When quantifiers do not agree: three systems

Basque weak quantifiers optionally agree with the inflected verb in number. This papers’ main aim is to study the dialectal variation shown by this phenomenon. The study will show that it is necessary to differentiate at least three systems: the western-central system, one that we will call the transition system, and the eastern system (souletin). The western-central system allows the presence of non-agreeing weak quantifiers in every case-marked position, ergative, dative or absolutive; the transition system does not allow it with ergative case arguments, and the oriental system allows it only with absolutive case arguments. In the latter system, the distribution of non-agreeing quantifiers is identical to that of bare nouns: bare nouns are only possible in those positions where absolutive case is assigned.


Introduction: the phenomenon
This rule has an exception in so called 'vague' weak quantifiers in Basque, which optionally agree in number with the inflected verb (2a-d) (see Rotaetxe 1979;Txillardegi 1977Txillardegi , 1978EGLU 1985;Etxepare, 2000). ii (2) a. Bezero asko etortzen da/dira halako egunetan customer many come-hab is/are such days-in 'A lot of customers come in such days' b. Bezero gehiegik eskatu du/dute arrain zopa customer too-many-erg asked aux-sg/aux-pl fish soup 'Too many customers asked for fish soup' c. Maiak lagun gutxi ikusi du/ditu gaur Maia.erg friend few seen aux.sg/aux.pl today 'Maia has seen few students today' The notion of what we mean by 'vague' weak quantifier can be intuitively grasped by means of the following contrast:

3
(3) a. Mila ikasle etorri dira/*da thousand student come aux-pl/aux-sg 'One thousand students came' b. Milaka ikasle etorri dira/da thousand-suffix student come aux-pl/sg 'Thousands and thousands of students came' Whereas (3a), which involves a 'definite' quantity, triggers plural agreement in the inflected verb, (3b), which involves a non-definite quantity (equivalent to thousands of in English), only optionally triggers agreement. Cardinal quantifiers always trigger plural agreement in standard basque. Vague quantificational expressions constructed out of them, on the other hand, may not.
This phenomenon is general in the Basque area, with some interesting and systematic dialectal variation that we will try to synthesize here. The present paper offers a descriptive account of the variation involved in optional number agreement in the Basque area, as well as some basic generalizations that provide syntactic cues for a unified analysis. A full syntactic explanation of the dialectal variation related to this phenomenon is beyond the scope of this paper. One solid conclusion that follows from our discussion is that non-agreeing quantificational expressions are not counting expressions, but rather expressions related to what Borer (2005) has called a "stuff divider": a functional head whose semantic contribution is to portion out the denotation of count terms so that they can interact compositionally with the counting function. In that context, vague quantifiers merely measure the noun. Measures constitute the other quantificational domain in Basque that presents an agreement alternation in number. iii 4 (4) Hiru litro ardo edan du/ditu three liter wine drunk aux-sg/aux-pl 'He/she drank three liters of wine' We may wonder at this point what the agreement alternation is: is it an alternation between plural number features and singular ones? Or is the singular agreement form just a default, selected in the absence of any number feature? It is not easy to answer to this question by looking at the inflected forms directly. However, if we move to other syntactic contexts, the answer seems to favor the conclusion that third singular agreement, in the context of vague quantifiers in Basque, is just a default, with no correspondence with actual number features.
One such context is provided by secondary predication, which requires agreement in number (see Artiagoitia, 1994). The example in (5) gives an illustrative example with a Small Clause complement.
(6) a. Liburu asko hondatuak ikusi ditut book many worn-out.pl seen aux.pl 'I've seen many books worn-out' b. *Liburu asko hondatua ikusi dut book many worn-out.sg seen aux.sg 'I've seen many books worn-out' Whereas a vague quantifier that agrees in plural with the inflected verb licenses a secondary predicate with a plural suffix -k on it, a vague quantifier that does not agree in plural can not license singular agreement in the secondary predicate either. The conclusion seems to be that agreement in singular with the quantifiers that do not agree in plural with the verb is impossible, and that therefore, the relevant quantifier forms must lack number features, either plural or singular. iv That the problem is in number agreement and not, say, in the ability of non-agreeing quantifiers to license a secondary predication is shown by the following fact: if we allow for a secondary predicate that does not have number, secondary predication with vague quantifiers becomes possible. One relevant configuration involves the [-ta] suffix, an adverbial ending that attaches to participles, which does not agree in number in Basque. When the participial substitutes for the [determiner+number] suffix, secondary predication with vague quantifiers becomes possible (7).
(7) Liburu asko hondatu-ta ikusi dut/ditut book many worn-out.part seen aux.sg/aux.pl 'I've seen many books worn-out' The paper is organized as follows: In section 2 we present the received analysis concerning the agreement alternation in Basque. Section 3 provides arguments against this view. Sections 4-7 show the properties of Basque non-agreeing quantifiers and their dialectal variation. We 6 distinguish three systems: (i) central, (ii) transition system (Lapurdian), (iii) eastern (Souletin). In section 8 we show the similarities between the non-use of the article in both Souletin and in some Romance languages. This section allows us to state a general syntactic condition on the non-agreeing cases. Section 9 concludes the paper.

A previous view: non-agreeing quantifiers are masses
The descriptive grammar of Euskaltzaindia (1985: 223-224) assimilates the absence of number agreement with weak quantifiers to the absence of number in mass terms. Take for instance the contrast in (8). The presence of number agreement in (8b) triggers a count interpretation of the mass term haragi 'meat', which comes to denote a set of individualized meat types. The grammar of the Academy suggests that the absence of number agreement with count terms has the opposite effect: it converts count terms into mass terms. The grammar comments on the following sentences in (9).
(9) a. Liburu asko erosi dut book many bought aux-sg 'I bought many books' b. Liburu asko erosi ditut book many bought aux-pl 'I bought many books' According to the Academy's grammar, (9a) and (9b) do not have the same interpretation: whereas "in the first case we consider a mass of books; in the other case we consider one book and then another one, and another one, and so on" (1985: 223). To make things clearer, the grammar presents the following case.
(10) a. Harri asko bota dute stone much thrown aux-sg 'They threw a lot of stone' b. Harri asko bota dituzte stone many thrown aux-pl 'They threw many stones' In (10a) harri 'stone' is taken to be non-count, as a big quantity of stone. In (10b) it refers to a big quantity of stones (as a count term). The Academy's grammar does not go beyond the intuition above. Although we will not pursue this line of analysis, we share the intuition that (10b) offers more opportunities for an individualized treatment of the stone than (10a b. Ikasle asko ikusi dut student many seen aux 'I have seen much student'

Are non agreeing quantifiers mass?
It can be shown however that non-agreeing quantifiers are not mass terms. As a starting point, we consider Pelletier's well known thought experiment (1975) to characterize mass terms. He proposes the existence of two imaginary machines, that he calls the Universal Grinder and the Universal Objectifier. For the Universal Grinder, we are to imagine a device which can grind anything, no matter how big or small. Into one end of the device "is inserted an object of which some count expression is true, and from the other end spews forth the finely-ground matter of which it is composed. So a hat is entered into the grinder and after a few minutes there is hat all over the floor" (from Pelletier and Schubert 1989:342). This is so despite the fact that we could also have said that there is felt all over the floor, using a mass expression. Examples of this type "show that many count expressions can be seen to already have within them a mass sense or a mass use" (ibidem: 343). Taking the word sagar 'apple' as our putative count term, we could take (13) to involve the mass coming out of the Universal Grinder.
(13) Entsaladak sagar pixkat dauka salad-D-erg apple bit has 'The salad has a bit of apple in it' Take, however, something like (14), with a non-agreeing vague quantifier.
(14) Ikasle asko ikusi dut gaurko batzarrean student a lot of seen I-have today's meeting-D-in 'I have seen a lot of students in today's meeting' The sentence in (14), with a non-agreeing quantifier, does not involve a mass term, in Pelletier's sense: what I have seen in (14) is not scattered pieces of student, but a number of students, all of them of a piece. True, the force of this argument against a mass-approach to non-agreeing quantifiers depends on the force of Pelletier's metaphor to characterize mass terms as a whole. We know that in this sense, the metaphor is not comprehensive enough.
Other mass terms appear to reflect objects that we would better locate in the entering side of the machine. This is the case of mass terms like furniture or crockery ): ground-up furniture and furniture do not mean the same, despite the mass status of the term.
In any case, even with simple ambiguous nouns such as apple, the mass-approach falls short of accounting for the range of interpretations that non-agreeing cases have. Consider a sentence like (15).
(15) Plater honetan sagar asko ikusten dut dish this-in apple many see aux-sg (i) 'I see a lot of apple in this dish' (ii) 'I see a lot of apples in this dish' As shown by the translations, non-agreeing quantifiers can be interpreted in two ways: either as mass terms, referring to a quantity of apple, or as referring to a plural set of (whole) apples. In other words: the sentence in (15) can be interpreted as making reference to, say, a dish containing a set of piled-up entire apples. The mass-approach has nothing to say about this second interpretation.
Other properties distinguishing mass terms from non-agreeing cases lead us to reject the mass approach to non-agreeing quantifiers. Lonning (1987) shows that masses cannot entertain a predication relation with non-homogeneous predicates. Homogeneous predicates are those that are both cumulative and divisive. The examples in (16) involve a nonhomogeneous predicate (to weigh more than 300 kilos). Whereas mass quantifications can not be the subject of the non-homogeneous predicate (16a), non-agreeing quantifiers with a count noun can (16b).
(16) a. *Ur askok 300 kilo baino gehiago pisatzen du water a lot of 300 kilo than more weight-hab aux '*A lot of water weights more than 300 kilos' b. Zaldi askok 300 kilo baino gehiago pisatzen du horse a lot of 300 kilo than more weight-hab aux 'A lot of horses weight more than 300 kilos' Finally, we note that some of the quantifiers that give rise to the alternation just cannot quantify over mass terms. This is the case of zenbait 'some' and hainbat 'a sizeable quantity'.
(17) shows that even the non-agreeing cases do not support a mass interpretation.
(17) a. Zenbait ardo edan dugu some wine drunk aux-sg * 'We drank some wine' √ 'We drank some wines' b. Hainbat haragi ekarri dugu some meat brought aux-sg * 'We brought some meat' √ 'We brought some meats' Up until now, we have concentrated on showing the differences that exist between nonagreeing quantifiers and mass terms. In the sections that follow, we will mainly concentrate on the dialectal variation that non-agreeing quantifiers show, and making as thorough a description as possible of this variation. As will be made clear, there are at least three systems in Basque when it comes to the distribution of non-agreeing quantifiers: Central-western, Transitional (Lapurdian), and Eastern (Souletin).

Syntactic distribution of non-agreeing quantifiers
In

The distributive nature of non-agreeing quantifiers
One of the characterizing properties of non-agreeing quantifiers (which further distinguishes them from mass terms) is their distributive nature (Etxepare 2000). They can only be interpreted distributively, and this sets certain restrictions on the kind of predicate they can attach to.

Distributive readings
Consider for instance the contrast between (19) and (20).  (19) is typical of count plural entities (see Krifka, 1992). Unlike (19), (20) only allows a strict distributive reading, where youngsters individually lift the stone, and several stone-liftings (as many as there are youngsters) occur.

Predicate classes
Non-agreeing quantifiers are incompatible with collective predicates (predicates that do not allow event distribution). The examples in (21) Having a meeting or arranging books in a certain order denote relations that require more than one individual and give rise to collective readings. Predicates that denote such a relation are incompatible with non-agreeing quantifiers.

Enumeration and anaphora
Another difference between agreeing and non-agreeing quantifiers is that the latter cannot make reference to specific individuals. Thus, non-agreeing quantifiers cannot be antecedent to anaphoras, in opposition to what happens with agreeing quantifiers, as the examples in (27) show.

Syntactic distribution of non-agreeing quantifiers
The transition system shows some differences compared to the Central-western system when it comes to the distribution of non-agreeing quantifiers. In the Central-western system non-agreeing quantifiers are grammatical in all grammatical functions, whereas in the Transition system this is not so: non-agreeing quantifiers can appear in S position, but only with absolutive case (29a), they don't accept to appear with the ergative case (29b); they can appear in IO position, with dative case (29c); and they can also appear in DO position with absolutive case (29d). Thus, non-agreeing quantifiers appear to be unable to appear with the ergative case. The sentence in (30), with a agreeing weak quantifier in subject position, can obtain two interpretations, a collective one and a distributive one (just as was the case in the Centralwestern system). Now, the fact that non-agreeing quantifiers do not accept ergative case makes it impossible to conclude whether there are any differences in this respect between the Central-western and the Transitional system.
However, the next two subsections make it clear that non-agreeing quantifiers in the transition system are also distributive.

Predicate classes
If non-agreeing quantifiers are really distributive, they will give an ungrammatical result Mikel-erg student many see aux.sg group single one forming 'Mikel has seen many students forming a single group'

Reciprocals
Reciprocals come to show exactly the same thing, that is, non-agreeing quantifiers have a distributive nature in this system. As was the case in the Central-western system, nonagreeing quantifiers are incompatible with reciprocals, as the ungrammaticality of (35b) shows. v (35) a. Gazte anitz joaten dira elkarrekin (ostatu horretara) youngster many go aux.pl together restaurant that-to 'Many youngsters go together (to that restaurant)' b. *Gazte anitz joaten da elkarrekin (ostatu horretara) youngster many go aux.sg together restaurant that-to 'Many youngsters go together (to that restaurant)'

Enumeration and anaphora
As expected, in the transition system non-agreeing quantifiers cannot make reference to specific individuals. As a consequence, non-agreeing quantifiers cannot be antecedent to anaphors (36b) and they don't allow the enumeration of individuals, i.e. it is possible to make reference to the members of the set we are talking about, (37b).
(36) a. Bezero anitz i sartu dira gaur. _ i ez dira oso pozik atera customer many come aux.pl today neg aux.pl very happy leave 'Many customers came today. They didn't leave very happy' b. *Bezero anitz i sartu da gaur. _ i ez da oso pozik atera customer many come aux.sg today neg aux.sg very happy leave 'Many customers came today. They didn't leave very happy' Thus, the main difference between the Central-western and the Transition system is related to the possibility of non-agreeing quantifiers to appear with the ergative case. Non-agreeing quantifiers in the former system have no problem to appear with ergative case whereas in the latter system, non-agreeing quantifiers cannot take ergative case. 6. Eastern system: "Souletin"

Syntactic distribution of non-agreeing quantifiers
For the third system, we follow the description provided by Coyos (1999) for the dialect of Arbailles. In this system, as was the case in the Transition system, non-agreeing quantifiers cannot appear in all grammatical functions. They can appear in S position, but only with absolutive case (39a), not with ergative case (38a). They cannot appear in IO position, with dative case (40). A quote by Coyos is in order here: "N avec le datif et déterminé par un quantificateur indéfini: si l'indice de datif est present dans le syntagme verbal, ce sera celui avec le pluriel". Finally, non-agreeing quantifiers have no problem to appear in DO position; in fact, the non-agreeing case is more commonly used than the agreeing one (41 (45)). x At this stage, a question that comes to our mind is the following: do Souletin BNs possess a non-overt plural marker? In Basque, the overt number marker is [-k].
The presence of this overt number marker is closely related to the presence of the definite article [-a] (cf. Etxeberria 2005Etxeberria , 2011. Etxeberria & Etxepare (2008, 2009 Basque) and by means of a classifier head which portions-out count BNs to make them countable in order to interact with the counting function. In Central and Western dialects, this covert function is only available when a vague quantifier is present. In Souletin, it would seem that this covert function can apply directly on the noun. This classifier does not have phonological realization, but it is able to pluralize BNs in this dialect. Note that non-agreeing weak quantifiers show exactly the same syntactic distribution as BNs. We leave the relationship between these two phenomena for future research (cf. Etxeberria & Etxepare, in prep).

Conclusions
Basque weak quantifiers display a number agreement alternation with the inflected verb.
This paper has investigated the dialectal variation of this phenomenon in Basque. We have seen that at least three different systems must be distinguished: central-western, transitional (Lapurdian), and eastern (Souletin). The paper has centered on the following grammatical issues and their geographical distribution: (i) syntactic contexts where the absence of agreement is allowed: in the central-western variety non-agreeing quantifiers can appear in all grammatical functions; the transitional system does not allow non-agreeing quantifiers with ergative case; and the eastern system only allows non-agreeing quantifiers in DO position; and (ii) the parallel distribution of non-agreeing quantifiers and bare nouns in Souletin: Souletin, and only Souletin, allows BNs in Basque, and the syntactic distribution of these BNs is parallel to the one shown by non-agreeing cardinal quantifiers.