The definite article in Romance expletives and long weak definites

This paper focuses on some issues involving expletive articles and long weak definites in Romance (mainly Spanish, Brazilian Portuguese and Catalan), in comparison to DPs that elicit a strong reading. We show the similarities between expletive definites and long weak definites, and we argue for an analysis in common to other polarity items in terms of polarity sensitivity. We reach the conclusion that the definite article in Romance comes in two variants: the referentially unique variant (to be translated as the semantic iota operator) and the polar variant, formally characterized with an abstract [+σ] feature, that encodes a weak bound reading (to be semantically translated by an existential operator).

Roughly speaking, strong DPs or regular definites show a cluster of properties that differentiates them from other DPs: they denote uniqueness (Kadmon 1990;Elbourne 2008;Roberts 2003), maximality (Sharvy 1980) and familiarity (Heim 1982;Simons et al. 2010), manifest scopelessness with respect to other operators (Carlson 1977), have no distributional restrictions within the sentence and combine with any sort of Noun (Longobardi 2001), refer to individual objects, show no discourse referential restrictions, and do not occur in existential/presentational constructions (McNally 1992). Long weak definites (Poesio 1994;Barker 2005), on the other hand, have been described as showing a contrasting cluster of properties: they denote non-uniqueness, manifest narrow scope with respect to other operators, have a restricted distribution to postverbal position, usually occur with relational nouns and with a PP expressing the possessor, and may occur in existential/presentational constructions. Finally, expletive definites, in the sense of Vergnaud and Zubizarretta (1992) are DPs that introduce a non-denoting determiner, show non-uniqueness, usually combine with body part nouns and express (extended) inalienable possession. 2 Let us illustrate these three types of DPs by means of the Spanish examples in (1), (2) and (3). Similar examples could be provided for Catalan. Brazilian Portuguese is illustrated from (4) to (6). (1) Strong reading Pedro limpiaba el baño todos los días. Pedro cleaned the bathroom all the days 'Pedro cleaned the bathroom everyday.' (2) Weak reading a. La mano del bebé cogía el dedo del cirujano. the hand of.the baby took the finger of.the surgeon 'The hand of the baby grasped the finger of the surgeon.' b. El conductor perdió el control del vehículo cuando explotó the driver lost the control of.the vehicle when exploded la rueda del camión. the wheel of.the truck 'The driver lost the control of the vehicle when the wheel of the truck exploded.' (3) Expletive reading a. Pedro se lavó la cara esta mañana. Pedro cl washed the face this morning 'Pedro washed his face this morning.' b. Pedro perdió el móbil esta mañana.
Pedro lost the cell this morning 'Pedro lost the cell phone this morning.' Notice that in (1) the interpretation corresponding to the overt DP el baño 'the bathroom' is that it exists a unique, specific bathroom that Pedro cleaned every day. The sentence in (2a) is true if the hand of the baby grasped not a specific finger but a finger of the surgeon, thus suggesting that the overt definite DP el dedo 'the finger' has a dependent reading with respect to the possessive PP del cirujano 'of the surgeon', which we will account for in terms of polar sensitivity. Similarly, the sentence in (2b) shows a postverbal subject with a bound definite DP: the definite expression la rueda 'the wheel' is interpreted as a bound variable, meaning a/some non-specific wheel of the truck. Finally, the meaning of the DPs in italics la cara 'the face' in (3a) and el mòbil 'the cell phone' in (3b) is that it corresponds to Pedro's face or to Pedro's cell phone.
This description reveals that, in contrast to strong DPs (cf. Abney 1987;Longobardi 1994;2001;Zamparelli 1995) long weak definites and expletive definites have interesting properties in common. Thus, we will pursue, in line with Beyssade (2013) and 2 Vergnaud and Zubizarreta (1992) use the concept of 'expletive' definite in semantic terms, to refer to the denotation of the definite article that specifies body-part nouns and extended inalienable nouns. They associate the meaning of expletive definites with that of generic expressions through a type-denotation of the Noun. See below, Section 3.2, for discussion on this issue. Longobardi (1994) uses the notion of expletive article to make reference to the morphosyntactic article that appears in many Romance languages, in contrast to English, in prenominal position of a proper name, in combination with generic plural expressions, and in combination with indefinite mass nouns. See Giusti (2015) for a review. Corblin (2013), a unified analysis for complex DPs involving a genitive phrase and expletive articles in Romance. From now on we will commonly refer to those DPs in italics in (2) and (3) as DPs with a bound reading.
In view of these claims, and given the common beliefs (i) that expletive articles are nondenoting determiners (Vergnaud & Zubizarreta 1992;Longobardi 1994); (ii) that weak definites (short or long) look like indefinites and presumably are close semantically to bare nominals (Carlson & Sussman 2005;Carlson et al. 2006); and (iii) that, in general, items that are optional are considered to have no meaning (Chomsky 1995;Groat 1995), in this paper we focus on three questions. We discuss: (a) what it means for a definite article to be semantically expletive; (b) to what extent the expletive and long weak reading of certain DPs is to be analyzed as a grammatical phenomenon; and, (c) whether the expletive/weak reading of certain DPs must be attributed to a semantic ambiguity of the N, referring to token objects and type entities, or rather to the existence of a bound reading of the definite article, a bound polar variant to be opposed to the strongly unique variant, commonly translated as a iota operator.
But before we deal with these questions we will show that an apparent optionality of the definite article, possible in some languages, must not be confused with the semantic expletiveness of the article.
As we have illustrated in (1) to (3), Spanish requires full DPs with an overt D in argument position. Interestingly, Brazilian Portuguese (BrP) 3 , by contrast, has been argued to allow bare nominals (BNs) in any argument position (Schmitt & Munn 1999 Full DP structures have been postulated in this language for canonical argumenthood: 4 if a language has determiners for canonical argumenthood, it may also have null determiners in the same structural position, which means that the D position is not optional, even though the article might not be expressed overtly at PF. 5 Now, this apparent optionality of the determiner is also observed in long weak definites and expletive definites. Consider in this respect the data in (5) and (6): the nominal expressions in italics also have what looks like an optional article instantiating the category D in syntax. These data show that BrP is a DP language, even though the article may not be overtly expressed, and that, in fact, true syntactic optionality of the syntactic category D does not exist.

(5)
Weak reading A mão do bebê agarrava (o) dedo do cirurgião. the hand of.the baby grasped the finger of.the surgeon 'The baby's hand grasped the finger of a surgeon.' 3 Whereas the data from Spanish corresponds to Peninsular Spanish, the data from Brazilian Portuguese corresponds to the variety spoken in the state of São Paulo (not the capital) and in the north of the state of Paraná. 4 See Munn and Schmitt (2005) for the hypothesis of a null D(eterminer) in BrP. Cyrino and Espinal (2015), following Ghomeshi et al. (2009), conceive canonical arguments in this same language as full DPs with a Number projection and a D projection above the NP, which saturate the predicate; thus, canonical arguments behave as both syntactic and semantic arguments. 5 Recall that a non-overt D has also been postulated for the licensing of bare plurals in object position in Romance (Longobardi 1994;Chierchia 1998 In order to address the three questions mentioned above, we will structure this paper as follows: in Section 2, we focus on the notion of semantic expletiveness, as applied to the definite article. We consider the semantic notions of uniqueness and expletiveness as applied to the definite article in expletive and weak contexts, and we show that "expletive" articles fail uniqueness because they are subject to a bound relationship. We support the assumption that the generation and composition of meaning of expletive articles and long weak definites is part of syntax. In Section 3 we discuss three approximations to the semantics of these constructions: Guéron's (1983;1986; anaphor binding, Vergnaud and Zubizarreta's (1992) non-denoting determiner and type-denoting noun, and Le Bruyn's (2014) unicity of the determiner and homonymy of the noun. In this section, we exclude a generic-type interpretation of the noun as the cause for expletiveness of the determiner. In Section 4, we show the structural similarities between expletives and long weak definites in the Romance languages under study. In Section 5, we present a new analysis of expletive articles and long weak definites based on the encoding of a possessive or "have" relationship either in the lexical entry of intrinsically relational nouns or in the syntactic structure of non-relational nouns. This proposal, formalized as a syntactic dependency between two constituents, postulates a binding relationship between a definite article and a c-commanding constituent that has the semantic property to which the definite article is sensitive. In this section, we aim at extending Espinal and Tubau's (2016a;b) account of the expletive head marker no, conceived as a polar item, to the functional category D. Finally, Section 6 concludes the paper.
Let us now move to Section 2, where we consider the notion of expletiveness. We show that there is no absence of meaning for what look like expletive and weak articles. We also show similarities between expletive definites and long weak definites.

Semantic expletiveness: Similarities between expletive definites and long weak definites
In this section we argue, along with Kayne (2016), against the idea that expletive elements are uninterpretable, that is, against the idea that they contribute nothing to the interpretation of the sentences in which they occur (Chomsky 1995;Groat 1995). We will show, furthermore, that in fact there are several similarities between expletive definites and long weak definites. These facts will pave the way for our proposal that the weakness of the article is to be characterized as the reflex of the fact they are bound expressions, to be presented in Section 4.
First, expletive articles have been identified with lack of uniqueness. On the one hand, Vergnaud and Zubizarreta's (1992) thesis for French is that expletive articles are nondenoting determiners, in the sense of non-referentially unique determiners. On the other hand, Beyssade's (2013) (following Coppock & Beaver 2012) claim for French expletive articles is not that they are semantically empty, but that they convey weak uniqueness: the definite article does not trigger a uniqueness presupposition but only a weaker presupposition, in which uniqueness depends on existence. 6 weak definites Art. 23,page 5 of 26 This property, i.e. lack of uniqueness, when applied to the Spanish example in (2a) implies that the baby grasped a finger of the surgeon, but not a specific one. Similarly, the example in (3a) entails that Pedro washed a face, which happens to be his face. By contrast, in (1) the interpretation is that Pedro cleaned some unique bathroom, conceived as a specific token entity. Therefore, the unspecific and possessive readings observed in definite DPs in examples (2) and (3) are inferable interpretations from the weak uniqueness presupposition hypothesis: expletive articles and long weak definite DPs don't presuppose uniqueness of tokens.
Second, and in relation to the lack of uniqueness, notice that both expletive articles and long weak possessive definites (in the sense of Poesio 1994;Barker 2005) may occur in existential/presentational sentences, a possibility excluded from DPs with a strong reading (i.e., DPs with a strong and a quantificational denotation). 7 This is exemplified in (7) and (8)  According to Milsark's (1974; findings, existential sentences generally allow weak determiners, but not strong ones, thus suggesting that the definite article preceding a proper name and the definite article preceding a long weak definite are semantically weak. Third, in possessive definites the definite article can be claimed to be expletive or weak because it is bound to another constituent (either a subject, a dative clitic pronoun, or a genitive/dative DP) (Vergnaud & Zubizarreta 1992). That is, expletive definites with a (non-)relational noun in the head position, as well as long definites, are weak as long as they are bound (i.e., anchored, Poesio 1994). This is illustrated in the Catalan examples in (9) with a part-body noun (cara 'face'), in (10) with an extended inalienable noun (mòbil 'cell phone'), and in (11) with a long weak definite (el moll del port 'the pier of the quay'). In (10d) we assume that the possessive determiner binds the article at a certain point in the derivation (we come back to this issue in (40) below). Similar examples could be provided for Spanish. 9 7 See also McNally and van Geenhoven (1997). 8 Notice that Catalan-like Italian and Brazilian Portuguese-also allows expletive articles preceding proper names (Longobardi 1994). However, this use of the definite article cannot be accounted for in terms of anaphor binding, and must be distinguished from the weakness-boundedness of the definite article in combination with inalienable nouns and in long weak definite constructions. 9 It should be noted that exactly similar examples to (9)-(11) are not possible in Brazilian Portuguese because (i) in this language there are no third person clitics, hence (10a) has no exact counterpart, and (ii) there are no constructions such as (10c). As for (9), a counterpart in Brazilian Portuguese is possible either with the reflexive pronoun, or without it, but not both: weak definites Els espies es van trobar al moll del port. the spies cl past.3pl meet at.the pier of.the quay 'The spies met at the pier of the quay.' Examples (9) and (10) illustrate the fact that the most prominent interpretation is the one where the definiteness of the body-part noun or of the extended inalienable noun is bound by another expression: either the subject en Joan 'D Joan' in (9), the dative clitic li 'to him' in (10a), the genitive complement del professor 'of.the professor' in (10b), or the dative adnominal complement al professor 'to.the professor', as in (10c). Overall, the interpretation is like the one expressed by the possessive determiner, overtly expressed in (10d): the semantic interpretation is one in which the possessee is bound by the possessor, and this binding relationship is licensed by the principle of c-command, as will be shown in Section 4.2.
In the case of the long weak definite in (11) the syntactic relationship between the PP del port 'of the quay' and the DP el moll 'the pier' determine the reading according to which the definite article of this DP is also bound. Therefore, we conclude that long weak definites show a syntactic dependency similar to the one made manifest by inalienable body part nouns and extended inalienables.
Fourth, and related to the previous property, it should be noted that expletive definites, like long weak definites, have a narrow scope interpretation with respect to other operators and quantified expressions. This means, that in the examples below, from Spanish, everybody raised his own hand, and each car stops at the traffic light on the corner, but not necessarily a unique traffic light, because there may be more than one and each car may stop at a different one. Under these readings we can attach weakness to boundness, since the expletive article in la mano 'the hand' and the long weak definite in el semáforo  (12) a. Todo el mundo i levantó la i mano. every the world raised the hand 'Everybody raised his hand.' b. Todos los coches i se detienen en el i semáforo de la esquina.
all the cars cl stop at the traffic.light of the corner 'All cars stop at the traffic light on the corner.' The narrow scope interpretation of the bound definite is still more evident when we consider their occurrence in downward entailing contexts, such as in negative environments, or in conditional and interrogative sentences. Consider first the meaning of expletive and long weak definites in the Catalan negative examples in (13) (cf. Giusti 2015: 179), where the definite article in object position co-occurs with a postverbal negative polarity item. This construction could be exemplified in other Romance languages as well, for example Spanish and Brazilian Portuguese.
(13) a. No enguixaré la cama de cap jugador. not plaster.fut the leg of any player 'I won't plaster the/any leg of any player.' b. No van veure la pel·lícula de cap director argentí.
not past watch the movie of any director Argentinian 'They did not watch the/any movie of any Argentinian director.' Examples (13a-b) illustrate the fact that the definite article of the DPs la cama 'the leg', la pel·lícula 'the movie' have a polar status, as shown by the fact that they do not stop an operation of negative concord from applying between the preverbal negative marker no and the postverbal negative polarity item cap 'any'. In fact, in these examples the definite article can itself be interpreted as a polarity item: 'I don't want to plaster any leg of any player' and 'They did not watch any movie of any Argentinian director'. 10 Similarly, in (14) we consider the possibility of a bound reading for the expletive and long weak definite in the context of other downward entailing contexts such as conditionals and interrogatives in Catalan, which shows that these definites we are considering have a polarity sensitive behavior with a narrow scope with respect to the conditional or interrogative operator. 11 (14) a. Si enguixo la cama de cap jugador, … if plaster the leg of any player 'If I plaster the/any leg of any player…' b. Has vist la pel·lícula de cap director argentí? has watched the movie of any director Argentinian 'Did you watch the/any movie of any Argentinian director?
Finally, notice that expletives, as well as long weak definites-as we have already mentioned-introduce a dependent relationship between two nominal expressions, the possessee being c-commanded by the possessor. 12 This phenomenon is illustrated by means of 10 See Poesio (1994: 285) for a similar claim, according to which in weak definites the definite article may have an existential, a free-choice and possibly also a negative polarity reading. 11 Notice that this construction is not possible in all Romance languages most probably due to the different characterizing properties of their n-words, as postulated in Espinal and Tubau (2016b). 12 Recall that this property was originally described for expletives by Vergnaud and Zubizarreta (1992) and for long weak definites by Poesio (1994).
the Spanish examples in (15), although examples from other Romance languages (French, Catalan and formal Brazilian Portuguese) could be provided as well.
(15) a. A los soldados i , les i radiografiaron la(s) i rodilla(s). to the.pl soldiers them X-rayed the.pl knee(s) 'The soldiers' knees were X-rayed.' b. La mano del niño cogió el i dedo del cirujano i . the hand of.the baby grasped the finger of.the surgeon 'The baby's hand grasped the finger of the surgeon.' In (15a) the dative clitic is understood as the possessor of the knees, whereas in (15b) the surgeon, although also understood as the possessor of the finger, does not directly c-command the possessee. (We will provide a solution to this problem in Section 4).
To sum up, so far we have shown that so-called long weak definites and expletive definites fail the uniqueness presupposition, in that they do not presuppose uniqueness of tokens, and express a possessive relationship. Therefore, expletive and weak articles are not devoid of meaning, since they have a bound interpretation. We have also shown that long weak definites and expletive definites may occur in presentational sentences, in the scope of quantifier expressions, and in c-command domains.
Given these considerations, we exclude the hypothesis that the indeterminate interpretation of long weak definites and expletive definites might be due to an ambiguity of the Noun, and we propose that the generation and composition of meaning of expletives and long weak definites is syntactically-driven and must be described as part of the grammatical system.

Three approximations to the semantics of constructions expressing inalienable possession
In this section we will review three previous approaches to the syntax and semantics of constructions that contain expletive articles and we will finally postulate the hypothesis that the Romance definite article comes in two variants: a lexical item that encodes a strong unique reading and a lexical item that encodes a weak bound reading. Guéron (1983; Jaqueline Guéron focuses on the possessive use of the definite article in French sentences of the sort in (16).

3.1
(16) Jean a levé la main. Jean has raised the hand 'Jean raised his hand.' This example is ambiguous between two readings: Jean raised the hand of a third person individual object (e.g., a doll), or Jean raised his own hand (the possessive reading). The first reading is what we have referred to as the strong, unique one, while the second reading corresponds to the anaphor binding one, which is to be associated with the concept of inalienable possession.
Guéron's proposal starts from the observation that the article in French shows morphosyntactic similarities with a third person clitic pronoun (it must be specified for grammatical Gender and Number: le livre 'the book', la rose 'the rose', les enfants 'the children'). It also starts from the hypothesis that the definite article in French participates in an anaphoric relationship by which the antecedent must c-command the anaphoric DP that contains a body-part NP. Guéron (2006: 591), furthermore, distinguishes between three constructions, exemplified in (17)  In (17a) the direct object denotes a body-part and the subject denotes the possessor. In (17b) the direct object denotes the body-part, but a dative nominal clitic denotes the possessor. In (17c) the direct object denotes the possessor while the body part DP is embedded in a PP adjunct to the VP.
Constructions I and II, exemplified in (17a) and (17b), exist both in French and in other Romance languages, but not in English, whereas construction III, exemplified in (17c), exists in both French and English.
Constructions I and II, which express inalienable possession, a case of Argument-binding, are subject to syntactic constraints: obligatory antecedent/possessor, locality (the possessor must be in the same minimal sentential domain as the body-part DP), and asymmetric c-command (the possessor must c-command the body part nominal or its trace). Interestingly, French differs from English in that the article in French, but not in English, is specified for Gender and Number features, and in French, but not in English, it can be interpreted as a pronominal element. 13 By contrast, construction III does not introduce a contrast between French and English, because it illustrates Adjunct (A'-) binding: the body-part DP is within an adjunct to the VP or sentence that contains the possessor DP.
Relevant to our analysis is the idea that the article in Romance (French in particular) can be interpreted as an anaphor, but we will claim that his is not a consequence of the article having morphosyntactic features (a property that can also be found in Dutch and German articles) but of its being characterized with an inherent semantic feature that specifies its polar status. See below Section 4.2.

Vergnaud and Zubizarreta (1992)
These authors focus on the expletive reading of non-denoting determiners, but for them the crucial factor is not the morphology of the article (having morphosyntactic features of Gender and Number, like a third person clitic pronoun in French), but a semantic con- John wouldn't lift a finger to help. b.
Can you give me a hand?
We thank a reviewer for pointing out to us that this binding relationship of the indefinite article in English is not limited to downward entailing contexts, as the following examples form the British National Corpus show: (ii) a.
The crewman waved, seaman to seaman, and Trent raised a hand in farewell. b.
As they approached, they could hear that he was whistling merrily as he walked and, upon seeing them, he raised a hand and gave them a frank open smile. Notice that, once we agree on the weakness of long weak definites and expletive articles, either we assume that this indeterminate interpretation is due to an ambiguity of the noun, referring either to a name type or a token (Vergnaud & Zubizarreta 1992;Beyssade 2013), or to an ambiguity of the Determiner, having either an expletive/weak meaning or a strong one (Longobardi 1994;Klein et al. 2013). According to the first view a semantic distinction is made between referential expressions that give rise to a type interpretation and those that give rise to a token interpretation: a type would correspond to a bare NP or DP headed by an expletive determiner (the situation exemplified by French), whereas a token would correspond to a DP headed by a non-expletive determiner (the situation exemplified by English). Vergnaud and Zubizarreta claim that type interpretations have a flavor of genericity, as suggested by one of the interpretations of the sentences with même (in French) and same (in English). Consider (19) (65)), under the reading that there may be a plurality of computers or cars which are of the same kind and appearance.
On a donné le même ordinateur à Sophie, à Justine, et à Cléa. someone has given the same computer to Sophie to Justine and to Cléa 'Someone gave the same (type of) computer to Sophie, Justine and Cléa.' I drive a Ford Falcon, and Tom drives the same car.
Moreover, Vergnaud and Zubizarreta (1992: 598) identify an NP with a type construal as a target of distributivity, and they identify distributivity with inalienable possession. Thus, in (21) the plural dative triggers a distributive construal of the type body-part noun: although in (21a) (their example (5a)) the noun estomac is singular, it is interpreted as having a plural reference, as made clear in (21b) (their example (4a)).
(21) a. Le médecin a radiographié l'estomac aux enfants. the doctor has X-rayed the.stomach to.the children 'The doctor X-rayed the stomach to the children.' b. Le médecin leur a radiographié l'estomac. the doctor them has X-rayed the.stomach 'The doctor X-rayed their stomach.' Let us consider some difficulties with this approach.
First, type-denotations are claimed to correlate with non-uniqueness. Note, however, that the problem with this association is that the non-unique reading of the determiner is in fact independent from the type analysis of the noun: (i) Non-uniqueness is a property of weak and expletive DPs, while type-denotation is a property of NPs.
(ii) Non-uniqueness is a property of DPs that are bound to other DPs or QP expressions, whereas the (sub)type reading of the Noun may or may not combine with a bound reading of the definite article. Thus, in (22) the type reading of the noun fusil 'rifle' is independent from the bound reading of the definite article in italics.
(22) A los soldados, les retiraron el fusil / el mismo tipo de fusil. to the.pl soldiers them took the rifle the same type of rifle 'Rifles were taken from the soldiers.' Second, regarding the type-token distinction and the generic interpretation, it is assumed that in English, where D functions as a full-fledged determiner (i.e., as an operator), not as an expletive, the presence of the definite determiner generally indicates that the nominal expression denotes a token, whereas in French, where the definite determiner may function as an expletive from the point of view of its denotation, DPs may refer to types. However, not all nominal expressions with an expletive article in French have a flavor of genericity. On the one hand, the test with même cannot apply to body-part nouns. On the other hand, it should be noticed that the attributed genericity that expressions with même are supposed to illustrate is distinct from the generic reading assigned to strict kind-referring expressions (i.e., arguments of kind-level predicates), encoded by default in Romance by means of definite DPs that have no Number specification (Borik & Espinal 2015). Consider in this sense (23) in Spanish: the DP in italics refers to the class of objects denoted by the noun i-pod, without considering any particular instantiations of it. 14 This kind reading is incompatible with the même/same test. In this example the kind reading of the DP el i-pod is incompatible with a bound reading, in such a way that this example cannot mean either 'Steve Jobs invented his i-pod'.

Le Bruyn (2014)
Bert Le Bruyn postulates a unified analysis of the definite article in French inalienable possessive constructions and English alienable constructions, and assumes that semantically the definite article is to be translated always as a iota operator. He faces the uniqueness challenge of the definite article in expletive constructions at the cost of postulating for relational nouns like hand a four ways' homonymy between a basic non-relational interpretation, a standard relational interpretation, an implicitly relational interpretation and an abstract relational interpretation. This complex lexical semantics for relational nouns is accompanied with a set of type-shifts that account for their combinations in the possessive domain with the definite article, the indefinite article and possessive determiners. Furthermore, Le Bruyn (2014: 328) faces the challenge of cross-linguistic variation by postulating a new relationality constraint defined as in (24).

(24)
The relationality constraint Definite inalienable possession constructions are only available when it is independently made clear that their nominal component has to receive a relational and not a non-relational interpretation.
Notice, however, that this lexical constraint is restricted to nouns with a relational interpretation, whereas-as discussed in Section 2-a common feature of expletive articles and the article of long-weak definites is that the possessor must c-command the possessee beyond its potential lexical specification for a relational interpretation. On the other hand, postulating that definite inalienable possession is semantically (un) constrained in English as it is in French, Le Bruyn cannot account for the morphosyntactic differences that exist between the in English and the pronominal like (anaphor bound) status of the article le/la/les in French. We take this to be an advantage of Guéron's approach over Vergnaud and Zubizarreta's and Le Bruyn's approaches, and it leads us to postulate that the Romance definite article comes in two variants: a lexical item that encodes a strong unique reading, and a lexical item that encodes a weak bound reading. 15 This hypothesis of ours is supported by the morphosyntactic and semantic contrasts between articles in French and English, which correlate with the referential properties of nominal expressions being identified on the determiner in French or on the noun in English (Bouchard 2002;Dobrovie-Sorin 2012). Furthermore, this hypothesis is supported by the fact that in English an overt possessive D is required while in French a reflexive pronoun or a dative pronoun instantiates a binding relationship; in Romance there is no need to postulate a null possessor since the article specified for phi-features of Gender and Number and interpreted as a bound polarity item should be contrasted with the overt possessive determiner that licenses a non-bound reading. 16 In the next section we will argue that our analysis is mediated by two additional hypotheses: (i) the postulation of a possessive relationship, specific of the lexical semantics of relational nouns, which makes explicit that both possessor and possessee stand in a dyadic relationship already encoded in the lexicon; and (ii) the postulation of a binding relationship, reflected in the syntactic derivation of all those nouns (beyond relational ones) that occur with expletive articles and long weak definites. This binding relationship specified in terms of asymmetric c-command holds between a definite article that has an abstract [+σ] and an exhaustifier σ operator adjoined to a c-commanding constituent (Chierchia 2006;. The examples in Section 4 are from Catalan for convenience, but the postulated analysis is predicted to extend to other Romance languages.

Lexical relationship
What makes a noun a good candidate for being interpreted as part of an expletive or a long weak definite expression is that it conveys a possessive relationship (Poesio 1994;Barker 2005;Beyssade 2013;Corblin 2013). As pointed out by Partee (1983/1997), Löbner (1985, Jensen and Vikner (1994), Partee and Borschev (2003), Barker (2011), among others, a small set of nouns, denoting family relations (e.g., brother, cousin) and body parts (e.g., head, leg), are properly (or inherently) relational, which means that they involve a relationship between two participants (i.e., are transitive, as opposed to sortal nouns, which have one participant and are intransitive). 15 We are uncertain at this moment about whether such analysis might be extended to Germanic languages such as English, German and Dutch. 16 One of the reviewers wonders whether an alternative analysis might consist in postulating a silent possessor pronoun (e.g. '…the their knees' for a sentence such as ( We will hereby represent this idea by postulating that relational nouns introduce a possessive (or have) relationship between two nominal arguments. Thus, in the case of the Catalan example (25a) above, partially repeated here, the noun cara 'face' encodes the inalienable information in (25b).
(25) a. afaitar-se la cara shave.cl the face 'to shave one's face.' b. λx[face(x) ∧ R HAVE (person,x)] 17 We assume that in the lexical representation associated with so-called relational nouns a R HAVE must be postulated between two entities: a possessor/haver and a possessee/havee. This association is inherently encoded in the lexical structure of these lexical items, and it connects a head noun to another nominal expression, either within the same nominal domain (as in the case of long weak definites), or within the same clause (as in the case of expletive definites).

Syntactic dependency
Besides what we have said in the previous section, the possessor must c-command the possessee. This is straightforwardly illustrated in examples like (26) and (27)  Notice that the syntactic binding relationship specified by means of indices is obligatory in the presence of a subject, an overt reflexive, a dative clitic or an overt possessive D. If we add an overt possessive D to (26) and (27a) the meaning is redundant and the sequences become (to a certain degree) ill-formed. (See also footnote 16.) (28) #En Joan i es i va afaitar la i seva i cara. D Joan cl past.3sg shave the her face (29) #Li i vaig agafar el i seu i mòbil. to.him past.1sg take the his cell This fact suggests that there is no need for postulating a null possessive D in these constructions, and that a binding relationship already takes place between the c-commanding 17 An alternative representation follows Landman (2004) in postulating an existential have, whose semantics is represented as in (i), where the verb introduces relations and existential closure applies to the internal argument. See Le Bruyn et al. (2013).
subject, reflexive and dative pronouns, on the one hand, and the expletive article specified for phi-features, on the other. In the case of non-relational nouns such as mòbil 'cell phone' or moll 'pier' the syntactic structure may show the merge of the possessor and the possessee in such a way that a c-commanding binding relationship does not overtly take place between these two constituents. Consider once again the examples in (10b-c) and (11), repeated as (30) Els espies es van trobar al moll del port. the spies cl past.3pl meet at.the pier of.the port 'The spies met at the pier of the quay.' In (30) the noun mòbil 'mobile' is associated with the extended inalienable information that the cell phone belongs to somebody and that the professor is the possessor of the cell phone. However, this information is encoded in syntax, by means of a prepositional complement introduced by the prepositions de 'of' and a 'to'. Similarly, in (31) the nonrelational noun moll 'pier' relates to port 'quay' by means of a relationship that is syntactically encoded between a DP and its PP complement. However, notice that the prepositional complements that introduce the possessor in (30) and (31) do not to c-command their specifiers, according to a classical definition of c-command (Reinhart 1976;. In order to circumvent this structural problem, we assume, following Kayne (1993;2000)'s proposal for possessive constructions, that the possessor, complement of the preposition de (as well as the adnominal preposition a in (30b) in long weak definites) crucially c-commands the possessee. The same will be true for sentences containing expletive definites. In both cases, the syntax provides the required c-commanding relationship for the interpretation according to which the definite reading of the DP in long weak definites and the extended inalienable noun in expletive definites are bound by a c-commanding possessor.
Assuming Kayne's proposal, take a sentence with a long weak definite such as (31), which contains a non-relational noun. According to Kayne, prepositional possessors are higher than the possessee in the DP structure. 18 This means that the derivation of the DP el moll del port 'the pier of the quay' in (31) should be analyzed as in (32), where the possessor undergoes movement to the specifier of a prepositional head that is merged during the derivation. In (32) the possessor is c-commanding the possessee at the beginning of the derivation. Since in this sentence there is a non-relational noun (i.e., moll 'pier'), we assume that the possessive relation is instantiated by the merger of the possessive preposition de in the syntactic derivation.
Our syntactic analysis, whereby the possessor must c-command the possessee, is able to account for the two French constructions in (17a, b) (repeated here as (33)) that were pointed out by Guéron (2006). 19 (33) a. Jean lève la main. Jean raises the hand 'Jean is raising his hand.' b. Je lui prends la main. I him take the hand 'I take his hand.' In all of these constructions, the possessor c-commands the possessee.
Let us now consider if it also accounts for our data in downward entailing contexts (Ladusaw 1980). Consider (34).
(34) Negation a. No van veure la pel·lícula de cap director argentí. not past watch the movie of any director Argentinian 'They did not watch the/any movie of any Argentinian director.' Conditional b. Si enguixo la cama de cap jugador, … if plaster the leg of any player 'If I plaster the/any leg of any player…' Interrogative c. Has vist la pel·lícula de cap director argentí?
have seen the film of any director Argentinian 'Have you seen the/any film of any Argentinian director?' In these contexts, according to our analysis, the possessor c-commands the possessee once it is reconstructed back in LF to its original position.  19 Regarding Structure III (17c), a contrast should be highlighted between the sort of example given by Guéron (1983: 31, example (39a)) and the one given in Guéron (2006: 591, example (9b)). In (i) the body part DP can be said to be c-commanded by the preverbal clitic pronoun, but in (ii) the body part DP, being embedded in a PP adjunct to VP, is not c-commanded. We assume, along with Guerón (1983: 32), that examples as (ii) constitute a different kind of construction, where a part-whole (and not a possessive) relationship is established. In such cases, the PP is not an argument of the verb, and, differently from what happens in inalienable possession, the antecedent may be [-animate]: (iii) Je l'ai pris par la poignée. I it.have picked by the handle 'I picked it up by the handle.' See also Guéron (2006) for a review of different analysis for these cases.
In (35a) the negation scopes over both the polarity item cap and the definite expression c-commanded by the possessor, and so the latter can also receive a negative polarity reading. Likewise,in (35b,c), since the negative polarity item cap c-commands the definite expression, and both are under the scope of a conditional or an interrogative operator. Now, along with the lexical and syntactic relationships just described, the analysis we offer for expletives and long weak definites is also inspired on existing work on E(xpletive) N(egation).
It is well known that the term EN refers to a pleonastic (paratactic or redundant) use of the negative marker, when it does not modify the truth-value of the proposition in which it appears (Jespersen 1917;Vendryès 1950;Horn 2010). 20 Informally speaking, expletiveness has been analyzed as a semantic effect that obtains at L(ogical) F(orm) when the semantic property of specific syntactic constituents (either the negative marker no or some other type of negative expression) is absorbed by the semantic contribution of an expression in a non-veridical context (Zwarts 1995). What this means is that EN is conceived as a specific type of polarity dependency. 21 More specific and important for our purposes is the idea that an expression may be inherently characterized with a semantically strong scalar feature [+σ] that is a requirement for its interpretation as a (negative) polarity item (Chierchia 2006): 22 scalar items that carry an abstract [+σ] feature activate alternatives within smaller domains, hence they are felicitous in downward entailing contexts. Furthermore, the abstract [+σ] feature associated with scalar items is to be conceived of as an uninterpretable feature that needs to be syntactically checked by an interpretable abstract freezing operator σ attached, according to Chierchia (2006), to a clausal node. This checking operation locks in the strongest meaning in the context, preventing further recalibration of meaning.
With these ideas and machinery in mind, we postulate the idea that the definite article that occurs in so-called expletive definites and long weak definites in Romance is a polar syntactic item, inherently characterized with a [+σ] abstract feature. Being polar 20 This phenomenon is found in many languages in the complement position of certain prepositions, in complement of verbs that express fear, doubt, refusal, and in wh-exclamatives. Characteristic examples of this construction in Catalan are given in (i), with the optional expletive negative marker in parentheses and the lexical trigger of EN in italics: (i) a.
La policia evità que (no) hi hagués un accident. the police stopped that not cl had.subj an accident 'The police prevented an accident.' c.
Gasta més ell en tres mesos que (no) tu en tot l'any. spends more he in three months that not you in whole the.year 'He spends more in thres months that you in a year.' (Espinal 2002(Espinal : 2777 (136a)) Various theoretical attempts to account for the expletiveness of the negative marker in such contexts are found in the literature. See, among others, Espinal (1992a;b;2000;2007), Krifka (2010), Delfitto (2013) and Delfitto and Fiorin (2014). 21 On classical studies on polar sensitivity see Ladusaw (1980), Progovac (1992 and Giannakidou (1998). 22 Chierchia (2006: 556) developed an approach in which a polarity item like any basically means some, with the addition that it forces us to consider not only the current pragmatic context, but the largest contextually relevant domain. Thus, by saying I didn't find any book, the speaker includes not only, say, a contextually salient set corresponding to lost books, but a larger set including, for example, library and academic books. This means that it is domain-widening, in the present world and in other words one might consider. Furthermore, any activates alternatives within smaller domains (i.e., the statement is pragmatically stronger than smaller domain alternatives). See also Espinal and Tubau (2016a) for discussion on the relationship between expletiveness and scalarity. means that its meaning is dependent on and is sensitive to another constituent present in a c-commanding position in the clause. Its interpretation is domain widening, in the sense that by saying El soldado levantó la mano 'The soldier raised his hand' in the weak reading the speaker means that the soldier raised one/any of his hands, and its meaning is close but not identical to that of an indefinite expression. But, moreover, the fact that an expletive article shows a polarity behavior in negative, conditional and interrogative contexts (see (34) above) proves that it activates alternatives in smaller domains. Thus, by saying Has vist la pel·lícula de cap director argentí? 'Did you watch the/any movie of any Argentinian director?', the speaker makes a pragmatically stronger statement than smaller domain alternatives conveyed, for example, by Has vist la darrera pel·lícula de cap director argentí? 'Did you watch the last movie of any Argentinian director?'. We therefore postulate, in parallel to a microparametric approach of n-words (Déprez 1997 and ff.;Labelle & Espinal 2014) and of the negative marker no (Espinal and Tubau 2016a;b), 23 that the definite article in Romance comes in two variants: a referentially unique variant (to be semantically translated by means of the iota operator, Partee 1987) and a polar variant (to be formally characterized by an inherent σ feature and semantically translated by an existential operator). In particular, we postulate that the definite article of expletive definites and long weak definites may share the analysis of indefinite polarity items and the expletive negative marker, inherently characterized as [+σ]. 24 Applying this analysis to expletive and long weak definite articles we predict that the definite article of these structures is a polarity sensitive item semantically characterized [+σ], which requires a σ operator adjoined to a c-commanding possessor constituent. This adjunction operation is necessary to obtain an appropriate interpretation, one according to which the definite article is bound to a possessor expression and does not convey a uniqueness presupposition, as normally expected for the default use of the regular definite article; in these well-defined structures there is a possessor constituent on which the definite article depends on.
A final consideration regarding the last claim is relevant at this point. We have postulated that the possessor c-commands the possessee in sentences containing expletive and long weak definite articles, and that such relationship is established at the beginning of the derivation. Additionally, we postulate the reconstruction of this relationship at LF, so that a σ operator may adjoin to the left periphery of the DP projection, which will provide an output where the possessor will bind the definite article carrying the [+σ] feature.
To see how this works let us consider, once again, the data in (9), (10) and (11), repeated in (36), (37) Déprez et al. (2015) and Tubau et al. (2016) for some empirical studies that support the hypothesis that Catalan no is ambiguous between a strong negative reading and a weak non-negative polar reading, a position defended also independently by Longobardi (2014). 24 One of the reviewers enquires how the two lexical items for the bound and not bound readings are related to one another, and whether the strong definite could just be the unbound version of the polar definite. Technically speaking, if an item specified with the [+σ] feature is unbound the derivation would crash. Therefore, we conceive the bound and not bound variants as being independent lexical items.
Els espies es van trobar al moll del port. the spies cl past.3pl meet at.the pier of.the port 'The spies met at the pier of the port.' In the case of (36) and (37a), we assume an LF structure where a DP constituent c-commands the D bound head, the former the subject or dative clitic and the latter the bound definite article. In (39a) a bound reflexive clitic is also part of the complex chain. As for (37d), which has an overt possessive determiner, we assume that the possessive pronoun, the possessor, is higher in the structure at the moment of the merge with the possessee. Consider (40). In the case of (37b, c) we assume, as seen above, that the possessor, being higher at the beginning of the derivation and having this position reconstructed at LF, c-commands the possessee, whose D bears the semantic feature [+σ]. This is shown in (41). Finally in the case of (38), we can also identify the necessary c-commanding relation between the possessor and the possessee at LF, after reconstruction. Likewise, a σ operator is adjoined to the possessor and is able to bind the definite article in the possessee phrase. Consider the partial LF structure represented in (42). Overall these binding relationships follow the schema in (43), where a constituent α is c-commanding a constituent β at LF.
(43) α [σ] … β [+σ] One generalization that we draw from this account is that only possessor constituents may have the σ operator attached to them. This includes subjects, possessor complements (dative clitics, genitive and dative adnominal complements), complements of relational head nouns, and possessive determiners. Once the σ operator is attached to these constituents the bound constituent is not required to satisfy uniqueness and several entities seem to satisfy the descriptive content of the bound polar nominal expression, behaving similar to a variable or a polarity item.
In the particular case that the possessor contains a negative polarity item, itself specified with a [+σ] feature, a complex chain needs to be identified, as shown in (44)  A second generalization driven from our formal analysis is that, whenever an article is defined [+σ], a σ operator is to be adjoined to a c-commanding possessive or genitive constituent, otherwise the derivation would crash. A clear advantage of this approach is that, even though the feature [+σ] is semantic, the relationship that holds between the DP whose head has this feature and the constituent that anchors it is syntactic. This means that the analysis of expletives and long weak definites we pursue is syntactic in nature. Being a structural phenomenon, the high productivity of sentences with bound definites comes as no surprise.
To sum up, in the previous sections we have presented two different relationships that are required in a grammatical analysis of expletive and long weak definites. On the one hand, we have postulated a R HAVE in the lexical semantics of relational nouns (which include body part nouns). On the other hand, we have postulated a syntactic dependency for structures with non-relational nouns. This structure, following Kayne (1993;2000) and Chierchia (2006), establishes a c-commanding relationship between the possessor and the possessee that licenses a semantic abstract inherent [+σ] feature postulated for definite articles in expletive and long weak definite expressions.

Semantic representation
Strong definite articles have a unique reference: thus in (1), repeated as (45), the definite article refers to the unique entity that has the property of being bathroom and that Pedro cleans every day, as represented in (45b). In contrast to this example that represents a strong reading of the definite article, in this section we show how the lexical relationship encoded in 4.1 and the syntactic dependency encoded in 4.2 combine and lead to full semantic representations. Bound definites, inherently specified with the semantic feature [+σ], are non-unique and have a dependent reading, giving as output reference to an entity bound by an existential quantifier.
Let us first consider example (9), repeated once more in (46a). Notice that in the formula in (46b) the bound definite article in la cara 'the face' is not translated as a iota operator; rather, an existential operator introduces the variable x that is the theme of the event and the internal argument of R HAVE. Thus, the formula asserts that Joan shaved an object x that has the property face and that he possesses it. the case of non-relational nouns. As a result of the postulated binding operation a possessive relationship is entailed, and uniqueness fails. Furthermore, we have argued that, being dependent expressions, expletives and long weak definites, in similarity with other polarity items, instantiate a polar variant of the definite article, one that is characterized by an inherent [+σ] strong scalar feature and that encodes a weak bound reading (to be semantically translated by an existential operator at the level of semantic representation). This variant should be contrasted with the referentially unique variant (to be translated as the semantic iota operator) that instantiates a strong reading of the definite article.