The Role of the Article in Patterns of Modification in Greek

Definition

The present entry focuses on definiteness agreement patterns in Modern Greek by exploring the function of the definite article in structures that involve a head noun and a modifier—typically an adjective, but not only—that is also accompanied by the definite article. Such structures have been variously dubbed polydefinite noun phrases/Determiner Phrases (or, simply, polydefinites), appositions, (pseudo)partitives, and evaluative appositives in the numerous studies that have been put forward to account for them over the past four decades or so. The peculiarity of these structures is the very presence of the second definite article that shows up whenever the modified noun itself is definite; at first sight, this ‘second’ article seems redundant, or expletive. Polydefiniteness has earned a privileged position in the literature, as has been discussed extensively and in depth over the past thirty years or so. The aim of this entry is to provide i. a description of the patterns that involve a noun and a modifier accompanied by its own definite article, and ii. a comprehensive survey of the relevant literature by highlighting commonalities and differences across the basic studies that have been written about polydefiniteness.

1. Introduction

In the present entry, three main and very common modification patterns in Modern Greek (henceforth Greek) are discussed with the focus on the presence of more than one definite article in all of them. The three patterns under examination manifest what is called definiteness agreement, the primary representative of which are the so-called polydefinite nominals. The aim is to present these patterns outlining their idiosyncrasies, the role of the multiple definite articles, and the similarities and differences among them. Rather than proposing a new analysis for these patterns, the existing analyses are presented in the framework of the rich bibliography that accompanies them. The ultimate goal of the entry is to show that polydefiniteness is not an isolated phenomenon, as has been assumed, but part of a broader system of definiteness agreement under DP-internal predication.

2. The Definite Article in Greek and Nominal Agreeing Modifiers

The definite article in Greek displays functions commonly found in languages with definite articles. It can be referential (o maθitis perase stis eksetasis, the student passed the exams), generic (o elefantas ine θilasitko, the elephant is mammal) or have no semantic content (o Janis, the John) ([1,2,3,4,5,6], a.o.). The definite article that is devoid of any content is called an expletive (Section 5 and Section 6); in Greek, proper names, place names, geographic terms, generic terms, and nouns with unique reference are always accompanied by the expletive article.

Morphologically, the definite article, like all nouns, displays different endings for number (singular, plural), gender (masculine, feminine, neuter) and case (nominative, genitive, accusative) with a relatively high degree of syncretism. For instance, for the neuter singular, there are two endings—one for the nominative and the accusative and one for the genitive, both in the singular and in the plural. We shall tackle the function of the article in Greek in Section 5 and Section 6.

2.1. Modification in Noun Phrases: Agreeing and Non-Agreeing Modifiers

The term ‘modification’ refers to the syntactic function of a lexical item (‘modifier’) on another category (‘modifiee’), modifying its original (lexical) meaning. Modification is optional, since it is not required by the argument structure of the modifiee. This does not mean that the modifier does not change the meaning of its modifiee. Adjectives, nouns (or noun phrases—NPs or DPs henceforth), numerals, quantifiers, prepositional phrases (PPs), genitive noun phrases, adverbials and relative clauses (RCs) are all typical modifiers of nouns and form part of the inclusive NP/DP.

The basic patterns of nominal modification in Greek that constitute the focus of this entry are as follows:

• Adjectives (Adj henceforth) and nouns in noun phrases (NPs/DPs)

i. to/ena kalo pedi (the/a good kid); ii. to peði to kalo (the kid the good)/to kalo to peði (the good the kid).

B.

Nouns/noun phrases (NPs) and nouns in noun phrases

i. to/ena potiri to krasi (the glass the wine); ii. o aðerfos tis o vlakas (the idiot the brother.hers).

C.

Prepositional phrases (PP) and nouns in complex noun phrases

o cirios me to kapelo (the gentleman with the hat); i ɣata me to luri (the cat with the leash).

2.2. Adjectives as Agreeing Modifiers

In all the above three patterns, there is agreement between the items involved, which, in Greek, is a major feature of nominal modification. Agreement concerns number, gender and case, visible on both adjectives (the agreeing category par excellence) and nouns, but regulated by the noun. Importantly, certain cases of Adj-N modification also involve definiteness agreement, i.e., instances in which both Adj and N are accompanied by the definite article. Some patterns of noun–noun modification (specifically in pattern B) involve case and definiteness agreement only. Even in pattern C, in which a noun is modified by a PP, definiteness agreement holds, too, as we will find out in Section 7.

In (1)–(6), agreement between the article, adjectives and the noun is exemplified:

(1)	oreo	vivlio
	nice sg, neut, nom/acc	book sg, neut, nom/acc

(2)	kali	anθropi
	good pl, masc, nom	people pl, masc, nom

(3)	ɣriɣori	ekseliksi
	quick sg, fem, nom/acc	development sg, fem, nom/acc development

The definite article also participates in the agreement pattern indicated in (1)–(3):

(4)	to oreo	vivlio
	the sg neut nom/acc nice sg, neut, nom/acc	book sg, neut, nom/acc

(5)	i	kali	anθropi
	the pl, masc, nom	good pl, masc, nom	people pl, masc, nom

(6)	i	ɣriɣori	ekseliksi
	the sg, fem, nom	quick sg, fem, nom/acc	development sg, fem, nom/acc

Interestingly, apart from adjectives proper, special subcategories of nouns, nam., those that denote age, profession/property/capacity, or nationality/origin can also modify a noun for as long as they display varied gender and, as a consequence, agreement with the modified noun in case, gender and number allows them to behave like Adj: o stratiɣós patéras tu the general father of his (both nouns share the same gender—masculine—case and number), i piitria fili tu the poetess friend of his (both nouns are feminine and share the same case and number). Even geographical names and names of days can modify nouns, if, again, their morphology forces agreement with the modified noun:

(7)	Na vaftistume ston Iorðani potamo (or: ston potamo Iorðani)
	to baptize-1pl in-the Jordan river
	Let us baptize ourselves in the river Jordan. (potamos is masculine, as is the name Iorðanis.)

(8)	ta Varðusia ori
	the mountain range Vardusia (Varðusia is neuter plural as is ori)
	Cirjaci mera
	Sunday day (Cirjaci and mera are both feminine singular)

(9)	papas anθropos
	man priest (priest)

Despite the involvement of two nouns in (7)–(9), there is a single noun head—hence, one inclusive noun phrase and, consequently, only one referent. The same holds for patterns A and B illustrated in Section 2.1. Interestingly, this does not hold for pattern C, as we shall see in Section 7.

As mentioned already, apart from agreement in morphosyntactic features, Greek further manifests agreement in definiteness. In certain complex structures, one can find two occurrences of the definite article agreeing in gender, case, number AND definiteness. Consider (10), where the adjective and noun each appear with their own definite article:

(10)	to vivlio to oreo
	the book the nic
	the nice book

There is agreement across the board in (10): all four categories also agree in definiteness, as witnessed by the presence of two definite articles.

Definiteness agreement is a very interesting grammatical phenomenon found in certain languages across the globe, including Greek. When a definite noun is modified by an adjective, as in pattern A, a noun, as in pattern B or a prepositional phrase, as in pattern C, the definite article may (and often must) appear not only with the head noun but also with the modifier.

Instances like (10) are called polydefinite noun phrases, or simply polydefinites. The present entry capitalizes on polydefinites and shows that it is not an isolated phenomenon but part of a broader system of definiteness agreement under DP-internal predication. We will see in detail in Section 5 and Section 6 how agreement in morphosyntactic features and definiteness has been implemented syntactically and in what ways polydefinite noun phrases are/may be different from simple (=non-polydefinite, or monadic [7]) noun phrases.

The rest of the entry is structured as follows. In Section 3, polydefiniteness, along with its basic properties, is introduced, given its leading role in definiteness agreement phenomena in Greek, two more of which are also discussed in this entry. Section 4 explores modes of adjective interpretation and adjective ordering issues. Section 5 presents a number of existing accounts of polydefiniteness, while at the same time it discusses the major analytical issues that concern the properties of polydefiniteness and of the definite article that precedes the Adj. Section 6 introduces cases of nominal apposition which are closely connected with polydefiniteness, and Section 7 presents, for the first time, a novel instance of definiteness agreement consisting of a complex NP with a modifying PP. Section 8 summarizes and Section 9 concludes the Entry.

3. Polydefiniteness (Or Determiner Doubling) in Greek (Pattern A)

3.1. General Remarks

The distinctive feature of polydefiniteness is the mandatory definite article in front of the modifying adjective when the containing noun phrase is itself definite:

(11a)	*to spiti paljo	vs.	to spiti to paljo
	the house old		the house the old→the old house

On the other hand, when the noun phrase is indefinite, the postnominal Adj does not need any article—more accurately, no article is allowed.

(11b)	spiti paljo	vs.	*spiti ena paljo
	house old		old house
	(Cf. also the examples (1), (2), (3) above.)

The name ‘polydefinite’ (noun phrase/(nominal) (construction)) has been given to sequences like (10) by Kolliakou [7]; Androutsopoulou calls the same sequence determiner/definiteness spreading [8], while the name (determiner) doubling is also occasionally used. Polydefiniteness is typically used to denote a special kind of modification, namely, Adj-N modification; in fact, polydefinite nominals are the primary representative of pattern A (Section 2.1). It is our purpose in this entry to provide evidence that polydefiniteness is more than that; it is a pervasive phenomenon of Greek nominal syntax which embraces a number of constructions, all of which share the reduplication of the definite article (definiteness agreement effect), as listed in Section 2.1.

3.2. Basic Properties

Polydefiniteness characterizes mainly the oral language. It has attracted (and is still attracting) the linguists’ interest. There is an impressively rich literature on the matter—both with regard to Greek but also to other languages that display multiple articles in one and the same noun phrase [9]. Manolessou was the first to approach polydefiniteness from a discourse–pragmatics perspective. She argues that it is a construction that has an informal character, in comparison to simple noun phrases with a prenominal Adj [10], but it does appear in written speech. Nonetheless, what really matters is the kind of register polydefinites may appear in Manolessou’s terms, style and register [10]. While polydefinites are rarely—if at all—used in written formal registers (science, essays, etc.), they are commonly used in narratives, theatrical plays, dialogues of various sorts, as well as in the lyrics of songs, where they even violate constraints on their occurrence so that they can align with the tune or music. As Manolessou points out, however, not all authors employ polydefinites with the same frequency, so one should take into account specific authors, era, the exact literary kind, etc. Polydefinites are further regularly used in people’s names—first and second names—something that has not been much discussed or explained in the literature (Section 5).

Tsiakmakis, based on the results of experiments he performed, shows further that polydefinites encode not only familiarity but also expressivity “possibly captured in terms of an expressive speech act” ([11], p. 118). Another interesting external property of polydefinites is the fact that native speakers’ judgements hardly converge when it comes to the—often subtle—nuances of the interpretation of the construction, something that is not usually the case with other areas/phenomena of modification. Intuitions are often sloppy and the relevant data lies in a rather gray zone between “always X” and “never X”. This sloppiness of judgments is reflected in the big number of studies available today in the literature on polydefiniteness. It seems that the safest way to approach (aspects of) the phenomenon is to say that the interpretation of a polydefinite (as X or non-X) is a matter of a (strong tendency) or preference. It also becomes clear that, given the fuzziness of the landscape, it is necessary for the researcher to perform experiments and use the results in order to build up a theory, something that happened just a few years ago (cf. [6]).

Within a polydefinite noun phrase, apart from the order Art < N < Art < Adj, the reverse order is also common: Art < Adj < Art < N. A property of polydefiniteness is that it ‘frees’ the Adj-N combination from the strict adjective ordering that is discussed in the following section. There is no fundamental, truth-conditional difference between the two orders, other than the stronger/contrastive stress that the Art < Adj < Art > N sequence (usually but not uniquely) conveys. Incidentally, this is one point of disagreement amongst both native speakers and researchers; for some, it is the Art < Adj < Art < N order that is the unmarked (in the sense of it being more frequent), the order Art < N < Art < Adj being the marked (rarer) one.

It is further possible to have a prenominal Adj and a polydefinite sequence in one and the same noun phrase (often with a slight intonational break between the head noun and the Art + Adj sequence. (See ref. [12] for details and discussion.)

(12a)	to oreo spiti to paljo
	the nice house the old

The reverse order of (12) seems awkward for many speakers (including the author of this entry but excluding many others). See refs. [12,13] for discussion:

(12b)	??to paljo to oreo spiti (it gets better if there is a slight intonational break between paljo and to oreo)
	the old the nice house

Several questions arise at this point. The first is whether there is any difference in interpretation between the pre- and the postnominal position of the adjective—i.e., between a monadic and a polydefinite, something that would render the polydefinite a marked and context (or otherwise) dependent construction; related to this is whether the two positions, and, therefore, the including constructions are equivalent, or free variants of each other. The second question is how the definite article emerges in front of the adjective in the polydefinite, and what its role is. Finally, one more question is why no article is required when the containing noun phrase is indefinite. In Section 4, Section 5 and Section 6, the answers that have been given to these questions over the past decades will be reviewed.

However, before that, a survey of the different modes of interpretation of the adjectives as modifiers of nouns that are extensively discussed in the literature on Adjs is in order.

4. Adjectival Modification and Interpretation

4.1. Adjective Hierarchy and Ordering

In Greek, all Adj types appear to the left of the noun within the noun phrase without any restriction. This is the default (and unmarked) direction for Adjs in Greek (like in English). This also holds when more than one Adj modifies the noun. Crucially, in the latter case, Adjs seem to obey a ‘canonical’ hierarchical order, as seen and exemplified in (13):

(13)	quant > subjective > size > shape (…) > classificatory-relational/provenance > N
	pola panemorfa mikra persika ɣatacia
	many most beautiful little persian kitten
	([1,13,14,15,16], a.o.)

The nature, or origin, of this hierarchical scale—whether it is cognitive, semantic or simply syntactic—is far from being an agreed-upon matter. Neither is it a settled matter whether the scale is a linguistic universal [14,17,18,19], or whether it simply represents a preference. In any case, as Panayidou points out, at some point it does get syntactically encoded [13]. And, apparently, this is a piece of evidence on which the majority of syntactic analyses are based (Section 5).

An issue interwoven with the issue of adjective (sub)categories concerns the patterns of modification. Hierarchical ordering, as exemplified above, hinges on the pattern whereby every leftmost adjective modifies the entire following [Adj[Adj[Adj(…) N]]] sequence. This pattern represents what is called scope modification, a manifestation of what Sproat and Shih call direct modification, as opposed to indirect [16]. Scope modification is the pattern par excellence that reveals the hierarchical ordering. The neutral/unmarked ordering may change only if an adjective receives more emphasis over the rest of the series, thus escaping from its original position:

(14)	(ta) pola persika mikra ɣatacia
	(the) many persian little kitten
	(Here persika receives emphatic stress.)

One more reason for an apparent violation of the Adj hierarchy is the intended change in scope:

(15)	(prosfata) katepsiɣmena freska laxanika
	(recently) frozen fresh vegetable

(16)	freska (prosfata) katepsiɣmena laxanika
	fresh (recently) frozen vegetable

As we shall point out in Section 5 and Section 6, polydefinites circumvent the hierarchical ordering of Adjs and scope interpretation.

Apart from scope modification, there is also the pattern called non-scope (or parallel) modification, a pattern that in traditional grammar is called ‘asyndeton’ (loose coordination) (figure of speech): In parallel modification each one of the adjectives directly modifies the noun—[Adj, Adj, (Adj…) N]—ignoring all subsequent [Adj [Adj N]…] combinations. Furthermore, the conjunction ce (and) may appear between any two adjectives (instead of the comma that also appears regularly). Adjectives in parallel modification (‘asyndeton’) can (in principle) be iterated ad infinitum.

(17)	(ta) omorfa, paraksena, spania, (…), aksiaɣapita (…) ɣatacia
	beautiful, weird, rare, (…), lovable little-kitten

4.2. Interpreting the Pre- and the Postnominal Position of Adjs

Given that in Greek adjectives may also appear postnominally, both in definite and in indefinite noun phrases, the critical question is whether there is any difference in interpretation between the pre- and the postnominal position of the adjective. A closer inspection of the relevant data reveals that prenominal Adjs in monadics often modify the noun in ways that they do not in polydefinites. Moreover, and importantly, while prenominal Adjs in Greek are often interpreted ambiguously, Adjs in polydefinites are not. In all the analyses of polydefiniteness, the discussion is carried out in terms of a number of interpretational oppositions established in the vast literature on adjectives by both syntacticians and semanticists. One such distinction is that between restrictive and non-restrictive Adjs. Adjs, in general, modify the noun either restrictively, if they contribute to the identification of the (intended) referent of the noun by restricting its denotation, or non-restrictively (or appositively), if they contribute “to the delineation of the whole set of objects that constitute the extension of the noun. In other words, a non-restrictive adjective forms an intrinsic part of the reference of the noun itself, whereas the restrictive adjective asserts a property for some referent, which exists independently of this property” ([20], p. 153). There is consensus among the researchers that in Greek, prenominal Adjs are ambiguous between the two interpretations ([1,7,9,21,22], a.o.):

(18)	Pare mazi su tin cenurja zaceta.→(restrictive and non-restrictive)
	take with you the new jacket
	Take with you the new jacket.

In contrast, in a polydefinite, the adjective with its own article may not be interpreted non-restrictively. It always acts as a restrictive modifier:

(19)	Pare mazi su ti zaceta tin cenurja.
	take with you the jacket the new
	Take with you the new jacket.

In the following example, due to its lexical meaning, the Adj can only be interpreted as non-restrictive [10]:

(20)	Ores cituse to katalefko çoni.→(non-restrictive)
	hours watched the all-white snow
	(S)he was watching the wholly white snow for hours.
	[snow is inherently white and there is no other kind of snow to be selected]

Here follows a representative example given by Kolliakou:

[Context: the financial situation of the company is really very bad]

(21a)	O ðiefθindis ðilose oti i ikani erevnites		θa apoliθun.
	the director declared.3sg that the efficient researchers will		will be.fired.3pl.pass
	The director said that the efficient researchers will be fired. (monadic)
(21b)	O ðiefθindis ðilose oti i erevnites→(polydefinite)	i ikani		θa apoliθun.
	the director declared.3sg that the researchers	the efficient.		will be.fired.3pl.pass
	The director said that the efficient researchers will be fired [7].

The interpretation in the first version of (21a) is what Kolliakou calls the ‘bad luck’ situation (all researchers have to be fired, including, unfortunately, the good ones), while the interpretation of (21b) is the ‘insane’ interpretation, whereby it is the efficient researchers specifically (the restrictive interpretation) that will be fired, quite paradoxically. (21) shows that restrictiveness is a key property of polydefinites, as what this construction achieves is the restriction of the possible referents to one—the intended one. Etxeberria and Giannakidou also show that polydefiniteness is associated with the restrictive interpretation [2].

It must be noted at this point that not all scholars agree on the fact that Adjs in polydefinites are uniformly interpreted restrictively. For instance, Manolessou shows that very often polydefinites are used descriptively (non-restrictively) ([10], pp. 151–160). See also the discussion in [6,12,23] for slightly different views.

Another interpretational opposition concerns whether modification is intensional or extensional. In the former case, modification applies to the sense or intension of the noun [24,25,26]; in the latter, it applies to its referent. The sense, or intension, of a noun consists of a set of interacting components that include, a.o.:

—a specification for a time interval i, at which f is supposed to hold:

(22)	xtesinos proeðros
	yesterday’s president

—an indication of the possible world w, which provides the means for knowing whether f holds in this or another possible world:

(23)	psetfices iðisis
	fake news

—most importantly, a variable assignment function g, which determines the truth value of the final formula ([25], pp. 7–8):

(24)	ena kalo (=koftero) maçeri
	a good (=sharp) knife

Intensional A are only allowed in monadics, i.e., prenominally:

(25a)	(o) ipotiθemenos kleftis
	(the) alleged thief
(25b)	*? (o ipotiθemenos) o kleftis o ipotiθemenos
	(the alleged) the thief, the alleged

An interpretational distinction closely related to the intensional/extensional one is what Larson calls individual vs. stage level interpretation ([27,28,29]). The individual interpretation reflects a property that is inherent to the referent, permanent and stable. The stage level interpretation reflects a non-permanent property of the referent—a property that may change over time or circumstances. An individual-level property is a property that uniformly characterizes the referent, in particular, its primary inherent property:

(26)	orati asterismi
	visible constellations

(a. constellations that are always visible, b. constellations that are visible at a particular time or under specific weather conditions)

It has been noticed that in contrast with prenominal Adjs, in a polydefinite, they are unambiguously stage-level [1]:

(27)	(i orati) i asterismi i orati ine poli (: those now visible)
	The visible constellations are numerous.

The next example illustrates one more interpretational distinction, one which concerns the opposition between intersectiveness vs. non-intersectiveness. In the first example of (28) below, there is ambiguity between the two interpretations: the singer may have sung beautifully (non-intersective reading: beautiful as singer), or she is a singer and she is beautiful (intersective reading). In the polydefinite counterpart of (28), only one interpretation exists—the intersective reading, i.e., the intersection of beautiful things with the things that sing [27,28,29]. Consider the following example:

(28a)	Oli θavmasame xtes tin orea traγuðstria (the beautiful singer/her beautiful singing)
	all admired.1pl yesterday the.fem nice.fem. singer
(28b)	Oli θavmasame xtes (tin orea) tin traγuðstria tin orea. [the beautiful woman]
	all admired.1pl yesterday the.fem (nice.fem.) the singer the nice

Yesterday we all admired the nice (beautiful) singer. (Cf. [2], pp. 34–35, ex.50).

Summarizing this subsection, Adjs in the prenominal position are, in general, ambiguous between the two possible readings of a given interpretational distinction. On the other hand, Adjs in polydefinites, i.e., postnominal Adjs, regularly miss one of the two readings and are interpreted unambiguously.

4.3. Comparing Greek with Romance

The facts described in the previous subsection are reminiscent of what is discussed by many authors with respect to the Romance languages ([1,9,14,25,30,31,32,33,34,35], a.o.), but also English ([24,27,28], a.o.). At the same time, it is true that the different interpretation of adjectives in pre- and postnominal position (in different languages of the Romance group) is not handled by linguists in a uniform way.

For instance, it has been extensively argued that in Italian, prenominal Adjs are interpreted as intensional, individual-level, non-intersective, non-restrictive, while postnominal ones are ambiguous: they can be extensional, stage-level, intersective, restrictive, but also intensional, individual-level, non-intersective, and non-restrictive [14]. Interestingly, this situation is the reverse of the one we just noticed for Greek. In the former, it is the postnominal position that gives rise to ambiguity; in the latter, it is the prenominal. In both cases, then, it is the default position, the position that does not require credentials from Adjs, which conveys ambiguity:

Prenominal:

(29a)	Le invisibili stelle di Andromeda esercitano un grande fascino. (Italian)
	the invisible stars of Andromeda have a great fascination
	(Cinque [14], 7 individual level)
(29b)	Maria ha intervistato ogni possibile candidato. ([14], p. 9)
	Maria has interviewed every possible candidate
(29c)	Un buon attacante non farebbe mai una cosa del genere.
	a good forward not would ever a thing of-the kind
	([14], 10 non-inters)
	A good-hearted forward would never do such a thing.

Postnominal:

(30)	Un attacante buon non farebbe mai una cosa del genere. (non-inters. and inters.)
	a good forward not would ever a thing of-the kind

The case of brutale is discussed by Cinque [31] and Crisma [32,33], and also by Laenzlinger [35] for French.

(31a)	la loro aggressione brutale all’ Albania
	the their attack brutal in-the Albania
(31b)	la loro brutale aggressione all’ Albania
	their brutal attack to Albania (brutale has a speaker-oriented/subjective reading)

Some adjectives change their lexical meaning according to whether they are in pre- or postnominal position:

(32a)	de nombreuses familles (French)
	numerous families
(32b)	les familles nombreuses
	families with lots of children (or, more generally, members)
(32c)	#une nombreuse famille (incoherent: Bouchard [25], p. 82)
	a numerous family
(32d)	numerosas familias (Italian)
	many/various families
(32e)	familias numerosas
	multi-membered families

(33)	una gran familia (an important/famous family)
	una familia grande (a family with many members)

(34)	un viejo amigo (an old friend/a friend for a long time) (Spanish)
	un amigo viejo (an elderly friend)

(35a)	une ancienne église ([25], p. 73)
	an old church
	former(ly) church
(35b)	une église ancienne
	old, ancient church

In contrast to what Cinque claims for Italian—namely, that the postnominal position, which is the default position in Italian, gives rise to ambiguity—Bouchard maintains that in French, there is no ambiguity: “There is always a meaning difference between the prenominal and postnominal constructions in French. (…) This difference is regular, predictable, because it is compositional and derives from the fact that the adjective modifies different elements in these two positions.” ([25], p. 72).

Going back to Greek, the differences we saw between prenominal adjectives and adjectives in the polydefinite are, by and large, similar to the differences between pre- and postnominal adjectives in Romance languages: the set of modes of interpretation of Adjs in the polydefinite corresponds to one of the two sets of modes of interpretation that in Italian (at least) are available in postnominal position. Interestingly, the ambiguity found with postnominal Adjs in Italian is found with prenominal Adjs in Greek.

This state of affairs allows us to draw the following gross generalization:

Default position→Ambiguity (postnominal for Italian, prenominal for Greek).

‘Marked’ position→No Ambiguity (postnominal for Greek (polydefinites), prenominal for Italian).

By ‘default position,’ the unconstrained position is meant, the position that accepts all Adjs independently of category and/or interpretation. The position that imposes restrictions on the Adjs that occupy it is marked. In Greek, the marked position requires article insertion and disallows certain adjective subcategories, like intensional and classificatory adjectives. It is also generally accepted that it does not involve movement (either of the N or of the Adj—see below, Section 4.4). On the other hand, the consensus about prenominal Adjs in Italian (probably in all Romance) is that noun movement is involved for the surface order to be derived. One can attribute the difference concerning the emergence of the marked position of adjective placement to some parametric variation.

If the above generalization is basically in the right direction, it necessarily leads to the assumption that polydefinites are indeed a marked case of modification compared to monadics. We shall discuss this more in Section 4.4.

According to the above discussion, the directionality of adjectives relative to the noun appears to have an effect on the interpretation of the Adj + N vs. N + Adj sequence. An interesting question at this point is whether the differences that we saw above between an Adj in a monadic phrase and the ‘definite’ Adj in a polydefinite should be attributed to the postnominal position itself, or to the (obligatory) definite article of the Adj. The discussion so far suggests that it can well be the postnominal position. The comparison with Romance and the similarities in interpretation we found between polydefinites (i.e., postnominal Adjs) and postnominal Adjs in Romance (in one of the two alternative modes) is further evidence in favor of the assumption that it is the postnominal position that is responsible for the interpretation of the definite adjective. In Section 5 and Section 6, the discussion about the adjectival article will lend additional support to this assumption, while in Section 4.5, we will see that with indefinites we find the same interpretive effects as in polydefinites, a fact that strengthens the impact of position on the way definite adjectives get their interpretation.

Up to now, we have been using the terms pre- and postnominal (adjectives/positions), but we have not really explicated their exact meaning. We shall do that in the following subsection.

4.4. More on Pre- and Postnominal Adjs

The dichotomy ‘prenominal’–‘postnominal (adjective/position)’, unless these terms are simply impressionistically used to denote linearization of the adjective with respect to the noun, clearly hinges on one’s theory (i) about the basic position of adjectives and (ii) about movement (of either the adjective or the noun, or both).

In Alexiadou et al., the terms reductionist and separationist approach to the issue of adjective placement (with regard to the noun) are discussed, echoing relevant literature rooted in the sixties [1]. According to the reductionist approach, the pre- and the postnominal positions are somehow related. According to the separationist one, the two positions are kept apart.

Concerning merging positions of the Adjs, within the cartographic framework, represented for Italian by Cinque’s accounts [14,30,31], Adjective Phrases (and Relative Clauses for that matter) are merged prenominally, the former lower than the latter and thus closer to the noun. Noun movement from its base position to higher positions in the tree structure is presupposed in order for the surface order of the categories involved to be derived [14,30,31]. A different view, still within the generative framework and the reductionist move, is taken by Larson [27,28] and by Larson and Yamakido [36]; all adjective modifiers are generated postnominally within a right-branching noun phrase. They are oblique complements of determiners and rise higher if nothing can give them a case in their base position.

In contrast to the above accounts, Bouchard assumes there is no noun movement [25]. As mentioned earlier, the two positions of the adjective contribute a different reading each and are not connected (via movement) [25]. (See also Sadler and Arnold for a similar line of argument [37].)

Concerning Greek, there is consensus that there is no noun movement ([1,21,38,39,40] for discussion; Karanassios argues for the opposite view [3]). When adjectives are described as pre- or postnominal, reference is made to their surface order, which, for Greek, is also the base order. Crucially, the fact that the adjectives that come closest to and to the left of the noun on the adjective hierarchy (e.g., provenance, classificatory, see Section 4.1) never show up to the right of the noun is strong evidence that the noun does not rise to a higher position.

In Section 4.5, below, we consider Adjs in indefinite noun phrases, comparing them to polydefinites. Do polydefinites have an indefinite counterpart?

4.5. Indefinites

Apparently, they do. It has been argued that postnominal adjectives in indefinite noun phrases parallel the adjectives in polydefinites in terms of interpretation and distribution [9,40]. Specifically, prenominal adjectives convey ambiguity—just as in their counterpart in definite phrases—and adjectives to the right of the noun are unambiguous, just like adjectives in polydefinites. At this point, there is a caveat. As mentioned already in Section 3, when the larger noun phrase is indefinite, no article may accompany the adjective. When this Adj comes before the noun in an indefinite nominal, we do not, in principle, know whether it forms part of a simple (’monadic’) nominal or a polyindefinite one (see also further below in this section).

Turning to the interpretation of an indefinite nominal with a postnominal Adj and a prenominal one, consider (36):

(36a)	θa ðioksi ipalilus ikanus (restrictive)
	will fire employees efficient
(36b)	θa ðioksi ikanus ipalilus (ambiguous)
	will fire efficient employees

(36a) matches the situation in a polydefinite with a postnominal definite Adj. (36b) matches a monadic with a prenominal Adj.

As in polydefinites, intensional/modal adjectives are excluded from the postnominal position in indefinites, too:

(37a)	*(enas) kleftis ipotiθemenos
	a thief alleged
(37b)	*(enas) erotas (tu) monos
	a love (his) sole
(37c)	*(enas) ðimarxos proin
	a mayor former

Likewise with classificatory adjectives: they may not appear postnominally [9,40].

(38a)	*enerjia pirinici (vs. pirinici enerjia ‘nuclear energy’)
	energy nuclear
(38b)	*centro aθlitiko
	center sports (vs. aθlitiko kentro ‘sports center’)
(38c)	*leoforio sxoliko
	bus school (vs. sxoliko leoforio, ‘school bus’)

In the following examples, the postnominal Adj is interpreted restrictively and intersectively in both the polydefinite and its indefinite counterpart:

(39a)	i orea epistimon i kalopliromeni
	the beautiful.fem scholar.fem the well paid. fem
(39b)	(mia) orea epistimon kalopliromeni
	(a. fem) beautiful.fem scholar.fem well paid

In both (39a) and (39b), the Adj is intersective: the referent is both a scientist and a well-paid person. Consider also the following:

(40)	θeli na aɣorasi (ena) forema kalocerino.
	want.3sg TO buy (a) dress.neut. summer(ish).neut

(41)	To forema to kalocerino pu aɣorase tis kostise 200 euro!!
	the dress.neut the summer(ish).neut that bought.3sg her cost 200 euro

In both (40) and (41), the postnominal Adj is interpreted restrictively.

We thus have more evidence for the assumption that what is responsible for the ways Adjs in polydefinites is the (postnominal) position (see Section 4.2).

The difference between definite and indefinite noun phrases lies in the fact that in the latter, the postnominal Adj is articleless. This derives from the fact that indefiniteness in Greek has no realization. What looks like an indefinite article is in fact a numeral or quantifier and, as such, is optional (see Section 6.2). That means that when there is a prenominal Adj in an indefinite nominal, it is either an adjective generated within a monadic, or, in theory at least, moved there from a postnominal position (as in polydefinites next to to vivlio to oreo the book the nice there is to oreo to vivlio the nice the book)—a kind of a ‘vacuous’ polyindefinite. Could the way such prenominal Adjs are interpreted tell us whether the prenominal Adj has moved or is generated there? Probably not, if our generalization about the default position of Adjs in Greek is correct: the prenominal position is ambiguous. This problem is an important one, but also one that cannot be solved in the present context. But it needs further research, too.

Interim conclusion: Greek has both polydefinites and polyINdefinites. Postnominal adjectives have been shown to get the same (modes of) interpretation in both polydefinites and in indefinite NPs. This is an important conclusion because it sets both types of nominal on an equal footing and renders dispensable with the (old) claim that there is an asymmetry as far as adjective placement is concerned in the Greek noun phrase (postnominal Adj possible in an indefinite but impossible in a definite) [3,41]. Even more importantly, it cancels the need to assume (short) noun movement in Greek in order to account for the occurrence of postnominal adjectives in indefinite NPs, as has been assumed [3] (cf. Section 4.3).

All in all, we do find postnominal adjectives in both types of Greek noun phrases. Moreover, these Adjs are interpreted in parallel as regards all the modes of interpretation they are prone to.

Closing this section and in the light of the above discussion about polydefinites and their indefinite counterpart (‘polyINdefinites’), we are able to answer the question whether polydefinites are equivalent to (or free variants of) monadic noun phrases. If the reasoning in the previous sections is on the right track, what a polydefinite seems to be a free variant of is the monadic, in which the prenominal Adj has the same interpretation (or set of modes of interpretation) as the Adj in the polydefinite. For example:

(42a)	θa ðioksi tus ipalilus tus ikanus (polydefinite, only restrictive)
	will fire3 sg the employees the efficient
(42b)	θa ðioksi tus ikanus ipalilus
	will fire3sg the efficient employees

When in (42b) the prenominal ikanus acquires the restrictive interpretation, then (42a) can be seen as a variant of (42b); otherwise, (42b) will have more interpretations than (42a).

The relation between the two types of noun phrase (monadics and polydefinites) was first discussed by Kolliakou [7]; truth conditionally, the two are identical, but polydefinites have additional pragmatic contribution. Quoting from Kolliakou (ibid.: 264) [7]:

(…) the special pragmatic import of the [polydefinite, MS] originates from an additional contextual restriction on the anchoring of the index (…)

Therefore, one of Kolliakou’s main points of analysis is that monadics and polydefinites can be treated uniformly.

All that said, we conclude that polydefiniteness is indeed context-bound. This provides more support to polydefiniteness as a marked type of adjective modification. We can further say that polydefinites—given their context-bound character—are a subset of the set of Adj-N modifiers.

In what follows (Section 5), I shall try for as many as possible of the analyses in the extensive literature on polydefinites to be (re)presented. The review is based on materials accessible to me, and my plan, far from being a detailed representation and review of the content of each one of the analyses (an unrealistic goal anyway), will be to underline their basic argument and focus of inquiry to the effect that their categorization, as syntactic, semantic, or pragmatic, gets reasonable justification. Moreover, while remaining primarily descriptive, I nonetheless adopt a unifying perspective in which predicative apposition holds a leading role in accounting for definiteness agreement as a major phenomenon within nominal modification, and as manifested in polydefinites, appositive nominals and PP-modifiers.

5. Theoretical Approaches to Polydefiniteness

A number of approaches feature in the literature on Greek polydefinites. Each focuses on a different property of the construction, though all contribute insightful global analyses to the field. They can be grossly categorized as follows: (a) those that make explicit claims about the status and role of the adjectival article; these can be further distinguished into syntactic and formal semantic, (b) those that focus on the structural properties of polydefiniteness and can be further split in clausal, predicative and appositional, and (c) a more recent approach to polydefinites reflects the attempt to incorporate into the structure semantic–pragmatic concepts incarnated in syntactic categories.

Androutsopoulou ([8,42]) belongs to the first category. She assumes a structure in which all functional heads, except topmost D(eterminer), are expletive. The adjectival article lexicalizes the head’s ‘definiteness’, while the topmost article position remains empty. In this analysis, the differences between articled As and simple prenominal Adjs are not accounted for. Neither is it explained why, in polydefinites, multiple occurrences of definite articles amount to, or imply, expletiveness.

Kolliakou’s extensive and detailed analysis is mostly semantically and pragmatically oriented. Focusing on the grammatical status and the function of the article in front of the Adj, she claims that the article is a phrasal affix, akin to weak form possessives, and further associates it with partitivity: the polydefinite picks out a proper subset of a set that has been previously established in the discourse. Kolliakou asides with the view of the definite article (of the Adj) as a definiteness agreement marker rather than a determiner-like element ([7], p. 280). She also makes the following interesting claim:

[…] from a semantic point of view, it is clear that the contribution of the definite article [she means the main article—Author] in both monadics and polydefinites is integrated into the meaning of the sentence just once. For instance, in both to kokino podilato and to kokino to podilato, the same property ‘the red bike’ is uniquely instantiated in a local setting […] ([7], p. 281).

The above statement implies that the adjective article does not have independent semantic import to the polydefinite.

More recently, Etxeberria and Giannakidou, too, resonating with Kolliakou’s claim in their formal semantics analysis, argue that the reduplicated article of polydefinites is a domain restrictor and not a referential determiner [2].

The D with the noun seems to form the referential core of the structure, i.e., the DP that refers to an object. The combinations of D with the additional adjectives are non-referring and perform D_DR ([2], p. 34).

The authors also associate polydefiniteness with partitive interpretation:

Our take on this idea is that one D is referential, the other(s) perform DDR. While the D plus NP introduces a referent, the additional D combining with adjectives performs domain restriction, and the multi-D structure is akin to a partitive ([2], p. 34).

The above approaches share the view that the article of the adjective in polydefinites is essentially devoid of the content regular definite articles have—namely, referentiality. Androutsopoulou calls it an expletive, Etxeberria and Giannakidou call it a domain restrictor. Kolliakou considers it a weak clitic, so a grammatical item different from the ‘regular’ definite article.

Category (b) is the largest category of all since the majority of the accounts are syntax-oriented. There is a common trait of the analyses in this category: all of them attempt to account for the syntactic aspects of polydefiniteness in terms of a clausal structure generated internally to the noun phrase. According to Alexiadou and Wilder [43] and Alexiadou [9,38,39], this clausal structure is a (Reduced) Relative in the framework of Kayne’s original work [44]. It contains a noun (along with its own article) and a (definite) adjective that stand for the subject and the predicate, respectively (see [12,13] for criticism on specific points of this analysis). Alexiadou and Wilder do not make explicit claims about the adjectival article [43].

Panayidou, placing her analysis in the framework of the cartographic theory, in particular Cinque’s 2010 theory of adjectives, argues that Adjs in polydefinites are generated inside Reduced Relatives, which she takes to be predicative phrases [13,14]. The article in front of the adjective is not a real article but the realization of predication. Panayidou’s proposal encodes the presuppositional character of polydefinites, which she associates with a definiteness head obligatorily realized in a polydefinite.

Turning to the genuinely predicative approach, Campos and Stavrou argue that the polydefinite contains an internal predicative (small) clause. This assumption stems from the premise:

Polydefiniteness and predication:

The adjective that appears postnominally, in either definite or indefinite DPs, forms part of a predicative clause [21].

This does not mean that ANY adjective that can appear in a predicative clause may also appear in a polydefinite, an issue that features—explicitly or implicitly—in nearly all the literature on polydefinites. Consider (43):

(43)	ena vivlio akrivo	to vivlio to akrivo
	a book expensive	the book the expensive
	to vivlio ine akrivo
	the book is expensive

Campos and Stavrou [45,46], and also Guardiano and Stavrou [47], who take up this analysis in order to explain contact phenomena in Italiot Greek, further support the role of predication in polydefiniteness by endorsing in their analysis the structure for (various types of) appositives, and placing it within the general theory of Den Dikken [48]. Their analysis will be further discussed and justified in Section 6, following the presentation of apposition.

The appositional type of analysis put forth by Lekakou and Szendroi relies on the association with ellipsis. Polydefiniteness involves two definite noun phrases referentially co-indexed that build the larger, containing nominal; crucially, the head noun of the nominal that contains the adjective remains silent [49]. This is how this approach links polydefiniteness with (close) apposition [12,49]. In the same vein, Panagiotidis and Marinis pair predication (the two definite nominals of polydefinites stand for the subject and the predicate) with noun ellipsis. In particular, they take the [art + Adj ] constituent to function quasi-pronominally, given that the head noun is ellipted [22].

The primary problem of the accounts relying on ellipsis is that they have to stipulate the obligatory omission of the noun inside the elliptical DP. In general, in nominal ellipsis, the omitted noun may be pronounced (e.g., if only for stylistic (i.e., strong emphasis) reasons) [50]. In polydefinites, this option is not available—the noun that would otherwise follow the Adj is never pronounced.

According to Ntelitheos, the nominals in polydefinites are also related to ellipsis and, crucially, to focusing. Adjectives move to a Focus Phrase where they encode contrastive focus. Since the head of this phrase carries the feature [definite], under agreement, the adjective is also marked as [+definite] [51].

Larson and Yamakido argue that the ‘definite’ article before the adjective in Greek polydefinites is needed in order for the adjective to get case. The assumption for this is that postnominal modifiers (like Adjs) originate in a structurally ‘deeper’ position from which they need licensing; the article before the Adj is precisely a marker of such a licensing. For Larson and Yamakido, licensing is essentially case licensing; however, it is not clear why adjectives need case, since they get it under agreement from the noun they modify [36].

Recently, Tsiakmakis puts forward a novel analysis of polydefiniteness which crucially relies on i. the expletive nature of the adjectival article (that he dubs ‘resumptive clitic’) and ii. that Greek polydefinites “weakly but systematically convey the speakers’ social and emotional closeness towards the addressee.” ([6], p. 114) The novelty of this analysis is that the author endorses in the structure he proposes the ‘expressive’ content of polydefinites in terms of an additional speech act found in the framework of Cohen and Krifka [52] and Krifka [53] and subsequent work. Tsiakmakis’ work can be categorized as pragmatic.

The interesting convergence point of almost all of the analyses is the claim that the adjectival ‘article’ is not an ordinary article—it is different from the article of the larger noun phrase and performs a specific mediating function: as a linker, as a predicator, and as a licenser. In all these disguises, the adjectival article is expletive. We shall say more about the definite article in the following sections.

The topic of Section 6 is apposition proper, namely, a modification pattern that justifies the association of polydefiniteness with a special type of predication and comprises a bunch of structures of which all manifest case and definiteness agreement.

6. Nominal Apposition (Pattern B)

6.1. Pseudopartives and Definite Appositives

In this section, we shall linger for a while on nominal apposition, given its obvious and tight relationship with polydefinites, as almost all of the analyses mentioned here reveal.

In traditional grammar, apposition occupies a central role within the domain of noun modification; in particular, in the domain called agreeing (‘same case’) modification [54]. Unlike polydefinites, where a noun and an adjective are involved, in appositives, there are two juxtaposed nouns, one of which modifies the other. The distinctive feature of appositives is that they contain two definite nominals but a unique referent. The uniqueness of the referent is signaled by the overt case agreement between the juxtaposed nouns (and the accompanying articles).

In contemporary linguistic theory, the connection between nominal apposition and polydefiniteness dates back to Horrocks and Stavrou [41]. Horrocks and Stavrou assume that underlying the polydefinite o Andreas o politikos (the Andreas the politician ‘Andrew the politician’) is the equational copulative clause o Andreas ine o politikos (the Andreas is the politician). This connection has been revived in more recent studies of polydefinites, Lekakou and Szendroi’s being the one that makes this connection the main focus of their analysis [49] (also [7,12,22,51]), as mentioned already in Section 5. It must be pointed out, however, that not all the analyses of polydefiniteness as involving apposition view apposition the same way. Lekakou and Szendroi rely on data that involves two exclusively (i.e., with no indefinite counterpart) definite noun phrases (also [55]):

(44a)	o	solomos	to	psari
	the.masc.sg.nom	salmon.masc.sg.nom	the.neut.sg.nom	fish.neut.sg.nom
	(=salmon)
(44b)	o	aetos	to	pexniði
	the.masc.sg.nom	eagle.masc.sg.nom	the.neut.sg.nom	toy.neut.sg.nom
	the eagle the toy (kite)

and further:

(44c)	i	Mirsina	i	vasilissa
	the.fem.sg.nom	Mirsina.fem.sg.nom	the.fem.sg.nom	queen.fem.sg.nom
	(Queen Myrsini)
(44d)	o	Ioaniðis	o	kaɵijitis
	the.masc.sg.nom	Ioanidis.masc.sg.nom	the.masc.sg.nom	professor.masc.sg.nom
	the Ioanidis the professor (Professor Ioanidis)

Slightly different appositional data constitute the basis of Campos and Stavrou’s [45,46] and Guardiano and Stavrou’s [47] analysis, which echoes Den Dikken’s [48,56] approach to nominal predication (also Corver’s [57] and Bennis, Corver and Den Dikken’s [58]). Unlike the instances in (44), they correspond (a) to the pseudopartitive construction (45), ([11,59], also [60]), and (b) to the N-of-N construction in English (46)–(47) ([61,62] for English; Napoli for Italian [63]).

(45a)	to		matsaci		o		maidanos
	the.masc.sg.nom		sprig.neut.sg.nom		the.masc.sg.nom		parsley.masc.sg.nom
	the sprig of parsley
(45b)	i	konserva		to		spanaci
	the.fem.sg.nom	tin.fem.sg.nom		the.neut.sg.nom		spinach.neut.sg.nom
	the tin of spinach
(45c)	to		bukali		i		kolonja
	the.neut.sg.nom		bottle.neut sg.nom		the.fem.sg.nom		perfume.fem.sg.nom
	the bottle of eau de toilette

(46)	o	vlakas o aðerfos tu/o aðerfos tu o vlakas
	the.masc.sg.nom	idiot.masc.sg.nom the brother.masc.nom his/the brother.his the idiot.masc.nom
	his idiot brother

What we see in all of (44)–(46) is a noun in a definite phrase modifying another noun in another definite phrase (definiteness agreement effect). Crucially, the two articles and the two nouns agree in case, and usually, though not necessarily, in number and only accidentally in gender, as the examples in (45)–(46) show, along with (47):

(47a)	i	alepu	o	aðerfos tu
	the.fem.sg.nom	fox.fem.sg.nom	the.masc.sg.nom	brother.msc.sg.nom his
	the fox of a brother of hers
(47b)	to	ðjamandi	i	jineka tu
	the.neut.sg.nom	diamond.neut.sg.nom	the.fem.sg.nom	wife.fem.sg.nom his
	that diamond of a wife of his

The set of nouns in each of the examples (44)–(47) agrees not just in case (hence, the rubric ‘same case’ modification in the grammars), but, crucially, also in definiteness. In sum, the appositional examples of this section behave in many ways similarly to the polydefinites; the basic difference is the fact that in polydefinites the definite modifier is an adjective, while in appositives it is a noun (there are restrictions on the kind of the modifying noun in these examples—for instance, it must carry a lexical feature [+evaluative] or something like that—but we are not entering an analysis for those appositives in the present context).

The advantage of unifying the analysis of polydefinites and appositives, and as we shall see in Section 8, of complex NPs with a PP modifier, is that there is no need for extra assumptions or stipulations.

6.2. Structuring Polydefinites According to Appositives

Having seen the basic facts of appositional nominals, we can now outline—ignoring technicalities and details—the analysis proposed by Campos and Stavrou for appositive nominals in Greek [45,46]. In the appositive nominal (like (46)–(47)), there is a predicate phrase generated in a low position inside the larger noun phrase that contains the two definite nouns. The head of this phrase is Pred(icate). This is essentially a functional head and, as such, it agrees with the higher D(eterminer). The high D is definite in this type of appositives, so, under agreement, Pred, too, is definite. The two definite heads are lexicalized as the definite article o (the), since the feature [+def] is realized by the morpheme for the definite article (by default). The o (the) that also realizes the predicative head, which, as said, is also definite, is Den Dikken’s and Corver’s ‘spurious’ article [48,57]. Schematically:

The analysis Campos and Stavrou (2004) [21] propose for polydefinites (see Section 5) is basically the same as the one in (48) (details aside):

The critical part of this analysis of both polydefinites and appositives boils down to agreement in definiteness and case being the effect of the spell-out of the feature [+def]. More precisely, agreement in definiteness can be seen as an epiphenomenon: it is the combined effect of the presence of a nominal copula in the structure and of the default realization of the [+def] feature (as the definite article); this ends up—via the appropriate mechanisms—in case agreement between subject and predicate. Let us recall at this point Larson and Yamakido’s assumption (Section 4.4) that the definite article in polydefinites is there to give case to the deeply embedded adjective. It is true that the postnominal Adj in polydefinites starts out in a deeply embedded position [36]. On the other hand, as already pointed out in Section 5, it is somewhat unorthodox to say that a modifying Adj needs to be assigned case, because it gets it from the subject under agreement (see Section 2). The analysis offered in this section fares better because it captures this ‘need’ of Adjs via an analysis that endorses properties of the language that exist independently (the feature +def), copulative clauses. We could, of course, say that the article in the Pred position mediates the relation between subject and predicate, but this is the role of the copula generally.

Having said all that, there is a further interesting consequence of the proposed analysis. We concluded in Section 4.2 that the Adj in the polydefinite gets its interpretation due to its position and not to the article. Now, we can say that the article appears in the structure for polydefinites and appositives because, essentially, of the way the predicative position is lexicalized when it carries the feature [+def] under agreement with the high article—so, basically, due to the lexicalization of the definiteness feature through the definite article as default in Greek. The most we can say about the definite article in polydefinites is its supportive role to the adjective—it does not have any impact on its interpretation.

Let us next see what happens with postnominal Adjs in indefinite NPs, like

(50)

(ena) vivlio akrivo

‘a book expensive’

Following the basic reasoning of the definite examples, we get:

Det in (51) lacks the [+def] feature, being [-def]. This automatically results in D remaining lexically empty, a correct prediction given that indefiniteness in Greek is realized by a zero morpheme; what is called the indefinite article is not an article (see for a different view [3]), but rather a quantificational item ([9,38,64], a.o.). Pred, too, lacks the feature specification [+def], in agreement with the higher D, and is, as a result, empty too. The overall effect of the structure in (51) is INdefiniteness agreement between the predicative head and the higher D (again via a series of appropriate mechanisms/syntactic movements).

An interesting prediction of the predicative approach to polydefinites and appositives is that it blocks the recursiveness of both of them—taking for granted that a definite adjective appears only once in a (definite) noun phrase. This is due to the independent fact that there is only one (primary) predication per clause or per noun phrase. However, at the same time, it contradicts the assumption that is common to almost all the analyses—that in a polydefinite, the ‘definite’ adjective can be repeated. Further research, primarily experimental, will hopefully show which side is justified.

The structural parallelisms of appositives with polydefinites are eloquent. In both, there is a small clause (predicative phrase) within the inclusive noun phrase. The adjective is the predicate, selected by the predicator. The subject of predication is either in the predicative clause or higher, in either case external to the predicate itself. In both cases, the subject technically and really agrees with the predicator in (in)definiteness.

In Section 6.1, we noted that appositives like o solomos to psari (the salmon the fish) (exs 44) do not have an exact indefinite counterpart. The same is true of (46) and (47). On the other hand, there is an indefinite homologue to (45), and this is also a pseudopartitive (52a) ([13,59,60,65,66,67] and references therein); (46)–(47), evaluative/emotional appositives according to Campos and Stavrou, may be paired with an indefinite counterpart but under certain conditions [45,46]:

(52a)	ena	bukali	kolonja
	a/one.neut.sg.nom	bottle.neut.sg.nom	perfume.fem.sg.nom
	a bottle of perfume
(52b)	?(mia)	alepu	jineka (also: jineka alepu)
	(a/one.) fem.sg.nom	fox.fem.sg.nom	woman.fem.sg.nom (woman fox)
	a fox of a woman
(52c)	vlakas ðiciɣoros (and: ðiciɣoros vlakas)
	idiot.masc.sg.nom lawyer.masc.sg.nom. (and: idiot lawyer)

In (52) the two nouns in ‘apposition’ agree in case and INdefiniteness, not necessarily in gender (cf. 52a).

We have, by now, a more or less clear picture of Patterns A and B of Section 2.1. This comprises polydefinites and different types of appositives. In Section 6.1 and Section 6.2, we provided the data that support a unifying analysis for the two patterns. There is no doubt that there are differences amongst all of the above modificational patterns ([9]), but there is also no doubt that their similarities are more and more important than their differences, allowing us to unify them in an appositional/predicative analysis. In Section 6.3, we shall make a brief comment on the role of the adjectival article as a nominal copula, thus rounding up the discussion of Patterns A and B.

6.3. The ‘Adjectival’ Article as a Nominal Copula?

We claim that the article in front of the Adj in the two patterns (A and B) we discussed so far must be considered as a predicative element—a kind of nominal copula. The question is, is that an idiosyncrasy of Greek? Are similar elements found in other languages? Let us notice first that in English there is the formative (term introduced by Jackendoff [65]) of that takes up this role: this monster of a guy, a pearl of a mother ([63]; Den Dikken’s [48] ‘spurious article’). Jespersen calls this ‘of’ partitive [62]). In Spanish, there is the preposition de (el burro de Sancho ‘the donkey of Sancho’) (di in Italian). All of them are functional elements, devoid of lexical content, but, crucially, they are all case assigners (or licensers).

As mentioned at the beginning of this entry, one of the functions of the definite article is expletiveness. It does not then come as a surprise that the predicative head in Greek nominal appositive structures reflects this basic property of the language: an element that carries case, has no semantic content, and which can serve several grammatical functions—as a predicator, a licenser, and a case assigner (as in Larson and Yamakido’s theory [36]; also [68,69,70]).

In the last section, we shall tackle one more modificational pattern that has escaped the attention of the scholars: definiteness agreement inside a noun phrase with a modifying prepositional phrase.

7. Complex Noun Phrases with a PP as Modifier (Pattern C)

7.1. The Data

We finally turn to pattern C listed in Section 2. It is represented by an underexplored phenomenon instantiating a novel pattern of definiteness agreement. Stavrou and Tsimpli were the first to notice that this pattern involves a weird kind of definiteness agreement whereby the noun phrase that is the complement of a DP internal P(reposition) obligatorily agrees in definiteness with the topmost article when the latter is definite, and under specific contextual conditions [11]. The relevant pattern is given in (53):

(53)	DP [DP D [def] NP [PP P DP[def]]]]

Let us consider the following dialogue:

(54)	Context: Enas cirios me mavri kabardina perimeni ekso ja ores.
	a man with black raincoat is waiting outside for hours

In the second mention of (54), in which the main article is definite (as it encodes familiarity), the noun inside the PP must be definite, too:

(55)	Ecinos o cirios me *(tin) mavri kabardina ine filos su?
	this the man with *(the) black raincoat is friend yours
	Is the man in the black raincoat a friend of yours?

Consider further the following example:

(56)	Context: Ena vivlio apo ɣnosto paleopolio puliθice xtes çilja evro.
	a book from well known antique shop was sold yesterday a thousand euro

(57)	To vivlio apo ??(to) ɣnosto paleopolio itan tu Tolstoi.
	the book from ??(the) antique shop was the Tolstoi.gen

The generalization seems to be that in the first-mentioned sentence, where the top article is indefinite, since the referent is new in the discourse, the embedded article is also indefinite—zero by default—as we saw above. In the second-mentioned sentence, where the top article is definite, the embedded article is definite, too. On the other hand, if supported by modification, the embedded article can have an interpretation of its own:

(58)	O cirios me to kapelo pu içe pari o aeras…
	the gentleman with the hat that was blown with the wind…

This is a paradoxical situation. One would expect that only the external noun phrase, in argument position, is interpreted according to context. The noun phrase that belongs to the modifying prepositional phrase is normally handled in the lower domain, since it is embedded under a preposition, which is a boundary for syntactic operations—it protects the embedded noun phrase from external influence. Let us note at this point that definiteness agreement of the kind just exemplified ‘is passed by’ if the entire DP is interpreted as generic:

(59)	I ɣata apo spiti ine xaðiara/*to’skase.
	the cat from house is cuddly/ran-away
	A pet cat is cuddly/*ran away.

7.2. The Analysis

Stavrou and Tsimpli account for this rather unexpected situation by assuming that the embedded noun phrase becomes visible due to the function of the properties of the predicative structure, which is presupposed, along with the properties of the Greek definite article (in particular its high degree of grammaticalization). Let us compare the third pattern of modification with the other two—A and B.

The obvious similarity with polydefinites and appositives is definiteness agreement between the two articles. On the other hand, there are a number of differences from the patterns A and B.

In the aforementioned definiteness agreement patterns, the two categories participating in the predication relation are both nominal, and in particular, either N-Adj or N-N. In the case of the complex noun phrase, a noun and a preposition are involved, two, at first sight, different categories. On the other hand, in both cases, the predicate is [-Verb]—it lacks verbal features. Moreover, in polydefinites and in appositive nominals, a single referent is implied (despite the presence of two noun phrases and two articles). In the complex NP-PP, there are (in principle) two referents, the primary referent and the referent of the embedded NP of the PP (man and raincoat in (54)).

The two nominals in all the other definite structures share the same case. In the complex NP-PP, there is a DP internal case assigner, P, and the two noun phrases have different cases (because there are two case assigners in the whole structure). The lower noun phrase gets accusative from the preposition that is a case assigner.

In contrast to polydefinites, where, as said in the previous sections, there are restrictions on their distribution, there are no (pragmatic or interpretational) restrictions in the case of the complex NP-PP.

Moreover, in the other definiteness agreement structures, the lower domain (subject or predicate) is not freely expandable through modification:

Appositives:

(60a)	*to bukali i orea/palia kolonjia
	the bottle the nice/old perfume
(60b)	*o vlakas o meɣalos aðelfos mu
	the idiot the big brother my
(60c)	?*ta plusia ta meɣala krati
	the rich the big states
(60d)	*o aetos to oreo puli (or *o meɣalos aetos to puli)
	the eagle the nice bird (or the big eagle the bird)
(60e)	to bukali i kolonia sto xriso peritiliɣma
	the bottle the perfume in the golden wrap

Stavrou and Tsimpli performed an offline acceptability judgment task to check the speakers’ preferences on the six conditions involving def-indef, def-def, def-indef, etc. The complex NPs were in subject position and the prepositions checked were me (with, together) and apo (from). The results showed that definiteness agreement with both prepositions is highly preferred, while the indef-def condition with the P me has a very low preference; apo has higher acceptability rates than me. Def-Def vs. Def-Indef was significant with me but not with apo; that means that when the higher definite article is not expletive, Def-Def is the default in the complex NP-PP examined (with both Ps). The most important findings referred to the Def-Null condition that received low scores in both Ps, and the Def-Null condition that was acceptable with a generic reading only (as a last resort).

In the rest of this section, we shall try to answer the question of how definiteness agreement between the two articles in such complex noun phrases is possible, given the intervening modifying PP.

First, as with pseudopartitives and appositives, a predication relation holds between the two definite nominals involved; therefore, here, too, there is a predicative clause inside the complex noun phrase. Moreover, the assumption shared by the majority of researchers, that the adjectival article has no semantic content, is also adopted for the case of the complex NP. Adopting a predicative analysis, we claim that the PP originates as the predicate of a small clause. The lower definite article appears because the predicative head agrees in definiteness with the higher determiner, just as in polydefinites and appositives.

The end result emerging from a set of assumptions and the corresponding syntactic mechanisms (the details of which need not occupy us here) is that a complex head P + D is created, and the preposition in combination with Dare phonologically merged: ap’ to (from-the) (but *mto is blocked for phonotactic reasons, as mt in Greek is unpronounceable (but cf. sto, ‘to/in/at.the’); Cf. also French au, aux; Italian agli, degli.

If in the Predicate Phrase there is an indefinite noun phrase, or a definite one with a semantically non-vacuous article, no ‘agreement’ issue arises, because the lower determiner will be interpreted according to context in its domain. In this latter case, it is P itself that lexicalizes Pred.

According to this analysis, all the structures that exhibit definiteness agreement—polydefinites and appositives on the one hand and complex NPs with a modificatory PP on the other, are aligned by the extension of the predicative analysis proposed for the former to the complex noun phrases of the type examined in this section.

An important conclusion that can be drawn at this point is that definiteness agreement, a pervasive feature of Greek nominal modification, is an effect of the high degree of grammaticalization of the definite article, a process that renders it one more morphosyntactic feature (cf. [68,69]). As already mentioned in Section 6.3, due to its high grammatical/functional character, the definite article is used as a means of lexicalization of the nominal predicative head that is assumed in the predicative analysis. In this section, we further saw that P can also lexicalize Pred, so P and the definite article are both morphemes that perform this grammatical function.

8. Summary

In this Entry, three common and central patterns of nominal modification in Greek were discussed. The first pattern we focused on was polydefiniteness, the reduplication of the definite article in front of a postnominal adjective. We saw that polydefinites have interpretational properties that differ from those of their monadic counterpart. Moreover, these properties are parallel to those of postnominal adjectives in Italian (and other Romance languages) (in one of the two sets of interpretation they have). We also found that polydefinites have an indefinite counterpart in which the adjective is not preceded by any article, though it is interpreted in the same mode as its counterpart in the polydefinite. In the light of these facts, we concluded that polydefinites are marked nominal modifiers, as compared to monadics, and their interpretation is due to their postnominal position rather than the article that precedes them, which, according to the majority of the analyses, is considered an expletive, without semantic content but with some grammatical role—broadly speaking, a mediating (licensing) predicative role. Reviewing the vast literature on polydefinities, it turned out that most of the proposed analyses appeal to the phenomenon of nominal apposition, another pattern of modification that displays definiteness agreement between the two nominals involved. Case agreement is also a typical feature of apposition—even if agreement in (the) other features fails, agreement in case is always active. The analysis we favored for both patterns is the predicative one, which, with no ad hoc stipulations, can unify polydefinite with appositive nominals.

In this Entry, a novel pattern was also presented, for the first time, which is not immediately clear how it fits in the rubric of agreeing modifiers. This pattern involves a complex DP that contains a PP modifier of the noun. In a paradoxical way, the main noun phrase and the noun phrase that is the complement of the modifying P appear to agree in definiteness (under certain conditions), so this complex noun phrase provides extra evidence for the pervasive role of definiteness agreement in Greek. It was shown that, with minor adjustments, the same analysis that is proposed for polydefinites and appositives can also account for such complex DPs.

All the above said, it is worth pointing out a difference between polydefinites on the one hand and appositives and complex NPs with PPs on the other. Polydefinites, as said in Section 3, Section 4 and Section 5, are marked instantiations of adjectival modification, being context dependent, and can be seen as a subset of it. On the other hand, appositives and complex NPs are not dependent on context. This difference can be seen as pragmatic/discourse dependent. A corpus-based approach is needed that will show how polydefinites are used in naturalistic data, preferably in well-defined genres, such as natural conversation, fiction, journalistic discourse, poetry, etc. That way, we will have a solid basis to better understand the constraints on the appearance of polydefinites and aspects of their use and interpretation that go well beyond formal syntactic constraints and (formal) semantic interpretation. Maybe that is the piece of the puzzle that is (still) missing and will enable us to complete the large picture, not only of polydefiniteness but also of the related constructions that it is inextricably related to.

9. Conclusions

Polydefiniteness, far from being an isolated phenomenon, is a more general property of Greek nominal syntax and, therefore, it is associated with a rich literature. It is also manifested in appositive nominals, with which it shares the feature of case and definiteness agreement. In both types of nominals, the adjective or the modifying noun is preceded by the definite article in agreement with the definite article that heads the entire phrase. The Entry argued for a unified analysis of these structures instantiating definiteness agreement, and further showed that another nominal—a complex noun phrase with a prepositional complement can be legitimately subsumed under the same umbrella as polydefinites and appositives.