on the
adverbial phrase, in the same way as a plural indefinite can be dependent on
another plural DP I will therefore assume that [+PL] can mark not only DPs, but
also all sorts of categories, including adverbials and aspectual/verbal projections.
In order to force un lapin to take wide scope in (17), I will assume that some
contexts not only allow but also FORCE scope-dependent readings to be
expressed by means of dependent plurals, e.g anti-license, among others,
singular indefinites.6
Of course, not all licensing contexts are also anti-licensing contexts for
singular indefinites:
(18) Tous les professeurs portent des cravates.

"All the professors wear ties."
(19) Tous les professeurs portent une cravate.
" All the professors wear a tie."
As shown by (18) and (19), tous les professeurs is able to license a number-
neutral des-D? without anti-licensing singular DPs, since (19) does not mean that
all the professors wear the same tie. But the fact that the anti-licensing contexts
are only a subset of the licensing contexts is not a problem in itself, since that is
exactly what we observe with other polarity phenomena:
(20) If Peter saw anyone yesterday, I will know it.

(21) If Peter saw someone yesterday, I will know it.
(20) shows that anyone can be licensed by being under the scope of a
conditional, while (21) shows that someone is not anti-licensed by this very
context which licenses anyone. That someone is not anti-licensed is clear from
the fact that (21) need not be interpreted as: "there is someone such that if Peter
saw him yesterday, I will know it".
4.2 Another event-dependency
(22) Dans ma vie, j'ai déjà vu des soldats tuer un prisonnier.

"In my life, I have already seen soldiers killing a prisoner."
(22) can be taken as true if I have been a witness of several events in which only
one soldier killed a prisoner. Des soldats is interpreted as dependent on some
aspectual plural operator, and this operator, on the other hand, distributes over un
6
This idea simply exploits the parallel with negative polarity items. For instance, while any, in
English, is licensed by certain environments, such as negation, some is anti-licensed by negation.
prisonnier. An account in terms of cumulativity would have to assume that, on

the one hand, events and soldiers are in a cumulative relation, while, on the other
hand, des soldats distributes over un prisonnier. But this distributive relation
would be quite a mystery, since it is not licensed in simple sentences:
(23) Des soldats ont tué un prisonnier > only one prisoner
"Some soldiers killed a prisoner."
Another possible analysis would be to assume that the subject of the

infinitive clause is a genuine plural which takes scope over the matrix verb,
yielding the following reading: "there is a set of soldiers such that for each of its
members, I have seen him killing a prisoner". In that case, the fact that each of
the killing-events may involve only one soldier would not be derived from a
number-neutral semantics for the des-DP, but rather from the fact that des-DPs
are able to take maximal scope and then to distribute over the whole sentence.
Such a mechanism is certainly at play in the following example:
(24) J'ai déjà vu trois soldats tuer un prisonnier.

"I have already seen three soldiers killing a prisoner."
(24) can mean something like "there are three soldiers such that I have seen each
of them killing a prisoner (possibly on three different occasions)".
However, it can be shown that such a 'wide-scope' analysis is not the right
one for (22): if it were, then the apparent number-neutral reading would imply
that the des-DP scopes over anything intervening between the matrix verb and
the infinitival clause. If such a wide-scope reading turns out to be impossible, an
apparent number-neutral reading should be impossible too. For instance, (25)
below would be predicted to have at most the two following readings:
(25) J'ai vu à plusieurs reprises des soldats tuer un prisonnier.

"I have seen repeatedly soldiers killing a prisoner."
(a) wide-scope reading: "There are several soldiers such that, for each of them, I
have repeatedly seen him killing a prisoner".
(b) narrow-scope reading: "I have repeatedly seen events in which several
soldiers killed a prisoner"
As a matter of fact, the (a) reading is impossible, while a narrow-scope
number-neutral reading turns out to be available: "I have repeatedly seen events
in which one or more than one soldier killed a prisoner". This provides support
for an analysis according to which des-DPs can be interpreted as number neutral
if they occur under the scope of a plural aspectual operator quantifying over
events.
4.3 Habitual and iterative sentences

Habitual sentences can be divided into two types, depending on whether
they are generalizations about situations or simply assert the existence of a
certain habit (sentences belonging to the second type will be referred to as
'simple habitual sentences' in what follows). These two types are illustrated
below:
(26) Quand Paul est nerveux, il fume des cigarettes.

"When Paul is nervous, he smokes cigarettes."
(27) Paul fume des cigarettes.
"Paul smokes cigarettes."
One of the differences between the two types is that the first one allows the
direct object to be a singular DP, while the second one does not:
(28) Quand Paul est nerveux, il fume une cigarette.

"When Paul is nervous, he smokes a cigarette."
(29) Paul fume une cigarette.
"Paul smokes a cigarette."
(29) can only be episodic, and its English counterpart is simply deviant.
This peculiar behaviour of simple habitual sentences is reminiscent of the
phenomenon of differentiated scope. If the comparison between simple habitual
sentences and those which show differentiated scope is sound, then habitual
predicates should not really disallow singular objects; rather, singular objects
should be allowed but should always take scope over the habitual predicate:
(30) Jean conduit une voiture.

"Jean drives a car."
Unless the previous discourse or the general context provides a contextual

restriction, (30) can only be interpreted as meaning that that there is a truck that
John drives. If I want to say that Jean is a car-driver, I have to utter (31):
(31) Jean conduit des voitures.

"Jean drives cars."
The reason why (29) cannot be an habitual sentence is then simply that its
meaning would be something like "there is a certain cigarette that Paul has the
habit to smoke", which is strongly deviant because "smoking a certain cigarette"
is a once-only predicate.
I therefore suggest the following mechanism: simple habitual predicates are
associated with an aspectual [+PL] operator. As such, they are able to license
des-DPs objects, which in turn are interpreted as number-neutral. Singular DPs,
on the other hand, are anti-licensed by [+PL], and therefore must take scope over
it.
It is interesting to note that numerical DPs must also take scope over
habitual predicates:
(32) J'enseigne à deux étudiants.

"I teach two students."
If there is no contextual restriction, (32) can only mean that there are two
students whom I teach, and not that I have the habit of teaching pairs of students,
even though there would be nothing unrealistic in having such a habit.
Numerical DPs display a similar behaviour in constructions which illustrate
differentiated scope:
(33) ??Le chasseur a tué deux lapins à plusieurs reprises.

"The hunter killed two rabbits repeatedly."
(33) is deviant because it entails that there are two rabbits that the hunter killed
several times. I am then led to assume that numerical DPs, and maybe other
types of DPs, are also anti-licensed by [+PL] in some contexts.
Concerning the semantics of simple habitual sentences, I will simply
assume that habitual predicates contain an habituality (Hab) operator which
asserts the existence of a plurality of events instantiating the predicate.7 A more
explicit semantic characterization would be required in order to account for the
dispositional flavour of habitual sentences. What is important is that there is a
similarity between the Hab operator and plural existential DPs. The Hab operator
can be naturally understood as marked as [+PL]. This analysis is actually not
restricted to habitual sentences and differentiated scope, but can be extended to
other types of iterative sentences. In all of them, we find plural objects which are
7
The semantic analysis proposed by Chierchia (1995, 1998) and Krifka et al. (1995) has been
convincingly criticized in Dobrovie-Sorin (2001); Delfitto (2000). Due to lack of space, I cannot
develop here my own criticism of Chierchia and Krifka et al.'s views.
actually interpreted as number-neutral, while singular indefinite objects are anti-

licensed:
(34) J'ai passé la journée à fumer des cigarettes.

"I spent the whole day smoking cigarettes."
(35) *?J'ai passé ¡ajournée à fumer une cigarette.
"I spent the whole day smoking a cigarette."
(35) entails that there is a single cigarette that I have been smoking during
the whole day.
Habitual and iterative predicates can also license [+pl] subjects and anti-
license singular indefinites subjects, if these subjects are postverbal, as in
locative inversion constructions:
(36) a. Dans ce restaurant mangent des gens célèbres.

"In this restaurant eat famous people."
b. Dans ce restaurant mange un homme célèbre.
"In this restaurant eats a famous man."
While (36a), if read as an habitual sentence, means that the restaurant is such that
events in which one or more famous people eat in that restaurant are usual, (36b)
can be habitual, but it then entails that there is a certain man who happens to eat
quite regularly in the restaurant. Post-verbal subjects thus turn out to behave
exactly like objects of habitual sentences.
My proposal also sheds light on the following contrast:
(37) ??Des romans américains se lisent généralement avec plaisir.

des novels American se read generally with pleasure
"When one read American novels, it is generally with pleasure.
(38) On lit généralement des romans américains avec plaisir.
one reads generally des novels American with pleasure
"When one read American novels, it is with pleasure."
A des-DP cannot easily be the subject of an habitual sentence. To the extent to

which (37) is acceptable, the subject is interpreted as generic. But it can be
shown that for a des-DP to be felicitously interpreted as generic, it must be the
case either that the generalisation is over groups of several individuals (which
motivates the use of a plural subject), as in (39), or that, for whatever reason,
plural morphology is 'neutralized', e.g., the des-DP is number-neutral (see,
among others, Dobrovie-Sorin 2002):
(39) Des enfants ensemble font beaucoup de bruit.

"Children who are playing together are noisy."
The contrast between (37) and (38) is therefore expected: in (37), the use of
a plural DP is not motivated, since the sentence clearly expresses a generalisation
about individual novels, not over groups of novels, while in (38), the des-DP
object occurs inside an habitual predicate, which makes it number-neutral. Since
plural morphology does not correspond anymore to semantic plural, and
therefore does not need to be semantically motivated, the sentence can then be
interpreted as: "When one reads an american novel, it is generally with
pleasure".
5. Bare Plurals in Spanish

Let us assume that Spanish Bare Plurals are just like French des-DPs,
except that the [+PL] feature cannot be introduced in a bare plural, not even as a
last resort. Namely, Bare Plurals would be marked as [+pl] (morphological
plural), which must be licensed by a c-commanding [+PL], but no [+PL] can be
introduced in a bare plural, not even as a saving device. While French allows a
[+PL] feature to mark a des-DP, Spanish has to resort to an overt determiner
(unos/unas). It is thus expected that whenever a French des-DP receives a
genuine plural interpretation and takes wide-scope (which means that [+PL] has
been introduced as a last resort), - then no bare plural can occur in the
corresponding Spanish sentence. This could explain why a 'de re' interpretation
is impossible for Spanish bare plurals which are c-commanded by an intensional
verb: while indicative mood forces, in (1), the introduction of [+PL] within the
des-DP, such a move is just impossible for Spanish Bare Plurals. Another
prediction is that predicates which denote single events should disallow bare
plural objects, except if there is a plural subject; this turns out to be the case:
(40) a. A las tres en punto, todos los policías estaban interrogando

sospechosos/unos sospechosos.
"At three o'clock exactly, all the policemen were questioning
suspects/some suspects."
b. A las tres en punto, cada policía estaba interrogando
sospechosos/unos sosphechosos.
"At three o'clock exactly, each policeman was questioning
suspects/some suspects."
In (40a), the subject is marked as [+PL], and is therefore able to license a

bare plural object. In (40b), the subject is not marked as [+PL], and, since the
predicate itself denotes a single event, no bare plural object is licensed.
Consider now the following contrast:
(41) a. Ya he visto policias/a unos policías.

"I have already seen policemen/some policemen."
b. Alas tresenpunto vi ** a policías/**policías/unos policías.
"At three o'clock exactly, I saw policemen/some policemen."
The fact that the past simple can only refer to a single event explains why no
bare plural object is licensed in (41b). The composed past tense in (41a) is
compatible with a plurality of events, and the predicate licenses a bare plural
object.
Another argument showing that Spanish Bare Plurals are, in some respect,
plural polarity items, comes from the following contrast:
(42) Unos soldados quemaron unas casas/*casas.

"Some soldiers burnt some houses/houses."
(43) Unos soldados quemaron casas, otros quemaron pisos.
"Some soldiers burnt houses, others burnt apartments.
An unos-subject cannot easily be distributive, unless it is 'contrastive', as in (43).

Only if it is able to distribute will it count as a licenser for a bare plural object,
since otherwise it cannot be said to take scope over it. The contrast between (42)
and (43) is therefore expected. In (43), casas is interpreted as number-neutral.
On the other hand, it is expected that DPs headed by unos/unas cannot be
number-neutral. This prediction seems to be realised. Consider for instance (44)
and (45):
(44) Ya he visto (a) soldados matar a un prisionero.

"I have already seen soldiers killing a prisoner."
(45) Ya he visto (a) unos soldados matar a un prisionero.
"I have already seen some soldiers killing a prisoner."
For (44) to be true, I must have seen several events in which one or more than
one soldier kills a prisoner, while (45) entails that what I have seen are events in
which a group of soldiers kills a prisoner.
Moreover, unos/unas-DPs are anti-licensed by the [+PL]-feature in the very
contexts which anti-license singular indefinites: this is why habitual and iterative
predicates disallow unos/unas-DPs:
(46) Fumo unos cigarillos.

"I smoke some cigarettes."
As expected, (46) only has an episodic reading, unless the object is understood as
taxonomic, in which case it must take wide-scope: "there are kinds of cigarettes
that I smoke".8 It is interesting to note that while Italian bare plurals do not seem
to obey the same restrictions as Spanish bare plurals (at least for most speakers),
dei/delle-DPs behave just like unos/unas-DPs, as shown by (47):
(47) Fumo deile sigarette.

'T smoke some cigarettes."
(47), if read as an habitual sentence, could only get a taxonomic wide-scope

interpretation: "there are kinds of cigarettes that I smoke". Unos/unas-DPs also
take wide-scope in differentiated-scope sentences:
(48) ?? Ese soldado ha matado repetidamente unos enemigos.

"This soldier has killed some enemies repeatedly."
(48) only yields the 'absurd' reading according to which some enemies got killed
repeatedly.
I conclude that viewing Spanish bare plurals as PLURAL-polarity items
sheds light on some aspects of their interpretation and distribution. Moreover, the
differences between them and French des-DVs can be attributed to the fact that
French has the last resort option of introducing a [+PL] -feature in a des-DP if
necessary, while this is impossible for Spanish Bare Plurals. It cannot be said,
however, that this view explains everything about the distribution of Spanish
Bare Plurals. For instance, it cannot explain why the following sentence, among
others, is grammatical:
(49) Conoce linguistas.

"He knows linguists."
My proposal is therefore only tentative, and needs to be supplemented with

a more comprehensive account. Yet, it at least captures certain similarities and
differences between French des-DVs and Spanish Bare Plurals, some ofwhich, to
the best of my knowledge, had not been noticed - in particular, the fact that plural
DPs and predicates which denote pluralities of events take part in the licensing of
bare plural objects is, as far as I know, a new observation.
8
Recall that I am only considering habitual sentences of the second type, 'simple habitual sentences',
not those which express generalizations over situations. The following discourse is of course
perfectly acceptable:
(i) "What do you generally do after dinner? I smoke some cigarettes."
Its Spanish counterpat is acceptable too.
6. Conclusion
My proposal is based on the following insight: there is an essential link
between the fact that des-NPs and Romance bare plurals can be interpreted as
number-neutral and their tendency to take narrow-scope with respect to all types
of operators. I have suggested that while des-DPs can be interpreted as true
plurals only at last resort, Spanish Bare Plurals never can (which certainly needs
to be qualified, especially when a relative clause is adjoined to a bare plural). I
have also emphasized that the licensing of number-neutral des-DPs and Spanish
Bare Plurals crucially involves plural quantifiers, intensional verbs and 'plural'
aspectuo-temporal operators. But it is clear that I have not captured all the
contexts which license Bare Plurals in Spanish.
My proposal may turn out to be fully compatible with another line of
research which proved successful, and according to which Bare Plurals denote
properties (e.g. are expressions of type <e, t>), so that they always need to be
'incorporated' into some other expression. What needs to be known is what
exactly the 'incorporating' contexts are. Besides those which have already been
identified (such as predicates which contain spatio-temporal variables), I have
shown that quantified plural objects and predicates which denote pluralities of
events also play a role in the definition of 'incorporating'contexts (in my terms,
'licensing' contexts). An hypothesis which needs to be made, then, is that
semantic incorporation of properties entails number-neutralization, as has been
suggested, for instance, by Dobrovie-Sorin (2002).
If these speculations are correct, my 'last resort' theory concerning French
could be derived from the following assumptions :
(a) Des-DPs are expressions of type <e,t>
(b) Type-shifting at last resort: a type shifting operation can affect a certain
expression only if it is the only way to fix a type-mismatch problem. For
instance, des-DPs are turned into standard plural indefinites only when they
cannot be incorporated
In Spanish, the availability of an overt determiner which is able to turn a
plural DP into an indefinite expression of type « e , t > , t> simply blocks the
possibility of a covert type shifting operation for Bare Plurals. This last idea
needs to be more carefully developed, and is inspired by Chierchia 1998, who
proposes a general principle according to which 'overt' type-shifters always
block 'covert' type-shifting operations.
References
Bosveld-de Smet, L. 1998. On Mass and Plural Quantification. The case of French
des/du-NPs. Doctoral dissertation. Groningen.
Carlson, G 1978. Reference to Kinds in English. Indiana: Bloomington.
Chierchia, G 1995. "Individual-Level Predicates as Inherent Generics''. The Generic Book
ed. by G Carlson & F.J. Pelletier, 176-223. Chicago: University of Chicago Press.
Chierchia, G 1998. "Reference to Kinds across Languages". Natural Language Semantics
6.339-405.
Chierchia, G 2001. "Scalar Implicatures, Polarity Phenomena, and the Syntax/Pragmatics
Interface", Ms., University of Milan.
Delfitto, D. 2000. Genericity in language. Issues of syntax, logical form and
interpretation. Ms.
Dobrovie-Sorin,  2001. "Adverbs of quantification and Genericity". Empirical Issues in
Formal Syntax and Semantics 4, Selected Papers from the Colloque de syntaxe et
sémantique à Paris (CSSP 2001) ed. by . Beyssade, . Bonami, P. Cabredo-
Hofherr & F. Corblin, in print. Paris: Presses Universitaires de la Sorbonne.
Dobrovie-Sorin,  2002. "Generic plural indefinites and (un)selective binding",
forthcoming.
Dobrovie-Sorin, . & . Laca 2000. "Les noms sans déterminant dans les langues
romanes", Ms. Paris7-Paris8.
Giannakidou, A. 1997. The landscape of Polarity Items. Doctoral dissertation. Groningen.
Kleiber, G. 1987. Du côté de la référence verbale : les phrases habituelles. Berne: Peter
Lang.
Kratzer, A. 1995. "Stage-level and Individual-level Predicates". The Generic Book ed. by
G Carlson & F.J. Pelletier, 125-175. Chicago: University of Chicago Press.
Krifka, M. & al. 1995. "An Introduction". The Generic Book ed. by G Carlson & F.J.
Pelletier, 1-124. Chicago: University of Chicago Press.
Laca, . 1990. "Generic objects: some more pieces to the Puzzle", Lingua 81.25-46.
Reinhart, T. 1997. "Quantifier Scope: how labor is divided between QR and choice
functions". Linguistics and Philosphy 20.335-397.
De Swart, H. 1991. Adverbs of Quantification: a generalized quantifier approach.
Doctoral dissertation. Groningen.
De Swart, H. 1996. "Indefinites and Genericity". Quantifiers, Deduction and Context ed.
by M. Kanazawa,  Pinon & H. de Swart, 171-194. Stanford: CSLI.
Szabolcsi, A. ed. 1997. Ways of Scope Taking. Dordrecht: Kluwer Academic Press.
ON THE STATUS OF THE PARTITIVE DETERMINER
IN ITALIAN*
GIANLUCA STORTO
University of California, Los Angeles
& University of Rochester
1. Introduction: Bare Partitives

Consider the object NPs in the sentences in (1):
(1) a. Ho incontrato degli studenti.

have .1.SG met of the students
"I met some students." (Italian)
b. Ho bevuto della birra.
have.1.SG drunk of the beer
"I drank some beer." (Italian)
 J' ai rencontré des étudiants.
I have met of the students
"I met some students " (French)
d. J' ai bu de la bière.
I have drunk of the beer
"I drank some beer." (French)
Traditional grammars of French or Italian treat NPs of this type as more or less
ordinary indefinite NPs: the noun is preceded by an indefinite determiner - the
partitive determiner - and the whole NP receives an indefinite interpretation.1
As shown by the glosses in (1), these NPs seem to be morphologically
related to partitive NPs: the partitive determiner can be taken to be a complex
morpheme composed of the partitive preposition de/di and the definite article. This
* I would like to thank Daniel Biiring, Ivano Caponigro, Alessandra Giorgi, Carson Schütze, and an
anonymous reviewer for their comments on the arguments presented in this paper. Thanks to Roberto
Zamparelli for sending me a draft of his paper (Zamparelli 2002b).
1
I do not distinguish between NP and DP in this paper.
316 GIANLUCA STORTO
morphological similarity is left unexplained under the traditional assumption that

the morpheme that precedes the noun is a lexical determiner.
The only difference between the NPs in (1) and partitive NPs like those in the
parallel examples in (2) seems to be that no quantifier or numeral appears before
the complex morpheme that precedes the noun.
(2) a. Ho incontrato due degli studenti.
have. 1 .SG met two of the students
"I met two of the students." (Italian)
b. Ho bevuto metà della birra.
have. 1. SG drunk half of the beer
"I drank half of the beer." (Italian)
 J'ai rencontre deux des étudiants.
I have met two of the students
"I met two of the students." (French)
d. J'ai bu la moitié de la bière.
I have drunk half of the beer
"I drank half of the beer." (French)
Building on these morphological properties - and following Chierchia (1998) -
let's adopt the label bare partitives (BPs) to refer to the class of NPs in (l). 2
The issue dealt with in this paper is the following: does the apparent
morphological similarity between BPs and (full) partitives shed light on the nature
of the partitive determiner? can we argue that the partitive determiner is the
morphological composition of the partitive preposition and the definite article?
I address these questions by examining the interpretive properties of BPs in
Italian. Interpretive data bear on the above questions under the following basic
assumption: the semantic interpretation of a phrase is compositionally derived in
terms of the semantic interpretation of the parts it is composed of and in terms
of the way in which these are combined. For the case at hand this assumption
leads to the expectation that if the partitive determiner is the composition of
the partitive preposition and the definite determiner, BPs should display those
semantic properties that, in (full) partitive NPs, can be attributed to the contribution
of the partitive preposition or to that of the definite article.
As a preview, the conclusions of this investigation are that Italian BPs fail this
test: NPs of this type do not display interpretive properties that one would expect
2
As will become clear later in the paper, this does not mean that I agree with Chierchia's proposal that
BPs are partitive NPs. I adopt Chierchia's terminology because the only alternative I know of-des/du-
NPs (e.g. Bosveld-de Smet 1998; Roy 2001) - is less "compact" and, while being quite transparent to
define the relevant class of NPs in French, is much less so when used to refer to Italian BPs.
THE PARTITIVE DETERMINER IN ITALIAN 317
to correlate with the interpretation of the partitive preposition and of the definite
article. This, I suggest, is an argument for doubting that the partitive determiner is
the morphological composition of the partitive preposition and the definite article.
In the first part of this paper I outline some basic interpretive properties of
Italian BPs (§2), and introduce Chierchia's (1998) analysis of Italian BPs and
discuss how it accounts for these properties (§3). In the second part of the paper I
present empirical evidence arguing against Chierchia's conclusion that Italian BPs
are unambiguously true partitives (§4), and point to further empirical evidence
arguing that even the weaker conclusion that Italian BPs can be ambiguously
analyzed as true partitives cannot be maintained (§5). I thus propose to maintain
the hypothesis that the partitive determiner is a lexical indefinite determiner (§6).
2. Are Italian BPs partitive NPs ?

The hypothesis that BPs are a particular type of partitive NP is, at first sight,
compatible with the general observation that both in French and in Italian BPs
are interpreted as indefinite NPs. In full partitives the quantificational force of
the whole NP is determined by the quantifier/numeral that precedes the partitive
preposition, thus the indefinite interpretation of BPs can be accounted for by
arguing that the absence of an overt determiner before the partitive preposition
triggers an existential interpretation of the whole NP.
However, it is known that partitives headed by an indefinite determiner -
differently from other indefinites - are interpreted necessarily as pre suppositional
strong indefinites: they display a series of semantic properties that distinguish
them from non-partitive indefinite NPs in general, and weak indefinites in
particular. If BPs are partitive NPs they should display the same semantic
properties that characterize (full) partitives as presuppositional strong NPs.3
2.1. A note on French BPs

It has been argued in various places in the literature (Delfitto 1993; Roy 2001,
a.o.) that BPs in French are not partitive NPs, because they display interpretive
properties - e.g. obligatory narrow scope with respect to other operators and well-
formedness in existential there constructions - that are parallel to the properties
displayed by bare nouns (BNs) in those Romance languages (e.g. Italian, Spanish)
3
The proper characterization of the weak/strong distinction is still a matter of debate in the literature
(for a brief overview see McNally & van Geenhoven 1998). For present purposes the only crucial detail
is that the ability to interact scopally with other operators, which weak indefinites lack, is partially
independent from the presuppositionality that characterizes (some) strong - i.e. non-weak - indefinites
(partitives, in particular). That is, the possibility is left open that certain indefinites do not qualify as
weak in that they can interact scopally with other operators but still are not necessarily presuppositional.
318 GIANLUCA STORTO
that license this second type of N P These properties argue that French BPs are
weak non-presuppositional indefinites, which rules out the possibility of analyzing
these NPs as partitives. 4
2.2. Italian BPs vs. Italian BNs

The case of Italian BPs is more controversial. Indeed, as shown by Chierchia
(1998), Italian BPs are not necessarily interpreted as weak indefinites: they display
interpretive properties that differentiate them from BNs, which instead display the
typical properties associated with weak indefinites.
(3) a. Non ho visto ragazzi.

not have.l.SG seen boys
"I haven't seen boys."
b. Non ho visto un ragazzo.
not have.l.SG seen a boy
"I haven't seen a boy."
 Non ho visto dei ragazzi.
not have. l.SG seen of the boys
'T haven't seen any/some boys."
In Italian BNs cannot scope above other operators in the sentence (3a) - a
property that Chierchia calls scopelessness - but other indefinites in general (3b),
and BPs in particular (3c), can do so. This observation, albeit sufficient to exclude
the hypothesis that Italian BPs must be interpreted as weak indefinites, does not
force the conclusion that they are presuppositional NPs.
Additional data discussed by Chierchia, however, seem to lead to the
conclusion that Italian BPs must be interpreted as presuppositional indefinites.
(4) a. Non ci sono folletti.
not there are.3.PL elves
"There are no elves." [locative or existential]
b. Non ci sono dei folletti.
not there are. 3. PL of the elves
"There are no elves." [locative only]
4
Ileana Comorovski, Petra Sleeman, and an anonymous reviewer (among others) pointed out to me
that the French facts are more complex, and some of the data proposed to support the conclusion that
French BPs are weak indefinites do not seem to survive further scrutiny. It is not my intention to address
the problem of the interpretation of French BPs in any detail (see Bosveld-de Smet 1998 for a thorough
discussion of French BPs). Above I just intend to point out that even the (probably incomplete) set of
data that support the hypothesis that French BPs are weak indefinites cannot be reproduced in Italian.
c. Non ci sono dei folletti capad di tanto.

not there are.3.PL of the elves capable of so much
"There are no elves that are capable... " [locative or existential]
The there sentence in (4b) licenses only one of two interpretations licensed by the
parallel there sentence in (4a). Both (4a) and (4b) can mean that no elves are in a
particular (contextually specified) place, a statement which is compatible with the
existence of elves in some other place; but only (4a) can alternatively convey the
stronger statement that elves do not exist altogether.
This difference can be explained along the lines of Zucchi (1993): existential
there sentences are infelicitous in those contexts where the intersection of the set
denoted by the N' in the postcopular NP and the set denoted by the coda is already
entailed to be empty or non-empty in the context. The prediction follows that
presuppositional NPs - NPs that introduce presuppositions on the set denoted by
the N' - can be used in existential there sentences only when a coda is present
such that the intersection between the set denoted by the coda and the set denoted
by the N' is not already entailed to be empty or non-empty.
No coda is present in (4b), thus the unavailability of the existential
interpretation must be due to the presuppositional nature of the postcopular NP
dei folletti? The incompatibility of Italian BPs with the existential interpretation
of there sentences leads to the conclusion that these NPs are necessarily interpreted
as presuppositional indefinites.
3. Chierchia's (1998) proposal

Summarizing, the interpretive facts about Italian BPs do not immediately
rule out the possibility that they are a type of partitive NP. If anything, the facts in
(4) seem to argue that Italian BPs are presuppositional indefinites, a property that
characterizes (indefinite) full partitives as well.
Building on the facts in (3) and (4) Chierchia (1998) argues that BPs in Italian
are true partitive NPs: the partitive determiner is the morpho-syntactic composition
of the partitive preposition and the definite determiner, and the interpretation of
BPs is compositional in the sense that the semantics of the partitive preposition
and of the definite determiner contribute in a transparent way to the semantics of
the whole NP.
5
The presence of a coda in (4c) explains the availability of the existential interpretation, in contrast
to (4b). Plausibly, the locative interpretation of there sentences corresponds to the existential
interpretation with the introduction of an implicit coda corresponding to the location.
320 GIANLUCA STORTO
3.1. Syntax
Chierchia proposes that BPs in Italian are structurally like full partitive NPs,
and differ from the latter only in that they are headed by an empty determiner:
The higher determiner position in BPs is filled by syntactic movement of material

from a lower position in the structure: the article incorporates into the preposition,
then the resulting complex incorporates into the phonologically empty noun
Ø[+part], and finally the result incorporates into the higher determiner position.
3.2. Semantics
The semantics of BPs is built compositionally from the semantics of the
elements that are present in the structure in (5b). Chierchia assumes that the
preposition di is semantically empty and that the part-of relation that characterizes
the interpretation of partitives {partitive relation henceforth) is provided by the
interpretation of the noun Ø[+part].6 The latter is interpreted as an entity of type
(e, (e, t)) that applies to the denotation of a definite NP and determines the set of
individuals that are part of the individual denoted by the definite NP.7
Incorporation of the definite determiner into (the preposition and) the
relational noun corresponds to the semantic composition of the partitive relation
and the meaning of the definite determiner:
6
Implicitly I have been and I will continue to assume in my discussion in the text that the partitive
relation in a partitive construction is contributed by the preposition di. For present purposes it is only
relevant that something in the structure of partitives provides the partitive relation, be it a null relational
noun or the partitive preposition. In either case it is expected that such a relation is present in the
semantics of BPs as well, if they are true partitives.
7
Chierchia assumes that the relational noun Ø[+part] imposes the restriction that its complement is a
definite NP. This restriction accounts for the so-called Partitive Constraint (Ladusaw 1982).
In addition, the following step (raising of the N-P-D complex to the higher D
position) involves type shifting of the meaning of the N-P-D complex via the
operator  (defined as  : P = λQx[P(x) ΛQ(x)]), which derives the meaning
of dei used as a determiner given in (6c). 8 This determiner applies to the meaning
of the embedded NP (folletti in (5b)) to obtain a generalized quantifier.
3.3. Consequences
Chierchia's proposal accounts for the facts in (3) and (4). Italian BPs are
existentially quantified NPs, so it is expected that they can interact scopally with
other operators in the sentence. In particular they are not expected to behave
like BNs, which denote kinds and receive an existential interpretation through the
process of Kind Derived Predication.9 This takes care of (3).
Furthermore, the contribution of the semantics of the definite article to the
interpretation of BPs explains the facts in (4): the meaning of the entity of type
(e, t) in (6b) is essentially equivalent to the meaning of the noun folletti with the
addition of the presupposition - introduced by the  operator - that the denotation
of this predicate is non-empty. BPs are thus presuppositional NPs, which explains
the lack of an existential interpretation for the there sentence in (4b).
Summarizing, the analysis proposed by Chierchia (1998)) maintains that
Italian BPs are true partitive NPs: the partitive article dei is the morpho-
syntactic composition of the partitive preposition and the definite article, and the
interpretation of BPs is the semantic composition of (i) the existential quantifier
introduced by the type-shifting operator 3, (ii) the partitive relation and (iii) the
definite article present in the (partitive) structure of these NPs.
4. Chierchia (1998) is not right

If Chierchia's arguments are correct Italian BPs are true partitives, and their
morphological similarity to full partitives is not accidental: the Italian partitive
determiner is the morpho-syntactic composition of the partitive preposition di and
the definite article i.
In this section I argue that this conclusion is not correct. In particular, I show
that two predictions of Chierchia's analysis - those in (ii) and (iii) above - are not
supported empirically. The semantic properties of Italian BPs - despite the facts
8
The function in (6a) applied to an NP meaning derives an entity of type (e, t) (6b), which cannot
constitute an argument. The type-shifted function in (6c), instead, derives an entity of type <<e, t>, t>,
which is a possible argument. Out of the various shifting operators from predicative into argumentai
types only 3 can apply in Italian BPs; see Chierchia (1998) for discussion.
9
This is Chierchia's explanation for the scopelessness of BNs. An alternative is the proposal by van
Geenhoven (1998) according to which BNs denote properties and receive an existential interpretation
by combining with the verbal predicate via the process of Semantic Incorporation.
322 GIANLUCA STORTO
in (3) and (4) - do not support the conclusion that the partitive relation and the
definite article are involved in the semantic composition of these NPs.
4.1. Against partitivity

Chierchia does not discuss explicitly one prediction of his analysis of
Italian BPs. If Italian BPs are true partitives they should not only trigger
the presupposition that the denotation of the embedded noun is not empty - a
presupposition triggered by the definite article in (5b) - but they should have a
partitive semantics. That is, they should display the semantic properties that in
full partitives are associated with the interpretation of the partitive relation.
In his discussion of partitives Chierchia assumes that Ø[+part] contributes the
relation < to the interpretation of the partitive construction. However it has been
argued by Barker (1998) that the semantics of partitives involves the proper part
relation < . This explains why partitives cannot be headed by a definite determiner
unless they are modified by a relative clause (7) (a property that Barker calls anti-
uniqueness).
(7) a. * the two of John's friends

b. the two of John's friends that you pointed out last night
The semantics of the relation < accounts as well for the entailment, which
partitives seem to convey, that the denotation of the embedded noun contains
additional elements that are not in the denotation of the partitive NP itself (8a):
(8) a. # Alcuni dei marziani che sono atterrati nel mío

[some of the Martians that have.3 .PL landed in the my
giardino mi hanno detto che loro sono gli ultimi
backyard] į to me have.3.PL told that they į are. 3. PL the last
rappresentanti della loro specie.
representatives of the their species
b. Dei marziani che sono atterrati nel mio giardino
[of the Martians that have.3 .PL landed in the my backyard]į
mi hanno detto che loro sono gli ultimi
to me have. 3. PL told that theyį are. 3. PL the last
rappresentanti della loro specie.
representatives of the their species
"Some martians who landed in my backyard told me that they are
the last representatives of their species."
(8a) requires the speaker to attribute contradictory beliefs to the relevant

group of Martians 10 which makes the interpretation of the sentence rather marked
(hence the diacritic). This is due to the entailment triggered by proper partitivity
that the denotation of marziani contains other individuals in addition to those
denoted by the whole partitive alcuni dei marziani.
On the other hand no similar semantic restriction holds for (8b): the speaker
can truthfully utter this sentence without attributing contradictory beliefs to the
relevant group of Martians. As (8b) shows, Italian BPs do not seem to be
characterized by the semantics of proper partitivity.
Italian BPs, thus, do not display properties that characterize partitive NPs in
general as a consequence of their partitive semantics. This is unexpected if they are
a type of partitive and the semantic composition of the Italian partitive determiner
involves the partitive relation <.
4.2. Against definiteness

Chierchia's analysis of Italian BPs accounts for the facts in (4) in terms of
the existence presupposition triggered by the semantics of the definite article in
the compositional derivation of the interpretation of the structure in (5b). However,
this presupposition does not seem to be comparable to the existence presupposition
that characterizes full partitive NPs; witness the contrast:
(9) a. # Mi piacerebbe trovare alcuni dei Dodo, ma
to me would please.3.SG find.INF some of the Dodos, but
so che o ramai sono estinti.
know. 1. SG that nowadays are. 3. PL extinct
b. Mi piacerebbe trovare dei Dodo, ma so
to me would please.3.SG find.INF of the Dodos, but know. l.SG
che oramai sono estinti.
that nowadays are.3.PL extinct
"I would like to find some Dodos, but I know that they are extinct
nowadays."
The existence presupposition triggered by Italian BPs - if present at all - is
more easily defeasible than the existence presupposition in parallel full partitives.
10
Or to attribute factually incorrect beliefs to them, if the speaker trusts the relevant individuals to
be speaking truthfully. The precise characterization of the semantic effect in (8a) is irrelevant for the
purpose of the argument in the text. Whatever the nature of this effect, it is clear that a similar effect
does not obtain in the case of (8b). Notice that (8a) has an (irrelevant) interpretation under which this
semantic effect does not arise: this is the interpretation according to which the pronoun loro refers to
the set of individuals denoted by the embedded NP i marziani rather than to the denotation of the whole
partitive NP alcuni dei marziani.
324 GIANLUCA STORTO
This is unexpected if both types of presuppositions have the same source in the
definite determiner embedded in the partitive structure as proposed by Chierchia.
4.3. What about (4)?

Summarizing, neither the prediction that Italian BPs involve the semantics
of the partitive relation, nor the prediction that they involve the semantics of the
definite article seem to survive a closer scrutiny of the interpretive properties of
Italian BPs. The data in (8)-(9) are problematic insofar as the semantics of BPs
is derived compositionally from a partitive structure like (5b). They thus argue
against the analysis proposed by Chierchia (1998).
An alternative analysis according to which the partitive determiner is a
lexical plural indefinite determiner does not lead to the expectation that Italian
BPs are characterized by the semantics of proper partitivity or by the existence
presupposition triggered by the definite determiner. But v/e saw above that the
presuppositional nature of Italian BPs was crucial within Chierchia's analysis to
deal with the data in (4). This explanation is lost under the alternative analysis that
rejects the structure in (5b) for Italian BPs.
In my opinion this does not constitute a shortcoming: Chierchia's account for
the facts in (4) cannot be correct in the first place. That the lack of an existential
interpretation for the sentence in (4b) cannot be due to the presuppositional nature
of the BP dei folletti is indeed argued by the fact that the same BP is compatible
with an existential interpretation of the there sentence in (10):
(10) So che ci sono dei folletti, e prima o poi

know. 1.SG that there are.3.PL of the elves, and sooner or later
ne troveró uno.
of them will find. 1 .SG one
'T know that there are some elves, and sooner or later I will manage to
find one."
The availability of an existential interpretation in (10) is unexpected if dei

folletti is a presuppositional NR Some other semantic property of the sentence
in (4b) must then be held responsible for the lack of an existential interpretation.
The there sentence in (10) differs from the sentence in (4) only with respect to its
polarity: the latter sentence constitutes a negative statement whereas the former
is a positive statement. This suggests the hypothesis that the unavailability of an
existential interpretation for (4b) is due to the negative polarity of this sentence.
Indeed, it seems to be a general fact about Italian that overt indefinites
license an existential interpretation in negative there sentences only when their
determiner/numeral is accented/marked as focus. The sentences in (11) uttered
with normal intonation in a neutral context are quite odd, but become perfect if a
contrast-set for the determiner/numeral is made salient in the context and/or by the
prosodie accent pattern: 11
(11) a. Non c' è ??(neanche) un folletto.

not there is ?? (even) one elf
"There isn't (even) one elf."
b. Non ci sono DUE/TRE/POCHI/MOLTI folletti.
not there are.3.PL two/three/few/many elves
"There aren't TWO/THREE/FEW/MANY elves."
An independent piece of data that becomes relevant in this connection is the

following observation: the Italian partitive determiner dei seems to resist focusing.
As far as I can see this property of the partitive determiner can be related to
the nature of the possible contrast-set for dei: dei folletti, intuitively, can refer
to a set of elves of any cardinality greater than 1, thus the contrast-set for dei
is given by cardinalities between 0 and 1. Somehow sets having one of these
cardinalities seem to be hard to construct for the speaker: 12 if the context does not
provide explicitly 0 or 1 as cardinality with which the denotation of dei should be
contrasted, it is hard for the speaker to construct a contrast-set that would license
focusing of the determiner (Schwarzschild 1993).
That something along these lines might explain the data in (4b) is argued by
the observation that when appropriate contrast sets for dei are provided explicitly
Italian BPs license existential interpretation in negative there sentences as well:
(12) a. Non ci sono DEI folletti, ce n' è soltanto uno.

not there are.3.PL of the elves, there of them is only one
"There aren't SOME elves, there is only one."
?
b. Non ci sono DEI folletti, infatti non ce n' è
not there are. 3. PL of the elves, indeed not there of them is
neanche uno.
even one
"There aren't SOME elves, indeed there isn't even one."
This provides further support for the claim that the facts in (4) should not lead
to the conclusion that Italian BPs are presuppositional. The unavailability of the
11
The examples in (11) should be read as follows: the sentences license an existential interpretation
but become bad - under the same reading - if the pitch accent does not fall on the determiner/numeral
or if the focus operator neanche is left out.
12
Possibly because they constitute degenerate pluralities.
326 GIANLUCA STORTO
existential interpretation for (4b) can be accounted for in terms of the interaction of
general properties of indefinites in negative there sentences and specific properties
of the Italian partitive determiner. And, as shown by (10) and (12), Italian BPs
are compatible with an existential interpretation of there sentences, which - if
the analysis proposed by Zucchi (1993) is correct - indicates that they are not
necessarily presuppositional.
5. What is the structure of Italian BPs?

Chierchia's conclusion that Italian BPs are true partitives seems to be too
hasty. Italian BPs do not display properties that are known to hold of full partitives
in general, and in particular they do not seem to display the semantic restrictions
that one would expect to associate with the semantics of the partitive relation and
with the semantics of the definite article.
But these facts do not force the conclusion that Italian BPs cannot have a
partitive structure: the data in (9)—(12) are problematic only for an analysis that,
like Chierchia's, claims that Italian BPs are unambiguously partitive NPs.
5.1. The first alternative: a syntactic ambiguity

Under the assumption that Italian BPs are structurally ambiguous one can
account both for the data discussed by Chierchia as evidence in favor of his
proposal and for the data presented in the previous section. Presuppositional
interpretations would arise from the possibility for Italian BPs to be assigned a
partitive structure à la Chierchia by the syntactic parser. Instead weak indefinite
interpretations would arise from - for example - their being assigned a structure
like the one proposed for French BPs by Delfitto (1993) or Roy (2001).
This analysis would be able to maintain the conclusion that - at least under
one of the two structures that can be assigned to Italian BPs - the morphological
similarity between BPs and full partitives is expected.
5.2. The second alternative: no ambiguity

Alternatively, one could suggest that Italian BPs are not ambiguous and
are never analyzed as having a partitive structure. The partitive determiner is a
lexical plural indefinite determiner and Italian BPs do not differ from other non-
partitive indefinite NPs like due/alcuni folletti 'two/some elves': they can enter
scope relations with other operators but are not necessarily presuppositional.
This second analysis would be preferable on theoretical grounds in that it
does not need to postulate a syntactic ambiguity to account for the interpretive
properties of Italian BPs. On the other hand this analysis could not maintain
even the weaker account for the morphological similarity between BPs and full
partitives in Italian that is available to the proponents of the first analysis.
5.3. Data from the interpretation of possessives

But are there empirical data that can distinguish between the two
alternatives? I think that there are: data from the interpretation of possessive NPs
seem to support the second analysis over the first one.
It is a well-known fact that the interpretation of possessive NPs can be
context-dependent. E.g. John's dogs in (13) can be interpreted as denoting the
set of dogs that attacked John, rather than the set of dogs belonging to him.
(13) Yesterday John and Paul were attacked by groups of dogs. Unfortu
nately John's dogs were not captured.
But free interpretations like the one exemplified in (13) are not available for
all types of possessive NPs (Storto 2000). In particular, the availability of free
interpretations seems to correlate with the presence of the definite article, as the
following Italian paradigm illustrates: 13
(14) leri Gianni e Paolo sono stati attaccati da gruppi di card;

yesterday Gianni and Paolo were attacked by groups of dogs
a. sfortunatamente i cani di Gianni non sono stati catturati.
unfortunately the dogs of Gianni not have been captured
b. (s)fortunatamente ognuno/uno/alcuni/molti/due dei cani di
(un)fortunately each one/one/some/many/two of the dogs of
Gianni non è/sono stati catturato/i.
Gianni not has/have been captured
 # (s)fortunatamente ogni/un/alcuni/molti/due cane/i di Gianni non
(un)fortunately each/a/some/many/two dog/s of Gianni not
è/sono stati catturato/i.
has/have been captured
Free interpretations are available for definite (14a) and partitive (14b)
possessive NPs, but they are not for (non-partitive) indefinite or quantificational
possessives (14c). Now, if Italian BPs can be assigned a partitive structure as
suggested by the first analysis sketched above it is expected that Italian possessive
BPs should license free interpretations. But, as (15) points out, this is not the case:
Italian possessive BPs do not license free interpretations.
13
The possessive NPs in (14c) cannot be construed as denoting part of the set of dogs introduced in the
context-setting sentence. This explains the degraded status of (14c) as a follow-up to the first sentence.
See Storto (2003) for an analysis of these facts.
328 GIANLUCA STORTO
(15) leri Gianni e Paolo sono stati attaccati da gruppi di card;

yesterday Gianni and Paolo were attacked by groups of dogs;
# sfortunatamente dei cani di Gianni non sono stati catturati.
# unfortunately of the dogs of Gianni not have been captured
The unavailability of free interpretations for Italian possessive BPs is

unexpected under the hypothesis that Italian BPs can be ambiguously analyzed
as partitives and as weak indefinites. Thus (15) argues in favor of the alternative
analysis: Italian BPs are not ambiguous, and should be analyzed in parallel
v/ith other overt non-partitive indefinite NPs. Like other non-partitive indefinites
possessive BPs do not license free interpretations.
6. Conclusions
Summarizing, I have suggested that - contra the conclusions drawn by
Chierchia (1998) on the basis of (3) and (4) - the postulation of a partitive structure
is not necessary to account for the semantic properties of Italian BPs.
In addition, I have illustrated cases in which Italian BPs do not show semantic
properties that one would expect to correlate with a structure like the one suggested
by Chierchia. Even if one wants to maintain that the structure proposed by
Chierchia constitutes a possible structural analysis for Italian BPs, a second non-
partitive structure is needed to account for those cases in which Italian BPs do not
trigger partitivity entailments and/or existence presuppositions.
Finally, with the data from possessive BPs I have argued that Italian BPs
cannot have a partitive structure at all: their morphological similarity with full
partitives does not indicate a structural similarity between the two types of NPs.
This leads to the conclusion that the interpretive properties of Italian BPs are better
accounted for under the assumption that they are overt non-partitive indefinites,
where dei is a lexical indefinite plural determiner, rather than the morpho-syntactic
composition of the partitive preposition and the definite article.
7. Afterthoughts (inspired by Zamparelli 2002b)

One analytic option that I have not considered explicitly is discussed by
Zamparelli (2002b). The hypothesis that Italian BPs have an underlying partitive
structure can be maintained by arguing that the partitive semantics determined by
this underlying structure does not apply to an ordinary definite nominal, but to a
kind-denoting definite nominal like the one in (16), which would be rendered in
English by a bare plural NP (see Zamparelli 2002a):14
14
This analytic option has been suggested to me independently by Dominique Sportiche (p.c.).
(16) ƒ dodo si sono estinti molti anni fa.

the dodos SELF are extinct many years ago
"Dodos have become extinct many years ago."
The existence of this alternative analytic possibility brings into question the
strength of the conclusions that I draw above: this alternative analysis does not fall
prey to one set of arguments that I raised against Chierchia's proposal.
The arguments in (9) are not arguments against a partitive structure tout-
court but arguments against a partitive structure built on ordinary definite NPs,
i.e. nominals that carry existence presuppositions. Kind-denoting definites do
not carry existence presuppositions, a property that within Zamparelli's proposal
immediately accounts for the differences between full partitives and BPs in (9).
Still, accounting for the facts in (8) within this alternative analysis is not as
straightforward: it is necessary to explain how the proper partitivity requirement
applies to a kind (the denotation of the definite NP that BPs are built on). I refer
the reader to Zamparelli (2002b, §5) for an explicit attempt at doing so.
Even leaving aside the task of deciding between the analysis of Italian BPs
sketched in this paper and the analysis proposed by Zamparelli, I take the main
(negative) conclusions of this paper to be correct: Italian BPs are different from
full partitive NPs. 15 Under either alternative analysis, the data in (14)—(15) argue
that BPs are not ambiguous: a structure isomorphic to that of full partitives is not
available to these NPs. 16 Thus, the postulation of such a structure is neither needed
nor desired in order to account for the semantic properties of Italian BPs.
References
Barker, Chris 1998. "Partitives, double genitives, and anti-uniqueness". Natural Language
and Linguistic Theory 16:4.679-717.
Bosveld-de Smet, Leonie 1998. On Mass and Plural Quantification. Ph.D. thesis,
Rijksuniversiteit Groningen.
Chierchia, Gennaro 1998. "Partitives, reference to kinds and semantic variation".
Proceedings of Semantics and Linguistic Theory VII, edited by Aaron Lawson, 73-98.
Ithaca, NY: CLC Publications, Cornell University.
Delfitto, Denis 1993. "A propos du statut lexical de l'article partitif en français: Quelques
hypothèses sur l'interaction entre morphologie et forme logique". Du lexique à
15
Zamparelli (2002b, §3) proposes interesting syntactic and semantic arguments in favor of an analysis
according to which the partitive determiner is derived in the syntax, rather than being a lexical plural
indefinite determiner. For reasons of space I cannot address those arguments here.
16
Within an analysis à la Zamparelli the data in (15) are not unexpected if the availability of free
interpretations for possessives depends on the presuppositions triggered by (ordinary) definite NPs.
See Storto (2003) for discussion.
330 GIANLUCA STORTO
la morphologie: Du côté de chez Zwaan, edited by Aafke Hulk et al., 103-116.

Amsterdam: Rodopi.
van Geenhoven, Verle 1998. Semantic Incorporation and Indefinite Descriptions. Stanford,
Calif.: CSLI Publications.
Ladusaw, William 1982. "Semantic constraints on the English partitive construction".
Proceedings of the First West Coast Conference on Formal Linguistics, 231-242.
Stanford, Calif.: CSLI Publications.
McNally, Louise & Verle van Geenhoven 1998. "Redefining the weak/strong distinction".
Manuscript, Universitat Pompeu Fabra and MPI Nijmegen.
Roy, Isabelle 2001. "Weak des/du-NPs in French and judgment forms". M.A. Research Pa
per, University of Southern California.
Schwarzschild, Roger 1993. "The contrastiveness of associated foci". Manuscript, Hebrew
University of Jerusalem.
Storto, Gianluca 2000. "On the structure of indefinite possessives". Proceedings of
Semantics and Linguistic Theory X, edited by Brendan Jackson & Tanya Matthews,
203-220. Ithaca, NY: CLC Publications, Cornell University.
—— 2003. Possessives in Context: Issues in the Semantics of Possessive Constructions.
Ph.D. thesis, University of California, Los Angeles.
Zamparelli, Roberto 2002a. "Definite and bare kind-denoting noun phrases". Romance
Languages and Linguistic Theory 2000: Selected Papers from Going Romance 2000,
edited by Frank Drijkoningen, Claire Beyssade, Paola Monachesi & Reineke Bok-
Bennema, 305-329. Amsterdam: John Benjamins.
—— 2002b. "Dei ex machina: A note on plural/mass indefinite determiners".
Manuscript, Università di Bergamo.
Zucchi, Alessandro 1993. "The ingredients of definiteness and the definiteness effect".
Natural Language Semantics 3.33-78.
DETERMINERS AND WEAKLY DISCRETISED DOMAINS*
LUCIA TOVENA
Université de Lille
1. The issue
There is a link between the structure of the domain of a noun and the types
of determiners it can combine with. Chierchia (1998) classifies determiners into
four classes on this basis, i.e. countable only determiners
—number sensitive and hence subdivided into singular determiners, e.g. each
and every, and plural ones, e.g. several—mass only determiners, e.g. much,
plural count and mass determiners, e.g. most, and unrestricted ones, e.g. the and
no. According to his analysis, a singular countable noun is associated with a set
of atoms, and the set-forming operator PL is used to enable us to talk about sets
of them. Conversely, the lexical entry of a mass does not single out the set of
atoms, but a sublattice, so that the difference between singular and plural is
neutralised, for the noun applies equally to atoms and sets thereof. Hence, plural
count nouns and mass are essentially the same, and it is claimed that no language
has determiners for the mass and singular count combination. See also Doetjes
(2001) for a similar claim. In Tovena (2001, 2002), however, I have shown that
this claim is not correct. There exist singular determiners applying also to mass
nouns, e.g. the Italian determiner nessuno "no/not any" combines with count
singular (1a) and mass (1b) but not with plural (1c) nouns.
(1) a. Non ha letto nessun libro.

"He did not read any book."
b. Non ha nessuna pazienza coi bambini.
"He has no patience with children."
 *Non ha letto nessun(i) libri.
"He did not read any books."
* Thanks to Francesca Tovena and Bernadette Vandenbossche for valuable discussions, to Ruth Huart
for much help and support, and to an anonymous reviewer for useful comments.
332 LUCIA TOVENA
I also pointed out a shortcoming in Chierchia's analysis. According to his

characterisation, mass nouns are expected to be all alike, as their domains have
only one type of structure. However, it is a linguistic fact that the class of mass
nouns may systematically split with respect to one determiner, cf. (2). Note that
the reading of (2a) is not the so-called taxonomic reading, exemplified in (3),
which is in general available for mass nouns thanks to an operation of coercion
that imposes a discretisation of the domain via the notion of species or kind
(Ojeda 1993).
(2) a.
Non ha mostrato nessuna pietà.
"He showed no mercy."
b. *Non ha bevuto nessuna acqua.
"He drank no water."
(3) Non beve nessun vino, né bianco, né rosso, né chiaretto.
"He does not drink any wine, be it white, red or rosé."
In Tovena (2001) I have argued that the subset of (abstract) mass nouns that
nessuno combines with are intensive quantities (IQ) (Van de Velde 1996),1 The
main peculiarity of IQs is their possibility of undergoing continuous increase or
contraction without a corresponding extension in space or time. The group of IQ
nouns has special syntactic/semantic behaviour with respect to several
determiners. I have hypothesized different possible levels of discretisation in the
domain of denotation of a noun. Strongly discrete units qualify traditional atoms,
and occur through the stipulation of the lexical entry. Thus, they are present in
the domain of countable nouns but absent from the domain of uncountable nouns
in general. As individuals, they can be 'seen' by quantifiers and be identified
directly or via anaphoric links. Weakly discrete units correspond to units that are
found in the domain of IQs and that can be 'seen' only in particular conditions.
No default visible units are present in mass domains.
This paper tests the idea of weakly discrete units on a number of cases and
tries to define conditions on their visibility to singular determiners. First, to this
aim, section 2 provides a formal characterisation of the notion, which exploits
the notion of equivalence classes. The members of such classes are
undistinguishable from one another, and can be pointed at only if their lack of
identity is respected. Then, section 3 discusses visibility conditions in the case
where a determiner does not partition its domain. Section 4 considers
1
Culioli (1982) mentions three properties that, when applied to common nouns, would give the
following classification: the term 'discret' ("discrete") covers count nouns, 'dense' ("dense") covers
mass nouns—and is understood as being in between—and 'compact' ("compact") covers IQs, which
are considered as absolutely non discretisable. However, I have been unable to trace a proper
discussion of this distinction in his work.
DETERMINERS AND WEAKLY DISCRETISED DOMAINS 333
determiners that do not return the same value for all the elements in the domain.
The consequent discrimination can be operated on weakly discrete units if it does
not lead to the identification of the referents of the witness sets (Barwise and
Cooper 1981; Szabolcsi 1997).
2. Units and atoms

The use of a lattice to describe the domain of denotation of a noun requires
the specification of a set of entities E on which an ordering relation R is defined.
In the case of countable nouns, the set E contains all the naturally discrete
entities denoted by the singular noun, and the ordering relation R is usually
constituted by the relation of membership, dubbed 'component relation' by
Chierchia. To be precise, Chierchia uses subset instead of membership as the
ordering relation, so that he never has sets of sets in his lattice. The lattice
represents the extension of the singular together with the plural forms. Each form
can be used to single out a portion of the lattice, the atoms for the singular and
the sets for the plural.
In the case of uncountable nouns, the whole lattice is the denotation of a
sole lexical form. As is traditionally done, the lattice is built on a set of elements
E—which are the minimal parts of the extension of a mass noun—and the
ordering relation R 'part-of', cf. Figure 1 from Szabolcsi and Zwarts (1993).
These minimal parts are represented as individual entities, as needed in order to
build the structure, but do not qualify as atoms in the sense understood for count
lattices. They are not singled out by the lexical form.
Chierchia proposes to use the same structure for count and mass nouns,
with the only difference being that the ordering relation is interpreted either as
'component-of' or 'part-of'. In Tovena (2001), I have reinterpreted his proposals
in terms of visibility of the units. According to my proposal, atoms ordered by
'component-of' qualify as strongly discrete units. They can be accessed directly
via the lexical entry, which makes them visible in all contexts, e.g.
334 LUCIA TOVENA
quantificational or anaphoric. As the entry of a mass noun does not directly

single out the mereological parts in the domain, no visible default units are
present in it. This is to say that atoms ordered via a 'part-of' relation are not
visible.
In the taxonomic reading of mass nouns, different kinds of N identify
different subdomains (Ojeda 1993), so that they constitute the atoms of the entire
domain, even when they denote subdomains of the atomless kind and can be
atomless themselves. They represent the countable sense of the mass noun.
Although no overt lexical material is needed to trigger the discretisation by
kinds, as is the case for count readings induced by classifiers in general, this
possibility is available only as secondary selection, and is typically used as a
rescue strategy, cf. (3) above.
However, the relevant reading of (2a) has to do with amounts/degrees of
mercy rather than with types. This difference is captured by assuming that there
are weakly discrete units, which correspond to units that are found in the domain
of IQs. These units are not induced by measuring, e.g. via classifiers, or
conversion, cf. the taxonomic reading, and are weak inasmuch as they can be
'seen' only in particular conditions. They do not carry an identity condition based
on the mereological extensionality of concrete mass nouns. I have proposed to
represent the denotation of an IQ via a structure of non-free join semilattice like
the one put forth by Szabolcsi and Zwarts (1993) to characterise amounts, cf.
Figure 2, and which is an abstraction of the free lattice used for masses.
As claimed by Szabolcsi and Zwarts, in the lattice in Figure 1 the elements

[a], [b] and [c] represent real stuff, and the sum of [a] and [b] needs to be
distinguished from the sum of [a] and [c]. Both sums are entities with a size of
two units, but they each have their own identity. The difference with the lattice in
Figure 2 is that here the identity of the bits of stuff has been taken away, and [a],
[b] and [c] are abstract unit-sized bits. "Fixing an arbitrary unit-sized [a] to start
with, [b] stands for the equivalence class of all unit-sized bits of stuff whose
addition to [a] yields a two-unit-sized bit" (Szabolcsi and Zwarts 1993:267), and
this two-unit-sized bit is noted as [a + b]. 2 Next, "[c] stands for the equivalence
class of all units whose addition to [a + b] yields a three-unit-sized bit", noted as
[a + b + c].
If p is a proper part of q, then there is some part of q (the witness) that does
not overlap with p (Landman 1991:314). Note that the sign '+' indicates a
progression along a continuum. Being abstract and allowing for the definition of
a scale, the lattice in Figure 2 resembles a chain, a special case of lattice used for
numbers, more than the standard representation for masses given in Figure 1, as
pointed out by Szabolcsi and Zwarts. On the other hand, they present the witness
property as a way to capture what distinguishes amounts from numbers. An
increased amount is obtained by considering more stuff than in the previous
amount, rather than just moving higher on a scale. But IQs are masses with no
standard extensional unit of measure.
Given this background, weakly discrete units can be understood as atoms
perceivable insofar as they are representative of their equivalence class but not as
distinct entities members of the same class. Next, IQs' lexical entries may be
associated with the units via the classes. This is to say that these lexical entries
may behave as count nouns in suitable contexts. The condition for the visibility
of weakly discrete units can be expressed in terms of a constraint according to
which a particular unit can be used if it holds that all the members of the same
class are equally good candidates for that use.
3. Constant functions
In this section we start our review of singular determiners and show that the
notion of weak unit can explain contrasts such as (2a,b), unaccounted for in
previous formal systems.
3.1 Negative and positive determiners

The level of discreteness provided by weakly discrete units is enough to
satisfy the requirement of a singular negative determiner such as nessuno for the
following reasons. Strongly discrete units are not needed because we are dealing
here with a negative quantifier, whose witness set is empty by definition. At no
stage in the interpretation of the quantified phrase will it be necessary to single
out a particular individual. All the elements are treated in the same way and this
determiner can be represented by a constant function, i.e. a function that maps all
the elements of the domain into the same value. The acceptability of (lb) and
(2a) is the result of the interaction between a weakly discretised domain and a
2
The equivalence class [a] is made up of a, b, c, etc. where a, b, c are arbitrary units of a given set,
and [a + b] is the equivalence class containing the sum of two arbitrarily chosen and disjoint units.
336 LUCIA TOVENA
determiner that does not partition the denotation of the noun, as it denotes an
empty intersection. The use/visibility of weak units is licensed by the constant
function determiner. The subdivision into weakly discrete units can be exploited
by Chierchia's domain restrictor S, a function defined only on atomic
denotations, so that the characterisation of nessuno as singular determiner can be
preserved.
The French negative determiner aucun "no/not any" provides another
example of a determiner that exhibits similar, apparently non homogeneous,
behaviour with respect to uncountable nouns, clustering IQs together with
singular countable nouns, as shown in (4).
(4) a. Il n'a pris aucun livre.

"He did not take any book."
b. Il n 'a montré aucune pitié.
"He didn't show any mercy."
 *Il n'a vu aucun étudiants.
"He didn't see any students."
d. *Il n'utilise aucun sable.
"He uses no sand."
This line of reasoning seems promising for all singular determiners whose
denotation can be represented by a constant function, not just for negative ones.
It solves the problem raised by sentences such as (5), predicted to be impossible
in Chierchia's account.
(5) He still had every confidence in her as a mathematician.
Sentence (5) says that the whole of his confidence applies to her. In this
case, the quantified NP has a unique witness set that coincides with the whole
denotation of the noun. Such a set may be identified without making use of
strongly discrete units as it doesn't partition the noun denotation. As noted in
Cooper (1990), every allows for the possibility of a general evaluation of the
witness set, 'in batch'. Such a possibility results in (5) in the reading as 'whole'
which is the only one available. This reading may be the result of a negative
proof, i.e. a search for contradiction on the restricted complement set, which is
empty.
On the contrary, the singular universal determiner each does not combine
with mass nouns of any type, cf. (6). As explained above, the taxonomic reading
is immaterial for the present study. The case of each can be dealt with by
assuming that the verification of the property of distributivity relies on the full
individuality of strong units.
(6) a. She put each apple in a different basket.

b. *He still had each confidence in her in every situation.
 * She summoned up each courage in the predicaments.
The analysis extends also to French polarity sensitive and free-choice item
le moindre "the least"3 Although this determiner is not a constant function,
strictly speaking, it points to an end-of-scale position and can be equated to a
universal or negative quantifier depending on the direction of the inferences on
the scale, cf. (7), a behaviour that justifies its inclusion in this section. Its
concessive value equates all individuals as possible referents of the det N'
expression. In these conditions, weak units can be seen and IQs are acceptable,
while (plurals and) mass nouns in general are out, cf. (8).
(7) a. Il n'a pas lu le moindre livre.

"He did not read a single book."
b. Il connaît le moindre recoin du village.
"He knows every nook and cranny of the village."
(8) a. Il n 'a pas montré le moindre courage.
"He did not show the least bit of courage."
b. *Il n 'a pas lu le moindre livres/les moindres livres.
"He didn't read any books."
c. *Il n 'a pas bu la moindre eau.
"He did not drink a drop of water."
3.2 Subject and object positions

The contrast in (9) may suggest that there is a subject/object asymmetry in
the acceptability of IQs with nessuno. However, as shown in (10), IQs are not
completely ruled out from subject position.
(9) a. *?Nessuna paura attanaglia Daniele.

no fear clamps Daniel
"No fear clamps Daniel."
b. Daniele non ha nessuna paura.
Daniel not has no fear
"Daniel is not at all afraid."
(10) Nessun coraggio basterà contro un nemico così perverso.
no courage suffices-FUT against an enemy so perverse
"No courage will suffice against such a twisted enemy."
3
For a discussion of the polarity sensitivity and free-choiceness of this item, see Tovena and Jayez
(1999); Jayez and Tovena (2000, forthcoming).
338 LUCIA TOVENA
Weak units cannot support the presupposition usually associated with a

topic position (Lambrecht 1994), whether it is called givenness (Schwarzschild
1999) or activation (Dryer 1996). The taxonomic reading, which would rescue a
sentence in the general case of mass nouns, seems not to be available and (9a) is
unfelicitous. In (10) the main predicate enhances the amount reading and così
"such a" also favours a comparison between degrees of qualities.
It could be asked whether IQs have absolutely no taxonomie reading. This
is the position of Van de Velde (2000), who argues against this type of reading on
the basis of French examples such as (11) containing the test expression quel
"what/which":
(11) ??Quelle admiration a-t-il pour son père?

"What admiration does he have for his father?"
(12) a. Nessuna ammirazione coglie Daniele quando parla col
presidente.
"No admiration seizes Daniel when he talks to the president."
b. Nessuna paura è cosi forte come quella di Daniele per le
for miche rosse.
"No fear is as strong as the one felt by Daniel for red ants."
In the light of Figure 2 and the sentences in (12), I suggest that, if such a
reading exists, it corresponds to a change in the ordering relation imposed in the
lattice, specific to this type of coercion, but it does not cause an alteration of the
structure of the lattice of IQs in the sense that it does not induce a reanalysis of
subdomains of it as atoms of a new structure. Hence it does not have the usual
dramatic impact as a rescuing strategy. If pragmatically plausible, types would
also correspond to a variation in intensity.4
Note that a classification based on an external factor—e.g. 'different fears'
can be named on the basis of their different causes, cf. (12b), not on the basis of
internal differences like a fear for 'sparkling' or for 'plain' water—involves a
countable use of the noun, which then can take the plural.
4
My position finds support in the Spanish data in (i), provided by a reviewer. The two sentences
present a partition of 'respect' into amounts that may be used also as subdomains for a taxonomy.
(i) a. Ana no tiene ninguna consideración con los animals.
"Ana has no consideration/respect for animals."
b. Ana no tiene ningún tipo de consideración con los animales.
"Ana has no consideration/respect for animals."
4. Non constant f unctions

Given the tack I have taken on constant functions, one would expect no
determiners characterized as non constant functions to combine with IQs. This is
not always the case. However, it will be clear from the discussion below that
existential determiners that combine with IQs are of a special kind that does not
disclose the identity of the elements in the denotation.
Let us look first at a case that meets our intuitive expectations. In the case
of the singular positive Italian determiner qualche "some", for instance, some
individuals would have to be singled out, because of the various witness sets
with cardinality greater than zero. As a result, the quantified NP cannot be built
with mass and is acceptable only with count nouns, cf. (13).
(13) a. *Ha risposto con qualche rabbia.

"He answered with some anger."
b. Qualche libro merita di essere letto.
"Some book deserves to be read."
French quelque/s N "some" is a more complex case. It is a countable

determiner that can be analysed as the combination of a singular and a plural
determiner subject to different distributional restrictions. The plural determiner
distributes freely, cf. (14).
(14) a. Il a rencontré quelques amis dans le train.

"He met some friends on the train."
b. A-t-il rencontré quelques amis?
"Did he meet some friends?"
 S'il a rencontré quelques amis dans le train, il rentrera tard.
"If he met some friends on the train, he will be home late."
d. ?Il n 'a pas rencontré quelques amis.
"He did not meet some friends."
The marginality of (14d) can be reduced to the general issue of the

expression of an existential quantifier in the scope of negation. Sentence (14d) is
felt to be awkward and, if ever, is used as denial. When the intended
interpretation is¬ French marks it overtly by using either the partitive form
des amis "friends" or a negative form such as aucun ami "no friend" that makes
clear that the expression has to be interpreted in the scope of predicate negation.
When the intended interpretation is ¬, more natural to prepose the
existential using the corresponding of athere-constructionas in (15).
340 LUCIA TOVENA
(15) Il y a quelques amis qu'il n 'a pas rencontré dans le train.

"There are some friends he did not meet on the train."
On the contrary, the singular determiner—a form which is rarely used in

spoken language, and is perceived as having a literary flavour also in written
language—undergoes restrictions due to the fact that, roughly put, the entities in
its denotation must remain nonspecific. In other words, assuming the standard
tripartite structure, the identity of the individuals in the denotation of the
restrictor is asserted not to matter. For this purpose, quelque "some" exploits
modality and aspectual features of the clause, when combining with countable
nouns, cf. (16b-e) below, and does not combine with concrete mass nouns.
However, it does combine with IQ nouns, and in such case the restrictions just
mentioned need not be met (Culioli 1982). Two different questions have to be
tackled to account for the distribution of singular quelque. On the one hand, one
should characterise the requirement provisionally labelled as nonspecificity
which imposes constraints on the context of occurrence. This is a problem that
concerns the determiner as a whole. 5 On the other hand, one should explain why
such a requirement appears to be waived specifically when this determiner
combines with IQs. Although I am concerned with the second question in this
paper, some remarks on the first one are in order.
Culioli identifies three types of environments suitable for' hosting singular
quelque. Environments of the first type have a 'valeur d'éventualité' ("possibility
value"), which could be reinterpreted as corresponding to most examples of
contexts containing a non-veridical operator (Zwarts 1995), cf. conditional in
(16b), epistemic in (16c), negative predicate in (16d), question in (16e).
(16) a. *Hier il a rencontré quelque ami dans le train.

"Yesterday he met some friend or other on the train."
b. Si quelque difficulté apparaît, il va abandonner l'affaire.
"If he meets any difficulty, he will drop the matter."
 Il aura rencontré quelque ami.
"He must have met some friend."
d. Je doute qu ' il ait rencontré quelque ami.
"I doubt he met some friend."
e. Est-ce que quelque incident serait survenu?
"Could any accident have occurred?"
Singular quelque is not natural in affirmative episodic sentences and does

not occur in the scope of overt negation, cf. (17). Thus, non-veridicality appears
3
A formal treatment is put forth in Jayez and Tovena (2002).
not to be a sufficient condition for capturing its distribution, which approximates

that of free choice items (FCI).
(17) *Il n 'a pas rencontré quelque ami.

"He did not meet some friend."
However, FC any is out in the epistemic rephrasing of (16c), cf. (18). This
contrast could be ascribed to the fact that epistemic must in (18) is an evidential
(Westmoreland 1995), i.e. a lexicalised label providing information about the
proposition, and contributes the information that the propositional content of the
sentence is inferred rather than just believed by the agent. Information is
presented with a stronger flavour of truth, although it is not overtly/plainly
known as a fact.
(18) *He must have met any friend.
Next, quelque does not occur under a modal of possibility, cf. (19). Note
that (20) does not prove that quelque occurs under a modal of necessity, which
would be in contrast with what happens with FCIs. In fact, (20) is interpreted as
a conditional.
(19) * Il peut rencontrer quelque ami.

"He may meet some friend."
(20) Il faut avoir quelque protecteur à la cour pour y faire quelque chose
(Culioli)
"You need to be sponsored by someone in the court to get anywhere."
The second type of environment identified by Culioli is claimed to have a

'valeur d'itération' ("repetition value"). It roughly corresponds to generic and
habitual sentences, cf. (21). Generic sentences are suitable contexts for FCIs too,
cf. (22).
(21) a. Tous les jours, en allant à l'école, je rencontrais, près du

ruisseau, quelque clochard endormi. (Culioli)
"Every day, on my way to school, I met a/some beggar sleeping
by the stream."
b. Chaque jour je butais sur quelque nouveau problème. (Culioli)
"Every day I stumble on a/some new problem."
c. Une rivière mène toujours à quelque endroit habité.
"A river always leads to an inhabited place."
342 LUCIA TOVENA
(22) a. N'importe quel étudiant sait ça.

"Any student knows that."
b. Any cat hunts mice.
However, the generic examples expose another difference between FC

items and quelque. In (22), and more generally with this type of context, phrases
containing FC items are themselves interpreted as generic. On the contrary, in
(21), quelque-phrases are not generic themselves, they are simply nonspecific.
These first two groups of contexts may look like reasonably close variants
of the same conditions of use because the det N' expression tends to be
interpreted as referring to a potential, not to a real specific individual, as claimed
in Van de Velde (2000). However, it is important to stress the fact that this idea of
potential individual is not incompatible with the existence of a referent. In (21a)
particular individuals are met. The sentence asserts that some particular events
occurred, hence cannot be characterised as strictly non-episodic. Actually, this
sentence is compatible with a situation where I always meet the same beggar.
What matters is that the identity of the individual is presented as non available,
just the general properties described by the N' can be taken into consideration as
a valid characterisation of the entity. On this non availability of individual
properties is constructed the generic reading, since the sentence is interpreted as
expressing a regularity and not a set of isolated facts.
Indeterminacy is emphasised by Van de Velde (2000) with respect to the
third group singled out by Culioli. He classifies examples such as (23) as cases
containing a modal marker expressing a 'valeur d'indéfinition' ("indefiniteness
value"), which seems to correspond to an indetermination value whereby any
instantiation is equally valid.
(23) a. Quelque vague général, qui se trouvait disponible, se vit

confier les rênes de l'Etat. (Culioli)
"Some sort of general, who happened to be available, was
entrusted with the reins of the state."
b. Ne l'avait-il pas traité, dans quelque feuille d'opposition,
d' "obsédé sexuel de la ligne droite"? (Culioli)
"Hadn't he called him a "sex maniac of the (straight?)/right line",
in a/some paper of the opposition?"
Up to here, quelque appears to meet the constraints imposed on FCIs and to

mimic their behaviour, e.g. in the case of negative sentences, except for the case
of modals of possibility and for its role in generic sentences. Furthermore, it
exhibits some sort of indeterminative value, which is an intuitive trademark of
free-choiceness. However, contrary to French FCIs such as n'importe quel "no
matter which" and le moindre - cf. the contrast between (16e), repeated here as
(24a), and (24b) with (25a) - it can occur in yes-no questions. In such a context
English any has a primary polarity sensitive reading, cf. (25b), and can acquire a
FC reading when bearing some stress. Furthermore, quelque is also not allowed
in comparatives, a traditional licensing context, cf. (26).
(24) a. Est-ce que quelque incident serait survenu?

"Could any accidents have arisen?"
b. Est-ce qu 'il a eu quelque incident?
"Has he had any accident?"
(25) a. ? Est-ce qu 'il a lu n 'importe quel livre?
"Did he readjust any book?"
b. Did he read any book?
(26) a. *Elle a mieux sauté que quelque (autre) fille de sa classe.
"She jumped better than some girl in her class."
b. Elle a mieux sauté que n 'importe quelle (autre) fille de sa classe.
"She jumped better than any other girl in her class."
 She did a better jump than any other girl in her class.
From these data we are led to conclude that quelque is out in all the
contexts where any receives universal interpretation. This difference also holds
with respect to a potential indiscriminative value (Horn 2000) of quelque (27) vs.
the traditional one of just any (28a). This value is also exhibited by n'importe
quel in contexts where it functions semantically like an adjective (28b), which is
not possible for quelque, cf. (27b). This is because with quelque the speaker does
not commit herself to any selection inside the category N', but from there it does
not follow that the identity of the individual does not matter.
(27) a. Quelque imbécile a dû fermer à clé la porte de la cave.

"Some idiot must have locked the door of the cellar."
b. *Ce n'est pas quelque acteur, c'est Monsieur X.
"It is not just any actor, it is Mister X."
(28) a. It is not just any book.
b. Ce n'est pas n 'importe quelle actrice, 'est Madame X.
"It is not just any actress, it is Madam X."
As said above, when quelque combines with IQ nouns, context restrictions

need not be met. The sentences in (29) show that positive contexts become
suitable environments, but not for mass nouns at large, cf. (29c).
344 LUCIA TOVENA
(29) a. Il a montré quelque courage.

"He showed some courage."
b. Il y a quelque hypocrisie à prétendre cela.
"It is a little hypocritical to claim that."
 *Il a bu quelque eau.
"He drank some water."
This limited waiving effect can be explained by recalling that the entities
denoted by IQ nouns cannot be specific in any type of context, because a weakly
discretised domain is structured as a non-free join semilattice, whose elements
have no individual identity. Thus, there is no need for modal or epistemic
contexts that suspend existential commitments and are referentially opaque. If
my hypothesis is on the right track, then the behaviour of quelque should be
distinguished from that of more traditional free choice items insofar as it does
not suggest that the choice of the referent is free because any member of the
class corresponding to N' satisfies the property of the nucleus. Rather, more
modestly, it suggests that the class corresponding to N' must contain a subset of
elements with cardinality > 1 that qualify as satisfiers of the property of the
nucleus. This suffices to leave the identity of the referent unspecified, but does
not bring about the load of full free choice, whereby, roughly put, the
interpretation of the determiner is infelicitous if it is built on the same set of
satisfiers in all the relevant worlds. For quelque to be felicitous, somewhere, in
some world, there must be the possibility of having an entity for which the
predicate is true, even if it is not in an accessible world. The result is an
indefinite like a without the possibility of being specific.
Recall that any also exhibits interpretive variations connected to the
structure of the domain of denotation of the noun. There is an increased facility
for licensing in the direct object position of a negative verb whenever the N is
mass, cf. (30a) which has a preferred FC reading of any and (30b) a preferred
polarity sensitive reading (Tovena 1998).
(30) a. He refused any offer of sympathy.

b. He refused any sympathy.
Finally, let us look at French indefinite un certain "a certain".6 When

combined with continuous entities, this determiner brings in discontinuity/
discretisation. In the general case, this discontinuity requires the presence of an
adjective modifying the noun that offers the qualitative criterion for subdividing
6
A full analysis of this expression is provided in Jayez and Tovena (2002).
the domain into species. No adjective is required with IQs and the discontinuity
is based on units interpreted as degrees of intensity, cf. (31).
(31) a. un certain courage

"a certain courage"
b. *une certaine eau
"certain water"
As weak units are directly accessible, no qualification is required. Given the

type of semilattice used for describing the structure of the domain of IQs, a
degree interpretation is possible if we conceive degrees as ordered intervals.
5. Summary and concluding remarks

In this paper, I have clarified the notion of weak units by making explicit its
formal characterisation and shown that it is possible to build a quantifier by
combining a singular determiner and a mass noun of the IQ type provided the
lack of identity of the weak units in the domain is respected. A series of singular
determiners characterised as denoting constant functions has been reviewed, that
use weak units because they never single them out. Therefore, weak units can be
'seen' in these cases. Then, I have reviewed a series of determiners characterised
as denoting non constant functions, that cannot use weak units except under the
tight condition of never relying on the identity of a potential referent.
In sum, it has been shown how the notion of weakly discrete units, visible
to determiners, provides help to account for the distribution of singular
determiners expressing constant functions, and allows a unified account of
several 'potential irregularities' in the distribution of singular existential
determiners.
References
Barwise, Jon & Robin Cooper 1981. "Generalized quantifiers and natural language".
Linguistics and Philosophy 4.159-219.
Cooper, Robin 1990. "Three lectures on Situation Theoretic Grammar". Research Report
HCRC 13, Edinburgh.
Chierchia, Gennaro 1998. "Plurality of mass nouns and the notion of 'semantic
parameter'", Events and Grammar, ed. by Susan Rothstein, 53-103. Dordrecht:
Kluwer.
Culioli, Antoine 1982. "A propos de quelque". Actes du Colloque franco-bulgare de
linguistique. Contrastive Linguistics, reprinted in Culioli Antoine (ed.) 1999. Pour
une linguistique de renonciation T.3, Ophyrs, 49-58.
Doetjes, Jenny 2001. "La distribution des expressions quantificatrices et le statut des
noms non-comptables", Typologie des groupes nominaux, ed. by George Kleiber et
al, 119-142. Rennes: PUR.
346 LUCIA TOVENA
Dryer, Matthew 1996, "Focus, pragmatic presupposition, and activated propositions".

Journal of Pragmatics 26,475-523.
Horn, Larry 2000. "Pick a theory, not just any theory". Negation and Polarity. Syntactic
and Semantic Perspectives, ed. by Laurence Horn and Yasuhiko Kato, 147-192.
Oxford: Oxford University Press,
Jayez, Jacques and Lucia M. Tovena 2000. "Free-Choiceness as non-locality
specification". talk delivered at "(Preferably) non-lexical semantics", Paris,
available from http://pweb.ens-lsh.fr/jjayez/.
Jayez, Jacques and Lucia M. Tovena 2002. "Determiners and (Un)certainty", SALT 12,
San Diego, Calif.
Jayez, Jacques and Lucia M. Tovena (forthcoming). "Free-Choiceness and Non
Individuation", Linguistics and Philosophy.
Lambrecht, Knud 1994. Information structure and sentence form. Cambridge: Cambridge
University Press.
Landman, Fred 1991. Structures for Semantics. Dordrecht: Kluwer.
Ojeda, Almerindo 1993. Linguistic Individuals. Stanford, Calif: CSLI.
Schwarzschild, Roger 1996. "Givenness, AvoidF and other constraints on the placement
of accent". Natural Language Semantics 7.141-177.
Szabolcsi, Anna 1997. Ways of scope taking. Dordrecht: Kluwer.
Szabolcsi, Anna and Frans Zwarts 1993. "Weak islands and an algebraic semantics for
scope taking". Natural language semantics 1.235-284.
Tovena, Lucia M. 1998. The fine structure of Polarity Sensitivity. New York, N.Y:
Garland.
Tovena, Lucia M. 2001. "Between mass and count". WCCFL 20.565-578.
Tovena, Lucia M. 2002. "Distributional restrictions on negative determiners". Meaning
Through Language Contrast, ed. by Katia Jaszczolt and Ken Turner, 3-28.
Amsterdam: Benjamins.
Tovena, Lucia M. and Jacques Jayez. 1999. "Any: from scalarity to arbitrariness".
Empirical issues informal syntax and semantics II, ed. by Francis Corblin et al., 39-
57. The Hague: Holland Academic Graphic.
Van de Velde, Danièle 1996. Le spectre nominal. Paris: Peeters.
Van de Velde, Daniele 2000. "Les indéfinis comme adjectifs". De l'indétermination à la
qualification. Les indéfinis, ed. by Leonie Bosveld et al. : 203-272. Arras: Artois
PU.
Westmoreland, Robert R 1995. "Epistemic must as evidential". Proceedings of the Tenth
Amsterdam Colloquium, 683-702. Amsterdam : ILLC.
Zwarts, Frans 1995. "Nonveridical contexts". Linguistic Analysis 25.286-312.
INDEX OF LANGUAGES AND DIALECTS
This index lists languages and dialects that are mentioned in the texts and
languages and dialects that are illustrated in the examples.
Afrikaans: 40-41 Seychelles Creole: 51-52, 55,

Aragonese, Old: 220 58-59,61,63,67
Castillian: 220 Dutch: 40, 76
Catalan: 216 English: 40, 42, 76, 78-80, 86, 196,
Old: 220, 225 281,292
Chinese: 131-133 Ewe: 52, 66
Creoles: 49-74 Fon Gbe: 66
Antillean Creole: 51-52, 53- French: 80-81, 155-156, 160-161,
55,63,68 164, 175, 177-191, 193-211,
Martinique Creole: 51, 216, 279, 281-282, 284, 292,
53, 55, 62-63, 68 315-318, 326, 336-339, 343-
St Lucie Creole: 51, 55, 344
68 Old: 219
Cape-verdean Creole: 33-34 French Lexifier Creoles: 49-74
Guadeloupean Creole: 51-52, German: 40, 76, 79-80, 217
54-56, 63-64, 68-69 Greek: 57
Guyanese Creole: 52, 54-56, Hungarian: 91
62,70-71 Icelandic: 40
Haitian Creole: 51-52, 54-55, Italian: 75, 76, 78, 81, 126, 183,
58,63-64,71-72 193-211, 216, 217, 250-252,
Louisiana Creole: 51, 69-70 254-256, 263-277, 316-329,
New Louisiana Creole: 331,339
51-52,53,56,63,66,70 Japanese: 132
Old Louisiana Creole: Latin: 249
51-52,54,56,58,69-70 Ligurian: 263-277
Mauritian Creole: 51-52, 55, Paduan: 274, 276
59-61,63-64,67-68 Piedmontese: 93
Reunion Creole: 53, 63 Portuguese: 184
348 INDEX OF LANGUAGES AND DIALECTS
Brazilian: 33-34, 131-152 Spanish: 1-16, 17-30, 75, 78-79,

European: 31-47, 75-76, 78, 87, 126, 184,317
81,87, 131-152 Old: 213-227
Old: 213-227 Swedish: 40-41, 91
Quechua, Imbabura: 131 Swiss German: 292
Romance, Old: 213-227 Turinese: 263-277
Romanian: 75-92, 113-129, 184 Wolof: 52, 66
Somali: 91
SUBJECT INDEX
This index does not claim to be exhaustive. Specific language items are printed
in italic.
Antisymmetry Theory: 49-74, 180

accomplishment: 181 A-position: 31-32,46
achievement: 181 argument structural approach to
adjacency: 36,38,43-45 focus structure: 76-91
adjective article, absence of: 250-253
adnominal: 194, 198,201,211 aspect: 177-191
agreement: 194, 197-202, 206, aucun'. 336, 339
211 auxiliary: 184, 186-188, 193-211
relational: 159, 164-165, 169-
170 
adverbs: 36, 38, 41-42, 44, 75, 79, Bare nouns: 317-318, 321, 328
83-84,86-88,90, 178-191 Basic-Operations Hypothesis: 280,
aspectual: 178-183, 187-190 286
manner: 37, 87, 181 boundedness: 181
negative: 180-185
subject-oriented: 36-38, 45 
agreement: 33 Case: 197-199,203,206,208-210
adjectival: 194, 197-202, 206, clause structure: 247
211 clitic: 113-128,213-227
past participle: 193-211 accusative: 195,203-210
predicate: 194, 98, 206-207, agreement: 193-211
210-211 cluster: 224, 225
subject: 193, 195, 197, 200, dative: 195,203-210
202,206-207,210-211 doubling: 113-128
agreement features: 94, 102-104 reflexive/reciprocal: 193-211
AGR morphology: 38-46 silent subject: 188-189
anaphoric operator: 113, 120-125, subject: 264-268
128
antecedents: 2-3, 6-7, 13-15
350 SUBJECT INDEX
CLLD: 32-35, 134-152, 213-216, E

221, 256-259, 284-286, 289- emphatic interjection: 250
291 en, quantitative: 178, 188-190
complementiser: 263-277 enclisis: 213, 216, 219, 221
conditionals: 1-16 EPP: 32-34, 36, 94, 101-102, 110
the modal analysis of: 4 expletive, overt: 194-197, 206-207
subjunctive: 4-7, 9-10, 14 expletive null subjects: 32-34
indicative: 4-5, 11-14
context change potential F
of/for: 9, 12-13 feature
de-: 2-6, 15 Case: 198,203,206,210
si-: 2-6 checking: 195,202-211
constant: 113, 120-123, 126-128 formal: 194-195, 202-203,
null: 113, 120, 122-123, 127- 208-211
128 gender: 194, 198,202-211
Copy Theory of movement: 144 number: 194, 198,202-211
count (nouns): 331, 333, 335, 339 uninterpretable: 196,202-204
focus: 15, 19-20, 26-29, 75-91, 113-
D 128
dative shift: 89-90 Basic Focus Rule: 77, 78
deaccenting: 76-81, 86-88 broad: 75-91
de-clauses,Spanish: 1-16 construction: 256, 259
definiteness: 323-324 contrastive: 75, 80, 115, 118,
D-linked quantifiers: 123-126, 128 125-126, 128
determiners: 49-74, 155-157, 170, Focus Prominence Rule: 79-
174-175, 331-333, 335-336, 80,88
339,345 identificational: 115-120, 124,
definite: 51-72 128
demonstrative: 51-72 information: 80, 90, 115, 119-
indefinite: 50-72,315-329 120
partitive: 315-329 narrow: 75-91
différent: 155-176 projection: 77-78, 81, 86
dislocation: 284 focus rules: 77, 90
Distributed Morphology: 39-40, Full Interpretation: 204-205
144
distributivity: 156, 173-174 G
domains, weakly discretised: 331 gerund: 35
double tensing errors: 281, 292
DP-structure: 49-74 H
Hypothesis on Computational
Complexity: 291
SUBJECT INDEX 351
I phrasal: 50, 64-72, 178, 188-

indefinites: 155, 157-158, 160 190
inflected infinitive: 44-45 remnant: 188-189
intensive quantities: 332 verbal: 180, 188-191
interrogative clause: 177-191 XP: 188-189
intonation: 81-91
Intrusive QUE: 279-283, 285, 287, N
290 negation: 18-20,23-29
I-to-C movement: 37-38, 43-45 Neg-raising: 20, 22-26, 29
inversion: 33-35, 37, 177-191 nessuno: 331-332,335-337
complex: 179 nominal projections: 49-74
islands: 134-136, 139-144, 151 Northern Italian Dialects (NIDs):
weak: 190 93-95, 97, 99, 263-277
nouns
 bare: 317-318, 321,328
kinds: 113, 126-128 count: 331,333,335,339
mass: 331-340, 343,345
L Nuclear Stress Rule: 76-91
language acquisition: 34-35, 279, numerals: 59-63
293
left dislocation: 32-35, 134-152, 
213-216, 221, 256-259, 284- object
286,289-291 direct: 193-195
left periphery: 94, 97-99, 101, 115, indirect: 194
188,263-277 null: 131-152
Optimality Theory: 94, 102, 110
M optionality: 93, 101-103, 107-110
mass (nouns): 331-340, 343, 345
mesoclisis: 220, 223 P
Minimalist Program: 94, 110 participle
Misplaced QUE: 279-280, 282-87, past: 193-211
290-293 partitive: 122
modal choice: 29 determiner: 315-329
mood: 1, 17-21,24-26,28-29 partitivity, proper: 322-324, 329
indicative: 6, 17-20, 26, 28- passive: 193-211
29 pe: 114, 122, 126-128
subjunctive: 17-22,24-29 performance errors: 279, 285, 290
conditional: 9,15 phrasal stress rule: 79, 81, 86-87
morphological merger: 38, 41-43 plurality- 155, 158-159, 166-168,
movement 170-173
head: 188, 190 plural markers: 51-72
352 SUBJECT INDEX
presupposition: 18-19, 28, 317-329 (simply-)split CP: 279, 282-283,

pseudo-relatives: 285-286, 288-289 286-288, 292
polarity subjunctive: 15-17, 19-21, split Comp field: 254-255, 258, 260
27,29 stress, retraction of: 253,260
possessive: 327-329 strong features: 50, 63-72
proclisis: 214, 216, 221 strong units: 337
pronouns subject: 31-47
oblique: 213-227 agreement with: 193, 195,
weak: 213-227 197, 200, 202, 206-207, 210-
prosody: 75-91 211
clitics: 93, 94, 264-268
Q inversion: 177-191
qualche: 339 null subject: 33-35, 46
quantifier null subject parameter: 31-32,
bare: 113, 122, 124, 128 46
D-linked: 123-126, 128 null subject languages: 31-34
quantities, intensive: 332 position: 31-47, 95-97, 101,
QUE, Misplaced: 279-280, 282- 272-277
287, 290-293 postverbal: 180, 189-190
QUE, Intrusive: 279-283, 285, 287, subjunctive: 270-272, 277
290 marker: 268-272
quelque: 339-344 polarity: 15-17, 19-21,26,29
subordinate, temporal: 177-191
R
reciprocal: 162, 168-169, 172-173 T
relational adjectives: 159, 164-165, tense morphology: 39-43
169-170 T-model: 94, 101
relatives Topic(s): 282-284, 287, 289-292
subjunctive: 128 topicalization: 138, 142, 147, 150-
remnant movement: 131, 138-152, 151
189 TP: 31-47
transitive expletive construction:
S 31,39-41,46
scrambling: 75-88, 147
semantics: 155-156, 161-162, 164, U
172, 175 unaccusative verb: 193, 197-198,
SpecAgrSP: 31, 36, 42-43, 45-46 201,207
Spec,IP: 32-36 unergative verb: 194
Spec,TP: 31-47 units, strong: 337
Spec,VP: 31-32,37 units, weak: 336-338, 345
specificity: 113, 121, 127
SUBJECT INDEX 353
V distribution of : 247, 254-255,

variable: 113, 120, 122-128, 131- 259
152 Vocative Phrase(s): 248-260
verb second: 213 truncation of: 252-253, 260
Vocative VP-ellipsis: 132, 144-146
Extradeictic: 248-249 VP-shell: 132-133, 131-144, 147
Infradeictic: 248-249 VP-topicalization: 139-152
Vocative Case: 247-260 VS(O): 31, 34-35
assignment of: 247, 250, 253
diagnostics for: 247-248, 251, W
253,259 Weak Cross Over: 110-122, 125
weak units: 336-338,345

(Current Issues in Linguistic Theory 245) Josep Quer (Ed.), Jan Schroten (Ed.), Mauro Scorretti (Ed.), Petra Sleeman (Ed.), Els Verheugd (Ed.) - Romance Languages and Linguistic Theory 2001

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(Current Issues in Linguistic Theory 245) Josep Quer (Ed.), Jan Schroten (Ed.), Mauro Scorretti (Ed.), Petra Sleeman (Ed.), Els Verheugd (Ed.) - Romance Languages and Linguistic Theory 2001

Uploaded by

Copyright:

Available Formats

ROMANCE LANGUAGES AND LINGUISTIC THEORY 2001

AMSTERDAM STUDIES IN THE THEORY AND

Series IV – CURRENT ISSUES IN LINGUISTIC THEORY

Advisory Editorial Board

Raimo Anttila (Los Angeles); Lyle Campbell (Christchurch, N.Z.)

Josep Quer, Jan Schroten, Mauro Scorretti, Petra Sleeman and

JOHN BENJAMINS PUBLISHING COMPANY

National Standard for Information Sciences — Permanence of Paper for Printed

Library of Congress Cataloging-in-Publication Data

Going Romance is a major European annual discussion forum for theoretically

Spanish De-Clauses Are Not Always in the Right Mood 1

Mood and Focus 17

Null vs Overt Spec,TP in European Portuguese 31

Determiner Architecture and Phrasal Movement in French Lexifier Creoles 49

On the Relation between Focus, Prosody and Word Order in Romanian 75

Economy of Structure. The Case of Subject Clitics in Piedmontese 93

Identificational Focus vs Contrastive Focus. A Syntactic Distinction 113

Null Objects and VP Ellipsis in European and Brazilian Portuguese 131

From Non-Identity to Plurality. French Différent as an Adjective and

On the Non-Unitariness of NP Subject Inversion. A Comparison of French

Deficient Pronouns and Linguistic Change in Portuguese and Spanish 213

Nominalizations of French Psychological Verbs. Syntactic Complements 231

Notes on Vocative Case. A Case Study in Clause Structure 247

Mapping out the Left Periphery of the Clause. Evidence from

The Left Periphery in Child French. Evidence for a Simply-Split CP 279

Plural Indefinite DPs as Plural-Polarity Items 295

On the Status of the Partitive Determiner in Italian 315

Determiners and Weakly Discretised Domains 331

Index of languages and dialects 347

Subject Index 349

(1) Iatridou's question: Is the tense/mood marking of both the antecedent

Suppose the inflection of the antecedent is semantically vacuous. Then, if a

(2) a. Si Caligula está muerto, me tendrán que ascender

The antecedent of a de-conditional (henceforth a de-clause) can host

(3) a. De haber matado a Caligula.

If de-clauses are truly uninflected (and if the preposition/complementizer is truly

The paper is organized as follows: section 2 presents the data. Section 3

(4) a. Si estámuerto, me tendrán que ascender.

He could also have uttered the sentence in (5a), or that in (5b).

(5) a. Si estuviera muerto (ahora), me tendrían que

b. De estar muerto (ahora), me tendrían que

De-conditionals can be paired up with either indicative or subjunctive si-

3. When is it appropriate to use a de-conditional?

3.1 Presuppositions of indicative and subjunctive si-conditionals

(6) Appropriateness conditions for indicative conditionals:

3.2 Back to Caligula

(5) a. Si estuviera muerto (ahora), me tendrían que ascender.

A counterfactual use of (5a) would be totally appropriate, as predicted,

(5) b. De estar muerto (ahora) , me tendrían que ascender

4. Revisiting the Caligula effect

4.1 Subjunctive (counterfactual) antecedents expand the existing domains

I assume that, in an initial context, the parties involved in conversation

(8) The initial context: < W, λw.{w} >5

Updating the initial context with a subjunctive antecedent amounts to expanding

Fig. 1 : Updating the context with a subjunctive antecedent

(9) <C,f> + ⌈si α [+subjunctive]⌉ =<C, f*>

What about the interpretation of subjunctive modals?7 Iatridou (2000)

(10) a. English: would= will (modal)+ 'past'

The feature whose phonetic realization we call 'past' provides, following

A natural interpretation of Iatridou's system for the modal uses of 'past'

(12)<C,f> + ⌈si α [+subjunctive]-ria ß⌉ =<C, f*>

(ii) C*= {w ε C: (f(w)Ç [α]) Ç [ß] = f(w) Ç [α]} 8

A subjunctive conditional then performs two operations: the antecedent

Assume [α]Ç [ß] = {w23,w43}