The Landscape of Greek Quantifiers

Abstract

After presenting some basic genetic, historical and typological information about Greek this chapter outlines the quantification patterns it expresses. It illustrates various semantic types of quantifiers, such as generalized existential, generalized universal, proportional, definite and partitive which are defined in the Quantifier Questionnaire in Chapter 1. It partitions the expression of the semantic types into morpho-syntactic classes: Adverbial type quantifiers and Nominal (or Determiner) type quantifiers. For the various semantic and morpho-syntactic types of quantifiers it also distinguishes syntactically simple and syntactically complex quantifiers, as well as issues of distributivity and scope interaction, classifiers and measure expressions, and existential constructions. The chapter describes structural properties of determiners and quantified noun phrases in Greek, both in terms of internal structure (morphological or syntactic) and distribution.

6.1 Introduction

In this article, we study the structures that the Greek language employs to express quantification. By Greek, I am referring to the contemporary Greek spoken in the countries of Greece and Cyprus (an estimated total of 14 million speakers), and Greeks in diaspora (an estimated 5–6 million). It has long been customary, especially in the study of classics, to use the term ‘Greek’ to refer to the ancient language – and for a while, linguists referred to the modern language as ‘Modern Greek’, or Koine Modern Greek (Κοινή Νεοελληνική; Babiniotis and Kontos 1967). However, ‘as a living language, contemporary Greek does not need to be qualified by an adjective which implies that it is somehow secondary to the ancient language’ (Holton et al. 1997: xiii). For this reason, it gradually became standard practice in linguistics to use Greek to refer to the modern language, adding the adjective ancient or modern only when these chronological stages need to be distinguished.

Greek is an Indo-European language, the sole descendant of Ancient Greek. Ancient Greek exhibited variation in its dialects – which, however, were always mutually intelligible and in later stages (e.g. in later antiquity and the Hellenistic period) developed into a common language koine (see among others Horrocks (1997)). It is now the standard view that ‘the vast majority of Greek speakers now speak a common language with only relatively minor dialectal variations. The only exception to this is the Greek Cypriots, many of whom ordinarily speak a dialect which, although linguistically close to standard Greek, presents some significant differences’ (Holton et al. 1997: xiii).

Until 1976, two versions of Greek co-existed: demotic (δημοτική), which was the actual spoken language at least since the turn of the twentieth century; and katharévousa (καθαρεύουσα), a hybrid made up of lexical, morphological, and syntactic features of Ancient and Modern Greek. ‘Katharévousa was used not only on most official occasions, but it was also the language of secondary and college education, the law, medicine, the church, armed forces, most newspapers, and even to a certain extent radio and TV broadcasting’ (Holton et al. 1997: xv). The title of the most authoritative earlier grammar of Greek – Νεοελληνική Γραμματική (της Δημοτικής) [Modern Greek Grammar (of Demotic), Athens 1941] – reflects precisely this context. Demotic became the official language in 1976, and since then, the Greek language ‘has come closer to developing a set of universally accepted norms than at any other stage in its history’ (Holton et al. 1997: xv). The grammar I will be using as reference in this chapter is Holton, Mackridge, and Philippaki-Warburton (1997), which describes what can be thought of as standard modern Greek, spoken at urban centers in Greece and Cyprus, which is based on demotic vocabulary, morphology and syntax, but does display a significant influence from katharévousa; for additional description, and more details in the history of Greek, see also the important works of Mackridge (1985) and Horrocks (1997).

Greek is a highly inflected language. The nominal system displays four cases (nominative, genitive, accusative, vocative), and there is agreement within the nominal, so all constituents are typically marked for case, number and gender. The verbal system is inflected for voice (active, medio-passive), tense (past, non-past), aspect (perfective-imperfective), and person, so verbal forms can be quite complex. We will not emphasize these morphological matters in this article, and recall them only when necessary. First, I briefly consider some basic facts about clause structure (6.1.1), and then I give some necessary background information about the DP structure (6.1.2). D plays an important role in the formation of quantifiers in Greek, as we will see.

6.1.1 Basic Facts About Greek Clause Structure

Alexiadou and Anagnostopoulou (1998) and others have cited Greek as underlyingly VSO, but I think that the most defining feature of Greek is that there is extensive word order freedom. In practice, the subject dominantly occurs sentence initially in affirmative declarative sentences, but the flexibility in word order allows constituent displacements such as topicalizations, focus, and wh-movement. Another distinctive property of modern Greek is that it does not have an infinitive, and therefore complementation is always finite.

We have three mood paradigms: subjunctive, indicative, and imperative. The imperative is used in main contexts only, and is marked with specific morphology on the verb (Mackridge 1985, Holton et al. 1997).¹

(1)	Pés	to.
	say.imperative.2sg	it
	Say it.

For the imperative, a special verb suffix is employed (-s in (1)), and a pattern of enclisis arises. In the indicative and subjunctive, mood marking does not happen with verbal inflection (as was the case in ancient Greek), but with sentential particles: the complementizers oti and pu mark the indicative in embedded clauses, but nothing special is used for indicative in main clauses. The subjunctive is indicated with the particle na. As a particle, na does not inflect and can be used in embedded as well as main clauses, preceding the inflected verb and clitic pronouns:

(2)	Na	to pis.
	subj	it say.perfective.nonpast.2sg
	You may say it.

These main subjunctives are used as requests, wishes, desires or orders, invitations. Na, in embedded clauses, is the typical subordinator after nonveridical verbs of volition, permissives, and the like – whereas indicative oti, pu follow veridical verbs (see Giannakidou (1998, 1999, 2009, 2010) for extensive description of mood choice in Greek based on the notion of non-veridicality). The verbal form employed with na in (2) is in the perfective nonpast (PNP), as indicated in the gloss, and cannot occur without na or the optative particle as:

(3)

* To pis.

(perfective nonpast: * on its own)

Holton et al. characterize this form as dependent, and besides na and as, it is licensed also after tha (future; Tsangalidis 1998), the conditional an, and other nonveridical and future oriented connectives such as prin ‘before' (Giannakidou and Zwarts (1999), Giannakidou (1998, 2009)).

(4)	{Tha/an}		to pis.
	Tha/if		it say.PNP.2sg
	You will say it./ If you say it.
(5)	Prin	to	pis,… .
	before	it	say.PNP.2sg

For a recent syntactic discussion of na, and survey of the literature, see Roussou (2000). Giannakidou (2009, 2010) argues that the verbal dependent – the PNP – is not a real present tense, but rather it is a temporally deficient form that needs the particles to supply a temporal anchor. The particles, including the subjunctive na, function as the present tense: they introduce the variable now in the syntax. Na is generated as a Mood head (Philippaki-Warburton 1993).

Greek diachronically possess negations that are heads (Ancient Greek ου, μην, Modern Greek dhen, min (δεν, μην). The modern Greek negations head their own projections NegP (Giannakidou (1998), see also Veloudis (1982)); but the Ancient Greek negators are argued to be phrasal (Chatzopoulou 2011). Dhen is used to negate indicative clauses, and min negates subjunctive clauses and gerunds. The correlation between negation and mood has been diachronically stable in the history of Greek, though not perfect (see Chatzopoulou forthcoming).

(6)	Na	min	to pis.
	subj	not	it say.1sg
	Don't say this.
(7)	Dhen	to	ipa.
	not	it	said.1sg
	I did not say this.

Now let’s look at the basic patterns of the Greek definite structures.

6.1.2 The D in Greek: Uses, Differences with English, and Genericity

Greek has a DP (Stavrou 1983, Stavrou and Horrocks 1989, Horrocks and Stavrou 1987), headed by the definite article. Like the noun and adjective, the article in Greek is fully inflected for gender, case, and number: o is masculine, i feminine and to neuter (in singular nominative). I will be using o in this chapter as the label for the definite article. The definite article is usually designated as D (Abney (1987); see Alexiadou et al. (2008) for a recent overview), and the demonstrative is generated in English also as D (thus *this the book). The English DP has the structure below; it produces typically a referential expression, a (maximal or unique) individual indicated with iota:

Demonstratives are generally thought of as definites that come with additional presuppositions of maximal salience or proximity (see Roberts (2002) for a comparison between definite descriptions and demonstratives in English). The DP produces the most basic argument e – which can be lifted up to the GQ type when necessary (modulo Partee’s 1987 type shifting rules).

6.1.2.1 The Greek Definite Article

The Greek article o is a D too, but it has a number of additional uses that are not observed in English, and which make it quite interesting.

6.1.2.1.1 (i) Definite Serializations

The article is used multiply in the so-called definite reduplication, or polydefinite structure (see Kolliakou 2004, Alexiadou and Wilder 1998, Campos and Stavrou 2004, Ioannidou and den Dikken 2009, Lekakou and Szendroi 2009).

(9)	a.	o		kókinos	o	tíxos
		the.nom.sg		red.nom.sg	the.nom.sg	wall.nom.sg
		the wall that is red
	b.	o	tíxos	o	kókinos
		the	wall	the	red

These serial [DP plus DP] structures are extremely common and productive in Greek. Often, they are thought to express a predication relation between the two DPs, as indicated above (the wall that is red), though the exact details are not crucial here. It is, however, important to note the possible permutation of noun and adjective. Also, there is no limit in how many DPs can be serialized. Consider the example below (ignoring agreement in the gloss):

(10)	to	palió	to	spíti	to	megálo	to	patrikó
	the	old	the	house	the	big	the	paternal
	the big old family house

Possessive pronouns (mas ‘ours’ below) can be added at any point:

(11)	a.	to	spíti		mas
		the	house		ours
	b.	to	palió	mas	to	spíti	to	megálo	to	patrikó
		the	old	ours	the	house	the	big	the	paternal
	c.	to palió to spíti mas						to megálo		to patrikó
	d.	to palió to spíti				to megálo mas			to patrikó
	e.	to palió to spíti				to megálo			to patrikó mas
		our big old family house

We see here that the possessive also appears as a definite description (with overt definite article) in Greek – again a major difference with English where the possessive does not, and cannot, contain the definite article.

6.1.2.1.2 (ii) Definite Article with Quantifiers

Another use of the Greek article which differs from English is when it appears to attach not to an NP, as is expected, but to a quantificational determiner (Giannakidou 2004). This is illustrated with the universal quantifier káthe ‘every’:

(12)	a. o		káthe fititís		(Giannakidou 2004: (32b))
	D.masc		every student.masc
	* the every student
	b. *	káthe	o	fititís
		every	D	student
(13)	i	káthe	fitítria
	D.fem.	every	student.fem

Giannakidou (2004) and Etxeberria and Giannakidou (2010) gloss D káthe lit. ‘the every’, as each. This use of D is observed in other languages too, e.g. Basque (Etxeberria 2005, 2009, Chapter 3, this volume) and Hungarian (Szabolcsi 1987; see also Szabolcsi 2010). The works cited propose that the article in this use modifies syntactically the quantificational determiner and not the NP. We come back to these uses of D when we discuss universal quantifiers later on. We will also find the definite article to interact with wh-quantifiers in Section 6.7, more specifically in the formation of free relatives and free choice items.

6.1.2.1.3 (iii) Definite Article with Proper Names

The Greek definite article is obligatory with proper names:

(14)	a.	o	Nikólas;	i	Ariádne
		the	Nicholas;	the	Ariadne
	b.	* Nikólas; *Ariádne

The article is dropped only with the vocative (Stavrou 2011): Nikóla! Ariádne!; but *o Nikóla! *i Ariádne! Otherwise, the Greek proper name looks like a definite description too.

6.1.2.1.4 (iv) Generic Reference in Greek Is Only Possible with the Definite Article

The Greek DP is the typical vehicle of genericity. Bare singular count nominals are not allowed in the language as arguments:

(15)	a.	*(I)	patáta	íne	laxanikó.
		the	potato	is	vegetable
		{Potato/the potato} is a vegetable.
	b.	*(I)	patátes		íne	laxaniká.
		the	potatoes		are	vegetables
(16)	a.	* (I)	patáta	itan	sápia.		(yesterday)
		the	potato	was	rotten
	b.	*Patátes		itan	sápies.
		potatoes		were	rotten

The first sentence is generic, and the second is episodic, as indicated and suggested by the predicates. We see that bare singular and plural count nouns are excluded in both cases. Generic reference is done via the definite determiner, in singular and plural. Compare the plural version to The potatoes are vegetables in English, which has a multiple kind reading (Krifka et al. 1995, Chierchia 1998). In Greek this sentence also has the kind denoting bare plural reading that the English definite plural lacks. Even singular mass nouns, which in English can be bare, cannnot appear bare as generic arguments:

(17)	* (I)	záxari	íne	glikiá.
	the	sugar	is	sweet
	Sugar is sweet.

So, Greek is very restricted in its use of bare nominals. Bare singulars are allowed only as predicate nominals as we see, in the existential structure (to be examined in Section 6.4), and in the object position, where it is has been argued that they contain a null D (Sioupi 1998, 2002, following Longobardi 1994, Chierchia 1998).

(18)	a.	Xriázome záxari.
		I need sugar.
	b.	O Jánis aftí ti stigmí diavázi {efimerída/periodiká}.
		the John this the moment read.imperf.3sg newspaper/magazínes
		John is reading {the newspaper/magazínes} right now.
	c.	O	Jánis	éxtise	{spíti/spítia}.
		the	John	built	house/houses
		John built a house.
	d.	I	mamá	éftiakse	{kéik/pítes}.
		the	mom	made	cake/pies
		Mom made {a cake/pies}.

The bare arguments in object position are all narrow scope indefinites, equivalent to existential bare plurals in English, and the singulars to a indefinites (see Sioupi’s work for more details). Bare singulars are also employed in minimizer negative polarity items (e.g. didn’t say a word), as we see in Section 6.6.

In subject position, bare singular existentials are out, but bare plurals are marginally allowed with existential, never generic, readings:

(19)	Gátes niaourízoun.
	Cats are meowing.

The point I want to make here is that bare arguments, to the extent that they are allowed, are equivalent to narrow scope indefinite existentials, and are never used generically. Generic reference in Greek is always via DP, regardless of mass/count differences.

6.1.2.2 Demonstratives

Finally, Greek possesses two demonstrative pronouns aftós, aftí, aftó ‘this’, ekínos, ekíni, ekíno ‘that’ – which, unlike English, must embed DPs (Stavrou 1983, Stavrou and Horrocks 1989, Alexiadou et al. 2008):

(20)	a. aftós	*(o)	fititís
	this	the	student
	this student
	b. ekínos		*(o)	fititís
	that		the	student
	that student

Horrocks and Stavrou argue that the demonstratives are not D heads in Greek, but phrases in Spec, DP. Other demonstratives in Greek are: the qualitative demonstrative tétjos ‘such’, and the quantitative tósos ‘that much’ (the latter related to the wh-word ósos ‘as much as’):

(21)	a.	Thélo	éna	tétjo. (with a pointing gesture; Holton et al.: 327)
		want.1sg	one	such
		I want one of these.
	b.	Dhen thélume tétja.
		We don’t want such things.
	c.	Íne tóso psilós!
		He is so tall!

Finally, anaphoric elements also appear as DPs in Greek: the word ‘self ’ – o eaftós mu ‘myself ’ – and o idhios lit. ‘the same one’, a long distance anaphor and a logophor in Greek (Iatridou 1986, Varlokosta and Hornstein 1993). This background on the Greek DP will suffice for our discussion of quantificational expressions. More details regarding the use of D will be pointed out as we move on.

6.1.3 Roadmap

Traditional grammars use the terms ‘pronouns’ as in definite (he, she, it) and indefinite (someone, something) pronouns, determiners, and quantifers to refer to what can collectively be understood as ‘quantificational expressions’. In this context, the word ‘determiner’ is understood descriptively as ‘a word that is not an adjective or a numeral but which accompanies a noun (e.g. “every”, “other”, “same”)’ (Holton et al. 1997: 303), hence quite differently from the way the term is used in the theoretical discussion in the syntax-semantics interface.

The background of our discussion here will be the generalized quantifier (GQ) theory (Montague 1974, Barwise and Cooper 1981, Zwarts 1986, Westerståhl 1985, Partee 1987, Keenan 1987, 1996, Keenan and Westerståhl 1997; for more recent works see Giannakidou and Rathert 2009, Szabolcsi 2010), which posits that there is a natural class of expressions in language, called quantificational determiners (designated as Qs), which combine with a nominal (NP) constituent (of type et, a first order predicate) to form a quantificational argumental nominal (QP). This QP denotes a GQ, a set of sets. In a language like English, the syntax of a QP like every woman is as follows:

The quantificational determiner Q every combines first with the NP argument woman, and this is what we have come to think of as the standard QP-internal syntax. The NP argument gives the domain of the Q, and the Q expresses a relation between this domain and the set denoted by the VP. QPs like every woman, most women, etc. are known as ‘strong’ (Milsark 1977), and they contrast with the weak Qs like some, few, three, many, etc., in that the the latter, but not the fomer, are admitted in the existential construction. Another element that combines with a domain set to give a nominal argument is the definite D, as we saw earlier. In Greek, like in English, the DP and the QP are the two argumental nominals – bare nominals are generally not allowed as arguments, as we saw, or if they do, they are thought to contain a null D.

The structure of this chapter will unfold as follows. We start first with existential QPs in Section 6.2. We present first the quantity denoting existentials such as numerals – including modified numerals (Section 6.2.2) and distributive numerals, and we also examine the indefinite QPs preceded by the Greek equivalent to some and those preceded by the indefinite article (Section 6.2.4). Then, I present the so-called value judgements existentials, i.e. those that express a subjective assessment of their quantity (equivalents to few, many, several, etc.). We will notice an interaction there between intonation and determiner, a pattern that we observe again later in our discussion of scope and negative polarity quantifiers in Section 6.6. We discuss also partitive structures, and in Section 6.2.7, the adverbial variants of existentials.

In Section 6.3, we move on to expressions of universal quantification and other strong quantifiers, where we observe the systematic interaction between D and Q mentioned earlier. Here we also discuss binominal each, floating quantifiers, and distributivity. In Section 6.4, we zoom in on the existential structure, and ask what kinds of quantifiers can appear there. It is hard to draw clear conclusions about the definiteness effect in Greek; also there is more than one variant of the existential structure in Greek. In Section 6.5, we discuss morphologically complex quantifiers such as comparative quantifiers, those created via boolean compounding (and, or, neither...nor..., and not), exception phrases (all but ten students), and bounding phrases (He exercised twice a day, six days a week for one year).

In Section 6.6, we discuss negative polarity quantifiers and negative concord in Greek, and consider some more general questions of scope in a bit more detail. We notice an interaction between scope and intonation in Greek that has been observed in the literature (Giannakidou 1998, 2000, Baltazani 2002). In Section 6.7, finally, we focus on wh-based quantification. Unlike English, there are three paradigms of wh-words in Greek: interrogative wh-words, relative wh-words, and a special wh-form for free relatives that employs the definite article. The form is also the one used as the basis for the formation of free choice quantifiers (Giannakidou and Cheng 2006), so we find again an interaction of D with quantifiers in free choice, suggesting the relevance of definiteness for the semantics of free choice.

We distinguish between D-quantifiers, i.e. those that we call QPs (formed by using the determiner Q), and A-quantifiers which are adverbial. The latter are mathematically less well understood, and morphosyntactically and semantically more variable than D-quantifiers. Finally, it is important to emphasize that, as just described, we take the basic semantic type of quantifiers to be a relation between two sets. Our classification is thus meaning based. Logically equivalent expressions in different languages may be syntactically non-isomorphic: e.g. each student in Greek appears as o káthe fititís, i.e. it is as a definite as mentioned earlier, but it will be classified as a universal based on its meaning.

My goal is to offer an accurate description of the Greek quantificational system, and it is my hope that this article will provide useful information to those interested in knowing what the landscape of Greek quantifiers looks like. The emphasis is therefore on broad empirical coverage and accuracy. However, connections to current theoretical discussions will also be made when they help the description – and, most importantly, when the lessons we draw from Greek can have implications for the analysis of quantification in general.

6.2 Expressions of Existential Quantification

We start with the examination of generalized existential (intersective) quantifiers (Keenan 1987, 1996). This is the class known as weak quantifiers, the Q expressing the intersection of their domain argument (NP) and the VP.

Often, existential Qs have been treated in the literature as ‘adjectival’, and therefore are not always considered syntactically Qs of type et,ett (cf. Link 1984, Partee 1988, Kamp and Reyle 1993, Krifka 1999, van Geenhoven 1998, Landman 2002). Ionin and Matushansky (2006) more recently argue that weak numerals, at least, are modifiers. Greek weak Qs are also argued to be adjectival as a class in Giannakidou and Merchant (1997), Stavrou and Terzi (2010). In what follows, I will generally refrain from syntactic questions, and consider primarily the semantic classification. So, what are called existential quantifiers below are simply relational expressions that are used in Greek to express existential quantification, regardless of whether they are syntactically quantificational determiners or not.

6.2.1 Indefinite Article and Numerals

A numeral is a word that expresses a number. Numerals are typically divided into cardinals (one, two, three) and ordinals (first, second, etc.). Ordinals in Greek behave like predicative adjectives and will not be considered here. Holton et al. state that ‘from the morphological point of view, Greek cardinal numerals may be divided into three categories: (a) indeclinable cardinals, (b) declinable cardinals, and (c) cardinals behaving like nouns’ (Holton et al. 1997: 294). Examples of declinable numerals are énas (masc.) mía (fem.) éna (neut.) ‘one’, trís (masc., fem.) tría (neut.) ‘three’, tésseris (masc., fem.) téssera (neut.) ‘four’, diakósi diakósies diakósia ‘two hundred’, xílji xíljes xílja ‘one thousand’. Ekatomírio ‘million’ behaves like a noun, and thus also declines (like all nouns in Greek). Indeclinable are the words designating the numbers 2, 5, 6, 7, 8, 9, 10, 11, 12 and the tens. Some examples are given below:

(23)	I María agórase ÉNA vivlío,	ke óxi pénde.
	Mary bought one book,	and not five [books]

(24)	a.	Tris ánthropi diamartiríthikan.
		Three people complaíned.
	b.	To tmíma mas tha dextí fétos xílius
		the department ours will. admit this year thousand.masc.acc.pl.
		diakósius néus fitités.
		two-hundred.masc.acc.pl new. masc.acc.pl students. masc.acc.pl
	c.	Ekremún apózimiósis enós ekatomiríu agrotón.
		Pend.3pl compensations.nom one.gen million.gen farmers.gen
		The compensations of one million farmers are still pending.
	d.	i xóra ton xilíon limnón
		the country the.gen.pl thousand.gen.pl lake.gen.pl
		the country of a thousand lakes

The bracketed part in the example (23) illustrates NP ellipsis which is generally available in Greek (Giannakidou and Stavrou 1999), and depends on contrastive focus. In the example, the numeral in stressed for this reason. Unstressed, the numeral is used as the indefinite article:

(25)	Skéftome na agoráso éna spíti.
	I am thinking of buying a house.

The use of numeral one as an indefinite article is very common in many European languages, and in Greek, the indefinite éna is a run of the mill unmarked indefinite with no preference for specific or non-specific readings (Giannakidou et al. 2011). For indefiniteness in the plural, the bare plural can be used, as indicated in the example below:

(26)	I	Maria	agorase	vivlia.
	the	Maria	bought	books
	Maria bought books.

As said earlier, the indefinite bare plural is always narrow scope and cannot be specific (unlike the singular which is neutral). For indefinite plural Greek also employs káti, which we will discuss later in this section, and which seems to be comparable to the use of unos in Spanish.

Greek numerals are also known to license null arguments:

(27)	I Éléna agórase tría vivlía, alá I María dhen agórase [e].
	Eléna bought three books, but María didn’t buy [any].

Giannakidou and Merchant (1997) call this ‘indefinite object drop’, and show that only indefinite existential quantifiers can serve as antecedents for indefinite object drop in Greek.²

Numerals can also be used in the so-called pseudopartitive structure (Stavrou 1983, 2003) which seems to be equivalent to a classifier structure:

(28)	a.	Xriazómaste		tría	bukália	krasí.
		need.1pl		three	bottles.acc	wíne.acc
		We need three bottles of wíne.
	b.	Dío	potíria		ximós	íne	arketá.
		two	glasses.nom		juice.nom	is	enough
		Two glasses of juice is enough.
	c.	tría métra ífasma
		three meters cloth

Like English, Greek is not a classifier language and uses containers and measure phrases to count units of mass nouns. We see here that no preposition is used – hence, pseudopartitive – but the two nominals agree in case (though not number, as the mass noun appears typically in the singular), and the case is determined by their grammatical function (object or subject).

6.2.2 Modified Numerals

Numerals can be modified by the following kinds of modifiers:

6.2.2.1 Quantity Bounding Modifiers

Quantity bounding modifiers are: tuláxiston ‘at least’, to polí ‘at most’, óxi parapáno apó ‘no more than’, akrivós ‘exactly’:

(29)	Írthan {tuláxiston/to polí/ óxi parapáno apó} diakósi fitités.
	came.3pl at least/at most/ no more than two hundred students
	{At least/ at most/no more than} two hundered students came.
(30)	To cake xriázete (akrivós) diakósia (akrivós) grammária vútiro (akrivós).
	The cake needs (exactly) two hundred (exactly) grams butter (exactly).

We see here that the modifier akrivós ‘exactly’ can float, and appear at the right or the left edge of the QP. Tuláxiston and to polí (lit. ‘the much’) typically precede the numeral but can also appear to the right: tría avgá to polí ‘three eggs at most’, tría avgá tuláxiston ‘three eggs at least’ but *tría avgá óxi parapáno apó ‘*three eggs no more than’ – notice the parallel with English. Importantly, the modifier can also ‘split’ the QP and appear to the right of the numeral, between the number word and the noun:

(31)	a.	Evgala	dhío	akrivós	fotografíes.
		took	two	exactly	pictures
		Lit. I took two exactly pictures.
	b.	Na vgális	dhío	{to polí/tuláxiston}		fotografíes
		Take	two	{at most/at least}		pictures.

As I mentioned at the beginning, Greek has great flexibility in word order, and this carries over to the QP internal structure. This flexibility in the positioning of modifiers suggests that they don’t just function as Q modifiers, but they may have flexible syntactic specification as Q or QP modifiers, something which is expected given that they are adverbial. Holton et al. (1997) discuss some of these modifiers as ‘adverbials within the noun phrase’ (1997: 337), along with the approximative and evaluative modifiers that we discuss next.

Another bounding modifier is móno(n) ‘only’. (N is added before a vowel for euphonic reasons only.) Móno shows exactly the same flexibility:

(32)	Evgala	(móno)	tris	(móno)	fotografíes	(móno).
	took.1sg	(only)	three	(only)	pictures	(only)
	I took (only) three (only) pictures (only).

Tuláxiston, to polí and mónon do not exclusively modify numerals, they can also modify e.g. proper nouns:

(33)	Tha milíso	{tuláxiston/ móno/ to polí} me ton Jáni.
	I will talk	{at least/ only/ at most} with John.

6.2.3 Approximative Modifiers

Typical approximative modifiers are perípu ‘around’, sxedhón ‘almost’:

(34)	Simetíxan		stis	diadilósis	{perípu/sxedhón}
	participated.3pl		in-the	demonstrations	approximately/almost
	tris	xiliades	fitités.
	three	thousand	students
{Approximately/ almost} three thousand students participate at the demonstrations.

Like the bounding quantifiers, perípu and sxedhón may also appear at the right edge of the QP:

(35)	Simetíxan stis diadilósis tris xiliádes fitités {perípu/ sxedhón}.
	Three thousand students approximately participated at the demonstrations.
	*[Three thousand students almost] participated at the demonstrations.

Notice the contrast with English almost that cannot be parsed as a constituent with the QP in this position. The intermediate position is also available: tris xiliádes {perípu/ sxedhón} fitités ‘three thousand {approximately/almost} students’.

Another class of approximative quantifiers is kamiá and kána. These sound like variants of perípu, but are morphologically related to the NPI kamía ‘NPI.any.fem.’ that we will discuss in Section 6.6. Kamiá is the feminine form, and kána is related to the masculine and neuter kanéna. As approximatives, kamiá and kána are used uninflected. Kamiá appears with a numeral that does not agree in gender/number, or with nouns ending in –ariá, which are classifying:

(36)	Tha prépi na íxes		{kamiá/kána}	déka tilefonímata	(oso elipes).
	you must have had		kamiá/ kána	ten phone calls.neuter	(while you were gone).
	You must have had about 10 phone calls while you were gone.
(37)	a.	Idha {kamiá/*kána} dekariá fitités.
		I saw about ten students.
	b.	Diávase	{kána/* kamiá} vivlío.
		Read.imperative.2sg	some book or other.

Dekariá is a classifying noun like ‘dozen’; - ariá and - ádha are very productive suffixes that create such classifying nouns: eksádha ‘six-piece’, ekatodádha ‘a mass of hundred’, penindariá ‘a mass of fifty’ (for a recent discussion see Stavrou and Terzi (2008, 2010)). The kamiá is not an NPI – given that it can be used in a positive veridical sentence in the past tense (37a). As for kána, we see that it cannot be used with the classifying nouns, but rather with a bare NP, or with the numeral ‘two’ (dhío), and with the bare noun, thus creating an indefinite noun phrase.

(38)	Tha agoráso {kána dio vivlía/ kána vivlío}.
	I will buy about two books/ a book.

Importantly, both kamiá and kána are not polarity sensitive, unlike their cognate kanénas, since they can be used in the veridical context of the simple past.

Numerals and modified numerals can all be used in the existential structure in Greek, which we consider separately in Section 6.3.

6.2.4 Indefinite QPs and Epistemic Judgement

An indefinite QP with the article éna can have specific or non-specific usages, as said earlier. Roughly, specificity means that the speaker has a particular individual in mind (in the ‘epistemic’ approach to specificity; Groenendijk and Stokhof (1981), Farkas (2002), Ionin (2006); for the choice function analysis see Reinhart (1997), Winter (1997). Specific indefinites refer to objects that are speaker identifiable but not part of the common ground; definiteness, on the other hand has to do with speaker and hearer reference, part of the common ground. Ionin (2006) argues that the QP is associated with a felicity condition that requires that the speaker be in position to identify the referent. This felicity condition is distinct from the presupposition of existence that a definite DP carries. It is clear then, that the specific use of an indefinite reflects an epistemic judgement on the part of the speaker. Such judgement often gets realized in the use of so-called ‘specificity’ markers such as certain in English.

Indefinite NPs sensitive to judgement or knowledge of the speaker exist in various languages, as noted by Haspelmath (1997) – and there exists a class of indefinites that appear to be the opposite of specific: they express uncertainty or indifference on the part of the speaker. They can only be used when the speaker does not know what their referent is. Examples of such indefinites are French un quelconque (Jayez and Tovena 2006), and Spanish singular algún. Jayez and Tovena call them epistemic, Alonso-Ovalle and Menéndez-Benito (2010) call them modal, but I will use the term referential vagueness (from Giannakidou and Quer 2011). Referential vagueness is an anti-specificity condition which says that the QP will be felicitous only if the speaker does not have a fixed value in mind.

Greek has two referentially vague determiners: the negative polarity kanenas series that we discuss in Section 6.6, and the non-polarity determiner that translates in English as some: kápjios, kápja, kápjo ‘some, someone, somebody’ – inflecting fully for φ-features (case, number, gender), and which can be used both as determiner and as full QP, as indicated.

(39)	kápjios, kápja, kápjo	‘someone, some N’
	káti	‘something’
	kápu	‘somewhere’
	kápote	‘sometime, once’
	kápos	‘in some way, in a certain way’
	kámboso	‘a certain amount’

This ká-series is composed morphologically by adding ká to a wh-word (the p-part and ti; see Section 6.7). The ká-indefinites, however, do not have wh- or interrogative uses in Greek.³ Some examples are given below:

(40)	a.	Idha kápjon na trexi sta skotiná.
		I saw someone run in the dark.
	b.	Kápja nosokoma tha ton kálmári.
		Some nurse will calm him down.
	c.	Fáe káti.
		Eat something.

(An older form, katití, also exists, but its usage is in decline). The ká-indefinite is typically used when the speaker does not have a specific referent in mind, or in situations where the speaker doesn’t care about the identity of the referent; empirical evidence for this comes from two experiments (Giannakidou et al. 2011) showing that kápjos is dispreferred situations where the speaker has one particular value in mind, such as below:

(41)	Thelo na miliso me kápjon glosologo. # Ine aftos o kyrios eki.
	I want to meet some linguist of other. # It’s that guy over there.
(42)	Thelo na miliso me kápjon glosologo. # To onoma tu ine Veloudis.
	I want to meet some linguist or other. # His name is Veloudis.
(43)	Thelo na milso me kápjon kathijiti. # Ine o proedros tu tmimatos filosofias.
	I want to talk to some professor or other. He is the head of the Philosophy Department.

The unmarked indefinite has no trouble in this fixed-value context:

(44)	Thelo na miliso me énan glosologo. Ine aftos o kyrios eki.
	I want to talk to a linguist. It’s that guy over there.
(45)	Thelo na miliso me énan glosologo. To onoma tu ine Veloudis.
	I want to talk to a linguist. His name is Veloudis.
(46)	Thelo na miliso me énan kathijiti. Ine o proedros tu Glosologikou.
	I want to talk to a professor. He is the head of Linguistics.

So, epistemic judgement does constrain the distribution of the ká-indefinite, albeit not in a polarity manner. The specific use of énas simply remains unmarked.

There is also a use of énas ‘someone’ as an independent QP. The example below is from Holton et al. 1997: 320):

(47)	Irthe {énas/kápjos}	ke se zituse.
	Someone came	looking for you.

This use of énas is equivalent to someone, as we see. For arbitrary reference ‘one’, Greek employs kanís (a cognate of the polarity kanénas that we examine later, Giannakidou (1994); kanénas itself can also, more markedly, be used in this context):

(48)	Anarotiete kanís an…
	One wonders whether… .

Finally, it is worth noting the use of káti – which means literally ‘something’ – as an indefinite determiner, akin to a plural indefinite article. In this use, káti combines with a plural NP and creates a plural indefinite:

(49)	Píran	tiléfono	káti	fitités.
	called	telephone	káti	students.
	Nomízo	oti	ítan	o	Pétros	ke	i	María.
	Think.1sg	that	be.3pl	the	Peter	and	the	Mary
	Some students called. I think it was Peter and Mary.

The singular *káti fititís is impossible. As a plural indefinite, káti appears invariant (not inflected for φ-features). Languages tend to not have a morphological plural indefinite article (a notable exception is Spanish unos), and supplement it with other forms, hence the use of káti in Greek. English weak sm has a similar use as in I saw sm students. As a plural indefinite, the káti indefinite conveys complete ignorance of reference, as is shown in the example above. There are cases where káti imposes anti-specificity in the sense that it cannot refer back to a discourse given set. Consider the following scenario (modeled after Martí’s (2009) example (1)):

(50)	Context: Teachers A and B are on an excursion with [a group of children]K. Teacher A comes to teacher B running:
	a. A: Akouses?		[Káti pedhiá]J, #K	xáthikan sto dásos.
		Did you hear?	Some children	were lost in the forest
		Eftixos pu ta diká mas ta kratísame edo!
		Thank God we kept ours here!
	b. A:	Akouses?	[Kápja pedhiá]#J, K	xáthikan sto dásos.
		Did you hear?	Some children	were lost in the forest

Káti pedhiá here cannot refer back to the discourse given set of children the teachers A, B were in charge of; Spanish unos has been claimed to have the same property (Gutiérrez-Rexach (2001), and this supports further a parallel between the Greek káti NP.plural and the plural indefinite article unos in Spanish, which remains non-specific in the plural. The b example with the plural kápja needs to refer to the previously introduced set, just like Spanish plural algunos (Martí 2008, 2009). So, unlike English, Spanish and Greek employ two indefinite paradigms in the plural: [A+NP.plural], and [SOME+NP.plural] – and these come with distinct patterns of context dependence: the former isn’t context dependent, but the latter is. The mystery is that in the singular we tend to have the opposite pattern, and this somehow needs to be explained.

Finally, in support of the equivalence of káti NP.plural to a plural SOME consider the following exclamative sentence:

(51)	Exi	káti kunímata!
	She’s got	SOME moves!

This is equivalent to the emphatic use of some, and makes a rather qualitative statement. Such uses provide further support for the idea that indefinites are generally associated with epistemic judgment.

6.2.5 Value Judgement Quantifiers, and the Role of Intonation

Value judgement quantifiers are those that come with some kind of judgement on the quantity they denote. Typical such examples in English are few, many, several, etc. I give below some examples with their Greek equivalents, all inflected for case number and gender. We start with those expressing a positive judgment on the quantity:

(52)	Idhame	{polés/arketés/káboses/merikés}	teníes	fétos to kálokéri.
	saw.1pl	many/several/several/a few	movies	this summer
	We saw {many/several/a few} movies this summer.

There are variants of MANY NP like [plíthos NP.genitive], as in plíthos tenión ‘crowd movies.pl.gen.’, and ‘ápires NP’ lit. ‘infinite.pl NP.pl’, as in ápires teníes ‘tons of movies’ – both designating quantities judged as very large. Polí is an adjective – the word for many and much in Greek – and it inflects, as above, where we find it as poles ‘many.fem.pl.acc’. The uninflected variant polí ‘many.neuter.sg’ is an adverb – equivalent to English ‘very’, ‘very much’ and ‘a lot’:

(53)	To	podílato	aftó	mu	arési	polí.
	the	bicycle	this	me.gen	like.3sg	much
	I like this bicycle a lot/very much.
(54)	I María íne polí kourasméni.
	María is very tired

As the English many and very, polí is generally emphatic, but in construals like polí kurasméni ‘very tired’ polí need not bear the main stress; stress could be on the adjective. If stressed, the emphatic variant of polí, which I will designate as POLI, delivers equivalence to English ‘too’ (Giannakidou 1997, 2000:465–466). An important diagnostic employed in Giannakidou was that emphatic POLI can license NPIs such as kanénan, but the unstressed polí cannot:

(55)	a.	I María íne POLI kourasméni	ja na milísi me kanénan.
		María is too tired	to talk to anybody.
	b.	* I María íne polí kourasméni	ja na milísi me kanénan.
		* María is very tired	to talk to anybody.

Hence, intonation realizes in Greek an otherwise lexical difference in English. This we find again with QPs expressing negative judgement such as líji, and emphatic LIJI. The adjective lígos(masc.) líji(fem.) lígo(neuter) means literally ‘small, little in size or quantity’ as is Thelo líji zaxari ‘I would like {a little bit of, some} sugar’. The examples and glosses below concerning intonation are from Giannakidou (2000), and the NPI típota serves as a diagnostic:

(56)	a.	LIJI	fitités	ípan	típota.
		few	students	said.3pl	anything
		Few students said anything.
	b.	*Líji	fitités	ípan	típota.
		a few	students	said	anything
		*A few students said anything.

Non-emphatic líji carries a more neutral judgement on the quantity like a few, and does not license the NPI. But the emphatic LIJI designates a quantity judged negatively as not much, or less than expected, like few, thereby allowing the NPI. Emphatic accent thus again marks an otherwise lexical distinction in English.

Another negative judgement value quantifier is eláxisti, literally the superlative of lígos, meaning ‘very few’:

(57)		Eláxisti	fitités	ípan	típota.
		very few	students	said.3pl	anything
		Very few students said anything.

Eláxisti allows for NPIs, as we see. More on the NPI facts in Section 6.6.

6.2.6 Existential Quantifiers in the Partitive Structure

All existentials mentioned in this section occur in the partitive – which in Greek involves using the ‘light’ proposition apó ‘of’, or ‘from’, plus a plural DP as is typically the case. When the existentials are used in the partitive, they receive proportional readings, as expected:

(58)	a.	Idha	{tris/kápjus/lígous/merikús/polús}		apó	tus	fitités.
		saw.1sg	three/some/a few/several/many		of	the	students
	b.	Idha	{to polí/tuláxiston}	pénde	apó	tus	fitités.
		saw.1sg	at most/at least	five	of	the	students

6.2.7 Existential A-Quantifiers

Adverbial expressions with existential force come in two basic varieties: quantificational adverbs (Q-adverbs) along with adverbials typically denoting frequency, and iterative adverbials (I-adverbials) that denote iteration.

6.2.7.1 Q-Adverbs and Adverbials with $-Force

(59)	O	Jánis	kapnízi	{sixná/spánia/póte- póte /káthe tris ke lígo}.
	the	John	smoke.imperf.3sg	often/rarely/when-when/every three and little
	John smokes {often/rarely/every now and then/very often}.

Notice the two idiomatic expressions: (a) the reduplication póte- póte (of the wh-word meaning when) meaning ‘every now and then’ or ‘occasionally’; and káthe tris ke lígo which involves a universal quantifier and a coordinate structure. These are perceived as conventionalized expressions. Reduplication, however, often creates distributive expressions in Greek as we will see soon. As indicated, the verb with Q-adverbs appears in the imperfective, since these sentences are habitual/generic statements and involve quantification of events (Krifka et al. 1995, Giannakidou 1995, 1997, 2009 for Greek). The Greek imperfective also has progressive usages that will not be relevant here.

Another group of frequency adverbials is those that are expressed in English with the prepositions on, in, at (on Monday, in the winter, at noon), or a bare plural naming a day of the week: On Monday, Mondays, Wednesdays. In Greek, these all appear as bare accusative DPs:

(60)	O	Jánis	érxete	{tis	kyriakés/	ta	vrádia}.
	the	John	comes.imperf.3sg	the	Sundays.pl.acc/	the	evening.pl.acc}
	John comes {on Sundays/at night}.

(61)	To	xióni	péfti		ton	ximóna.
	the	snow	falls.imperf.3sg		the.sg.masc.acc	winter.masc.sg.acc
	The snow falls in the winter.
(62)	O	pyretós	anevéni		ti	níxta.
	the	fever	rises.imperf.3sg		the.sg.acc	night.fem.sg.acc.
	The fever rises at night.
(63)	Ti níxta, ta pedhiá kimúnde.
	At night, the children are asleep.

Finally, another group denoting frequency appears in the form n forés + accusative ‘three times a week/a month’, etc.:

(64)	Ton	vlépo	tris	forés	{tin	evdomáda/	to	mína}.
	Him	see.imperf.1sg	three	times	the.acc	week.acc/	the.acc.	month.acc
	I see him three times {a week/a month}.

So, we see a systematic use of DP in frequency adverbials, where English uses an indefinite QP.

6.2.7.2 Iterative Adverbials

These are: tris forés, pénde forés, etc.:

(65)	Milísame	pénde	forés	fétos	to	kálokéri.
	talked.perfective.1pl	five	times.acc	this-year the		summer
	We talked five times this summer.

The iterative sentence contains a verb in the perfective, as we see in the gloss. Also noteworthy is the deictic word fétos ‘this year’ – a constant meaning ‘the year of utterance’. Greek employs likewise pér(i)si for the year prior to the year of utterance, and tu xrónu for the year after. Crucially, these are not indexicals, unlike next year, last year in English which can be anchored to the year of the clause and not necessarily the utterance:

(66)	I María ípe to 2007 oti tha édíne eksetásis tu xrónu.
	Mary said in 2007 that she would take the exam in 2011.
	(utterance year: 2010)
	Not: Mary said in 2007 that she would take the exam in 2008.

This concludes our presentation of existential structures in Greek. We move on now to universal quantifiers.

6.3 Universal Quantifiers, Distributivity, and Interaction with D

In this section, we discuss strong quantifiers in Greek: universals, and the quantifiers equivalent to both, most. Greek has two expressions of universal quantification: ólos(masc.), óli(fem.), ólo(neut.) ‘all’, and the indeclinable káthe ‘every’ – a distributive universal, variants of which appear with the definite article, e.g. o káthe ‘each’. D is also involved in the formation of both, most. We discuss the two major groupings in turn.

6.3.1 Ólos

Ólos fully inflects for φ-features, but I will be referring to the whole paradigm as ólos following the grammars. Ólos means both ‘all’ and ‘whole’ in Greek. Like in English, it appears in the periphery of a DP, and cannot follow the D:

(67)	a.	Psífisan		óli	i	fitités.
		voted.3pl		all.pl	the.pl.	students
		All the students voted.
	b.	*i óli fitités
	c.	Éfage	óli		tin		túrta.
		Ate.3sg	all.acc.fem		the.fem.acc		cake
		He ate all the cake/the whole cake.
	d.	Émine		óli		tin		óra.
		stayed.3sg		all.acc.fem		the.fem.acc		hour
		He stayed the whole hour./ He stayed the whole time.

In the use as ‘whole’, ólos can in fact follow the D, and seems to be equivalent to the adjective olókliros (i, o) ‘whole’.

(68)	a.	{i	óli /	óli	i}	sizitisi
		the	whole/	whole	the	discussion
		the whole discussion
	b.	olókliri	i	sizitisi
		whole	the	discussion

Unlike English, ólos cannot appear with a bare NP:

(69)	Óla *(ta)	pediá	íne	xaritoména.
	all D	children	are	cute
	All children are cute.

We see that even in generic contexts, as the sentence above could be, ólos must be followed by a DP, as DPs are the typical vehicles of generic quantification in Greek. Given this requirement for a definite DP, Giannakidou (2004) I argues that ólos is not a quantificational determiner, since all Qs in Greek take NP complements. Rather, I suggested treating ólos as an adverbial, i.e. an exhaustivity modifier of the DP, as has been suggested for all in English, and similar items in other languages. More supporting evidence for a non-determiner analysis comes from the fact that ólos, like all, but unlike the determiners we saw earlier, can float:

(70)	a.	I	fitités	éfigan	óli	norís.
		The	students	left	all	early.
	b.	I fitités éfigan noris óli.
	c.	I fitités óli éfigan noris.
	d.	Ta pediá ta ída na févgun óla norís.
		The children, I saw them all leaving early.
(71)	a.	* Fitités tris éfigan norís.
		Students three left early
	b.	Fitités éfigan tris norís

We see here that ól i can appear in various places: in the periphery of the DP to the right, after the verb, at the right edge of the sentence. These are unacceptable positions for the existential quantifiers we discussed in the previous section which all appear pre-NP and seem to form a constituent with the NP – they can only be separated via split topicalization, which is marked by intonation breaks, indicated here with comma⁴:

(72)	Vivlía, agorasa polá. Periodiká, polí líga.
	Books, I bought many; magazínes, very few.

Hence, the mobility of ólos, in conjunction with its extraordinary behavior of combining with a DP allows us to think of it as an adverbial modifier of the DP, maybe a kind of exhaustivity marker or designating good fit, as suggested in Brisson (2003) for English all.

Finally, ólos, does not occur in partitives:

(73)	a.	* óli apó tus fitités
		all of the students
	b.	(o) káthe énas apó tus fitités
		each one /every one of the students
	c.	o kathénas apó tus fitités
		everyone of the students

From the universals, the variants of (o) káthe énas (discussed next) occur in the partitive. ‘Most’ is fine too – i perisoteri apó tus fitités ‘most of the students’, but ‘both’ is not so good (but still usable): ? ke i dhio apó tus fitités ‘both of the students’.

6.3.2 Káthe, kathénas: Distributivity, and D-Universals

Káthe appears to be a universal distributive quantificational determiner in Greek. Holton et al. (1997: 313) characterize it as a distributive determiner too – like every. It appears to be a singular uninflected determiner, combining only with a singular argument:

(74)	a.	káthe	fititís;	káthe	fitití;	*káthe	fitités
		every	student.sg.nom;	every	student.sg.gen;	every	student.nom.pl
	b.	káthe fitíria(fem);			káthe fitítrias(fem.gen)

So, unlike olos, the definite and indefinite article, the demonstrative, and the existential quantifiers we discussed earlier which appear to inflect fully (with the exception of a few numerals), káthe is morphologically set apart from adjectives and D in the language. But káthe can combine with énas, and in this case it inflects:

(75)	a.	kathénas;	kathenós
		every-one.nom.sg.masc	every-one.gen.sg.masc
	b.	kathemía;	kathemías
		every-one.nom.sg.fem;	every-one.gen.sg.fem
	c.	kathéna;	kathenós
		every-one.nom.sg.neut.;	every-one.gen.sg

It seems natural to think of káthe as ‘every’ and kathénas as ‘everyone’. However, there are certain facts that set kathénas apart from ‘everyone’. I discuss them in detail next.

6.3.2.1 The Presence of D and Context Dependence

The first difference is that kathénas is always construed with the definite determiner o. I give below examples in a generic and episodic context:

(76)	a.	O kathénas gnorízi oti i ji íne strogilí.						generic
		Everyone knows that the earth is round.
	b.	?? Kathénas gnorízi oti i ji íne strogilí.

(77)	a.	O	kathénas	éfere	apó	éna	vivlío.	episodic
		the	everyone	brought	of	one	book
		Everyone brought one book each.
	b.	* Kathénas éfere apó éna vivlío.

We see here that o kathénas receives both generic and episodic uses – in the latter referring to a discourse specific set of entitites which renders the D-káthe QP context dependent. Giannakidou (2004) and Etxeberria and Giannakidou (2010) claim that D in this case does not function as e-forming, but as a modifier that does not saturate (in the sense of Chung and Ladusaw (2003)) the NP: it composes with Q, via an operation called D-domain restriction (DR), designated in d below. D-DR can be thought of as a morphological or lexical operation on the Q, and semantically it contributes the context set variable C. (Westerståhl (1984, 1985) claimed that the definite article contributes C anyway). C renders the QP anaphoric to a salient discourse set (property). So, for laguages that employ D-restriction, contextual restriction is grammaticalized, and is not merely a matter of pragmatics.

(78)	a. [QP o D + káthe Q [NP fititís N]]
	b. o káthe fititís = [(C) káthe] (student)	‘each student’

The result of D-DR is a presuppositional Q, i.e. a Q imposing on the context the constraint that there be a non-empty set to quantify over. Similar D- universals are observed in Basque, Salish, and Hungarian. Etxeberria and Giannakidou further suggest that each has a structure parallel to the Greek [D-every]; only with each, D is covert. This idea is supported also by the parallels between D káthe and each in the domain of distributivity that we discuss next. Finally, Matthewson 1998, 2001 also documents interactions of D with quantifiers in Salish.

The context dependent and therefore presuppositional nature of D-universals means first that these QPs will not be able to quantify over empty sets. This is indeed what we observe. Notice the contrasts below, and the parallel of D- káthe and each (the examples are from Etxeberria and Giannakidou (2010)):

(80)	#An vris to káthe láthos, tha su dhóso bonus; but there may be no mistakes at all.
	#If you find each mistake, I’ll give you a bonus; but there may be no mistakes at all.
(81)	If you find every mistake, I’ll give you a bonus; but there may be no mistakes at all.
	An vris káthe láthos tha su dhóso bonus; but there may be no mistakes at all.
(82)	If you find all (the) mistakes, I’ll give you a bonus; but there may be no mistakes at all.
	An vris óla ta láthi tha su doso bonus; but there may be no mistakes at all.

Unlike óla and bare káthe, o káthe and each presuppose the existence of mistakes, and Giannakidou (1997, 1999) characterizes o káthe as veridical for this reason.

Second, D-universals cannot refer to non-existing kinds:

(83)	a. Káthe monókeros éxi éna kerato.	(Etxeberria and Giannakidou 2010)
	Every unicorn has one horn.
	b. # O káthe monókeros exi éna kerato.
	# Each unicorn has one horn.

Again, notice the parallel with o káthe and each: they both cannot refer to non-existing kinds – but káthe ‘every’ can be used for non-existing kind reference (for each, see Beghelli and Stowell (1997)).

However, D-universals are fine in characterizing sentences:

(84)	a. Sto programá mas, o káthe fititís prépi na epiléksi dhío mathimata simasiologías.
	b. In our program, each student must choose two semantics classes.

What is crucial is the restriction ‘in our program’, which renders the example not a predication of a kind, but a characterizing sentence that expresses a generalization about a particular set of students in our program. O káthe and each can be used here.

This section ends with two more points supporting the composition of D and Q. First, D is used to form other presuppositional determiners in Greek: those equivalent to both, and most:

(85)	Xriázome	ke	ta	dhío	vivlía.
	need.1sg	and	the	two	books
	I need both books.

(86)	Agórasa	ta	perisótera	vivlía.
	bought.1sg	the	more	books
	I bought most books.

‘Both’ in Greek is literally ‘and the two’ – and likewise, we can build presuppositional partives of the form ‘all n of the NP’ (e.g. all three of the books) in a parallel fashion:

(87)	ke	ta	tría	vivlía;	ke	ta	íkosi	vivlía, etc.
	and	the	three	books;	and	the	twenty	books
	‘all three books’;				‘all twenty books’

These QPs presuppose their quantitity, and the use of ke ‘and’ can be seen as a join operation, along with the use of D. Likewise, ‘most’ is decomposed in Greek into D and the comparative of polí ‘many, much’ – perisotera. So Greek appears to use D systematically in the formation of strong quantificational determiners, and not just universal ones. The same pattern is observed in Basque, see Etxeberria’s earlier work, and Chapter 3, this volume.

Second, D plus Q really results in a complex Q, rather than a DP. The competing DP structure is also available, typically with weak quantifiers, in Greek:

(88)	[I [tris fitités	pu írthan sto parti]],	ítan endelos methisméni.
	[The [three students that came to the party]] were completely drunk

These structures are DPs, as indicated in the brackets, and are interpreted like regular definite descriptions: the denotation of three students will be a familiar and unique set of three students. The output of these structures is then of type e, and not a GQ.⁵ Giannakidou and Etxeberria offer two arguments that the D-universal structure is not a DP of this kind. First, [o káthe NP] cannot co-occur with the demonstrative pronoun (aftós ‘this’, ekínos ‘that’) – which in Greek must embed DPs as we noted at the beginning:

(89)	a. aftós	*(o)	fititís
	this	the	student
	this student
	b. ekínos	*(o)	fititís
	that	the	student
	that student
(90)	a. {aftí/ekíni}	i	tris	fitités
	these/those	the	three	students
	b. {aftós/ekínos}	o énas fititís
	this/that one student
(91)	*{aftós/ekínos}	o	káthe	fititís
	this/that	the	every	student

The demonstratives aftos/ekinos require a DP. Since the demonstrative cannot occur with o káthe, we must conclude that the o káthe constituent doesn’t count as a DP.

The second piece of evidence that [o káthe NP] does not behave syntactically as a DP comes from the fact that it cannot be used in the polydefinite structure that we mentioned at the beginning; but a numeral under D is no problem:

(92)	o	kókinos	o	tíxos
	the	red.nom	the	wall.nom
	the wall that is red
(93)	a. * o káthe o fititís
	b. o énas o fititís			‘the one the student’
	c. i tris i fitités			‘the three the students’

In a language where DPs duplicate easily and routinely, the impossibility of D-spread with o káthe suggests again that o káthe does not create a DP.

6.3.2.2 D-Universals, Distributitivity, and Distirbutivity Markers

D-universals are distributive. Although the DP with oli can have collective or distributive readings, the káthe QP, with or without D, does not have collective readings. We see below that all variants of (o) káthe are incompatible with a collective predicate like ‘meet’:

(94)	a.	Oli	i	fitités	sigendróthikan.
		All	the	students	gathered. (collective)
	b.	*{Káthe fititís / o káthe fititís / o kathénas} sigendrothike.
		every student/each student/everyone gathered. (distributive)

There are, however, degrees of distibutivity. For instance, unlike everyone, o kathénas is awkward without an overt distributor. In our example earlier which I repeat here, we had apó éna vivlío, a typical distributive PP formed with the preposition apó; without the proposition, with a simple accusative, the reading strongly preferred is the collective one, which renders the sentence odd again:

(95)	a.	O	kathénas	éfere	apó	éna	vivlío.
		The	everyone	brought	of	one	book
	b.	# O kathénas éfere éna vivlío.
		Everyone brought one book

So, o kathénas really needs a distributive phrase to be well-formed. In the absence of an explicit phrase, e.g. when we use an intransitive verb, or an individual level predicate (that cannot distribute over events) as in the next examples, the result is problematic for o kathénas – but not for káthe NP and ‘everyone’:

(96)	a.	#O kathénas íne 7 xronón.
		Everyone is 7 years old.
	c.	# O kathénas kimíthike.
		Everyone slept.
		#Each one slept.
(97)	a.	Káthe fititís íne 20 xronón.
		Every student is 20 years old.
	b.	Káthe fititís kimíthike.
		Every student slept.

This contrast suggests that o kathénas is strongly distributive, and cannot be used without a distibutive phrase. In English, each has been claimed to be so (Beghelli and Stowell 1997) – notice the parallel ill-formedness of #Each one slept. If each is also a D-universal, then the distributivity property must related to the use of D. Káthe, on the other hand, and everyone, seem to have no sensitivity to the presence of a distributive phrase and they do not contain D. We can think of them as weakly distributive.

Support for both (a) strong distributivity of o kathénas and (b) the parallel between D-universals and each comes from the fact that o kathénas itself is used as a distributive phrase, quite like binominal each in English:

(98)	Fagame	(apó)	tría	míla	o kathénas.
	ate.1pl		three	apples	each
	We ate three apples each.

We see here that o kathénas is used as a distributor of the object QP (with only optional addition of apó; recall that apó is necessary for distributivity otherwise), just like each.

Interestingly, another distributive construal with káthe involves káthe énas, and no D, which I think can be best thought as ‘each one’.

(99)		Context: I met with a group of students.
	a.	Káthe énas ixe káti endiaferon na mu pi.
		Each one had something interesting to say.
	b.	O kathénas ixe káti	endiaferon na mu pi.
		Each one had something	interesting to say

The presence of énas renders both construals anaphoric in the sense that they need an antecedent, hence the strong requirement that there be a context with explicit mention of students. This requirement of explicit previous mention is not present with káthe or o káthe, since these can either not be context sensitive (káthe), or their domain extension can be accomodated (o káthe).

With indefinites, the way to create distributivitity markers is by reduplicating:

(100)	a.	I	fitités	bikan	dhío-dhío.
		the	students entered		two-two
		The students entered in twos/two by two.
	b.	O	Jánis	éfage	ta	sokolatákia	éna-éna.
		the	John	ate	the	chocolate. PL	one-one
		John ate the chocolates one by one.
	c.	*O	Jánis	éfage	to	sokolatáki	éna-éna.
		the	John	ate	the	chocolate.SG	one-one

Reduplicated numerals and indefinites in Greek are distributivity markers. Such reduplication seems to be a strategy for distributivity crosslinguistically – e.g. the Hungarian reduplicated egy-egy (Farkas 1998) is likewise distributive. Distributive indefinites obviously depend on a plurality to be able to distribute, so they are out with singulars, as we see above in c.

6.3.2.3 D-Universals and Indiscriminative Free Choice Readings

We discuss free choice phenomena in detail in Section 6.7, but here it is important to note that the Greek D-universal, but not káthe, has the so-called indiscriminative reading (Horn 2000) that appears in English with just any. The Greek free choice item opjosdhípote (Giannakidou 2001) can also co-occur with o (Lazaridou-Chatzigoga (2007), see also examples in Vlachou (2007)). Giannakidou and Etxeberria are the first to note the indiscriminative reading with o káthe:

(101)	a.	Tin períodho ton eksetaseon		erxete	o káthe fititís
		the period the.gen. exas.gen		comes	the every student
		ke me enoxlí	me anoites erotísis.
		and me bothers	with silly questions
	b.	Tin períodho ton eksetaseon		erxete	o opjosdhípote fititís
		the period the.gen. exams.gen		comes	the any.FC student
		ke me enoxlí	me anoites erotísis.
		and me bothers	with silly questions
		During the exam period, just about any student may come by and bother me with silly questions.

Here o káthe makes reference to a salient set in the discourse – the students of the speaker – and expresses a generalization about this set, while also being indiscriminative (in the sense of Horn 2000, 2006): o káthe fititís is read like any random student of the speaker, as suggested above by using just about any in the translation. We have thus restriction to a particular set (my students), and an arbitrary/pejorative reading at the same time, a reading that often arises with free choice items. Importantly, the pejorative reading does not arise with bare káthe:

(102)	a.	(Stis meres mas), o kathénas	borí na vgali dhíploma odhígisis.
		(Nowadays), just anyone	can get a driver’s license
	b.	(Stis meres mas), káthe enílikas borí na vgali dhíploma odhígisis.
		(Nowadays), every adult can get a driver’s license.

The a example, with o kathénas, creates a context in which the driving test is simply too easy, and even bad drivers can pass it. But the b sentence with káthe is simply a statement that it is possible for every adult to take the exam and get a license.

6.3.3 Universal A-Quantifiers

The word for the Q-adverb ‘always’ in Greek is panda, or the slightly higher register pandote:

(103)	a.	I Ariadne	{pánda/pándote}	ksexnái na fái.
		Ariadne	always	forgets to eat.
	b.	I Ariadne	{pánda/ pándote}	kimate noris.
		Ariadne	always	sleeps early.

Pánda/pándote belong to the Ancient Greek adjectival paradigm pas (masc.), pása (fem.), pan (neuter) glossed in Holton et al. as ‘each, all’ (1997: 312) – ote in pándote is the Ancient Greek word for when. In Modern Greek, the pas paradigm is still used, again in combination with the D; and it belongs to a slightly formal register:

(104)	a.	Irthan	i		pándes.			(*pándes)
		came	the.masc.pl		all.masc.pl
		Everybody came.
	b.	Kséri	ta		pánda.			(*pánda)
		knows `the.neut.pl			all.neut.pl
		He knows everything.
	c.	Ise	to	pan		ja	ména.	(*pan)
		be	the.neut.sg	all.neut.sg		for	me
		You are everything to me.

The expressions i pándes, to pan, ta pánda are perceived as everybody and everything – the word for universe is synpan (syn- ‘con’). Notice that unlike káthe, pas actually declines. The adverb panda is the plural neuter, following a common strategy for creating adverbs from adjectives in Greek.

Other adverbial expressions of universal quantification are formed with káthe: káthe kyriakí ‘every Sunday’, káthe mina ‘every month’, káthe xrono ‘every year’, etc.

(105)	(Káthe Kyriakí) páme stin eklisía (káthe kyriakí).
	Every Sunday we go to church.
(106)	Káthe mína prépi na plirónume tus loghariasmús.
	Every month we have to pay the bills.

Káthe can also combine with a clause introduced by the complementizer pu, and it means ‘every time that’:

(107)	Sinxízome	káthe (forá) pu ton vlépo.
	I get-upset	every time that I see him.

The verb contains imperfective aspect, since these are habitual sentences. We see also that káthe can be followed by the word forá ‘time, course’, which allows us to think that the noun is dropped when not present. Adverbs of nonuniversal habitual reference are: siníthos ‘usually’, sixná ‘often’, and the lower frequencies we discussed earlier with existentials. D never appears in adverbial use with káthe: tin káthe kyriakí would be impossible:

(108)	* Tin káthe kyriakí páme stin eklisía.
	Every Sunday we go to church.

This concludes our discussion of universal quantifiers in Greek. Now that we have the basics nailed down, we move on to see what kinds of quantifiers occur in existential structures.

6.4 Existential Structures in Greek

Existential structures in English appear in the form There BE in XP, where in XP is a locative phrase. The study of these structures has a long and venerable tradition in English (Milsark 1977, 1979, Keenan 1987, more recently McNally 1992, Francez 2007, 2009). A main claim has been that the existential structure exhibits the so-called definiteness effect, i.e. it excludes definite DPs, universal and other strong quantifiers, and allows only the (weak) intersective quantifiers. Recent literature on existential structures, however, has made it clear that we need to reconsider the so-called definiteness effect. Here are some examples with definites, each, and a proper name in the English existential:

(109)	a. There is Fred in the garden.	(McNally 1992: (8))
	b. There was the table in the garden.
	c.  There was each faculty member at the meeting.

At the worst, these may be slightly unnatural, and at best, they are fine sentences of English. In Greek there appear to be three structures that can be thought of as equivalent to the English existential: one that involves the verb BE (íne ‘be.3sg/pl); one that involves the verb HAVE (éxi ‘have.3sg) and which seems to be the one exhibiting the strongest definiteness effect; and one that employs the verb exist (iparxi ‘exist.3sg). I will present the data in turn.

6.4.1 BE-Existential

The BE-existential accepts intersective quantifiers, but also definites, demonstratives, and names – though not universals, including D-universals. This again can serve as an argument for the non-DP nature of the D-universal:

(110)	a.	Íne	{tría/polá/káti/meriká/ LIGA/ ∅}	vivlía	páno sto trapezi
		There are	three/many/a.plural/several /few/∅	books	on the table
	b.	Íne	{tuláxiston/to polí/móno}	tría vivlía	páno sto trapezi.
		There are	at least/at most/only	three books	on the table.

(111)	a.	Íne ta pedhiá	sto grafío	ke se periménun.
		There are the kids	at the office,	and they are waiting for you.
	b.	Íne	ola ta pedhiá	sto grafío	ke se periménoun.
		There are all the kids at the office, and are waiting for you.
	c.	# Íne	káthe pedhí	sto grafío ke se periméni.
	There is every child	at-the office waiting
	d.	# Íne	to káthe pedhí	sto grafío ke se periméni.
	There is every child	at-the office waiting

(112)	a.	Íne o Jánis sto grafío ke se periméni.
		There is the John in the office waiting for you.
	b.	Íne aftos o enoxilitikos typos eki.
		There is this annoying guy over there.

The BE-existential is dispreferred with mass nouns, even when combined with plausible quantifiers:

(113)	a.	#Íne	záxari	ston	kafé.
		is	sugar	in the	coffee
		There is sugar in the coffee.
	b.	#(Dhen)	íne	polí	záxari.
		(not)	is	much	sugar
		There is much sugar. There isn’t much sugar.

6.4.2 HAVE-Existential

In the éxi structure, the quantifier is in the accusative case, so it does not function as the subject of the sentence (as with the BE existential), but as the object. I am not going to indicate case marking in the examples below to keep the glosses simple. The éxi structure is by far preferred with mass nouns:

(114)	a.	Éxi	zaxari	ston	kafé.
		has	sugar	in the	coffee
		There is sugar in the coffee.
	b.	(Dhen)	éxi	polí	záxari.
		(not)	has	much	sugar
		There {is/isn’t} much sugar.

More examples with intersective quantifiers:

(115)	a.	Éxi	{ tría/polá/káti/meriká/ LIGA/ ∅} vivlía páno sto trapezi.
		There are three.many/a.plural/several /few/∅ books on the table
	b.	Éxi {tuláxiston/to polí/móno} pende vivlía páno sto trapezi.
		There are at least/at most/only five books on the table.

Definites, again, are not impossible with éxi – though they are very marginal. But names, universals, and MOST can’t be used:

(116)	a.	?? Éxi ta pedhiá sto grafío ke se periménun.
		There are the kids at the office waiting for you.
	b.	Éxi afto to pedhí sto grafío ke se periméni.
		There is this boy in the office waiting for you.
	c.	?? Éxi ola ta pedhiá sto grafío ke se periménun.
		all the children
	d.	# Éxi káthe pedi sto grafío ke se periméni.
		every child
	e.	# Éxi to káthe pedi sto grafío ke se periméni.
		each child
	f.	# Éxi ton Jáni sto grafío ke se periméni.
		the John
(117)	# {Éxi /íne} ta perisotera vivlía páno sto trapezi.
	# There are most books on the table

The judgments here are subtle, and one may expect considerable speaker variation. But as I said earlier, the HAVE-existential seems to be exhibiting the definiteness effect.

6.4.3 EXIST-Existential

This is a personal structure: the XP is the subject of the verb and there is agreement (unlike with exi where the XP is syntactically the object, and íne where the 3sg and 3pl are the same form). Here are some examples, first with mass nouns:

(118)	a.	Ipárxi	záxari	sto	spiti.
		exists	sugar	in the	house
		There is sugar in house. (No need to buy more).
	b.	(Dhen)	ipárxi	arketi	záxari.
		(not)	exists	much	sugar
		There is enough sugar. There is not enough sugar.

More examples with intersective quantifiers:

(119)	a.	Ipárxun	{tria/polá/káti/meriká/ LIGA/ ∅} vivlía páno sto trapézi.
		exist.3pl	three.many/a.plural/several /few/∅ books on the table
	b.	Ipárxi	{tulaxiston/to polí/mono} éna vivlío páno sto trapézi.
		Exist.3sg	at least/at most/only one book on the table.

Definites, names, universals, and MOST are impossible:

(120)	a.	#Ipárxun ta pedhiá sto grafio ke se periménun.
		Exist.3pl the kids at the office waiting for you.
	b.	#Ipárxi afto to pedhí sto grafío ke se periméni.
		Exist.3sg this boy in the office waiting for you.
	c.	#Ipárxun ola ta pedhiá sto grafío ke se periménoun.
		Exist.3pl all the children at the office waiting for you
	d.	# Ipárxi káthe pedhí sto grafio ke se periméni.
		Exist.3pl every child at the office waiting for you
	e.	# Ipárxi to káthe pedhí sto grafio ke se periméni.
		Exist.3pl each child at the office waiting for you
	f.	# Ipárxi o Jánis sto grafio ke se periméni.
		Exist.3sg the John at the office waiting for you

(121)	# Ipárxun	ta perisotera	vivlía páno sto trapezi.
	exist.3pl.	MOST	books on the table

With ipárxi, there seems to be a definiteness effect, but there is a question here to what extent this structure is a true existential, and not simply an existence predication.

Finally, all three variants exclude the partitive:

(122)	a.	#Èxi tría apo ta pedhiá	sto grafío ke se periménun.
		has.3sg three of the children	at the office waiting for you.
	b.	#Ine tría apo ta pedhiá	sto grafío ke se periménun.
		Is three of the children	at the office waiting for you.
	c.	# Ipárxun tría apo ta pedhiá sto grafío ke se periménun.
		Exist.3pl three of the children at the office waiting for you.

The contrast of the partitive with the simple cardinal tría pedhiá or the modified existentials, which are all good, is really striking – and a question worth examining.

6.5 (More) Morphologically Complex Quantifiers

We have already seen that morphological complexity is involved in the formation of presuppositional and distributive universals in Greek (D-universals), and in the formation of other strong quantifiers meaning ‘both’ (ke i dhío), and ‘most’ (i perisóteri). This overt D-deployment for strong quantifiers is a specific property of Greek (and Basque, see Etxeberria, Chapter 3, this volume); but the complexity we are going to examine now concerns more ‘expected’ complex quantifiers such as comparative quantifiers, those created via boolean compounding (and, or, neither...nor..., and not), exception phrases (all but ten students), and bounding phrases (He exercised twice a day, six days a week for one year). I present each in turn.

6.5.1 Comparative Quantifiers

‘More than’ in Greek is typically formed with parapáno, an adverb meaning ‘further, above’, as in Meni dhío tetragona parapáno ‘He lives two blocks further up’ or the plain adverb páno ‘above’, plus the preposition apó ‘of’ (used also in the partitive and in phrasal comparatives). Greek, therefore, unlike English, does not simply employ the comparative MORE (pio, perisotero) for the more-than quantifier. The perisotero can also be used, as we shall see, but is less preferred. Another difference from English is that the NP appears typically in the plural with MORE THAN ONE:

(123)	a.	Aghorasa	parapáno		apó	éna	{vivlía/?vivlío}.
		bought.1sg	more		than	one	book.pl/book.sg
		I bought more than one {book/*books}.
	b.	Parapáno	apó	énas	fitités		irthan.
		More	than	one	student.pl		came.3pl
		More than one student came.
	c.	??/*Parapáno	apó	énas	fititis		irthe.
		more	than	one	student.sg		came.3sg

The contrast with English, which does not allow the plural, is sharp, and suggests that in Greek ‘more than n’ could be treated in the grammar as a plural determiner. The singular improves typically with temporal expressions, or if we replace parapáno with páno:

(124)	Perimena páno apó mia	{ora/??ores}.
	I was waiting more than one {hour/*hours}.

With numbers larger than one, as expected, only the plural becomes possible.

The MORE perisotero variant is also possible. It is an adjective, thus an agreeing form, and when used, there is strong preference for the plural. Notice below that the singular is indeed ungrammatical:

(125)	a.	Aghorasa	perisótera	apó	éna vivlía.
		bought.1sg	more.pl.	than	one books
	b.	*Aghorasa	perisótero	apó	éna vivlío.
		bought.1sg	more.sg.	than	one book
		I bought more than one book/*books.

Here we see that the MORE part also shows plural morphology, agreeing wih the NP. In this comparative structure we find a strong pattern with the plural, stronger than with the adverbial.

Another kind of comparative quantifier is ‘more girls than boys’. In Greek this appears as {pio polá/perisótera} koritsia apóti agória, lit. ‘{more much/more} girls than boys’ – apóti being one of the words for THAN that Greek employs (there is a bunch of them: apó for phrasal comparatives, apóti for clausal comparatives, and pará for metalinguistic comparatives; see Giannakidou and Stavrou (2009), Giannakidou and Yoon (2011), Merchant (2009) for more details). Apó is strongly dispreferred, and the comparative clause can separate, as in English:

(126)	a.	Irthan	perisótera		koritsia	{apóti/*apó}	agória.
		came	more.pl		girls	than.clausal/of	boys
		More girls came than boys.
	b.	Perisótera		koritsia	irthan	apóti	agória.
		more.pl		girls	came	than.clausal	boys
		More girls came than boys.

The fact that the QP is discontinuous, and the use of clausal than apóti, suggests that maybe the [perisótera NP apóti NP] is not a constituent – and the comparative part is clausal comparative with TP ellipsis (which is the standard analysis of the apóti clause in Greek, Merchant (2009)).⁶

The comparative quantifiers occur uneventfully in the existential structure:

(127)	Exi/Íne	perisotera		koritsia	apóti	agória	stin	taksi mas.
	has/is	more.pl		girls	than	boys	in-the	class ours
	There are more girls than boys in our class.

(128)	a.	{Exi/Ine}		perisotera apó éna		vivlía	sto trapezi.
		has/is		more than one		books	in-the table
		There is more than one book on the table.
	b.	{Exi/Ine}	parapáno apó mia		{óra/??óres}	pu	se	periméno.
		has/is	more than	one	hour/hours	that	you.acc	wait.1sg
		There is more than one hour that I am waiting for you!

Again, the plural is the expected form, with the exception of the temporal expression where the singular is preferred.

6.5.2 Boolean Compounding

Boolean compounding is generally possible. Here are some examples:

(129)	a.	Parapáno apó 5	alá to polí 10 fitités				tha jínun dektí.
		More than 5	but at most 10 students				will be admitted
	b.	Perisoteri apó 5	alá to polí 10 fitités				tha jínun dektí.
		More than 5	but at most 10 students				will be admitted
	c.	{Parapáno/perisoteri} apó 5 alá óxi páno apó 10 fitités
		More than 5 but no more than 10 students
		tha jínun dektí.
		will be admitted

Notice again the use of both the adverbial form and MORE. In the negative (‘no more than’) version, we also use óxi which is Greek constituent negation (Giannakidou 1998, Veloudis 1982).⁷ It is also used in other but-compounds, as well as constituent negations of quantifiers:

(130)	a.		o Jánis alá óxi i María		‘John but not Mary’
	b.		Irthan	óxi	óli	i	fitités.
			came.3pl	not	all	the	students
			Not all the students came.

(131)		Efxaristíthikan poli		ala	óxi	óli	i	kalesmeni.
		enjoyed.3pl	many	but	not	all	the	guests
		Many but not all the guests had a good time.

Neither…nor construals are formed by oúte…oúte. Oúte is the NPI-EVEN in Greek (Giannakidou (2007); Greek also has a positive EVEN akomi ke). In addition to being a focus particle, the lexical item oúte is also used as cross-categorial coordinator. The examples below are from Giannakidou (2007: (45)):

(132)	a.	Sto	párti	o	Jánis	oúte	efage	oúte	ípje.
		at-the	party	the	John	neither	ate.3sg	neither	drank.3sg
		At the party John neither ate nor drank.
	b.	*(Dhen)	milisa	oúte	me	to	Jáni	oúte	me	ti	María.
		not	talked.1sg	neither	with	the	John	neither	with	the	María.
		I talked to neither John nor María.
	c.	Oúte	i	María	(dhen)	írthe.
		even	the	María	didn’t	come
		Mary didn't come either.

In the last example, oute (dhen) is used as not either (for more details see Giannakidou (2007)).

6.5.3 Exceptive Phrases

Typical exceptive phrases in Greek are formed with ektós, which is an adverb meaning literally out/outside, or beside, as in Afto íne ektós thematos ‘This is beside the topic’, plus our familiar preposition apó:

(133)	Irthan	óli	ektós	apó	to	Jáni.
	came	all	apart	from	the	John
	{All/everyone} came but John.
(134)	Irthe	káthe	fititís	ektós	apó	to	Jáni.
	came	every	student	apart	from	the	John
	Every student but John came.

There is also a more formal version with the genitive: ektós tu Jáni. As in English, the ektós constituent can be separated:

(135)	Káthe	fititís	írthe	ektós	{apó	to	Jáni	/tu Jáni}.
	every	student	came	apart	from	the	John/	John.gen
	Every student came but John.

Another way to form the exceptive phrase is via alá óxi ‘but not John’: óli i fitités alá óxi o Jánis ‘all the students but not John’. Again, separation is possible:

(136)	a.	Idha ólus tus fitités alá óxi ton Jáni.
		saw.1sg all the students but not the John
	b.	Olus tus fitités ídha alá óxi to Jáni.
		I saw all the students but not John.

6.5.4 Bounding Phrases

These are adverbial phrases like ‘twice a day’, ‘six days a month’. In Greek these appear with accusative DPs:

(137)	a.	dhío	forés	tin	iméra		‘two times a day’
		two	times	the.acc	day.acc
	b.	éksi	méres	to	mína		‘six days a month’
		six	days	the.acc	month.acc
	c.	eptá	forés	to	xróno		‘seven times a year’
		seven	times	the.acc	year.acc

Notice that there is no special word for twice, dhío forés is ‘two times’. There is a more archaic paradigm ending in –is: dhis (twice, from dhío), tris (thrice, from tría), tetrákis (four times, from tessera), and also polákis ‘many times’; this paradigm, however is not very productive in Modern Greek, and is used only in very high registers.

We move on now to polarity quantifiers.

6.6 (Negative) Polarity Quantifiers

In this section we discuss polarity sensitive quantifiers – also known also as negative polarity items (NPIs). Some of these appear only in negative contexts, but others have a broader distribution in nonveridical contexts, i.e. they are sensitive to whether a truth or existence inference is available (Giannakidou 1997 et sequel; Zwarts 1995). The examples here are mostly from my previous work on Greek NPIs.

In the literature on English, any is often quoted as an NPI, though any is known to have two readings, the NPI reading (with negation) and the free choice reading (with modal verbs and imperatives).

(138)	a.	I didn’t buy any books.	NPI
	b.	Any book can be useful.	Free choice
	c.	Press any key.	Free choice

The NPI reading is an existential quantifier in the scope of negation, but the free choice reading conveys freedom of choice (Vendler 1967), and it may look like it involves universal quantification – but look at the c example (from Giannakidou (2001); see also Horn (2000, 2006) for arguments against the universal analysis of any). Greek, like many other languages, employs distinct lexical items for the NPI-existential and the free choice quantifier (for more data from other languages, see Haspelmath (1997)). We discuss free choice in Section 6.7, along with wh-quantifiers because the free choice quantifier (but not the NPI) is wh-based in Greek.

With negation, Greek employs what appears to be one lexical NPI, but it comes in two intonational variants: an emphatic and a non-emphatic version (Veloudis 1982, Giannakidou 1994, 1997, 1999, 2000). The emphatic one seems to be a strong NPI, licensed only in the scope of negation and antiveridical expressions such as without, but the non-emphatic appears in the whole range of non-veridical environments which include, but are not limited to, some (but not all) downward entailing quantifiers. I will start by describing the NPIs with negation. I also discuss minimizers and negative concord in this context. Then, I illustrate the difference in distribution between the emphatic and non-emphatic NPIs. I also compare the non-emphatic NPI to any, and show empirical differences suggesting that any is not always licensed but can be rescued in semantically non-licit environments such as only, the complements of factive verbs, and comparatives (Giannakidou 2006, Giannakidou and Yoon to appear). Finally, there will be some observations showing a correlation between intonation and quantifier scope in Greek.

6.6.1 Emphatic and Non-emphatic NPIs in Negative Contexts, and Negative Concord

Greek has the two paradigms of NPIs illustrated below (Veloudis 1982, Giannakidou 1997 et seq., Tsimpli and Roussou 1996). The glosses are suggestive only:

(139)	kanénas/KANÉNAS	‘anyone, anybody/no-one, nobody’
	kanénas N/KANÉNAS N	‘any N/no N’
	típota/TÍPOTA	‘anything/no thing’
	poté/POTE	‘ever/never’
	puthená/PUTHENA	‘anywhere/nowhere’
	kathólu/KATHOLU	‘at all/not at all’

Upper-case letters indicate emphatic accent. Kanénas is the masculine, kamía is the feminine, kanéna is the neuter. The accent is not related to focus for reasons that have been discussed elsewhere (Giannakidou 1997, 1998: 227–231). Given the quantifiers polí and líji, which also come in emphatic and non-emphatic variants, I suggested that it is best to handle emphatic n-words as lexically distinct from non-emphatic ones, so emphatic accent functions as morphological marking.

The NPI series uses a variety of morphological sources including existential quantifiers (énas), universals (ólu), and wh- (pu, poté, with stress shift from póte ‘when’). Under negation and antiveridical without both paradigms are licensed:

(140)	a. Dhen	ídhe	{típota/TÍPOTA}	o		Jánis.
	not	saw	anything	the		Johh
	John didn’t see anything.
	b. * Idhe		{típota/TÍPOTA} o Jánis.
(141)		*xoris	na	dhi	{kanénan/KANÉNAN}.
		without	subj	see.3sg	n-person
		without having seen anybody.

So, both paradigms – emphatic and non-emphatic – are NPIs and need negation. The version with the emphatic is a negative concord structure, i.e. since it contains negation plus an NPI that itself appears to be negative – i.e. it can answer negatively as a fragment:

(142)	Pjon ídhes?	“Who did you see?”
	{KANÉNAN/*kanénan}
	Nobody/*Anybody.

The ability to answer negatively while participating in negative concord is the hallmark property of NPIs known as n-words (Laka 1990, Giannakidou 2006). Two things are important to note here. First, the emphatic NPI gives a negative answer, and second, the non-emphatic NPI cannot do that. In Giannakidou (1998, 2000) I argued that the fragment NPI is the remnant of an elliptical answer that has undergone ellipsis, and ‘given that the remnants in fragment answers are accented, non-emphatics are excluded because they are not accented. Considering that utterances with non-emphatics typically involve pitch accent on negation, we may argue alternatively that ellipsis excludes non-emphatics because the accented negation itself must be deleted’ (Giannakidou 2000: 469).

Another difference between emphatic and non-emphatic NPIs with negation concerns locality. Non-emphatic NPIs, but not emphatics, are licensed in syntactic islands. The example below illustrates this with a relative clause (but other examples are given in Giannakidou (1998); see also Quer (1993) for a similar observation about Catalan n-words):

(143)	Dhen	prodhosa	mistiká	[pu	ekséthesan	{kanénan/*KANÉNAN}]
	not	betrayed.1st	secrets	that	exposed.3pl	n-person
	I didn’t reveal secrets that exposed anybody.

In this respect, non-emphatics are like any, which is also licensed in islands as we see in the translations. Given that non-emphatics appear in islands, it is not surprising that they also appear long-distance, again like any. Notice too the contrast with the emphatic NPI:

(144)		I	Ariadne	dhen	ípe	oti	ídhe	{típota/*TÍPOTA}.
		the	Ariadne	not	said.3sg	that	saw.3sg	n-thing
		Ariadne didn’t say that she saw anything.

The observed locality of the emphatic NPI is again typical of negative concord, and is reminiscent of universal quantifier dependencies, which are also clause-bounded (for Greek, see Farkas and Giannakidou (1996)). Three things are additionally important to note here. First, Greek exhibits what I called strict negative concord, i.e. it always requires the presence of negation for the licensing of the emphatic NPI:

(145)	a.	KANÉNAS	*(dhen)	ípe	TÍPOTA.		Greek
		n-person	not	said.3sg	n-thing
		‘Nobody said anything.’
	b.	Nikt	*(nie)	uderzyl	nikogo.		Polish
		n-person	not	hit.3sg	n-person
		‘Nobody hit anybody.’
	c.	Balázs	*(nem)		beszélt	senkivel	semmiröl.	Hungarian
		Balázs	not		spoke.3sg	n-person	n-thing
		‘Balázs didn’t talk about anything with anybody.’

Greek, Hungarian, Japanese (Watanabe 2004), and Slavic languages form a natural class in terms of strict negative concord, and require sentential negation even when more than one n-word occurs in a sentence. In some Romance languages, the presence of negation is not obligatory, and two n-words may co-occur without it as long as one of them is preverbal (Zanuttini 1991):

(146)		Nessuno		ha	letto	niente.		Italian
		n-person		have.3sg	read	n-thing
		‘Nobody read anything.’

So negative concord in Romance is not strict. Given examples like the above, it is conceivable that these Romance n-words form branching negative quantifiers (de Swart and Sag 2002), but it is implausible to argue this for Greek or other strict negative concord varieties, where the NPI n-words alone do not suffice for negative meaning.

Another piece of evidence against negativity of n-words in Greek is that emphatics do not give rise to double negation readings (Giannakidou 2000, 2006):

(147)		KANÉNAS	dhen	ípe	TÍPOTA.
		n-person	not	said	n-thing
		Nobody said anything.
		Nobody said nothing.

The sentence does not have a double negative reading, as we would expect under the hypothesis that the n-words are negative (e.g. Nobody said nothing). The strict concord property, locality, and the absence of double negation readings, along with a number of other diagnostics employed in my earlier work, led me to the conclusion that Greek emphatics are not negative quantifiers, but rather, universal quantifiers that need to be interpreted outside the scope of negation (Giannakidou 1998: chapter 4, 2000). Such universal NPI n-words have since then been identified in Korean (Yoon 2008), Japanese (Yoshimura 2007), and one variety of Hungarian n-words (Surányi 2006). These n-words, crucially, also have emphatic intonation. Puskás (1998) in particular argues for Hungarian that ‘This stress [i.e., the accent observed in Hungarian n-words] cannot be assimilated with the stress assigned in FP [Focus Phrase] which has strong emphatic or identificational reading. Therefore it cannot be argued that Hungarian negative phrases carry the feature [+f]’ (Puskás 1998: 199). Szabolcsi (l981: 530–532) also observes that Hungarian n-words, on a par with universal quantifiers, ‘may not fill the F-position’. If these n-words are universal quantifiers, the fact that the accent is not focus ties in with their semantic function as universals.

6.6.2 Negation, Intonation, and Scope in Greek

Since we are talking about emphatic NPIs scoping above negation, it is relevant to note the following generalization (Giannakidou 1998: 71–73, 2000).

(148)	The scope-negation generalization
	In Greek, a pitch-accented quantifier takes wide scope over negation.

This is a general observation about quantifier and negation interaction, and I am relying here on discussion from Giannakidou (2000: 480–481). Consider the sentences below:

(149)	I	Anna	dhen	parakolúthise	PARAPÁNO	apó	tría	mathímata
	the	Anna	not	attended.3sg	more	from	three	classes
	Anna didn’t attend more than three classes.

The English version of this sentence has two possible readings, depending on whether more than three classes scopes over negation or not. The first possibility is illustrated in the LF where more than three classes has adjoined to IP, and takes wide scope over negation. The second possibility indicates adjunction of more than three classes to VP, below negation.

(150)	a.	[IP more than three classes1 Anna didn’t [VP attend t1]]
	b.	[IP Anna didn’t [VP more than three classes1 [VP attend t1]]]

Under the a reading, we know that there were more than three classes from which Anna was absent, and we have no idea how many classes she actually attended. In the b reading, on the other hand, with negation taking wide scope, Anna attended no more than three classes, and we don’t know how many classes Anna was absent from. Hence, the two readings are true under distinct circumstances.

Now, the Greek sentence, with the accented QP, has only the wide scope QP reading, whereas accent on negation dhen permits only the wide scope negation reading. The use of ‘accent’ here is a bit impressionistic, but see Baltazani (2002) for a more phonologically informed description. So, accent seems to indicate the element taking wide scope. The point can be further illustrated with the interaction between negation and kápjon fititi ‘some student’:

(151)	a.	DHEN	idha	kápjon	fititi.
		not	saw.1sg	some	student
		I didn’t see any student.
	b.	Dhen	idha	KÁPJON	fiti.
		not	saw.1sg	some	student
		There was some student that I didn’t see.

The sentence a has only the wide scope negation reading below, and the b sentence can only have wide scope kápjon fititi:

(152)	a.	¬∃x[student(x) & saw(I, x)]
	b.	∃x[student(x) & ¬saw(I, x)

A final point before closing is that another NPI, the minimizer, is formed in Greek with an emphatic bare singular. Bare arguments (singulars and plurals as we see) are generally allowed under negation and their interpretation is a narrow scope existential (as Carlson (1977) observed for English bare arguments under negation). This is an interpretation akin to that of the non-emphatic NPI-existential:

(153)	a.	Dhen	efaje	BUKIA.
		Not	ate	bite
		He didn’t eat a bite.
	b.	Dhen agorasa vivlía.
		I didn’t buy books.

Minimizers, interestingly, also bear accent (see the a example above) – but cannot be argued to scope above negation. Notice however, that the bare argument is not a quantifier; so, we can still maintain Giannakidou’s generalization that pitch accent on the quantifier indicates wide scope. The accent on the minimizer can be taken to constitute a marking of the conventionalization of the item as a minimizer NPI – maybe an overt reflex of NPI-EVEN oute, which can also be used in the minimizer NPI (Giannakidou 2007). Notice that the bare plural in the b example does not bear accent and is not conventionalized as an NPI.

6.6.3 Nonveridical Contexts: Only Existential NPIs

As mentioned earlier, the non-emphatic NPI is a narrow scope existential inside the scope of negation, so it is the Greek counterpart to NPI any – and just like any, its distribution is not limited to the scope of negation. The existential NPI appears in a broad variety of non-veridical contexts including questions, conditionals, modal verbs, the future, imperatives, subjunctive complements of non-veridical verbs. The emphatic NPI in these environments is systematically ruled out:

(154)	Píjes	{poté/*POTE}			sto	Parísi?
	went.2sg	ever			in-the	Paris
	Have you ever been to Paris?
(155)	An dhis	tin Eléna		{puthená/*PUTHENA}, na tis milísis.
	If you see	Eléna		anywhere, talk to her.
(156)	Elpízo	na	emine		{kanéna/*KANÉNA}			komati.
	hope.1sg	subj	remained.3sg		any			piece
	I hope there is a piece left.
(157)	Pare		{kanéna/*KANÉNA}			mílo.
	take.imp.2sg		any			apple
	Take any apple.
(158)	Borí	na	írthe		{kanénas/*KANÉNAS}
	can.3sg	subj	came.3sg		any person.
	It is possible that anyone/someone came.					(epistemic modal)

The nonemphatic NPI is further licensed in disjunctions, with various modalities, and habitual sentences. With a few exceptions (noted in the literature), these are also licensing contexts for any, though the free choice reading is considerably preferred (see my earlier work for extensive data). The Greek NPI does not have a free choice reading, and it is also non-scalar (Giannakidou 1997, 1998, 2009). Rather, it seems to be a narrow scope indefinite that contains a dependent variable, i.e., one that can never be interpreted as a free variable (Giannakidou 2011), and which therefore needs to be licensed via binding (either via ∃-closure under negation and nonveridical operators, or via binding by a Q-operator). Additionally, kanénas brings in a condition of referential vagueness (Giannakidou and Quer 2011), and Holton et al. (1997) characterize it as ‘non-specific’.

As far as downward entailing (DE) contexts go, NPIs are OK with negative value judgement quantifiers, e.g. emphatic LIJI or elaxisti ‘very few’, but not with something more neutral (in terms of judgement) as at most n:

(159)	a.	{Eláxisti/?LIJI}	ánthropi ídhan típota.
		Very few/Few	people saw anything.
	b.	* To poli 5	ánthropi ídhan típota.
		At most five people saw anything.

Notice the relative awkwardness of LIJI ‘few’ – the judgements I have collected through the years vary a lot regarding this quantifier. Given the impossibility of AT MOST, we must conjecture that the negative judgement is important for licensing, and not DE per se.

Finally, it is important to add that that there are environments where any is fine, but the Greek NPI cannot occur. Some such environments are only, the complements of emotive factive verbs, and comparatives. We review these next, in connection with minimizer NPIs.

6.6.4 Non-licensing Environments for Greek NPIs

In English, any and minimizers like say a word are cited as appearing in the complement of emotive factive verbs (positive and negative), with only, and in comparatives:

(160)	a.	I am glad he said a word!
	b.	I’m glad we got any tickets.	(Kadmon and Landman 1993).
	c.	Mary regrets that she lifted a finger.
	d.	Only Mary {gives a damn/said anything}.
Comparatives
(161)	a.	Roxy is prettier than anyone of us. (phrasal)
	b.	Roxy ran faster than anyone had expected. (clausal)
	c.	He said the sky would sooner fall than he would budge an inch.

The Greek NPI and the minimizer – formed with a bare nominal, as must be recalled – are excluded from these contexts (see Giannakidou (2006), and Giannakidou and Yoon (2010), where the comparative examples are drawn from):

(162)	a.	*Xérome pu dhinis dhekára.
		I am glad you give a damn.
	b.	* Metániosa pu ipa típota.
		I regret that I said anything.
	c.	*/# Móno i María {dhini dhekára/idhe típota}.
		Only Mary gives a damn/said anything.
	d.	#I María metániose pu ipe leksi.
		Mary regrets that she said a word.
(163)	I María tréxi grigorótera apó {opjondhípote/*KANÉNAN/*kanénan}.
	Mary runs faster than anybody.
(164)	*I María	diavase	perisótera arthra	apóti	tis ixe protini kanénas kathijitís.
	Mary	read	more articles	than	suggested any professor
	Mary read more articles than any professor has suggested.

So we see this asymmetry between Greek NPIs and minimizers, on the one hand, and English any and minimizers, on the other, as regards the possibility of rescuing (Giannakidou 2006), i.e. sanctioning by global pragmatic inferencing. Rescuing is a secondary sanctioning mechanism, which legitimizes NPI in violation of LF licensing: NPIs here are found in a veridical context without an ‘official’ licenser.

This concludes our discussion of NPIs. We move now to wh-quantifiers, our final topic.

6.7 Wh-Based Quantifiers and Free Choice

Greek has three paradigms of wh-quantifiers: one for interrogatives, one for relative clauses, and one for free relatives (called ‘correlative’ quantifiers in Holton et al. (1997)). In the relative clause, we see again the workings of the definite (D) article o, since it appears on top of the wh-component, either forming a unit with the wh-word (free relatives), or in addition to it (relative clauses). The free relative, D-containing construal is the source for free choice quantifiers (Giannakidou 2001, Giannakidou and Cheng 2006). So, overall we observe a manipulation of wh-forms (and meaning) by definiteness – a fact challenging the link, found in recent works (e.g. Kratzer and Shimoyama 2002), between interrogative (propositional) semantics and free choice. The Greek wh-patterns suggest a richer interaction between wh-words and definiteness that fits better classical models where the wh-words are sets of individuals (Cooper 1983), operated upon directly via e.g. exhaustification and intensionalization.

6.7.1 The Morphological Paradigms

The interrogative wh-paradigm is given below, followed by examples. I give the labels in the nominative, but bear in mind that wh-words, like the other quantifiers in Greek, also inflect for gender, number, case. I am also giving the variants in the Greek alphabet to see the relations between paradigms. We can think of the interrogative paradigm as the p-paradigm:

(165)	pjos, pja, pjo	‘who’	Greek: ποιóς, ποιά, ποιó
	pjos, pja, pjo N	‘which N’
	ti	‘what’	τί
	ti N	‘what kind’
	póte	‘when’	πóτε
	pu	‘where’	πού
	pos	‘how’	πώς
	póso	‘how much’	πóσο
	jatí	‘why’	γιατί

(166)	a.	Pjos efaje ti supa?
		Who ate the soup?
	b.	Pja mitera den irthe?
		Which mother didn’t come?
(167)	a.	Ti efages?
		What did you eat?
	b.	Ti anthropos íne?
		What kind of man is he?
(168)	a.	Pu ton idhes?
		Where did you see him?
	b.	Poso káni?
		How much does it cost?

The p-paradigm is used only with interrogative meaning. For relative pronoun use, the definiteness marker—which is the invariant form of the definite article as a bound morpheme (Giannakidou and Cheng 2006): opios, opia opio, etc.:

(169)	opíos, opía, opío			‘who.Rel.’		Greek: oποίoς, οποία, οποίο
	opíos N
	opóte			‘when.Rel.’			οπóτε
	ópos			‘how.Rel.’			óπως
	ópu			‘where.Rel.’			óπου

(170)	a.	o	ándras	*(ton)	opío		agapá	i	María
		the	man.nom	the.acc	which.masc.sg.acc love.3sg			the	María
		the man that Mary loves
	b.	i	jinéka		*(i)	opía		diamartirithike
		the	woman.sg.fem.nom		the.acc	which.masc.sg.acc complaíned
		the woman who complaíned

(171)	a.	to meros ópu sinandithíkame
		the place where we met
	b.	(We agreed to meet at 9 pm), opóte ke pigame
		We agreed to meet at 9, which is when we went

Greek also has an uninflected relative pronoun pu (που) which is used in more colloquial speech as a relative that:

(172)	a.	o	ándras	pu	agapá	i	María
		the.masc.sg.nom	man.msc.sgnom	that	love.3sg	the	María
		the man that Mary loves
	b.	i	jinéka		pu	diamartirithike
		the.sg.fem.nom	woman.sg.fem.nom		that	complained
		the woman who complained

Greek employs yet a separate paradigm for free relative and correlative structures, which consists of the definite marker o plus the interrogative p-word – and a stress shift to ó, as indicated. O appears again as a bound morpheme on the p-word and remains invariant. I am using below the (ever) paraphrase as a handy way to show that this is free relative use, i.e. the complement of this pronoun is always a clause, just like with the wh-ever paradigm in English:

(173)	ópjos, ópja, ópjo	‘who(ever)’	Greek: óποιος, óποια, óποιο
	óti	‘what(ever)’	ó,τι
	óti N	‘what(ever) N’
	ópote	‘whenever’	óποτε (vs. Rel: οπóτε)
	ópu	‘wherever’	óπου
	ópos	‘whichever way’	óπως
	óso	‘as much as’	óσο

Some examples:

(174)	a.	Parigila	óti	parigile	o	Jánis.
		ordered	what	ordered	the	John
		I ordered what John ordered.
	b.	Káne	óti	su	pi	i	mitera	su.
		do	what	you	tell	the	mother	your
		Do what your mother tells you.
	c.	Ópjos	irthe sto parti		efxaristithíke.
		Whoever	came to the party		had a great time.
	d.	Kándo	ópos		thélis.
		Do it	whichever way		you want.
	e.	Ópu pao,	me akoluthi.
		Wherever I go he follows me.
	f.	Fae	óso		thélis.
		Eat	as much as		you want.
(175)	*Dhiavase		ópja	efimerida.
	read		wh-ever	newspaper
	*Read whichever newspaper.

Note the inability of the free relative p-word to take an NP complement – it always requires a clause. Free relatives in English have been analyzed as definite descriptions by Jacobson (1995), who argues for a covert iota operator on top of the wh-set. In Greek, Alexiadou and Giannakidou (1998) argue that o is the overt counterpart of iota, hence the Greek free relative is overtly a definite description.

In English – ever is obligatory for free relative use – whoever came to the party, but not *who came to the party – but in Greek plain free relatives are possible, as we saw in the examples above. A free choice variant of the free relative p-word can be formed by adding the free choice marker –dhípote (Giannakidou 1997, 2001), which then bears the main stress in the word. The addition of free choice marking to a wh-form for free choice is a common strategy cross linguistically.

(176)	opjosdhípote, opjadhípote, opjodhípote	‘whoever’	οποιοσδήποτε
	opjosdhípote, opjadhípote, opjodhípote N	‘whichever’
	otidhípote	‘whatever’
	otidhípote N	‘whatever N’	ο,τιδήποτε
	ópotedhípote	‘whenever’	οποτεδήποτε
	ópudhípote	‘wherever’	οπουδήποτε
	óposdhípote	‘definitely’	οπωσδήποτε
	ósodhípote	‘no matter how much’ οσοδήποτε

There is a long-standing debate on whether free choice quantifiers are variants of universal quantifiers or not. Giannakidou (1998, 2001) argues that Greek FCIs are best analyzed as variable contributing elements without force of their own – i.e. indefinites (see Horn (2000, 2006) for a similar analysis of English any). Giannakidou and Cheng (2006) further identify free choice free relatives as definite FCIs, relying on the analysis of free relatives as plural definites of Jacobson. So jointly, the universal effect of FCIs, when it arises, can be accounted for by (in)definiteness and there is no need to recourse to a universal analysis.

6.7.2 Distribution of FCIs: Polarity and Variation

Greek FCIs appear to have limited distribution too, and are excluded from positive veridical sentences (in the simple past). So FCIs are polarity items in this broad sense. Unlike NPIs, however, FCIs do not improve with negation, as long as the sentence remains episodic (Giannakidou 1997, 1998, 2001). I give below examples from Greek and Spanish, Catalan:

(177)	a.	* Idha		opjondhípote.		(Greek; Giannakidou 2001)
		saw.perf.1sg		FC-person
		‘*I saw anybody.’
	b.	* Dhen		idha		opjondhípote.
		not		saw.perf.1sg		FC-person
		Intended: ‘I didn’t see anybody.’8

(178)	* (Non) Expulsaron del partido a cualquier disidente. (Spanish)
	not expel.3pl from-the party acc FC dissident
	Intended: ‘They didn’t expel any dissident from the party.’ (Quer 1999)
	‘*They expelled any dissident from the party.’
(179)	* (No) Li va comprar qualsevol ram. (Catalan)
	Not her/him aux.3sg to.buy FC bouquet
	Intended: ‘S/he did’t buy him/her any bouquet.’ (Quer 1998)
	‘*S/he bought him/her any bouquet.’

Rather, FCIs are licensed via binding: they contain a world variable that needs to be bound, so they must be found in the scope of intensional and modal operators (all nonveridical) that can bind it. This is why FCIs cannot be used in an episodic context. I give some examples here. Notice that I am using the –or other paraphrase to get the difference between the FCI and the NPI:

Protasis of conditionals
(180)	An	kimithis	me	{opjondhípote/kanénan}	tha	se	skotoso.
	if	sleep.2sg	with	FC-person/NPI-person	fut	you	kill.1sg
	If you sleep with anybody, I’ll kill you.

Directive intensional verbs (selecting subjunctive )
(181)	I	Ariadne	epémine	na	afisoume	{opjondhípote/kanénan}
	the	Ariadne	insisted.3sg	subj	let.1pl	FC-person/NPI-person
	na		perasi	mésa.
	subj		come.3sg	in
	Ariadne insisted that we allow anyone in.
	With kanénan: ‘Ariadne insisted that we allow someone or other to come
	in.’

(182)	Borí	na	ánapse	{opjosdhípote/kanénas}		to	fos.
	can.3sg	subj	lit.3sg	FC-person/NPI-person		the	light
	Anyone may have turned on the light.
	With kanénas: ‘Someone or other must have turned on the light.’

(183)	Borís	na	dhanistis		{opjodhípote/kanéna}		vivlío.
	can.2sg	subj	borrow.2sg		FCI /	NPI	book
	You may borrow any book.
	With kanéna vivlío: ‘You may borrow some book or other.’

(184)	Dhiálekse		{opjodhípote/kanéna}			vivlío.
	choose.2sg		FCI /	NPI		book
	‘Choose any book.’
	With kanéna vivlío: ‘Choose some book or other.’
(185)	Opjadhípote ghata kinigai pondikia.
	Any cat hunts mice.

For the differences between FCIs and Greek NPIs in non-veridical contexts, see Giannakidou and Quer (2011), and Giannakidou (2011).

6.8 Epilogue

Greek and English, both Indo-European languages, obey the basic GQ syntax and employ quantificational determiners that select NP arguments. The two languages, however, were found to exhibit some interesting differences in the morphological make-up of quantificational determiners that, if adequately appreciated, can be instructive for uncovering what we can think of as the finer structure of quantification. One fact that needs to be singled out, and which impacts a number of areas, is the systematic employment of the definite article in quantifier composition. The involvement of the definite article in wh-formation and with universal quantifiers, has been a constant in the diachrony of Greek (Tzartzanos 1945). Regarding D-universals, if the suggestion that D expresses domain restriction (Giannakidou 2004, Etxeberria and Giannakidou 2010) is correct, then Greek grammaticalizes the contextual domain restriction argument, so domain restriction is not merely a factor in pragmatics. Concerning wh-words, the involvement of D can offer valuable guidance in assessing current ideas about the nature of quantification, especially when it comes to proposals that establish a link of ‘classical’ quantification with interrogative semantics via Hamblin alternatives (Kratzer and Shimoyama 2002). Any such attempt to use propositional alternatives as the source of quantification would be challenged by a language like Greek, where we see overtly individual-based operations on the wh-words, such as definiteness, domain restriction, or exhaustification.