draft of November 2011
draft of November 2011
draft of November 2011
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
Topic-strategies and the internal structure <strong>of</strong> nominal<br />
arguments in Greek and Italian ∗<br />
Theodora Alexopoulou and Raffaella Folli<br />
Cambridge and Ulster<br />
ta259@cam.ac.uk and r.folli@ulster.ac.uk<br />
Draft <strong>of</strong> <strong>November</strong> <strong>2011</strong><br />
Abstract<br />
The paper investigates investigates how Italian and Greek employ Clitic Left Disloca-<br />
tion (clld) to encode discourse topics. Greek is sensitive to the definiteness/referentiality<br />
<strong>of</strong> the topic, employing clld exclusively for referential topics and resorting to Topi-<br />
calisation for non-referential/property denoting topics. By contrast, clld is the main<br />
topic-strategy in Italian including non-referential/property denoting topics. This contrast<br />
is shown to mirror variation in more general patterns <strong>of</strong> anaphoric relations in the two<br />
languages, which, in turn, relate to the structure <strong>of</strong> nominal arguments: the fact that<br />
Greek non-referential indefinites systematically involve bare nouns which may be dropped<br />
rather than picked up by a pronoun in intrasentential anaphora contexts; the fact that<br />
Italian indefinites always involve a determiner and Italian pronouns can support property<br />
anaphora.<br />
The hypothesis is that the crosslignuistic variation in the anaphoric and clld patterns<br />
is the consequence <strong>of</strong> structural variation in the syntax <strong>of</strong> nominal arguments. Greek<br />
∗ We would like to thank the audiences <strong>of</strong> the following workshops and conferences: Funny Indefinites, June<br />
2008, Berlin, LAGB, September 2009, Edinburgh, Bare Nouns, <strong>November</strong> 2009, Paris, Syntax Workshop <strong>of</strong> the<br />
Irish Network on Formal Syntax, February 2010, Belfast, Meditereanean Syntax, October 2010, Athens, On<br />
Linguistics Interfaces, December 2010, Belfast and Workshop on Current Issues in Semantics and Pragmatics,<br />
June <strong>2011</strong>, York. Special thanks to George Tsoulas for discussion.<br />
1
nominal arguments lack a D-layer; rather they are Number Phrases. On a par with Italian,<br />
Greek nouns are predicative; however, while the nominalisor in Italian is D, in Greek it<br />
is Number. The hypothesis allows for a three-way distinction between Italian, Greek and<br />
English which explains the properties <strong>of</strong> bare nouns in the these languages and further<br />
patterns relating to the availability <strong>of</strong> Indefinite Argument Drop and Subnominal Deletion.<br />
It further suggests that D and referentiality are not intrinsically linked. Finally, under this<br />
analysis, variation in the realisation <strong>of</strong> topic-structures is not confined to PF (i.e. whether<br />
the in-situ element is a gap/null epithet or a pronoun); rather the specific properties <strong>of</strong><br />
the pronominal are crucial, in turn interacting with the syntax <strong>of</strong> nominal arguments in<br />
each language.<br />
2
1 Introduction<br />
Clitic Left Dislocation (clld) is standardly assumed as the main topic-strategy in Italian and<br />
Greek (Cinque 1990; Anagnostopoulou 1994; Rizzi 1997; Tsimpli 1995). On a par with English<br />
Topicalisation, the (A ′ ) topic-operator involved in clld is analysed as anaphoric in the sense<br />
<strong>of</strong> Lasnik and Stowell (1991); rather than binding a variable (like a quantificational operator),<br />
is linked anaphorically to the in-situ element <strong>of</strong> the dependency, through co-reference. The<br />
pronominal in clld is <strong>of</strong>ten analysed as an overt counterpart <strong>of</strong> the gap element in Topical-<br />
isation (Rizzi 1997; Tsimpli 1999). The crosslinguistic variation then is confined to PF, the<br />
choice between a pronominal element and a gap (null epithet for Lasnik and Stowell 1991).<br />
Under this view, crosslinguistic variation is primarily related to the nature <strong>of</strong> the operator,<br />
anaphoric vs. quantificational while variation in the nature <strong>of</strong> the in-situ element has been<br />
much less in focus. Current assumptions make the prediction that the range <strong>of</strong> intepretations<br />
available for Topicalisation should be available for clld structures; in addition, no variation<br />
is expected within clld since, ultimately, the involvement <strong>of</strong> the pronominal is taken as a PF<br />
realisation <strong>of</strong> the gap involved in Topicalisation. 1<br />
Indeed, clld-ed indefinites in Italian allow both the de re (wide scope) reading in (1a) as<br />
well as a de dicto reading, as indicated by the continuations in (1b) and (1c). If clld involves<br />
movement just like Topicalisation, then the indefinite in (1a) should reconstruct under the<br />
scope <strong>of</strong> the intensional predicate.<br />
(1) a. una gonna rossa la cerco da un po’<br />
a red skirt her.cl look-for-1sg for a while<br />
A red skirt I’ve been looking for a while...<br />
b. ma non ne ho trovata nessuna che mi piaccia<br />
but not <strong>of</strong>-them.cl have-1sg found none-fem that me please-3sg.subj<br />
... but have not found anyone that I like.<br />
c. ma non riesco a ricordarmi dove l’ho<br />
messa<br />
but not reach-1sg to remember where her.cl-have-1sg put<br />
puffle ... but I cannot remember where I’ve put it.<br />
1 To be precise, Lasnik and Stowell (1991) argue that the gap element in Topicalisation is a null epithet rather<br />
than a variable. The fact remains that the pronominals in clld should allow the same range <strong>of</strong> interpretations<br />
as the null epithet in Topicalisation.<br />
3
Indeed, this is the analysis proposed by Cecchetto (2001) who discusses cases <strong>of</strong> clld-ed<br />
indefinites interpreted in the scope <strong>of</strong> IP internal quantifiers. For instance, (2) is ambiguous<br />
between a wide scope reading for the indefinite and a wide scope reading for the universal.<br />
(2) a. un articolo di Chomsky ogni studente l’ha letto<br />
an article <strong>of</strong> Chomsky every student it.cl-has read<br />
An article <strong>of</strong> Chomsky, every student has read it.<br />
However, there are two unexpected facts for the view that clld and Topicalisation are<br />
different (PF) realisations <strong>of</strong> the same structure. The first is that Italian clld is rather<br />
exceptional in allowing ambiguous readings for clld-ed indefinites. For instance, in Greek,<br />
clld-ed indefinites can only admit wide-scope interpretations (Iatridou 1995; Alexopoulou<br />
and Kolliakou 2002) as indicated in (3). Mia kokini fusta resists the opaque or de dicto reading,<br />
as indicated by the infelicity <strong>of</strong> the continuation in (3b) (Iatridou 1995; Anagnostopoulou 1994;<br />
Alexopoulou and Kolliakou 2002). 2 Similarly, ena artrho tu Chomsky cannot be interpreted<br />
under the scope <strong>of</strong> the universal in (4).<br />
(3) a. mia kokini fusta tin psahno edho ke meres<br />
a red skirt it look-for-1sg here and days<br />
I’ve been looking for a red skirt for a few days ...<br />
b. =ke de boro na vro kamia pu na m’aresi<br />
and not can-1sg subj find-1sg none<br />
... and I cannot find any that I like.<br />
that subj me-please-3sg<br />
c. ke de boro na thimitho pu tin eho vali<br />
and not can-1sg subj remember-1sg where her.cl have-1sg put<br />
... and cannot remember where I put it. (Alexopoulou and Kolliakou 2002)<br />
(4) ena artrho tu Chomsky to diavase kathe fititis<br />
an article the-gen Chomsky it-cl read-3sg each student-nom<br />
There’s an article <strong>of</strong> Chomsky that every student read (only wide scope for indefinite).<br />
Second, Greek employs Topicalisation when a non-referential or de dicto interpretation <strong>of</strong> the<br />
indefinite topic is intended.<br />
2 We mark infelicity with =.<br />
4
(5) a. Fetos i moda ine apesia; idika i bluzes ine aparadektes<br />
I hate this year’s fashion; the blouses are especially outrageous.<br />
b. mia kokini bluza psahno edo ki ena mina ke de boro na vro<br />
a red blouse her.cl look-for-1sg here and one month and not can subj<br />
puthena kamia pu na m’aresi<br />
find-1sg anywhere anyone that subj me like-3sg<br />
A red blouse I’ve been looking for for a month now and I cannot find one that I<br />
like.<br />
(Alexopoulou and Kolliakou 2002, ex.51)<br />
So, while the hypothesis that clld and Topicalisation are different PF realisations <strong>of</strong> the<br />
same structure can elegantly capture the variation between English and Italian, the Greek<br />
facts are problematic, since this language has two distinct syntactic structures for topics,<br />
namely Topicalisation and clld, a fact that indicates that the presence <strong>of</strong> the pronominal<br />
has interpretative consequences. We turn to the discussion <strong>of</strong> the differences between these<br />
two languages next.<br />
The minimal pairs in Italian (1) and Greek (3)&(5) illustrate a contrast between Italian<br />
and Greek topic-strategies. Italian employs clld as the main topic-strategy, regardless <strong>of</strong> the<br />
referentiality <strong>of</strong> the dislocated topic. Greek, by contrast, employs clld only for referential top-<br />
ics; non-referential topics undergo Topicalisation. The present paper investigates the source<br />
<strong>of</strong> this contrast. Our analysis builds on the insight <strong>of</strong> Rizzi (1997),Cinque (1990), Anagnos-<br />
topoulou (1994) and Tsimpli (1999) according to which clld is characterised by the absence<br />
<strong>of</strong> quantifier-variable chain and the presennce <strong>of</strong> an anaphoric link between the dislocated<br />
phrase and the pronominal. Indeed we demonstrate that clld mirrors general anaphoric<br />
possibilities between pronouns and their antecedents in the two languages. We show that<br />
the contrast regarding clld is not an isolated fact, but rather, is linked to a further set <strong>of</strong><br />
empirical contrasts regarding the availability <strong>of</strong> Indefinite Argument Drop, bare subnominal<br />
ellipsis and the availability <strong>of</strong> bare nouns in the two languages. The emerging patterns indi-<br />
cate strongly that variation is these contexts is linked to variation in the internal structure <strong>of</strong><br />
the nominals and pronouns involved. Our central claim is that all relevant contrasts can be<br />
reduced to one: the absence <strong>of</strong> a D-layer from Greek nominal arguments and the hypothesis<br />
5
that Greek arguments instantiate Number Phrases. This hypothesis has consequences for the<br />
structure and interpretation <strong>of</strong> pronominals, which, in turn, directly affects the interpretative<br />
possibilities in clld in the two languages. Crucially, the clitic in clld cannot be just an overt<br />
PF realisation <strong>of</strong> the epithet assumed in Topicalisation, since its specific properties interact<br />
direclty with the topic-strategies in the two languages. Variation in topic-structures then can-<br />
not be confined to PF, but rather, interacts with the structure <strong>of</strong> nominal arguments in the<br />
two languages. 3 Further, the hypothesis allows a three way typological distinction between<br />
English, Italian and Greek in relation to bare nouns. Unlike English, Italian and Greek do not<br />
have argumental nouns (in the sense <strong>of</strong> Chierchia 1998). Rather nouns are predicative. But<br />
while in Italian D acts as the nominalisor turning predicative nouns to arguments, in Greek<br />
the nominalisor is Number.<br />
In section 2 we briefly review previous approaches to the interpretations <strong>of</strong> clld-ed in-<br />
definites. Section 3 presents a set <strong>of</strong> facts that correlate with the clld contrasts, drawing<br />
from contexts <strong>of</strong> Indefinite Argument Drop, Bare Subnomianal Ellipsis and the distribution<br />
<strong>of</strong> bare nominals in the two languages. Our analysis <strong>of</strong> these facts is presented in section 4.<br />
2 Indefinite Topics: Previous Analyses<br />
Though no previous analysis has focused on the crosslinguistic variation between Greek and<br />
Italian clld, the main facts presented earlier have been discussed for each language. Two<br />
kinds <strong>of</strong> analyses have been proposed to explain the wide scope readings <strong>of</strong> Greek clld-ed<br />
indefinites and the ambiguity <strong>of</strong> their Italian counterparts: (i) scope based analyses, where the<br />
relevant readings are linked to the underlying derivations and (ii) Topichood-based analyses,<br />
where referential readings are viewed as a consequence <strong>of</strong> the discourse function <strong>of</strong> clld-ed<br />
phrases as topics. We review these analyses briefly below.<br />
Scope based analyses Iatridou (1995) and Anagnostopoulou (1994) take the impossibility<br />
<strong>of</strong> interpreting clld-ed indefinites within the scope <strong>of</strong> IP internal operators as evidence for a<br />
base-generation analysis <strong>of</strong> Greek clld. Similarly, Cecchetto (2001) takes the Italian facts as<br />
3 A preliminary discussion and analysis <strong>of</strong> this set <strong>of</strong> facts is presented in Alexopoulou and Folli (<strong>2011</strong>).<br />
6
evidence <strong>of</strong> reconstruction, and, therefore proposes a movement analysis for Italian clld-ed<br />
DPs. Italian and Greek clld then involve distinct derivations, movement and base generation<br />
respectively. However, there is no independent evidence for this derivational contrast. In<br />
fact, in both languages the structures display many standard clld properties (no wco, no-<br />
parasitic gaps, sensitivity to islands, unavailability <strong>of</strong> clld-ed downward entailing quantifiers<br />
see Cinque 1990; Anagnostopoulou 1994; Tsimpli 1995; Rizzi 1997; Alexopoulou, Doron, and<br />
Heycock 2004). 4<br />
In addition, the ”scope” facts are rather complex, posing a problem for the derivational<br />
approach. Consider (6a); while the de dicto reading is excluded, this example allows the so-<br />
called ”3rd-reading” (Fodor 1970; von Fintel and Heim 2009), according to which, I’m looking<br />
for a specific type <strong>of</strong> an Armani skirt, e.g. one that I’ve seen in a brochure, but still not for<br />
a token. 5 Such examples could be taken as evidence that clld may at least allow partial<br />
reconstruction.<br />
(6) a. mia fusta tu Armani tin psahno edho ke meres<br />
a skirt the-gen Armani her.cl look-for-1sg here and days<br />
An Armani skirt I’ve been looking for for a few days ....<br />
b. ke de mporo na vro kamia pu na mu kani<br />
and not can-1sg subj find-1sg none that subj me fit-3sg<br />
... and I cannot find anyone that fits me.<br />
In addition, there is evidence from both Greek and Italian, that the properties <strong>of</strong> the dislocated<br />
element affect the ”scope” possibilities. For example, the plural clld-ed interrogative in (7)<br />
allows only a wide scope reading. 6<br />
(7) ?quanti pazienti ritieni che li debba visitare ogni medico?<br />
how-many patients think-2pl that them should visit each doctor<br />
How many patients do you think each doctor should visit? (no wide scope for universal)<br />
Unlike clld-ed indefinites, Greek definites may be interpreted under the scope <strong>of</strong> the universal<br />
4 But see Haegeman (pear) for some differences.<br />
5 Thanks to S. Iatridou for this observation.<br />
6 This example is due to Longobardi (1986) cited by Cinque (1990).<br />
7
quantifier in (8).<br />
(8) to aftokinito tu to asfalise kathe fititis<br />
the car his-gen it insured-3sg each student-nom<br />
Each student insured his car (distributive reading available).<br />
Finally, clld-ed PPs don’t reconstruct in Italian (Cecchetto 2001).<br />
In sum, there are at least three factors that may affect reconstruction: the number <strong>of</strong> the<br />
dislocated element (7), its definiteness (8) and its category. 7 Under the scope approach, these<br />
interactions are mysterious.<br />
Topichood analyses Topics are <strong>of</strong>ten taken to necessarily involve well established dis-<br />
course antecedents and, as a result, be compatible only with referential interpretations (Rein-<br />
hart 1982; Alexopoulou and Kolliakou 2002; Endriss 2006). This explanation works for Greek<br />
clld, since clld-ed phrases are necessarily referential topics (Philippaki-Warburton 1985;<br />
Anagnostopoulou 1994; Tsimpli 1995; Rizzi 1997). However, Italian clld challenges the gen-<br />
eralisation that topics are exclusively referential as we have seen in (1); the Greek Topicali-<br />
sation examples in (5b) are equally problematic.<br />
3 Beyond topics: nominals and anaphoric relations in Italian<br />
and Greek<br />
In this section we show that the clld facts mirror more general anaphoric possibilities in the<br />
two languages, which in turn relate to the make up <strong>of</strong> the nominals in the two languages.<br />
In particular, we draw evidence from the availability <strong>of</strong> Indefinite Argument Drop, Bare<br />
subnominal deletion and bare nominals. The most important shortcoming <strong>of</strong> the scope-based<br />
and topichood analyses reviewed in the previous section is that they could not capture the<br />
more general patterns discussed below. We begin by looking at the basic contrast between<br />
clld and Greek Topicalisation in some more detail.<br />
7 The presence <strong>of</strong> modal operators further affects readings (Alexopoulou 2009).<br />
8
3.1 Greek manipulates structure while Italian manipulates Ds<br />
Let us focus on the non-referential indefinites in some more detail. Consider the minimal pair<br />
in (9).<br />
(9) a. Una segrataria *(la)-trovi facilmente/ prima o poi una segretaria *(la)-trovi<br />
A secretary her.cl-find easily/ sooner or later a secretary her.find<br />
A secretary, you fill find her easily/ sooner or later you will find her<br />
b. Gramatea tha (*ti) vrite sigura<br />
Secretary will her.cl find-2pl certainly<br />
A secretary you will find her certainly.<br />
As already seen, Greek employs Topicalisation when a non-referential interpretation is in-<br />
tended while Italian clld. Thus, the clitic ti is ungrammatical in Greek but la obligatory<br />
in Italian. However, (9b) points to a further difference between the two languages: the dislo-<br />
cated phrase in Greek is a bare noun while in Italian the indefinite article is indespensible.<br />
Example (10), where the bare noun is preceded by mia is possible in Greek, but there is a<br />
strong preference to read mia as one. 8<br />
(10) mia grammatea tha vrite sigura<br />
one-fem secretary will find-2pl certainly<br />
It seems then that Greek is sensitive to the (in)definiteness <strong>of</strong> the topic and has two<br />
distinct topic strategies: indefinite topics are typically bare nominals linked to a gap, while<br />
definite/referential topics are preceded by an article and linked to a pronominal clitic (11).<br />
By contrast, Italian indefinite topics as in (9a) are structurally very similar to the clld-ed<br />
definite topic in (11b); both types involve clld.<br />
(11) a. ti Maria tha ti vrite sigura sto grafio<br />
the-acc Maria will her.cl find-2pl certainly at-the <strong>of</strong>fice<br />
Maria you’ll certainly find her at the <strong>of</strong>fice.<br />
b. il libro che cerchi lo trovi sul tavolo<br />
the book that are-looking-for.2sg it.cl find-2sg on-the table<br />
8 Greek has no indefinite article distinct from the numeral one.<br />
9
The book you are looking for you’ll find it on the table.<br />
The pattern becomes somewhat more complicated in Italian when plurals are considered.<br />
As can be seen in (12a) and in (7) repeated as (12b), the plural clitic le resists a non-referential<br />
antecedent, just like the Greek clitics.<br />
(12) a. *Segretarie le<br />
trovi facilemente.<br />
Secretaries them.cl.fem find easily<br />
Secretaries, you will find them easily.<br />
b. ?quanti pazienti ritieni che li<br />
debba visitare ogni medico?<br />
how-many patients think-2sg that them.cl.msc should visit each doctor<br />
How many patients do you think each doctor should visit? (no wide scope for<br />
universal)<br />
However, unlike Greek, Italian does not resort to an alternative structure (Topicalisation)<br />
to realise plural indefinite topics. Rather it employs an alternative clitic, namely ne as shown<br />
in (13). Importantly, while a bare plural is possible in these structures, a DP involving the<br />
bare partititive construction delle N is also available.<br />
(13) a. (Delle)-segretarie ne trovi facilemente.<br />
Secretaries <strong>of</strong>-them.cl find.easily<br />
Secretaries, you will find them easily<br />
b. (Delle)-segretarie se *(ne) trovano facilmente<br />
Secretaries refl <strong>of</strong>-them.cl find easily<br />
Secretaries, can be found easily.<br />
In sum then, Greek non-referential/indefinite topics are structurally different from their<br />
definite/referential counterparts; they involve bare nominals linked to a gap. By contrast,<br />
Italian systematically employs clld as its topic strategy but manipulates the type <strong>of</strong> D-<br />
elements (pronominals/articles) involved.<br />
10
3.2 Indefinite pronominal antecedents and Indefinite Argument Drop<br />
(IAD)<br />
The range <strong>of</strong> available interpretations for an indefinite clld-ed phrase in the two languages<br />
mirrors the possibilities available in intrasentential anaphora. Compare (9a) with (14). The<br />
pronoun la in (14a) and lo in (14b) can admit a non-referential antecedent just like the clld<br />
example in (9a). Notice again that la and lo cannot be ommitted in (14).<br />
(14) a. A:Maria ha trovato una baby sitter? B:Si, l’ha<br />
trovata<br />
A:Maria has-3sg found a baby sitter? B:Yes, her.cl’has-3sg found<br />
A:Has Maria found a baby sitter? B: yes, she found one<br />
b. Gianni sta cercando un idraulico ma non lo trova<br />
Gianni is looking-for a plumber but not him.cl find-3sg<br />
Gianni is looking-for a plumber but cannot find one.<br />
Compare now the Greek Topicalisation example in (9b) with (15). The pronouns ti and<br />
ton are incompatible with the indefinite bare noun antecedents, just as in (9b). In fact, there<br />
is no pronoun in these examples; rather the (object) argument is dropped; the phenomenon<br />
was first observed by Dimitriadis (1994) who also drew a correlation between the availability<br />
<strong>of</strong> Greek Topicalisation and argument drop in examples like (15). 9 (See also Giannakidou and<br />
Merchant (1997), Tsimpli and Papadopoulou (2005) and Panagiotidis (2002)):<br />
(15) a. A:vrike dada I Maria? B:ne, (*ti) vrike<br />
Q:found-3sg nanny the-nom Maria A:yes, (*her) found-3sg<br />
Has Maria found a nanny? Yes, she found.<br />
b. o Yanis psahni idravliko alla dhe (*ton) vriski puthena<br />
the-nom Yanis look-for-3sg plumber but not (him) find-3sg anywhere<br />
Yanis is looking for a plumber but cannot nd one anywhere.<br />
Note that the problem in (15) is not the bare nominal per se, but the necessarily non-<br />
referential interpretation <strong>of</strong> bare nouns which bars them as pronominal antecedents. As we<br />
shall see in the next section, bare nouns can be discourse transparent, i.e. act as pronoun<br />
antecedents. But in (15) the interpretations are necessarily non-referential/weak. When enan<br />
9 He refers to what we call Topicalisation as Empty clld.<br />
11
(=a/one) is involved in (16), ambiguity arises. The presence <strong>of</strong> the pronominal in (16a) forces<br />
a referential interpretation for the indefinite, exactly as in the clld case in (3). The indefinite<br />
interpretation is only available when the object is dropped in the second clause (16b), on a<br />
par with (5).<br />
(16) a. o Yanis psahni enan idravliko, ala dhe ton vriski puthena<br />
the-nom Yanis look-for-3sg one plumber but not him find-3sg anywhere<br />
Yanis is looking for a plubmer but cannot find him anywhere. ∃ > look-for<br />
b. o Yanis psahni enan idravliko, ala dhe vriski puthena<br />
the-nom Yanis look-for-3sg one plumber but not find-3sg anywhere<br />
Yanis is looking for a plubmer but cannot find one anywhere. look-for > ∃<br />
The facts reviewed so far indicate that: (i) the relation between a clld-ed phrase and the clitic<br />
mirrors intrasentential anaphora in the two languages; (ii) Greek pronouns systematically<br />
resist non-referential antecedents while Italian clitics can accept such antecedents; (iii) Greek<br />
allows Indefinite Argument Drop (IAD) where Italian systematically employs pronouns.<br />
Before moving on it is worth pointing out that IAD is available also with subjects in Greek<br />
as shown in (17) (and noticed by Giannakidou and Merchant 1997). (17b) can be a felicitous<br />
exchange in a context where a new helpline is set up at a university, available to faculty,<br />
students and the general public, but what is <strong>of</strong> interest is if students specifically use it.<br />
(17) a. A:irthe kanis? B:ne irthe<br />
A:came-3sg anyone? B:yes, came-3sg<br />
A:Did anyone come? B:Yes, someone did.<br />
b. A:tilefonisan fitites? B:ne tilefonisan<br />
A:phoned-3pl students? B:yes, phoned-3pl<br />
A:Did students phone? B:Yes, some did.<br />
By contrast, such indefinite subjects cannot be dropped in Italian.<br />
(18) A:Ha telefonato qualcuno B:Si, qualcuno ha telefonato/*Si, ha telefonato<br />
A:Has phoned someone? B:Yes, someone has phoned/*Yes, has phoned<br />
A: Has anyone phoned? B: Yes, someone has phoned.<br />
12
3.3 Bare nouns<br />
As has become evident, many <strong>of</strong> the critical examples in Greek involve bare nouns. In fact, bare<br />
nouns are very widely available in Greek (19), in contrast to Italian, where their distribution<br />
is restricted (Benincà 1980; Longobardi 1994; Chierchia 1998). Crucially for the present<br />
discussion, a bare noun is ungrammatical in examples like (20) where Greek naturally prefers<br />
a bare noun.<br />
(19) a. i Maria vrike dada gia ta pedhia<br />
the-nom Maria found nanny for the children<br />
Maria found a nanny for the children.<br />
b. mpikan kleftes ke sikosan ta pada<br />
broke-in-3pl burglars and lifted-3pl the all<br />
Burglars broke in and took everything.<br />
c. diadilotes pirpolisan magazia ke aftokinita stus dromus yiro apo<br />
demonstrators set-on-fire shops and cars in-the streets around from<br />
to Politehnio<br />
the Politehnio<br />
Demonstrators set on fire shops and cars in the streets around the School <strong>of</strong><br />
Engineering.<br />
The Italian counterparts <strong>of</strong> Greek (19a) and (15) resist bare nouns as shown in (20).<br />
(20) a. Maria ha trovato una baby-sitter/*baby-sitter per i bambini<br />
Maria has found a nanny/*nanny for the children<br />
Maria found a nanny for the children.<br />
b. Gianni sta cercando un idraulico/*idraulico.<br />
Gianni is looking-for a plumber/*plumber<br />
Gianni is looking for a plumber.<br />
3.4 Bare subnominal ellipsis<br />
The final set <strong>of</strong> observations comes from another context <strong>of</strong> intrasentential anaphora which<br />
gives rise to the deletion <strong>of</strong> the nominal antecedent. Consider (21). In the answer, the noun<br />
tavolo is elided; the nominal is headed by uno which consists <strong>of</strong> the indefinite article and the<br />
classifier o, followed by the adjective (Alexiadou and Gengel 2008). In addition, the elliptical<br />
13
nominal is doubled by the pronoun lo, which in this case is non-referential. 10<br />
(21) A:Vorrei un tavolo grande B:Mi spiace. Non lo abbiamo, uno<br />
A:would-like-1sg a<br />
grande<br />
big<br />
table big B:Me displeases-3sg. Not it have-1pl, a<br />
I would like to buy a big table. I’m sorry. We do not have a big one.<br />
(From Alexiadou and Gengel 2008, attributed to V.Samek-Lodovici)<br />
Compare now with the Greek counterpart <strong>of</strong> (21) in (22). As in Italian, the noun can be<br />
elided, but the adjective alone is enough; there is no element heading the nominal and, in<br />
addition, there is no doubling. All we have is a bare adjective holding the place <strong>of</strong> a nominal<br />
object.<br />
(22) a. thelo afti ti fusta se kitrino<br />
want-1sg this the skirt in yellow<br />
I would like this skirt in yellow.<br />
b. Distihos dhen eho kitrini. (Mono mavres mu ehun<br />
unfortunately not<br />
mini)<br />
left)<br />
have-1sg yellow-fem.sg (Ony black-fem.pl me have-3pl<br />
Unfortunately I don’t have a yellow one. ( Only black ones are left).<br />
The facts presented in this section are summarised in Table 1. All critical cases involve<br />
non-referential or weak indefinites. The emerging pattern is that for such nominals, Italian<br />
10 Belletti and Rizzi (1981) argue on the basis <strong>of</strong> the examples below where ne is obligatory and uno un-<br />
grammatical, that ne involves less structure than DP. However, Cardinaletti and Giusti (1990) have argued<br />
against this position and analyse ne as a head Q, although in their paper they do not take an explicit position<br />
on whether Q is a functional category different from D or it is the head <strong>of</strong> D. In the discussion to come we will<br />
adopt this latter position.<br />
(i) a. ho letto un lunghissimo libro<br />
have-1sg read a very-long book<br />
I have read a very long book.<br />
b. ne/*uno ho letto uno/*un (lunghissimo)<br />
14
systematically employs articles and pronouns where Greek resorts to bare structures. The<br />
analysis then needs to capture the following: (i) Italian nominals are systematically DPs<br />
where Greek nominals involve less structure; (ii) Italian pronouns are compatible with non-<br />
referential antecedents where Greek pronouns necessarily involve referential antecedents.<br />
Structure Italian Greek<br />
(A) Indefinite Topics clld <strong>of</strong> an indefinite Bare nominal<br />
nominal (1) linked to a gap (5,9b)<br />
(B) CLLD-ed De dicto and de re Only de re (3)<br />
indefinites available (1) and 3rd reading (6)<br />
(C) Bare nouns Limited distribution; Productive use<br />
bare partitive <strong>of</strong> bare nouns (9b,20)<br />
construction (12,13)<br />
(D) Indefinite Unavailable Obligatory with<br />
Argument Drop (14,19) (some) weak indefinites<br />
and bare nouns (16,17b)<br />
(E) Bare Subnominal Unavailable May involve<br />
Ellipsis always a D element (22) bare adjective (23)<br />
Table 1: Summary <strong>of</strong> the crosslinguistic contrasts in the realisation <strong>of</strong> weak indefinites<br />
15
4 Analysis<br />
4.1 Greek nominal arguments are Number Phrases<br />
Since Horrocks and Stavrou (1987), Greek nominal arguments are assumed to be DPs (see<br />
also Stavrou 1991). This view remains dominant in the Greek literature to date, modulo<br />
Kolliakou’s work on Greek definites (Kolliakou 2003) and two proposals for treating some<br />
cases <strong>of</strong> Greek bare nouns as NPs by Tomioka (2003) and Tsimpli and Papadopoulou (2005).<br />
In this section we provide arguments that Greek nominals are Number Phrases without a<br />
D layer. We begin our discussion with bare nouns in section 4.1.1 and then move to nominals<br />
with articles and quantifiers in section 4.1.3.<br />
4.1.1 Bare nouns in Greek<br />
We hypothesise the structure in (23). We need to show two things: first, that bare nouns are<br />
arguments and not incorporated properties. Second, that no D layer is needed.<br />
(23) a. aghorasa vivlia<br />
bought-1sg books<br />
I bought books.<br />
VP<br />
✟❍<br />
✟<br />
✟ ❍<br />
❍<br />
b. V NumP<br />
aghorasa<br />
✟ ✟ ❍<br />
❍<br />
Num NP<br />
+Pl<br />
N<br />
vivlia<br />
Morphology In terms <strong>of</strong> their morphology, bare nominals are marked for case, gender<br />
and number just like any other argument, in contrast to incorporated arguments which,<br />
crosslinguistically, may show reduced morphology (Farkas and de Swarts 2003).<br />
16
Syntax They can be left dislocated (24a), undergo (focus-)movement (24b) or be pas-<br />
sivised (24c) like any other argument. 11<br />
(24) a. danio, kserume pia trapeza tha mas dosi<br />
loan, know-1pl which bank will us give-3sg<br />
A loan, we know which bank will give it to us.<br />
b. gamo theli, ohi tsilimpurdismata<br />
marriage want-3sg not affairs<br />
S/he’s after marriage, not affairs.<br />
c. plastes taftotites ekdothikan mono stin Katohi (ohi ston<br />
fake identity-cards were-issued only in-the Occupation (not in-the<br />
efmilio)<br />
civil-war)<br />
Fake identity cards were issued only during the Occupation period (not during<br />
the Civil war).<br />
In addition, bare nouns can be modified like their non-bare counterparts.<br />
(25) a. agorase akrivo aftokinito<br />
bought-3sg expensive car<br />
She bought an expensive car.<br />
b. theli dada me ptihio<br />
want-3sg baby-sitter with degree<br />
She wants a babysitter with a degree.<br />
c. kalos yatros ton exetase (min anisihis)<br />
good-nom doctor-nom clhim examined-3sg (not<br />
A good doctor examined him, don’t worry.<br />
worry-2sg)<br />
11 Panagiotidis (2003) points out that a predicate like perno tilefono (=take phone) is ambiguous between<br />
I get(=buy/fetch) a phone and make a phonecall. Interestingly, even under the latter interpretation, where<br />
tilefono could be taken as semantically incorporated to the meaning <strong>of</strong> the whole predicate, the bare noun can<br />
be dislocated as in (i).<br />
(i) tilefono de mpori na pari i marina; ine mikro pedhi<br />
phonecall not can subj take-3sg the-nom Marina; is small child<br />
Marina cannot make phonecall; she’s only a child.<br />
17
Thus, the morphosyntactic properties <strong>of</strong> bare nouns provide evidence for an argument analysis<br />
since there is no evidence <strong>of</strong> incorporation.<br />
No number neutrality for bare singulars Semantic evidence also indicates that bare<br />
nouns are arguments and excludes an incorporation analysis. A characteristic property <strong>of</strong><br />
incorporated bare nouns is number neutrality, even when they are marked with singular<br />
morphology. Number neutrality entails compatibility with both atomic (singular) and plural<br />
interpretations (Farkas and de Swarts 2003; Espinal 2010). Greek bare singulars are only<br />
compatible with an atomic interpretation. Thus, (26a) denotes reading <strong>of</strong> one newspaper;<br />
characteristically, (26c) is ungrammatical with the singular, exactly because the predicate<br />
necessitates a plural interpretation (compare with stamp collector in English).<br />
(26) a. dhiavase efimeridha<br />
read-3sg newspaper<br />
S/he read a newspaper. (reading <strong>of</strong> one newspaper)<br />
b. dhiavase efimeridhes<br />
read-3sg newspapers<br />
She read newspapers. (reading <strong>of</strong> more than one newspapers)<br />
c. mazevi *gramatosim-o/gramatosim-a<br />
gather-3sg stamp-sg/stamps-pl<br />
She collects stamps.<br />
Greek bare singulars cannot license plural interpretations in (27) and (28) (adapted from<br />
Espinal 2010, ex.4a). The second sentence in (27) is infelicitous; Greek contrasts in this re-<br />
spect with languages like Catalan, where bare nouns license plural interpretations in contexts<br />
like (27) (Espinal 2010).<br />
(27) psahno aftokinito; = ena mikro gia tin poli ki ena fortighaki ya ekdromes<br />
look-for-1sg car; = one small for the city and one van for trips<br />
I’m looking for a car. = a small one for the city and a van for trips.<br />
18
Further, Greek bare nouns have atomic interpretations in contexts like (28), where their<br />
Catalan counterparts are number neutral and compatible with plural readings. 12 - 13<br />
12 With focal stress on the verb, we can get the implicature <strong>of</strong> more flowers, accounts or houses in (28). But,<br />
this is just the fact that, any indefinite interpreted existentially is true even if more than one such entities<br />
exist. In other words, the Greek examples in (i) are no different from their English translations in this respect.<br />
13 Espinal (2010) further notes that bare singulars in Catalan are restricted to predicates that allow an<br />
interpretation where the predicate (verb+bare singular) denotes a characterising property <strong>of</strong> the subject. This<br />
assumption explains the contrast between (ia) and (ib). (ia) involves a characterising property <strong>of</strong> the external<br />
argument, that <strong>of</strong> car-owner, while (ib) does not. Building on Espinal and McNally (2007), Espinal (2010)<br />
assumes that only ”have”-predicates are compatible with these characterising interpretations. Example (ib),<br />
then, is bad because it cannot be analysed as a ”have” predicate.<br />
(i) a. Tengo choche<br />
have car<br />
I have a car. (It could be one or more than one; I am a car-owner).<br />
b. =Limpio choche<br />
clean car<br />
I’m cleaning a car.<br />
(From Espinal 2010, ex.18)<br />
Greek examples like (ia) can certainly be interpreted as providing a characterising property <strong>of</strong> the subject.<br />
Crucially, though, bare singulars can appear as objects <strong>of</strong> a wider range <strong>of</strong> predicates, that resist this interpre-<br />
tation (15) but can also appear as subjects themselves (25c). Further note that while (iib) is not felicitous in<br />
the minimal context <strong>of</strong> (iia), it can be felicitous in (iii); suppose that speaker B owns a company which takes<br />
on cleaning <strong>of</strong> public buildings such as schools, churches and gyms. Assuming shared knowledge between A<br />
and B that cleaning a church is the most difficult and time consuming <strong>of</strong> these jobs, B’s reply is natural, since<br />
”cleaning a church” is a predicate that is implicitly contrasted with ”cleaning a gym or school”.<br />
(ii) a. ti kanis?<br />
what doing-2pl<br />
What are you doing?<br />
b. =?katharizo aftokinito<br />
clean-1sg car<br />
I am cleaning a car.<br />
(iii) a. A:pu vriskete o Yanis<br />
A:where is-3sg<br />
Where is Yanis?<br />
the-nom Yanis-nom<br />
b. B:katharizi eklisia; katalavenis; tha ton dume se kamia bdomada pali<br />
B:clean-3sg church; understand-2sg; will him see-1pl in a week again<br />
19
(28) a. i amigdalia ebgale luludi<br />
the-nom almond-tree made-3sg flower<br />
The almond tree had a flower.<br />
b. eho logariasmo stin ethniki<br />
have-1sg account in-the national<br />
I have an account in the National Bank.<br />
c. eho spiti<br />
have-1sg house<br />
I have a house.<br />
Free adjectival modification Bare singulars in Catalan can only combine with classifying<br />
modifiers as in (29a) but resist qualitative and descriptive adjectives is in (29b) an (29c).<br />
Llarga, escocesa and de quadres denote a subtype <strong>of</strong> skirt while alta in (29c) can only modify<br />
individual entities. This contrast indicates that in Catalan bare singulars denote properties<br />
and cannot denote individuals.<br />
(29) a. Per a aquest espectecle necessitareu faldilla llarga/escocesa/ de quadres<br />
for to this event need-fut skirt long/kilt/plaid<br />
For this event you will need a long skirt/a kilt/ a plaid skirt.<br />
b. *Necessiten<br />
need<br />
faldilla<br />
skirt<br />
c. *Té parella alta/malalta<br />
has parner tall/ill<br />
feta a Singapur/neta<br />
made in Singapore/clean<br />
(From Espinal 2010, ex.8,9)<br />
Again, Greek bare singulars exhibit properties <strong>of</strong> arguments denoting individuals, as in-<br />
dicated by the availability <strong>of</strong> the descriptive and qualititative adjectives in (30b) and (30c).<br />
(30) a. tha hriastite makria/skotzesiki/plise fusta<br />
will need-2pl long/scotish/plaid skirt<br />
You will need a long skirt/a kilt/a plaid skirt.<br />
b. tha hriastite fusta rameni stin India/kathari fusta<br />
will need-2pl skirt sewn in-the India/clean skirt<br />
He’s cleaning a church. You know; we won’t see him for a week.<br />
20
You will need a skirt sewn in India/ a clean skirt.<br />
c. ehi arosto pedhi/ehi psilo gkomeno<br />
has ill child/has tall boyfriend<br />
S/he has an ill child/a tall boyfriend.<br />
Discourse transparency Finally, bare nouns can introduce discourse referents (31), that<br />
is, they can be discourse transparent (Farkas and de Swarts 2003), a fact which, as pointed out<br />
by Kolliakou (2003), indicates that D is not necessary in Greek for referential interpretations.<br />
(31) telika vrike dada; ti gnorisame htes sto parti tis<br />
finally found-3sg nanny; her.cl met-1pl<br />
Yotas<br />
Yota-gen<br />
yesterday at-the party the-gen<br />
He finally found a nanny; we met her yesterday at Yota’s party.<br />
The facts reviewed so far indicate that bare nouns in Greek are fully fledged arguments<br />
both in terms <strong>of</strong> their morphosyntax as well as their semantic interpretations. In addition,<br />
the contrasts with Catalan, clearly indicate that not only plurals, but singular bare nouns<br />
have Number. We have to assume then that these nominals are minimally Number Phrases<br />
(alternatively, they can be Noun Phrases necessarily marked for Number).<br />
The next question is whether we should assume a null D head above NumP. Our answer<br />
is negative. The first set <strong>of</strong> facts arguing against a null D comes from scope.<br />
Scopal inertia Greek bare nouns exhibit the scopal inertia standardly exhibited by their<br />
crosslinguistic counterparts (Chierchia 1998; Farkas and de Swarts 2003).<br />
Examples involving the indefinite article like (32b) and (33b) are ambiguous: the former<br />
allows a transparent or de re reading and an opaque or de dicto reading while (33b) additionally<br />
allows the 3rd reading according to which Maria is looking for a specific type <strong>of</strong> Armani skirt.<br />
On the other hand, the bare nouns in the (a) examples allow only the opaque readings.<br />
(32) a. i Maria theli na padrefti Italo<br />
the-nom Maria want-3sg subj marry-3sg Italian<br />
Maria wants to marry an Italian. (only opaque reading)<br />
21
. i Maria theli na padrefti enan<br />
the-nom Maria want-3sg subj marry-3sg one-acc<br />
Maria wants to marry an Italian. (ambiguous)<br />
(33) a. i Maria theli n’agorasi fusta tu<br />
the-nom Maria wants subj-buy-3sg skirt the-gen<br />
Maria wants to buy an Armani skirt. only opaque<br />
Italo<br />
Italian<br />
Armani<br />
Armani<br />
b. i Maria theli n’agorasi mia fusta tu<br />
the-nom Maria wants subj-buy-3sg one skirt the-gen<br />
Maria wants to buy an Armani skirt. 3-way ambiguous<br />
Armani<br />
Armani<br />
If the bare nouns involved a null D, we would expect exactly the same range <strong>of</strong> interpre-<br />
tations as in the case where the purported D element is overtly present since, the overt/null<br />
alternation should be a PF alternation and should not affect LF interpretations, as is the<br />
case in Italian. But (33) and (32) indicate that this is not so. This is then our first argument<br />
against postulating a null D for Greek bare nouns.<br />
The contrast between bare and non-bare nouns extends beyond the above intensional<br />
contexts to interactions with the universal quantifer (34) and with negation (35). Bare nomi-<br />
nals cannot take scope over the universal quantifier (34a) (see Farkas and de Swarts 2001 for<br />
similar facts in Hungarian) or over negation (35a) (Chierchia 1998).<br />
(34) a. kathe episkeptis diavase efimeridha/efimeridhes<br />
each visitor read-3sg newspaper-sg/newspapers-pl<br />
Each visitor read a newspaper/newspapers. Only ∀ > ∃<br />
b. kathe episkeptis diavase mia efimeridha/kapies efimeridhes<br />
each visitor read one newspaper-sg/some newspapers-pl<br />
Each visitor read a newspaper/some newspaper. ∀ > ∃ or ∃ > ∀<br />
(35) a. dhen idhe rogmes sto tavani<br />
not saw-3sg cracks in-the ceiling<br />
S/he didn’t see cracks in the ceiling. Only ¬ > ∃<br />
b. den idhe mia lakuva sto dromo<br />
not saw-3sg a hole in-the street<br />
S/he didn’t see a hole in the street. ∃ > ¬ or ?¬ > ∃<br />
c. den idhe lakuva sto dromo<br />
not saw-3sg hole in-the street<br />
S/he didn’t see a hole in the street. Only ¬ > ∃<br />
22
4.1.2 Towards a crosslinguistic typology <strong>of</strong> bare nouns<br />
In his seminal paper on kinds, Chierchia (1998) proposes a semantic parameter according to<br />
which languages vary in the way their nouns may be allowed to function as arguments or<br />
not. Accordingly, there is variation across languages in the denotation <strong>of</strong> nominal categories:<br />
nouns may be classified by means <strong>of</strong> two features, [+/-pred] and [+/-arg]. In Italian, for<br />
instance (and Romance more generally), noun phrases are [+pred],[-arg], hence are always<br />
headed by a determiner head D to turn into arguments. By contrast, English nouns are<br />
optionally [+pred] or [+arg]. Thus, a D <strong>of</strong>ten heads a predicative noun to form an argument,<br />
but English allows nouns to function as arguments without D, as in the case <strong>of</strong> bare plurals.<br />
The hypothesis accounts for a number <strong>of</strong> contrasts between Italian and English. Where English<br />
allows bare plurals as in (36), Italian necessarily involves a definite article (37a), or some<br />
indefinite determiner (37b) or the bare-partitive construction (37c). Bare nouns in Italian are<br />
assumed to involve a null D (Longobardi 1986,1994, Chierchia 1998) and are restricted to<br />
some special governed positions, a syntactic condition necessary for licensing the null D head.<br />
(36) a. Lions are wild animals.<br />
b. Dogs are barking in the courtyard.<br />
c. Water is dripping from the faucet.<br />
(37) a. i leoni sono animali selvaggi<br />
the lions are animals <strong>of</strong> wild<br />
Lions are wild animals.<br />
b. alcuni cani stavano giocando nel giardino<br />
some dogs were playing in-the garden<br />
Some dogs were playing in the garden.<br />
c. Del vino si e’<br />
<strong>of</strong>-the wine refl past<br />
Some wine got spilled.<br />
rovesciato<br />
spill<br />
23<br />
(Chierchia 1998)<br />
(Adapted from Chierchia 1998)
So how does Greek fit into this typology? Unlike Italian, Greek bare nouns are much more<br />
productive. 14 However, unlike English, Greek bare nouns exclude kind interpretations, since,<br />
as pointed out by Roussou and Tsimpli (1994), established kinds as in (38) are necessarily<br />
definite.<br />
(38) a. i/*∅ dinosavri ehun eksafanisti<br />
the-nom/*∅ dinosaurs-nom have-3pl disappeared<br />
Dinosaurs are extinct. (bare nominal ungrammatical under the kind reading)<br />
b. ta/*∅ skilia ine katikidhia zoa<br />
the/*∅ dogs are domestic animals<br />
Dogs are domestic animals.<br />
c. ?dinosavri ehun eksafanisti<br />
dinosaurs-nom have-3pl disappeared<br />
Dinosaurs have disappeared (bare nominal possible only under the existential<br />
reading).<br />
In addition, bare nominals in Greek may be singular as well as plural.<br />
In sum, we need a three-way distinction between Italian, Greek and English. We propose<br />
that Italian and Greek nouns are both [+pred], [-arg]. The difference lies in the element that<br />
turns the predicative noun into an argument, i.e. on the ”nominalisor”; we propose that it<br />
is D in Italian but Number in Greek. The first consequence <strong>of</strong> this difference is the wider<br />
availability <strong>of</strong> bare nouns in Greek. If number is enough to turn a predicative noun into an<br />
argument in the absence <strong>of</strong> D, bare nouns are expected to be more widespread. In addition,<br />
if number is the nominalisor, both singular and plural bare nouns are also expected. Finally,<br />
since Greek nouns are not [+arg] they cannot freely shift to kinds and, thus, established kinds<br />
cannot be bare nouns (38), in contrast to English.<br />
The strength <strong>of</strong> this view lies in the fact that it allows us to give a three way varia-<br />
tion relying on existing analytical tools: Chierchia’s semantic parameter and the possibility<br />
that different individuating heads (D or Num) may act as nominalisors crosslinguistically.<br />
Of course, it raises a host <strong>of</strong> questions: why number in Greek but D in Italian? What is the<br />
14 To give a quantitative perspective, according to Marinis (2003), around 45% <strong>of</strong> target like child Greek<br />
involves bare nouns.<br />
24
semantics <strong>of</strong> Greek Number? We will provide some speculative answers to these questions in<br />
our concluding section.<br />
The crosslinguistic variation provides our second counter argument to the hypoethesis<br />
that Greek bare nouns involve a null D. Quite apart from the Greek internal facts reviewed<br />
in section 4.1.1, it is hard to see how the variation between Italian and Greek is to be accounted<br />
for if in both languages bare nouns are DPs with a null D. At this point, it is worth considering<br />
one argument presented in favor <strong>of</strong> the null D hypothesis in Greek. Sioupi (2001) notes that<br />
bare nominals are excluded from subject positions as in (39). She takes such examples to<br />
indicate that there are special structural conditions licensing bare nominals, namely that the<br />
bare nominal be governed. Such structural restrictions are evidence for a null D, which, as in<br />
Italian, is not freely available, but needs to be structurally licensed.<br />
(39) *pedia efagan to psari<br />
children ate the fish<br />
Children ate the fish. (From Sioupi 2001, ex.4a)<br />
However, as we have seen already, bare nouns can appear in subject (non-governed) posi-<br />
tions; this is further illustrated by the examples below.<br />
(40) a. itan enas hamos; yinekes epsahnan ta pedia tus mes ta<br />
was a disaster; women were-looking-for the children their in the<br />
halasmata; pedia kitazan yiro tus sastismena<br />
ruins; children were-looking around them startled<br />
It was a mess; women were looking for their children in the ruins; children were<br />
looking around startled.<br />
b. ton gratzunisan gates tu dromu<br />
him.cl scratched-3pl cats the-gen street-gen<br />
Street cats scratched him.(Kolliakou 2003)<br />
c. ta hronia ta palia, varia fortia fevgan ya tin America<br />
the years the old, heavy loads were-leaving for the America<br />
In old times heavy loads (<strong>of</strong> immigrants) were leaving for Americal (popular song<br />
by D.Papakonstantinou).<br />
d. alepudes irthan ke perisi<br />
foxes came-3pl and last-year<br />
Foxes appeared last year as well.<br />
25
(41) a. karharias ehi na emfanisti s’afti tin periohi apo to 2002<br />
shark-nom has subj appear-3sg in-this the region<br />
A shark has not appeared in this area since 2002.<br />
since the 2002<br />
b. ton exetase yatros<br />
him.cl examined-3sg doctor-nom<br />
A doctor examined him.(Kolliakou 2003)<br />
c. kleftis de spai tetia klidaria me tipota<br />
thieve-nom not break-3sg such lock with nothing<br />
There’s no way a thieve can break such a lock.<br />
Why then the badness <strong>of</strong> (39)? Preverbal subjects in Greek are standardly viewed as topics<br />
(Philippaki-Warburton 1985; Alexiadou and Anagnostopoulou 1998). 15 The problem with (39)<br />
is that it is hard to see how a bare indefinite which normally is interpreted as a weak indefinite<br />
can be a topic in a sentence where the subject/topic has eaten a specific/definite fish. Consider<br />
now the contrast between (39) and (40a); what is their difference? Intuitively, yinekes and<br />
pedhia are felicitous topics because what is at issue is not specific sets <strong>of</strong> women or kids,<br />
but some representatives <strong>of</strong> each kind. By contrast, in (39) the reading that the bare noun<br />
forces, some representatives <strong>of</strong> the kind <strong>of</strong> kids ate the fish is an odd one. The more natural<br />
interpretation which is some (specific) kids ate the fish would necessitate an explicit determiner<br />
like kati (=some or other) or kapia (=some). Importantly the fact that these overt elements<br />
are necessary here in order to obtain the right interpretation is evidence that bare nouns<br />
cannot involve a null D since they always appear to have distinct interpretations from their<br />
counterparts with overt prenominal determiners.<br />
This explanation may appear in contradiction with the fact that in examples like (31)<br />
the bare noun introduces a discourse referent whereas in (39) this seems not possible. The<br />
difference lies in the topic interpretation such examples elicit. Bare nouns in subject positions<br />
can introduce discourse referents as shown by yatros in (41b) and (25c), where kalos yatros<br />
is preverbal. The contrasts between (39) and (41b) require a systematic investigation <strong>of</strong> the<br />
interaction between the semantics <strong>of</strong> bare nouns and topichood which is beyond the scope <strong>of</strong><br />
this paper. However, the data presented here indicate clearly that preverbal bare nouns can<br />
15 See though Roussou and Tsimpli (2006), who by and large accept the view that preverbal subjects are<br />
predominately topics but also argue for a preverbal subject position.<br />
26
e felicitous subjects, even under referential interpretations.<br />
Returning to our original set <strong>of</strong> questions, we have shown so far that the indefinite nominals<br />
that are involved as topics in clld and Topicalisation have a different internal syntax in the<br />
two languages. In Italian they systematically involve DPs while in Greek they surface as bare<br />
nouns which we analyse as Number Phrases. We now turn to the properties <strong>of</strong> the pronominal<br />
element resuming the clld-ed phrase so as to understand how the structural difference in<br />
the indefinite antecedents involved in clld and topicalisation interact with the pronominal<br />
clitic. The properties <strong>of</strong> the Greek pronominal clitic are inseparable from the properties <strong>of</strong> the<br />
definite article since the two elements are morphologically identical (in all genders, numbers<br />
and cases) and the pronominal is standardly assumed to be <strong>of</strong> the same categorical status<br />
with the article (see Alexiadou, Haegeman, and Stavrou (2007) and references therein). We<br />
therefore turn to the analysis <strong>of</strong> Greek definites next.<br />
4.1.3 Greek definites<br />
We have established so far that Greek arguments need not be DPs. The question is whether<br />
Greek nominals are ever DPs and the obvious case to be considered is definites. The analysis<br />
<strong>of</strong> the definite article has been a matter <strong>of</strong> controversy in the Greek literature, not the least<br />
because it does not exhibit standard properties <strong>of</strong> a determiner head. First <strong>of</strong> all, the arti-<br />
cle is not in complementary distribution with demonstratives; in fact, it is obligatory with<br />
demonstratives (42), a fact which has been accounted for by assuming that the article realises<br />
a distinct head, Def (Definiteness), selected by D (determiner) which hosts the demonstra-<br />
tive in an example like (42) Androutsopoulou 1994, 1995. (Alternatively, the article has been<br />
viewed as agreement—Karanassios 1992, Stavrou 1996, Mathieu and Sitaridou 2002).<br />
(42) afto *∅/to vivlio<br />
this the<br />
this book<br />
book<br />
In addition, the article can co-exist with weak possessive pronouns that are attached to the<br />
right <strong>of</strong> nouns or adjectives (Alexiadou and Stavrou 2000).<br />
27
(43) a. to vivlio mu<br />
the book<br />
my book<br />
my<br />
b. to palio mu aftokinito<br />
the old my car<br />
my old car<br />
Perhaps most problematically for a head analysis <strong>of</strong> the article, Greek licenses polydefinites<br />
or determiner spreading structures like (44a) along the monadic ones (44b) (for detailed<br />
discussions <strong>of</strong> polydefinite structures see Androutsopoulou 1994; Alexiadou and Wilder 1998;<br />
Kolliakou 2003,2004, Lekakou and Szendröi 2010 among others and Alexiadou, Haegeman,<br />
and Stavrou 2007 for an overview).<br />
(44) a. to kokino to podhilato to kenurjio<br />
the red the bike the new<br />
the red new bike<br />
b. to kokino kenurjio podhilato<br />
the red new bike<br />
the red new bike<br />
Both the monadic and polydefinite above refer uniquely to one bike; thus, the polydefinite<br />
structures are not multiple definite phrases, as a head analysis would predict. To account for<br />
this, many analyses stipulate that the article can be expletive, following Androutsopoulou<br />
(1994). However, such analyses do not clarify when the article is expletive and when not, as<br />
pointed out by Kolliakou (2003). Crucially, this view is misguided. The main piece <strong>of</strong> evidence<br />
provided in support <strong>of</strong> the expletive analysis <strong>of</strong> the Greek article is its obligatoriness with<br />
proper nouns as in (45).<br />
(45) mu aresi o Messi/*Messi ala protimo to<br />
me-gen like-3sg the-nom Messi<br />
Maradona/*Maradona<br />
Maradona<br />
I like Messi but I prefer Maradona.<br />
but prefer-1sg the-acc<br />
28
However, the data below indicate that the article is not expletive even with proper names.<br />
Consider the contrast in (46) and (47). Examples (46a) and (47a) are specificational sen-<br />
tences whereas (46b) and (47b) are equatives. (46a) means Messi does not have properties <strong>of</strong><br />
Maradona, e.g. he is not as good a player. Similarly, (47a) means that Greece does not have<br />
properties <strong>of</strong> European countries while (47b) means that the referent <strong>of</strong> Europe is not the<br />
same as the referent <strong>of</strong> Germany (there are more countries in Europe). The only difference<br />
between these minimal pairs is the definite article preceding the proper names Maradona and<br />
Evropi.<br />
(46) a. o Messi dhen ine Maradona<br />
the-nom Messi not is Maradhona<br />
Messi is not a Maradona/like Maradona.<br />
b. aftos dhen ine o Messi; ine o Maradona; tus<br />
this-nom not<br />
mperdhepses<br />
mixed-up-2sg<br />
is the-nom Messi: is the-nom Maradona; them.cl<br />
He is not Messi; he is Maradona; you mixed them up.<br />
(47) a. i Eladha dhen ine Evropi<br />
the-nom not is Europe<br />
Greece is not (like) Europe.<br />
b. i Germania dhen ine i Evropi<br />
the-nom Germany not is the Europe<br />
Germany is not the whole <strong>of</strong> Europe.<br />
The contrast extends beyond predicative structures. Example (48a) means that someone<br />
with the properties <strong>of</strong> Judas (e.g. a traitor) advised the subject <strong>of</strong> the verb (him) while<br />
(48b) means that the object <strong>of</strong> the verb (him) was examined by someone with properties <strong>of</strong><br />
Kassandra (e.g. someone who only makes negative predictions for the future). If these proper<br />
nouns were accompanied by a definite article, they would refer to individuals.<br />
(48) a. ton simvulepse Iudas<br />
him.cl advised-3sg Judas<br />
He was advised by someone like Judas.<br />
b. ton exetase Kassandra<br />
him.cl examined-3sg Kassandra<br />
29
He was examined by someone like Kassandra.<br />
The above examples indicate that even with proper names the use <strong>of</strong> the definite article is not<br />
expletive. If the article is not expletive, it is hard to see how it can be a head in polydefinite<br />
structures. Note also that the additional article in polydefinites is not expletive. As discussed<br />
in detail by Kolliakou (2003), the ”second” definite in a polydefinite restricts the range <strong>of</strong><br />
the first one, even when proper names are involved. Thus, (49a) presupposes more than one<br />
Christinas and the polydefinite in (49b) picks the one with the surname Sevdali.<br />
(49) a. Pia<br />
Christina tha erthi;<br />
who-fem.nom Christina will come-3sg?<br />
Which Christina will come?<br />
b. i Christina i Sevdali<br />
the-nom Christina the-nom Sevdali<br />
Christina Sevdali.<br />
When such restrictive modification is not possible for pragmatic reasons, the polydefinites are<br />
infelicitous as shown by (50).<br />
(50) a. Taksidhepse ston plati Iriniko<br />
travelled-3sg in-the wide Pacific<br />
She travelled in the wide Pacific.<br />
b. =Taksidhepse<br />
travelled-3sg<br />
ston Iriniko ton plati<br />
in-the Pacific the wide<br />
(From Kolliakou 2003, ex.14)<br />
Research from a different vein indicates that Greek definite phrases behave like noun phrases in<br />
articless languages rather than DPs in languages with articles. In particular, Bo˘sković (2008)<br />
proposes that nominals in languages without articles show properties that systematically<br />
distinguish them from nominals in languages with articles. The contrast can only be explained<br />
if the former instantiate Noun Phrases rather than DPs with null Ds. Interestingly, Greek<br />
nominals exhibit some <strong>of</strong> the properties <strong>of</strong> noun phrases in articless languages. 16 The first<br />
16 Boscovic is aware <strong>of</strong> the fact that Greek is a potential exception to his generalisations,ibid, fn.3.<br />
30
generalisation is that only languages without articles, allow Left Branch Extraction (LBE);<br />
thus, LBE is ungrammatical in English (51a), but available in Serbo Croatian (51b).<br />
(51) a. *Expensive/*ThatI he saw [ti car]<br />
b. Skupa/Tai je vidio [ti kola] (Serbo Croatian)<br />
expensive/that is seen car<br />
(From Bo˘sković 2008, ex.3-4)<br />
Greek does allow LBE (as also noted by Boskovic). The grammaticality <strong>of</strong> (52a) is not<br />
surprising, given that we are analysing bare nominals like akrivo aftokinito as NumPs. 17 The<br />
crucial assumption is that there is no DP layer, which, according to Bo˘sković (2008) is a phase<br />
blocking extraction.<br />
(52) akrivo aghorase aftokinito<br />
expensive baught-3sg car<br />
He bought an expensive car.<br />
Cases with a demonstrative are more interesting since, as we’ve seen, a demonstrative like<br />
afto (=this) necessarily involves the definite article. As can be seen in (53), the demonstrative<br />
can be extracted out <strong>of</strong> the nominal, indicating that afto to aftokinito is in fact a noun phrase.<br />
(53) afto aghorase to aftokinito<br />
this bought-3sg the car<br />
She bought this car.<br />
The crucial fact here is that the presence <strong>of</strong> the article does not affect the extraction possibili-<br />
ties. So if we consider a nominal involving a demonstrative and an adjective like afto to akrivo<br />
aftokinito (=this the expensive car) we see in (54) that the demonstrative+article+adjective<br />
can undergo left extraction.<br />
(54) afto to akrivo aghorase aftokinito<br />
this the expensive bought car<br />
17 All LBE examples are pragmatically marked; here we assume that at least one element <strong>of</strong> the extracted<br />
phrase bears the sentential stress as indicated by the small caps.<br />
31
Note further that the pattern is exactly the same if the nominal involves a numeral+adjective<br />
as in (55).<br />
(55) a. ena kalo thelo krayon<br />
one/a good want-1sg lipstick<br />
b. dhio<br />
two<br />
kala<br />
good<br />
thelo paradighmata<br />
want-1sg examples<br />
In sum, not only Greek allows LBE despite the fact that it has articles, but, in addition,<br />
definite and indefinite phrases behave alike, indicating that what is <strong>of</strong> relevance to Boskovic’s<br />
generalisation is not whether a language has a definite article but whether this article is a D<br />
head. The evidence so far indicates that Greek definites are not DPs, but, rather behave like<br />
NPs.<br />
Bo˘skovi˘c’s second generalisation is that languages without articles allow adjunct extraction<br />
as in (56a) while languages with articles disallow it (56b). 18<br />
(56) a. *From which cityi did Peter meet [NP girls ti ]?<br />
b. Iz kojeg gradai je Ivan sreo [NP djevojke ti] (Serbo Croatian)<br />
From which city did Ivan meet girls?<br />
Again, Greek nominals pattern with an articleless language like Serbo Croatian rather than<br />
English since they allow adjunct extraction as in (57).<br />
(57) apo pia poli gnorise koritsia o Petros?<br />
from who-fem city met-3sg girls<br />
Petros met girls from which city?<br />
the-nom Petros-nom<br />
Note that, as shown by Horrocks and Stavrou (1987) , Greek allows possessor extraction; if<br />
DPs are phases and Boskovic’s argumentation is valid, then the extraction in (58) indicates<br />
18 Though both English and Serbo Croatian involve a NP here and, therefore there is no DP phase, only<br />
the SC NP allows AP extraction because AP is heading the NP in English while in SC it sits on Spec,NP.<br />
This difference derives from the fact that English has DPs and SC NPs as arguments. If he is right, then the<br />
evidence indicates that in Greek as well APs are at Spec,NumP; this is exactly the assumption we’ll be making<br />
in the next section when we consider IAD and bare subnominal ellipisis.<br />
32
that the definite is a noun phrase. 19<br />
(58) pianu martira arnithikan na eksetasun tin katathesi<br />
who-gen witness-gen refused-3pl subj examine-3pl the-acc testimony<br />
Whose witness testimony did they refuse to examine?<br />
The facts reviewed in this section 20 indicate strongly that the Greek article does not behave<br />
like a D head while there is evidence that definite nominals allow extraction possibilities typical<br />
<strong>of</strong> noun phrases that are not DPs. 21 We will, therefore, assume that the article is not a D<br />
head, but rather a prenominal modifier and that Greek nominals, definite and indefinite, are<br />
uniformingly Number Phrases. 22 Definites then are just definite Number Phrases. Further,<br />
19 Horrocks and Stavrou (1987) in fact use this type <strong>of</strong> evidence to argue for a DP; they link long possessor<br />
extraction as in (58) with focus movement within the nominal as in (i). They argue that tu protu martira<br />
in (i) moves to a position internal to the nominal exactly like the wh-phrase in (58) moves to CP. They take<br />
the article to be a D head allowing focus-movement to its Spec. While examples like (i) necessitate movement<br />
internal to the nominal, it is not necessary that this is to Spec,DP as we will see shortly.<br />
(i) arnithikan na eksetasun tu protu martira tin katathesi<br />
refused-3pl subj examine-3pl the-gen first-gen the-acc testimony<br />
They refused to examine the first witness’s testimony.<br />
20 We note that, unsurprisingly, Italian patterns with English and DP languages with regard to the diagnostics<br />
proposed by Bo˘sković (2008).<br />
21 Bo˘sković (2008) discusses some further generalisations which are either not relevant for Greek (e.g. supe-<br />
riority effects for multiple wh-fronting) or are trivially relevant: for instance, Greek most is i perissoteri (=the<br />
most), that is it implicates the definite article and has the expected reading <strong>of</strong> more than half; in addition,<br />
Greek allows clitic doubling since it has articles. Boskovic links clitic doubling to the existence <strong>of</strong> DPs and,<br />
indeed takes the referentiality <strong>of</strong> clitic doubling structures as a consequence <strong>of</strong> the involvement <strong>of</strong> D. The<br />
facts discussed in this paper are a counter example to these correlations. Firstly, Italian, which clearly has<br />
DPs allows non-referential readings, whereas Greek, which certainly allows NPs does not allow non-referential<br />
readings in doubling structures.<br />
22 This is not too far from Kolliakou (2003) who takes the definite to be an argument <strong>of</strong> a noun appearing at<br />
its Spec. Further, she assumes that definite and indefinite nominals are all noun phrases; in her HPSG analysis,<br />
any lexical category specified for the head feature nom (in turn specified for number, gender and case) can<br />
project a nominal; apart from nouns, articles, numerals and adjective share this head feature reflecting the<br />
fact that any <strong>of</strong> these categories can project a nominal argument on its own.<br />
33
following Giannakidou and Merchant (1997), we assume that numerals including ena are also<br />
prenominal adjectives. This analysis correctly predicts that, bar semantic anomaly, more than<br />
one <strong>of</strong> these elements may appear prenominally as in (59).<br />
(59) a. afto to ena aftokinito<br />
this the one car<br />
this one car<br />
b. afta ta dhio kokina aftokinita<br />
these the two red cars<br />
these two red cars<br />
Further, the noun can be elided from all these structures as in (60) (Kolliakou 2003; Gian-<br />
nakidou and Merchant 1997; Giannakidou and Stavrou 1999). We see below that an adjective<br />
alone (60a) or a numeral and an adjective (60d) may be the only (overt) part <strong>of</strong> a NumP.<br />
The definite article is no different in (60b) and (60c), except for the fact that in these uses<br />
it is, descriptively a pronoun. In other words, the pronominal clitic involved in clld is an<br />
elliptical NumP. 23<br />
(60) a. tu klepsane to aftokinito ke pire kenurjio<br />
his.cl stole-3pl the car and bought new<br />
His car was stolen and he baught a new one.<br />
b. ta dhio prota vivlia ine tis Marias; fer’ta mu se parakalo<br />
the two first books are the-gen Maria-gen; bring them.cl you beg<br />
The first two books belong to Maria. Bring them to me please.<br />
c. aghorase kenurjio aftokinito ke to efere na to dhume<br />
bought-3sg new car and it.cl brought subj it.cl see<br />
d. psahname ya kero mathimatiko ya ti desmi alla kataferame ke<br />
looking-for-1pl for time mathematician for the ”a-levels” for<br />
vrikame enan ekseretiko<br />
quite a<br />
bit <strong>of</strong> time, but managed-1pl and found-1pl an excellent<br />
We were looking for a mathematician for A-levels but we managed to find an<br />
23 Of course, a definite NumP is very different from a ”red” NumP both in semantic terms but also in PF<br />
terms, since the article/pronoun always cliticises on a host adjective/noun or verb. The point though is that<br />
there is no structural difference between a definite NP and any other NumP either in terms <strong>of</strong> extraction<br />
possibilities or the distribution <strong>of</strong> the article against all other prenominal elements.<br />
34
excellent one (so the time we took looking was well spent).<br />
We can now return to our original set <strong>of</strong> questions 24 and see how the assumption that<br />
Greek nominals involve NumPs can account for the facts presented in section 3, summarised<br />
on table (22). We should point out that, even though we advocate here the total absence<br />
<strong>of</strong> DPs from Greek, a mixed analysis which would assume that bare and indefinite nominals<br />
are NumPs but definite nominals are DPs would also account for the anaphoric patterns and<br />
the clld/topicalisation facts; we will point this out where relevant in the coming sections.<br />
Before we turn to our original set <strong>of</strong> questions though, we’ll briefly discuss the crosslinguistic<br />
implications <strong>of</strong> our hypothesis for the nominal system in Italian and Greek.<br />
4.1.4 Definites in Italian and Greek<br />
The main consequence <strong>of</strong> our analysis so far is that Number is the main nominalisor in<br />
Greek (including definites) while it is D in Italian that plays this role. In fact, this conclusion<br />
can shed light to a further set <strong>of</strong> crosslinguistic differences between the two languages. First<br />
and foremost, the systematic need <strong>of</strong> D in Italian to construct a nominal explains the wide<br />
range <strong>of</strong> D elements available in the language; the article and the pronouns retain distinct<br />
morphology; there is a distinction between a range <strong>of</strong> definite D elements (il, lo/la, gli/li...)<br />
and indefinite ones (un, bare partitive construction, ne) as we’ve seen bare nouns are very<br />
restricted, while the partitive construction is systematically employed for indefinites and mass<br />
nouns. By contrast, Greek has only one definite element which has both article and pronominal<br />
use; it characteristically lacks a separate indefinite article or pronoun.<br />
24 Examples like (i), where the definite article is involved in CP nominalisation may support the view that<br />
the article is a head afterall. We speculate that the article contributes nominal phi-features to C but still is<br />
not heading the structure.<br />
(i) (to) pios tha kerdisi tis ekloyes tha eksartithi apo to pos tha pai i<br />
the who-nom will winn-3sg the-acc elections will depend-3sg from the how will go-3sg the-nom<br />
ikonomia<br />
economy<br />
Who will win the elections will depend on how things go with the economy.<br />
35
Secondly, the Greek article has strong definite and referential interpretations, relevant<br />
even in the case <strong>of</strong> proper names. By contrast, the Italian, article appears to have weaker<br />
definite semantics. For instance, non referential nouns like gli occhiali in (61) or l’orologio and<br />
la macchina in (62) are definite in Italian.<br />
(61) A:Porti gli occhiali? B:Si, li porto<br />
A:wear the glasses? B:Yes, them.cl wear<br />
Do you wear glasses? Yes, I do.<br />
(62) a. Porti l’orologio?<br />
wear-2sg the<br />
Do you wear watch?<br />
watch?<br />
b. Guida la macchina?<br />
drive-3sg the machine<br />
Does he drive a car?<br />
Unsurprisingly, the Greek countrparts <strong>of</strong> these examples involve bare nouns.<br />
(63) foras (*ta) yialia/ (*ti) vera/ (*ta) takunia<br />
wear-2sg (*the) glasses/ (*the) wedding-ring/ (*the) high-heels<br />
Do you wear glasses/a wedding ring/high heels?<br />
(64) odhigis aftokinito<br />
drive-2sg car<br />
Do you/can you drive a car?<br />
Further, as pointed out by Giusti (2010), the definite article is preferred in cases like (65)<br />
where the interpretation is an indefinite one.<br />
(65) Scommetto che non troverai mai la/?una segretaria di un onorevole che<br />
bet-1sg that not will-find never the/a secretary <strong>of</strong> a depute who<br />
sia disposta a testimoniare contro di lui<br />
can-subj submit-subj a testimony against <strong>of</strong> him<br />
I bet you’ll never find the secratary <strong>of</strong> a depute who can testify against him. (Giusti<br />
2010)<br />
Again, the definite article is banned from such environments in Greek, since it cannot head<br />
relatives rendered in subjanctive.<br />
36
(66) *den prokite na vri ti yineka pu na tu kani ola ta<br />
not going-to-3sg subj find-3sg the-acc woman that subj him.cl do-3sg all<br />
hatiria<br />
favours<br />
He’s not going to find the woman that will satisfy every whim <strong>of</strong> his.<br />
the<br />
This set <strong>of</strong> contrasts can be understood, if, as proposed by Giusti (1993, 1997, 2002, 2010),<br />
the main role <strong>of</strong> the Italian article is that <strong>of</strong> a syntactic/grammatical morpheme acting as the<br />
nominalisor <strong>of</strong> a predicative noun, building a DP argument. By contrast, in Greek the article<br />
appears only when needed for semantic/pragmatic reasons, since Number is the nominalisor,<br />
hence the stronger definite interpretations when the article is involved.<br />
The crosslinguistic contrast also indicates that the more referential/definite interpreta-<br />
tions arise in the language where the nominal has less structure and, indeed, is not a D head,<br />
i.e. Greek. This is expected since the definite article is implicated for interpretation. By con-<br />
trast, in Italian, D is syntactically necessary as the nominalisor, and, therefore, weaker/less<br />
referential readings arise when D is needed as a nominalisor in contexts which are not clearly<br />
definite/referential like (65) or under the de dicto reading <strong>of</strong> (67) or even the use <strong>of</strong> lo in the<br />
predicative structures in (68). It is worth pointing out that our view departs from the propos-<br />
als <strong>of</strong> Longobardi (1994) and Déchaine and Wiltschko (2002) in one crucial respect. What<br />
these two proposals share is that D and referentiality are intrinsically linked. Both analyses<br />
assume a structural ambiguity for articles and pronouns, depending on their interpretation.<br />
For instance, lo under the de dicto reading in (67) and in the predicative structure in (68)<br />
lacks a D layer, and is just a NP. 25 We will discuss these examples in more detail in the next<br />
section.<br />
(67) Gianni sta cercando un idraulico ma non lo trova<br />
Gianni is looking-for a plumber but not him.cl find-3sg<br />
Gianni is looking-for a plumber but cannot find one.<br />
(68) a. Bella la/lo e’<br />
beautiful it.cl is<br />
Beautiful she is.<br />
25 Longobardi pc after related talk <strong>of</strong> his at Cambridge University, <strong>2011</strong>.<br />
37
. belle lo sono?<br />
beautiful it are<br />
Beautiful they are.<br />
In sum then, the hypothesis that D is the nominalisor in Italian while Greek nominals<br />
are NumPs can account for the rich inventory <strong>of</strong> D elements in Italian as well as the weaker<br />
definite semantics <strong>of</strong> D elements in this language.<br />
4.2 Indefinite Argument Drop and Bare Subnominal Ellipsis<br />
As Giannakidou and Merchant (1997) show, Greek Indefintie Argument Drop (IAD) in-<br />
volves recycling <strong>of</strong> the descriptive content <strong>of</strong> the antecedent (or property anaphora in terms<br />
<strong>of</strong> Tomioka 2003).Crucially, it does not pick the discourse referent <strong>of</strong> the antecedent. 26 Con-<br />
sider (69). Example (69a) does not allow a reading in which Nafsika dried the dishes Napoleo-<br />
das washed, a reading available in (69b) where the pronoun is used. On the basis <strong>of</strong> the<br />
obligatoriness <strong>of</strong> the disjoint reading in (69a), Giannakidou and Merchant (1997) conclude<br />
that an empty pro analysis, proposed by Dimitriadis (1994) is not possible, since it would<br />
predict the co-referential reading in (69b). What IAD involves in (69a) is the recycling <strong>of</strong><br />
the descriptive content <strong>of</strong> the antecedent, i.e. piata but not the specific set introduced by the<br />
antecedent.<br />
(69) a. o Napoleodas epline piata ke i Nafsika skupise<br />
the-nom Napoleodas-nom washed-3sg dishes and the-nom Nafsika dried-sg<br />
Napoleon washed dishes and Nafsika dried dishes. (Disjoint reading)<br />
26 See Giannakidou and Merchant (1997) for a detailed discussion <strong>of</strong> the properties <strong>of</strong> IAD. Panagiotidis<br />
(2002) excludes a VP-ellips analysis <strong>of</strong> the phenomenon; his main argument is that all restrictions relevant to<br />
IAD involve the nominal antecedent and never any verbal element. Further, examples like (i), where only the<br />
object is dropped but other parts <strong>of</strong> the VP are overt indicate that what is dropped is just the object.<br />
(i) ti mia mera vrike dulia o Yorgos stu Zografu ke to epomeno proi<br />
the one day found-3sg job the-nom Yorgos-nom at-the Zografu and the next morning<br />
vrike i Maria stin Kesariani<br />
found-3sg the-nom Maria at-the Kesariani<br />
One day Yorgos found a job at Zografu and the next morning Maria found one at Kesariani.<br />
38
. o Napoleodas epline (ta) piatai ke i Nafsika tai<br />
the-nom Napoleodas-nom washed the<br />
skupise<br />
dried-3sg<br />
dishes and the-nom Nafsika them<br />
Napoleon washed (the) dishes and Nafsika dried them.<br />
That these cases involve property anaphora is further confirmed by the fact that adjec-<br />
tives like tetios/tetia/tetio standardly used for concept or property anaphora (Kolliakou 2003)<br />
license argument drop.<br />
(70) a. vrikes teties<br />
(melitzanes)?<br />
found-2sg such-acc.fem.pl (aubergines)<br />
Did you find such ones/aubergines?<br />
b. ne, vrika<br />
yes, found-1sg<br />
Yes, I found.<br />
Further, Giannakidou and Merchant (1997) establish that IAD is licensed by weak indefinite<br />
quantifiers while strong quantifiers necessitate a pronoun.<br />
(71) a. Q:Efere o Adreas ola ta/ke ta dio/ta perisotera vivila<br />
Q:Brought-3sg the-nom Adreas all the/and the two/the most books<br />
Did Adread bring all/both/most books?<br />
b. A:Ne, *(ta) efere<br />
A:Yes, *(them) brought-3sg<br />
Yes, he brought them.<br />
(Giannakidou:Merchant:1997)<br />
(72) a. Efere o Adreas merika/kapja/liga/deka/tulahiston<br />
brought-3sg the-nom Adreas several/some/a-few/ten/at-least<br />
tria/parapano apo tria/tipota/∅ vivlia<br />
three/more from three/any/∅ books<br />
Did Andreas bring several/some/a few/at least three/more than three/any/∅<br />
books?<br />
b. Ne, (*ta) efere e.<br />
Yes, (them) brought-3sg e<br />
Yes he brought several/some/a few/ten/at least three/more than three/some/∅<br />
books.<br />
39
(Giannakidou:Merchant:1997)<br />
Returning to our analysis, we argue that the weak indefinites in (72) are NumPs; we, then,<br />
analyse IAD as a case <strong>of</strong> NumP ellipsis (Tomioka 2003). 27 A NumP ellipsis analysis <strong>of</strong> IAD<br />
accounts for the availability <strong>of</strong> IAD with subjects in Greek—see (17) repeated as (73). We<br />
also explain why IAD is unavailable in Italian, since there are no NumP arguments in the<br />
language.<br />
(73) a. A:irthe kanis? B:ne irthe<br />
A:came-3sg anyone? B:yes, came-3sg<br />
A:Did anyone come? B:Yes, someone did.<br />
b. A:tilefonisan fitites? B:ne tilefonisan<br />
A:phoned-3pl students? B:yes, phoned-3pl<br />
A:Did students phone? B:Yes, some did.<br />
One question is why the whole NumP cannot be elided with definites as in (71). The<br />
reason is interpretative. Absence <strong>of</strong> definite marking gives rise to indefinite interpretations.<br />
Consider for instance (74). The answer in (74b) involves a weak indefinite with an elided noun<br />
despite the definite antecedent; a definite is not appropriate in this case.<br />
(74) a. tis eferes tis valitses<br />
them.cl brought-2sg the-acc suitcases<br />
Did you bring the suitcases?<br />
b. efera (kamposes); mu ehun mini tris teseris akoma<br />
brought-1sg (many); me have-3pl left three four still<br />
I brought quite a few; but still have three or four left.<br />
Let us now reconsider examples like (75), discussed earlier. Following Zimmermann (1993)<br />
let us assume that idravliko and dada in (75) denote properties. The examples indicate that<br />
Greek pronouns resist property anaphora; the pronominal requires a referential antecedent. 28<br />
The property anaphora effect is achieved through IAD, i.e. recycling the antecedent noun.<br />
27 The ellipsis analysis preserves the basic intution <strong>of</strong> Giannakidou and Merchant (1997) who also propose<br />
that IAD involves NP deletion; however, they assume that the elided NP is headed by a null D.<br />
28 Greek pronouns can take natural functions as their antecedents as shown in examples like (i) discussed in<br />
Alexopoulou and Heycock (2003). Such antecedents though are still extensional.<br />
40
(75) a. o Yanis psahni idravliko alla dhe (*ton) vriski puthena<br />
the-nom Yanis-nom look-for-3sg plumber but not (him) find-3sg anywhere<br />
Yanis is looking for a plumber but cannot nd one anywhere.<br />
b. i Maria epsahne<br />
dada ena hrono ke telika (*ti)<br />
the-nom Maria was-looking-for-3sg nanny one year and finally (*her)<br />
vrike meso mias gnostis<br />
found-3sg through an acquaintance<br />
Maria was looking for a nanny for a year and in the end she found one through<br />
an acquaintance.<br />
As noted earlier and illustrated again in (76), the pronoun can take an indefinite an-<br />
tecedent (76a), including a bare noun (76b), as long as the antecedent is interpreted referen-<br />
tially.<br />
(76) a. o Yanis psahni enan idravliko, ala dhe ton vriski puthena<br />
the-nom Yanis look-for-3sg one plumber but not him find-3sg anywhere<br />
Yanis is looking for a plubmer but cannot find him anywhere. ∃ > look-for<br />
b. o Napoleodas epline (ta) piatai ke i Nafsika tai<br />
the-nom Napoleodas-nom washed the<br />
skupise<br />
dried-3sg<br />
dishes and the-nom Nafsika them<br />
Napoleon washed (the) dishes and Nafsika dried them.<br />
By contrast, lo and la in Italian admit property-denoting antecedents (77). Under our<br />
analysis, the crosslinguistic contrast in property anaphora is a consequence <strong>of</strong> the structural<br />
contrast between DPs and NumPs. The obligatoriness <strong>of</strong> a D element in (77) means a weak-<br />
ening <strong>of</strong> the definite/referential interpretation <strong>of</strong> the pronominal. Note that we do not predict<br />
that every language that has DPs will allow a ”definite” D in examples like (77); an indefinite<br />
D element may be used instead and it remains an open question how property anaphora is<br />
(i) a. i yineka pu misi kathe adrasi ine i pethera tui<br />
the-nom woman that hates-3sg each man is the-nom mother-in-law his<br />
The woman every man hates is his mother in law.<br />
b. tin kopela pu efere kathe fititis tii valame na katsi dipla tui<br />
the-acc girl that brought-3sg each student-nom her.cl put-1pl subj sit-3sg next his<br />
The girl each student brought, we put her to sit next to him.<br />
41
dealt with in a language. The point we are making here is that the weakening <strong>of</strong> the definite<br />
semantics <strong>of</strong> the Italian lo/la is due to the obligatoriness <strong>of</strong> DPs in Italian. Note further that<br />
these non-referential uses involve clear verbal arguments, arguing against an intrinsic link<br />
between D, argumenthood and referentiality in Italian as proposed by Longobardi (1994). If<br />
lo were to be analysed as involving less structure, e.g. as a np, the immediate question is why<br />
an NP argument is possible at all in Italian all <strong>of</strong> a sudden and how it is different from Greek.<br />
(77) Gianni sta cercando un idraulico ma non lo trova<br />
Gianni is looking-for a plumber but cannot find one.<br />
Note that even Italian, does not universally allow ”definite” pronouns to support property<br />
anaphora. For instance, when the object is plural, as in (78), the plural <strong>of</strong> lo/la, is not<br />
acceptable. Instead, the clitic ne needs to be used. 29<br />
(78) a. Q: Maria ha trovato delle aiutanti?<br />
Q: Maria has found <strong>of</strong> helpers?<br />
Has Maria found helpers?<br />
b. A: No, ?le/ne ha trovate<br />
A: No, them.cl/ne has found<br />
No she has not found.<br />
(79) Gianni sta cercando degli aiutanti per l’ufficio ma non ?li/ne<br />
Gianni has-been looking-for <strong>of</strong><br />
trova<br />
find-3sg<br />
helpers for the-<strong>of</strong>fice but not them.cl/ne<br />
Gianni has been looking for assistants for the <strong>of</strong>fice but cannot find any.<br />
We will speculate in section 4.4. on the effect <strong>of</strong> number, but, for the moment, suffice to say<br />
that Italian systematically involves a D element even for arguments that are not referential.<br />
This D element may even be an apparently definite D like lo/la or an indefinite one like<br />
ne. Greek pronouns do not support property anaphora. Instead, the relevant readings are<br />
obtained through IAD which involves NP ellipsis.<br />
29 Notice that the plural le can be used in (78) if the answer is positive, but it forces a discourse transparent<br />
reading,which is not the relevant one here.<br />
42
Let us now turn to subnominal ellipsis; the relevant examples are repeated below. The<br />
crosslinguistic pattern follows staightforwardly from the contrast between DPs and NumPs.<br />
Despite the non-referential antecedent, Italian obligatorily requires a D element, uno in the<br />
elliptical structure (note that uno grande is doubled by lo). By contrast, in Greek the bare<br />
adjective is enough. 30<br />
(80) A:Vorrei un tavolo grande B:Mi spiace. Non lo abbiamo, uno<br />
A:would-like-1sg a<br />
grande<br />
big<br />
table big B:Me displeases-3sg. Not it have-1pl, a<br />
I would like to buy a big table. I’m sorry. We do not have a big one.<br />
(From Alexiadou and Gengel 2008, attributed to V.Samek-Lodovici)<br />
(81) a. thelo afti ti fusta se kitrino<br />
want-1sg this the skirt in yellow<br />
I would like this skirt in yellow.<br />
b. Distihos dhen eho kitrini. (Mono mavres mu ehun<br />
unfortunately not<br />
mini)<br />
left)<br />
have-1sg yellow-fem.sg (Ony black-fem.pl me have-3pl<br />
Unfortunately I don’t have a yellow one. ( Only black ones are left).<br />
Before we move on, note that the essence <strong>of</strong> the analysis presented here does not rely critically<br />
on the assumption that the Greek article and pronominal do not project DPs. The critical<br />
element <strong>of</strong> the analysis is that weak indefinites, the ones that license IAD and bare subnominal<br />
ellipsis are NumPs. The incompatibility <strong>of</strong> pronouns with property-denoting antecedents is<br />
orthogonal to the categorical status <strong>of</strong> these elements, since a pronoun can accept a bare<br />
nominal antecedent (i.e. a NumP), as long as it is referential (76b) or indeed a non-bare<br />
one (76a).<br />
30 We do not discuss here the conditions <strong>of</strong> such subnominal ellipsis; for a detailed discussion see Giannakidou<br />
and Stavrou (1999). The point is that such subnominal ellipsis can take place within a bare nominal in Greek.<br />
43
4.3 Back to CLLD and Topicalisation<br />
Let us now turn to our very first question, why Italian but not Greek clld-ed elements can<br />
be ambiguous between a referential and non-referential reading. It has now become clear that<br />
the clld facts mirror the anaphoric possibilities in the two languages. Our original example<br />
in (1) repeated in (82) is as ambiguous as the example in (77). The available interpretations<br />
are exactly those allowed in Italian between an indefinite antecedent and lo/la.<br />
(82) a. una gonna rossa la cerco da un po’<br />
a red skirt her.cl look-for-1sg for a while<br />
A red skirt I’ve been looking for a while...<br />
b. ma non ne ho trovata nessuna che mi piaccia<br />
but not <strong>of</strong>-them.cl have-1sg found none-fem that me please-3sg.subj<br />
... but have not found anyone that I like.<br />
c. ma non riesco a ricordarmi dove l’ho<br />
messa<br />
but not reach-1sg to remember where her.cl-have-1sg put<br />
puffle ... but I cannot remember where I’ve put it.<br />
Similarly, Greek (3) repeated in (83), is compatible only with the referential interpretation<br />
since, as we have seen, the Greek pronoun resists property anaphora.<br />
(83) a. mia kokini fusta tin psahno edho ke meres<br />
a red skirt it look-for-1sg here and days<br />
I’ve been looking for a red skirt for a few days ...<br />
b. =ke de boro na vro kamia pu na m’aresi<br />
and not can-1sg subj find-1sg none<br />
... and I cannot find any that I like.<br />
that subj me-please-3sg<br />
c. ke de boro na thimitho pu tin eho vali<br />
and not can-1sg subj remember-1sg where her.cl have-1sg put<br />
... and cannot remember where I put it.<br />
The possibility <strong>of</strong> property anaphora in Italian further allows clld examples where the dis-<br />
located element is not an indefinite, but a predicate like bella in (84). Unsurprisingly, this<br />
possibility is not available in Greek.<br />
(84) a. Bella lo é<br />
beautiful it.cl is<br />
44
Beautiful she is.<br />
b. belle lo sono?<br />
beautiful it are<br />
Beautiful they are.<br />
While clld mirrors the anaphoric possibilities in the two languages; but Topicalisation<br />
does not, at least not directly. The interpretative possiblities in Topicalisation are the ones<br />
predicted by a movement derivation. Thus, the Topicalisation example in (85) is ambiguous<br />
like the corresponding example involving Focus-movement (86). In this, (85) contrasts with<br />
examples like (16b), repeated as (87), where IAD forces a de dicto interpretation <strong>of</strong> the<br />
antecedent. 31<br />
(85) enan idravliko psahni o yanis<br />
one-acc plumber look-for-3sg the-nom Yanis-nom<br />
A plumber Yanis is looking for. look-for > ∃, ∃ > look-for<br />
(86) enan idravliko psahni o yanis<br />
one-acc plumber look-for-3sg the-nom Yanis-nom<br />
A plumber Yanis is looking for. look-for > ∃, ∃ > look-for<br />
(87) o Yanis psahni enan idravliko, ala dhe vriski puthena<br />
the-nom Yanis look-for-3sg one plumber but not find-3sg anywhere<br />
Yanis is looking for a plubmer but cannot find one anywhere. look-for > ∃<br />
These facts point to the following conclusions. First, they reinforce the anaphoric nature<br />
<strong>of</strong> clld. At the same time they indicate that the properties <strong>of</strong> the pronominal system <strong>of</strong><br />
languages is a crucial element in the crosslinguistic variation <strong>of</strong> anaphoric chains like clld<br />
and Topicalisation. 32 Second, Topicalisation, at least in Greek, involves a proper copy/trace<br />
rather than some null pronominal element. Taken together, these conclusions indicate that<br />
the the variation in the topic strategies adopted crosslinguistically cannot be understood just<br />
31 This contrast between IAD and Topicalisation argues against the proposal by Dimitriadis (1994) to assume<br />
that the Topicalisation example is in fact a case <strong>of</strong> Empty Clitic Left Dislocation involving the same pro element<br />
implicated in IAD.<br />
32 This is unlike non-anaphoric or true resumption where a line <strong>of</strong> research systematically relates crosslin-<br />
guistic variation to the properties <strong>of</strong> C rather than the pronominal (Sells 1984; McCloskey 1990; McCloskey<br />
2002; Shlonsky 1992; Rouveret 2002; Alexopoulou 2006).<br />
45
as PF variation, since these topic strategies interact with deeper aspects <strong>of</strong> the grammars <strong>of</strong><br />
languages, in this case the structure <strong>of</strong> nominals and pronouns.<br />
4.4 Why D in Italian but Number in Greek?<br />
Our main claim is that the variation in the topic-strategies <strong>of</strong> Greek and Italian interacts<br />
with the structure <strong>of</strong> nominals. As shown, the relevant structures reflect the anaphoric pos-<br />
sibilities in the two languages which, crucially, are governed by the structure <strong>of</strong> the relevant<br />
nominal antecedents and pronouns. However, once we move away from topic-strategies and<br />
anaphoric relations to the nominals themselves, the question is whether the variation in nom-<br />
inals correlates with further contrasts in the two languages, or, to put the question somewhat<br />
differently, what enables Number to be a nominalisor in Greek while D is necessary in Italian?<br />
Our speculation is that the morphological and featural make up <strong>of</strong> nominal categories in the<br />
two languages plays a role in this respect. Greek nominal elements such as nouns, adjectives,<br />
numerals, quantifiers and the definite article show morphological case, gender and number.<br />
Any <strong>of</strong> these elements then can project a NumPhrase. Additionally, while number is crosscat-<br />
egorial in the sense that both verbal and nominal categories have number, case and gender<br />
are only nominal. Thus, case and gender morphology provide categorical marking crucial for<br />
the identification <strong>of</strong> syntactic complements, 33 while Number provides relevant individuating<br />
semantics that apply to predicative nouns to turn them into arguments. These three features<br />
then appear to work in tandem to provide syntactic marking and semantics for a nominal<br />
argument. By contrast, Italian nominals are not marked for case. Number marking in Italian<br />
is not as systematic since some D elements like ne and the bare partitive construction do not<br />
bear morphological number marking. It then seems that what Italian lacks in case (and possi-<br />
bly number) morphology, it makes up in the range <strong>of</strong> D elements that are central to building<br />
arguments; 34 Greek on the other hand, relies on the features carried by nominal elements<br />
which are crucial for the projection <strong>of</strong> nominal arguments. The definite article/pronoun are<br />
33 Indeed, Lekakou and Szendröi (2010) propose that Greek arguments are headed by a Kase head.<br />
34 This again echoes Giusti’s position that the primary role <strong>of</strong> the definite article as a functional head is<br />
syntactic, to assign case to its complement NP (Giusti 1993,1997, 2002).<br />
46
just semantically definite elements but do not act as D-heads.<br />
A further question is whether the different role <strong>of</strong> Number in the two languages means that<br />
Number has different semantics/interpretation in the two languages. Greek number indeed<br />
has some unexpected properties; first, as seen already,Greek allows bare singular arguments.<br />
In addition, mass nouns in Greek may show plural morphology as demonstrated by Tsoulas<br />
(2008) (see also Alexiadou 2010). In some intuitive sense then, number appears to be ”active”<br />
in every instance <strong>of</strong> nominal arguments (e.g. including mass nouns), a fact which can be linked<br />
to its nominalisor role <strong>of</strong> number.<br />
(88) a. trehun nera apo to tavani<br />
drip-3pl water-pl from the-sg ceiling-sg<br />
Water is dripping from the ceiling.<br />
b. to patoma itan gemato nera<br />
the-sg floor-sg was full<br />
The floor was full <strong>of</strong> water.<br />
waters-pl<br />
(From Tsoulas 2008, ex.9,10)<br />
By contrast, bare singular arguments or plural mass nouns are not available in Italian. In<br />
addition, Italian seems to have at its disposal a set <strong>of</strong> clitic pronouns which are number<br />
neutral (lo, la (in some varieties), si, ne). For instance, when these pronouns take predicates<br />
as antecedents, they are not sensitive to number as seen in (89).<br />
(89) belle lo sono?<br />
beautiful it are<br />
Beautiful they are.<br />
The same number neutrality shows up with mass nouns and reflexive predicates as in (90).<br />
(90) a. di carne ne mangia<br />
<strong>of</strong> meat ne eats<br />
b. Gianni e Maria si lavano spesso<br />
Gianni e Maria self wash <strong>of</strong>ten<br />
Finally, Italian allows some <strong>of</strong> the number neutral singular bare nouns like Catalan as in (91).<br />
47
(91) il pero e’ in fiore<br />
the pear-tree is in flower<br />
The pear tree is blossoming.<br />
A systematic investigation <strong>of</strong> these facts is beyond the scope <strong>of</strong> this paper. Note though that<br />
if Italian does allow number neutral DPs and this were to be captured by an absent Num<br />
head in the internal structure <strong>of</strong> the DP, then our analysis would entail less structure for non-<br />
referential uses <strong>of</strong> la/lo, just as proposed by Longobardi (1994) and Déchaine and Wiltschko<br />
(2002) do. However, the optional head would be Num in our analysis, not D.<br />
Finally, an understanding <strong>of</strong> the precise semantic contribution <strong>of</strong> Greek number may lead<br />
to an even simpler structure for Greek nominals where they are just Noun Phrases with a<br />
Number feature that allows them to be interpreted as arguments.<br />
5 Conclusion<br />
clld has been assumed to be the main topic-strategy in Greek and Italian; the topic oper-<br />
ator assumed in these structures is linked anaphorically to the in-situ element <strong>of</strong> the A-bar<br />
dependency (rather than through binding as is the case in quantificational A-bar chains).<br />
clld and Topicalisation have been viewed as distinct PF realisations <strong>of</strong> topic-operator struc-<br />
tures, with the pronominal in clld analysed as an overt realisation <strong>of</strong> the gap/null epithet<br />
involved in Topicalisation. We showed that the relation between the clld-ed phrase and the<br />
pronominal clitic mirror general anaphoric patterns in the two languages, therefore, reinforc-<br />
ing the view <strong>of</strong> these structures as involving an anaphoric operator. But, at the same time,<br />
we showed that the properties <strong>of</strong> the prononimal are crucial since they affect the anaphoric<br />
possibilities in each language and, as a result, the range <strong>of</strong> available readings in clld. Cru-<br />
cially, the properties <strong>of</strong> the pronominal have consequences for interpretation indicating that<br />
the crosslinguistic variation in the realisation <strong>of</strong> topic-operator structures is not confined to<br />
PF variation regarding the overtness <strong>of</strong> the in-situ element. Indeed, our main claim is that<br />
the variation in topic-structures interacts with the internal structures <strong>of</strong> nominal arguments<br />
in the two languages.<br />
We focused on the realisation <strong>of</strong> indefinite topics which is where the contrast in the<br />
48
topic-strategies between the two languages surfaces; Italian allows property denoting phrases<br />
(indefinite DPs or indeed adjectives) to be clld-ed and resumed by a clitic, while Greek<br />
clld is restricted to referential topics; non-referential/property denoting topics are necessarily<br />
topicalised. We showed that this contrast reflects a more general pattern according to which<br />
Greek pronouns resist property anaphora where Italian systematically allows a pronominal to<br />
take a non-referential antecedent. We argued that the contrast stems from the fact that Greek<br />
nominal arguments lack a D-layer and are instead NumPs. As a result, property denoting<br />
indefinites in the relevant contexts systematically involve bare nouns in Greek but DPs in<br />
Italian. These non-referential DPs can be antecedents <strong>of</strong> D-pronouns in Italian while in Greek<br />
the effect <strong>of</strong> property anaphora is achieved through NumP ellipsis (IAD) which allows the<br />
”recycling <strong>of</strong> the descriptive content <strong>of</strong> the antecedent”.<br />
The variation in the topic-strategies then is reduced to variation in the syntax <strong>of</strong> nominal<br />
arguments in the two languages. We argued that nouns are [+pred] and [-arg] in both Greek<br />
and Italian, but the two languages differ in the nominalising head, D for Italian and Num<br />
for Greek. We linked this hypothesis to the rich morphological make up <strong>of</strong> Greek nominals<br />
and the poor inventory <strong>of</strong> D-elements—-indeed we argued that even the Greek definite article<br />
is not a D head— which contrasts with the poorer case and number morphology <strong>of</strong> Italian<br />
nominal elements but the wider range <strong>of</strong> D elements. In addition, we linked the D/Num<br />
contrast to the ”weaker” semantics <strong>of</strong> the Italian definite Ds and the stronger readings <strong>of</strong> the<br />
Greek definite clitic.<br />
A number <strong>of</strong> questions remain open. How do the Italian facts compare with other Ro-<br />
mance languages? Do other Romance languages allow weaker readings <strong>of</strong> definite Ds and<br />
property anaphora? Does availability <strong>of</strong> NP arguments predict IAD? English is a potential<br />
counterexample here since NP ellipsis is ungrammatical in the English *John is looking for<br />
dinosaurs but cannot find, though dinosaurs is a NP. While these answers await a systematic<br />
investigation, our stronger prediction is that the finer variation in the interpretative possi-<br />
bilities <strong>of</strong> clld and topicalisation structures across Romance should reflect variation in the<br />
nominal structure.<br />
49
References<br />
Alexiadou, A. (2010). Plural mass nouns and the morpho-syntax <strong>of</strong> Number. In Proceedings<br />
<strong>of</strong> West Coast Conference on Formal Linguistics. Cascadilla Press.<br />
Alexiadou, A. and E. Anagnostopoulou (1998). Parmetrizing agr: word order, verb move-<br />
ment and EPPchecking. Natural Language and Linguistic Theory 16 (3), 491–539.<br />
Alexiadou, A. and K. Gengel (2008). Classifiers as morphosyntactic licensors <strong>of</strong> np ellipsis:<br />
English vs. Romance. In NELS 39.<br />
Alexiadou, A., L. Haegeman, and M. Stavrou (2007). Noun Phrase in the Generative Per-<br />
spective, Volume 71 <strong>of</strong> Studies in Generative Grammar. Mouton de Gruyter.<br />
Alexiadou, A. and M. Stavrou (2000). Adjective-clitic combinations in the Greek DP. In<br />
B. Grlach and J. Grijzenhout (Eds.), Clitics in Phonology, Morphology and Syntax.<br />
John Benjamins.<br />
Alexiadou, A. and C. Wilder (1998). Adjectival modification and multiple determiners. In<br />
A. Alexiadou and C. Wilder (Eds.), Possessors, Predicates and Movement in the DP,<br />
pp. 303–332. Amsterdam: John Benjamins.<br />
Alexopoulou, T. (2006). Resumption in relative clauses. Natural Language and Linguistic<br />
Theory 24 (1), 57–111.<br />
Alexopoulou, T. (2009). Binding illusions. In J. H. Claire Halpert and D. Hill (Eds.),<br />
Proceedings <strong>of</strong> the 2007 Workshop in Greek Syntax and Semantics at MIT, Volume 57,<br />
pp. 33–48. MIT Working Papers in Linguistics.<br />
Alexopoulou, T., E. Doron, and C. Heycock (2004). Broad subjects and clitic left disloca-<br />
tion,. In D. Adger, C. de Cat, and G. Tsoulas (Eds.), Peripheries: syntactic edges and<br />
their effects, pp. 329–358. Kluwer.<br />
Alexopoulou, T. and R. Folli (<strong>2011</strong>). Indefinite topics and the syntax <strong>of</strong> nominals in Italian<br />
and Greek. In M. B. Washburn, S. Ouwayda, C. Ouyang, B. Yin, C. Ipek, L. Marston,<br />
and A. Walker (Eds.), Proceedings <strong>of</strong> West Coast Conference on Formal Linguistics 28.<br />
Cascadilla Press.<br />
50
Alexopoulou, T. and C. Heycock (2003). Quantifier scope in relative clauses and definiteness<br />
effects. In C. Beyssade, O. Bonami, P. C. H<strong>of</strong>herr, and F. Corblin (Eds.), Empirical<br />
Issues in Formal Syntax and Semantics, Volume 4, pp. 81–96. Presses de l’Université<br />
de Paris-Sorbonne.<br />
Alexopoulou, T. and D. Kolliakou (2002). On Linkhood and Clitic Left Dislocation. Journal<br />
<strong>of</strong> Linguistics 38 (2), 193–245.<br />
Anagnostopoulou, E. (1994). Clitic Dependencies in Modern Greek. Ph. D. thesis, Univer-<br />
sity <strong>of</strong> Salzburg.<br />
Androutsopoulou, A. (1994). The distribution <strong>of</strong> the definite determiner and the syntax <strong>of</strong><br />
Greek DPs. In Papers from the 30th Annual Meeting <strong>of</strong> the Chicago Linguistic Society.<br />
University <strong>of</strong> Chicago Press.<br />
Androutsopoulou, A. (1995). The licensing <strong>of</strong> adjectival modification. In Proceedings <strong>of</strong> the<br />
14th West Coast Conference on Formal Linguistics, Stanford, California. CSLI Publi-<br />
cations.<br />
Belletti, A. and L. Rizzi (1981). The syntax <strong>of</strong> ”ne”: some theoretical implicatins. The<br />
Linguistic Review 1, 117–154.<br />
Benincà, P. (1980). Nomi senza articolo. Rivista di grammatica generativa 5, 51–63.<br />
Bo˘sković, ˘ Z. (2008). What will you have, dp or np? In E. Elfner and M. Walkow (Eds.),<br />
Proceedings <strong>of</strong> NELS 37. Booksurge.<br />
Cardinaletti, A. and G. Giusti (1990). Partitive ”ne” and the QP hypothesis. a case study.<br />
ms, University <strong>of</strong> Venice.<br />
Cecchetto, C. (2001). Syntactic or semantic reconstruction? evidence from pseudoclefts and<br />
clitic left dislocation. In C. Cecchetto, G. Chierchia, and M. T. Guasti (Eds.), Semantic<br />
Interfaces. CSLI.<br />
Chierchia, G. (1998). Reference to kinds across languages. Natural Language Semantics 6,<br />
339–405.<br />
Cinque, G. (1990). Types <strong>of</strong> A-bar Dependencies, Volume 17. MIT Press. Linguistic Inquiry<br />
51
Monographs.<br />
Déchaine, R.-M. and M. Wiltschko (2002). Decomposing pronouns. Linguistic In-<br />
quiry 33 (3), 409–442.<br />
Dimitriadis, A. (1994). Clitics and object drop in Modern Greek. In Proceedings <strong>of</strong> the<br />
Sixth Student Conference in Linguistics (SCIL-6), University <strong>of</strong> Rochester, Volume 23<br />
<strong>of</strong> MITWPL, pp. 1–20.<br />
Endriss, C. (2006). Quantificational topics, a scopal treatment <strong>of</strong> exceptional wide scope<br />
phenomen. Ph. D. thesis, University <strong>of</strong> Potsdam.<br />
Espinal, M. T. (2010). Bare nominals in Catalan and Spanish. their structure and meaning.<br />
Lingua 120, 984–1009.<br />
Espinal, M. T. and L. McNally (2007). Bare singular nominals and incorporating verbs. In<br />
G. Kaiser and M. Leonetti (Eds.), Proceedings <strong>of</strong> the III NEREUS International Work-<br />
shop. Definiteness, Specificity and Animacy in Ibero-Romance languages, Universität<br />
Konstanz, pp. 45–62. Arbeitspapier 122.<br />
Farkas, D. F. and H. de Swarts (2001). Bare nominals in the typology <strong>of</strong> indefinites. Reader<br />
for ESSLLI’01 course on the Typology <strong>of</strong> Noun Phrases.<br />
Farkas, D. F. and H. de Swarts (2003). The semantics <strong>of</strong> incorporation: from argument<br />
structure to discourse transparency. Stanford monographs in Linguistics. CSLI Publica-<br />
tions.<br />
Fodor, J. D. (1970). The linguistic description <strong>of</strong> opaque contexts. Ph. D. thesis, MIT,<br />
Published in 1976 by Indiana University Linguistics Club and in 1979 in the Series<br />
“Outstanding Dissertations in Linguistics” by Garland.<br />
Giannakidou, A. and J. Merchant (1997). On the interpretation <strong>of</strong> null indefinite objects<br />
in greek. In Studies in Greek Linguistics, Volume 17, pp. 141–155. Aristotle University.<br />
Giannakidou, A. and M. Stavrou (1999). Nominalization and ellipsis in the Greek DP.<br />
Linguistic Review 16 (4), 295–331.<br />
Giusti, G. (1993). La sintassi dei determinanti. Padova:Unipress.<br />
52
Giusti, G. (1997). The categorial status <strong>of</strong> determiners. In L. Haegeman (Ed.), The New<br />
Comparative Syntax. London: Longman.<br />
Giusti, G. (2002). The functional structure <strong>of</strong> noun phrases. a bare phrase structure ap-<br />
proach. In G. Cinque (Ed.), Functional Structure in DP and IP: The cartography <strong>of</strong><br />
syntactic structures, Volume 1, pp. 54–90. Oxford University Press.<br />
Giusti, G. (2010). The syntax <strong>of</strong> the definite article at the interfaces. University <strong>of</strong> Venice,<br />
ms.<br />
Haegeman, L. (to appear). Argument fronting in English, Romance CLLD and the left<br />
periphery. In Negation, Tense and Clausal Architecture: Cross-linguistic Investigations.<br />
Georgetown University Press.<br />
Horrocks, G. and M. Stavrou (1987). Bounding theory and Greek syntax: evidence for<br />
wh-movement in NP. Journal <strong>of</strong> Linguistics 23, 79–108.<br />
Iatridou, S. (1995). Clitics and Island Effects. Penn Working Papers in Linguistics 2, 11–31.<br />
Karanassios, G. (1992). Syntaxe Comparé du Groupe Nominal en Grec Moderne et dans<br />
d’Autre Langues. Ph. D. thesis, Université de Paris VIII.<br />
Kolliakou, D. (2003). Nominal constructions in Modern Greek: Implications for the archi-<br />
tecture <strong>of</strong> Grammar. Studies in Constraint-based Lexicalism. CSLI Publications.<br />
Kolliakou, D. (2004). Monadic and polydefinites: their form, meaning and use. Journal <strong>of</strong><br />
Linguistics 40 (2), 263–323.<br />
Lasnik, H. and T. Stowell (1991). Weakest Crossover. Linguistic Inquiry 22 (4), 687–720.<br />
Lekakou, M. and K. Szendröi (2010). Eliding the noun in close apposition, or Greek poly-<br />
definites revisited. ms, Meertens Instituut, Amsterdam and University College London.<br />
Longobardi, G. (1986). L’estrazione dalle ”isole” e lo scope dei sintagmi quantificati. In<br />
Parallela 2:Aspetti della sintassi dell’italiano contemporaneo. Tübingen:Gunter Narr.<br />
Longobardi, G. (1994). The structure <strong>of</strong> dps: Some principles, parameters and problems.<br />
Linguistic Inquiry 25, 609–665.<br />
53
Marinis, T. (2003). The acquisition <strong>of</strong> the DP in Modern Greek, Volume 31 <strong>of</strong> Language<br />
Acquisition and Language Disorders. John Benjamins.<br />
Mathieu, E. and I. Sitaridou (2002). Split wh-constructions in Classical and Modern Greek.<br />
In Linguistics in Potsdam, Number 19, pp. 143–182.<br />
McCloskey, J. (1990). Resumptive pronouns, ā-binding and levels <strong>of</strong> representation in Irish.<br />
In R. Hendrick (Ed.), Syntax <strong>of</strong> the modern Celtic languages, Volume 23 <strong>of</strong> Syntax and<br />
Semantics, pp. 199–248. New York and San Diego: Academic Press.<br />
McCloskey, J. (2002). Resumption, Successive Cyclicity, and the Locality <strong>of</strong> Operations.<br />
In S. D. Epstein and D. T. Seely (Eds.), Derivation and Explanation in the minimalist<br />
program, pp. 184–226. Oxford: Blackwell.<br />
Panagiotidis, P. (2002). Pronouns, Clitics and Empty Nouns: ‘Pronominality’ and licensing<br />
in syntax. John Benjamins Publications.<br />
Panagiotidis, P. (2003). Empty nouns. Natural Language and Linguistic Theory 21, 381–<br />
432.<br />
Philippaki-Warburton, I. (1985). Word order in Modern Greek. Transactions <strong>of</strong> the Philo-<br />
logical Society (83), 113–143.<br />
Reinhart, T. (1982). Pragmatics and Linguistics: an Analysis <strong>of</strong> Sentence Topics. Indiana<br />
University Linguistics Club.<br />
Rizzi, L. (1997). The Fine Structure <strong>of</strong> the Left Periphery. In L. Haegeman (Ed.), Elements<br />
<strong>of</strong> Grammar: Handbook <strong>of</strong> Generative Syntax, pp. 281–337. Dordrecht:Kluwer.<br />
Roussou, A. and I.-M. Tsimpli (1994). On the interaction <strong>of</strong> case and definiteness in Modern<br />
Greek. In I. Philippaki-Warburton, K. Nicolaidis, and M. Sifianou (Eds.), Themes in<br />
Greek Linguistics., pp. 69–76. Amsterdam: John Benjamins.<br />
Roussou, A. and I.-M. Tsimpli (2006). On Greek VSO again! Journal <strong>of</strong> Linguistics 42,<br />
317–354.<br />
Rouveret, A. (2002). How are resumptive pronouns linked to the periphery? In Linguistic<br />
Variation Yearbook, pp. 123–184. John Benjamins.<br />
54
Sells, P. (1984). Syntax and Semantics <strong>of</strong> Resumptive Pronouns. Ph. D. thesis, university<br />
<strong>of</strong> Massachusetts at Amherst.<br />
Shlonsky, U. (1992). Resumptive Pronouns as a Last Resort. Linguistic Inquiry 23 (3),<br />
443–448.<br />
Sioupi, A. (2001). On the semantic nature <strong>of</strong> Bare Singular nps in Greek. In C. Clairis<br />
(Ed.), Recherches en Linguistique Grecque II. Actes du 5e Colloque International de<br />
linguistique Grecque, Volume 2, pp. 231–234. Sorbonne: Paris:Harmattan.<br />
Stavrou, M. (1991). Nominal apposition: more evidence for a DP analysis <strong>of</strong> NP. In J. Payne<br />
(Ed.), Empirical approaches to Language Typology. Berlin: Mouton de Gruyter.<br />
Stavrou, M. (1996). Adjectives in Modern Greek: an instance <strong>of</strong> predication or an old issue<br />
revisited. Journal <strong>of</strong> Linguistics 32, 79–112.<br />
Tomioka, S. (2003). The semantics <strong>of</strong> null arguments in japanese and its cross-linguistic<br />
implications. In K. Schwabe and S. Winkler (Eds.), Interfaces, pp. 321–339. Amsterdam<br />
and Philadelphia: John Benjamins Publications.<br />
Tsimpli, I.-M. (1999). Null operators, clitics and identification: a comparison between Greek<br />
and English. In A. Alexiadou, G. Horrocks, and M. Stavrou (Eds.), Studies in Greek<br />
syntax, Volume 43 <strong>of</strong> Studies in Natural Language and Linguistic Theory, pp. 241–262.<br />
Kluwer Academic Publishers.<br />
Tsimpli, I.-M. and D. Papadopoulou (2005). Aspect and argument realisation: A study on<br />
antecedentless null objects in greek. Lingua 116, 1595–1615.<br />
Tsimpli, M. I. (1995). Focusing in Modern Greek. In K. Kiss (Ed.), Discourse Configura-<br />
tional Languages, Oxford Studies in Comparative Syntax, pp. 176–206. Oxford Univer-<br />
sity Press.<br />
Tsoulas, G. (2008). On the grammar <strong>of</strong> number and mass terms in greek. In E. Anagnos-<br />
topoulou and S. Iatridou (Eds.), Proceedings <strong>of</strong> the 2007 Workshop in Greek Syntax and<br />
Semantics at MIT, Volume 56, pp. 131–46. MIT Working Papers in Linguistics.<br />
von Fintel, K. and I. Heim (2009). Intensional semantics. Lecture Notes.<br />
55
Zimmermann, T. E. (1993). On the proper treatment <strong>of</strong> opacity in certain verbs. Natural<br />
Language Semantics 1 (2), 149–179.<br />
56