In this paper we discuss the by now famous Bare Singular (BS)1 in Brazilian Portuguese (BrP). BSs are very productive in BrP and they are restricted in the other romance languages, including European Portuguese2. Our aim is to deeply investigate the consequences of Pires de Oliveira & Rothstein’s (2011) hypothesis that the BS is “mass”, i.e. it is derived from a lattice structure where atoms are vague - the stuff, so to say-, and denotes the kind (or the concept) – Chierchia’s view on bare nominals in Chinese. According to them, the BS in BrP is just like the bare nominal in languages such as Mandarin. Chierchia (1998) allows the down operator to apply to predicates that are either plural or mass, i.e. a semi-lattice structure built from (vague) atoms. In Chinese all nouns are mass. The difference between count and mass nouns is not due to different structures but rather to different ways of organizing the lexicon: English is [+arg; +pred], Chines is [+arg; - pred]. Pires de Oliveira & Rothstein (2011) propose a fine-grained parameter, according to which thelexicon in BrP is not the same as in English: English is an either or language, whereas in BS, roots are always available. If this is so, then BrP shares something with several indigenous languages in Brazil: it has a bare nominal (not a Bare Singular, because singular does not mean an atomic count noun). We explore this view in this paper.

The first section briefly reviews Pires de Oliveira & Rothstein claim that the BS livro (‘book’) in João comprou livro (João bought book) denotes the kind, and as such is it not derived from a count noun. The authors rely on Rothstein’s (2010) formal apparatus, briefly introduced in this section. The prediction is that this sentence is true in a situation where João bought books by kilo or volume, since only mass nouns can have non-cardinal interpretation. The authors claim that this is precisely the case.

Beviláqua’s (2015) experiment is presented in the second section. Its aims was to empirically verify the claim that muito livro has mass interpretation. The results are compatible with the idea that one may measure an alleged count noun like livro. However, these results are not conclusive, because the data can be explained by grinding, a semantic operation that coerces counts into mass. We present theoretical arguments that sustain a distinction between grinding and measuring.

Since our aim is to understand the semantics of muito, plurality is an important piece. Plural nouns only have cardinal interpretation, although as shown by Barner & Snedeker (2005) some mass nouns as furniture in English may have cardinal readings. The third section discusses the relation between muito and muito-PL. We argue that the plural morpheme presupposes an atomic domain, i.e. a predicate of type <exd, t>. This is crucial for us to understand why there are no BSs in English: in English the count distinction is at the noun level, while in BrP it is morphological (as the plural morpheme, i.e. it is higher up in the derivation). In the fourth section we explore Rothstein & Pires de Oliveira’s (in press) proposal, according to which the root predicate allows for different arrangements of the lattice structures. It is beyond the aims of this paper to present a formal apparatus, we just want to point that it should be possible to count mass nouns, as long as there are units available in the context. Barner & Snedeker (2005) have shown that in English there are some mass nouns that may be counted, the furniture noun type. But they have also shown that this is not the case with other types of mass nouns, in particular substances. In English as in BrP, one cannot count water or honey. Thus, at first sight it seems that Rothstein & Pires de Oliveira make wrong predictions. However, Lima (2014) shows that in Yudja (a Juruna language from Tupi stock), one counts honey or water.

In the conclusion, we assume that the mass and count distinction is linguistic, and that the ontological domain is made of stuff. We tentatively propose an explanation for why in BrP it is not possible to linguistically count mass nouns. This will be the opportunity to try to clarify the cross-linguistic picture. BrP, and several other languages around the world, including several indigenous Brazilian languages, may refer to the concept directly. This explains why a bare noun in Karitiana or in BrP or in Chinese may “denote” different instantiations of aKind. English cannot do that directly, it must do it through number; so it has a Bare Plural (BP). But it is also the case that we don’t count mass nouns in BrP, and Yudja people do. We briefly discuss Pires de Oliveira, Lima & Rothstein (2014) suggestion that the presence of a classifier system is crucial in the explanation.

It is certainly the case that the mass and count distinction surfaces in BrP, as can be attested by the weirdness of the sentences below:

(1)

a. # João viu areia-S.3

‘João saw sand-PL’4

b. # Maria comprou duas farinha-S.

‘Maria bought two flour-PL’

.

One may find contexts where (1.a) and (1.b) are acceptable, i.e. one where different types of a particular stuff, or portions of that stuff. The former is called the “universe sortal”, the latter, “universal packager”; but these are mechanisms of coercion which apply to save the derivation. We come back to this issue in the next sections. On the other hand, count nouns are fine without appeal to any sort of coercion:

(2)

a. João viu cavalo-S.

‘João saw horse-PL’

b. Maria comprou dois livro-S5.

‘Maria bought two book-PL’

.

Thus, in BrP mass nouns do not combine with plural morphology (1.a), and cannot be counted (1.b); count nouns can (2a.b.).

Relying on the literature about BrP, Pires de Oliveira & Rothstein (2011) observed that muito is a mass quantifier because it combines with a mass noun, engenders non-cardinal interpretations, and is incompatible with the bare plural. These properties are exemplified below. In (3) the most natural interpretation is volume:

(3)

Tem muita água nessa reserva.

There is muito6 water in this reserve.

.

(3) is true if the volume of water in the reserveis above a contextual pattern.7 Crucially it cannot be interpreted as: the number of units of water is greater than the number of units of water that is considered normal. This is expected, since água ‘water’ is mass and we all know that substances are not “atomic”. The literature on cognition has shown that human infants, as young as 3 months, as well as non-humans do distinguish between substance and units (see, for instance Carey (1985), Carey and Spelke (1996) and Barner et al. (2008)). Thus, (3) makes perfectly sense. Muito is a mass quantifier which measures volume (not unities).

And more, muito cannot combine with the bare plural:

(4)

* Tem muita gata-S na sala.

have muito cat-PL in+the room.

.

Sentence (4) is ungrammatical in BrP. This is expected if muito is a mass quantifier: it does not combine with plural nouns.

Pires de Oliveira & Rothstein (2011) predict that if their hypothesis that the BS is mass is correct, then it should compose with muito, and give rise to a non-cardinal interpretation. They exemplified the discussion with a “homemade” example. Suppose that we know that João has back problems and he is packing his backpack with one very thick book, and someone comments:

(5)

É muito livro pra você carregar.

It is muito book for you to carry.

.

They predicted that (5) can be true in such an extreme situation where there is just one object, but it is thick (or heavy). Notice that in this scenario it cannot be the case that the number of books that he has is greater than the normal standard, because he has just one. Whatever is compared, it cannot be the cardinality.

However, it is a consensus that the BS can be used in a situation where more than one individual is involved. Nobody doubts that sentence (5) also has a cardinal interpretation, i.e. it is true in a situation where the number of books that João is carrying is greater than the number of books taken as the standard. There are two issues here: (i) is it really the case that a volume interpretation is available for the BP? Is the sentence true if João has less number of units but they are thicker or heavier? Pires de Oliveira & Rothstein (2011) argue that this is the case. Thus the sentence in (5) can be interpreted by comparing the volume, the weight, and the number of individuals. (ii) How do we account for the cardinal reading if muito is a mass quantifier?

Let’s first turn to the empirical data.

Beviláqua’s experiment was designed to evaluate whether the hypothesis that muito BS, exemplified in (5), can have a volume interpretation. In this experiment, the ultimate goal was to verify whether in BrP bare noun phrases behave differently in contexts which made salient the volume or the number of individuals Very briefly, in the task participants were asked to judge the truth of a statement based on visual inputs. The target-sentences, recorded in audio samples to diminish the influence from the written variety, asked about the adequacy (or not) of the sentence to describe the situation in a photography. For instance, the picture in figure 1 was introduced while the participant heard the question Tem muito livro na sacola? (Is there much/many book in the bag?), and he had to choose “yes” or “no”:

BrP has agreement and plural morphology, thus muito is part of a grammar where it is in contrast with muito-PL. If muito is mass and muito- PL is plural, we have just met our old friends: much and many. Unfortunately this is not so! As we said, English is a well behaved language: much only combines with mass, see (6), and many only combines with plural count nouns:

(11)

a. * Many water.

b. * Many waters

c. * Many book.

d. Many books

.

Thus, in English both much book and many book are out, and the reason seems to be clear: a cumulative denotation is needed in other to compare, and book in English is singular, a non-cumulative predicate, thus it cannot be compared.

The scenario inBrP is less clear. Given what we said so far, the form muito livro can be translated by much book and also by many book. However, this is not the best way of describing the situation, since we are dealing with different grammars. In English, book is indeed (semantically) singular, whereas this is not the case with livro in BrP, which in this particular resembles bare languages as Chinese, Karitiana, Yudja. But in both English and BrP, the plural morpheme requires individualization (and by consequence atomic sums), which explains why in both languages these phrases only compare the number of individuals. We come back to the cross linguistic issue in the conclusion. For now, let’s dig even deeper into the nominal system in BrP.

The literature on number agreement in BrP (See Schere (1988) Scherre & Naro (1998)) claims that plural morphology must be marked at least at the right most constituent in the noun phrase, thus (12.a) is ungrammatical, but all other forms are possible:

(12)

a. *Muito livro-PL.

b. Muito-PL livro

c. Muito-PL livro-PL

.

Suppose this is so. Thus, we don’t expect muito livro to be plural, since there is no plural mark. But it does have number interpretation. Consider the two grammatical possibilities of marking plurality, (12.b) and (12.c). There seems to be no difference in interpretation; they are true if the number of units of books is greater than the standard number of books. How do we know that livro is plural?

If we trust the data so far, the main difference is muito livro, on the one side, and muito-PL livro(-PL), (12.b) and (12.c), on the other; both can be used in counting contexts, whereas only muito livro can be used to compare volume. Comparing muito-PL and muito we arrive at the following generalizations:

.

(I) Muito-PL compares by number

(II) Muito allows comparison by number and by volume

.

A straight (and inadequate) solution is to propose that muito is ambiguous: muito-C where c means count, and muito-V, where v means volume. Since there seems to be no lexical restriction to BSs in BrP, in the sense that any lexical item can receive a mass interpretation, that leads to assume that all count nouns in BrP are ambiguous between count and mass: livro-C (book) is a count noun, and we have the cardinal reading, and livro-V is the mass noun, and the volume reading is allowed.12 However, the hypothesis that there is a muito-C is not economical, and is challenged by the ungrammaticality of (12). If (13) is not grammatical, then we can be sure that plurality is given by the morpheme, and that (12c) is just agreement:

.

(13)

*Tem muito livro-S na sacola.

There is muito book-PL in+the bag

.

If muito were ambiguous, then it should combine with bare plural nouns, like livros in (13). But it does not.

So how can we account for these facts? And what consequences such an account brings about?

The bare mass in BrP behaves as English mass phrases, i.e. it does not combine with the plural:

(16)

a. * Muita-S água

b. * Muita-S água-S

c. Muita água

.

The hypothesis is that água (water) in (16.c) also denotes the kind and is shifted to the root predicate, which denotes a lattice built from vague atoms. This is pretty much the picture that Chierchia (1998) offers for mass nouns. If this is so, and if it is the case that muito is underspecified with respect to the scale, then we expect that it is also possible to map water into a scale that has unities, in the same way it is possible with livro.

Thus, on the one side, we expect languages that count mass nouns, a prediction that is banned in the literature - mass nouns cannot be counted is the mantra! It is true that substances are not composed of natural unities and one could justify the inexistence of the unity reading for mass nouns due to its natural properties, as Rothstein (2010) and others suggested. Nouns as furniture can be counted because they have natural units. But if there are units in the context, what would block such a possibility for substances? Why we do not normally count water? Isn’t a drop a natural unity? The prediction is that this is indeed possible, and it happens in some languages. Lima (2014) argues that Yudja is such a language. In Yudja it is possible to count mass without any classifier, as exemplified below:

(17)

txabïu uda awïla wï

‘three someone honey bring’

‘Someone brought three (portions of) honey’

.

If a count reading is available for the root noun livro, one expects that is it also available for root nouns that denote substances as areia (‘sand’). This is the mirror situation of the volume reading of BS; so one has to check whether the counting of substances is not due to coercion via packaging. This was not done yet for BrP. But Gomes & Lima (2015) found some preliminary results for BrP that confirm that it is possible to “count” mass the same way we “count” books in muito livro. The idea is that one can felicitously use (18) to describe a situation where the number of units of sand in a context is greater than the number of units of sand in the other relevant situation, given that the volume is the same:

(18)

Essa sala tem mais areia que aquela.

This room has more sand than that.

.

Suppose this is so, then we have support for the hypothesis that “Bare” bare nominals, i.e. nominal without determiners and without any morphology, as they happen in Chinese, Karitiana, Yudja and BrP, denote the root noun. This explains the similarities between such different languages. But now we have to come back to our first example and explain why one cannot say (1.b) in BrP, repeated here as (19) for convenience:

(19)

# Maria comprou duas farinha-S.

‘Maria bought two flour-PL’

.

The literature claims that (19) may be interpreted as types of flour – white flour and integral flour – or packaging: two conventionalized packages of flour. This leads us to our last section where a cross- linguistic perspective is proposed.

In their 2011 paper, Pires de Oliveira & Rothstein advance the hypothesis that English does not have the BS, only the BP, because it is an either/or language, that is a noun is either count or mass, and once a count operation is performed on the root noun, it is no longer available for other operations, for instance the down operator to generate the kind. They suggested that there languages where the root noun is always available, as BrP and Chinese; in these languages the root noun is always available, allowing structures like the BS in BP. But, as discussed in Pires de Oliveira (2014), these are also very different languages, since BrP has the plural morpheme which forces the distinction between plural and non-plurality. Chinese, on the other hand, has a very complex system of classifiers, which does not seem to match our binary distinction between plural and root.

This hypothesis allows us to see the similarities between BrP and Yudja when the nouns are bare, but it does not account for the difference between them, i.e. it is not possible to directly count mass in BrP, as exemplified in (19). Pires de Oliveira, Lima & Rothstein (2014) suggest that the difference may be explained by the classifier system. BrP has a classifier system that is obligatory in order to count substances. The consequence is that if (19) is possible it is through coercion: the data is reanalyzed so that units are provided via a hidden classifier, that we know it is there, because there are classifiers in BrP. Yudja has no classifier system, and no number morphology; thus the bare noun phrase is its only resource (Yudja is barer than Chinese, in this respect).

Thus it is fine to count substances, given that there are units in the context; no coercion is involved. This is a hypothesis that is still under investigation, built as part of a castle the foundations of which are still very shaky; to say the least we lack a formal definition of vague atoms. No doubts much more investigation is needed, but we hope to have contributed to clarifying the mass hypothesis in Pires de Oliveira & Rothstein (2011).