Jump to ContentJump to Main Navigation
Nominal Arguments and Language Variation$

Li Julie Jiang

Print publication date: 2020

Print ISBN-13: 9780190084165

Published to Oxford Scholarship Online: November 2020

DOI: 10.1093/oso/9780190084165.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (oxford.universitypressscholarship.com). (c) Copyright Oxford University Press, 2021. All Rights Reserved. An individual user may print out a PDF of a single chapter of a monograph in OSO for personal use. date: 28 July 2021

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Chapter:
(p.23) Chapter 2 Bare Numeral Classifier Phrases and Bare Nouns in Mandarin
Source:
Nominal Arguments and Language Variation
Author(s):

Li Julie Jiang

Publisher:
Oxford University Press
DOI:10.1093/oso/9780190084165.003.0002

Abstract and Keywords

Chapter 2 examines Mandarin numeral classifier phrases. It begins with a discussion of a list of tendentially universal properties of numeral-noun phrases in number marking languages (NMLs) like English and French and argues for a D-less analysis of them. It then shows that although numeral containing phrases in Mandarin differ a great deal from those in NMLs in the internal nominal domain, their scope behavior, interpretations, and distribution are rather similar to those of NMLs. It argues that the D-less analysis of numeral-noun phrases in NMLs can be extended to Mandarin numeral classifier phrases and further argues for a kind-referring analysis of Mandarin bare nouns. The proposed analysis of numeral classifier phrases correctly predicts the scope behavior of bare nouns in Mandarin and allows us to account for its numeral-less classifier phrases. This chapter concludes that it is not necessary to stipulate an empty D in Mandarin

Keywords:   Mandarin, numeral classifier phrases, number marking languages, scope behaviors, bare nouns, kinds, properties, numeral-less classifier phrases, DP, D-less

2.1 Introduction

This chapter investigates the syntactic and semantic properties of Mandarin numeral classifier phrases as well as their internal domain. Mandarin numeral classifier phrases consist of a numeral, a classifier, and a noun in the order [Num Cl N].

I compare numeral classifier phrases in Mandarin with numeral-noun phrases in number marking languages, that is, languages with obligatory morphological exponents of grammatical number, like English, French, and Russian, focusing on their behavior in the internal nominal domain, as well as on their syntactic and semantic behaviors at the clausal level. I begin with a discussion of a list of tendentially universal properties of numeral-noun phrases in number marking languages, which serves as the basis for a general assumption about the semantics and syntactic status of numerals. I argue for a D-less analysis of numeral-noun phrases, wherein numerals are phrasal in the syntax and are lexically ambiguous between a modifier and a modifier with a built-in choice function variable. It will be shown that the proposed D-less analysis of numeral-noun phrases can capture their remarkable cross-linguistic argumental behavior in a straightforward way.

Looking into numeral containing phrases in Mandarin, I observe that although they differ a great deal from those in number marking languages in terms of the internal nominal domain, their external syntax and semantics is rather similar to that of number marking languages concerning their long-distance scope behavior, semantic interpretations, and syntactic distributions.

I argue that the D-less analysis of numeral-noun phrases in number marking languges can be extended to numeral classifier phrases in Mandarin to account for the cross-linguistically uniform behaviors of numeral containing phrases. I further argue for a kind-referring analysis of Mandarin bare nouns, as proposed in Krifka (1995) and Chierchia (1998b); language variation in the nominal domain, under such an analysis, is primarily located in two interrelated factors: what nouns denote (p.24) and what lower functional heads (i.e., canonical number morphology and classifiers) denote.

Crucially, the proposed analysis of numeral classifier phrases in Mandarin correctly predicts the scope behavior of its bare nouns. It also allows us to examine and account for the numeral-less classifier phrases [Cl N] in Mandarin, which have rather restricted distributions but also exhibit the long-distance scope ability of numeral-noun phrases. Last but not least, the account proposed in this chapter has further implications for other types of nominal phrases in Mandarin, both with and without classifiers. I elaborate on the details in Chapter 3.

The chapter is organized as follows. Section 2.2 discusses a list of tendentially universal properties of numeral-noun phrases in number marking languages, such as English, French, and Russian, and briefly reviews previous analyses of numerals and numeral-noun phrases with respect to their syntax and semantics. I argue for a D-less analysis of numeral noun phrases. Such an analysis of numeral provides the basis for understanding numerals classifier phrases in in Mandarin.

Section 2.3 reexamines numeral classifier phrases in Mandarin with regard to their internal properties, scopal abilities, semantic interpretations, and syntactic distributions. We will see that Mandarin numeral containing phrases differ a great deal from those in number marking languages in their internal nominal domain but are rather similar to them in the external syntax and semantics. I show that although previous analyses have provided us with important insights about Mandarin numeral classifier phrases, they are unable to capture all of the peculiarities in behavior that Mandarin numeral classifier phrases exhibit.

Section 2.4 argues for a D-less analysis of numeral classifier phrases in Mandrin and a kind-referring analysis of Mandarin bare nouns. I show in this section that the proposed analyses successfully capture the differences and similarities between numeral-containing phrases in Mandarin and those in number marking languages.

The analysis of bare nouns in Mandarin as kinds predicts that bare nouns in Mandarin should behave like bare nominals in English, which are also kind-referring (Carlson 1977a, b) and to exhibit the property of narrowest scope, unlike English indefinites. We will see in Section 2.5 that the predicted scope behavior of Mandarin bare nouns is borne out, as confirmed by observations made in Yang (2001) and X. Li (2011, 2013). In the rest of Section 2.5, I discuss the range of interpretations of Mandarin bare nouns and illustrate that these different interpretations can be derived from their basic kind level meanings within a Neocarlsonian account of bare nominals, which includes a set of ranked type-shifting operations and a principle blocking covert type-shifting operations in the presence of corresponding overt correspondents. We further see that bare nouns in Mandarin cannot project a DP. Thus, numeral classifier phrases and bare nouns in Mandarin, under the analysis presented here, are both D-less.

Section 2.6 introduces and discusses the numeral-less classifier phrases in Mandarin [Cl N], focusing on their scope behavior, semantic interpretations, and syntactic distributions. I argue that Mandarin [Cl N] is the result of phonologically deleting one from [one Cl N] and has the full structure of the numeral classifier phrases, by adopting the one-deletion view as first argued in Lü (1944). Section 2.7 (p.25) discusses the interpretational restriction for nominals in the sentence initial position unique to Mandarin. Section 2.8 summarizes.

2.2 The external syntax and semantics of bare numeral-noun phrases: some tendentially universal patterns

This section examines the external syntax and semantics of bare numeral-noun phrases, like three cats, in number marking languages, that is, languages with obligatory morphological exponents of grammatical number.1 The reason for doing so is to show that their distribution and interpretation display a formidable uniformity across languages, regardless of whether or not the language has an overt article D(eterminer) (Section 2.2.1). Because this behavior appears to be universal, when a bare numeral and a nominal are combined, all else being equal, they should have uniform syntax and semantics. I will argue for such an analysis in Section 2.2.2, and this will constitute my baseline. From it, I arrive at understanding the syntax and the semantics of numeral classifier phrases in Mandarin in Section 2.3.

2.2.1 Cross-linguistic properties of numeral-noun phrases

As is known, numerals can have two basic types. First, numerals can denote properties of type <e, t> (1ia) or property modifies of type <<e,t>,<e,t>> (1ib); second, numerals can denote entities of type <e> (1ii).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.26) The numeral “3” in (1ia) is property-denoting, serving as the predicate of “those”; “3” in (1ib) servers as a property modifier, modifying “boys.” The numerals in (1ii), on the other hand, denote entities. The numeral “5” in (1iia) denotes entities and directly serves as the argument. In “the weight of John measured in pounds is 100” in (1iib), the numeral “100” also denotes an entity. If adopting the view that gradable adjectives have a degree-based semantics (e.g., see Kennedy 1999; Kennedy and McNally 2005), the numeral in (1iic) can also be regarded as entity-denoting, and (1iic) can be paraphrased as “the degree of John’s height measured in feet is five.”

Concerning bare numeral-noun phrases, they exhibit a list of key properties in English: (i) predicative, (ii) serve as restrictors of definites, (iii) receive a narrow scope existential reading, (iv) exhibit long-distance scope and island escape-ability (e.g., Fodor and Sag 1982), (v) receive a generic/individual level interpretation (e.g., Krifka et al. 1995), (vi) lack of anaphoric use, and (vii) have entity level use.2 Examples to illustrate these properties are given as follows.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.27)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Compared with other properties, the fourth property (long-distance scope and island-escaping ability of numeral-noun phrases) (5) is relatively less transparent; I briefly provide the background of this property here.

Since May (1977), a large part of the literature on Logical Form has assumed that a covert operation can produce a syntactic and thus semantic representation that cannot be observed in the surface phonetic form of a sentence; this operation is generally referred to as Quantifier Raising (QR). One important argument in favor of a syntactic scope mechanism such as QR is that it is subject to island constraints (Ross 1967) just like overt syntactic extractions. For instance, in the following sentence the quantifier phrase “every woman” can receive either a wide scope or a narrow scope interpretation.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

In (10a), however, only the narrow scope interpretation within the adjunct if-clause is available for “every woman.” This sentence cannot receive a wide scope reading out of the adjunct clause; (10a) will not be interpreted as “that the arrival of any woman will be sufficient for John to be glad” (Winter 1997: 401). The unavailability of the wide scope interpretation of “every” out of the adjunct clause in (10a) is parallel to the impossibility of extracting a wh-phrase out of an adjunct island, as exemplified in (10b). The identical behavior of (10a) and (10b) thus shows that QR is subject to structural restrictions.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.28) Nevertheless, since Fodor and Sag (1982) and much subsequent literature, the exceptional island-escaping behavior of indefinites (e.g., “some,” “a”) has received much attention. Indefinite NPs, unlike other quantifier phrases, can escape all islands for extraction (Ruys 1992: 102–103). For instance, in the example in (11), besides the narrow scope interpretation, the indefinite “some woman” can escape from the adjunct island to receive a wide scope reading over the if-clause:

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Farkas (1981), Ruys (1992), and Abusch (1994) have further argued that indefinites can escape islands without having the widest sentential scope. For example, in (12), the indefinite phrase “a student of his” can receive a narrowest scope interpretation within the if-clause: “for every professor, the cheating on the exam by any student of his will be sufficient to make him rejoice” (12i). In addition to this narrowest scope interpretation, “a student of his” can also have a scope narrower than the sentence initial quantifier phrase “every professor” but wider than the if-clause. The interpretation of such a scope can be paraphrased as “for every professor, there is a specific student of his; if this student of his cheats on the exam, he will rejoice” (12ii). This scope of indefinites which is narrower than the sentential scope-bearing element but wider than another scope-bearing element is referred to as Intermediate Scope.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Regarding numeral-noun phrases, they behave like indefinites “a” and “some” with respect to the ability to escape islands and long-distance scope interpretations (Farkas 1981; Ludlow and Neale 1991; Ruys 1992; Winter 1997, 2001a, 2005; Kratzer 1998; Ionin and Matushansky 2006). Some similar examples are given here:

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

In (13), if the numeral-noun phrase “one woman” has wide scope over the if-clause (13i), it refers to one specific woman and the sentence is interpreted as “there exists one specific woman whose arrival to the party will make John glad.” In contrast, when (p.29) “one woman” receives a narrow scope interpretation within the conditional clause (13ii), it refers to any woman and this interpretation is nonspecific. Accordingly, we interpret the sentence as “the arrival of any woman will be sufficient for John to be glad.” Likewise, in (14), when “three students of his” receives an intermediate scope interpretation (14i), it refers to three specific students and the sentence is interpreted as “for every professor, there are three specific students of his; if these three specific students cheat on the exam, he will rejoice.” On the other hand, when “three students of his” is within the scope of the if-clause, it receives a nonspecific indefinite interpretation such as “any three students.” Therefore, the sentence is interpreted as “for every professor, the cheating on the exam by any three students of his will be sufficient to make him rejoice.”

This exceptional long-distance scope behavior as well as the island-escaping ability of indefinites headed by a or some and numeral indefinites, as shown in Kratzer (1995), Reinhart (1997), and Winter (1997), require an analysis distinguishable from standard quantification, which I will address in Section 2.2.3.

Returning to the list of properties of numeral-noun phrases in (2) through (8), these properties are attested in other number marking languages, independent of “how much D” they have.3 I elaborate this point with French and Russian here.

In a language like French which has overt article determiners, bare nominals are banned in argument positions, and the article determiners are always required (e.g., Vergnaud and Zubizarreta 1992; Chierchia 1998b):

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

In such a “strict” DP language in which overt Ds always project in the syntax, the same behavior of numeral-noun phrases in (2) through (8) is observed.

(p.30) Specifically, French also freely allows numeral-noun phrases to be arguments and does not require article determiners to appear with them (16). Its numeral-noun phrases can receive a narrow scope existential reading (16a, b), exhibit island-escaping ability (16c), have long-distance scope (16c, d), and receive a generic interpretation (16e). French numeral-noun phrases can also serve as restrictors of definites (17); they also lack anaphoric use (18) but have entity use (19).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.31)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

In contrast, Russian does not have overt evidence of article determiners and freely allows bare arguments (e.g., Dayal 2004; Bošcovic 2005), as shown in (21).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Russian numeral-noun phrases also exhibit the same properties in (2) through (8), just like their English and French counterparts, as demonstrated in (22) through (26).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.32)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Although number marking languages differ in how freely they allow bare arguments and how “much D” they have, their bare numeral-noun phrases appear to have identical distributions and interpretations with respect to all of the seven properties considered in the data set just cited. These facts lead us to the following generalizations.4

(p.33)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

The question is then what would an analysis of these very general facts look like. In the next subsection, I briefly review previous assumptions about, and analyses of, numerals with respect to their semantics and their syntactic status. Based on some of these assumptions, in Section 2.2.3, I will argue for a lexical analysis of bare numerals and a D-less analysis of numeral-noun phrases, which captures these uniform properties.

2.2.2 Assumptions about the semantics and syntax of numerals

There are three main views about the semantics of numerals. The first view attributes the existential force to numerals as part of their semantics and treats numerals as determiners; under this view, numerals are functions from predicates to generalized quantifiers, of type <<e,t>,<<e,t>,t>> (Montague 1974; Bennett 1974; Barwise & Cooper 1981, among others). However, this view of treating numerals as determiners faces various empirical problems (Winter 1997: 407; Ionin and Matushansky 2006: 319–320). In particular, as argued in Ionin and Matushansky (2006: 320), if simple numerals (e.g., two) are determiners of type <<e, t>,<<e, t>,t>>, then there is no semantic rule for combining two determiners of this type to form a complex numeral like two hundred. Furthermore, combinations of two determiners, such as *the every book, *no these books, are disallowed.

The second view treats numerals as predicates, type <e, t> (Partee 1986). As in the first view, the semantic composition of complex numerals would fail unless we assume semantic composition of two predicates as conjunction (Heim and Kratzer 1998). However, as Ionin and Matushansky (2006: 321) have argued, treating complex numerals as conjunction yields incorrect truth conditions. Specifically, if we treat complex numerals as conjunction, two hundred books will require “books” simultaneously having the cardinality 100 and the cardinality 2.

The third view, which is the one that I pursue, regards numerals as predicate modifiers, type <<e,t>,<e,t>> (Link 1987; Verkuyl 1993; Carpenter 1997; Winter 1997, 2001; Landman 2003; Ionin and Matushansky 2006; Bale et al. 2011, among many others). This view of numerals requires the nominals that they combine with to denote a set containing either atoms or sets. Compared with the previous two views, this view has various advantages. To name a few, when occurring with other determiners or in a predicative construction, numerals behave semantically more like adjectives, such as “the three/tall girls”; “they are three/tall” (Winter 1997: 406). (p.34) Empirically, this modifier view is supported by evidence not only from well-studied languages like English, but also from less-studied languages such as Finish, Turkish, Hungarian, and Western Armenian (see Ionin and Matushansky 2006; Bale et al. 2011 for discussion of these languages). In addition, the semantic type of modifiers enables us to have an analysis of nominal phrases containing complex numerals (e.g., two hundred books) that is both compositional in the semantics and recursive in the syntax (Ionin and Matushansky 2006: 317–319).

Under the predicate modifier view, the semantics of a numeral can be defined as a restrictive adjectival modifier (Link 1983; Ionin and Matushansky 2006; Bale et al. 2011, among others):5

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Turning now to the syntactic status of numerals, two possibilities are available: as heads (Num) or as phrases (NumP):

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Both possibilities in (29) have been pursued in the literature. The head analysis of numerals has been pursued by Ritter (1991), Giusti (1991, 1997), Zamparelli (1995, 2002), and Simpson (2005), among others. The phrasal analysis of numerals has been proposed by Selkirk (1977), A. Li (1999), Haegeman and Guéron (1999), Borer (2005), Ionin and Matushansky (2006), Di Sciullo (2012), and Zhang (2013), among many others,.

The head analysis of numerals (29a), as discussed earlier, faces theoretical problems, e.g., there is no semantic rules for combining two determiners for complex numerals if numerals are treated as determiners. In addition, the head analysis is challenged empirically. It has been observed that numerals cross-linguistically can undergo coordination to form complex numerals and can be case-marked in some languages (Ionin and Matushansky 2006; Zabbal 2005; Di Sciullo 2012; among others). For example, coordination of numerals with overt coordinators is attested cross-linguistically, as illustrated in (30) and (31).6

(p.35)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

As we can see in (30) and (31), simple numerals may form complex numerals (involving addition or/and multiplication) through conjunction; the coordination between numerals can proceed recursively as with any other phrases. As for case-marking on simplex and complex numerals, I refer the readers to Ionin and Matushansky (2006) and Zabbal (2005) for detailed discussions.

Given the just-cited theoretical and emprical considerations, I pursue the view in (29b) for numerals, treating them as phrases. Such assumptions about numerals as predicate modifiers in the semantics and phrases in the syntax allow for an analysis with compositional semantics and recursive syntax of nominal expressions containing either simplex or complex numerals.7

The analysis of numerals as predicate modifiers in (28) naturally captures two of the cross-linguistic properties of numeral-noun phrases: the predicative use (cf. (2), (17), (23)) and the use as restrictors of definites (cf. (3), (18), (24)). I illustrate them in (32) and (33) with examples from English.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

8

(p.36)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

9

10

Regarding the structures and the semantic derivations in (32) and (33), there are three points that I will address. First, I adopt the analysis of the canonical number morphology as a head (Div0), along the lines argued in Borer (2005). Such a head (Div0) has also been argued to host the Chinese-type classifiers, which I will discuss shortly in Section 2.2.3. Second, I adopt the assumption that bare nouns in English and other number marking languages denote properties, along the lines pursued in Chierchia (1998b, 2010) and Dayal (2004); further discussion on bare nominals in number marking languages will be provided in Chapter 6, Section 6.2. Third, I adopt the analysis that the definite article the in English is interpreted by the iota operator “ι‎” in the semantics, which shifts properties to arguments with a definite interpretation (33b) (e.g., Sharvy 1980).

In addition to being used in predicate positions and with definite determiners or demonstratives, numeral-noun phrases can also be used in argument positions cross-linguistically, as we saw in Section 2.2.1. The analysis of numeral-noun phrases as predicates in (32) and (33) is not able to capture their argumental uses. In the following subsection, I will briefly review previous analyses of numeral-noun phrases in argument positions, some of which I will use to argue for a D-less analysis of argumental numeral-noun phrases, i.e., there is no need to assume a null D in the syntax to argumentize the numeral-noun phrases.

(p.37) 2.2.3 A D-less analysis of numeral-noun phrases and an ambiguity approach to numerals

Numeral-noun phrases can also appear on their own, without a definite determiner or a demonstrative, in argument positions. They can receive an indefinite interpretation, exhibiting long-distance scope and island-escapting abilities; they can also receive generic reference in generic sentences (cf. (4), (5), (6), (16), (22)). Let us first consider their use as indefinites. I start with a brief review of some of the representative analyses of the structure of indefinite numeral-noun phrases, followed by a review of how numeral-noun phrases derive their indefinite interpretation and long-distance scope and island-escaping abilities. Building on previous analyses, I then argue for a D-less analysis of indefinite numeral-noun phrases.

Zamparelli (1995, 2000) extends the DP hypothesis in Longobardi’s (1994) work and proposes a two-layer DP hypothesis, in which the higher DP is referred to as “StrongDP” (SDP in short) and the lower DP is referred to as Predicate DP (PDP in short). SDPs are assumed to be the only locus for an interpretation of semantic type <e>, and PDPs are assumed to be the locus of quantificational force. The Two-layer DP hypothesis includes three main principles, as shown in (34).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

The two-layer DP hypothesis not only accounts for bare noun phrases without numerals, but also it addresses bare numeral-noun phrases. In particular, numerals are treated as heads and can appear either in the head position of PDP or that of SDP. When numerals are in the head of PDP, they must be interpreted as weak cardinality predicates and cannot be interpreted referentially or specifically (Zamparelli 2000: 239). In contrast, numerals in the head of SDP are strong quantifiers or referential elements. Thus, in argumental numeral-noun phrases (e.g., John saw two boys), numerals are determiners, like generalized quantifiers, type <<e,t>,<e,t>,t>>. Zamparelli further explains the scope behavior of numeral-noun phrases with additional semantic assumptions.

This two-layer DP hypothesis not only accounts for the similarities and differences between Germanic languages (e.g., English) and Romance languages (e.g., Italian) with regard to the behavior of their noun phrases, but also it discusses and explains more nominal constructions than those investigated in Longobardi (1994 et seq). The broader empirical coverage and stronger theoretical explanations make Zamparelli’s work important in the literature. Although Zamparelli’s hypothesis mainly focuses on Germanic and Romance languages, there are attempts to extend a modified version of his two-layer DP hypothesis to classifier languages like Mandarin, e.g., Liao (p.38) (2011); however, as we will see shortly in Section 2.3, such attempts face both theoretical and empirical challenges.

Different from the previous DP hypotheses in Longobardi (1994, 2001) and Zamparelli (1995, 2000), Borer’s (2005) analyzes a much wider range of languages from number marking languages to classifier languages and from well-studied languages like English and Mandarin to less-studied languages like Armenian and Hebrew. Her analysis of nominal phrases (both noun phrases with numerals and those without) is built on her theory of “Exo-Skeletal Syntax” which contains the following main assumptions.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.39) Building on the Exo-Skeletal Syntax, the structure of a functional projection, according to Borer, is assumed to be the one in (36), in which <e>F is some functional open value of type F; R(F) is a range assignor to <e>F; co-superscriptiong (“2” in this example) notates range assignment.11

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Concerning the nominal domain, Borer assumes that the internal structure of nominal arguments has three functional layers above the bare noun (Nmax): classifier phrases (Clmax), quantity phrases (#max), and determiner phrases (DPs) and that nominals in all languages share an uniform structure, as shown in (37).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

The lowest functional projection, the classifier phrase (CLmax) in (37), headed by the open value <e>DIV (div. for “division”), is assumed to be responsible for portioning out nouns. Plural inflection, overt classifiers, and indefinite articles are assumed to base-generate in the same CLmax domain and can accomplish the portioning-out function.

Specifically, in number marking languages with overt determiners, such as English, both plural inflection and the indefinite determiner (e.g., a/an in English) can assign range to the open value <e>DIV and accomplish the portioning-out function. In classifier languages like Mandarin, it is the classifiers that assign range and accomplish that function. (Borer 2005: 93). Thus, numeral-noun phrases in English and Mandarin numeral classifier phrases share the same structure. I illustrate it with two examples in (38a) and (38b); they share the same structure in (39).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.40)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

A key element in Borer’s nominal structure is that plural inflection is classifier inflection. Specifically, plural inflection is assumed to be a distinct instaniation of the classifier system and appear in the same position as classifiers. As a consequence, “not only is it the case that classifier languages do not (appear to) have plural inflection, but languages which mark plural do not appear to have classifier inflection” (cf. Borer 2005: 92). This prediction by Borer is stronger than the one-way prediction made by Chierchia (1998b) that classifier languages do not (appear to) have plural inflection. The main motivation for this prediction is from the observation that classifier inflection and plural inflection appear in complementary distribution made by T’sou (1976) (Borer 2005: 93). I will elaborate on this issue in Secion 2.3 as well as in Chapter 3, Section 3.5.12

In this work, I adopt Borer’s analysis that overt classifiers and plural inflection appear in the same position. I also argue for an uniform structure of numeral (p.41) containing phrases across languages; however, the uniform structure to be argued for is not a DP but a D-less structure, to be presented shortly.

Turning now to the semantics of indefinite numeral-noun phrases. Some authors treat numerals as determiners of type <<e,t>,<<e,t>,t>> (Bennett 1974; Scha 1981; van der Does 1992, 1993 among others). Such an analysis of numerals attributes their existential force to the numeral semantics. However, we saw that treating numerals as heads faces both theoretical and empirical problems (cf. Section 2.1.2). More importantly, the exceptional long-distance scope property and the island-escapting ability of indefinites requires an analysis of which is distinguishable from standard quantification (Kratzer 1996; Reinhart 1997; Winter 1997).

An analysis of indefinites in terms of choice functions has been pursued in Kratzer (1996), Reinhart (1997), and Winter (1997, 2001, 2005), which captures their island-escaping abilities. In these works, indefinite numeral-noun phrases are analyzed on a par with a/some indefinites, which are assumed to lack quantificational force of their own. Specifically, indefinites are assumed to involve a free function variable in their semantics that assigns an individual to the restriction of the predicate (cf. Winter 1997: 409; see also Reinhart 1997); this function is a choice function that is existentially closed. A choice function, according to Reinhart (1997) and Winter (1997: 409), can be informally understood as a function that chooses any member from any non-empty set. Furthermore, the quantificational procedure introduced by the choice function is assumed to apply at any compositional level (see Winter 1997: 409–411 for details). For instance, for “three students,” the choice function which is existentially closed at any compositional level, applies to the set denoted by the nominal and picks from it a plurality consisting of three individuals each of which is a student; this plural individual is argumental, of type <e>.

I illustrate this with an example to show how the choice function analysis in Reinhart (1997) and Winter (1997) works for indefinite numeral-noun phrases. In the sentence in (13), (repeated in (40)), if the existential closure of the choice function is performed within the antecedent adjunct clause (40ia), the narrow scope interpretation is obtained and can be paraphrased as (40ib). If the existential closure is performed outside the conditional (40iia), it gives rise to a wide scope reading paraphrased as (40iib), where CH(f) means that f is a choice function.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.42) Very crucially, the interpretation of the indefinite is determined by the stage at which existential closure is applied, and in both cases in (40) no mechanism extracts the indefinite numeral-noun phrase one women out of the island (cf. Winter 1997: 411). Thus there is no island constraint violation under the choice function analysis.

The choice function analysis of indefinites enables the existential closure of choice functions to apply at any stage of the compositional derivation; it thus captures their long-distance scope ability. I will adopt the choice function analysis of numeral indefinites, because it not only provides a uniform account of different indefinites (a-, some-, and numeral indefinites), but also it captures the long-distance scope behavior of numeral-noun phrases cross-linguistically (we will see further emprical evidence from Mandarin in Section 2.3 and from Nuosu Yi in Chapter 4). I refer the choice function variable that is subject to existential closure as “f.”

A question immediately arises for this choice function analysis, namely, where and when exactly is the choice function introduced in the numeral-noun phrases? Some authors assume that the choice function is linked to some functional head, e.g., a null D (as pursued in Winter 2001, 2005; Ionin and Matushansky 2006 for English). Here I will puruse a different approach and propose a simple-minded lexical analysis that attributes the choice function to numerals as part of their semantics: numerals come with a built-in choice-function variable (see also Dayal 2012, 2014 for a similar view). Specifically, numerals first look for properties, and the choice function variable introduced in numerals, which needs to be existentially closed, then applies to the numeral-modified properties. I demonstrate this lexical view of choice function of numerals in (41).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

The structure of indefinite numeral-noun phrases (41b), under the proposed lexical analysis of the choice function, is the same as that of the predicative numeral-noun phrases (32c).

I consider one empirical argument and one theoretical argument for this lexical view of choice functions and the D-less analysis of indefinite numeral-noun phrases.

Its empirical motivation comes from the remarkable cross-linguistic properties of numeral-noun phrases that we saw in Section 2.2.1: they can freely appear in argument positions with an indefinite or a generic reference across languages regardless of whether a language is a number marking language or a classifier language and also regardless of whether a language has overt Ds or not.

(p.43) Take number marking languages first. Numeral-noun phrases are freely argumental in languages where no evidence of overt D is detected, such as Russian; they are also freely argumental in strict DP languages where bare arguments are banned and determiners must always occur with bare nominals, as in French.

Looking into French; bare nominals are turned to arguments obligatorily via an overt D regardless of whether they are bare singulars, bare plurals or mass nouns. For example, in (15) (as repeated in (42)), the bare plural baleines “whales” cannot appear in argument position without the determiner les “the” even if the definite D does not directly contribute to its interpretation as kinds (42a).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Nevertheless, even in a language that strictly disallows bare nominals without D like French, its numeral-noun phrases still can freely appear in argument positions without a determiner. French numeral-noun phrases also display all other cross-linguistic properties. Examples are repeated here.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.44) The fact that the remarkable cross-linguistic argumental behavior of numeral-noun phrases is still detected in a strict D language like French suggests that numeral-noun phrases must have a different source to form arguments than bare nominals.

Look at another numeral marking language Russian. As we have seen in Section 2.2.1, numeral-noun phrases in such a language that has no overt evidence of D manifest the same properties as their counterparts in other number marking languages with overt D (e.g., English and French). They freely appear in argument positions, receiving either a generic interpretation (44a) or an indefinite interpretation (44b/c) and exhibiting the same long-distance scope ability (44d/e).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Similarly, in classifier languages, numeral classifier phrases which contain a numeral, a classifier, and a noun, are freely allowed in argument positions with either an indefinite or a generic interpretation; the same long-distance scope property is detected in languages without overt D, such as Mandarin (as we will see shortly in Section 2.3) as well as in languages with overt D, such as Nuosu Yi (as we will see in Chapter 4).

Regardless of how much languages may differ with respect to argument formation of their bare nominals (e.g., obligatorily via an overt D, optionally via an overt D, or not through an overt D), they uniformly allow bare numerals containing phrases to appear in argument positions with the same properties. These remarkable cross-linguistic properties of numeral-containing phrases need to be acknowledged; the uniform argumental behavior of numreal containing phrases across languages suggests that numeral-containing phrases must have a different source to form arguments from bare nominals which vary cross-linguistically.

(p.45) Turning to the theoretical consideration of the lexical view of numerals and the D-less analysis of numeral indefinites, it provides a uniform analysis of the cross-linguistic behavior of bare numeral containing phrases. The choice function, which is built into the lexical entry of numerals and contributes to the long-distance scope interpretation, and is not subject to the existence of any functional head (e.g., D) which might be subject to parameterization. This lexical view, therefore, can reduce unnecessary theoretical assumptions and parameterization in the narrow syntax concerning numeral-containing phrases, helping us to maintain a minimalist fashion of linguistic inquiry.

Admittedly, one can accommodate the choice function variable in the syntax. For example, one can assume that the choice function is linked to some functional head (which might be subject to parameterization), e.g., a null D (as pursued in Winter 2001, 2005: 770; Ionin and Matushansky 2006: 322 for English). If we pursue this syntactic analysis of the choice function variable, we then need to make other assumptions as to where the choice function variable is realized in languages without such a functional head (i.e., languages without Ds). Even if we assume that languages without overt determiners also project null DPs in the syntax, we then need to explain why bare nouns, such as English dogs, do not have the same scope behavior as bare numeral phrases like two dogs since they might as well be argumentized via a null D. As we know, bare nominals differ from ordinary indefinites in that the former allows only narrow scope indefinite readings, while the latter participates in scope interaction (see, e.g., Carlson 1977a, b), as illustrated here.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Of course, one can further assume two types of null Ds: one is linked to the choice function variable and merges with bare numeral-containing phrases, leading to the long-distance scope behavior; the other merges with bare nouns and only contributes to the narrowest existential reading (e.g., the Longobardi’s (1994) style of null D). However, such a parametric approach of null Ds not only doesn’t show advantage over the lexical approach to numerals, but also it inevitably add unwanted stipulations and complications in the syntax.13

Regardless of what assumption about numeral-(classifier)-noun phrases one makes, it has to capture their uniform argumental behavior across languages as well (p.46) as their long-distance scope ability. The proposed lexical analysis of choice function straightforwardly captures this cross-linguistic uniformity; it also shows theoretical advantages over a null DP analysis which inevitably would be more stipulative. Regarding the implementation in the syntax, we can view that an agreement relation exists between the abstract existential closure ∃ and the choice function in the lexical entry of numerals (e.g., analogous to the analysis of negative concord in Zeijlstra 2004).

Consequentially, under the proposed lexical analysis of choice functions, numerals have a predictable lexical variant in which they are property modifiers, as demonstrated in (46).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

In addition to allowing us to account for the island-escaping and long-distance behavior of numeral-noun phrases as well as their indefinite interpretation, the proposed lexical view of choice functions also allows us to derive a generalized quantifier (GQ) variant of numeral-noun phrases. If considering the Principle of Same-Type Coordination in Partee 1987, i.e., only categories of the same semantic type can be coordinated, the phrase “three boys” in (41) which coordinates with a GQ “every girl” must also be regarded as a generalized quantifier, type <<e,t>,t>.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

With regard to the GQ use of numeral-noun phrases, it is widely assumed that the quantificational force of numerals is related to their modifier use. The quantificational force can be assumed to be gained via a global existential closure (Heim 1982), a covert existential quantifier (Link 1983, 1987; Krifka 1999), a type-shifting principle (Partee 1986; Landman 2003), or simply via our lexical analysis of choice function. I remain agnostic as to how exactly this quantificational force of numerals is obtained, but the proposed lexical analysis of choice functions is still compatible with any standard view one would adopt.

Now let us turn to the generic interpretation of numeral-noun phrases in argument positions; some examples are repeated here.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.47) The sentences in (48) all report a general property of the numeral-noun phrases; these sentences are referred to as “generic sentences” (or “characteristic sentences”) as in Krifka et al. (1995: 2–3). For instance, “three boys” in (48a) does not refer to three specific boys or any three boys in a particular event. Instead, the sentence expresses a statement over events and reports a property of any three boys in general, namely, “generally, any three boys can lift a piano.”

Generic sentences contrast with sentences that express statements about particular events and properties of particular objects, which are generally referred to as “episodic sentences” (Krifka et al. 1995). In (49), the two sentences are episodic, describing specific episodes or isolated facts; the two phrases “three boys” and “two canaries” have referents within their particular events.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Regarding what contributes to the generic quality of the numreal-noun phrases in generic sentences, Krifka et al. (1995: 14–21) argue that “the locus of the genericity in generic sentences is not in the nominal subject but rather in the sentence itself.” The authors show that generic sentences put no restriction on what types of nominal phrases may occur in them. For example, in English, all of the following NPs can participate in generic sentences: proper names, indefinite NPs, definite singular NPs, quantified NPs, bare plural NPs and bare singular NPs (Krifka et al. 1995: 8). Some examples are given here.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

The variety of nominals acceptable in generic sentences makes it implausible that this type of genericity is conditioned by the meaning of an NP.

As for how genericity arises, it has been assumed that the generic meaning is contributed by an operator Gen; this operator quantifies over situations as well as objects (Lawler 1972; Schubert and Pellitier 1989).14 Generally, the Gen operator can be viewed as a universal quantifier (∀) quantifying over situations or possible worlds (e.g., see Kratzer 1981; Chierchia 1995; Krifka et al. 1995: 45–63); this is the view that I adopt for the semantics of generic sentences.

(p.48) Syntactically, this Gen operator needs to be located in a structurally high enough position to quantify over both the external argument and the internal argument. Accordingly, it should be introduced at a position above vP (assuming the VP-internal Subject Hypothesis; see, e.g., Koopman and Sportiche 1991); this leaves us with limited choices: to merge Gen with IP or to merge Gen with TP, AspP, or ModP (with the canonical assumptions that all of these phrases are located above vP, and with the split IP analysis).15 Here I will simply assume that the Gen operator merges with IP without going into details as to which projection within IP Gen exactly merges with since it does not make much difference for the purposes of the current discussion (or see Chierchia 1995 for the assumption that Gen merges with AspP). By merging in a high enough position, the Gen operator can quantify over the numeral-noun phrases in both the external and internal argument positions in generic sentences. I illustrate the structure and the semantics of generic sentences with the example in (51) (with irrelevant details omitted).16

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

To summarize Section 2.2, we observed a series of tendentially universal properties of numeral-noun phrases in number marking languages: (i) numeral-noun phrases are systematically ambiguous (in the Universal Lexicon) between a predicate and an indefinite variant, (ii) in their indefinite incarnation, they are arguments with long-distance scope properties, (iii) in their predicate incarnation, they act as restrictors of Ds, demonstratives, quantifiers and the generic operator, and (iv) Properties i—iii (p.49) are stable, regardless of whether a language has (overt) D or not. I argued for a lexical analysis of numerals, namely that numerals are lexically ambiguous between a modifier and a modifier with a built-in choice function variable. I showed that this analysis of numerals allows us to capture the remarkable uniform properties of numeral-noun phrases across languages. It is conceivable that these properties can be explained in different ways, but this simple lexical approach of numerals suffices regardless of whether the language has a (overt) D or not. In Section 2.3, I will show that this lexical analysis of numerals suffices for Mandarin, a classifier language, as well.

2.3 Reexamining bare numeral classifier phrases in Mandarin

2.3.1 Similarities and differences between Mandarin and number marking languages in the nominal internal domain

Mandarin numeral classifier phrases share some similarities with numeral-noun phrases in number marking languages, such as English and French, with respect to their internal structure.17 For instance, in English, the plural number marking -s/-es is obligatory when a numeral (other than “one”) combines with a count noun. Although Mandarin does not have the same canonical number marking as English, it has classifiers, which are also obligatory when a numeral combines with a noun. This similarity is shown in (52) and (53). Without the plural morphology -s and without the individual classifier “ge” (as defined and discussed in Chapter 1), both (52a) and (53a) are ungrammatical.18

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Setting aside the cited similarity, number morphology and classifiers are quite different in other respects. For illustrative purposes, I will use English to demonstrate (p.50) their differences, but very importantly, variation among number marking languages (e.g., among English, French, and Russian) is also observed, which I will address in Chapter 6.

The first difference shows up when the numerals are removed from the nominal domain. When numerals are absent, number morphology still can mark count nouns, e.g., books. Bare plural Ns in English can freely occur in argument positions (cf. Section 2.2.1).19 However, in Mandarin, only when the numeral yi “one” is absent can the bare classifier noun phrase, [Cl N], be allowed in very restricted argument positions.

For example, compare the English bare plural students in (55a) with the Mandarin [Cl N] phrase ge xuesheng “Cl student” in (56a), the former can freely appear in the external argument position, but the latter cannot. Similarly, English bare plurals can freely appear in the internal argument position (55b), but the bare classifier phrase in Mandarin cannot do so (56b). Only in exteremly restricted positions, e.g., in certain post-verbal positions, bare classifier phrases in Mandarin are allowed (56c) (e.g., Lü 1944; Chao 1968; Yang 2001). I will further discuss the [Cl N] phrase in Mandarin in Section 2.7, where I argue that this phrase is not driven by the syntax but by semantic and prosodic factors (e.g., Lü 1944; Li and Feng 2013).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

When both numerals and number morphology/classifiers are absent together, we observe another difference between number marking languages and Mandarin. In (p.51) English, bare nouns without number morphology and without numerals (e.g., student) cannot occur as bare arguments (cf. Section 2.2.1).20 Examples from English are given in (58). Interestingly, bare nouns in Mandarin without numerals and classifiers are grammatical expressions with different interpretations (e.g., Chao 1968; Cheng and Sybesma 1999; Yang 2001; X. Li 2011, 2013) (57b), and they can always merge directly with a verb and occur freely as bare arguments (59).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Furthermore, number morphology is number specific, i.e., it either singular or plural (60a), but classifiers are not, i.e., they remain the same regardless of singularity or plurality (60b).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

The last difference between number morphology and classifiers is that the classifier system is rich (e.g., there are various different types of classifiers as we saw in Chapter 1), and classifiers are not treated as bound morphemes like -s/-es in Mandarin, as shown here.

(p.52)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

This discussion on number marking morphology and classifiers has shown to us that there are internal differences between numeral-containing phrases in canonical number marking languages and those in classifier languages, even though they do share some similarity.

In the following section, I provide data from Mandarin and aim to show that regardless of their various differences within the nominal domain, numeral classifier phrases in Mandarin and numeral-noun phrases in number marking languages behave similarly at the clausal level with regard to their scope behavior. In particular, I will show that Mandarin numeral classifiers behave like (numeral) indefinites in number marking languages: they can receive a wide scope interpretation and exhibit the island-escaping ability; moreover, they can also receive an intermediate scope, just like (numreal) indefinites in English. In Section 2.3.3, I further show that numeral containing phrases in Mandarin and those in number marking languages are also similar with repect to interpretation and distribution.

2.3.2 Scope behavior of Mandarin bare numeral classifier phrases

There has been a common observation that surface word order is correlated with scope relations in Mandarin; examples used to demonstrate this observation frequently involve two or more of the following elements: scope-bearing adverbs (e.g., changchang ‘often’, dou ‘all’), negation (e.g., bu ‘not’, mei ‘not’), the existential marker you ‘exist’, and universal quantifiers (e.g., mei ‘every’, suoyou ‘every’) (e.g., S.F. Huang 1981; J. Huang 1982; T. Lee 1986; A. Li 2014). Two examples are given here.21

(p.53)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

However, the just mentioned quantifiers form only a subset of quantifiers; if bare numeral classifier phrases are brought into the picture, we observe different scope behaviors. J. Huang (1982: 214–220) has observed that Mandarin bare numeral classifier phrases can escape from islands (complex NP islands specifically) to undergo QR. One of his examples is given in (63).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

In (63), the numeral-classifier phrase san ge ren “three Cl men” can either receive a wide scope reading out of the complex NP without violating island constraints or receive a narrow scope reading within the complex NP.

In additon to escaping from the complex NP islands as observed in J. Huang (1982), numeral classifier phrases are also able to escape from other islands. An example involving an adjunct clause is given here.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Similar to numeral-noun phrases in number marking languages, numeral indefinites in Mandarin can also receive either a wide scope or a narrow scope interpretation with respect to a quantifier c-commanding it. Aoun and Li (1989: 142, 1993a: 12) observe that numeral classifier phrases in passive sentences allow both a narrow scope interpretation and a wide scope interpretation with respect to a universal quantifier (see also Jiang 2012: 110–113, 191; A. Li 2014: 231), as illustrated in (65).

(p.54)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

In each example in (64), the universal quantifier c-commands the numeral classifier phrase, but the numeral classifier phrase still can receive a wide scope interpretation.

Regarding Mandarin numeral classifier phrases in active sentences, F. Liu (1997: 54–57) and X. Li (2011: 63) observe that they also behave like English indefinites in allowing both a narrow scope interpretation and a wide scope interpretation with respect to a universal quantifier.22 Three examples from F. Liu (1997) and X. Li (2011) are given in (66); in (67), I provide three more examples.

(p.55)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

On the top of these observations, I observe that, the same as numeral-noun phrases in number marking languages (cf. (12), (16), (22d)), numeral classifier phrases in Mandarin can escape islands without having the widest scope, i.e., they can receive an intermediate scope interpretation, which is a unique property of indefinites, as we saw in Section 2.2.1. In (68), a universal quantifier c-commands a complex NP which contains a numeral classifier phrase, and another quantifier phrase “most” is above the universal quantifier. We detect an intermediate scope interpretation of the numeral classifier phrase, scoping over “every” but under “most.”

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.56)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Similarly, in (69), an intermediate scope interpretation of the numeral classifier phrase, scoping over the closest “every” but under the sentence initial “every,” is available.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.57)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Note that, the intermediate scope reading can be difficult to get especially in an out-of-the-blue situation, and this reading might not be the preferred reading either. However, what is crucial here is that it is still possible to get the intermediate scope reading even though it can be very difficult sometimes. If we provide contexts/scenarios that help establish the intermediate scope reading, such a reading can be easier to get. To understand better the intermediate scope reading of the numeral classifier phrase, I will examine one of the just-cited examples (68a) closer because the intermediate scope reading will be important in the later discussion on numeral-less classifier phrases (Section 2.6.2).

The sentence in (68a) can be interpreted as follows. Every student (let’s say student 1, student 2, student 3, student 4, and student 5) was conned out of money by one swindler (different ones) (i.e., student 1 by swindler X, student 2 by swindler Y, student 3 by swindler Z, student 4 by swindler W, and student 5 by swindler V); for most teachers, they consoled every student who was conned out of money by one swindler. Under such a reading, the numeral indefinite “one swindler” has narrow scope under the universal quantifier “every student” and receives a non-specific indefinite reading. In the meantime, the sentence in (68a) can also receive the following interpretation. There is a particular swindler X who tricked money from every student (p.58) (let’s also say student 1, student 2, student 3, student 4, and student 5); for most teachers, they consoled every student who was conned out of money by this swindler X. In such a reading, the numeral indefinite “one swindler” scopes over the universal quantifier “every student” but under the other quantifier “most teachers” (i.e., intermediate scope) and receives a specific indefinite reading. Crucially, two scenarios can establish the second reading of (68a) (i.e., the intermediate scope reading of “one swindler”). The first scenario is that there is a particular swindler X who tricked money from every student, the students are aware of the fact that a single and particular swindler tricked money from all of them, and most teachers consoled every student who was conned out of money by this swindler X. The second scenario is similar to the first one except that the students do not know that they were all conned out of money by the same particular swindler X and mistakenly think that different swindlers tricked money from each of them; however, those swindlers for the students happen to be one and the same person X. One reviewer asks whether the intermediate scope reading under this second scenario is a special case of the narrow scope reading. But as we can see, for the narrow scope reading of “one swindler” there are indeed different swindlers for the students (i.e., swindler X for student 1, swindler Y for student 2, swindler Z for student 3, swindler W for student 4, swindler V for student 5); however for the intermediate reading of “one swindler,” there is indeed only one particular swindler for the student (i.e., swindler X for student 1, 2, 3, 4, and 5), regardless of whether the students are aware of this fact or not. These two readings are independent of each other, and the sentence in (68a) will be judged to be true under either of the two readings.

The preceding discussion has demonstrated that Mandarin bare numeral containing phrases also exhibit the long-distance scope ability in the same way as those in number marking languages. In the following section, we will see that numeral classifier phrases in Mandarin and numeral-noun phrases in number marking languages are similar with respect to their interpretation and distribution in the clausal domain.

2.3.3 Interpretation and distribution of Mandarin numeral classifier phrases

Numeral classifier phrases in Mandarin can appear in predicate positions just as those in number marking languages (cf. (2), (17), (23) in Section 2.2.1). An example to illustrate this is given in (70).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Mandarin numeral classifier phrases can also appear in argument positions, as we just saw in Section 2.3.2. The distribution and interpretation of Mandarin numeral classifier phrases have been discussed in very early work such as Chao (1968: 76) (p.59) and Li and Thompson (1981: 91, 167–168), which both share the observation that indefinite nominal expressions in Mandarin (including numeral classifier phrases) generally do not occur in sentence initial position. What we are going to see is that numeral classifier phrases in Mandarin are allowed in sentence initial position in both generic sentences and episodic sentences. We will start with generic sentences that contain numeral classifier phrases in the sentence initial position.

As observed in A. Li (1997, 1998), numeral classifier phrases can freely appear in the sentence initial position if they receive a “quantity-denoting” interpretation (see also Li and Thompson 1981: 167). The contrast is given in (71).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

The sentence in (71a) sounds unnatural to many speakers because they are not clear who “three students” is referring to. In order to make this sentence natural, a verb you “exist/have” can be added in the sentence initial position to assist the numeral classifier phrase to receive a non-specific indefinite interpretation (72), namely, that it could be any three students. In contrast, the numeral classifier phrase in (71b), according to Li, is not individual-denoting and receives a “quantity-denoting” reading instead, which merely expresses the quantity information denoted by “three.”

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Li’s contrast in (71) points to a very important semantic difference between these two types of numeral classifier phrases. The numeral classifier phrases in (71a/72) are object-referring, namely, they refer to three boys in a particular situation/event. In contrast, the numeral classifier phrase in (71b) does not refer to any individual in a particular event/situation.

If the numeral classifier phrase in (71b) is as “quantity-denoting,” one might expect it to behave differently from numeral-noun phrases in number marking languages. However, the reference of the numeral classifier phrase in (71b) accurately should be captured as generic rather than “quantity-denoting”: the sentence is a general statement about the number of boys that is not sufficient across situations/possible worlds. This indeed is the definition of generic sentences as in Krifka et al. (1995: 2–3), as discussed in Section 2.2.3.

Once the refernce of numeral containing phrases in (71b) is captured as generic, we expect that numeral containing phrases in Mandarin should behave similarly to (p.60) those in number marking languages in generic sentences. I will demonstrate this point below.

The generic sentence in (71b) is the same as its counterpart in number marking languages (cf. (6), (16e), (22e), (48)). Some examples from English are repeated in (73), with the Chinese counterparts provided in (74).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

To further demonstrate the similarities between generic sentences in Mandarin and those in number marking languages, I will address two properties of generic sentences which capture previous observations of numeral classifier phrases in Mandarin.

A. Li (1998: 700–701) observes that numeral classifier phrases with generic reference (“quantity-denoting” in her term) cannot interact with other quantificational phrases. For instance, the following sentence in (75a) cannot receive an interpretation as “five children cannot finish, among them, 10 bowls of rice” but the sentence in (75b) can receive such a reading with a sentence initial you “exist.”

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

The contrast in (75), indeed, is the plurality problem of numeral containing phrases: distributive or collective. The Mandarin example in (75a) reflects one of the (p.61) properties of generic sentences: sentence initial nominals with numerals cannot receive a distributive interpretation in generic sentences; they have to be interpreted collectively (Krifka et al. 1995: 35–55). To illustrate this point, consider the following generic sentences with sentence initial numeral-noun phrases in English. When the predicate is a distributive predicate (i.e., a predicate that can only apply to individuals) (76), the sentence is awkward with the numeral-noun phrase twelve cats in the sentence initial position (76b), in contrast with (76a). However, with collective predicates (i.e., predicates that can only apply to groups or sums) (77), sentence initial numeral-noun phrases are acceptable.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

A possible explanation for why generic sentences must have collective reading rather than distributive reading with numeral subjects is provided by Declerck (1988), namely, distributive predicates require unboundedness. This unboundedness requires that the restrictive indefinites must not be numerically specified. In contrast, with collective predicates, the number-specified numeral-noun phrases in the restrictor become acceptable as the number specification is essential (Declerck 1986, 1988; cf. Krifka et al. 1995: 35).

This is to say, if a predicate is neutral between distributive and collective and if it is used in generic sentences with a numeral subject, the numeral containing phrase receives a collective interpretation. An example to illustrate is the following one from Krifka et al. (1995: 55): “Six apples cost one dollar.” This sentence can only be interpreted as “six apples in total cost one dollar” rather than “six apples, each of them costs one dollar.”

Returning to the Mandarin examples in (75), the sentence in (75a) is a generic sentence, and the numeral classifier phrase can only receive a collective reading, namely, “five children all together, cannot finish ten bowls of rice over all events/situations,” as a result of “unboundeness” requirement for indefinites, similar to the English case in (78). In contrast, in (75b), after adding an overt existential operator you “exist,” the sentence is not a generic sentence anymore and becomes an episodic sentence. The unboundeness constraint is irrelevant in episodic sentences; therefore, the distributive predicate “eat” allows the numeral classifier phrase “five children” to be interpreted distributively: “there are five children, each of them cannot finish ten bowls of rice; thus there are fifty bowls in total.”

(p.62) A. Li (1998) further observes another contrast involving numeral classifier phrases between sentences in (78) and sentences in (79).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

In (78), the numeral classifier phrases cannot be referred anaphorically or referentially by a pronoun, whereas the numeral classifier phrase in (79a) can.

This contrast indeed is a contrast between generic sentences and episodic sentences cross-linguistically and can be understood by the differences between them. The examples in (78) are generic sentences; the numeral classifier phrase san ge ren “three Cl people” in (78a) cannot be referred to anaphorically or referentially by a pronoun since it does not have a reference in a particular event/situation, nor does it refer to any individual. Regarding the logophor ziji in (78b), it also cannot be bound by a generic referring numeral classifier phrase which has to be interpreted collectively in generic sentence, as we saw previously. In contrast, the two sentences in (79) are episodic sentences, and the numeral classifier phrase “three Cl people” in (79a) does have a reference in a particular situation/event; therefore, the pronoun tamen “they” can refer anaphorically/referentially to it.

In addition to A. Li’s (1997, 1998) work on generic numeral classifier phrases, Tsai (2001) examines in detail constructions in which you “exist” need not occur and numeral classifier phrases can freely occur in sentence initial positions. All constructions observed by Tsai fall into the category of “generic sentences” as described in Krifka et al. (1995). These constructions, as first observed in Tsai (2001) and modified by Liao (2011), are illustrated in (80).

(p.63)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

23

All numeral classifier phrases in these sentences do not refer to a plural individual in a particular event but a nonspecific plural individual across events, and it is the genericity of these sentences that gives rise to the generic interpretation of the numeral classifier phrases (cf. Section 2.2.3).

We saw earlier that generic sentences allow numeral classifier phrases to appear in the sentence initial position in Mandarin; what we are going to see that episodic sentences can also admit numeral classifier phrases in sentence initial positions in Mandarin.

(p.64) As observed in Chao (1968: 76), it is possible for indefinite expressions to appear in the subject position, as shown in (81a), although the preferred (i.e., more frequently occuring) forms would be the one in (81b).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

A number of scholars further observe a variety of contexts that allow an indefinite numeral-containing phrase to appear in the subject position: (i) when it is used referentially by modifying the NP with a vivid description (Lee 1986) (82a), (ii) when it is used with an intransitive verb that takes a complex form (Fan 1985) (82b), (iii) when it is used in very short sentences, expressing unexpected new discoveries and depending on the current scene or a known knowledge background (X. Zhu 1988) (82c), and (iv) when it is used with stage-level predicates instead of individual-level predicates (Shyu 1995) (82iv) (cf. Huang, Li and Li 2009: 318–320).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.65) On the top of Shyu’s (1995) observation, Huang, Li and Li (2009) notice that only some sentences with stage-level predicates are acceptable with indefinite subject and some others are not, as illustrated in (83). According to them, the addition of the existential you “have” in (83) makes a clear difference in acceptability.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Very crucially, Huang, Li, and Li point out that the contrast between sentences in (82) and those in (83) can be captured by the notion of thetic and categorical judgment in Kuroda (1972, 1992). The thetic judgment conforms to the traditional paradigm of subject-predicate and is a single cognitive act that recognizes the existence of an entity or event; in contrast, the categorical judgment respresents the recognition or rejection of material of a judgment and consists of two cognitive acts: the recognition of the subject (in the logical sense) and the acknowledgement (affirming or denying) of what is expressed by the predicate about the subject (Kuroda 1992: 22–23). A thetic judgment is expressed by a sentence that describes what is perceived by the speaker and can have an indefinite NP as its subject, the referent of which is only relevant to the unique current perceptual cognition. On the other hand, the categorical judgment has a “presupposed” subject in the sense that a precondition for making the judgment is that “the mind of the judger must be directed ftrst to an individual, before the predicate can be connected to it,” it cannot be nonspecific indefinite (Kuroda 1972, cf. Ladusaw 1994: 222–223).

Returning to the sentences in (82) and (83), the acceptable ones in (82) all describe/report perceptions of situations/events (e.g., the speaker can directly observe someone is looking for someone, jumping, fighting, or reading), whereas none of the unacceptable or degraded ones in (82) and (83) describe perceived situations. Specifically, the sentence in (82ivb) contains an individual-level predicate, which denotes permanent properties of an individual rather than expressing a thetic judgment; in the sentences in (83), a speaker genernally does not directly observe someone else’s earlier experiences, inaction, future plans (cf. Huang, Li and Li 2009: 323). Huang, Li, and Li then draw the generalization that indefinite numeral containing phrases can appear in the subject position when it occurs in a sentence expressing thetic judgment.

Although the distinction between thetic judgment and categorical judgment captures the contrast between (82) and (83), Huang, Li, and Li (2009: 323, fn. 32) also acknowledge that under certain circumstances, especially when a specific context is provided, it is possible to allow numeral indefinites to appear in subject position in sentences like those in (83). They take the sentence in (83a) as an example and provide the context in (84a); under this context, the sentence becomes (p.66) acceptable. A. Li (1998: 694, fn. 3) and Jiang (2012) make a similar observation. A. Li (1998: 694, fn. 3) notes that the reason why sentences with numeral classifier phrases in the sentence initial position like the one in (71a) are usually marked as unaccepted or unnatural in the literature is perhaps because these examples are cited out of context, and because such sentences are not acceptable unless there is a clear linguistic context for quantity. Jiang (2012) observes that if a sentence is not uttered in a “out-of-the-blue” situation but with specific contexts provided, indefinite numeral classifier phrases can appear in the sentence initial position with a specific indefinite reading, as illustrated in (84b, c).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

What this discussion shows is that in episodic sentences (in contrast with generic sentences), a numeral classifier phrase can appear in subject/sentence initial position with an indefinite interpretation in Mandarin, but the sentence containing the numeral indefinites needs to expresses thetic judgment or be uttered with specific contexts.

In object positions, indefinite numeral classifier phrases are freely allowed (e.g., Chao 1968; Li and Thompson 1981; Cheng and Sybesma 1999, 2005, among many others):

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.67) In addition to receiving generic and indefinite readings in argument positions, numeral containing phrases in Mandarin can also be used with definite elements to appear in argument positions in the same way as those in number marking languages (cf. (3), (18), (24)). Mandarin does not have overt definite determiners (the same as number languages like Russian), but it has demonstratives which can appear with numeral classifier phrases to yield a definite interpretation (86a); numeral classifier phrases in Mandarin can further be used with quantifiers such as mei “every” in argument positions, as illustrated in (86b).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Numeral classifier phrases in Mandarin lack the anaphoric use, just like numeral-noun phrases in number marking languages (cf. (7). (19, (25)). In (87a), ba ge nansheng “eight boys” cannot receive a definite reading and cannot anaphorically refer to the plural individual in the antecedent clause. After a demonstrative is added, the [Dem Num Cl NP] phrase receives a definite interpretation and can be used anaphorically (87b).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Last, the same as numeral-noun phrases in number marking languges (cf. (8), (20), (26), numeral classifier phrases in Mandarin also have entity level use:

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.68) The preceding discussion has demonstrated that Mandarin numeral classifier phrases behave quite similar to numeral-noun phrases in number marking languages with respect to distribution and interpretation. Mandarin numeral classifier phrases can appear in both predicate positions and argument positions. In argument positions, they behave like numeral indefinites in number marking languages and can receive long-distance scope interpretations and escape island constraints. Mandarin numeral classifier phrases can also receive a generic interpretation and freely appear in argument positions; they can be used with a quantificational expression like “every” to serve as generalized quantifiers or be used with a demonstrative to receive a definite interpretation.

There is, however, a distributional difference between numeral classifier phrases in Mandarin and numeral-noun phrases in number marking languages, namely, it is unnatural for Mandarin indefinite numeral classifier phrases to appear in sentence initial position unless a specific context is provided or a sentence expresses thetic judgment (cf. (82)/(84)).

In (89), I summarize argumental numeral classifier phrases in Mandarin with respect to their interpretation and distribution. In (90) through (94), I summarize the examples demonstrating all the properties of numeral classifier phrases in Mandarin that have been discussed.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.69)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.70)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

With a fuller picture of the interpretation and the distribution of Mandarin numeral classifier phrases and a better idea of their internal properties, I would like to understand a main puzzle: how can numeral classifier phrases in Mandarin be so similar to numeral-noun phrases in number marking languages in the clausal domain with respect to their syntax and semantics (cf. Section 2.3.2 and Section 2.3.3) but so different from them in the nominal internal domain (cf. Section 2.3.1)? Specifically, I would like to account for the following properties of numeral classifier phrases in Mandarin:

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

In the next subsection, I review previous analyses of Mandarin numeral classifier phrases to examine what issues have been solved and what issues remain unexplained.

(p.71) 2.3.4 Previous analyses of Mandarin numeral classifier phrases

The previous research on Mandarin numeral classifier phrases mainly focuses on explaining their distributional and interpretational differences, and those differences receive mainly structural accounts. I will review three representative analyses by A. Li (1997, 1998, 1999), Tsai (1999, 2001), and Liao (2011).

2.3.4.1 A. Li (1997, 1998, 1999): DP and NumP Analysis

As we have seen, A. Li (1997, et seq.) observes that Mandarin numeral classifier phrases can appear in sentence initial positions when they receive a generic interpretation (the so-called “quantity-denoting” use, in her terms), in contrast with indefinite numeral classifier phrases. Li proposes that the semantic interpretational differences correlate to two different syntactic structures which lead to their distributional differences.

According to Li, numeral classifier phrases with indefinite readings (“object-referring indefinite” or “individual-denoting” in her terms), which require you in the sentence initial position, project a DP with an empty D. Li assumes that the empty D contributes to the existential reading of numeral classifier phrases and is subject to the ECP, following Longobardi’s (1994) government-based DP hypothesis. This treatment ties the indefinite reading of numeral classifier phrases to a fixed syntactic position—D (97a).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

24

Regarding numeral classifier phrases with the generic interpretation (“number denoting” in her term), Li assumes that an intermediate projection exists between DP and NP—a NumP (97b). She compares nominal structure with clausal structure (p.72) by equating DP with CP (both functioning as arguments) and NP with VP (both functioning as predicates) and proposes that an intermediate structure between DP and NP, i.e., NumP, can occur without a dominating DP, and this NumP is equivalent to IP at the clausal level (IP can occur without a dominating CP, as in exceptional-Case-marking or raising structures). Li argues that NumPs on their own can also be argumental in Mandarin in addition to argumental DPs. Given that there is no empty D, the generic NumP is not subject to any syntactic constraint and can freely be argumental in any position. Li further provides three empirical arguments to support the differences between the indefinite DP and the generic NumP with respect to co-occurrence with operators like you “exist” and dou “all,” coreference and binding possibilities, and scope properties.

In addition to the observational contribution of numeral classifier phrases, as we saw in Section 2.3.3, Li’s work is important at least in two respects. The first is that her observation and analysis provide an empirical argument for the claim that the functional projection DP could exist in a language without overt determiners, such as Mandarin, which strengthens the DP hypothesis. The other is that she proposes that NumPs can be argumental on their own without a DP projection. I adopt her insight concerning argumental NumPs for Mandarin numeral classifier phrases. However, we shall see that the government-based DP analysis does not suit Mandarin numeral classifier phrases. Next, I consider three arguments against applying Longobardi’s DP hypothesis to Mandarin numeral indefinites (see also Yang 2001).

First, linking the indefinite readings of numeral classifier phrases to a syntactic position D which is subject to a licensing condition (i.e., government) is incapable explaining their interpretations and distributions in Mandarin.

Longobardi’s (1994) government-based DP hypothesis is designed to explain why Italian bare plurals and mass nouns, which can only receive a narrow scope existential interpretation (not a generic one), are only allowed in object positions not in subject positions:

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.73) The empty D, under Longobardi’s approach, contributes to the existential interpretation of Italian bare plurals and mass nouns and is subject to syntactic licensing.25 However, as we have seen earlier, numeral indefinites in Mandarin receive both specific and nonspecific interpretations, in addition to a generic interpretation. Further, generic and specific indefinite numeral classifier phrases both can appear in the sentence initial position (e.g., (70b), (82), (84)), and only nonspecific numeral classifier phrases are usually not preferred in this position (e.g., (70a)). Without making changes to the original DP hypothesis, the approach based on the empty D is unlikely to account for the interpretation and distribution of Mandarin numeral classifier phrases.26

Second, it is impossible to account for the long-distance scope interpretation and the island-escaping ability of indefinite numeral classifier phrases in Mandarin by assuming an empty D as in Longobardi 1994 or even in his later work. We elaborate this later.

Another important component of Longobardi’s DP hypothesis is that the empty D receives the narrowest possible scope (i.e., default existential) interpretation (Longobardi 1994: 618/641). This component finds its basis in the empirical facts concerning bare plurals/mass nouns in both Italian and English: they behave the same with respect to their interpretations and scope behavior. As Longobardi notes, Italian bare plurals/mass nouns receive only opaque interpretations and exhibit the narrowest scope behavior, just like English bare plurals/mass noun (Carlson 1977b). This empty D for bare plurals/mass nouns also differs from other overt indefinite Ds in that the latter can receive both opaque and transparent interpretations with long-distance scope, as we saw in (45) (repeated in (100)).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.74) Mandarin does not have overt indefinite Ds as Italian/English does, which were assumed to contribute to the long-distance scope interpretation of indefinites by Longobardi. Assuming an empty D projecting above numeral classifier phrases in Mandarin, one would expect numeral-classifier phrases in Mandarin to behave like bare plurals and mass nouns in Italian/English, which only receive a narrowest scope (existential) interpretation because the empty D only contributes to the narrow scope existential interpretation in Longobardi’s framework.27 However, this is not the case: Mandarin indefinite numeral classifier phrases behave like some/a and numeral indefinites in English regarding long-distance scope (cf. Section 2.3.2).

2.3.4.2 Tsai (1999, 2001): Extended Mapping Hypothesis

Tsai (1999, 2001) examines the environments where numeral classifier phrases can appear in the sentence initial position without the verb you “exist.” He observes an interaction between modality and generic interpretations (his “quantity-denoting” reading) of the numeral classifier phrases in all of these environments. Tsai’s observation indeed captures the fact that the interpretation of numeral classifier phrases in generic sentences is contributed by something in the clausal level rather than in the Num-ClP itself.

Based on the observation of the correlation between modality and sentence initial numeral classifier phrases, Tsai proposes an extended mapping hypothesis based on Diesing’s (1992) Mapping Hypothesis. In Tsai’s hypothesis, it is a VP-level existential operator that is responsible for the interpretation of the numeral classifier phrases (both indefinite and generic). In addition, whether verbs can undergo head movement to the head of IP or to that of ModP to bind the subject numeral classifier phrases is crucial in his account. In the case of generic sentences, according to Tsai, verbs can undergo V-to-Mod movement so that the existential operator associated with the verb can bind the numeral classifier phrase in the subject position. In contrast, in episodic sentences, verbs cannot undergo V-to-I movement in Mandarin, which, he assumes, is due to the lack of any agreement morphology in Mandarin. As a consequence, the existential operator on V cannot license the numeral classifier phrase in the subject position in episodic sentences, and the vacuous quantification is rescued by an overt existential operator you.

Tsai’s analysis provides a straightforward account for the reason why the verb you is required for sentence initial numeral classifier phrases in episodic sentences. However, it still faces three challenges. First, it is unclear how Tsai’s analysis explains the long-distance scope property of indefinite numeral classifier phrases (cf. Section 2.3.2) given that the existential closure operator is assumed to be local within the VP domain. Second, as we have seen, in episodic sentences, numeral classifier phrases can also appear in subject position when contexts are provided (p.75) (cf. (84)). Third, as for generic sentences, the reference of the numeral classifier phrases is not existential but generic, and numeral classifier phrases do not have existential interpretations in any situation/possible world, as we have discussed in Section 2.3.3 (see also E. Tsai (2011) for a review of Tsai (2001)).

2.3.4.3 Liao (2011): Parallel Merge and DP analysis

Liao (2011) proposes a parallel merge analysis with a modified version of Zamparelli’s (2000) two-layer DP hypothesis for Mandarin numeral classifier phrases. In brief, Liao separates both the indefinite and the generic force from numeral classifier phrases and attributes them to the inherent features [+existential] in Asp and [+Gen] in Mod, respectively. He assumes that D always projects above numeral classifier phrases in Mandarin and undergoes Parallel Merge with the head of AspP or that of ModP in narrow syntax. Furthermore, Liao follows Mergerdoomian (2008) and Tenny (2000) in assuming that AspP in Mandarin is located lower than vP but higher than VP. According to him, the head of AspP is associated with the existential closure operator with the [existential] feature if the Asp head is overt. As for ModP, which is located above vP, its head is associated with the Generic operator, along with the [Gen] feature in syntax. Liao explains the generic reading (“quantity denoting” in his work) and the existential reading of numeral classifier phrases through Multiple Agreement (both downward and upward Agree) (as in Pesetsky and Torrego 2007) and operator-variable unselective binding (as in Aoun and Li 1993a), with a series of other assumptions. In what follows we consider some of his other assumptions about the syntax and semantics of these constructions.

First, Liao assumes that an empty D head always projects with NumP in Mandarin and that numerals are the head of NumP in the syntax but a modifier of N in the semantics. He identifies his NumP to the PredDP in Zamparelli’s (2000) work and his DP to Zamparelli’s SDP (Liao 2011: 242).28

(p.76) Second, Liao further identifies his version of SDP with the DP in Li’s work and his PredDP to Li’s NumPs, namely DP receives an individual denoting reading whereas NumPs receive a quantity-denoting one.

Third, in the semantics, Liao adopts Lin’s (2004) analysis of the existential and generic operators in Mandarin, together with some further assumptions. Lin (2004) argues that the null Asp in Mandarin brings about a habitual/generic modal meaning, while dynamic aspects are associated with existential closure. Liao adopts Lin’s view and further assumes that the generic operator resides in the lower modal and that this ModP necessarily appears with a null Asp which does not have quantificational force (Liao 2011: 244). Liao assumes that the existential operator is realized on the overt AspP head and that the Generic operator is realized on the ModP head; these two operators bind their indefinite variables respectively through unselective binding, as illustrated in (101).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Syntactically, existential and generic force are realized as syntactic inherent features in Asp and Mod respectively. These features are valued through Spec-head agreement through feature specifications to D that have undergone parallel merge.29 Specifically, the null D carries an unvalued (but interpretable) quantificational feature, i[_ QF], looking for the closest possible valued feature (either upwardly or downwardly), and agrees through the Spec-head relation either with the head of AspP or ModP.

With these assumptions, Liao explains the number denoting reading (i.e., the generic reading of numeral classifier phrases) and the existential force of the DP in subject and object positions as follows. The quantificational force of quantity denoting numeral classifier phrases is valued by the feature [Gen] (a universal quantification) in Mod, which necessarily appears with a null Asp. As for the number denoting DP in object position, as the null Asp is assumed not to have quantificational force it enables the unvalued feature on D in the object position to continue looking upward until it encounters a valued feature to check with; this is the [Gen] feature on Mod. For the quantity denoting DP in subject position, its D directly merges with Mod and checks its Q feature with [Gen] on Mod. In order to enable the quantity denoting reading of both subjects and objects, Liao assumes Multiple-Agree in syntax and unselective binding in semantics (Liao 2011: Section 5.3.2). With regard to the existential reading of DPs in subject and object positions, Liao (2011: 253) suggests that sentence initial you and the subject DP are introduced in independent domains and that D is valued by the quantificational feature associated with you. For object DPs with existential interpretations, the uninterpretable [QF] feature in D is valued by Asp.

(p.77) Liao’s analysis correctly captures the source that contributes to the generic interpretation (his “quantity-denoting” reading) of numeral classifier phrases in generic sentences: the genericity in the sentence. Liao (2011: 50) adopts Kratzer’s 1981 work on the Gen operator, which, under his approach, is realized in Mod, quantifying over the sentence and contributing a universal quantification over possible worlds. However, Liao’s analysis faces two main emprical problems.

First, although it is correct to attribute the generic reference of numeral classifier phrases to a Gen operator in generic sentences that is introduced in a fixed high enough syntactic position (i.e., Mod) to quantify over both subjects and objects; however, it is incorrect to tie numeral classifier phrases’ indefinite interpretations tightly and solely to a fixed low syntactic position below vP (i.e., in AspP). To be concrete, this very local existential operator on the head of AspP which is assumed to be lower than vP in Liao’s work cannot provide the correct interpretation for numeral classifier phrases in the preverbal position in episodic sentences (cf. (82)/(84)). Second, this local existential operator lower than vP also cannot allow for the long-distance scope interpretation of numeral classifier phrases outside of vP.30

In addition to the above two empirical problems, there are two theoretical puzzles in Liao’s analysis. First, one may wonder how we should justify the language-specific assumptions about Mandarin made in Liao’s work (e.g., AspP locates lower than vP, the null Asp head does not have quantificational force but an overt Asp does, etc.).31 More importantly, these language-specific assumptions about Mandarin make it difficult to understand the rather similar behavior of its numeral containing phrases and those in other languages. Either we need to apply what Liao has assumed to other languages as well, or we need to make further language-specific assumptions for other languages.

The second puzzle is a broad one, namely, how we understand the assumptions and stipulations in the narrow syntax in Liao’s work. As Liao is pursuing a Minimalist approach for syntax, the purpose of which is a “slimmer and slimmer” UG by reducing most if not all principles/conditions/constraints to a third factor explanation (Chomsky (2005, sub seq.)),32 we need to know where we can attribute these assumptions and stipulations to: UG, the third factor, or somewhere else?

To summarize this section, previous analyses provide us with important observations and insights into the properties of numeral classifier phrases in (p.78) Mandarin and help us address some of issues we would like to understand in (96). Nevertheless, as we have seen, lots of puzzles still remain unsolved.

In the next section, I will provide an alternative analysis of numeral classifier phrases in Mandarin, which, as we will see, can derive in a straightforward way the similarities between numeral-containing phrases in Mandarin and those in number marking languages in the clausal domain as well as their differences in the nominal internal domain. Crucially, the analysis correctly predicts the scope behavior of bare nouns in Mandarin, as I will demonstrate in Section 2.5; it also has further implications for the numeral-less classifier phrase [Cl N] in Mandarin (to be shown in Section 2.6).

2.4 A D-less analysis of Mandarin numeral classifier phrases

In this section, I will start with the syntax of numeral classifier phrases in Mandarin with the goal to understand the uniform properties that numeral classifier phrases and number-noun phrases share cross-linguistically (i.e., issues in (96i)). Then I will move on to the semantics of bare nouns and that of classifiers in order to understand the differences between Mandarin and number marking languages in the internal norminal domain (i.e., issues in (96ii)).

2.4.1 Uniform D-less analysis of bare numeral-(classifier)-noun phrases

Section 2.2 and Section 2.3 showed to us that the behavior of bare numeral-noun phrases in number marking languages and that of numeral classifier phrases in Mandarin exhibit tremendous uniformity with respect to their scope behavior, distribution, and interpretation, so one may wonder whether the D-less analysis of numeral-noun phrases argued in Section 2.2.3 (cf. (41) and (46)) can be extended to numeral classifier phrases in Mandarin. Indeed, it is desirable to analyze bare numeral containing phrases across languages in a similar way since such an analysis not only accounts for their cross-linguistically uniform behaviors, but it also avoids language specific assumptions about bare numeral containing phrases. Under such an approach, let us first settle on the syntax of numeral classifier phrases.

Regarding the internal domain of numeral classifier phrases, many authors establish a connection between the number morphology and the classifier (Greenberg 1972; Sanches and Slobin 1973; T’sou 1976; Doetjes 1997, 2012; Chierchia 1998b; Cheng and Sybesma 1999; Fukui and Takano 2000; Borer 2005, Watanabe 2006, among others). They either observe that the use of number morphology and that of classifiers are in complementary distribution (Sanches and Slobin 1973: 4; T’sou 1976: 1216), or they argue that their roles are parallel, i.e., classifiers and the number morphology both signal the presence of minimal parts and that “numerals need the presence of a syntactic marker of countability which can be either individual (p.79) classifiers or number morphology” (Doetjes 1997: 35). Based on either their complementary distribution or their parallel role, a number of authors identify classifiers with number morphology and propose that they should appear in the same position in the structure (Doetjes 1997; Chierchia 1998b; Cheng and Sybesma 1999; Fukui and Takano 2000; Borer 2005; Watanabe 2006, among others). Although the details of their analyses of the internal structure of numeral classifier phrases may differ, these authors seem to agree on a structure roughly like the one in (102) (I refer the readers to Dékány 2011: 232–233 for a review of their work).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

In (102), the number morphology and the classifier appear in the same position, i.e., the head of the Division Phrase Div (or the head of the Classifier Phrase Cl); the Div head merges with an NP, forming a larger phrase (i.e., Div’ or Cl’) and further merging with a numeral n.33

I will adopt the structure in (102) to maintain the connection between the number morphology and the classifier. Regarding numerals, as argued in Section 2.2.2 and Section 2.2.3, they are phrasal in the syntax and property-seeking functions in the semantics across languages. The structure in (102), together with the proposed analysis of numerals in (46) (as repeated in (103b)), helps us arrive at a uniform D-less structure of bare numeral-(classifier)-noun phrases across languages (103a).34

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.80) In the rest of this work, I will use ClP instead of DivP in the discussion. Since the classifier-noun phrase Cl’ is able to combine directly with numerals, it follows naturally that Cl’ [Cl N] should denote properties. Then what remains unknown is the semantics of classifiers and that of bare nouns in ClPs, which I will discuss shortly in Section 2.4.2. Although we have not fully understood the internal domain of numeral classifier phrases in Mandarin yet, this analysis in (103) is sufficient to account for the uniform behavior of numeral-(classifier)-noun phrases in Mandarin and number marking languages at the clausal level without additional language specific assumptions. I will demonstrative this in what follows.

Let us first consider the predicative use of Mandarin numeral classifier phrases in (69), as repeated in (104); its analysis is given in (105) (with the unknown parts in the semantics of classifiers and bare nouns underlined).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

The predicative status of the numeral classifier phrase in (105) allows it to serve as the restrictor of demonstratives, as shown in (106) and (107).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.81)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

In (107), there are two points that I will address. First, I follow Kaplan (1989), Wolter (2006), and Dayal (2012) in assuming that demonstratives are property-seeking functions with an indexical specification. Second, I treat demonstratives syntactically differently from article determiners: the former appears in the specifier positions (of ClPs), whereas the latter is the head of DPs. In the following, I provide my motivation for treating demonstratives differently from article determiners.

In the syntax, demonstratives and determiners have traditionally been assigned to the same structural position (e.g., Jackendoff 1977). This analysis aimed to capture the fact that in languages like English article determiners and demonstratives cannot co-occur (i.e., they have the same distribution), as illustrated in (108).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Despite the fact just cited, many authors have argued that demonstratives and determiners do not occupy the same structural position (Giusti 1997, 2002; Brugé 2000, 2002; Brugé and Giusti 1996; Panagiotidis 2000; Grohmann and Panagiotidis 2004; Shlonsky 2004; Alexiadou et al. 2007, among others). One of their arguments for distinguishing the structural position of demonstratives from determiners is that in many languages, such as Romanian, Hungarian, Spanish, Greek, the two elements can co-occur, as exemplified in (109).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.82) These authors further argue that demonstratives are base-generated in a lower position between NP and DP and can undergo movement to a higher specifier position, e.g., Spec DP (e.g., Giusti 2002: 71–72; Alexiadou et al. 2007: 109).

Take Romanian as an example; its demonstrative can appear in the position preceding the bare noun (110a); when the enclitic article determiner is merged (110b), the demonstrative cannot appear in the position preceding the noun (110c) but only in the position following the noun (110d). Giusti (2002) argues that the position of the demonstrative in (110d) is an intermediate position lower than DP. She proposes the structure in (111) to derive the Romanian examples.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

In addition to this syntactic motivation, demontratives and the definite determiner also differ semantically. One diagnosis that can differentiate a genuine definite determiner from demonstratives is proposed in Löbner (1985) (“consistency” test in his term) and further discussed in Robinson (2005):

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

In (112), noun phrases with a definite determiner yield a contradiction when appearing with a predicate and its negation; whereas noun phrases with a demonstrative admit a sensible interpretation (113). Mandarin demonstratives zhe “this” and na “that” behave semantically on a par with demonstratives in English:

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.83) Based on this, I adopt the views that article determiners occur in the structural position in D and that demonstratives occur in a specifier position.

Next let us move on to the argumental use of Mandarin numeral classifier phrases. As we saw in Section 2.3.2 and Section 2.3.3, just like number marking languages, Mandarin allows numeral classifier phrases to appear in argument positions with an indefinite interpretation. The indefinite interpretation, as we have shown in Section 2.2.3, is brought about by the choice function variable. The choice function is attributed to the lexical entry of numerals in the proposed analysis (cf. Section 2.2.3). Consider the narrow scope existential reading of numeral classifier phrases in Mandarin first. One example is repeated in (115), with the analysis in (116).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

As we can see, the analysis in (116) allows Mandarin numeral classifier phrases to be argumental, leaving no room to project a DP in the syntax for indefinite numeral classifier phrases. Thus, Mandarin is D-less at the numeral classifier phrase level.

The lexical analysis of choice function also allows Mandarin numeral classifier phrases to have the long distance interpretation if a sentence contains other scope-bearing element(s). This long-distance ability can be explained in the same ways as it is in number marking languages (cf. (40) in Section 2.2.3). A Mandarin example is repeated in (117).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

The syntax and semantics of the example in (117) is as follows. In (118ia), the existential closure is within the antecedent of the conditional, a narrow scope reading derived, which is paraphrased as in (118ib) (with irrelevant details omitted). If the (p.84) existential closure is performed outside the conditional (118iia), a wide scope interpretation is achieved, as paraphrased in (118iib).

Very crucially, in the wide scope reading in (118ii), the choice function is existentially closed (∃f) outside the conditional (the island-escaping behavior) without extracting the indefinite numeral classifier phrase yi ge xuesheng “one student” out of the adjunct island; thus island constraints are not violated under this choice function analysis of numerals. I illustrate its structure in (118iii) (with irrelevant details omitted).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

As we can see in (118iii), the indefinite ClP yi ge xuesheng “one Cl student” still remains in the adjunct clause. The choice function variable f introduced in the numeral is existentially closed (∃f) outside the conditional. Since there is no movement (overt or covert) out of the adjunct clause, island-constraints are not violated. Regarding the implementation in the syntax, we can view that an agreement relation exists between the abstract existential closure ∃ and the choice function in the lexical entry of numerals (e.g., analogous to the analysis of negative concord in Zeijlstra 2004).

Given that numeral classifier phrases receive an indefinite interpretation in episodic sentences, their lack of anaphoric use is expected. One example is repeated below:

(p.85)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Besides receiving an indefinite interpretation, numeral classifier phrases in Mandarin can also receive a generic interpretation (as we saw in Section 2.3.3). Based on the analysis of generic sentences adopted in Section 2.2.3, I analyze an example from Mandarin (120a) with its structure and semantics illustrated in (120b/c) (with irrelevant details omitted).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

35

In (120), the generic numeral classifier phrase in Mandarin receives the same analysis as its English counterpart in (51) without any language-specific stipulations.

As we can see, the properties that numeral classifier phrases in Mandarin share with numeral noun phrases in number marking languages (that we saw in Section 2.3.2 and Section 2.3.3) can be accounted for via the proposed uniform analysis of (p.86) numeral containing phrases in (103). However, as we saw in Section 2.3.3, indefinite numeral classifier phrases in Mandarin do differ from their counterparts in number marking languages in one respect. That is, in Mandarin, sentences containing indefinite numeral classifier phrases in the sentence initial position are usually not that natural; a sentence either needs to have a sentence initial you “exist” or a specific context or to express thetic judgment in order to allow a numeral indefinite in the sentence initial position (cf. (82–84)). In contrast, there is no such restriction for number marking languages such as English and French. Some examples from Mandarin are repeated here.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

The distributional restriction of Mandarin numeral classifier phrases in (121a), as we saw in Section 2.3.3, is closely related to their interpretation. Indeed, a similar phenomenon is observed when Mandarin bare nouns appear in the sentence initial position, as we will see in Section 2.5. I will discuss and account for this Mandarin-specific interpretational restriction on nominals in sentence initial positions in Section 2.7.

In the next section, I will analyze the internal domain of numeral classifier phrases in Mandarin, focusing on the issues that have not been solved in (96ii), i.e., the function of classifiers and the semantics of bare nouns.

2.4.2 The function of classifiers and the semantics of bare nouns

This section examines bare nouns and classifiers with the goal to understand the issues regarding the differences between Mandarin and number marking languages in the nominal internal domain (96ii) (as repeated in (122)).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.87)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Numerals across languages, as argued in Section 2.2, are functions from properties into quantized properties or functions from properties to entities. When numerals combine with ClPs, it follows naturally that ClPs should denote properties, type <e, t>. What we need to figure out next, then, is the semantics of classifiers and that of bare nouns in ClPs. If we assume that bare nouns in Mandarins are of type α‎, the detail of which will be determined later, the type of classifiers might be of <α‎, <e,t>> (as determined by the semantics of the numeral n). Hence, the internal structure and semantics of numeral classifier phrases in Mandarin can be treated as the one here.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

To implement the idea in (123), I will consider two approaches. The first approach treats bare nouns in Mandarin and classifier languages in general as property-denoting of type <e, t>, as suggested in Krifka (2004) and Borer (2005), to be discussed in Section 2.4.2.1. The second approach treats bare nouns in Mandarin and classifier languages in general as kind-denoting of type <e>, as proposed in Krifka (1995) and Chierchia (1998b, 2010), to be presented Section 2.4.2.2. I will argue for the second approach (i.e., bare nouns in classifier languages are kind-referring) and show that the first approach is problematic theoretically and empirically, even though both approaches can derive the differences between Mandarin and number marking languages in the nominal domain.

2.4.2.1 Approach 1 (bare nouns are property-denoting) and its problems

Let me first consider the view that bare nouns in Mandarin and classifier languages in general denote properties, type <e, t> (e.g., Krifka 2004; Borer 2005). I will illustrate how this view derives the differences between Mandarin and number marking languages in the nominal internal domain.

Properties are generally viewed as the attribute of objects. For instance, the bare noun “dinosaurs” can express properties of dinosaurs (as in “Stegosaurus and (p.88) Diplodocus are dinosaurs.”), and the properties are attribute general to any individual instance of dinosaurs as well as a plurality of instances of dinosaurs.

If bare nouns in Mandarin are property-denoting, type <e, t>, then they will have to be true of something. There are two logical possibilities.

One is to assume that these nouns are mass only properties (e.g., Krifka 2004: 193). If so, classifiers are needed to quantize these nouns, i.e., to turn mass properties into natural subproperties (atomic or non-atomic), given that numerals need to combine with sets containing atoms/sets (cf. Section 2.2.2). Another possibility is to assume that bare nouns in Mandarin are properties which are underspecified/undifferentiated in their mass versus count denotation (e.g., Borer 2005: 94). Under this view, every noun can apply to either whole individuals or to their parts. Hence, a noun like “shrimp” in Mandarin, for example, will be true in a world of “shrimps” or “their parts” (i.e., shrimp meat). The same would have to be true of “dog,” “table,” “water,” “blood,” or any other noun. Since their denotations are unspecified, nouns in Mandarin cannot combine directly with numerals (i.e., numerals need to combine with sets containing atoms/sets). Similarly, it is reasonable to conjecture that some function becomes necessary in Mandarin to connect numerals with unspecified nouns. Classifiers provide such a function. The purpose of classifiers, under such an analysis, is to atomize or quantize the “underspecified/undifferentiated properties” (mass or count), i.e., to enable numerals to combine with a set containing atoms/sets.

As we can see, the property-denoting approach to bare nouns can provide a natural role for classifiers and addresses the issues regarding why numerals cannot combine with bare nouns directly in Mandarin and why classifiers are always required (i.e., the properties in (124a/b)). Two examples from Mandarin that follow illustrate how this approach works.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.89)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

In (124a), the bare noun pingguo “apple” is property-denoting (124ai); the individual classifier ge in (124av) applies to a mass property or a property unspecified in mass or count and extracts the set of the corresponding atoms, providing the correct semantics for the numeral san “three” to combine with. The atomizing function AT in the semantics of the individual classifier in (124a) takes a property and returns the set of atoms in the extension of such property (see, e.g., Chierchia 2008, 2010). The analysis of individual classifiers in (124a) can be extended to other types of classifiers as well. For instance, in (124b), the measure classifier bang “pound” combines with the undifferentiated property yingtao “cherry,” yielding a property of “cherry” measured in pound (124biii); this newly created property which contains countable atoms/groups allows the numeral to combine with them (124biv).

Given the non-argumental type of bare nouns under this property-denoting approach, to derive bare arguments in Mandarin (e.g., (59), as repeated in (125a/b)), an argumentizing operation becomes necessary for nouns in these languages. Such an operation is not hard to justify: we can assume either a covert D in the syntax (e.g., Longobardi 1994; Borer 2005), or a type-shifting operation in the semantics (TS in short) (e.g., Chierchia 1998b; Dayal 2004, 2010), or a combination of the two (e.g., Chierchia 1998b), as illustrated in (126a/b).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.90)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

By assuming the argumentizing operation(s) in (126a/b), the puzzle, namely why a bare noun can always directly merge with a verb and freely be argumental in Mandarin (i.e., (125c)), can be accounted for.

The property-denoting analysis of bare nouns, however, does require additional assumptions or analyses to account for the last puzzle in (122d), namely why numeral-less classifier phrases [Cl N] in Mandarin are so restricted in argument positions as we saw in (56) (repeated in (127)).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Specifically, since the classifiers serve to atomize or quantize the “mass-only properties’’/”underspecified properties” (mass or count), one may expect that the bare classifier-noun phrase [Cl N] in Mandarin, which is property denoting (e.g., (124avi)), can be turned into arguments via the argumentizing device availabe in this language, e.g., a null D or a type-shifting operation, in the same way as its bare nouns are turned into arguments (126a/b). Contrary to this expectation, numeral-less classifier phrases in Mandarin are not freely argumental but very restricted in argument positions (127). It is important, however, to acknowledge that the second approach to bare nouns also faces this puzzle, as we will see in Section 2.4.2.2. I will put aside this unsolved puzzle for now and will return to it in Section 2.6.

Although the “property” thesis of bare nouns just cited might seem appealing since it can derive most differences between Mandarin and number marking languages in the nominal internal domain, it runs into a series of problems.

First, a property-denoting analysis of bare nouns cannot provide good reasons to justify classifiers in general.

As both theoretical and experimental studies have argued, nouns in Mandarin and those in classifier languages in general are lexically distinguished between mass and count even though they are not overtly marked with such a distinction (e.g., (p.91) Soja, Carey and Spelke 1991; Imai and Gentner 1997; Cheng and Sybesma 1999; Cheng, Doetjes and Sybesma 2008; P. Li et al. 2009; Doetjes 2012, among others).

Back in the 60s, a widespread view was that the pre-linguistic child does not make the ontological distinction between objects and substances and that the ontology underlying natural language is induced in the course of language development (Quine 1960, 1969). Quine’s view meshed well with the then popular claim (often associated with the Sapir-Whorf hypothesis) that languages embody indefinitely varying ontological/conceptual systems that in turn reflect the forms of life of particular groups of speakers (cf. Chierchia 2010: 102). However, Quine’s view has been challenged by the work of cognitive psychologists since the 80s (e.g., Carey and Spelke 1991; Imai and Gentner 1997; P. Li et al. 2009, among others). In particular, Li et al. (2009: 518) argue that the count-mass distinction is a pre-linguistic ontological distinction central to our deep conceptual system, universally available for all speakers and resistant to linguistic or cultural influence.

Besides the work of cognitive psychologists, linguists have also provided arguments for the mass-count distinction in Mandarin and classifier languages in general (e.g., Doetjes 1997, 2012; Cheng and Sybesma 1999; Cheng, Doetjes and Sybesma 2008). In particular, Doetjes (1997: 35) argues that classifiers and the number morphology both signal the presence of minimal parts and that “numerals need the presence of a syntactic marker of countability which can be either individual classifiers or number morphology.” Doetjes’s (1997) insight captures the fact that both number morphology and individual classifiers indicate notional count nouns in contrast to mass nouns. Cheng and Sybesma (1999) further argue that the mass-count distinction manifests itself through the classifier system: one class of classifiers, i.e., “individual classifiers” (they are “count-classifiers” in Cheng and Sybesma 1999), must combine with nouns that are notional count. In contrast, the other classes of classifiers, such as “measure classifiers” and “container classifiers,” do not have such a restriction, i.e., they can combine with either notional count or notional mass nouns, as examplified here.36

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.92) The behavior of the individual classifiers presupposes that nouns in Mandarin are lexically divided to count and mass, e.g., it is the lexical property of shui “water” and mianfen “flour” that prevents them from combining with individual classifiers like ge in Mandarin.

If bare nouns in Mandarin are mass-only properties or are undistinguished in mass properties and count properties, the restriction on individual classifiers would have no force, and one would wrongly expect that individual classifiers like ge should work with both types of nouns just like other types of classifiers do. Hence, if nouns are properties in Mandarin, there is no good reason to justify classifiers. On the contrary, a kind-denoting analysis of bare nouns in classifier languages (Krifka 1995; Chierchia 1998b, 2010) will provide us a natural account for the obligatory existence of classifiers, as we will see shortly in Section 2.4.2.2).

Second, given the conclusion that most people are reaching, namely that the mass-count distinction is present at the lexical level or some pre-syntactic level, as we just discussed, one might expect there to be classifier languages that disallow (certain kind of) bare arguments. In particular, there ought to be an analogue of French among classifier languages that always disallows bare arguments, and overt Ds always project above bare nouns. There also ought be an analogue of English among classifier languages that disallows singular count nouns as bare arguments. Nevertheless, so far as we know, this does not happen in any classifier languages. On the contrary, the kind-denoting analysis of bare nouns in classifier languages would predict that such languages should not exist (to be shown in Section 2.4.2.2).

Admittedly, so far we only observe one language, namely French, which obligatorily projects overt Ds above all bare nominals (bare singulars, bare plurals, and mass nouns) (cf. (15) in Section 2.2.1). If languages disallowing bare arguments to the degree that French does are already rare, very little can be concluded from the fact that a language which is a combination of Mandarin and French does not exist. Nevertheless, there should be other languages like French which disallow bare arguments (i.e., the requirement imposed by French is not actually so rare), but we should not expect these languages to be classifier languages as well. It of course remains to be seen whether this prediction is borne out as classifier languages are investigated further.

Third, if we assume that bare nouns in Mandarin and classifier languages in general are property-denoting and may require a null D in the syntax to turn them into arguments (126a), we would predict that if overt determiners develop in the grammar of a classifier language, they should combine with bare nouns directly in the same way as the determiner the in English combines with bare nouns (e.g., the dog, the water). However, as we will see in Chapter 4, this is disallowed in Nuosu Yi, a classifier language with an overt definite determiner.

These three problems lead us to an alternative analysis in which nouns in Mandarin and classifier languges in general are kind-denoting.

2.4.2.2 Approach 2 (bare nouns are kind-denoting)

This section discusses the view that bare nouns in Mandarin and other classifier languages are kinds (i.e., entities), as proposed in Krifka (1995) and Chierchia (1998b, (p.93) 2010).37 I will illustrate how this view derives the differences between Mandarin and number marking languages in the nominal internal domain in (122).

Kinds are generally viewed as regularities occurring in nature and a plurality (or totality) of instances sharing some properties (Carlson 1977b: 173). For instance, in addition to denoting properties (as in “Stegosaurus and Diplodocus are dinosaurs.”, cf. Section 2.4.2.1), the bare noun “dinosaurs” can also refer to an entity (as in “Dinosaurs are extinct.”), a plurality/totality of instances with certain properties that distinguish itself from other kinds of animals, such as “fish.”

Very crucially, to treat bare nouns in Mandarin and classifier languages in general as kinds does not mean that nouns are all mass or that nouns do not make a distinction between mass and count in these languages. Instead, there can be count kinds (e.g., apple-kind) and mass kinds (e.g., water-kind) just like there can be count properties and mass properties; whether it is a count kind or a mass kind depends on whether the instances of the kind are atomic/whole objects or not (cf. Chierchia 2010: 131; see also Lima 2014). In other words, nouns in Mandarin do make a lexical distinction between mass and count. Such a conclusion has been argued for by both theoretical work and experimental work (e.g., Soja, Carey and Spelke 1991; Imai and Gentner 1997; Cheng and Sybesma 1999; Cheng, Doetjes, and Sybesma 2008; P. Li et al. 2009; Doetjes 2012, among others), as we just saw in Section 2.4.2.1.

I adopt the view that the mass-count distinction exists not only in number marking languages but also in classifier languages like Mandarin as argued for by the authors just cited. Under such a view, individual classifiers can be viewed as atomizing functions from count kinds to sets of atomic entities, type <ek, <e,t>> (e.g., Chierchia 2008, 2010; Dayal 2012, 2014).38 Let us look at an example to see how the kind-referring analysis works. In (130a), the numeral-classifier phrase san ge apple “three apples” has the semantics and structure in (130b) and (130c), respectively.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.94)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

In (130bi), the bare noun pingguo “apple” is kind-denoting. In (130biii), the “up”-operator predicativizes kinds and maps kind-level individuals to properties (Chierchia 1984, 1998b; Partee 1986). The individual classifier ge in (130biii) applies to a count kind and extracts the set of the corresponding atoms (i.e., the atomic instances of the apple-kind), providing the correct semantics for the numeral san “three” to combine with (Chierchia 2010). The atomizing function AT in the semantics of the classifier in (130ciii) takes a property and returns the set of atoms in the extension of such property. One consequence immediately following from the analysis in (130) is that individual classifiers will not apply to notional mass nouns (cf. (128b)) because mass nouns do not contains atoms.39

Note that, this analysis of individual classifiers in (130) can apply to other types of classifiers as well. Specifically, standard measure classifiers, like bang “pound,” measure parts under sum; kind classifiers, like zhong “kind,” measure subkinds; container classifies, like wan “bowl,” measure parts via a fill-in-relation with respect to the noun; group classifiers like qun “group,” measure sets formed as groups; partitive classifiers, such as ceng “layer,” measure partitions.40 Syntactically, all these classifiers are heads, merging with a kind-referring noun and forming a larger phrase Cl’, which further merges with a numeral. The semantics of each type of classifier and the internal structure of the classifier phrase are illustrated in (131).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.95)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

In addition to providing a uniform analysis of all types of classifiers in Mandarin, the analysis in (130) and (131) has two advantages. One, it allows a common syntax for classifiers in Mandarin in which all classifiers can appear in the position preceding a noun and following a numeral (132). To demonstrate the common structure of classifiers, I repeat some examples from Chapter 1 in (133).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.96)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Two, given that each type of classifiers has its own semantics which distinguishes it from the others (131), we expect that these classifiers, despite sharing the same structure in (132), may differ in their behaviors in other structures; this prediction is borne out.

One such variation is demonstrated in Chao (1968) (see also D. Zhu 1956, 1961, 1982; Chao 1968; Lu 1987; Cheng and Sybesma 1998, 1999; Tang 1990, 2005; X. Li 2007, 2011, 2013; Hsieh 2008; Jiang 2009; Liao and Wang 2011; D. Tsai 2011; Li and Rothstein 2012; S. Huang and Jenks 2014; Jin 2016, among many others). Chao (1968) observes that Mandarin individual classifiers behave differently with respect to de-insertion (de can be understood as a modification marker; e.g., see Chao 1968: 588). Individual classifiers resist de-insertion, but other classifiers allow it. A contrast between container classifiers and individual classifiers is illustrated here.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

De-construction can receive an interpretation similar to the psuedo-partitive construction in English (cf. Chapter 1); it describes the quantity of books as measured in the number of boxes those books fill (e.g., Cheng and Sybesma 1998; Tang 2005; X. Liao 2007, 2011, 2013; Hsieh 2008; Jiang 2009; Liao and Wang 2011; Li and Rothstein 2012). Such a construction excludes the appearance of individual classifiers (132b). Regarding how this contrast can be acconted for, I refer the readers to X. Li (2007, 2011, 2013) and Jiang (2009) for analyses compatible with the kind approach to bare nouns and to Cheng and Sybesma (1998), Tang (2005), Hsieh (2008) and Liao and Wang (2011) for different analyses.

Some variation is observed in Tang (2005), X. Li (2011, 2013) and Hsieh (2008), namely, that with a certain context of quantification of “aboutness” rather than “exactness” of the quantity of the noun, some individual classifiers can enter de-construction if the numeral is large enough.

Other variation can be seen in Jiang (2009) (see also Tang 2005, Hsieh, 2008; Li 2011, 2013; Tsai 2011; Zhang 2013; Jin 2016; among others). Jiang observes that de-construction can express a quality-denoting interpretation with standard measure classifiers in addition to the quantity-denoting interpretation. She argues that the two different interpretations of standard measures correspond to two different internal structures: an attributive modifier structure or a pseudo-partitive structure, as Schwarzschild (2006) argues for in English. Other variation among different types of classifiers is also observed and can follow from the analysis of classifiers in (128) and (129); however, illustrating this variation here will take us too far afield. I refer (p.97) the readers to Zhang (2013: 157–172) and Jiang et al. (2020) for other semantic and syntactic differences among different types of classifiers.41

In addition to these advantages, the analysis of bare nouns in Mandarin as kind-referring (128) has three immediate consequences. First, it explains why numerals cannot directly combine with a noun in Mandarin (i.e., the first puzzle in (122a)). One example is repeated in (135a). In the syntax, the numeral merges with the bare noun (135b); when numerals, which are property-seeking, combine with kind-referring nouns in the semantics, a type-mismatch arises, preventing numerals from combining directly with nouns (135c).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Second, the kind approach to bare nouns in Mandarin provides a natural account for the obligatory existence of classifiers (i.e., the first puzzle in (122b)): classifiers turn kind-referring nouns into properties so that the type-mismatch can be resolved (Chierchia 1998b, 2010).

Third, the analysis of bare nouns in classifier languages as kinds explains why bare nouns always directly merge with a verb and can freely be argumental in Mandarin (59) (i.e., the third puzzle in (122c)). As being kind-referring, Mandarin bare nouns denote entities which satisfy the semantic requirement of a verb; therefore, they are predicted to always be argumental. I illustrate this with the following example.42

(p.98)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

This kind-denoting analysis of bare nouns would predict that bare arguments are always allowed in classifier languages (Chierchia 1998b, 2010). In other words, we would expect a classifier language that disallows bare arguments should not exist given that its bare nouns are inherently argumental.

Under this analysis of bare nouns and the proposed analysis of numeral classifier phrases in Section 2.4.1, bare nouns and numeral classifier phrases in Mandarin are both inherently argumental; hence it is no longer necessary, nor even possible to assume a null D in order to account for the behavior of the nominal arguments. As a result, nominal arguments in Mandarin are D-less:

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

The kind-denoting approach to bare nouns, however, does require additional assumptions or analyses to account for the third puzzle in (122d), just like the property-denoting approach to bare nouns does. Specifically, since the semantics of classifiers serves to turn kind-referring nouns to sets that contain atoms/sets, such an analysis predicts that numerals are optional and that the bare classifier-noun phrase [Cl N] can be turned into arguments via an argumentizing device availabe in this language (e.g., a null D or a type-shifting operation). However, numeral-less classifier phrases in Mandarin are so restricted in argument positions as we saw in (56) (as repeated in (138)). I will return to this issue in Section 2.6.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.99)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

It is important to point out that the kind-denoting approach to bare nouns in classifier languages does face empirical challenges.

A widespread view is that classifier languages do not have overt determiners (Chierchia 1998b; Bošcović 2012a, 2014). In particular, Chierchia (1998b) speculates that classifier languages should not develop article determiners in their grammar: since nouns in classifier languages are argumental, considerations of economy rule out the presence of a determiner. However, as we will see in Chapter 4, the data from Nuosu Yi alters the empirical generalization regarding classifier languages and calls for an explanation. I will show in Chapter 4 that a modification of Chierchia’s (1998b) framework is needed and that the modified framework explains why classifier languages with overt Ds are possible but rare as well as allowing us to make further predictions about classifier languages (to be shown in Chapter 5).

To summarize this subsection, I have considered two different analyses of bare nouns in Mandarin with the goal to account for the differences between Mandarin and number marking languages in their nominal internal domain (i.e., the puzzles in (124)). I settled on an analysis in which nouns in Mandarin are kind-denoting, as proposed in Krifka (1995) and Chierchia (1998b, 2010). Classifiers, under this analysis, are required to turn kind-referring nouns into sets that contain atoms or sets. Thus, Mandarin, under this analysis, is a language with nouns of type <e>.

2.4.3 Summary and predictions

To summarize Section 2.4, I argued that the D-less analysis of numeral-noun phrases argued in Section 2.2.3 (cf. (41) and (46)) can be extended to numeral classifier phrases in Mandarin (cf. (103). As we saw, it is desirable to analyze bare numeral containing phrases across languages in a similar way since such an analysis not only accounts for their cross-linguistically uniform behaviors (as we saw in Section 2.3.2 and Section 2.3.3), but it also avoids language specific assumptions about bare numeral containing phrases. To account for the differences between Mandarin and number marking languages in the nominal internal domain (as we saw in Section 2.3.1), I argued for a kind-referring analysis of Mandarin bare nouns, as proposed in Krifka (1995) and Chierchia (1998b). Specifically, I argued that language variation in the nominal domain is primarily located in two interrelated factors: what nouns denote and what lower functional heads (i.e., number morphology and classifiers) denote. I demonstrated that such an analysis of Mandarin bare nouns provides rather straightforward accounts for the key puzzles of Mandarin numeral classifier phrases—”why numerals cannot combine directly with a bare noun,” “why classifiers are obligatory,” and “why all bare nouns can freely and directly merge with a verb.”

Importantly, bare nouns in Mandarin, under our analysis, are expected to behave like bare nominals in English (bare plurals/mass) which are also kind-referring with respect (p.100) to their scope behavior. Specifically we predict that Mandarin bare nouns receive the narrowest scope existential interpretation, same as bare plurals as observed in Carlson (1977b) rather than English indefinites. This prediction is borne out by the observations made in Yang (2001) and X. Li (2011, 2013) about bare nouns in Mandarin, which I will illustrate and discuss in the next section. I will also demonstrate how the different interpretations of Mandarin bare nouns are brought about with our analysis.

2.5 Bare Nouns in Mandarin

2.5.1 Scope behavior of Mandarin bare nouns

As discussed earlier, it has been observed in Carlson (1977b) that indefinites (e.g., “a guest” and “some guests”) and bare nominals (bare plurals like “guests” or mass nouns as “cream”) behave differently with regard to their scope behavior. Indefinites can receive both narrow scope and wide scope interpretations, while bare nominals, which are kind-referring, can only receive a narrow scope existential interpretation (cf. (45)). One example is repeated in (139).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Since Carlson (1977b), bare nominal argument terms have been treated to refer to kinds (e.g., Carlson 1977b; Krifka 1988; Wilkinson 1991; Chierchia 1998b; Dayal 2004). If we assume bare nouns in Mandarin are uniformly kind-referring, we predict that their scope behavior should resemble that of English bare nominals rather than that of its indefinites. This prediction is borne out.

Yang (2001), Dayal (2004), and X. Li (2011, 2013) examine bare nouns in Mandarin in detail; they observe that bare nouns in Mandarin do not exhibit any of the long-distance scope behavior of English indefinites like “some NPs” or “an NP.” Instead, Mandarin bare nouns behave like English bare nominals and receive the same narrowest scope interpretation in relation to other scope-bearing elements.

Let us first look at one example from Yang (2001). In (140a), the bare noun wenti “problem” can receive a definite reading (140ai) or a narrow scope existential reading (140aii);43 however it cannot receive an intermediate scope reading like indefinites “some problem” or “a problem,” as paraphrased in (140aiii). Look at another example by X. Li (2011). In (140b), the (modified) noun xin “letter” cannot receive the wide scope interpretation in (140biii). The only possible readings are the narrow scope reading (140bii) or the definite reading in (140bi).

(p.101)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

The preceding examples containing bare nouns contrast with the examples containing bare numeral classifier phrases, which can receive a wide scope or an intermediate scope reading in addition to a narrow scope reading as we saw Section 2.3.2 (cf. (62)–(68)). Some examples are repeated in (141i), with the bare noun counterparts provided in (141ii)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.102)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

The analysis that Mandarin bare nouns are kind-referring correctly predicts their scope behavior, which resembles English bare nominals and not English indefinites. In the next section, I briefly summarize the interpretations and the distribution of Mandarin bare nouns; I illustrate in Section 2.5.3 that the analysis of bare nouns in Mandarin as kinds can also derive their other interpretations in a principled manner.

(p.103) 2.5.2 Interpretation and distribution of Mandarin bare nouns

Mandarin bare nouns can appear in both pre-verbal and post-verbal positions. As noted in the literature (e.g., Krifka 1995; F. Liu 1997; Cheng and Sybesma 1999; Yang 2001), bare nouns in Mandarin can freely serve as arguments, appearing with kind-level predicates (142i), in generic sentences (142ii), as well as in episodic sentences (142iii).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

In episodic statements like (142iii), bare nouns receive narrow scope existential and definite readings, as we also saw in the previous section (Section 2.5.1). The ability of bare nouns to function as definites is also illustrated in (143), in which the linguistic context brings out their anaphoric use (e.g., Dayal 2004, 2011; X. Li 2011, 2013; Jiang 2012).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.104) Recently, Jenks (2018) claims that bare nouns in Mandarin cannot be anaphoric (144a–c); the one exception he allows are bare nouns in subject/topic position (144a, d, e).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

According to Jenks, the bare nouns in (144b, c) are judged infelicitous in object and indirect object positions, and demonstratives must be used, but in (144d, e), anaphoric bare nouns are possible in subject position.

Dayal and Jiang (submitted) argue against the claim in Jenks (2018) that Mandarin bare nouns cannot be anaphoric definites on the basis of elicited as well as naturally occurring data. In (145), I provide some of their examples from Beijing Language and Culture University Contemporary Chinese Corpus (BCC Corpus in short, see Gou et al. 2016):44

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.105)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

In (145a, b), the bare nouns nanhai “boy” and gou “dog” appear as objects in the second clause, referring anaphorically to indefinite expressions in the antecedent clauses. In (145c), the bare noun nühai “girl” in indirect object position can also refer anaphorically to the indefinite in the antecedent clause.

Based on the preceding data, I will maintain the generalization that Mandarin bare nouns can be anaphoric, appearing in both subject and object positions.45

Note that, it has been observed in the earlier work that in subject position Mandarin bare nouns have a strong tendency to be interpreted as definite (Chao 1968; Li and Thompson 1981). Some scholars further assume that indefinites are excluded from subject position in Mandarin (e.g., Cheng and Sybesma 1999). Some examples to illustrate this are given in (146).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.106)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

When adding the verb you “exist/have” in the sentence initial position, the bare noun in the subject position receives an existential interpretation, as illustrated in (147).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

In the later work, however, it has been observed that various examples do not following the tendency observed in the earlier work (e.g., J. Huang 1997; Yang 2001). J. Huang (1997) observes that bare nouns in sentence initial positions can receive an interpretation similar to English bare plurals, as exemplified in (148a, b). Yang (2001) further observes that the existential reading of bare nouns become immediately available on pre-verbal bare nouns, once they occur in the context of left peripheral locatives, as shown in (142iiib) (as repeated in (148c)).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Importantly, the behavior of Mandarin bare nouns as presented here is analogous to that of numeral classifier phrases in sentence-initial positions that we saw in Section 2.3.3, namely that without contexts provided it is unnatural for Mandarin indefinite (p.107) numeral classifier phrases to appear in the sentence initial position and that adding the verb you “exist/you” in the sentence initial position can make such sentences natural. A set of examples are repeated here.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

The interpretational tendency of bare nouns and the related phenomenon regarding numeral classifier phrases as seen in (146) to (148) will be further discussed and accounted for in Section 2.7. In the following section, I will illustrate that the analysis of bare nouns as kinds not only captures their scope behavior but also derives their other interpretations in (142).

2.5.3 Derive other interpretations of bare nouns from kinds

In this section, I illustrate that the object level meanings of bare nouns in Mandarin can be derived from their basic kind level meanings within a Neocarlsonian account of bare nominals (Carlson 1977b, 1989; Chierchia 1998b, 2010, 2016; Dayal 2004, 2011a).46 This approach has been pursued in Yang (2001) and X. Li (2011, 2013) for Mandarin. The Neocarlsonian account includes a set of ranked type-shifting operations and principle blocking covert type-shifting operations in the presence of corresponding overt correspondents. Justification for the choices made here over other possibilities will be presented in Chapter 4, Section 4.8.6, after I discuss a typologically unusual classifier language Nuosu Yi.

(p.108) I give below the specific version of the Neocarlsonian approach adopted in this work, due to Chierchia (1998b), with the specific modification of Rank of Meaning from Dayal (2004).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

47

In the two kind-related type shifts in (150a), the “up”-operator predicativizes kinds and maps kind-level individuals to properties. In (150b), the three canonical argument forming operations turn properties into arguments. The “down”-operator in (150bi) nominalizes, mapping those properties that correspond to kinds to kind individuals (e.g., Chierchia 1984; Partee 1987). Importantly, plural properties can be turned into kinds, but singular ones cannot. This is so because the semantics of singularity clashes with the conceptual notion of a kind which corresponds to the plurality of all instances of the property (Dayal 1992; Chierchia 1998b). In (150bii), the iota operator “ι‎” shifts properties to arguments with a definite interpretation and is used to interpret the definite article the in English (e.g., Sharvy 1980); in (150biii), “∃” shifts properties into existential generalized quantifiers and is traditionally taken to be the meaning of the indefinite article a/an in English (e.g., Montague 1974: 216).

In Chierchia (2016), a uniform term “ARG” is used to represent covert argument-forming operations which turn properties into arguments, i.e., a null D in the syntax and an invisible type-shift in the semantics. ARG can be viewed as a variable ranging (p.109) over the three canonical argument forming operations: , ι‎, and ∃ in (150b), and its interpretation is subject to the two principles in (151), Ranking of Meaning and Blocking Principle. I adopt Chierchia’s term “ARG” in this work and use it to represent either a null D in the syntax and an invisible type-shift in the semantics, both of which are covert argument forming operations.

A crucial point for the Neocarlsonian view is the difference between the indefinite readings of bare plurals and ordinary indefinites, as observed in Carlson (1977a, b), as we saw in (45). I repeat the examples in what follows, in which bare nouns allow only narrow scope indefinite readings, while indefinites participate in scope interaction:

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Furthermore, indefinites can also receive an intermediate scope interpretation, as we have seen in Section 2.2.1.

Regarding the narrow scope existential reading of bare nominals in episodic sentences (152a)/(153a), Chierchia (1992, 1998b) proposes a sortal adjustment operation, Derived Kind Predication (DKP in short). DKP adjusts predicates when a sentence has a predicate that is primarily object-level with a kind-level argument by introducing a local existential quantification over instances of the kind, which derives the existential interpretation of bare nominals from their kind denotation. DKP is defined in (154a) and illustrated in (154b–d).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

As this existential quantification in (154) is locally introduced, DKP has the advantage in explaining the narrow scope behavior of bare plurals, like the one in (152)/(153).

(p.110) The difference between bare nouns and indefinites in (152) and (153) is then explained via the sortal adjusting operation DKP in (150aii) and the (choice function variable that is subject to) existential closure ∃ in (150biii): the former derives the obligatory narrow scope indefinite reading of kind-denoting bare nominals (153a)/(153a), and the latter derives the flexible scope interpretations of ordinary indefinites (152b)/(153b) (as we saw in Section 2.2.3).

DKP bears some analogy with noun-incorporation (van Geenhoven 1995, 1998; McNally 1995; Dayal 2011a) in that it behaves like a lexical operation on the predicate. The noun incorporation analysis assumes that nouns simply denote properties. The mismatch between property-denoting nouns and verbs, which take objects as arguments, is solved with a lexical operation on the predicate which introduces an existential quantifier over instances of the property. I illustrate the effects of this operation on a simple verb in (i). However, as argued in Chierchia (1998b), DKP is more explanatory than noun-incorporation for the following reason. DKP, a type shifter that applies on demand, can explain sentences like the ones in (ii) which receive either a kind-oriented reading or an object-oriented reading. In contrast, noun incorporation, a lexical operation on the predicate, can only account the object-oriented reading in (ii) (see Chierchia 1998b: 365–366 for details).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

In addition, DKP can also account for the suspension of scopelessness of modified bare nominals as the one in (157). The property-denoting nominal “roses that I was looking for” cannot be defined as a kind (as typical kind-level predicates are marginal with nominals of this sort), so after the kind operator applies to it, “(roses that I was looking for)” cannot be well defined (as kinds). Accordingly, the only other way to make this property-denoting nominal argumental is through a regular existential generalized quantifier ∃, which should be expected to interact scopally with other scope-bearing element. This, indeed, is the case: the modified nominal in (157) can take wide scope over negation, thus suspending the scopelessness of bare arguments, in contrast with (153).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Next, let me briefly discuss the reason for the ranking in (151a) that I adopt. In Chierchia (1998b), “” ranks over “ι‎” and “∃” (151ai); this ranking is motivated by the fact that (English) plurals generally favor the kind interpretation over the indefinite one (158a). Chierchia claims that “∃” comes into the picture when “” is undefined (158b, c).48

(p.111)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

A further explanation is that “” only changes the type of its arguments without changing the information associated with it, but “∃” introduces quantificational force in addition to changing the type of its arguments. Kind formation “,” therefore, is more meaning-preserving than “∃” and should get picked whenever possible. Dayal (2004), however, notes that Chierchia’s ranking in (151ai) would block bare nominals in determiner-less languages from having any object level meaning, definite or indefinite. She also notes that the same reasoning that favors “” over “∃” should apply to “ι‎” as it also merely changes the type of its arguments without adding quantificational force. The revised ranking (151aii) explains the fact that bare nominals can denote kinds as well as contextually-salient entities in languages without definite determiners, such as Hindi and Russian. That is, definite readings are never blocked by kind formation in such languages. Ranking “∃” below “ι‎” is based on her claim that bare nouns in such languages are not bona fide indefinites and that their indefinite readings are derived from their kind-level meaning:

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

The last piece of the Neocarlsonian account of bare nominals that will be relevant to us is the Blocking Principle in (151b) that favors overt type-shifting operations over the corresponding covert ones. The Blocking Principle is what explains the difference between the anaphoric potential of bare nominals in languages like English as opposed to languages like Hindi or Mandarin, for example:

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

With this background in place, let us return to Mandarin. Yang (2001) and X. Li (2011) examine bare nouns in Mandarin in detail and argue that bare nouns in Mandarin denote kinds in the same way as English bare nominals do. As they are kind-referring, bare nouns in Mandarin can also achieve their narrow scope existential reading via DKP; thus their scope behavior is correctly captured and predicted (p.112) by the original assumption that bare nouns in Mandarin are kind-referring. The generic interpretation of bare nouns in Mandarin can also be explained in the same way as for English bare nominals, i.e., Gen can operate directly on kind-denoting nominals and introduce variables ranging over instances of the kind, so the generic references of the bare nominal in generic sentences is predicted and captured (Chierchia 1998b).

In (142), we saw that Mandarin bare nouns receive a range of different interpretations; in (161), I repeat these examples and present the account in Yang (2001) that derives the different readings of bare nouns from their kind reference:

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Note that, although DKP can derive the narrow scope existential reading of bare nouns direct from kinds, their “definite” interpretation still needs to be accounted for. Yang (2001: 31–32) proposes that the definite interpretation of Mandarin bare nouns is given by the iota operator in the semantics, as shown in (161cii). Yang first applies the type-lifting up-operator “” to the Chinese kind-denoting bare noun to turn it into properties, and then applies the iota operator to the property created by the up-operator, as demonstrated in (162).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Although the approach in (162) can derive the definite reading from the kind reference, it faces one theoretical problem and one emprical problem, which I will present in what follows. These problems will lead us to pursue a simpler way to derive definites from kinds.

Theoretically, the approach in (162) is not economic since it involves redundant computational steps, which is a departure from computational efficiency, a third (p.113) factor principle (Chomsky’s 2005 term) not specific to but also applicable to the human faculty of language. To illustrate, type-shifting or D is supposed to repair type mismatch and make the “unsaturated” predicative noun saturated and argumental (as in Higginbotham 1987; Szabolcsi 1992). Nevertheless, in (162) type mismatch is “created” on purpose just so it can feed the use of type-shifting in semantics. Furthermore, an already argumental noun is “forced” to turn non-argumental and then back to argumental in (162), which involves two redundant operations in that it gives rise to the same “argumental” result. Although these two operations return an individual different from the one that we start with, they are still less economic if we can achieve the same result with only one operation.49

In addition to being theoretically disfavored, the approach in (162) is also not supported by cross-linguistic data. In Chapter 4, a classifier language Nuosu Yi will be shown to have an overt definite determiner su which has the same functions as the determiner the in English. If the assumption in (162) is not language specific but apply universally, one would predict that the definite determiner su in Nuo Yi should be able to combine with kind-denoting bare nouns after they are turned into properties by the type-lifting up-operator “” and derive a definite interpretation, like (163a). However, this prediction is not borne out: su can never be used directly with a bare noun in Yi without a classifier (163b). Crucially, bare nouns in Yi behave like those in Mandarin and can also receive a definite interpretation in addition to the definite classifier phrases with su (the details will be provided in Chapter 4). If the assumption in (162) is language-specific that only applies to some classifier languages like Mandarin, we would like to know why it does not apply to some other classifier languages, like Nuosu Yi, and how the definite interpretation of bare nouns in Nuosu Yi is obtained.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.114) Given the earlier challenges, let us consider a simpler approach to derive definites from kinds. A direct way to obtain the definiteness, as proposed in Jiang (2012, 2014, 2018), is by plugging into the kind a situation variable, provided by the context. This operation applies to kinds by restricting them to specific situations so that the definite interpretation can be obtained, as demonstrated in (164).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

From here on, let Situation Restriction (SR in short) refer to the above strategy. Since kinds are functions from situations to (possibly plural) individuals (e.g., see Chierchia 1998b: 349), by applying them to a “resource” situation (an analogue of Domain Selection, e.g., see von Fintel 1994; Elbourne 2001) we get an individual e. Situation Restriction can be viewed as a function from kinds to situation restricted kinds, of type <ek, e>; it restricts kinds to a specific situation and obtains the maximal members in that situation provided the context. Importantly, the function of SR is the same as that of the iota operator (ι‎), the canonical Frege-Russell definition of “the largest member of X if there is one (else, undefined)” (cf. Chierchia 1998b: 346 and see Sharvy 1980 for the same analysis for English definite determiner). Thus, via SR, a definite interpretation of bare nouns can be achieved. (One might also note that reference to the maximal sum in a situation here plays a role similar to reference to stages in Carlson’s 1977 a theory.)50

Note that, the proposed approach to derive definites from kinds for Mandarin bare nouns is along the same lines in Tue (2011) and Dayal (2011b, 2012). In particular Dayal (2012) derives the definite reading of Bangla bare nouns via taking the extension of the kind at the evaluation index, as illustrated below:

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

I illustrate how Situation Restriction works with an example repeated below. The bare noun nansheng “boy” in (166a) receives a definite interpretation that refers anaphorically to the noun “a boy sitting in the house” in the antecedent sentence. The definite interpretation of nansheng “boy” is achieved by applying Situation Restriction to the extension of nansheng “boy<ek>” and obtaining the maximal member that instantiates the boy-kind in a situation s′. Syntactically, the bare noun nansheng does not have any further functional projection above it—it is simply a bare noun, as shown in (166b). The bare noun is interpreted in the semantics as definite through Situation Restriction (166c), and the whole sentence in (166a) has the semantics in (166d).

(p.115)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

The proposed Situation Restriction shares similarities with the E-type anaphora which is analyzed as definite articles followed by an NP which is deleted in the phonology in Elbourne (2001). SR plugs into kinds a situation variable provided by the context, deriving a definite interpretation of bare nouns. As for the E-type anaphora in donkey sentences like “if a farmer owns a donkey, he beats it,” the definites (i.e., the anaphora) in the consequent referring to entities introduced in the antecedent have the situation variables of predicates in the antecedent (cf. Elbourne 2001: 260).

Crucially, this “maximal member” view of definiteness in Mandarin bare nouns differs from what X. Li (2011) and Li and Bisang (2012) propose for definiteness in Mandarin, according to whom definiteness is characterized by the pragmatic notion of “identifiability.” Their view of definiteness might be too weak to capture the anaphoric use of bare nouns. Bare nouns in Mandarin can anaphorically refer to the maximal plural member in an antecedent clause, as illustrated here.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Merely being able to identify with some members rather than the maximal member in a specific situation is not strong enough to capture the anaphoric use of Mandarin bare nouns which can denote the largest members in a domain. This anaphoric use of (p.116) bare nouns in Mandarin requires the source that contributes to its definiteness to function as the canonical Frege-Russell definition of the iota operator (ι‎).

The identifiability and maximality difference in Mandarin bare nouns is reminiscent of a similar phenomenon in English, which is not fully understood. As observed by Condoravdi (1997), English bare plurals can be used to refer to a context-anchored particular plural individual as in (168a). Similar examples have been provided by Dayal (2013), as in (168b, c):

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

The referential ability of the bare plurals in (168) seems to suggest that bare plurals in English behave like definite plurals and also seems to suggest the operation of iota (cf. Dayal 2013). Nevertheless, we cannot equate bare plurals in English in (168) with definite plurals. There are at least two reasons discussed in Dayal (2013). First, there are structurally parallel sentences such as the ones in (167) where the bare plurals appear infelicitous and definite plurals have to be used (cf. Dayal 2013: 58).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Second, bare plurals in English do not admit anaphoric definite readings (cf. Dayal 2013: 54). For example in (170), the bare plural “children” cannot be used anaphorically to refer to the maximal member in the antecedent clause, and the definite determiner “the” has to be used. The anaphoric ability in (170) is enforced and defined by the iota operator and brought up via the definite determiner (the), and this is an ability that bare plurals lack.51

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Turning to bare nouns in Mandarin, they have a wider range of usages than bare plurals in Englishi, i.e., they receive a definite interpretation and behave like definites in English in addition to the existential and kind interpretations, as we saw in (161). In what follows we provide three examples from Mandarin in (171) to (173) which (p.117) correspond to the three cases in (168) to (170) above. We can see that bare nouns in Mandarin can be used in all of the three preceding cases.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

These examples show that bare nouns in Mandarin can behave like definite plurals in English, in addition to behaving like English bare plurals (as we saw in Section 2.5.1). The proposed Situation Restriction can be viewed as analogous to the iota operator and can be used to account for the cases in (171) to (173).

Although the present analysis of definite bare nouns in Mandarin differs from the one proposed by Li (2011) and Li and Bisang (2012), this view is compatible with theirs given that the notion of “maximality” is related to “identifiability.” In order to refer to the maximal sum of individuals in a situation, these individuals should be identifiable first.

Before I end my discussion on Mandarin bare nouns, I would like to address one last issue, namely the use of Mandarin bare nouns with demonstratives. I will show in the following section that the proposed analysis of Situation Restriction on bare nouns in (164) has important consequences for demonstratives.

2.5.4 Bare nouns with demonstratives

As we saw in Section 2.3.3, demonstratives in Mandarin can combine with numeral classifier phrases (cf. (92a)). It has been widely observed in the literature that demonstratives can combine directly with bare nouns in Mandarin (Chao 1968; Tang 1990, 2007; A. Li 1999; Cheng and Sybesma 1999, 2014; Yang 2001; Wang 2005; H. Yang 2005; X. Li 2011, 2013, among others), as illustrated here.

(p.118)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

As far as I am aware, the preceding pattern in Mandarin is representative of classifier languages generally. The need for a classifier to mediate between a noun and a demonstrative has been previously noted for Burmese-Yipho languages in general (Xu 2001: 205), Cantonese (Matthews and Yip 2011: 107), Wu (X. Li 2011: 6, fn 3), Southern Min (Tang 2007: 980), and Nuosu Yi, which I will discuss in Chapter 4 (Section 4.6).

The demonstrative-noun phrase [Dem NP] in Mandarin differs from genuine definites semantically. Recall the test from Löbner (1985) that differentiates genuine definite determiner from demonstratives, as we saw in Section 2.4.1:

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

In (175a), the genuine definites, i.e., the noun phrases with a definite determiner, yield a contradiction when appearing with a predicate and its negation (175a); in contrast, in (175b), the noun phrases with a demonstrative admit a sensible interpretation (175b). Mandarin [Dem NP] phrases behave semantically on a par with [Dem NP] phrases in English as well as those in other languages that freely allow bare nouns as arguments as examined in Dayal (2004). Look at the example in (176a), when appearing with a predicate and its negation, the demonstrative-noun phrases admit a sensible interpretation, whereas in (176) the bare nouns with a definite interpretation yield a contradiction when appearing with a predicate and its negation.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.119) Regarding the syntactic status of demonstratives in Mandarin, I adopted in Section 2.4.1 the view that demonstratives and article determiners do not occupy the same structural position in D that demonstratives occur in specifier positions. Such a view is based on the synatic and sementic differences between demonstratives and article determiners, as argued by many authors (Löbner 1985; Giusti 1997, 2002; Brugé 2000, 2002; Brugé and Giusti 1996; Panagiotidis 2000; Grohmann and Panagiotidis 2004; Shlonsky 2004; Alexiadou et al. 2007, among others) (cf. (111)–(115), Section 2.4.1). Under this view, demonstratives are phrasal and can merge with either higher nominal phrases like numeral classifier phrases (177a) or with lower nominal phrases like bare nouns (177b).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

As for the semantics of demonstratives, I adopt the view that demonstratives are property seeking functions with an indexical specification, as in Kaplan (1989), Wolter (2006), and Dayal (2012) (cf. Section 2.4.1). Under the analysis of bare nouns as kinds in Mandarin, I propose that demonstratives, in addition to being obtained from properties, can also be obtained from kinds via the help of indexically individuated situations. Specifically, I propose that the same strategy to derive definites from kinds via Situation Restriction (164), plus an indexical component, can be exploited to obtain the definite reading for [Dem NP] in Mandarin. For example, if the child-kind c is a function from situations s to the maximal entity that instantiates the child-kind in s, zhe haizi “this child” in (176) can be represented as “haizi(theren),” where theren denotes the (distal) situation the speaker is pointing at (178c).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

That is to say, unlike definite determiners which universally combine with properties and return entities of type <<e,t>, e>, demonstratives can be a function contextually restricted by an indication of the speaker; this function picks a salient individual that the speaker is pointing at, type <<e,t>, e> (179a), or demonstratives can be a function from kinds to individual entities of type <e, e>, restricting kinds to the situation that the speaker is pointing at, type <ek, e> (179b).

(p.120)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

To summerize Section 2.5, I have shown that the other interpretations of Mandarin bare nouns can be derived from their kind interpretation within a Neocarlsonian account of bare nominals, as pursued in Yang (2001) and X. Li (2011, 2013). I proposed an alternative approach to derive the definite interpretation from the kind interpretation via Situation Restriction. I showed that the proposed account for the definite interpretation of Mandarin bare nouns can also account for the demonstrative-noun phrase in Mandarin. As we will see later in Chapter 5, this Neocarlsonian account of bare nominals discussed in this section can be further extended to bare nominals in other classifier languages like Nuosu Yi as well. In the following section, I will discuss the last type of bare nominal arguments in Mandarin, i.e., the bare classifier-noun phrase. I will show that the proposed analysis of bare numeral classifier phrases in Mandarin as shown in Section 2.4 also has important implications for the bare classifier-noun phrase.

2.6 Mandarin is not so bare: one-deletion of numeral classifier phrases

In addition to bare nouns and bare numeral classifier phrases, Mandarin allows another type of nominal arguments which consists only of a classifier and a noun but no numeral [Cl N] (Lü 1944; Chao 1968; Paris 1981; A. Li 1997; Cheng and Sybesma 1999; Borer 2005; J. Huang 2009, 2014; Cheng et al. 2012; Li and Bisang 2012; Zhang 2013; Jiang 2014, 2015; Li and Feng 2015, among others). Numeral-less classifier phrases, which are also referred to as bare classifier phrases (i.e., bare ClPs) in the literature, receive an interpretation similar to [one Cl NP], as illustrated in (180).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

As early as in Lü (1944), it was convincingly argued that the Mandarin [Cl N] expression is the result of deleting the numeral one from the numeral classifier phrase [one Cl N] (see also Chao 1968: 554; A. Li 1997; Borer 2005; J. Huang 2009, 2014; Jiang 2015; Li and Feng 2015). Besides the one-deletion view, there is another view which (p.121) argues against a relationship between [Cl N] and [one Cl N]. The view opposed to one-deletion claims that [Cl N] receives only a nonspecific interpretation, and therefore cannot be treated as the deletion of one in [one Cl N] as the latter can receive specific interpretation in addition to the nonspecific one (Cheng and Sybesma 1999; Cheng et al. 2012; Li and Bisang 2012; Zhang 2013).

In this section, I am going to defend the one-deletion view of the bare classifier phrase in Mandarin and argue that the bare classifier phrase is not really bare in the syntax; instead, it has the structure [one Cl N]. Specifically, I will argue for the following four points: (i) Mandarin [Cl N] can appear in environments that require specific interpretations (Section 2.6.1), (ii) Mandarin [Cl N] behaves like [one Cl N] and indefinites in general in allowing long-distance scope interpretations (Section 2.6.2), (iii) the arguments against the one-deletion analysis of Mandarin bare ClPs (e.g., Cheng and Sybesma 1999; Cheng et al. 2012; Li and Bisang 2012) do not hold (Section 2.6.3), and (iv) the one-deletion analysis not only captures the syntactic and semantic behaviors of Mandarin bare classifier phrases but also logically accounts for why [Cl N] can only be interpreted as singular instead of “two,” “three,” or “some” (Section 2.6.4). Let us start with the interpretation and distribution of Mandarin [Cl N].

2.6.1 Interpretation and distribution of [Cl N]

Regarding the interpretation of numeral-less classifier phrases [Cl N] in Mandarin, it is widely accepted that they usually receive a prominent nonspecific indefinite interpretation (e.g., see Lü 1944; Chao 1968; A. Li 1997; Cheng and Sybesma 1999; P. Chen 2004; J. Huang 2009, 2014; Li and Bisang 2012, among many others), as exemplified in what follows.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Numeral-less classifier phrases, however, can also receive a specific indefinite interpretation, and such a interpretation can emerge easily in constructions that require specific interpretations (e.g., see Lü 1944; P. Chen 2004; J. Huang 2009, 2014; Jiang 2012, 2015). I illustrate this point with two tests for specificity in Mandarin.

The first test for specificity comes from Lü (1944), Sybesma (1992, 1999), and Chen (2004), who illustrate that Mandarin ba-construction requires the nominals which follow ba to be interpreted as either definite or specific. Lü (1944: 161), Chen (2004), and J. Huang (2014) have demonstrated that [Cl N] in Mandarin can appear (p.122) in ba-construction and can be interpreted as specific. I provide two of their examples in (182).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

In addition to the above examples obseved by Lü (1944), Chen (2004), and J. Huang (2014), examples of this sort can be widely seen in daily life as well as in literary work. In what follows I provide some of the examples found in Beijing Language and Culture University Contemporary Chinese Corpus (BCC Corpus in short):

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.123)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

In all of the preceding examples, the [Cl N] phrase appears in the ba-construction and needs to be interpreted as specific rather than nonspecific.

Another test for specificity comes from J. Huang (1987), who showed that Mandarin bare nouns cannot appear in the secondary predication sentences as in (184a), but the numeral classifier phrases [one Cl N] can (184b). This test shows that bare nouns behave differently than numeral indefinites that they do not allow a specific interpretation.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

The [Cl N] expression can appear in the secondary predication sentence which behaves like [one Cl N] rather than bare nouns (Jiang 2012, 2015), as exemplified in (185).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

In (185), the numeral ‘one’ yi can be omitted, and both [one Cl N] and [Cl N] receive a specific indefinite interpretation.

Similarly, examples of this sort can be widely seen in daily life as well as in literary work. I provide some of the examples found in the BCC Corpus here.

(p.124)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

What the preceding examples showed to us is that the numeral-less classifier phrase [Cl N] can appear in the environments that require a specific interpretation and that the numeral-less classifier phrases in these examples have to be interpreted as specific indefinite.

Although the numeral-less classifier phrase [Cl N] and the numeral classifier phrase [one Cl N] share the nonspecific and specific indefinite interpretations, they (p.125) do differ in one respect: the former cannot express the meaning which emphasizes the numeral information of one, as first observed in Lü (1944: 166–167) and further discussed in Li and Bisang (2012). In particular, Lü (1944: 166–167) observes that when the numeral information is addressed in a sentence, yi “one” cannot be omitted. He examines and describes a series of contexts where yi “one” can never be omitted. I generalize and summarize some of them as the following: (i) in sentences with negation meaning similar to “not a single” (187i); (ii) in sentences when “one” is focused (187ii); (iii) in dou-sentences which emphasize “the whole” (187iii); (iv) in sentences with contrastive focus (187iv).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Li and Bisang (2012) provide additional examples to illustrate that if the numeral information in a sentence is important and emphasized, it will prevent [Cl N] from appearing in the sentence without one. Their examples are given in (188).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.126) Next, let us look at the distribution of [Cl N]. As we saw in Section 2.3.1 (cf. (56)), Mandarin [Cl N] phrases only appear in extremely restricted positions. It has been observed in the literature that they are banned in sentence initial position (e.g., Lü 1944; Chao 1968: 554; Cheng and Sybesma 1999; Li and Bisang 2012) (189a), but they are also disallowed quite generally in the second position of a sentence (189b). In order to make both sentences in (189) grammatical, the numeral one has to be used (190).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

In addition, [Cl N] never appears in a topic position or the position after topic but before the verb, as observed in Li and Bisang (2012).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

In post-verbal positions, [Cl N] is allowed and many examples have already been given to that effect (e.g., (181), (182), (185)). However, it cannot freely occur in the post-verbal position, as we have seen in Section 2.3.1. Lü (1944: 170–171) examines and describes a series of contexts where [Cl N] is disallowed in the post-verbal position. One of these contexts, which always holds, can be characterized along the following lines: [Cl N] can never occur in coordination structures after the first conjunct in a listing situation where there are multiple nominals (see also Yang 2001: 69). We have already seen examples of this sort in (56b) (as repeated in (192)); I provide one more example from Yang (2001) in (193).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.127)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Interestingly, [Cl N] seems to be incompatible with certain word preceding it, such as guo, an aspect marker:

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

If we replace the aspect marker guo with the aspect marker le, the numeral-less classifier phrase [Cl N] will be allowed in the above examples in (194).

In this section, we saw that the numeral-less classifier phrase [Cl N] shares the interpretations with the numeral classifier phrase [one Cl N]: [Cl N] can receive a nonspecific indefinite interpretation; it can appear in the environments that require a specific interpretation. However, [Cl N] does differ from [one Cl N] in that it cannot express the meaning which emphasizes the numeral information of one. In terms of its disbribution, [Cl N] is extremely restricted in Mandarin. [Cl N] is banned in the sentence initial position as well as the second position of a sentence. [Cl N] is allowed in the post-verbal position; however, even in the post-verbal position, it is not freely alowed. In particular, in the post-verbal position, [Cl N] is banned in coordination structures after the first conjunct in a listing situation where there are multiple nominals; it is also banned in the position following certain elements such as the aspect guo. The following section discusses the scope behavior of Mandarin [Cl N]; as we will see, Mandarin [Cl N], just like indefinites, can receive a long-distance scope interpertation and exhibit the island escaping ability.

2.6.2 Scope behavior of [Cl N]

Mandarin [Cl N] phrases exhibit the same island escaping ability as indefinites like [one Cl N] and can receive the same long-distance scope interpretation. In (195), yi ge xuesheng “one-Cl-student” and ge xuesheng “Cl-student” are the internal arguments of the predicate in the adjunct clause. Both of them can have a wide scope interpretation escaping from the adjunct if-clause in addition to the narrow scope interpretation.

(p.128)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Under the wide scope reading, (yi) ge xuesheng “(one) Cl student” is interpreted as a specific student, and the speaker will pay for the hearer’s drink if the hearer brings that specific student to the party. In contrast, under the narrow scope reading, (yi) ge xuesheng “one Cl student” is interpreted as nonspecific, and the hearer’s drink will be paid as long as he/she can bring a student (any one) to the party.

Admittedly, the wide scope reading of [Cl N] in (195) can be difficult to get, especially in an out-of-the-blue situation, and the prominent interpretation of the [Cl N] phrase is the narrow scope nonspecific reading. However, what is crucial for us is that it is still possible to get the wide scope reading even if it can be difficult. If we provide contexts/senarios that can help establish the wide scope reading of [Cl N], such a reading will be easier to get. For example, in a scenario in which the speaker has a specific student in mind that he/she would like the hearer to bring to the party, but the speaker does not want to express it explicitly to the hearer and use the sentence in (195b) instead.

As we saw in Section 2.3.2, numeral classifier phrases can take wide scope over a scope-bearing element. Two examples are repeated in (196i). The same as numeral classifier phrases, numeral-less classifier phrases [Cl N] in Mandarin can also take wide scope over the quantifier most or every, receiving a specific indefinite interpretation: a specific kid in (196iia) and a specific swindler in (196iib).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.129)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Note that all the examples in (195) and (196) involve the individual classifier ge, and one may wonder whether the scope behavirors still hold if we consider other classifiers, as pointed out by a reviewer. In (197), different individual classifiers are used in the examples, and the bare classifier phrase can still receive either a specific indefinite reading, scoping over the other scope-bearing element (e.g., “if” or “every”), or a nonspecific indefinite reading, scope under the other scope-bearing element. These examples show that the scope behavior of Mandarin [Cl N] is not restricted to a particular individual classifier but is observed among other individual classifiers.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.130) In addition, bare ClPs in Mandarin can also escape islands without having the widest scope, similar to (numeral) indefinites. In the following examples in (198), [Cl N] can receive an intermediate scope interpretation, behaving like numeral indefinites in (68) and (69) (some examples are repeated in (199)).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.131)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Having seen the interpretation, distribution, and scope behavior of the numeral-less classifier phrases in Mandarin, let us move on to the previous analyses of them.

2.6.3 Previous analyses of Mandarin numeral-less classifier phrases

Two main views have been proposed for Mandarin [Cl N] which are quite opposite in nature. One treats [Cl N] as one-deletion from [one Cl N] (e.g., Lü 1944; Chao 1968; A. Li 1997; Borer 2005; J. Huang 2009, 2014; Jiang 2012, 2015; Li and Feng 2015, among others), the other argues against one-deletion analysis and treats [Cl N] as an independent phrase with no relation to [one Cl N] (e.g., Cheng and Sybesma 1999; Cheng et al. 2012; Li and Bisang 2012; Zhang 2013). The main reason that the (p.132) opponent of the one-deletion analysis claims the one-deletion analysis is wrong is that they think [Cl N] can only be interpreted as nonspecific indefinite in Mandarin while [one Cl N] can be either specific or nonspecific.

In this section, I am going to review Cheng and Sybesma (1999) and Li and Bisang (2012) and show the following points: (i) their arguments against the one-deletion analysis of Mandarin [Cl N] do not hold; (ii) the facts that Mandarin [Cl N] phrases behave like [one Cl N] in their ability to appear in environments that that require specific interpretations and their ability to receive a long-distance scope interpretation (cf. Section 2.6.1 and Section 2.6.2) cannot be explained by the analyses in Cheng and Sybesma (1999) and Li and Bisang (2012), and (iii) the one-deletion analysis logically explains why [Cl N] can only be interpreted as singular rather than “two,” “three,” or “some.”

2.6.3.1 Cheng and Sybesma (1999) against one-deletion: Empty Num Analysis

Cheng and Sybesma (1999) analyze Mandarin [Cl N] and Cantonese [Cl N] in a similar way and claim that both of them cannot be viewed as the phonological deletion of [one Cl N] to simply [Cl N]. Their main reason of believing the one-deletion analysis is wrong is the following:

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Based on the interpretation of Cantonese [Cl N], Cheng and Sybesma (1999) predict that the interpretation of Mandarin [Cl N] is also nonspecific only; in particular they predict that “in contexts where only an indefinite specific interpretation is possible, Mandarin [Cl N] phrases should not be able to surface” (Cheng and Sybesma (1999: 525). They provide two tests to argue that [Cl N] in Mandarin cannot be specific.

Before reviewing their tests, I would like point out some problems with the main reason why they believe the one-deletion view of Mandarin [Cl N] is wrong, i.e., it is illegitimate to make predictions about the interpretation of Mandarin [Cl N] based on that of Cantonese or to assume that Mandarin [Cl N] should be interpreted like Cantonese [Cl N]. Cheng and Sybesma (1999) have shown that Cantonese [Cl NP] and Mandarin [Cl NP] differ not only in their semantic interpretations but also in syntactic distributions. In particular, the Cantonese [Cl N] can receive a definite reading in addition to the indefinite reading and can freely appear in argument positions. On the other hand, Mandarin [Cl N] can never receive a definite interpretation and cannot appear in pre-verbal positions but only in extremely limited positions. Given these differences, there is no good reason to assume the interpretation of [Cl N] in Mandarin is the same as that in Cantonese.

Returning to their test, the first one involves bounded predicates. As argued by Sybesma (1992:176–178), bounded predicates, such as resultative predicates and the ba-construction, force either a definite or a specific interpretation of the object. (p.133) They provide several examples to illustrate that [Cl N] cannot appear with bounded predicates. Coincidentally, they also use the same ba-construction as in Lü 1944, but observe a single case disallowing [Cl N] (202i). Besides the ba-construction, they use resultative predicates to illustrate that they are also incompatible with [Cl N], as given in (201ii). Based on the contrasts they have observed, they conclude that the “[Cl N] phrase must be nonspecific and that there is no phonological reason why ‘one’ could not be suppressed” (Cheng and Sybesma 1999: 525–526).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

There are two main problems with the examples they provide and the conclusion they draw from them.

First, there is no direct logical relation between the unacceptability of [Cl N] in the examples in (201) and the conclusion that [Cl N] must be nonspecific.

Second, the argument Cheng and Sybesma provide faces vast counterexamples observed in the literature and widely seen in literary work. As we saw in Section 2.6.1, as early as in Lü (1944: 160–161), it has been observed that Mandarin [Cl N] can appear in the ba-construction, receiving a specific interpretation (see also P. Chen 2004; J. Huang 2009, 2014).52 I repeat some of their examples in what follows.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.134) In addition to the preceding examples obseved by Lü (1944), Chen (2004), and J. Huang (2014), I provided further examples of this sort widely seen in daily life and literary work in (183) (as repeated in (203)).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.135)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

The preceding examples in (202) and (203) have shown that the numeral-less classifier phrase [Cl N] in Mandarin can appear in the environment that requires a specific interpretation; such a fact has to be acknowledged.53

Their resultative predicate test in (201ii) also encounters counterexamples easily found in daily conversation as well as in newspaper reports. I provide some of the examples found in the BCC Corpus here.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.136)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Turning to their second test, Cheng and Sybesma (1999) also use secondary predication from J. Huang (1987) (cf. (184), Section 2.6.1) and provide two examples where [Cl N] is unacceptable with secondary predication.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

There are two issues with regard to this test. First, examples in which [Cl N] appears with secondary predication are wided seen in daily conversation as well as in literary work, as we have seen in (185) and (186) (as repeated in (206) and (207)).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.137)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Second, the two unacceptable cases in (205) both contain guo, an aspect marker, immediately preceding [Cl N]; such a morpheme seems to reject [Cl NP] regardless whether a sentence contains a secondary predication or not, as illustrated in (208).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.138)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

In contrast, all the grammatical sentences in (206) and (207) contain another aspect marker le which does not cause a problem when it occur with [Cl N] and a secondary predication. The contrast between (205) and (206)/(207) perhaps suggests a phonological condition for one-deletion, as I will discuss in Section 2.6.4.

Although Cheng and Sybesma disagree with the one-deletion analysis of Mandarin [Cl N], they propose a structure in which there is a null numeral projection NumP above ClP for Mandarin numeral-less classifier phrases, as illustrated here.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

According to Cheng and Sybesma, the empty head needs to be lexically governed; thus accounting for the distribution of Mandarin [Cl N]. However, to posit a null numeral for [Cl N] predicts that a numeral, i.e., yi “one,” can be recovered in [Num Cl N], and this indeed echoes the one-deletion analysis of [Cl N] which views that there is an omitted numeral yi “one” in [Cl N] (cf. Li and Bisang 2012: 347).

In the next subsection, I consider arguments and tests from Li and Bisang (2012) and show that their arguments and tests also do not succeed in supporting their claim that [Cl N] in Mandarin can only be nonspecific and that [Cl N] cannot be treated as deleting one from [one Cl N].

2.6.3.2 Li and Bisang (2012) against one-deletion: VP Internal Existential Quantification Analysis

Li and Bisang (2012) use a set of tests to argue for three points:

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Based on their tests they reach the following conclusion: given that indefinite [Cl N] is only nonspecific, but [one Cl N] is three-way ambiguous, [Cl N] cannot be the result of one-deletion from [one Cl N].

The last two points in (210b/c) that Li and Bisang (2012) argued have also been illustrated in Lü (1944), and we have seen some of the examples in (187) (cf. Section (p.139) 2.6.1), so no review of that is necessary. On the other hand, their first point in (210a) and the conclusion that they reached are flawed, so this section reviews arguments and tests Li and Bisang have provided for (210a) and the reasoning behind their conclusion.

Li and Bisang (2012: 345–346) use three tests to support their claim that [Cl N] can only be nonspecific: (i) modification by a relative clause, (ii) creation verbs, and (iii) modification by a secondary predicate.54 Their third test is the same as Cheng and Sybesma’s (1999) second test in (205), as we just saw in Section 2.6.3.1, and I have shown many examples in which [Cl N] is compatable with a secondary predicate in (206) and (207), so this test will be excluded from the current discussion.

Let us look at their first test; Li and Bisang (2012) use the fact that bare ClPs cannot serve as the head of a relative clause (RC in short) to argue that they do not have a specific interpretation. Zhang (2006) has argued that when the RC precedes and modifies the numeral classifier phrase, the whole [RC Num Cl N] phrase is only specific; she claims that the different word order of the RC and the [NumP-Cl] with regard to the head noun corresponds to different interpretations. Specifically, according to Zhang, the [RC Num Cl N] phrase only receives a specific interpretation, but the [Num Cl RC N] phrase can either receive a specific or a nonspecific interpretation. Li and Bisang (2012) adopt Zhang’s claim as a criterion. They show that when [one Cl N] is modified by a RC, the modified phrase [RC [one Cl N]] is grammatical and receives a specific interpretation only (211a), but [Cl N] cannot be modified by the RC—the sentence containing [RC [Cl N]] is ungrammatical (211b).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Li and Bisang (2012: 346) use the ungrammaticality of (211b) to illustrate that the [Cl N] in Mandarin does not express specificity.

Li and Bisang contrast (211) with (212). In (212a), the [one [Cl [RC N]]] phrase can receive either specific or nonspecific; in (212b), the sentence containing [Cl [RC N]] is acceptable. According to Li and Bisang (2012: 346), the reason why (212b) is acceptable is because “the noun in [Cl N] is modified by the RC so that the whole phrase has a nonspecific reading.”

(p.140)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Li and Bisang’s first test and their reasoning are problematic for three main reasons.

First, the acceptablity of [Cl [RC N]] in (212b) does not support their claim that [Cl N] can only be nonspecific and never be specific. In particular, [Cl [RC N]] can appear in environments that force a specific indefinite interpretation, such as ba-sentences (e.g., Lü 1944; Sybesma 1992, 1999; P. Chen 2004) and sentences with resulative predicates (see Sybesma 1992: 176–178). Some of the examples found in the BCC Corpus are given in what follows.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.141)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

The preceding examples in (213) and (214) illustrate that the [Cl [RC N]] phrase can receive a specific indefinite interpretation, contrary to Li and Bisang’s claim that such a phrase only receives a nonspecific interpretation.

Second, when we place a scope-bearing element above [Cl [RC N]], [Cl [RC N]] behaves like indefinites and can receive not only a narrow scope interpretation but also a wide scope interpretation (215).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

When [Cl [RC N]] takes a wide scope in (215i), it receives a specific interpretation, namely “a specific foreigner who wanted to learn Chinese.”

Third, the unacceptability of [RC [Cl N]] in the sentence in (211b) does not support Li and Bisang’s conclusion that [Cl N] does not express specificity. The ungrammaticality of (211b) could arise for various reasons. For example, if the head nouns in (211) are raised from within the relative clause, we would expect that they can be reconstructed back into the relative clause; however reconstructing [Cl N] back into the RC is unacceptable (216b), which contrasts with (216a). This suggests that the (p.142) sentence in (211b) is ungrammatical to begin with and that its ungrammaticality is not related to the interpretation of [Cl N].

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Turning to Li and Bisang’s second test; they use creation verbs to show that these verbs accept both [Cl N] and [one Cl N] and that these two phrases only receive a nonspecific interpretation with these verbs (217).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

The preceding two sentences in (217) can only illustrate the point that Mandarin [Cl N], the same as [one Cl N], can receive a nonspecific interpretation; however it cannot support the claim that Mandarin [Cl N] can only receive a nonspecific interpretation.

Li and Bisang further show that [Cl N] and [one Cl N] differ in whether they can express addressed numeral information; I showed in Section 2.6.1 that Lü (1944), who argues for the one-deletion view of [Cl N], also reaches such a conclusion, and examples have been provided to illustrate this point in (187). Importantly, such a difference between [Cl N] and [one Cl N] alone does not present an argument against the one-deletion analysis of [Cl N].

Regarding the analysis of Mandarin [Cl N], Li and Bisang propose that the classifier phrase ClP is the maximal projection of the indefinite [Cl N] with no higher projection above it, as shown in (218).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

They propose that indefinite [Cl N] denotes a set of atomic entities, i.e., a set of singularities based on the assumption that bare nouns denote kinds (e.g., Chierchia 1998b). They argue that classifiers in [Cl N] apply to a kind term in the denotation of NP and return a subset of atomic individuals, i.e., at type of <k, <e, t>>. Their analysis of the semantics of individual classifiers is given in (219a), with the bare CLP ben shu “Cl book” analyzed in the (218c) (INST stands for the instantiation operation).

(p.143)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Based on the semantics of [Cl N] in (219c), the bare ClP is interpreted as a predicative nominal phrase of type <e, t>. Li and Bisang assume that indefinite [Cl N] introduces variables that have to be bound by the Heim’s (1982) type of existential quantifier. Consequently, if [Cl N] falls into a post-verbal position, the existential quantifier comes in by default existential quantification over the VP (see Diesing 1992), thus accounting for the indefinite interpretation of [Cl N] in post-verbal positions.

Li and Bisang’s analysis of bare ClPs in Mandarin is similar to Tsai’s (1999, 2001) Extended Mapping Hypothesis for numeral classifier phrases in Mandarin (see Section 2.3.4). However, there are two main problems with their analysis. First, Li and Bisang’s analysis expects [Cl N] to be acceptable in post-verbal position; however, as we have seen in Section 2.6.1, [Cl N] cannot freely occur in post-verbal position in Mandarin. In particular, it cannot occur in coordination structures after the first conjunct in a listing situation where there are multiple nominals (e.g., Lü 1944; Yang 2001), as we saw in (192) and (193) (as repeated in (220) and (221). [Cl N] is also incompatible with certain word preceding it, such as the aspect guo, as we saw in (193) (as repeated in (222).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.144) Second, Li and Bisang’s analysis based on the local existential quantifier cannot account for the long-distance scope behaviors of Mandarin [Cl N] as we saw in Section 2.6.2. Some examples are repeated here.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

To summarize Section 2.6.3, I reviewed the arguments and tests from Cheng and Sybesma (1999) and Li and Bisang (2012) against the one-deletion approach to Mandarin [Cl N] and showed that their arguments and tests do not support their claim that [Cl N] in Mandarin only receives a nonspecific interpretation nor the claim that [Cl N] cannot be treated as deleting one from [one Cl N]. In particular, we saw many exmaples from previous literature and corpora which demonstrate that bare classifier phrases, [Cl N] as well as [Cl [RC N]], can appear in envorinments that require a specific indefinite interpretation, such as ba-construction, sentences with resultative predicates, and sentences with secondary predicates.

In the following section, I will show that the one-deletion analysis of Mandarin bare classifier phrases, as first argued in Lü (1944), not only can capture their syntactic and semantic properties that we have seen, but also it can logically account for why [Cl N] in Mandarin can only be interpreted as singular, instead of “two,” “three,” or “some.”

2.6.4 One-deletion analysis of Mandarin [Cl N]

The previous sections showed us that Mandarin numeral-less classifier phrases [Cl N] and the numeral classifier phrase [one Cl N] share significant similarities regarding (p.145) their interpretations and scope behaviors. These similarities between [Cl N] and [one Cl N] lead us to the one-deletion analysis of Mandarin bare classifier phrases as first proposed in Lü (1944) and further elaborated by many others (e.g., Chao 1968; A. Li 1997; Borer 2005; J. Huang 2009, 2014; Jiang 2014, 2015; Li and Feng 2015). Consequently, bare ClPs in Mandarin are not really bare in the syntax; instead they have the structure of the numeral classifier phrase [one Cl N], and it is the process of eliding the numeral one at PF that leads to the surface bare form of [Cl N], as illistrated in the following.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Let us first look how the one-deletion analysis accounts for the similarities between [one Cl N] and the [Cl N] in Mandarin.

In Section 2.4, I showed that the proposed lexical view of numerals allows us to derive a general quantifier (GQ) variant of numeral classifier phrases. The quantificational force could be gained via a global existential closure (Heim 1982), a covert existential quantifier (Link 1983, 1987, Krifka 1999), a type-shifting principle (Partee 1986, Landman 2003), or simply via the proposed lexical analysis of choice function. Based on the analysis of numeral classifier phrases developed in Section 2.4, a possible theory of specificity that captures [one Cl N] and its relation with the bare ClP can be the following one.

We can view that it is the GQ reading that leads to the nonspecific interpretation of [one Cl N] and that it is the choice function in the lexical entry of the numeral yi “one” that contributes to its specific/weak presuppositional interpretation, allowing long-distance scope interpretations. Accordingly, the nonspecific and the specific interpretation of the numeral classifier phrase [one Cl N] can be characterized in the way illustrated in (225).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Concerning the interpretation of bare ClPs in Mandarin, they uncontroversially receive the nonspecific interpretation, i.e., the GQ reading in (225a), and the interpretational issue of bare ClPs is whether the specific reading in (225b) is available after eliding one. Previous sections show that Mandarin bare ClPs do allow the speific interpretation, so bare ClPs is the same as [one Cl N] in terms of specificity. (p.146) Accordingly, we can view that it is the elided one that contributes to the nonspecific reading as well as the specific reading of bare ClPs, as shown in (226).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Next, let us consider the differences between the bare ClP and [one Cl N]. As shown in Section 2.6.1, the difference between the bare ClP and [one Cl N] has to do with their distribution ((189)–(194), as summarized in (227)) as well as whether they can be used in a clause where the numeral “one” information is important and stressed (187)/(188), as summarized in (228).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.147)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

To account for the distributional difference between bare ClPs and [one Cl N], two possible answers can be provided. One is that the restricted distribution of bare ClPs is the result of syntactic restrictions, e.g., government, as proposed in Cheng and Sybesma (1999) and J. Huang (2009). The other is that the distributional restriction is due to phonological/prosodic restrictions, as proposed in Lü (1944) and Li and Feng (2015). In the rest of this section, I am going to briefly illustrate how the restricted distribution of bare ClPs that we have seen can be accounted for via a phonological account (I refer the reader to Li and Feng (2015) for a more detailed discussion of the phonogical account for Mandarin bare ClPs).

Phonetically, Mandarin “one” yi is a syllable only consisting of a high front vowel /i/ without onset or coda. If /i/ is not stressed, it could easily be omitted in a fast speech when a stressed word precedes it. In other words, the phonetic characteristics of yi “one” in Mandarin yields the optional deletion.

(p.148) Lü (1944: 174) describes a necessary phonological condition for one-deletion: deleting yi “one” in Mandarin is possible when yi (/i/) is unstressed (qinyin hua “lightened/unstressed”) and preceded by a stressed word (zhongyin “heavy/stressed syllable”). I adopt Lü’s condition and modify it slightly: deleting yi “one” in Mandarin is possible when yi (/i/) is unstressed and when there is a stressed word closely preceding it. In addition to this condition, I conjecture that the motivation for deleting one is to facilitate efficiency in speech, especially, in fast colloquial speech. With the conditions and the motivation for one-deletion in mind, let me illustrate how they account for the restricted distribution of Mandarin [Cl N].

In a sentence, verbs usually bear main stress; when [one Cl N] follows a verb forming a constituent with it, a possible environment is provided to delete one. This is exactly the case where [Cl N] occurs (e.g., (227i)). However, in the post-verbal position, if [one Cl N] is separate from the stressed word by pause (i.e., in listing situation), the condition for deleting one is not met. This accounts for why [Cl N] is disallowed in coordination structures after the first conjunct in a listing situation (e.g., (227ii)). In sentence initial position/pre-verbal position, there is no word immediately preceding yi “one,” the condition for one-deletion is not satisfied either; thus accounting for why [Cl N] is unacceptable in (227iii). In cases where all the conditions for deleting one are met, if deleting one does not facilitate computational efficiency (i.e., makes speech efficient) but rather creates more difficulties in speech (i.e., causes pronunciation difficulties or violates phonological rules), one-deletion should not be expected to occur. This could be the reason why [Cl N] are unacceptable in sentences like the one in (227iv).

The phonological account for the distribution restrictions of bare ClPs can makes a prediction about one-deletion in Mandarin. If a stressed word, not even a verbal element, immediately precedes [one Cl N], we should expect the deletion of “one” can happen. This prediction is borne out. In (229), when [one Cl N] is followed by demonstratives or universal quantifiers which normally bear stress, one can be deleted. The examples that follow show that the deletion rule is prosodically conditioned.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

55

As for the semantic difference between bare ClPs [Cl N] and [one Cl N] in Mandarin, it can be understood by placing a semantic condition for eliding one: one can be elided from [one Cl N] to derive the bare ClP when the numeral information is not stressed/ (p.149) important. Accordingly, in the environments in which the numeral information is stressed/focused as we saw in (228), we should not expect the eliding of one to occur.

As we can see, the one-deletion analysis can derive the interpretations, distribution, and scope behaviors of Mandarin bare ClPs that we saw in the previous sections. More importantly, the earlier one-deletion analysis also logically explains why Mandarin [Cl N] can only be interpreted as singular rather than “two,” “three,” or “some”: since it is yi “one” that is deleted from [one Cl N], the [Cl N] phrase can only be interpreted as singular one. In what follows I summarize the one-deletion analysis of Mandarin bare classifier phrases.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

56

2.7 Interpretational restrictions on sentence initial nominals

This section discusses the interpretational tendency of bare nouns in the sentence initial position and the related phenomenon regarding numeral classifier phrases, as we saw in Section 2.3.3 and Section 2.5.2. The goal is to provide a uniform account for this property of nominal arguments unique to Mandarin.

(p.150) 2.7.1 The interpretational tendency

As we saw in Section 2.5.2, in very early work, such as as Chao (1968) and Li and Thompson (1981), it has been observed that in subject positions, Mandarin bare nouns have a strong tendency to be interpreted as definite. In order to receive an existential interpretation, a sentence initial verb you “exist/have” is needed. Two examples are repeated in (231).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

In later work, however, it has been observed that various examples do not following this tendency (e.g., J. Huang (1997); Yang (2001)). In particular, Yang (2001) observes that the existential reading of bare nouns become immediately available on pre-verbal bare nouns, once they occur in the context of leftperipheral locatives (as seen in (148c) and repeated in (232)).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

The behavior of Mandarin bare nouns is analogous to that of numeral classifier phrases in sentence initial position that we saw in Section 2.3.3, namely that without contexts provided it is unnatural for Mandarin indefinite numeral classifier phrases to appear in the subject position. In order to make such sentences natural, a sentence initial you “exist/have” can be added or a (perceivable) situation/context can be provided. Some examples are repeated here.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.151) Two main puzzles arise from the preceding examples. The first puzzle is related to how we should understand the interpretational tendency for nominals in the subject position in Mandarin (231)/(233a); the second is about how we should understand the examples that do not follow this tendency (e.g., (232)/(233c)).

Yang (2001) re-analyzes the tendency for nominal arguments in sentence initial positions to be interpreted as definites rather than indefinites in Mandarin with a typological account based on Li and Thompson’s (1976, 1981) categorization of Chinese as a topic-prominent language. I adopt Yang’s account for the pre-verbal nominal arguments in Mandarin and provide some more details for such an account in the following section.

2.7.2 Towards an account

Li and Thompson (1976) in their cross-linguistic work argue that languages can be typologically categorized as topic-prominence languages and subject-prominent languages (234). They argue that Mandarin is a topic-prominent language (see also Chao 1968; Xu and Liu 1997 for a similar view for Mandarin).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

A key argument from Li and Thompson concerns the double subject phenomenon, which is only attested in topic-prominent languages:

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.152) Sentences of the sort in (235) have been referred to as “topic-comment” structures (e.g., Li and Thompson 1976, 1981). Topics in these structures (e.g., “that tree” and “fish”) do not have a selectional relationship with the predicate and are unlikely to derive from movement out of the comment. As for the comment, it is a full-fledged sentence containing a subject (e.g., “leaves” and “red-snapper”). This type of sentences is not attested in subject-prominent languages. I adopt Li and Thompson’s (1976) view and treat Mandarin as a topic-prominent language.

Concerning what a topic is, it could be simply understood as that what is being talked about (i.e., “aboutness”) (Li and Thompson 1976; Chafe 1987; Lambrecht 1994, cf. Li and Bisang 2012). With regard to what kind of nominals can serve as licit topics, I adopt the traditional view in Lambrecht (1994: 262): “a topic constituent must have a referent, and this referent must be identifiable and have a certain degree of pragmatic salience in discourse.”57 Lambrecht (1994: 77–78) further elaborates on the idea of what an identifiable referent: “(it) is one for which a shared representation already exists in the speaker’s and the hearer’s mind at the time of utterance, while an unidentifiable referent is one for which a representation exists only in the speaker’s mind.” Nominals’ references, according to Lambrecht, are arranged in a scale based on their acceptability as topic, as given in (236). (237) provides a definition of each point on the scale.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(p.153) Nominals with the reference in the first three scales “active,” “accessible,” and “ unused” are the most likely to function as a topic according to Lambrecht (1994: 262). With this understanding of topics and the assumption that Mandarin is a topic-prominent language, let us look at the sentence initial nominals in Mandarin. For the sentence initial nominal XP, theoretically it can have one of the following syntactic representations.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

In (238a), the sentence initial XP is a base-generated topic, either with or without a second nominal (YP) in the preverbal position; in (238b) the sentence initial XP is a topic moved from within the IP domain; the sentence initial XP in (238c) is not in the topic position but simply in the pre-verbal position.

By assuming that Mandarin is a topic-prominent language, we can exclude the possibility in (238c), so the sentence initial XP in Mandarin is either moved to the topic position from within the IP domain (238b) or it is simply base-generated as a topic (238a). Furthermore, if the XP within the IP is blocked from moving to the Spec TopP position (i.e., via overt existential operator you “exist/have” or if the Spec TopP is filled), we should expect the XP to stay within the IP domain, as illustrated in (239).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

For the semantics of an XP in topic position as in (239), it should satisfy the identifiability requirement in (236), namely, its reference should be either “active,” “accessible” or “unused.”

The next section outlines the consequences of this assumption about sentence initial XP in Mandarin.

2.7.3 Accounting for the interpretational tendency

I will start with numeral classifier phrases in the sentence initial position first. According to the analysis outlined in (238) and (239), the sentence initial numeral classifier phrases should always be in the Spec TopP position, either as the result of movement from within the IP domain or as the result of base-generation. However, if it is interpreted as nonspecific, which refers to any plural/singular individual not (p.154) identifiable to the hearer (e.g., the sentence is uttered in an out-of-the-blue situation), we would expect this sentence to be unacceptable because it doesn’t have the characteristics of “topics” and fails to satisfy the requirement of a topic as in (236). This prediction is borne out, capturing (233a), as demonstrated in (240).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

If a nonspecific numeral classifier phrase can stay within the IP domain as the result of being blocked from moving to the topic position, we should predict that such a sentence is acceptable. This again is true: in (236b), when the sentence has a sentence initial you serving as an existential operator, it prevents the numeral classifier phrase san ge xuesheng “three student” from moving to the TopP domain, as illustrated in (241).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

If the numeral classifier phrase is identifiable to the speaker, we should expect it can appear in the sentence initial position. This is exactly what happens in (233c), namely san ge haizi “three kids” refers to three specific individual identifiable to the speaker. The possible structure for this example is given in (242).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Similarly, bare nouns in the sentence initial position, under the topic-prominent analysis of Mandarin, should always stay in the Spec TopP position. Given that Mandarin bare nouns can be interpreted as definite, we should expect that bare nouns in sentence initial positions should receive a default definite interpretation. This predication is also borne out and correctly captures the interpretational tendency in Mandarin (231a) as illustrated in (243).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Likewise, if the bare noun has to stay within the IP domain because of some element that blocks it from moving into the topic domain, we should expect that the bare noun can easily receive an existential interpretation besides the definite reading. This again is confirmed by (231a) and (232) and illustrated in (244) and (245).

(p.155)

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

In (244), the bare noun ke “guest” cannot moves to the sentence initial position as the movement is blocked by the sentence initial verb you “exist.” In (245), the movement is blocked by a locative element waimain “outside” (see also Yang 2001 for the same analysis for the sentence in (245)).

Furthermore, the preceding analysis predicts that if the referent represented by a nominal is active (see (236)/(237), namely, a person focusing at a particular moment (e.g., via hearing or seeing), the nominal should still appear in the sentence initial position and serve as topic. This predication is also true, as supported by (147b, c) in Section 2.5.2 (as repeated in (246)).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

In these two examples, the speaker is the witness of the event either through seeing or hearing, the bare noun’s referent is active which ranks it highest in the scale of Topic Acceptability (236).

As we have seen, the interpretational restriction on sentence initial nominals, under the typological view of Mandarin, receives a uniform account. Furthermore, this analysis has an important implication, namely, in Mandarin both definite and specific nominals can serve as topics appearing in the sentence initial position (either through movement or base-generation).

2.8 Summary

This chapter investigated the internal domain as well as the external syntax and semantics of bare numeral containing phrases in Mandarin and compared them with those in number marking languages. It was argued that the syntax and the semantics of bare numeral containing phrases are universal. I showed that although numeral classifier phrases in Mandarin and numeral-noun phrases in number marking languages differ a great deal in their internal nominal structure, they share strong (p.156) similarities at the clausal level concerning their long-distance scope behavior, semantic interpretations and syntactic distributions. Specifically, bare numeral containing phrases are systematically ambiguous (in the Universal Lexicon) between a predicate and an indefinite variant: (i) in their indefinite incarnation, they are arguments with long-distance scope properties; (ii) in their predicate incarnation, they act as restrictors of determiners, demonstratives, quantifiers, and the generic operator. We saw that these two properties are stable, regardless of whether a language has (overt) D or not.

I argued for a D-less analysis of bare numeral containing phrases, in which numerals are phrasal in the syntax and are lexically ambiguous. The proposed D-less analysis captures the remarkable cross-linguistic argumental behavior of bare numeral containing phrases in a straightforward way and also accounts for their long-distance scope ability. I further argued for a kind-referring analysis of Mandarin bare nouns, as proposed in Krifka (1995) and Chierchia (1998b). Regarding language variation in the nominal domain, I argued that it is primarily located in two interrelated factors: what nouns denote (kinds or properties) and what low function heads (i.e., Div and Cl) denote (a function from properties to properties or a function from kinds to properties).

The proposed analysis of numeral classifier phrases correctly predicts the scope behavior of bare nouns in Mandarin: as kind-denoting, they exhibit the narrowest scope ability like English bare nominals rather than English indefinites. The other interpretations of Mandarin bare nouns, as I showed, can be derived from their kind references. Mandarin bare nouns, therefore, also do not need to project a DP in syntax. We reached the conclusion that it is not necessary to stipulate an empty functional category D in Mandarin in order to account for the syntactic and semantic properties of nominal arguments in this language. This is arguably a simpler analysis of Mandarin nominal arguments since it avoids stipulating the presence of invisible projections that otherwise have no overt manifestation in this language. Thus, Mandarin is argued to be a classifier language without D.

The proposed analysis of Mandarin numeral classifier phrases has also allowed us to examine and account for the numeral-less classifier phrases [Cl N] in Mandarin, which have rather restricted distributions but also exhibit the same long-distance scope behavior of numeral classifier phrases. I argued that Mandarin [Cl N] is the result of phonologically deleting one from [one Cl N] and has the full structure of the numeral classifier phrase, by adopting the one-deletion view as first argued in Lü (1944). In addition, the account developed in this chapter has also helped account for the interpretational restriction on nominal arguments in the sentence initial position in Mandarin.

We shall see that the account proposed in this chapter has further implications for other types of nominal arguments in Mandarin, both with and without classifiers. I elaborate on the details in Chapter 3.

Notes:

(1.) “Number marking languages” in this work refers to standard number marking languages such as Romance and Germanic languages where number morphology is obligatory when a numeral (except for “one”) combines with any count noun. Number marking languages in this book do not include languages like Turkish, Hungarian or Western Armenian, which have number morphology on bare nouns without the occurrence of numerals but ban number morphology or optionally allow it on nouns when they appear with a numeral. These languages will be referred to as non-canonical number marking languages. The variation between number marking languages and non-canonical number marking languages in the literature has been attributed to differences with respect to syntactic agreement (Ritter 1991; Frank 1994), or a different semantics for their numerals (Bale et al. 2011), or the different semantics of the functional head determines number morphology on nominals (Sauerland 2003; Sauerland et al. 2005; Scontras 2011, 2014). I leave these non-canonical number marking languages for future research.

(2.) In this book, I restrict my discussion to bare numeral-noun phrases and do not concern about modified numeral-noun phrases, such as exactly two books, about four books, or other amount expressions, like many books and few books. Modified numeral-noun phrases have differrent scope behaviors from those of bare ones discussed in book, i.e., the former lack the long-distance scope behavior (see Kamp and Reyle 1993; Beghelli 1995; Beghelli and Stowell 1997; Szabolcsi 1997, 2010; Winter 2001, 2005, a.o.).

(3.) Concerning what instantiates the functional category “D” (e.g., determiners and/or demonstratives), this book holds the following view. In a broad sense, demonstratives (e.g., English this/that) would fall under the D-family, together with determiners, such as English the/a. But in a narrow sense, the functional category D includes determiners but not demonstratives. That is to say, demonstratives like that/this in English and zhe/na “this/that” in Mandarin would not fall under the category D in the narrow sense. The motivation for “the category D in the narrow sense” is based on the differences between determiners and demonstratives with respect to their semantics and structural positions, which I will address in Section 2.4.1. In this book, I will only focus on D in the narrow sense, i.e., I view that the functional category D includes article determiners but not demonstratives.

(4.) Note that, these properties of numeral-noun phrases are not limited to number marking languages, we also observe them in identical ways in numeral classifier phrases of classifier languages like Mandarin and Nuosu Yi, to be discussed in Section 2.3 and Chapter 4, respectively.

(5.) The details of the numeral semantics are negotiable since the core differences among various analyses do not significantly change what I will propose.

(6.) Note that in Ionin and Matushansky (2006: 340–341) complex cardinals involving addition are treated either as right-node raising or NP deletion, so under their account the coordination is between two numeral phrases containing two NPs, rather than between two cardinals. However, they treat complex cardinals involving multiplication as phrases which occupy the specifier position of Num (±plural) (Ionin and Matushansky 2006: 328).

(7.) Further discusions on complex numerals, such as one hundred and sixty, and phrases containing complex numerals will be provided in Chapter 3, Section 3.5.4.

(8.) Although the structure in (30) accomadates languages like Englih, it is challenged by languages which have one more type of plurals in addition to the canonical plural morphology and allow the occurrence of double plurals, such as Breton (Acquaviva 2008: 206) and Amharic (Kramer 2009, 2010, 2016). Discussions will be provided to address such issues in Chapter 3 (Section 3.5), where plurlity in Mandarin is addressed.

(9.) More accurately, in Borer (2005), the lower functional projection that hosts plural inflection and overt clasifiers is the classifier phrase (CLmax), the head of which is the open value <e>DIV (div. for ‘division’). I will provide more details of Borer’s (2005) analysis in the following subsection.

(10.) Note that in languages like English, the derivation of numeral-noun phrases can be more complex than simply merging a numeral with a noun, as shown in Kayne (2016), and silent categories have been proposed, leading to more refined structures of numeral-noun phrases. Discussions on such refined structures are beyond the scope of this book, and I refer the readers to Kayne (2016).

(11.) Note that, the range assignor R(F) may merge with some higher open value, assigning range to it.

(12.) Regarding Language variation, according to Borer, it mainly lies in the arrange assigners and how the open value of each head is assigned. For instance, in languages with overt Ds like English, the definite determiner can assign range to the head of DP, #max (and possibly to Clmax as well) (e.g., Borer 2005: 164). In some ClLs like Cantonese, classifiers can also assign range to D via movement from within the ClMAX domain to the DP domain (e.g., Borer 2005: 186). Hence, although on the surface we observe a determiner-noun phrase (ia) and a classifier-noun phrase (ib), they are the same in the syntax as a DP (ii).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

I will address issues on language variation in Part III of this book.

(13.) A potential question may arise here, namely, why the source of choice functions cannot be a covert semantic operation like type-shifting. This semantic assumption is theoretically plausible. However if we allow such a covert semantic operation for NCs in French, say, one would like to know why it is not available for bare nouns. Admittedly, one can always make some assumptions to explain this puzzle; however, we believe that adding further assumptions would inevitably result in a more stipulative theory than the proposed lexical analysis.

(14.) This reference to situations is similar to Carlson’s 1977 references to states.

(15.) I am aware of some assumptions in the literature that AspPs are structurally lower than vP (such as in Mergerdoomian 2008: 90 and Tenny 2000: 326); however, it still remains unknown as to whether they are well justified theoretically and empirically. Accordingly, we will simply adopt the traditional analysis of AspP and assume it is structurally higher than vP.

(16.) Note that the numeral-noun phrases “three boys” in (51) can either remain as a predicate of type <e, t> or can be shifted through the choice function f to an argument. The differences lie in whether the operator Gen binds the predicative “three boys” or quantifies over the nonspecific “three boys” and over situations. The semantics of these two cases are both possible and are compatible with the analysis developed here.

(17.) As noted in fn. 1, “number marking languages” in this work refers to standard number marking languages such as Romance and Germanic languages where number morphology is obligatory when a numeral (except for “one”) combines with any count noun. Number marking languages here do not include languages like Turkish, Hungarian, or Western Armenian, which have number morphology on bare nouns without the occurrence of numerals but ban number morphology or optionally allow it on nouns when they appear with a numeral.

(18.) A number of authors establish a connection between the number morphology and the classifier (Greenberg 1972; Sanches and Slobin 1973; T’sou 1976; Doetjes 1997; Chierchia 1998b; Cheng and Sybesma 1999; Borer 2005, among others). I will address this issue in Section 2.4.1.

(19.) In other number marking languages like Russian, bare plural Ns can also freely occur in argument positions, but they cannot do so in some other number marking languages like French (cf. (15), (21)).

(20.) While bare nouns without number morphology and with numerals cannot occur as bare arguments in English as well as in English, they can do so in Russian, as we saw in (21). Discussions on variation among number marking languages will be provided in chapter 7.

(21.) Two examples that demonstrate this observation are given below:

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

(22.) Aoun and Li (1989: 142) mark a canonical active sentence with a universal quantifier in subject subject position and a numeral classifier phrase in object position as umambiguous (i); however, A. Li (2014: 242, fn. 29) notes that when the context is clear, the numeral classifier phrase can refer to a specific individual, receiving a wide scope interprettion (ii).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

Aoun and Li (1989: section 5.3) further note a contrast between dative and double object constructions with regard to the interpretation of numeral indefinites (see also S. Huang 1996; A. Li 2014).

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

The scope freezing effect of numeral indefinites in double object constructions (iiib) is not unique to Mandarin but also attested in English as well as other languages (e.g., see Larson 1988, 1990, Aoun and Li 1993a; Marantz 1993; Bruening 2001, 2010; Antonyuk 2015).

(23.) Note that “generic sentences” as describing (43f) has a much narrower meaning compared to the term in Krifka et al. 1995; their use of the term covers all the constructions in (43) including (43f) which reports the biological characteristics of a creature.

(24.) Note that numerals are in the Spec position of NumPin A. Li’s (1998, 1999) work but in its head position in her (1997) work.

(25.) Regarding the syntactic distribution and interpretations of bare plurals and mass nouns (i.e., existential and generic) in English, Longobardi (1994 sub seq) assumes that those bare plurals/mass nouns project a null D with an existential operator as well. But unlike Italian, the null D in English is assumed not to be subject to licensing constraint in the syntax; therefore, English bare plurals can appear freely in any argument position. The generic reading and the existential reading of bare plurals/mass nouns in English are further explained by Diesing’s (1992) Mapping Hypothesis: the generic reading will emerge if bare plurals map into the restrictor in generic sentences; when bare plurals map into the nuclear scope in episodic sentences, it will lead to an existential reading. I refer the readers to Dayal (2011b) for a detailed review of Longobardi (1994 sub seq).

(26.) A possible change to the original assumptions of the DP hypothesis can be that the empty D contributes to nonspecific indefinite interpretations only but not specific indefinite or generic interpretations in Mandarin; then, the subject-object asymmetry can be explained via some syntactic licensing constraints (e.g., ECP). Nevertheless, by doing so, some other puzzles emerge. To name a few: why do only numeral indefinites with nonspecific interpretations project D? What contributes to the specific and generic interpretations?

(27.) Note that, Longobardi (1994 et seq.) does not discuss numeral noun phrases, so it remains unknown whether or not a DP will project above numeral noun phrases in Italian/English within his framework.

(28.) In Zamparelli’s 2000 work, numerals are always heads in DP (p. 251), either in a higher DP position (his StrongDP (SDP)) or a lower DP position (his Predicate DP (PDP)). Numerals can be heads of PD or SD. When in PD, they must be interpreted as weak cardinality predicates and cannot be interpreted referentially or specifically (p. 239). In contrast, numerals, which are stronger quantifiers or referential elements, are in the SD position. Thus, for argumental numeral expressions, numerals are determiners in Zamparelli’s work, as generalized quantifiers, of type <<e,t>,<e,t>,t>> (see Ionin and Matushansky 2006 for clarification and discussion of problems associated with this treatment of numerals). Furthermore, Zamparelli proposes the following principles: 1. Only SDPs can appear in Argument position (p. 121, ex (322)); 2. All and only non-referential SDPs undergo QR at LF (p. 119, ex (319); 3. A filled SDmax layer triggers presuppositions of existence (p. 121: ex (321). Zamparelli explains the different scope interpretations of numeral phrases together with some semantic assumptions. However, in Liao’s work, determiners and numerals are two separate, independent lexical items in argumental numeral expressions. Accordingly, Liao’s identification of his version of DP/NumP with Zamparelli’s SDP and PredDP are inappropriate. Even for the non-argumental numeral expressions, numerals in their work differ semantically in nature: in Zamparelli’s work they are predicates while in Liao’s they are modifiers of N.

(29.) If D is a head merging with another head (Mod or Asp), we are left with the question of why the agreement between these two heads is established through Spec-head agreement. We should not be able to get a concept of “specifier” here as the merger occurs between two LIs rather than two SOs.

(30.) In addition to the puzzles surrounding the indefinite numeral classifier phrase, Liao’s analysis of generic numeral classifier phrases is also challeged empirically. According to Liao, the generic operator necessarily requires a null Aspect to co-occur with it in Mandarin. If the head of AspP is overt in a generic sentence, it will agree and check the [existential] feature with D in the object position, and this will block the object D from “looking up” to agree and check with the [Gen] feature on Mod. Nevertheless, we do observe examples of generic sentences with overt aspect markers, which demonstrate that generic sentences do not exclude overt aspect in Mandarin (cf. (73)/(74) in Section 2.3.3).

(31.) Liao follows the assumption by Mergerdoomian (2008: 90) and Tenny (2000: 326) that AspP is lower than vP; however, this assumption was proposed without justification in these works as well.

(32.) The goal is to enable us to understand the evolutionary puzzle of human languages (see Chomsky (2005 et seq) for details).

(33.) Counterexamples in which classifiers and number morphology co-occur have been observed (e.g., Sanches and Slobin 1973; Aikhenvald 2000; Zaavala 2000; Wiltschko 2008; De Belder 2008; Gebhardt 2009). I will address this issue in Chapter 3, Section 3.5.

(34.) The locality of this Spec-Head relation between a numeral and a classifier in (103a) is also seen in Cheng and Sybesma (1998: 406), Watanabe (2006; 2010), and Zhang (2013: 213).

(35.) The Num-NP “three boys” here can either remain as a predicate with the Gen binding it (e.g., see Chierchia 1995), or it can be shifted through the choice function f to an argument with a nonspecific interpretation and being quantified over situations (e.g., see Krifka et al. 1995). The semantics of these two cases are both possible and are compatible with our analysis.

(36.) The distinction between count-classifiers and mass-classifiers in Cheng and Sybesma (1999) will be discussed in more details in Section 2.4.2.2.

(37.) Regarding why kinds are entities rather than sets, I refer the readers to Carlson (1977b: 159–173).

(38.) I superscript “ek” to indicate that reference is to kind level individuals and use “e” for reference to object level individuals.

(39.) Atoms can be simply defined in the traditional way: as definite singular count nouns (e.g., the desk, the concept) which are relevant only to the plural domain but not to the mass domain (e.g., Link 1983). Or, assuming that the mass domain and the plural domain both contain atoms and only differ in whether they contain minimal stable atoms (i.e., in the mass domain (e.g., water), one cannot define the minimal stable atoms; however, atoms can be well defined in the plural domain (e.g., apples)) (e.g., see Chierchia 2010), the atoms checked by individual classifiers can be defined as the minimal stable ones, and these atoms still populated the denotations of definite singular count nouns, as in the traditional view.

(40.) I simplify the semantics of each type classifiers, as the details would take us too far afield. I refer the readers to Dayal (2004) for a discussion of “sub-kinds,” Link (1983) for “groups” and Ionin and Matushansky (2006) for “partition.”

(41.) Note that, to account for the complex facts regarding nominal phrases containing classifiers in Mandarin, there is growing consensus that a non-unified analysis of numeral classifiers is needed. In such analyses, both left and right branching structures are possible (e.g., X. Li 2011, 2013; Jiang 2009; Jin 2013, 2018; Zhang 2013; Jiang et al. 2020):

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

While the above authors agree that both structures are needed, there are different proposals pertaining to the details of these structures and their distribution. I refer the readers to Zhang (2013) and Jiang et al. (2020) for a review of different proposals.

(42.) Regarding how the other interpretations of bare nouns in Mandarin come about, I will discuss it in Section 2.5.

(43.) If not considering the view that definite expressions do not enter scope relations, there is another view by Reinhart (1997) and Ruys (1992) which treats definites to take the maximal/widest sentential scope. Yang (2001) indeed treats definites in this way. However, whether definites enter scope relations is not quite relevant for the purpose of discussion here.

(44.) BCC Corpus is one of the major Chinese corpora in Mainland China with diverse writing genres (newspaper, literature, Weibo “microblogs,” etc.), with a total number of characters around 15 billion (see Gou et al. 2016).

(45.) I refer the readers to Dayal and Jiang (submitted) for more data on anaphoric bare nouns in Mandarin and further discussion on this issue.

(46.) Note that bare nouns in Mandarin can appear in predicate position in addition to serving as arguments. Such a predicative use of bare nouns in classifier languages can be obtained via type-shifting mechanisms of the kind discussed in Partee (1986) and Chierchia (1982, 1984, 1998b).

(47.) For simplicity, I avoid reference to the world argument. For instance, I treat , ι‎ and ∃ as functions of type <<e,t>, e/<<e,t>,t>> rather than <<s, <e,t>>, e/<<e,t>, t>>.

(48.) Such readings are further discussed in Dayal (2013).

(49.) As suggested by Amy Rose Deal, the two steps in (162) indeed could be a one-step operation, like the one that follows.

Bare Numeral Classifier Phrases and Bare Nouns in Mandarin

However; it is unclear how these two operations and ι‎ are “fused” into one step: do they apply simultaneously or are there additional assumptions to make here? Even if the two operations apply at once and count as one step, they are always two operations. If we can achieve the same result with one operation, that would be a simpler analysis.

(50.) Note that there could be a number of ways of obtaining definiteness from kinds. SR is one such way, so is a new type of iota of type <ek, e> which directly gives rise to the definiteness of kind-referring bare nouns. However, they are both the same idea, and the “new iota” will be undistinguishable from the situation restriction that we proposed.

(51.) Regarding how to account for the phenomenon in English in (1) to (3), I refer the readers to the discussion in Condoravdi (1997) and Dayal (2013).

(52.) Indeed, Lü (1944: 160–161) and Chen (2004) have illustrated that Mandarin [Cl N] can receive a specific indefinite interpretation under various circumstances besides the ba-construction.

(53.) Regarding why the examples with [Cl N] appearing in the ba-construction in (202i) are marked as unacceptable in Cheng and Sybesma (1999) (indeed they are acceptable to me as well as my informants), I will leave it for further research.

(54.) Li and Bisang (2012: 339) treat ba-construction with [Cl N] as marginal because they find that no consensus among their informants regarding the grammaticality of ba-sentences with [Cl N]. Nevertheless, as we have seen in (182) and in (183), examples in which [Cl N] appears in ba-construction are widely accepted by variouls scholars and are also easlily found in daily life and literary work. These examples have to be acknowledged and cannot be simply treated as marginal as Li and Bisang (2012).

(55.) Besides eliding yi “one” completely, yi /i/ can be pronounced with the demonstrative zhe “this” or na “that,” yielding the pronunciation zhei or nei.

(56.) We can view that the numeral one in bare ClPs is either realized in a phonetically null form (i.e., [onenull Cl N]) or it is being deleted in the phonology after narrow syntax; the difference between these two does not make a difference for the purpose of our discussion.

(57.) I am aware of other approaches to the semantics of topicalization (e.g., Buring’s (1995, 1997) alternative semantics) but choose to adopt the traditional view of topic for reasons of simplification.