Professional Documents
Culture Documents
Volume 245
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
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
Past Participle Agreement with Pronominal Clitics and the Auxiliary Verbs
in Italian and French 193
Paul Law
Lucia Tovena
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.
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
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
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).
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.
SPANISH DE-CLAUSES 5
the proposition expressed by the antecedent be inconsistent with the context set
(Kratzer 1979).
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:
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
SPANISH DE-CLAUSES 7
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
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 ' }
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.
SPANISH DE-CLAUSES 9
(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)
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.
10 LUIS ALONSO-OVALLE
,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.
SPANISH DE-CLAUSES 11
Assume V ε C:max≤w(C[α])[ß]={w)
In symbols:
(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.
SPANISH DE-CLAUSES 13
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.
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
14 LUIS ALONSO-OVALLE
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.
16 LUIS ALONSO-OVALLE
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):
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:
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:
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:
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.
20 CLAUDIA BORGONOVO
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.
(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
protasis or the apodosis of a counterfactual, SMC obtains, but not with a realis
clause, as in (10a):
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
a: NPI licensing :
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:
With strong NPIs (such as NPI idioms) the sentences behave as expected when
negation is in the upstairs clause.
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.
24 CLAUDIA BORGONOVO
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):
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.
26 CLAUDIA BORGONOVO
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.
(19) offers additional evidence for the claim that subjunctive marks focus.
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):
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:
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:
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
#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.
30 CLAUDIA BORGONOVO
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)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
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:
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
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.
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:
. *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."
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.
NULL VS OVERT SPEC,TP IN EUROPEAN PORTUGUESE 35
life (Adragäo 2001) reveals that inversion is highly marked and rare in the child's
early productions:
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:
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.
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.
NULL VS OVERT SPEC,TP IN EUROPEAN PORTUGUESE 37
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:
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
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):
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
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
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:
sag-en sag-te-n
sag-t sag-te-t
sag-en sag-te-n
smaka-r smaka-de
smaka-r smaka-de
smaka-r smaka-de
werk
werk
werk
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
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
falá-mos
fala-stes
fala-ram
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
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):
NULL VS OVERT SPEC,TP IN EUROPEAN PORTUGUESE 45
4. Conclusions
The study developed in this paper permits drawing the following
conclusions:
46 JOAO COSTA
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.
NULL VS OVERT SPEC,TP IN EUROPEAN PORTUGUESE 47
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:
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.
• 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
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"
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
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.
DETERMINER ARCHITECTURE AND PHRASAL MOVEMENT 53
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.
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.
DETERMINER ARCHITECTURE AND PHRASAL MOVEMENT 55
• 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
• 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"
Demonstrative Phrase
(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
DETERMINER ARCHITECTURE AND PHRASAL MOVEMENT 59
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:
Similarly, in binominal constructions, both the head noun and the complement
noun can be independently determined (29). For ban, this is impossible (30):
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
DETERMINER ARCHITECTURE AND PHRASAL MOVEMENT 61
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:
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
DETERMINER ARCHITECTURE AND PHRASAL MOVEMENT 63
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.
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:
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
DETERMINER ARCHITECTURE AND PHRASAL MOVEMENT 65
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
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.
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
DETERMINER ARCHITECTURE AND PHRASAL MOVEMENT 69
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:
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:
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).
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.
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
DETERMINER ARCHITECTURE AND PHRASAL MOVEMENT 71
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.
1) The Spec vs. head status of functional elements: filled heads force local
movement to Spec, filled Spec prevents local movement to Spec.
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).
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
is no internal argument with which it could form one focus domain; here the
subject, verb and adverbial constitute their own focus domains.
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
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
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.
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.
FOCUS, PROSODY AND WORD ORDER 79
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.
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.
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)
FOCUS, PROSODY AND WORD ORDER 81
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.
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).
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).
H* H+!H* L%
(22) (Ai auzit vestea?) VIne MAma.
("Have you heard the news?") come.PRES.3SG mother.the
"MOTHER's coming."
FOCUS, PROSODY AND WORD ORDER 85
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."
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.
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.
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).
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.
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?")
FOCUS, PROSODY AND WORD ORDER 89
(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.
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.
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
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.
FOCUS, PROSODY AND WORD ORDER 91
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 &
Philadelphia: John Benjamins.
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
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).
(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:
(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."
ECONOMY OF STRUCTURE 97
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?
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:
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:
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:
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.
ECONOMY OF STRUCTURE 99
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.
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
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:
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.
ECONOMY OF STRUCTURE 101
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.
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
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.
ECONOMY OF STRUCTURE 103
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
" 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).
ECONOMY OF STRUCTURE 105
[+(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
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
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)
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
ECONOMY OF STRUCTURE 107
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:
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
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:
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.
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
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.
the preposition pe. Pe must be used with person denoting proper names and
personal pronouns and is optionally used with all person-denoting nouns.
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).
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
(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
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.
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.
IDENTIFICATIONAL FOCUS VS CONTRASTIVE FOCUS 117
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.
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.
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
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.
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
IDENTIFICATIONAL FOCUS VS CONTRASTIVE FOCUS 119
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.
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.
(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.
IDENTIFICATIONAL FOCUS VS CONTRASTIVE FOCUS 121
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.
(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.
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
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.
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).
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.
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.
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).
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.
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.
IDENTIFICATIONAL FOCUS VS CONTRASTIVE FOCUS 129
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
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).
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).
Raposo's movement, or variable analysis accounts for the island effects found in
this variety of Portuguese as can be seen below:
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
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.
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
NULL OBJECTS AND VP ELLIPSIS 135
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.
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.
But the movement analysis does not account for sentences like (11) in BP, where
there is an apparent DP-extraction:6
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):
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.
NULL OBJECTS AND VP ELLIPSIS 137
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
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
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):
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.
Differently from EP, VP-shell constructions can appear topicalized from any
kind of islands in BP:13
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).
NULL OBJECTS AND VP ELLIPSIS 143
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.
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
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
16
A more radical proposal would be to consider that EP always moves the VP-shell in its remnant
form.
NULL OBJECTS AND VP ELLIPSIS 145
When the empty category is a null object (A-NO), the governing verbs are
different, as in (33):
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.
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
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:
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.
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
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
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
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.
NULL OBJECTS AND VP ELLIPSIS 149
(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
Even with idioms containing light verbs, topicalization is possible from islands in
BR
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:
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 &
Philadelphia: John Benjamins.
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.
NULL OBJECTS AND VP ELLIPSIS 153
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
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.
(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.
(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.
(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"
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)).
FROM NON-IDENTITY TO PLURALITY 163
naturally interpreted as asserting that the books Pierre reads are of a different
kind than those read by Paul (van Peteghem 2001):
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:
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).
164 BRENDA LACA & LILIANE TASMOWSKI
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):
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,
FROM NON-IDENTITY TO PLURALITY 165
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."
166 BRENDA LACA & LILIANE TASMOWSKI
(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}
(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.
FROM NON-IDENTITY TO PLURALITY 169
(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]
170 BRENDA LACA & LILIANE TASMOWSKI
(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."
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:
This is a property D-différents shares with certains (see Corblin 2001) and
with non-intersective (strong) determiners:
Note, furthermore, that des N différents allows for the collective generic
interpretations that are excluded for D-différents:
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:
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
176 BRENDA LACA & LILIANE TASMOWSKI
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".
* 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
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
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.
3.1 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 ]]]]]
(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.
ON THE NON-UNITARINESS OF NP SUBJECT INVERSION 181
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
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):
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):
Furthermore, with respect to déjà "already", these four adverbs occur to its right,
as the contrasts (15) and (16) show:
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"
ON THE NON-UNITARINESS OF NP SUBJECT INVERSION 183
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).
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
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...
ON THE NON-UNITARINESS OF NP SUBJECT INVERSION 185
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
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:
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:
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".
ON THE NON-UNITARINESS OF NP SUBJECT INVERSION 187
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:
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:
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
ON THE NON-UNITARINESS OF NP SUBJECT INVERSION 189
(32b). Then, the whole IP is moved leftward past the subject (remnant
movement) to the specifier position of a higher functional projection GP (32c):
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.
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
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
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
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.
The auxililary "be" (French être and Italian essere) occurs in passive and with
unaccusative verbs, where the participle agrees with the surface subject.
* 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
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.
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),
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).
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
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.
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).
PAST PARTICIPLE AGREEMENT IN ITALIAN AND FRENCH 197
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).
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.
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).
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).
PAST PARTICIPLE AGREEMENT IN ITALIAN AND FRENCH 199
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.
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).
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.
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.
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.
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
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.
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
PAST PARTICIPLE AGREEMENT IN ITALIAN AND FRENCH 205
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.
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
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.
PAST PARTICIPLE AGREEMENT IN ITALIAN AND FRENCH 207
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)).
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
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.
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).
PAST PARTICIPLE AGREEMENT IN ITALIAN AND FRENCH 209
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.
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
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).
PAST PARTICIPLE AGREEMENT IN ITALIAN AND FRENCH 211
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.
Cambridge: Cambridge University Press.
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
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
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
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
216 ANA MARIA MARTINS
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).
DEFICIENT PRONOUNS AND LINGUISTIC CHANGE 217
('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)
218 ANA MARIA MARTINS
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).
The facts considered above show that the oblique pronouns i and ende
qualify as deficient pronouns like the accusative/dative/se pronouns. However
DEFICIENT PRONOUNS AND LINGUISTIC CHANGE 219
the two groups of deficient pronouns diverge from each other in the following
aspects:
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.
220 ANA MARIA MARTINS
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
(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)
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.
224 ANA MARIA MARTINS
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.
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.
228 ANA MARIA MARTINS
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.
230 ANA MARIA MARTINS
—.._____ 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.
* 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.
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.
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.
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.
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.
NOMINALIZATIONS OF FRENCH PSYCHOLOGICAL VERBS 237
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).
(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."
238 JUDITH MEINSCHAEFFER
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).
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).
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).
NOMINALIZATIONS OF FRENCH PSYCHOLOGICAL VERBS 241
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).
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).
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.
244 JUDITH MEINSCHAEFFER
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
NOMINALIZATIONS OF FRENCH PSYCHOLOGICAL VERBS 245
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.
246 JUDITH MEINSCHAEFFER
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
* 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.
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
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:
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
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:
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.
NOTES ON VOCATIVE CASE 253
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.
NOTES ON VOCATIVE CASE 255
(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:
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
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.
NOTES ON VOCATIVE CASE 257
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:
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":
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:
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
NOTES ON VOCATIVE CASE 259
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):
The crucial piece of evidence, however, comes from testing coordination of two
distinct Vocative Phrases:
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):
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
NOTES ON VOCATIVE CASE 261
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
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) . ,
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:
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
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:
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.
(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.
MAPPING OUT THE LEFT PERIPHERY 267
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:
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:
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.
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:
triggered whenever the higher verb selects a finite embedded clause, and its
realisation is obligatory/
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:
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:
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
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
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:
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:
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
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.
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
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
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.
282 DORIAN ROEHRS & MARIE LABELLE
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.
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).
THE LEFT PERIPHERY IN CHILD FRENCH 285
(19) a. ... [Cp DPj [- QUI, [IP ec¡... ]]] (Côté 1998: 154 )
b. ...[FinpDPitFin'QUiitipeCi ...]]]
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:
(21) ForceP
THE LEFT PERIPHERY IN CHILD FRENCH 287
Misplaced QUE constructions follow from improper Chain Reduction and failure
in the application of FF-Elimination:
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.
288 DORIAN ROEHRS & MARIE LABELLE
(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..]]
THE LEFT PERIPHERY IN CHILD FRENCH 289
(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 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:
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]]]]]
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.
THE LEFT PERIPHERY IN CHILD FRENCH 291
(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."
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.
THE LEFT PERIPHERY IN CHILD FRENCH 293
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.
294 DORIAN ROEHRS & MARIE LABELLE
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."
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) 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) 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) 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):
(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.
(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):
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):
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
PLURAL INDEFINITE DPS AS PLURAL-POLARITY ITEMS 299
(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).
300 BENJAMIN SPECTOR
(13) Quand des gens sont venus l'an dernier, je les ai rencontrés.
"When people came last year, I met them."
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):
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:
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."
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".
PLURAL INDEFINITE DPS AS PLURAL-POLARITY ITEMS 303
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)).
(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):
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
304 BENJAMIN SPECTOR
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:
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) 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".
(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.
PLURAL INDEFINITE DPS AS PLURAL-POLARITY ITEMS 305
(23) Des soldats ont tué un prisonnier > only one prisoner
"Some soldiers killed 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:
(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
306 BENJAMIN SPECTOR
if they occur under the scope of a plural aspectual operator quantifying over
events.
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:
(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:
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:
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) 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.
308 BENJAMIN SPECTOR
(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:
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:
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".
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:
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:
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):
(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:
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.
PLURAL INDEFINITE DPS AS PLURAL-POLARITY ITEMS 313
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
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).
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
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.
THE PARTITIVE DETERMINER IN ITALIAN 319
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.
3.1. 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:
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).
THE PARTITIVE DETERMINER IN ITALIAN 321
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 = λQx[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.
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.
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):
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.
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
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.
(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
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
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.
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
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.
* 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
(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).
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
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.
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).
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.
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.
DETERMINERS AND WEAKLY DISCRETISED DOMAINS 337
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).
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
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."
DETERMINERS AND WEAKLY DISCRETISED DOMAINS 339
3
A formal treatment is put forth in Jayez and Tovena (2002).
DETERMINERS AND WEAKLY DISCRETISED DOMAINS 341
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.
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.
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).
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.
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).
6
A full analysis of this expression is provided in Jayez and Tovena (2002).
DETERMINERS AND WEAKLY DISCRETISED DOMAINS 345
the domain into species. No adjective is required with IQs and the discontinuity
is based on units interpreted as degrees of intensity, cf. (31).
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
This index does not claim to be exhaustive. Specific language items are printed
in italic.