Ensaio teórico ON DENOTING ABSTRACT ENTITIES FOLTRAN Maria José arthur@ufpr.br RODRIGUES Patrícia arthur@ufpr.br Universidade Federal do Paraná (UFPR) 12 1 ON DENOTING ABSTRACT ENTITIES 269 291 269-291 http://creativecommons.org/licenses/by/4.0/

This work is concerned with copular clauses in Brazilian Portuguese, as in (i) Maria bêbada é chato ‘Maria drunk is annoying’ and (ii) Crianças é divertido ‘children is fun’, in which the predicate exhibits an unmarked form for gender and number (masculine singular), in spite of the feminine and/or plural forms of the nouns in the subject constituent. We claim that a unified analysis is not available for such clauses. In the first case, the mismatching agreement pattern is the result of a neuter agreement with a small clause subject. In the second case, the lack of agreement is due to the presence of a DP subject which lacks agreement features.

 Resumo

Este trabalho examina construções copulares do português brasileiro, como (i) Maria bêbada é chato e (ii) Crianças é divertido, em que o predicado exibe uma forma não marcada para gênero e número (masculino singular), apesar da forma feminina e/ou plural do nome no constituinte em posição de sujeito. Defende-se que não é possível propor uma análise unificada para tais sentenças. No primeiro caso, a aparente falta de concordância é resultado da concordância neutra com uma small clause na posição de sujeito. No segundo, é resultado da presença de um DP defectivo com relação a traços de concordância.

 Concordância neutra Entidades abstratas Semântica Small-clauses Introduction

This paper deals with the semantics and the syntax of some copular clauses of Brazilian Portuguese (BP) of the kind shown in (1).

.

(1) a. A Maria bêbada é chato.

the FEM Maria drunk FEM is annoying MASC

‘The situation of Mary being drunk is annoying.’

b. Alunas que bebem é chato.

students FEM/PL who drink PL is annoying MASC/SING

‘The situation of students drinking is annoying.’

c. Crianças pequenas é divertido.

children FEM little FEM/PL is SING fun MASC/SING

‘Some activitiy involving little children is fun.’

.

The predicates of these sentences exhibit an unmarked form for gender and number (masculine singular), in spite of the feminine and/ or plural forms of the nouns in the subject constituent.

The sentences in (1) have a different interpretation from their agreeing counterparts, shown in (2).

.

(2) a. A Maria bêbada é chata.

the FEM Maria drunk FEM is annoying FEM .

‘Maria drunk is annoying.’

b. Alunas que bebem são chatas.

students FEM/PL who drink PL are annoying FEM/PL

‘Students who drink are annoying.’

c. Crianças pequenas são divertidas.

childrenFEM littleFEM/PL are funFEM/PL

‘Little children are fun.’

.

Concerning the reading of these sentences, we can observe that, in (1), the predicate applies to a situation or an eventuality involving the subject and, in (2), the predicate assigns a property to an individual. Hence, while the agreement in (2) is a case of trivial agreement between a predicate and a DP subject, the agreement facts in (1) claim for an explanation. To account for the neutral agreement of the sentences in (1), we will be concerned, in particular, with the following questions: a) can the mismatching agreement facts in (1) have a unified explanation? b) how is the mismatching agreement in (1) related to the observed readings? c) to which categories do non-agreeing subjects belong?

The answer to the first question is no. Our claim is that the subject in the sentence (1a) is a small clause (SC), and that the agreement is a neutral agreement of the kind observed with clausal subjects. The sentences in (1b-c), however, have a DP in the subject position, not a clause. Some works in the literature claim that this DP is in fact the object of an elliptic infinitive verb, and consequently, the observed agreement facts are also the result of a neutral agreement with a clause. We will argue against this hidden infinitive approach and defend, in accordance with Weschsler (2011) and Danon (2012), that the subject in these sentences are actually DPs which lack the necessary features to trigger agreement.

Regarding the actual readings of these sentences – the fact that the subject denotes or is involved in a kind of eventuality or situation to which the predicate applies – we will argue that they are a consequence of the semantic selection made by the predicate. In accordance with proposals that predicates take a semantic category as one of their arguments (Asher 1993; Hegarty 2003; Rochette 1988), we argue that the SC and DP subjects in these sentences are the realization of a semantic category – a situation – selected by the predicate; the SC, being sentential, can denote a situation, while the DP could refer to a situation by virtue of the absence of some agreement features – this would allow shifting the literal denotation of the subject to a contextual property associated to the noun.

The paper is organized as follows. Section 1 briefly presents Hegarty’s (2003) work about abstract entities. Section 2 discusses the properties of sentence (1a) and argues that its subject is a SC. Section 3 is devoted to the discussion of sentences (1b-c). We argue against the hidden infinitive approach to explain these sentences and assume the analysis of their subjects as DPs. We also present some of the semantic properties of these sentences. In the conclusion we discuss some remaining issues and present our final remarks.

 1 Abstract entities

Several studies have postulated different types of ontological categories to account for properties of arguments (Asher 1993; Hegarty, 2003; Rochette 1988). In these approaches, predicates take, as one of their arguments, a semantic category – a state, a proposition, an event, etc; – and each category presents a particular type of syntactic realization. We adopt this view to frame our proposal in this paper. In order to make it clear, we will briefly survey Hegarty’s (2003) work on abstract entities.

Hegarty, based on Asher (1993), places propositions, facts, situations, reasons, events on a spectrum of world immanence, “according to the degree to which they are entities in the world, interacting with other entities in space and time” (p. 893). For him, events have a high world immanence status, while propositions have a status of low world immanence. Facts, situations and reasons occupy an intermediate position on the spectrum. From this point of view, “situations have locations in time and space, rather like events and states, but situations include ramifications to an extent that events and states do not, making them less definitively bounded in space and time” (p. 892).

One way to distinguish these different kinds of abstract entities from one another is to look at the predicates with which expressions referring to these entities can combine. Hegarty (2003) proposes that predicates like (in)tolerable, (ir)reversible or can(’t) be remediated take a situation as argument, since a situation can include ongoing ramifications, lingering effects, which makes it reasonable to say that the situation is or is not, for instance, tolerable. On the basis of this definition, we can add to this list other predicates such as temporary and inconceivable. Therefore, in line with the studies on abstract entities, we assume that predicates can be grouped in classes on the basis of kinship in meaning, so that we can say that a class selects for a particular ontological category. Without adopting any specific assumption on a specific set of semantic categories and their canonical syntactic realization, we assume that the predicates (in)tolerable, (ir)reversible, temporary, inconceivable and similar ones belong to a semantic class that selects for a situation. Moreover, this semantic category situation can be realized, as we will discuss below, as a SC or a DP which lacks agreement features.

 2 The SC subject

In this section we examine the sentence in (1a), repeated here for ease of exposition. The only possible reading for this sentence is the one in which the adjective chato ‘annoying’ ascribes a property to a situation.

.

(1) a. A Maria bêbada é chato.

the FEM Maria drunk FEM is annoying MASC

‘The situation of Mary being drunk is annoying.’

.

As we can see in (3), the subject a Maria bêbada ‘the Maria drunk’ can combine with predicates that select for a situation.

.

(2) A Maria bêbada é intolerável / inconcebível / passageiro / irremediável.

the Maria drunk is intolerable / inconceivable / temporary / can’t be remediated

‘Maria (being) drunk is intolerable/inconceivable/temporary/ can’t be remediated’.

.

On the other hand, the examples in (4-6) illustrate the fact that this subject cannot combine with predicates that select only for an individual, like lazy or proud (4), nor with predicates that are more appropriate to events, like happen (5), or with predicates more appropriate to propositions, like true (6).

.

(3) *A Maria bêbada é orgulhoso / preguiçoso.

the FEM Maria drunk FEM is proud MASC / lazy MASC

.

(4) *A Maria bêbada aconteceu ontem.

the Maria drunk happened yesterday

.

(5) *A Maria bêbada é verdade.

the Maria drunk is true

.

For the sentence in (4) to be grammatical, there must be agreement between a Maria and the predicate orgulhoso ‘proud’ or preguiçoso ‘lazy’, as in (7)1.

.

(6) A Maria bêbada é orgulhosa / preguiçosa. the FEM Maria drunk FEM is proud FEM / lazy FEM ‘Maria drunk is proud / lazy.’

.

Therefore, we assume that the subject a Maria bêbada ‘the Maria drunk’ in (1a) denotes a situation, and, on the basis of the semantic selection concept discussed in section 1, we propose that the predicate chato ‘annoying’ can select semantically for a situation. Moreover, we argue that this situation can be realized syntactically as a SC. In other words, the subject of the sentence (1a) is the small clause [a Maria bêbada] ‘the Maria drunk’, which denotes a situation. In this way, assuming the widely held analysis that the copula selects a SC, the sentence in (1a) can be represented as in (8): the subject of the SC selected by the copula is also a SC. The agreement in this case is the kind of default agreement observed with clausal subjects.

.

(7) Ser [SC [SC a Maria bêbada] chato]]

Be [SC [SC the Maria drunk] annoying]]

That the sequence a Maria bêbada ‘the Maria drunk’ in (1a) is a constituent can be shown by some constituency tests, as in (9)2. In (9a), the sequence is replaced by a pronoun; in (9b), this sequence answers a wh-question formed with the expression o que ‘what’; (9c) shows the clefting test.

.

(8) a. Isso é chato.

this is annoying

‘This is annoying.’

b. O que é chato? [A Maria bêbada].

what is annoying? the Maria drunk

‘What is annoying?’ Mary drunk.

c. É [a Maria bêbada] que é chato.

(it) is the Maria drunk that is annoying

‘It is Mary drunk that is annoying.’

.

In short, it seems reasonable to assume that the sequence a Maria bêbada ‘the Maria drunk’ in (1a) refers to a situation and that it is a SC. To summarize our proposal, we argue that the predicate chato ‘annoying’ can select semantically for a situation, which can be realized as a SC. Hence, the SC selected by the copula in (1a) has another SC as its subject, which denotes a situation. It is the clausal subject that triggers the default agreement, as in any other clausal subject. In the next section, we will discuss the sentences in (1b-c). Unlike the sentence in (1a), these sentences have a DP in the subject position.

 3   Another way of denoting a situation

The sentences (1b-c), repeated here for convenience, also exhibit the same mismatching agreement observed in sentence (1a).

.

(1) b. Alunas que bebem é chato.

students FEM/PL who drink PL is annoyng MASC/SING

‘The situation of students drinking is annoying.’

c. Crianças pequenas é divertido.

children FEM little FEM/PL is fun MASC/SING

‘Some activity involving little children is fun.’

.

In these sentences, the constituents in the subject position are formed by nouns modified by a relative clause in (1b) and by an attributive adjectival phrase in (1c). Nonetheless, the predicates chato ‘annoying’ and divertido ‘fun’ apply to an understood situation rather than to the individuals alunas ‘students’ and crianças ‘children’. For instance, we can see in (10) that the subject of these sentences can combine with predicates that select for a situation, like intolerável ‘intolerable’.

.

(10) a. Alunas que bebem é intolerável.

students FEM/PL who drink PL is intolerable SING

‘The situation of students drinking is intolerable.’

b. Crianças (pequenas) é intolerável.

children FEM (little FEM/PL ) is intolerable SING

‘Some activity involving little children is intolerable.’

.

This kind of structure disallows then the possibility of analyzing these constituents as SCs; if they were SCs, the ungrammaticality of (11a-b), in which a pronoun substitutes for the noun phrase, would be unexpected. This shows that the modifiers and the noun form a single phrase, in this case, a DP.

.

(11) a. *Elas que bebem é chato.

they who drink is annoying

b. *Elas pequenas é divertido.

they little is fun

.

On the other hand, these sentences exhibit a subject internal agreement: the verb bebem ‘drink’ in the relative clause in (1b) is inflected for number and person to agree with the noun alunas ‘studentsFEM’ and the adjective pequenas ‘little’ in (1c) exhibits a marked form for number and gender to agree with the noun crianças ‘childrenFEM’.

These sentences, as we will see in the next section, are similar to Scandinavian ‘pancake sentences’ and to Modern Hebrew ‘ze-clauses’. The analyses presented in the literature for this kind of structure are twofold: a) those that consider these subject constituents as reduced infinitival phrases (Faarlund 1977 apud Enger 2004, Josefsson 2009); b) those that analyze them as a DP (Weschler 2011, Danon 2012). In this paper, we will adopt the second alternative: we claim that the subjects in (1b-c) are DPs.

The outline of this section is as follows. First of all, we will present the analyses proposed for this kind of construction on Scandinavian languages and Hebrew. Then, considering Brazilian Portuguese data, we will explain our analysis arguing against the hidden infinitival approach and discussing the semantic properties of the DPs in subject position.

3.1 Previous works

The Scandinavian sentences in which the predicative adjective disagrees with its subject are traditionally referred to as ‘pancake sentences’, named after the examples studied. The sentence in (12) is an example from Norwegian.

.

(12) Pannekaker er godt.

pancakes PL are good NEUT/SING

‘Pancakes are good.’

.

Enger (2004) claims that the use of the neuter form in pancake sentences is a manifestation of semantic agreement3. He recuperates the traditional idea that subjects in this kind of sentences refer to propositions4 and suggests a slightly different course5: these subjects display a low degree of individuation, i. e., these subjects are far from prototypical subjects – proto-agents in the line of Dowty (1991). Accordingly, agentive subjects trigger syntactic agreement. Nouns in pancake sentences are usually not animate, and if they look animate, they will not typically be interpreted as such. So they have a low potential for agentivity, and consequently will not trigger syntactic agreement.

Enger assumes the continuum of individuation (Sasse 1993, apud Enger 2004), a notational variant of the more familiar animacy hierarchy.

The idea is that when noun phrases whose reference is low on the individuation scale are used, a pancake sentence then results: high degree of definiteness correlates with agreement, low degree, with apparent absence of agreement. Enger also observes that the category neuter generally ranks low on the individuation hierarchy. So typical subjects in pancake sentences are noun phrases referring to inanimates, abstract nouns and mass nouns, i.e., entities on the right side of the individuation continuum. In this way, abstract nouns, mass nouns and propositions are very similar.

Josefsson (2009) questions Enger’s work – mainly the idea that the agreement can occur due to semantic reasons – and argues that in Swedish (another Scandinavian language) there are two different constructions involving neuter agreement: Construction NOM(inal) and Construction PROP(ositional), exemplified in (13) and (14), respectively.

.

(13) Senap är gul-t.

mustard COMMON is yellow NEUT

‘Mustard is yellow.’

.

(14) Två älskare är omoralisk-t.

[two lovers] COMMON/PL is immoral NEUT

‘To have two lovers is immoral’.

.

The author proposes that the topmost XP of the subject phrase of both constructions contains a null neuter element that accounts for the neutral predicative agreement. So he can dismiss the semantic agreement. To explain the reading in (14), he needs to project a vP, the head of which is a null light verb.

Nevertheless, Josefsson’s approach is equally costly: to account for the same agreement facts in two different constructions, he needs to postulate that both subjects have the same maximal projection: SemP – Semantic Phrase. This projection hosts the neuter feature that triggers neuter agreement on the predicative adjective. Moreover, he needs to admit many different light verbs6 to explain the variety of readings that can be inferred from sentences like (14).

Wechsler (2011) also discusses the nature of agreement in ‘pancake sentences’, regarding Swedish data. The author’s proposal is that grammar chooses among three kinds of agreement: a) grammatical agreement – when the target form depends on the controller’s phi features; b) semantics agreement – the target form depends on the controller’s meaning; c) failure agreement – target takes the default form. In the case of ‘pancake sentences’, Wechsler argues that grammatical agreement is blocked by the lack of phi features and not by the presence of a silent verbal head or a clausal structure in the sentences under analysis. Moreover, ‘pancake-NPs’ include entity-denoting nominals exhibiting logical metonymy7, so that they denote eventuality. Hence, the neuter singular form of the predicate follows from a failure of agreement (the default form is used), and not from a semantic agreement – the situation or event reading arises from the metonymic effect.

According to Wechsler, the lack of phi features that mark the referential index of pancake-NPs8 gives rise to agreement failure. In this case, subjects are not accessible to predicates on eventuality. The advantage of Wechsler’s approach is that it unifies the subjects of these sentences (they are all DPs) and dismisses the need for semantic agreement.

In the same direction, Danon (2012) deals with copular clauses in Hebrew involving the non-agreeing copula ze. The author explains that, among the different types of copula in Modern Hebrew, there are two types of pronominal copulas that are used in present tense verbless clauses: i) hu/hi/hem/hen , which is homophonous with a 3rd person pronoun, and which agrees in number and gender with the subject (15); and ii) ze/zot/ele, which is homophonous with a demonstrative and which never agrees with the subject (16).

.

(15) yeladim hem macxikim /*macxik.

children- M-P copula M.-P funny M-P / funny M-S

‘Children are funny.’

.

(16) yeladim ze macxik / *macxikim.

children- M-P copula M.-S funny M-S / funny M-P

‘Something (contextually-determined) involving children is funny.’

.

The sentence in (16) has a “hidden eventuality” reading, i.e., the predicate applies to an understood eventuality related to the overt subject rather than to the literal denotation of the subject, as in (15). Danon also points to another interpretation found with the copula ze: the sentence in (17) has what he refers to as a ‘classification reading’ - in this case “the subject is classified as having something to do with a red form”.

.

(17) tlunot ze tofes adom. complaintsFEM/PL copulaM.S. formMASC/SING red MASC/SING ‘Complaints (should) involve a red form.’

.

Moreover, Danon notes that the subjects of ze clauses can be generic (18), non-specific indefinite (18) or referential definite (20).

.

(19) nemerim (ba-bayit) ze nexmad. tigers-M-P in-the-house ZE-M-S nice-M-S ‘(Having/dealing with) tigers (at home) is nice.’

.

(20) šney orxim ze me’acben. two guests-M-P ZE-M-S good-M-S ‘(Having) two guests is annoying.’

.

(21) ha-bibliyografya ze tov the-bibliography-F-S ZE-M-S good-M-S ‘(Having/doing) the bibliography is good.’

.

For the author, ze-clauses must have a unified explanation, regardless the reading they can obtain, be it a “hidden eventuality” reading or a “classification” reading. He emphasizes that these sentences are not actually predicational sentences, since the predicate in fact does not apply to the subject. The question he formulates is why the syntax of ze-clauses correlates with this observed semantics, since it is impossible for agreement itself, or lack of agreement, to have any kind of semantic effect. For him, there must be an independent factor that is responsible both for the lack of agreement and for the observed semantic effects. He claims that this factor is the absence of interpretable features on the subject itself and that this can happen only to non-thematic subjects.

Danon suggests that two possible approaches could be postulated to account for this lack of interpretable features on the subject. One is to presume that a DP consists of multiple layers, each with its own set of features, and that, in ze-clauses, D lacks its features, what makes the DP ‘invisible’ to DP external agreement. The other is to assume the index/concord distinction of HSPG (Kathol, 1999 and other references therein). NPs normally carry two sets of agreement features: a) concord: used in NP internal agreement, i.e., features that are DP-internal); b) index: used in NP external agreement. Non-agreeing subjects in ze- clauses have concord, but lack index features. For Danon, these two approaches capture the idea that the subject of a ze clause is defective in what concerns the agreement features, and, in his work, he uses the term ‘INDEX features’ to refer to both proposals. He concludes that this lack of INDEX features has two kinds of consequences: a syntactic one (there is no external agreement) and a semantic one (any encoding of argumenthood is impossible). Since these features are used as argument identifiers, lack of INDEX features entails that the subject is not marked as an argument. Ze subjects would then be predicates and the sentence would be interpreted as expressing a subset relation between predicates. To explain the hidden eventuality reading, Danon proposes that it is “the result of shifting the denotation of the subject to a contextually- determined property P related to the overt nominal”.

Concerning our data, the proposal that these subjects are in fact hidden infinitival clauses would be more advantageous, because in this way we could unify the analysis for the three sentences in (1): we could say that the apparent lack of agreement follows from the fact that there would be a clausal subject in the three sentences. Nevertheless, we have some evidence to reject this generalization. Therefore, for sentences in (1b-c), we adopt the DP analysis proposed by Danon.

3.2 Explaining Brazilian Portuguese Data

In the previous sections, we presented constructions of Brazilian Portuguese that show a mismatching agreement between the subject and the predicate (cf. (1)). Although semantically all these subjects refer to situations, we reject the view that the agreement in these sentences is an instance of semantic agreement. Furthermore, despite the fact that these sentences show the same reading effects, we assume that syntactically they are instances of two different structures: the subject in the sentence (1a) is a SC, and this explains the agreement facts: clausal subjects triggers default agreement – in the case of BP, masculine/singular. As we have already mentioned, it would be theoretically much more interesting, or less costly, if we could generalize the clausal approach for the subjects in (1b-c). Nonetheless, we have some evidence for rejecting this analysis. Following Danon (2012), we argue that the subjects in (1b-c) are DPs that lack index features. For this reason, they also trigger default agreement. In what follows, we will detail our reasons to avoid the hidden infinitive approach, then we will present some semantic properties of the subjects in (1b-c), followed by a discussion about the analysis we propose.

3.2.1 Against the hidden infinitive approach

The main idea behind the hidden infinitive approach is that subjects like those in (1b-c) can be paraphrased by an infinitival clause. The sentence in (1c), for example, would have the paraphrase in (21).

.

(21) Taking care of/dealing with little children is fun.

.

If we postulate an infinitive clause underlying the subjects in (1b- c), the default agreement would naturally result. For some authors (Josefsson 2009; Wechsler 2011; Danon 2012) this option is far from being simple. Below we present some arguments against this hypothesis, all involving Brazilian Portuguese data. Some of them are borrowed from Wechesler’s and Danon’s work.

Firstly, the hidden infinitive analysis cannot explain the restriction against definite subjects (see next section), which are perfectly acceptable in overt infinitive clauses:

.

(22) Beijar minha mulher é divertido. to kiss my wife is fun. ‘Kissing my wife is fun.’

.

(23) *Minha mulher é divertido.

my wife FEM is fun MASC .

.

Another argument comes from a remark by Wechsler (2011) on Swedish ‘pancake sentences’. According to him, this approach makes the wrong prediction in that the noun phrases in ‘pancake sentences’ would have the distribution of an infinitival clause. He shows that the subject of these sentences cannot be extraposed as an infinitival clausal subject would. This is also the case of the sentences in (1b-c), as illustrated in (24-25) for (1c).

.

(24) *É divertido crianças pequenas.

(it) is fun children little

.

(25) a. Cuidar de crianças pequenas é divertido.

to care of children little is fun

‘To take care of little children is fun.’

b. É divertido cuidar de crianças pequenas.

(it) is fun to care of children little

‘It is fun to take care of little children.’

.

Another evidence is due to Hellan 1986 (apud Danon 2012): if we add a complement to the post-copular predicate, a paraphrase with an overt infinitive is impossible.

.

(26) Água mineral é bom para lavar o cabelo. water mineral is good to wash the hair ‘Spring water is good for washing the hair.’

.

Concerning this evidence, Danon (2012) observes that clauses that have a classification reading, as in (17), do not have a reasonable paraphrase with an infinitival sentence either. These kinds of sentences, with a classification reading and a mismatching agreement, are also possible in BP, as illustrated in (27). They are not, however, the same type of sentences found in (1b-c), in that their predicates are not adjectival and they cannot receive the situation reading. Although we do not examine these sentences in this work, we think the observed mismatching agreement might have a similar explanation as in the sentences with the situation reading, since both kinds of constructions have a DP as their subject. Hence, the fact that we cannot have an infinitival paraphrase for sentences in (27) could be additional evidence against the hidden infinitive approach.

.

(27) a. Caixas de leite é da Tetra-Pak.

boxes of milk is of+the Tetra-Pak.

‘The only milk box that is worth considering is Tetra- Pak box.’

b. Reclamações é no formulário vermelho.

complaints is in+the form red.

‘The red form is for complaints’.

.

Finally, as we will discuss in the next section, the copula ESTAR cannot be used in the constructions in (28). This fact cannot be derived from the hidden infinitive hypothesis, since ESTAR occurs freely in overt infinitive clauses (29).

.

(28) a. *Alunas que bebem está chato.

students who drink ESTÁ annoying.

b. *Crianças está divertido.

children ESTÁ fun.

.

(29) Comprar uma casa neste bairro está impossível.

buy a house in+this neighborhood is impossible

‘To buy a house in this neighborhood is impossible’.

.

We consider that the facts discussed in this section expose the weakness of the hidden infinitival approach. For this reason, we reject this approach and adopt Danon’s idea that the subject of these clauses is actually a DP that lacks ‘index’ features.

 3.2.2 The DP subject

In this section, we will first point out some of the semantic properties of the copular clauses in (1b-c), in order to find a descriptive generalization about these sentences. To close the section, we will discuss which would be an explanation for their actual readings.

Firstly, these noun phrases cannot be definite, as we can see in (30). In this case, BP is more restrictive than Modern Hebrew, which accepts definite subjects in ze-clauses, as we have seen in (20). Nonetheless, it is interesting to note, as stressed by Danon (2012), that D in Hebrew has been analyzed as an abstract head, rather than the locus of overt determiners or articles. Hence, this difference between the two languages could be related not to the property of definiteness per se, but to the fact that, in BP, D hosts the article.

.

(30) a. *As alunas que bebem é chato.

the students FEM who drink PL is annoying MASC/SING

b. *As crianças pequenas é divertido.

the children FEM little FEM/PL is fun MASC/SING

Secondly a quantified noun can only get something similar to a collective reading or a non-specific reading:

.

(31) Quatrocentos convidados é chato.

four hundred guests is annoying SING

‘Having four hundred guests is annoying’.

.

The sentence in (31) means that the situation of having four hundred guests is annoying, and could not be used to express the idea that there are four hundred specific guests that are annoying. In this case, we should use the sentence in (32).

.

(32) Quatrocentos convidados são chatos.

four hundred guests are annoying PL

‘Four hundred guests are annoying’.

.

Finally, considering the copula used in these sentences, we have noted that the copula SER is preferred to copula ESTAR9. About the sentences analyzed in this paper, most speakers reject them with ESTAR. In conclusion, the sentences that present a mismatching agreement between a DP subject and a predicate are preferentially construed with the copula SER, and this subject must be generic, but, as we have seen, the predicate does not apply to the generic subject itself. This leads us to ask how a DP could denote a situation. Following Danon (2012), we assume that the DPs in these copular clauses lack ‘index’ features, and for that reason, they cannot trigger agreement. Hence, it would be the lack of agreement features that would account for the ability of the DP to refer to a situation. As proposed by Danon, there would be a shifting the denotation of the subject to a contextually-determined property related to the nominal.

Even if we cannot explain how exactly this is so, we think this phenomenon is probably related to the fact that only generic subjects can trigger a default agreement. We will leave this issue for future work.

 Conclusion

The goal of this paper was to account for the mismatching agreement facts observed in the sentences in (1). At first, we observed that these sentences have similar readings: their predicates apply to a situation involving the subject. We chose then to adopt an approach that involves the concept of semantic selection: a predicate selects as one of its arguments a semantic category, which presents a particular syntactic realization.

Nevertheless, we have seen that the subjects of the sentences in (1) could not be subsumed under the same analysis. While the subject of (1a) is a SC, the subjects of (1b-c) are DPs. In line with Danon’s (2012) work, we assumed that these DPs lack the features needed to trigger external agreement.

Our hypothesis is that there is a class to which predicates like (in) tolerable belong, and that this semantic class selects for a situation, a semantic category whose syntactic realization can be a SC or a DP which lacks agreement features.

To conclude, we would like to call attention to some remaining issues, namely: (i) why only generic DPs are possible in constructions like (1b-c)?; (ii) how is it possible that generic DPs can be present in agreeing and in non-agreeing constructions? (iii) what triggers the possibility of shifting the denotation of these DPs into a contextually- determined property?; and (iv) what is the role of the copula SER in these constructions? We also think that the machinery that allows a DP internal agreement, but precludes a DP external agreement, must be unfolded.

Foltran et alii (2012) propose that the subject of the copula in sentences like (7) is the DP a Maria ‘the Maria’ and that the adjective bêbada ‘drunk’ is a secondary predicate.

Foltran et alii (2012) argue more extensively for the SC status of the subject in (1a).

Enger explains that, in Norwegian, the adjective has the masculine/feminine form and the neuter form. He analyzes sentences in which the adjective remains in the neuter form, despite the masculine/feminine form of the noun in the subject position. The verb form ‘er’ is the same in singular and plural forms.

This analysis pursues the idea that these subjects are really objects: they are ‘relics’ of an underlying infinitive phrase, in which they function as objects. We will turn to this analysis in section 3.2.1.

For Enger, the analysis of these subjects as propositions is not general enough, because in Norwegian, the use of neuter can have nothing to do with propositions, as we can see in the following example:

(i) Halm er gult.

straw MASC is yellow NEUT/SING

Josefsson says that this light verb is HAVE, but it could also be construed as GET, GIVE, PERCEIVE, TAKE, HOLD and PUT.

He uses the notion of ‘logical metonymy’ following Pustejovsky 1995.

These NPs include entity-denoting nominals exhibiting logical metonymy, so that they denote eventualities.

In Brazilian Portuguese there are two copulas: ‘ser’, for permanent properties, and ‘estar’, for transitory properties (roughly speaking).

 Dordrecht Kluwer Academic Publishers 1993 ASHER N Reference to Abstract Objects in Discourse 85 n. 59 108 vol. 1-2 2012 DANON G Acta Linguistica Hungarica Nothing to Agree on? Non-agreeing subjects of copular clauses in Hebrew 546 n. 3 619 vol. 67 1991 DOWTY D LanguageLanguage Thematic Proto-Roles and Argument Selection 5 n.1 34 vol. 27 2004 ENGER H. O Journal of Linguistics Scandinavian pancake sentences as semantic agreement XIX Workshop on Formal Linguistics Rio de Janeiro, Universidade Federal do Rio de Janeiro 2012 FOLTRAN M. J et alii Complex Small Clauses Constructions in Brazilian Portuguese 891 927 vol. 113 2003 HEGARTY M Lingua Semantic types of abstract entities 89 122 Dordrecht Foris 1986 HELLAN L MUYSKEN P van RIEMSDIJK H Features and Projections The Headedness of NPs in Norwegian 35 72 vol. 32 2009 JOSEFSSON G Nordic Journal of Linguistics Peas and Pancakes: on apparent disagreement and (null) light verbs in Swedish 223 274 Cambridge Cambridge University Press 1999 KATHOL A LEVINE R GREEN G Studies in Contemporary Phrase Structure Grammar Agreement and the syntax-morphology interface in HPSG Cambridge (MA) MIT Press 1995 PUSTEJOVSKY J The Generative Lexicon Cambridge MIT 1988 ROCHETTE A Semantic and syntactic aspects of romance sentential complementation 646 686 Berlin de Gruyter 1993 SASSE H.-J JACOBS J et alii Syntax: Ein internationales Handbuch zeitgenossischer Forschung Syntactic categories and subcategories 23 SEPT Paris CSSP 2011 WECHSLER S The structure of Swedish pancakes