Zurich Open Repository and

Archive
University of Zurich
Main Library
Strickhofstrasse 39
CH-8057 Zurich
[Link]

Year: 2008

The role of the plural system in Romance

Stark, E

Posted at the Zurich Open Repository and Archive, University of Zurich

ZORA URL: [Link]
Book Section
Published Version

Originally published at:

Stark, E (2008). The role of the plural system in Romance. In: Detges, U; Waltereit, R. The paradox of
grammatical change: perspectives from Romance. Amsterdam/Philadelphia: Benjamins, 57-84.
 THE PARADOX OF
GRAMMATICAL CHANGE
PERSPECTIVES FROM ROMANCE

Edited by

ULRICH DETGES
Ludwig-Maximilians- Universitiit Miinchen

RICHARD WALTEREIT
Newcastle University

JOHN BENJAMINS PUBLISHING COMPANY

AMSTERDAM/PHILADELPHIA
 CONTENTS THE ROLE OF THE PLURAL SYSTEM IN ROMANCE1

Introduction
ELISABETH STARK
Ulrich Detges and Richard Waltereit 1
Freie Universitiit Berlin

Syntactic change from within and from without syntax:

A usage-based analysis
This paper presents a diachronic fonnal m orphosyntactic analysis of the role of th e
Richard Waltereit and Ulrich Detges 13 functional proj ec tion PI/PI* in Romance indefin ite nominals , responsi ble for number
and the countability distinction. Reint erp reting the complex system of indefinite
On explaining the rise of c ' e st- clefts in French nominal determination in two central Romance languages, viz. French and Italian,
Andreas Dufter 31 which both feature an indefinite article and a 'partitive article' as a dev ice of
'nominal classification' in a broad sense in contrast to Romance languages without
such an element, viz. Spani sh, it argues that this 'classification system' arose when
The role of the plural system in Romance
nominal declension in Latin was partially or completely lost. The application of the
Elisabeth Stark 57 latest minimalist assumptions on agreemellt processes in the syntax both to modem
Romance languages and to (Late) Latin allows us to describe and explain the
Morphological developments affecting syntactic change obvious differences between French, Italian and Spanish and to relate them to the
Maria Goldbach 85 interaction of gender and number marking in Romance indefinite nominals.

Grammaticalisation within the IP-domain 1. Introduction: Different systems of indefinite nominal determiners in
Susann Fischer 107 Romance nominals as results of language change
2
As I have repeatedly shown, there is considerable cross-linguistic variation
Imperfect systems and diachronic change in indefinite nominal determination in Romance, despite some well-known
Giampaolo Salvi 127 generalizations assuming homogeneous semantic and syntactic systems of
nominal determination for all Romance languages (e.g. Chierchia 1998,
From temporal to modal: Divergent fates of the Latin Longobardi 200la). See the data in (1 a-f):
synthetic pluperfect in Spanish and Portuguese
Martin Becker 147
1 This article was pres ented in two v ersions , the first at the Workshop on Grammatical
Language Change and its Explanations in Saarbrticken in September 2005 and the second at
Non-lexical core-arguments in Basque, Romance and German
the lntemational Workshop on the Typological Comparison of Italian and Ge rm an at the
Hans-lngo Radatz 181 Johann-Wolfgang-Goethe Universitat, Frankfurt am Main, in February 2006. S pec i a l thanks for
very h elpful comments go to Guido Mensching and Natascha Pomino from Freie UniversiHit
Towards a comprehensive view of language change: Berlin as well as to Ulrich Detges, Richard Waltereit, Michele Loporc aro , Ulrich Wandruszka,

Three recent evolutionary approaches Giinter Grewendorf, Alessandra Tomaselli, Cec i l ia Poletto, Carme Picallo and one anonymous
reviewer. All errors and short comings are, of course, mine.
Esme Winter-Froemel 215
2 Cf. some earlier work on the cross-linguistic divergence in Romance indefinite nominals and
its syntactic i mpl ications , cf. Stark (2006 and forthcoming).
Subject Index 251
 ELISABETH STARK THE ROLE OF THE PLURAL SYSTEM IN ROMANCE 59

a. Has vista *(un) tiguila? (Spanish) (2) . . . ubi se tamen monies illi, inter quos ibamus, ape rieb ant et faciebant vallem infin itam
As-tu vu *(un) aigle? (Fre nch) ingens, pl anis sima et valde pulc hram, et trans vallem appa rebat mons sanctus Dei Sy na .
Hai visto *(un) aquila? (Ital ian) " ... where in tbe meantime the mountains, between which we were going, opened out and
"Did you see an eagle?'' formed an endles s valley - hu ge , very flat and very beautiful - and across tbe valley there
b. Compro pan.
appeared Sinai, the holy mountain of God."
(Spanish)
(Vincent 1997:156, from tbe Peregrinatio Aegeriae, late 4'h ce ntury AD)
J'achete *(du) pain. (Frenc h )
Compro (del) pane. (Italian )
"I buy (some) bread" The first mention of the valley (et faciebant vallem . . 'and they formed a
.

c. Mefalta agua. (Spanish)

valley' ) in a postverbal object does not show any indefinite marking (via
It me fa ut *(de /')eau. (French) aliquis, quidam 'anyone' , ' s omeone' or else), and even when this discourse
Mi occorre (dell')acqua. (Italian) referent is referred to anaphorically (et trans vallem apparebat . . . 'and across
"I need (some) water" the valley there appeared .. . ' ) in a prepositional adverbial, it is left without a
d. Veo (a unos) estudiantes en el edificio. (Spani sh) 3
definite determiner and not referred to via (clitic) pronouns. In short: Latin did
Je vois *(des) etudiants dons le batiment. (French) ·
not need (obligatory) overt marking of nominal determination in argumental
Vedo (degli) studenti nell' edificio. (Italian)
noun phrases and did not have third person pronouns (cf. Vincent 1997 and
"I see (some) students in tbe building".
others). The Romance languages now do have overt nominal determination,
In fact, there are two main differences to be observed: which is obligatorily overt in certain syntactic contexts, and they also have
Firstly, bare noun phrases in argument position occur in Spanish and Italian (clitic) third, person pronouns. This change is of considerable typological
(and other Romance languages) under very restricted grammatical conditions: impact and can be summed up under diverse headings such as 'from synthetic
bare plurals surface mostly postverbally with subject and object function to analytical coding' , from 'dependent-marking' to 'head-marking' or, in a
regardless of the lexical entry of the noun (normally with non-specific minimalist framework, as changes in the lexical feature structure (see below) of
interpretation of the nominal, c f. Id). Bare singulars are also possible with universal functional heads such as C, T, v and D. Alternatively, one could
'mass-denoting nouns' (not with 'entity-denoting nouns' in so-called singular think, as Vincent 1997 did for definite nominals, of the rise of new functional
countable NPs, see la) postverbally with subject and object function (see lb categories in Romance via reanalysis of ambiguous syntactic input or, less
and le), again with non-specific interpretation of the nominal. Conversely, the radically, of a different parameterized projection of functional categories in
only Romance language which almost never permits bare noun phrases in different languages. Whatever name we give to the change, it has to be
argument position is French. described and explained in a w ay that satisfies both the aspect of the initial
Secondly, every Romance language possesses a so-called indefinite article innovation, which is most probably linked to universal cognitive principles, and
derived from the Latin numeral unus, "one", which accompanies singular count the later individual mutations of the different Romance linguistic systems.
noun phrases. However, only two of the three Romance languages discussed These differ to a considerable extent, both in their history and in their present
here, i.e. French and Italian, have a further indefinite determiner, [Link]-called morphosyntactic make-up, and their linguistic evolution can therefore not be
'partitive article' derived from the composition of Latin de and the definite subsumed under some vague general functional coding or decoding strategies,
article. It marks indefmite non-countable singular noun phrases, usually with as will be shown below. In the following I shall in particular argue in favour of
'mass-denoting nouns ' , in pre- and postverbal subjects and objects. It is a formal account of the morphosyntactic changes leading to the new feature
obligatory in French and (still) optional in Italian, where its use is increasing structure of a functional head 'Pl ' in addition to the already (now widely
constantly, however (cf. Korzen 1996). Functionally, its morphological plural accepted) projection NumP inside nominals 4 and their· language-specific
,

(Fr. des, It. dei, degli) is to be considered as the plural of the indefinite article features triggering different overt· realizations of nomin al determination in
(see examples l b-d). different Romance languages.
If we compare this picture to Latin, the main differences or changes within
3 Prepositional p hrases are typi ca l syntactic contexts in which nominal determination does not
nominals become immediately obvious: for indefinite nominals as well as
need lo be marked overtly, even in languages w i th nominal determination, cf. modem
definite ones, obligatory determination has become a property of Romance Romanian. However, modern French, Italian or Spanish would require a definite determiner in
languages in contrast to Latin. S ee example (2): tbe case in q uest i on .
4 Cf. especially Ritter (1993, 1995), Picallo (1991, 2005), M unn & Schmitt (2005) and
Zamparelli (2004, 2005) for the discussion ofNumP.
60 ELISABETH STARK
THE ROLE OF THE PLURAL SYSTEM IN ROMANCE 61

The remainder of this article is structured as follows: section 2 gives an

neuter gender are 'mass-denoting nouns' .7 In addition, (3c) shows the well­
overview of the nominal morphology and indefinite determiners in French,
known 'collective ' semantics of the Latin neuter plural ending in -a (cf. SchOn
Italian and Spanish compared with the Latin declensional system in order to
1 97 1 , Windisch 1 973). Although theses oppositions are not systematic, the
reveal two fundamental correlations. Section 3 introduces the formal model,
Latin neuter and especially the Latin neuter plural in -a both unambiguously
-
following Chomsky's most recent proposals for modelling agreement processes
. marked in spoken and written varieties - can be re-interpreted as a partly gene­
and for deriving the correct morphological and semantic quantification of
ralised 'classification system' mainly denoting the opposition between ' single,
nominals (section 3 . 1 ), before applying it to indefinite nominals in modem
contoured object' (e.g. one piece of cheese, one olive, one berry) and 'non­
French, Italian and Spanish. Section 4 tries to retrace the morphosyntactic
contoured substance' (e.g. cheese, oil) or 'collective' (grapes).
evolution from Latin to modern Romance languages by applying the 'probe
Furthermore, this important semantic opposition is as much related to
model' to questions of diachronic syntax, especially to the functional proj ection
gender as to number: the Latin plural is neither automatically interpreted as
PI(*), after which the conclusions given in section 5 will summarize the
additive (cf. Link 1 99 1 ) nor restricted to 'entity-denoting nouns ' . See examples
findings and relate them to some parallel approaches in the recent generative
(4a-b):
literature.
(4) a. frig ora caloresque "an intense heat and cold": plural indicating intensification
2. Afunctional 'explanation': Romance indefinite determiners as b. acquae "waters", cerae "wax tablets": different appearances of a substance8
'classification devices'
I have repeatedly pointed to the fact tbat Romance indefinite determiners Although the Latin plural can have a sortal reading, bare plurals of abstract
can be understood as 'classification devices' in a broad sense, coding what or 'mass denoting-nouns' are not automatically re-categorized as they for
Seiler in his functional universal dimensions calls "apprehension":5 instance are in modern Romance languages (compare Fr./wile "oil", des huiles
"different smts of oil"). Virtually any Latin noun can be pluralized, and
APPREHENSION is the universal operational dimension with corresponding frequent occurrences of plurals of 'mass-denoting nouns' or abstract nouns, as
subdimensions which explicate the grasping and representation of concepts
in (4) above, are in fact attested. 9 This indicates that Latin had no
corresponding to objects or things by means of language. (Seiler 1986:145)
gramrnaticalized 'countability distinction' at the level of noun phrases.
In Latin, the basic notional distinction between an individualized obj ect and In contrast to Latin, modern Standard French, Italian and Spanish have
some uncontoured substance could be coded by using gender and number, almost completely lost the possibility of 'classification' via gender/number
especially by using the semantic and morphological residue of the lndo­ alternations (and declension class). The main morphological changes in the
European opposition between neuter and masculine/feminine. See examples nominal system from Latin to Romance are well-known and comprise the
(3 a-c):
7 In describing the facts in these terrns, following Lobe] (1993:192ff.) and with reference to
Gil's (1987) typology I assume a fundamental difference between the lexical categories 'mass­
(3) a. caseus "one single (piece of) cheese",
denoting', 'entity-denoting' and 'abstract noun' (N), which derive f rom characteristics of the
olea "olive" /"olive tree"
potential (extra-linguistic) referents and are based on denotational properties of the head noun,
b. caseum"cheese as a substance", and the countability or non-countability of entire noun phrases. This last opposition is a
oleum "oil" grammatical category or a feature which depends on the internal syntactic structure of the noun
c. acinus/acinum "berry", phrase and is mainly characterized by compatibility with certain indefinite determiners
acina "grape", (French/Jtalian: unluno ' one' , ' a', [Link] vs. duldel 'of. the ' , PART= partiti ve article).
frumentum "wheat", This assumption is justified by the fact that virtually any noun in Romance languages (like in
frumentu "com"6 any language with a grammaticalized countability distinction in this sense) can in principle
appear in any kind of noun phrase, cf . German Nach dem Unfall lag jede' Menge Auto he rum,
(3a) shows lexical roots with masculine and feminine gender, resulting in "After the accident, there was much car lying around".
8 Cf. Kiihner & S tegmann el955:69, 73), Hofmann & Szantyr ((1997) [1965]:18, 21).
'entity-denoting nouns', whereas the nouns from the same root in (3b) with
9 Cf. lturrioz Leza (1986:295f.): ''This individualization strategy [= pluralization of abstract
nouns, E.S.] is more widespread in the classical languages (Greek, Latin) than in modem
5Cf. Seiler (1986) and Stark (2005, 2006, and forthcoming). Gerrnan or any other European language; it is thus often difficult to translate an abstract [plural,
6Cf. in detail Hofmann & Szantyr ((1997) [1965]:7-10), Meisterfeld (l998:56ff.) and for Late E. S.] NP without changing its number: [ . . . ] Asperitates viarum et angustiae [ . .. ] ' The
Latin analogical neuter plurals following the same pattern cf. Morani (2000:228). roughness(es) and narrowness(es) of the ways'".
62 ELISABETH STARK
THE ROLE OF THE PLURAL SYSTEM IN ROMANCE 63

complete loss of morphological case, a reduction of gender (especially loss of

muro "wall" etc.), have a plural form in -a when denoting a plurality, body
the neuter), contrasted with a solid formal preservation of number, but now
parts or a 'collective reading'. However, they form a plural in -i when used
with only the 'additive plural meaning' left. In addition, modem Standard
metaphorically to denote something similar in form but without a collective
French shows the complete loss of the declension classes (already lost in Old
denotation (e.g. It.: le braccia denotes both arms of an animate being, whereas i
French, cf. Delfitto & Schroten 1 99 1 : 180f.). Gender and number are usually bracci denotes the arms of a river, It. le mura denotes the city wall, whereas i
marked (in the phonetic code) by prenominal determiners only. Modem
muri denotes the single walls of a building). Unlike Italian, Spanish does not
Standard Italian is different from French in that it has preserved 3 main
seem to preserve a certain 'classification potential' in nominal (declensional)
declension classes, 2 overtly marked genders and overt number marking. endings; the only slight 'classification potential' left in Spanish is the
However, the declensional endings -a and -e are far from being unambiguous
erroneously named 'neuter' (deriving from the Latin neuter singular) in the
markers of singular or plural, as they can either indicate feminine singular,
pronominal system which refers to CPs, i.e. quotations, statements of fact etc.
(rarely) masculine singular or feminine plural (-a) or masculine singular or
(cf. Picallo 2002) . 10
feminine plural (-e). The morpheme -o unambiguously indicates singular, but
If we try to relate the findings concerning the paradigm of indefinite
both masculine and (rarely) feminine gender. Just like Italian, modem
determiners with the nominal morphology shown by French, Italian and
(European) Standard Spanish has 3 main declension classes, 2 overtly marked
Spanish up to now, we can observe two important correlations:
genders and overt number marking. It is unambiguous with respect to number
Firstly, the possibility of having bare arguments, especially bare plurals in
marking ("plurality is manifested consistently with the suffix 1-sf', Harris
argument position (Italian, Spanish vs. French), correlates with the formal
1 992:67). See (5a-c): preservation of the Latin 'neuter' in nouns or pronouns, not only with the possi­
11
bility of overt plural marking. Secondly, the presence of a 'partitive article'
(5) a. un ami (m) I une amie (f.)-des ami(e)s (French)
[renami I ynami -dezami] explicitly coding non-countability (Italian, French vs. Spanish) correlates with
12
·

"a male friend"I"a female friend", "male/female friends" ambiguous plural marking.
If we try to explain the general language change from Latin towards
b. Sg. : -o I -a
I -e (Italian)
PI.:-il-a/-e Romance in relation to the re-structuring of nominal morphology and the
grammaticalization of nominal determination for indefinite argument nominals
Iibr-a- libr-i (m.) mano- mani (f.) bracci-o- bracci-a (m.-f.)
too, we could now easily take up Givon' s observations about the widespread
"book"-"books" "hand" - "hands" Hann"-"anns"
cas-a- cas-e (f.) poet-a-poet-i (m.) grammaticalization of the numeral "one" into an indefinite article (cf. Giv6n
"house''-"houses" "poet"-"poets" 1981) and thus an indicator of countability via implicature ('if something can
can-e- can-i (m.) be quantified as being one element out of a set in contrast to two, three and so
" dog"-udogs"
on, it is a member of a set of countable discrete entities'). This element could
c. Sg.: -o I -a I -e; (Spanish) thus have compensated for the loss of the complex Latin declension and
PI.: -s: assumed the function of 'classification' in a broad sense. Even if this
pas-a- pas-os (m.) man-o- man-os (f.) 'explanation' seems convincing at first sight, it leaves us with a lot of open
"step" -"steps" "hand"-"hands" questions and does not provide a satisfying structural descriptive analysis of the
pas-a- pas-as (f. ) map-a- map-as (m.) individual phenomena occurring in the morphosyntactic structure of the
"raisin"-"raisins" "map"-"maps" different Romance languages. In particular, the new syntactic property of
jef-e-jef-es (m.) nub-e -nub-es (f.)
obligatory coding of countability in Romance indefinite nominals in contrast to
"chief'-"chiefs" "cloud"-" clouds":

10 Spanish personal pronouns and demonstratives show a threefold morphological opposition,

Please note in (5b) the residue of the original Latin classification potential
with fonns ending in -e for masculine s ingul ar, -a for feminine singul ar , and -o for the so­
of the neuter plural in -a as opposed to a regular plural form in -i (originating in called 'neuter' (e.g . span. no quiero pensar mds en el/el/alel/o "I do not want to think about
Late Latin, cf. Hofmann & Szantyr ( 1 997 [ 1 965] :2 1 ) and reanalysed as him/her/it any longer").
11
feminine (plural), always indicating a collective or at least 'pair' reading). Conlrary to what is postulated in Schroten (2001).
12
Some nouns ending in -o (masculine singular), usually denoting concrete Cf. Stark (in prep . ) , with reference also to older stages of the R omanc e languages and the
simultaneous development of the ' partitive article' and the loss of overt or unambiguous plural
objects like body parts (It.: ginocchio "knee", orecchio "ear" and so on, also
marking in French and Italian.
64 ELISABETH STARK
THE ROLE OF THE PLURAL SYSTEM IN ROMANCE 65

Latin remains unexplained, and the considerable differences between the single commanded domain for a suitable goal, i.e. an item with matching valued
Romance languages which we showed above in section 1 are completely features. B oth, the probe and its goal must be 'active' for agree to apply, i.e.
neglected - as are the specific properties of many other languages which have they must contain at least one uninterpretable!unvalued feature.15 After match,
grammaticalized "one" into an indefinite article and are nevertheless very the operation value takes place: the valued features of the goal are copied onto
different from Romance in many respects (English, German, Hebrew and so the probe, which can, but does not have to, instantiate an unvalued feature of its
on). In addition, we have to bear in mind that "[u]seful or needed things are not goal. After this agreement process the goal is mobile for syntactic movement.
sufficiently explained by their usefulness or the need for them" (Haspelmath In order to apply this model, originally conceived for agreement processes
1 999 : 188). at the sentence level (e.g. person and number agreement and nominative case
assignment between T and the subject DP), to the internal structure of nominals
3. A formal account of Romance indefinite nominals (gender and number agreement and its interaction with indefinite determiners)
3.1 'Manufacturing plurality' and in order to derive their correct quantificational interpretation, we will use a
Following the formal modelling of agreement in the latest versions of the modified version of Heycock & Zamparelli ' s (2003) proposal for English
13
minimalist program, lexical items and more generally heads "interact through indefinite nominals. Following Heycock & Zamparelli (2003), we assume first
agreement to influence the shape of syntactic structure and the process of that nouns do not have a specific interpretation with regard to 'count' or 'mass'
semantic interpretation" (Pesetsky & Torrego 2004: 1 ) . Lexical items can be or anything else but that their denotation is merely a set of singleton elements
described as having ce11ain bundles of grammatical features, so-called ((J­ irrespective of their morphological number (cf. Heycock & Zamparelli 2003:13
features, some valued, others unvalued (e.g. the number feature Nb on nouns is for a detailed argumentation). The interpretation of a nominal as being
normally valued and can have the value singular, S ING, or plural, PLUR, 'semantically pluralized', i . e. as denoting "a set composed in all the ways in
whereas it does not come already valued from the lexicon for determiners or which these [elements, E.S.] can be grouped together into pluralities" (as
adjectives, which accordingly have to agree with the noun in nominals, cf. "having a join semilattice structure", with the notation [+LATT]) is created by
Pesetsky & Torrego 2004: 1 f. ) . Features can thus be described as attribute-value merge of a noun with a functional head, 'PI*' which operates semantic
pairs. Moreover, features are considered to be either interpretable or pluralization (i.e. creates a 'semilattice interpretation ' , cf. Link 1983). The
uninterpretable: "This distinction is concerned with a different question: semantic feature [±LATT] is introduced by Heycock & Zamparelli (2003) in
whether or not a feature of a particular lexical item makes a semantic order to describe the contribution of the functional head PI* to the
contribution to the interpretation of that item" (Pesetsky & Torrego 2004:2). interpretation of the whole nominal. Pl* creates sets of sets of different sizes
The number feature is generally considered to be interpretable on nouns, out of the set of singleton elements denoted by the NP:
whereas gender in modern Indo-European languages does not seem to make a
semantic contribution to the respective lexical items (but see Radford For example, with 4 individual tall boys a, b, c, d in the domain, [NP tall boys] will
2004: 288, Picallo 2005 and the discussion below). There is a direct relation denote {{a), {b), {c), {d)), and PIP will be:
between the notion of (un)valued features and (un)interpretable features: if a 11 [p1p tall boys] !I= {{a, b, c, d), {a, b, c), {b, c, d){a. c, d), {a, b, d), {a, b), { a, c},
{a, d), {b, c), {b, d) , {c, d), {a}, {b }, {c), {d)}.
feature (e.g. number) is interpretable under a given category, it enters the At the next level up, NumP behaves as a filter. It regulates the cardinality of the PIP
derivation with a specified value (e.g. plural). In contrast, if the feature is denotation, eliminating from it all the pluralities with the wrong number of atoms [ . . . ].
uninterpretable, it i s not specified; it will be valued via agree. That is, (Heycock & Zamparelli 2003: 13)
uninterpretable features must be deleted before Spell-Out, and in order to be
deleted, they must be valued by agree, so that agreement can be understood as This denotation of plural noun phrases is parallel to the one of mass terms,
directly related to the interpretability of syntactic structures by the semantic which also denotes sets of sets of portions of the same substance (cf. Link
component. 14 Agreement processes are conceived of as 'probing' processes: 1983). Later we will mention some structural parallels between plural noun
Unvalued features on a given category item act as a probe and search in their c- phrases and non-countable nominals, independent of morphological number
and demonstrating the relevance of the [LATT]-feature in the nominal
" Cf. Chomsky (2000, 2001, 2005), Radford (2004, chapter 8), Heycock & Zamparelli (2003),
derivation. The functional head Pl(*) is also responsible for the morphological
Zamparelli (2004, 2005), Pesetsky & Torrego (2004), Mensching (2005), Mensching & number agreement inside the nominal . Heycock & Zamparelli (2003) work
Remberger (2006).
14 For details see Radford (2004:289f.).
15 Cf. Radford (2004:289ff.).
66 ELISABETH STARK
THE ROLE OF THE PLURAL SYSTEM IN ROMANCE 67

within a framework which still assumes feature checking, as in earlier versions NumP. Numerals are lexical ly specified for their semantic [±LATT] feature
of minimalism, and they apply the checking mechanism both for and have to possess the same [±LATT] feature as Pl(*) in order to derive a
morphological/grammatical and semantic features (essentially number and grammatical structure; the numeral "one" has [-LATT] , numerals above "one",
[±LATT]) , whereas we will apply the probe-model and assume agree processes [+LATT] . Heycock & Zamparelli 2003 ( 17f.) assume here a traditional feature
only between grammatical attribute-value pairs. Thus, in our framework the checking mechanism between Num and Pl(*), which avoids, e.g., the
number feature of Pl(*) is unvalued ([Nb: ]) and acts as a probe. This feature ungrammatical structure *one oils. We will not go into any further detail here,
will be valued by agree with the lexical noun (cf. 6a) and gets deleted after given that the problem of explicit quantification by numerals is not central to
valuation (cf. 6b), i.e. it gets invisible to the semantic, but remains visible to the our discussion. Also, NumP controls number agreement outside the nominal ,
phonological component. The interpretation of the nominal depends ultimately e.g. in cases where the respective nominal is the sentential subject and has to
on the different propetties of the heads in the derivation and on the interaction agree in number with the finite verb .
of different agreement processes in the structure: There are several arguments in favour of assuming that two (rather than
only one 18) functional projections are involved in the plural marking of a
(6) a. probe and match b. value and delete
nominal. One important argument comes from languages like Hungarian,
PIP PIP which combines singular N with cardinals higher than "one":
� �
N N
I , .......,. [Nb:
PI* Pl*
(8) a. probe and match b. value and delete
probe ] [Nb: PLUR] [[Link]: PLUR][Nb: PLUR]
: TT]
[+LA .
i.�........
.
oils ... .. .. ..... . .
[+LATT] oils
;
!... .. ... ..... . ..l �
NumP

PIP Num
probe & match value & delete PIP
� n= [5] �
PI* N at PI* N
Note that, in this derivation, N does not have any uninterpretable features I probe I··> [Nb: ] [Nb: SING] ('five') [Nb: SING] [Nb : SING]
and is thus theoretically not 'active' in the sense described above. On the other [+LATT] haj6t [+LATT] haj6t
l!
. .. . . .. ... . . . . ..... .. .
('ship') � ('ship')
hand, the probe in PI(*) has to be instantiated for morphological number - and
L. . ... . . . . .. . J L,___,_ . . .. . .. ._.1
•

there is only one matching element in its c-commanded domain, N, with a probe & match value & delete

valued Nb feature. For English nominals, therefore, we have to assume16 that

agree takes place in this way, given the fact that the probing operation cannot
If morphological number agreement and 'semantic pluralization' plus
take place erroneously between a probe and an inadequate goal. indication or filtering of the correct cardinality were situated in the same
In the next step, PIP can optionally be merged with Num, the head that functional proj ection, say Num, the Hungarian structure could not be accounted
hosts cardinals, weak quantifiers etc . : 17 for, as Num would host two contradictory features, i.e. the cardinality "five",
thus selecting a set of five elements out of a set of sets and regulating morpho­
(7) merge with Num
logical plural agreement outside the nominal, and a probe being valued with
NumP morphological singular number inside the nominal .
� Most importantly for our argumentation, by separating morphological and
Num PIP
semantic number (SING vs. PLUR and [±LATT]), Heycock & Zamparelli
n=[5J �
(jil'e) PI* N (2003) can account firstly for the difference between ' count' and ' mass'
[1>!8: PLUR][Nb: PLUR] readings of singular nouns in languages like modern English, German or
[+LATI] oils
Spanish and secondly for the striking parallels in the syntactic behaviour of sin­
gular mass and plural nominals in many languages: In these languages, it is not
Since NumP acts as a filter regulating the cardinality of the PIP denotation the noun itself that carries [±LATT] information, nor is it the morphological
(see above), only sets containing five elements will be left in the denotation of number of a nominal, as virtually any noun can appear in both a countable or
16
18
Just as Pesetsky & Tonego (2004:1) seem to assume tacitly for Latin nominais as well. Like Ritter (1993 and subsequent work) with NumP only.
17 Cf. He ycoc k & Zamparelli (2003:llff.).
68 ELISABETH STARK THE ROLE OF THE PLURAL SY STEM IN ROMANCE 69

non-countable nominal. It is merge of N with PI* irrespective of the morpho­ 3.2 Romance languages: gender, number and 'classification'
logical number of N which creates a [+LATT] denotation. 19 How can we now relate this model to our initial problems, i.e. the striking
Now, what happens if N is not merged with PI* and is therefore not seman­ differences between French, Italian and S panish with regard to their systems of
tically pluralized? In the case of a singular N, a default element, the indefinite indefinite determiners, the possibility of having bare nominals in argument
article, has to enter the enumeration in order to derive the [-LATT] denotation: position and the respective language change from Latin to Romance? First of
all, Romance languages and Latin have overt gender as an additional cp feature
(9) a. probe and match b. value and delete besides number, a feature which is usually considered to be uninterpretable. As
NumP gender is expressed overtly at least on some cardinals, weak quantifiers and the
� indefinite article and as it has to be valued there in order to be deleted,22 there
PIP
has to be a probe for gender in Num, 23 and there will be two subseq uent agree
Num
PIP
n=[I] � operations to carry out, the first one for morphological number between PI* and
PI N
� [l'lll: S�lG] [Nb: SING] N (but see below) and the second one for morphological gender between Num

[,£�},,
PI N
[-LATT] oil and N (PI* in Spanish is not a matching goal for the probe for gender). Num
u·� [Nb: ] [Nb: SING]
* (an) boy controls gender agreement inside the nominal. A structure like the following
[-LATT] ail
.
boy
*(a) J one must accordingly be assumed for Spanish:
.. 1
. ... . . . .-.......J
L ... . ..
... !. .... .J
probe & match value & delete
,_ ············-···-··..···

(10) a. agree between PI* and N

PIP
According to Heycock & Zamparelli (2003), the obligatory presence of a(n) �
in a structure of this type is due to the fact that PI would not contribute any­
I ''"""">
PI* N
thing semantically to the denotation of the whole nominal. They further assume probe [Nb: J [Nb: PLUR]
·· _. ..
that a functional head which is semantically inactive and phonologically empty
_ •.. . · · .··
[+LATI] [Gd: FEM]
should not exist in the structure. 20 Alternatively, we could say that the l aguas
1 �
L ..... .... .................:
[±LATT]-feature in PI(*) needs an overt expression in a language for one or the probe & match
other value, the opposite value being implicated in the absence of overt coding.
Furthermore, PI(*) is syntactically necessary for phrase internal morphological
(IO)b. agree between Num and N
number agreement, which does not automatically coincide with external
NumP
morphological number agreement (cf. Fr. la plupart des gens fument, Engl.
this!*these committee are to meet tomorrow) . This is enough reason to assume ------
Num
I 1•- ...
PIP
its existence inside (indefinite) nominals (see below). Be it as it be, in the case probe ·>[Gd: J �
of the absence of Pl*, the Pl position in English is filled with an element n>[l] PI* N
providing the [-LATT] interpretation for the whole nominal, thus (muchas)[l'll'l: PWR][Nb: PLUR]
disambiguating ovet1ly the semantic interpretation of the whole structure. [+LATT] [Gd: FEM]
Contrary to the Romance languages and their indefinite article deriving from aguas
the Latin numeral unus (see below), English a is merged in PI on the basis of
probe & match
different distributional properties (cf. the impossible literal translation into
1
Romance of the English many a time, too tall a boy etc. ) ? Functionally, the
obligatory indefinite article in Romance countable nominals assumes the
22
parallel coding of [-LATT] , among others functions (see below). Cf. Hey cock & Zamparelli (2003: 19f.).
23 This, then, is an explanation for the necessary location of gender in Num: due to the
mechanism of agree and feature deletion immediately after merge (cf. the 'Earliness Principle',
19 Cf. Heycock & Zamparelli (2003:16f.).
which requires operations, especially agree, to apply as early as possible in a derivation, cf.
2° Cf. Heycock & Zamparelli (2003: I 7). This is also in complete accordance with Economy
Radford 2004:282), gender cannot be a probe together with number in PI(*), as it would get
Principles assumed already in Chomsky 1995.
21 valued and deleted.
Cf. Heycock & Zamparelli (2003: 17f .).
70 ELISABETH STARK THE ROLE OF THE PLURAL SYSTEM IN ROMANCE 71

(lO) c. resulting configuration situated in Num, and agree will therefore take place not between Num and/or
NumP
PI(*) with N as in English or Spanish. but between Num and PI(*) only. We
� should remember that, in Romance, NumP is the place for gender agreement
Num PIP inside and outside the nominal (cf. Heycock & Zamparelli 2003:19f.).
[[Link]] � Additionally, we assume that, for French at least, number is valued (and inter­
t1> [ I] PI* N
(muchas)[�la: PLUR][Nb: PLUR] pretable) in NumP and has to be assigned to PI(*) after the agree operation
[+LATT] [Gd: FEM] (remember that fully valued probes can after value assign unvalued features to
aguas
their goals, like the person-number-gender probe in T assigns nominative case
to its 'goal DP', see above, the beginning of section 3.1). Evidence for this
In the case of agua, we would get un(a) for PI [-LATT] just like in English. comes from the morphological difference between plural and singular
Yet, in Spanish and in Romance in general the indefinite article is located in determiners in French - as on N in Spanish, the inflectional endings of the
Num, because it also realizes the gender feature of this category. In Spanish, French determiners indicate unambiguously whether a nominal is plural (obli­
after full instantiation of the probes, N can additionally move from its position gatorily ending in [-e)) or not. PI(*) is visible/active to the probe in Num
at least up to Num and further to D (as only DPs are arguments in Romance, cf. because its number feature is again not valued, but it is not itself a probe. For
Chierchia 1998 and Longobardi 2001b) and can thus appear as a bare plural in this unvalued feature we will use the notation ([Nb: A]).24 After valuation of
argument position (cf. also Picallo 2005:110, though assuming a slightly dif­ the gender probe of Num, the number value can be assigned to PI(*) (and
ferent denomination and function for PI(*), see below). The gender feature of N deleted afterwards). The derivation of French nominals is as follows:
remains undeleted in this and the following derivations, because it is consi­
dered to be interpretable, roughly following Picallo 2005, who assumes that (11) a. agree between Num and PI*
"gender declension is the morphological exponent of an abstract interpretable
NumP
feature in the functional layer of nominal structures" (Picallo 2005: 107ff.), an
'abstract classification feature' (cf. also Harley & Ritter 1999, 2002a,b for a ------
Nu m
!
PIP
feature·geometric approach to pronouns, where gender is the specification I probe + mJ>[Gd: l �
under a 'classification node'). Note that, theoretically, a probe for gender could [Nb: PLUR] PI* N
n [Gd: FEM] [ o]
also be situated in Pl(*) together with the number probe (as proposed in Picallo
I Nb: A] (eau/ xf)
2005 for Spanish and Catalan). However, the below discussion of Italian and [+LATT]
especially French will show that the assumption of two different probes in two
probe & match
different heads, one for gender in Num and one for number in Pl(*), makes
clearer predictions for the syntactic behaviour of indefinite nominals in the
respective languages. We will try to show in the following that the gender (1 1) b. resulting configuration
feature is ultimately indirectly related to the semantic [±LATT] feature
NumP
introduced by Heycock & Zamparelli 2003. The problem is that the location of
gender varies in the Romance languages, and the complete syntactic and Num �p
�
semantic structure of Romance nominals can only be obtained by assuming a --
PJU; ... L:ilVI]
complex interplay between the functional heads Num and PI(*) and their [Nb: PLUR] PI* N
respective number and gender features. n [Gd: FEM] [o]
As has been pointed out in section 2, French nouns do not have a morpho­ *(des) [Nb. PWR] (eau[x])
logical marker for gender and number. They could be analysed as having lost [+LATT]

any grammatical features and as not contributing to an agreement phenomenon

of any kind. The only information they provide is their descriptive semantic
content. Accordingly, morphological number must be specified and interpreted 24 Note that this unvalued feature does not act as a probe . It is comparable to the unvalued Case

elsewhere, for example in NumP, by obligatorily inserted lexical items which feature of N.

bear a valued number but an unvalued gender feature. A probe for gender is
 THE ROLE OF THE PLURAL SYSTEM IN ROMANCE 73
72 ELI SABETH STARK

(13) b. resulting configuration:

During this agree operation the unvalued feature for number of Pl *, which
doesn't act as a probe, is instantiated. That is, the agree operation also serves N u mP
for assigning number to PI*. Thus, as French nouns do not participate in
agreement processes at all, they remain totally disabled for movement to D - Nu � PIP

for this reason, bare noun phrases cannot appear in argument position in
French. In contrast to Spanish, and because of the lack of number information
([Link]]
[Nb:SING]
n
�
[Gd : FEM]
N
eau
in N and PI(*), French indefinite singular noun phrases exhibit a morphological *(une) (�tb. 8Rl6] ([o])
[-LATT]
element for singular nominals with a [+LATT] interpretation, the 'partitive
article', indicating morphological number and [+LATT]:
Just as in English, the indefinite atticle un(e) has to appear obligatorily with
(12) a. agree between Num and PI* b. resulting configuration French singular countable noun phrases; however, it has to do so in order to
NumP
indicate the [-LATT] interpretation (it also appears higher in the structure than
� English a, cf. Heycock & Zamparelli 2003:17f.) and to indicate singular
Num PIP NurnP number.
I probe j..,.......i»[Gd: ] �
' ·' ·' ,., ·
[Nb: SING] PI* N Nu � PIP
Italian now has overt expression of number, but this is quite ambiguous,
whereas gender, at least for the overwhelming majority of nouns ending in -o
n �--;---I--.:::ll'l'lj
[Gd: FEM]
[Nb: A]
eau
( [ o]) [Nb: SING]
* n
p� N and -a, is marked on the noun and therefore assumed to be a valued feature on
[+LATT] [Gd: FEM] eau N here:
..
L.... ... _.,_
_.l
probe & match
(de I') [�lb. 8Rl6]
[+LATn
(14) a . agreement:
NumP
Please note that, in the derivations (11) and (12), the number feature
�
remains undeleted in Num, just like the gender feature in PI*. We will come Num PIP
back to this later. I probe 2 1:.......- ...-l>[Gd: ] �
. . ,,
· .· ·• · [Nb: SING] PI* N

) . .. . . .. . . . .. .-·J>[Gd:
French countable nominals are to be derived in the following way: -!J. ....

. . ..
] [Gd: FEM]
�
__

[N b: A] acqua

(13) a . agree between Num a n d PI: ... I [+LATn

! t
L---··�---·-t !..__q·-·-. .1
NurnP probe & probe &
match match
�
Nurn PIP
I probe f· . .. .. J>[Gd: ] �
. . ' [Nb: SING) PI N (14)b. resulting configuration:
n [Gd: FEM] eau
[Nb: A] ([o])
! [- TT] �
.. . . .
!...... . .. .. ... ._,_,;
probe & match

/1
(dell')

This makes two subsequent agree operations necessary: first, one between
the gender probe of Pl* and N, and second, one between the gender probe of
 THE ROLE OF THE PLURAL SYSTEM IN ROMANCE 75
74 ELISABETH STARK

( l 5)b. resulting configuration:

Nom and Pl* . 25 During this second agree operation the unvalued feature for
number of PI*, which doesn' t act as a probe, is instantiated. That is, the second
agree operation also serves for assigning number to PI* , like in French. Just as
in Spanish, there are two agree operations, but this time two for gender. The (Gti: FEM]
main difference to Spanish lies in the fact that morphological number
agreement takes place between Num and Pl* and not between Pl * and N . u ova
(delle) (NI!. PLUR]
Here we can find a structural explanation for the empirical correlation of [+LATT]
unambiguous number marking on the one hand and the absence of a 'partitive
article' on the other. Thus, if agree has to take place between Num and Pl(*),
Finally, with a countable nominal:
the former possesses an element marking singular or plural, especially for
nomi nals with a [+LATT] interpretation. Furthermore, only if morphological
( 1 6) a. agreement:
number cannot be obtained at all via the complex interplay between gender and
number features in the structure, as it still can be in Italian (the ending -e e.g. is NumP
disambiguated as to its number feature, when the feminine or masculine gender ........-----_
Num PIP
is clear), this 'partitive' element will be obligatory, as in French. We can also I pro b e 2 b . ....... .....;,.[Gd: ] �
already understand the second correlation, the one between bare nominals ·"· · • • •· ··,., :·•. [Nb: SING] PI N
_!1...... .... .....,. [Gd:
. . . . . . .............. I . . [Nb: A]
(plurals) in argument position (Italian, Spanish vs. French) and the expression ] [Gd: FEM]
�. . ::-;t "' . .
r:::;: acqua
of gender as an overtly coded feature on N. In this case, N takes part in an . t
i [-LATT]
agree operation between N and Pl(*), and even in one between N and Num or L_!L...J L_JL...J
Num and PI(*) - and after these operations have taken place, it can move probe & probe &
match match
upwards in the structure into a functional proj ection compatible with argument­
hood in Romance languages like D.
As a fmther illustration of this, let us have a look at the derivation of Italian ( l 6) b . resulting configuration :

indefinite plural nominals:

( l 5) a. agreement:
�--�[Link]
NumP
........-----_
Num PIP un '
I probe 2 f . . .... ..... .. l>[Gd: � [ -LATT]
. .. .
. ]
[Nb: PLUR] Pl* N
·

I probe • [ .. .. . . -.. . . . . .!Ir ·--1> ��� Al

..
[ G� �Ml
o Again, there is an obligatory insertion of the indefinite article in order to
i [+LATT] t derive the [-LATT] property of the whole nominal in Num, but not in Pl
L----�-�-_t L. ,______l because of the gender feature.
probe & probe &
___

match match

4. From Latin to Romance: what happened in the structure ?

As Latin nouns are inherently specified (and marked) for gender and
25 Please note that, according to the "Feature Visibility Convention" in Radford (2004: 289), a number and as, at least for concrete denotations, the combination of gender
deleted feature like the gender feature in PI(*) after merge and agree with N remains visible to (especially neuter) and number gives relatively clear-cut infonnation on the
the syntax and PF even if it becomes invisible for the semantic component after deletion, so that 'classification ' of the intended referent (cf. section 2), we assume that the
the gender feature there can represent a matching feature for the gender probe in Num after
[±LATT] feature is a lexical feature on nouns in Latin, the functional projection
merge with PIP.
 76 ELISABETH STARK THE ROLE OF THE PLURAL SYSTEM IN ROMANCE 77

PI having only a probe for number. Indefinite nominals in Latin thus have the (1 8) a. b.

following structure: NumP

NumP
� �
Num PIP
( 1 7) a. neuter nouns: Num PIP
[Gd: ] � [Gd: ] �
NumP n PI N ?
• n PI N
� [Nb: ) [Nb: A]
[-LATT] [Nb: ] [Nb: A]
Num PIP [Gd: B)
un- [Gd: B]
[Ga: NEUTR] � [-LATT]
n PI N ole-
ole-
wwm [Nb. SING] [Nb: SING]
[Gd: NEUTR]
[+LATT]
Reanalysis defined in its original acceptation as a "change in the structure
oleum
of an expression or class of expressions that does not involve any immediate or
intrinsic modification of its smface manifestation" (Langacker 1977:58) could
( 1 7) b . feminine nouns:
be seen here in the rearrangement in the feature structure of Num. B ut there are
NumP considerable problems for the modelling of gender and number agreement
� (unvalued gender feature in Num, no value for the number feature etc.) inside
Num PIP indefinite Late Latin and future Romance nominals, so that the right-hand
�Y·J � derivation is simply not acceptable.
11 PI N
una fNb. SH!Gl [Nb: SING] What exactly happened during the syntactic evolution from Latin to
[Gd: FEM I Romance, then? How can we adequately describe and explain the actual
[-LATT] syntactic make-up of Romance together with its differences from Latin, which
olea
are also topologically relevant? Please have a look at the following structure,
which evolved from the left-hand one in the above Latin derivation:
If we abstract away from the Latin case marking system, there is a gender
probe in Num in this derivation, as Latin indefinite determiners and quantifiers, (19) a. b. French:
including numerals, can be marked for gender, j ust like it has been assumed for NumP
modern Romance languages. However, PI does not have a [±LATT] feature �
responsible for 'semantic pluralization' - the interpretation of the denotational Num PIP
characteristics of the whole nominal is a feature of N. This is reflected by the [Gel. FEM] �
[Nb: SING] PI N
fact that there is no direct association in Latin between morphological plural
11 [Gd: FEM] eau
and [+LATT] , as has been shown above in section 2. Thus, the number probe U/1 acqua une
[r<U-.--U'"IT"f'
- CC"J [Nb. SH!G] ( [ o] )
and the [±LATT] feature have to be separated in languages without a gramma­ [-LATT] [-LATT]
ticalized 'countability di stinction ' .
Now, imagine a system which loses to a considerable extent its gender- and The reanalysis of "one" as a countability marker and the whole
number marking devices through the (partial) loss of inflectional endings - like 'countability grammaticalizati on' that took place after the Latin declension
Late Latin : not only case, but also number and especially gender, carrying the system broke down led to the dichotomic coding of [±LATT] in Pl( *) as the
most important conceptual distinction ('apprehension' , 'classification ' ) besides new location for ' semantic pluralization' , but also as a key projection for the
(in)definiteness for a nominal to be an argument. Of course, the speakers of this internal gender and number agreement inside Romance nominals. 26 In central
system can 're-analyse' its numeral(s), especially "one", as a 'countability mar­
ker' and as a singular marker and thus compensate for the inflectional classifi­ 26
Cf. the definition of grammatica/ization in Roberts & Roussou 2003 : "[ . . . ]
cation system. However, this appealing functional explanation is too simplistic grammaticalization is reanalysis 'upwards' along the functional structure. Since movement is
when it comes to the details of syntactic computation: always local and upward, categorial reanalysis is also local and upward" (Roberts & Roussou
2003 : 7 1 ).
 THE ROLE OF THE PLURAL SYSTEM IN ROMANCE 79
78 ELISABETH STARK

(20) lbero Romance

':Romance languages like French and Italian, where number marking is
problematic on N (gender marking as well in French) , the number feature NurnP
appears in PI(*), as has been postulated for English or Hungarian on the basis �
of independent facts by Heycock & Zamparelli (2003) (see section 3.1 above) . Nurn PIP

This number feature is unvalued and has to receive a value by agree, but cannot
fCd: FEM] �
n > [l] PI * N
be a probe, and for thi s reason, an additional agree operation has to take place (mucha) f�lb: S�iq [Nb : SING]
between Num and PI( * ) without directly or exclusively involving N (as in [+LATT] [Gd: FEM]
Latin) . There is now a valued number feature in Num, together with the agua
ori ginal gender probe, while a valued gender feature for French is situated in
PI( * ) . This structure, derived from the defective one in Late Latin, now with a
One, for example a functionalist, could n o w ask why the assumption o f a [±LATT]-feature in PI(*) like in English, is fundamentally different from the
functional proj ection, PI(*), with a [±LATT] feature, would be necessary in central Romance one : just as in Italian, two agree operations have to take place,
order to account for the morphosyntactic facts in modern Romance. The and the result is the possibility of moving N to a higher proj ection in order to
location of a valued number feature in Num seems to be enough, together with make it available for argument position even if it is bare. In contrast to Italian
the re-interpretation (rather than the re-analysis) of "one" as a countabil ity and French, however, there is no association of gender with the [±LATT]
marker. Now, besides the fact that the assumptioq of creating (morphological feature in PI( * ) , and there is therefore no ' classification' system and no explicit
and semantic) plurality in Num and filtering it at the same time has been marking for the [+LATT] value in the singular case - the 'partitive article' .
rejected on independent grounds by Heycock & Zamparelli (see section 3.1 Thi s could not be explained if we assumed that gender and number were both
above), i t also does not solve our agreement problems, and, more importantly situated in Num and functioned as a probe (Spanish) , as a partial probe (Italian)
still, it does not allow us to understand the interaction of noun morphology, or as valued features (French) .
nominal classification and nominal determination discussed in section 2 above.
It would not give evidence for the obligatory. insertion of the numeral "one" in 5. Conclusion
Num in indefinite singular countable nominals. Gender seems to play a key In this paper, I have tried to describe and explain the grammatical change in
role in this story, and, both historically and functi onally, it makes perfect sense the evolution of some devices of 'nominal classification' in a broad sense from
that it should do so in relation to a head with a [±LA'IT] feature: gender classes Latin inflectional nominal endings (overt gender and number marking) to
are assumed to have been directly related to extra-linguistic referent classes in Romance indefinite determiners . Comparing the noun morphology and the
Proto-Indo European (cf. S chOn 1 97 1 , Windisch 1 973). Also, in our model the need for indefinite determination in argument position in French, Italian and
gender feature is situated in PI( * ) , either as a probe (Italian) or as a valued Spanish allowed us to observe two main correlations : firstly, the possibility of
feature (French) - the functional proj ection now responsible for ' semanti c bare arguments (Italian and Spanish vs. French) correlates with the preservation
pluralization ' and, more than anythin g else, for the correct derivation of a of a quite clear-cut overt gender marking morphology on the noun (in contrast
'mas s ' or ' count' reading of the (singular) nominal . All of this is only indirectly to Delfitto & Schroten 1 99 1 , who see an interdependency with number
related to the filtering of the correct quantity that is done in Num (see section marking) ; secondly, unambiguous number marking correlates with the absence
3.1 above). Also, if we compare the structures of ( 12b) and ( 1 4b) above with of a classifier for ' mass readings' (the 'partitive article ' : Spanish vs. French and
the one for Spanish (and other Romance languages with unambiguous plural Italian) . If the loss of nominal declensional endings in Latin is analysed as the
marking) in (20), we can derive the possibility or the need for a 'partitive loss of the marking of the fundamental 'mass-count distinction ' via noun
article ' in Italian and French : morphology, the rise of the numeral "one" as an indefinite article indicating
' countability' , a functional and typologically consistent analysis, is easily moti­
vated. However, only fine-grained formal analyses of the different agreement
proces ses inside indefinite nominals in different Romance l anguages based on
Chomsky ' s 'probe model ' have revealed the syntactic change in Romance
nominals which, just as it did for definite nominals with overt D-marking, led
to the assumption of new features in PI( * ) (and Num) , as a head responsible
80 ELISABETH STARK THE ROLE OF THE PLURAL SYSTEM IN ROMANCE 81

Finally, if N i s not involved i n an agree operation of any kind, it will not be

now for the semantic [±LATT] feature. This assumption is consistent with
other syntactic accounts of the change from Latin to Romance, the great enabled for syntactic movement to a higher proj ection, and consequently bare

maj ority of which show a tendency towards the explicit marking of former nominals will not be able to appear in argument position, which is true for
27 French but not for Italian and Spanish.
inherent (semantic or pragmatic) features .
The formal account presented only as a sketch in this paper and based on
PI(*) and its feature make-up seems to play a central role in determining the
morphological structure and syntactic behaviour of Romance indefinite independent assumptions has allowed us to derive and explain the syntactic

nominal s : if there is a gender feature, whether unvalued or valued (i.e. inter­ behaviour of Romance indefinite nominals in opposition to the morpho­

pretable, see also section 3 . 2 above) in PI( *) (besides Num), explicit markers of syntactic regularities of Latin, and this in a much more detailed way than any
(diachronic) functionalist description of the facts would have been capable of.
[±LATT] for the singular exist (the indefinite and the 'partitive' article) and are
Functional considerations are important for uncovering the general cognitive
merged in Num, inducing an agree operation between Num and Pl ( * ) for
number, and, in Italian, one for gender between Pl(*) and N (thereby valuing motivations for innovations in l anguages and changing systems ; they are

the former uninterpretable gender feature; PI(*) is therefore always the related to the interaction of the interfaces with the syntax, but in principle they
are too general and extra-linguistically driven to be able to explain
syntactic location with an interpretable gender instance) . The question of the
morphosyntactic regularities in language change related to the 'narrow syntax ' .
interpretability of features would thus be in line with the conception proposed
in Pesetsky & Torrego 2004: "an uninterpretable feature must enter an Agree
relation with an interpretable counterpart" and the requirement that "every
feature must have at least one interpretable instance" (Pesetsky & Torrego
REFERENCES
2004 : 8 ) . This analysis is also quite parallel to the one presented in Picallo
Chierchia, Gennaro.1998 . "Reference to kinds across languages" . Natural
2005 , who assumes a functional projection c ( ' class ' ) directly above N in which
gender and number features are situated together; Picallo further assumes that
Language Semantics 6 :4. 339-405 .
Chomsky, Noam . 1995 . The minimalist program. Cambridge, Mas s . : The MIT
"gender declension is the morphological exponent of an abstract interpretable
Press.
feature in the functional layer of nominal structures" (Picallo 2005 : 1 07), this
feature being non-overt in Romance (cf. Picallo 2005 : 1 09) . Now , while ----------. 2000. ''Minimalist Inquiries : The framework". Step by Step. Essays
Heycock & Zamparelli 2003 concentrate on morphological number in relation on Minimalist Syntax in Honor of Howard Lasnik, ed. by Roger Martin,
to 'nominal classification ' , Picallo 2005 considers gender to be central here, David Michaels & Juan Uriagereka, 8 9- 1 5 1 . Cambridge, Mass : The MIT

assuming that number features always depend on the expression of gender Press .

(which i s true for Spanish and Catalan but much les s obvious for Italian or ----------. 200 1 . "Derivation b y phase". Ken Hale: A Life in Language, ed. by
French). By separating gender and number agreement processes in indefmite Michael Kenstowicz, 1 -52. Cambridge, Mas s : MIT Press.

nominals in Romance on the one hand, as I propose here in contrast to Ritter

----------. 2005 . On phases. Ms .
1993, and observing their interaction on the other, as also proposed here, the Delfitto, Denis & Jan Schroten. 1 99 1 . "B are plural s and the number affix in
derivation of both intranominal morphological agreement and indefinite DP" . Probus
3 :2. 1 55-185.
determination becomes obvious . If PI(*), which is responsible for morpho­ Gil, David.1987. "Definitenes s , noun phrase configurationality, and the count­
logical number agreement inside the nominal, hosts the gender information in mass distinction" . The representation of (in)definiteness, ed. by Eric J.

order to cause the uninterpretable gender feature in Num to be valued and Reuland & Alice G . B . ter Meulen, 254-269. Cambridge, Mas s . : MIT Press.

deleted, there will be an overt element for thi s for either case (the indefinite Giv6n, Talmy. 1981. "On the development of one as an i ndefinite marker".

arti cle or the so-called 'partitive article' ) . This is done via the number probe or Folia Linguistica Historica IT. 35-5 3 .
the number feature valued in Num, thus triggering the agree proces s . If Pl (*) Harley, Heidi & Elizabeth Ritter. 1 999. Meaning in Morphology: Motivating a

only hosts a number feature, with gender and number features being present on feature-geometric analysis of person and number. M S , University of
the N as in Spanish, there will be no overt element for the semantic feature Calgary & University of Pennsylvania.
[+LATT] in PI * for the singular. ----------. 2002a. "Person and number in pronouns : a feature-geometric
analysis". Language 78. 482-526.
27 Cf. e.g. Longobardi 200 1 a for definiteness and D. ----------. 2002b. "Structuring the bundle: a universal morphosyntactic feature
geometry." Pronouns - Grammar and Representation, ed. by Horst J.
82 ELISABETH STARK THE ROLE OF THE PLURAL SYSTEM IN ROMANCE 83

S imon & Heike Wiese, 23-39. Amsterdam & Philadelphia: John ed. by Mark B altin & Chris Collins, 5 62-603 . Oxford & Cambridge, Mass. :
B enj amins. Blackwell.
Harris, James W. 1 992. "The form classes of Spanish substantives". Yearbook ----------. 200 l b : "How comparative is semantics? A unified theory of bare
of Morphology 1 991, ed. by Geert Booij & Jaap van Marle, 65- 8 8 . nouns and proper names" . Natural Language Semantics 9:4. 335-369.
Dordrecht: Kluwer. Meisterfeld, Reinhard. 1 998. Numerus und Nominalaspekt. Eine Studie zur
Haspelmath, Martin. 1 999. "Optimality and diachronic adaptation". Zeitschrift romanischen Apprehension. [Link]: Niemeyer.
filr Sprachwissensclzaft 1 8 : 2. 1 8 0-20 5 . Mensching, Guido. 2005 . "Sonden und Phasen in romanischen Sprachen" .
Heycock, Caroline & Roberto Zamparelli . 2 003 . Friends and colleagues: Deutsche Romanistik - generativ, ed. by Georg Kaiser, 1 23- 1 4 3 . Tiibingen:
Plurality, coordination, and the structure of DP. Ms, University of Narr.
Edinburgh & Universita di B ergamo. ---------- & Eva Remberger. 2006. "Probes: lack of agreement in romance".
Hofmann, Johann B aptist. 1 997 (2nd reprint of the first edition of 1 9651 1 972). Studies on Agreement, ed. by Joao Costa & Maria Cristina Figueiredo
Lateinisclze Syntax und Stilistik: mit dem allgemeinen Teil der lateinischen Silva, 1 73-20 1 . Amsterdam & Philadelphia: Benj amins.
Grammatik. Neubearbeitet von Anton Szantyr. [Link]: Beck. Morani, Moreno. 2000. 1ntroduzione alia linguistica latina. Miinchen: Lincom.
lturrioz Leza, Jose Luis. 1 986. "lndividuation and determination Ill : The Munn, Alan & Cristina Schmitt. 2005 . "Number and indefinites". Lingua 1 1 5 .
concept of verbal plurality and the pluralization of abstractives". Funci6n 8 2 1 -85 5 .
1 : 2 . 20 1 -308. Pesetsky, David & Esther Torrego. 2004. The Syntax of Valuation and the
Korzen, l!l)rn. 1 9 96 L 'articolo italiano fra concetto ed entitii. Uno studio
.
Interp retability of Features. Unpublished M S . Boston: MIT.
semantico-sintattico sugli articoli e sui sintagmi nominali italiani con e Picallo, Carme. 1 99 1 . "Nominals and nominalizations in Catalan". Probus 3 .
senza determinante - con un 'indagine particolare sulla distribuzione del 279-3 1 6.
cosidetto 'articolo partitivo '. 2 vols. Kopenhagen: Museum Tusculanum ----------. 2002 . . "Abstract agreement and clausal arguments" . Syntax 5 :2. 1 1 6-

Press. 1 47 .
3 ----------. 2005 . "Some notes on grammatical gender and [-pronouns".
[Link], Raphael & Carl Stegmann. 1 95 5 . Ausfiilzrliche Grammatik der
lateinisclzen Sprache. Satzlehre. Erster Teil. Leverkusen: Gottschalksche Specificity and the evolution/emergence of nominal determination systems
Verlagsbuchhandlung. in Romance. Selected papers from the international workshop NEREUS li,
Langacker, Ronald W. 1 977. "Syntactic Reanalysis". Mechanisms of Syntactic October 2004 in Berlin (= Konstanzer Arbeitspapiere zur Sprachwissen­
Change (Austin), ed. by Charles N. Li. 57- 1 39 . S anta Barbara: University schaft, 1 1 9), ed. by Klaus von Heusinger, Georg Kaiser & Elisabeth S tark,
of California. 1 07- 1 2 1 . Konstanz: Universitlit Konstanz.
Lehmann, Christian. 1 99 1 . "The Latin nominal group in a typological Radford, Andrew. 2004. Minimalist Syntax. Exploring the structure of English.
1h Cambridge: CUP.
perspective". New studies in Latin linguistics. Selected papers from the 4
international colloquium on Latin linguistics, Cambridge, April 1 987, ed. Ritter, Elizabeth. 1 993. "Where ' s Gender?". Linguistic Inqui1y 82. 1 46- 1 50.
by Robert Coleman, 203-23 2. Amsterdam & Philadelphia: John Benj amins. ----------. 1 995. "On the syntactic category of pronouns and agreement".
Link, Godehard. 1983. "The logical analysis of plurals and mass terms" . Natural Language and Linguistic Theory 13 . 405-443.
Meaning, use and interpretation of language, ed. b y Rainer B liuerle, Roberts, Ian & Anna Roussou. 2003. Syntactic Change. A Minimalist
Christoph Schwarze & Arnim von Stechow, 302-323 . Berlin: de Gruyter. Approach to Grammaticalization. Cambridge: Cambridge University Press.
----- ----. 1 99 1 . "Plural". Semantik. Ein internationales Handbuch der
-
Schon, llse. 1 97 1 . Neutrum und Kollektivum. Das Morphem -a im Lateinischen
zeitgenossischen Forschung, ed. by Arnim von Stechow & Dieter und Romanischen. Innsbruck: Institut fiir Vergleichende Sprach­
Wunderlich, 4 1 8 -440. Berlin & New York: Mouton de Gruyter. wissenschaft der Universitat Innsbruck.
Lobel, Elisabeth. 1 993. "On the parametrization of lexical properties". The Schroten, Jan. 200 1 . "L' absence de determinant en espagnol". Typologie des
parametrization of universal grammar, ed. by Gisbert Fanselow, 1 8 3- 1 99. groupes nominaux, ed. by Georges Kleiber, Brenda Laca & Liliane
Amsterdam & Philadelphia: John Benj amins. Tasmowski, 1 89-203 . Rennes : Presses Universitaires de Rennes.
Longobardi, Giuseppe . 200 l a. 'The structure of DPs: some principles, Seiler, Hansjakob. 1 986. Apprehension. Language, object, and order. Part Ill:
parameters and problems". The handbook of contemporary syntactic theory, The universal dimension of apprehension . Tiibingen: Narr.
84 ELISABETH STARK

Stark, Elisabeth. 2002. "Indefiniteness and specificity in old Italian texts".

Jou rnal of Semantics 1 9 . 3 1 5-332.
----------. 2005 . "Explaining article grammaticalization in Old Italian".
Romanistische Korpuslinguistik 11. Korpora und diachrone
Sprachwissenschaft, ed. by Claus D. Pusch, Johannes Kabatek & Wolfgang
Th e paper used in this publication meets the minimum requirements of American National
Raible, 455-468 . Tiibingen: Narr.
----------. 2006. lndefinitheit und Textkohiirenz. Entstehung und semantische §™ Standard for Information Sciences - Permanence of Paper for Printed Library Materials,
ANSI Z39·48-1984.
Strukturierung indefiniter Nomina/determination im A ltitalienischen
( Beihefte zur Zeitschrift fiir romanische Philologie, 336), Tiibingen:
=
·

Niemeyer.
----------. 2007 . "Gender, number, and indefinite articles - about the
'typological inconsistency' of Italian" . Evolution and .function of nominal
determination (= Studies in Language and Linguistic Typology), ed. by
Elisabeth Stark, Wemer Abraham & Elisabeth Leiss, 49-7 1 . Amsterdam &
Philadelphia: John Benjarnins.
�c---�-�--. Forthcoming. "Typological conelations in nQminal determination in

Romance". Proceedings of the Copenhagen Determination Symposium

(August 2004), ed. by Alex Klinge et � - . Amsterdam & Philadelphia:
Benj amins.
Van Kemenade, Ans & Nigel Vincent. 1 997 . Parameters of morphosyntactic
change. Cambridge: Cambridge University Press.
Vincent, Nigel. 1 997. "The emergence of the D-system in Romance". Van
Kemenade & Vincent 1 997. 1 49- 1 69 .
Windisch, Rudolf. 1 9 7 3 . Genusprobleme i m Romanischen. Das Neutrum im
Rumiinischen. Tiibingen: Narr.
Zamparelli, Roberto. 2004. "Manufacturing Plurality" . Talk given at the
Copenhagen Determination Symposium, 26.-28 . August 2004.
----------. 2005 . "The structure of (in)definiteness" . Lingua 1 1 5 . 9 1 5-936. Library of Congress Cataloging-in-Publication Data

The paradox of grammatical change : perspectives from romance I edited by Ulrich Detges, Richard
Waltereit.
p. cm. -- (Amsterdam studies in the theory and history of linguistic science. Series rv, Current issues
in linguistic theory, ISSN 0304-0763 ; v. 293)
Includes bibliographical references and index.
1. Linguistic change. 2. Grammar, Comparative and general. 3. Romance languages-- Grammar, Histori­
cal. I. Detges, Ulrich. II. Waltereit, Richard.
P142.P365 2008
440' .045--dc22 2007044470
ISBN 978 90 272 4808 4 (Hb; alk. paper)

© 2008 - John Benjamins B. V.

No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means,
without written permission from the publisher.

John Benjamins Publishing Co. • [Link] 36224 • 1020 ME Amsterdam • The Netherlands
John Benjamins North America • [Link] 27519 • Philadelphia PA 19118-0519 • USA