Classifiers in East and Southeast Asian languages Counting and beyond

74
Classifiers in East and Southeast Asian languages Counting and beyond Walter Bisang 1. Introduction 1.1. General outline of the paper The function of classifiers is to make count nouns enumerable by individual- izing and classifying them. This is more or less the standard view on classifi- ers. In fact, there are many languages such as Modern Standard Chinese and Japanese in which this is their main function. However, this is not the only function that classifiers can display in the languages of East and Southeast Asia. In some languages, classifiers go beyond counting. As we shall see from Vietnamese, they can even display the functions of individualization and classification without being primarily involved in counting (§ 1.2, §2.3.2.2). In addition to the functions of individualization and classification, classifiers are also used in the context of reference (function of referentialization) and/ or in the context of possession and relative clauses (function of referentiali- zation). The languages to be discussed for these additional functions are Thai, Hmong, Cantonese, and the Miao language of Weining (Guizhou Province, China). If we want to understand how it is possible for one and the same linguistic item to be maximally used for all of these four functions we have to look at the common cognitive basis that allows classifiers to further expand their functional range. Since this is a challenge which needs more than just a paper and since my primary aim is to present empirical evidence, I shall only give a sketch of the basic cognitive operations that may be involved in classi- fiers and their functional expansion (cf. §1.2, §2.2.3 and §2.3.1). The fact that classifiers can go beyond counting provides a valuable and fascinating em- pirical basis for understanding what counting has in common with other gram- matical functions. As will be shown in this paper, there is an enormous range of functional variation among language specific classifier systems of which most com- parative studies on classifiers do not seem to be aware. Nevertheless, the whole range of variation seems to be connected to only two diachronic proc- Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois Library Authenticated | 172.16.1.226 Download Date | 8/8/12 2:08 AM

Transcript of Classifiers in East and Southeast Asian languages Counting and beyond

Classifiers in East and Southeast Asian languages Counting and beyond

Walter Bisang

1. Introduction

1.1. General outline of the paper

The function of classifiers is to make count nouns enumerable by individual-izing and classifying them. This is more or less the standard view on classifi-ers. In fact, there are many languages such as Modern Standard Chinese and Japanese in which this is their main function. However, this is not the only function that classifiers can display in the languages of East and Southeast Asia. In some languages, classifiers go beyond counting. As we shall see from Vietnamese, they can even display the functions of individualization and classification without being primarily involved in counting (§ 1.2, §2.3.2.2). In addition to the functions of individualization and classification, classifiers are also used in the context of reference (function of referentialization) and/ or in the context of possession and relative clauses (function of referentiali-zation). The languages to be discussed for these additional functions are Thai, Hmong, Cantonese, and the Miao language of Weining (Guizhou Province, China). If we want to understand how it is possible for one and the same linguistic item to be maximally used for all of these four functions we have to look at the common cognitive basis that allows classifiers to further expand their functional range. Since this is a challenge which needs more than just a paper and since my primary aim is to present empirical evidence, I shall only give a sketch of the basic cognitive operations that may be involved in classi-fiers and their functional expansion (cf. §1.2, §2.2.3 and §2.3.1). The fact that classifiers can go beyond counting provides a valuable and fascinating em-pirical basis for understanding what counting has in common with other gram-matical functions.

As will be shown in this paper, there is an enormous range of functional variation among language specific classifier systems of which most com-parative studies on classifiers do not seem to be aware. Nevertheless, the whole range of variation seems to be connected to only two diachronic proc-

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

114 Walter Bisang

esses of development, an item-oriented process and a category-oriented proc-ess. Both processes can be areally located. The former process leads to the Chinese system of classification, the latter process is involved in the classi-fier systems of Southeast Asian languages such as Vietnamese, Thai, and Hmong (§3). Both processes bear different consequences to be described in §3.3. One of the consequences may be that classifier use in the context of counting is not necessarily linked to the category-oriented development (cf. Vietnamese). Another consequence may be that the category-oriented proc-ess tends to make a noun syntactically referential (Löbel 1996, to appear; on Vietnamese, cf. §2.3.2.2) and thus paves the way for classifiers to occur in the function of referentialization and relationalization. The functional expansion of classifiers may therefore be due only to the category-oriented process.

The structure of this paper will be as follows: In § 1.2,1 shall start to sketch the cognitive basis that allows the functional expansion of classifiers, and I shall present a typological framework for comparing classifiers which will be based on the four functions discussed above. Some basic information on word order will be presented in §1.3. The whole of §2 will be dedicated to a de-tailed description of the four functions of classifiers and the languages in which they occur. The next section, i.e. §3, will deal with the two processes involved in the development of classifiers.

7.2. A typology of classifier languages

In East and Southeast Asian languages, a noun expresses only a mere concept of an object which can be further specified - if necessary - by various opera-tions. Thus, an expression like e.g. Chinese xin can mean 'letter, letters, a letter, the letter, etc.' according to a given context. This high degree of inde-terminateness of nouns is a very important characteristic of East and South-east Asian languages (Bisang 1993,1996) which is crucial to the existence of classifiers.

Since Greenberg (1974: 25), one particular instance of indeterminateness, i.e. transnumerality, is generally linked to the existence of classifiers in a given language.

Numeral classifier languages generally do not have compulsory expression of nominal plurality, but at most facultative expression.

In a language with transnumerality there is no pragmatic constraint to mark plurality, so there is no Horn scale of the type <PL, 0> (Horn 1984, 1988).

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 115

Furthermore, there is no Quantity Implicature from non-plural to singular. According to Greenberg (1974) a noun in a transnumeral language cannot occur in immediate combination with a numeral - it first has to be individual-ized by the classifier in order to be countable or enumerable. Thus, Chinese xin 'letter' has to be individualized with the classifier feng if it is used in the context of counting:

(1) Modern Standard Chinese a. sän ßng χιη

three CL letter 'three letters'

b. *sän χιη three letter

In most classifier languages, we find several classifiers which form a seman-tic system to classify nouns (§2.2). Thus, individualization is not the only function of classifiers; they also display the function of classification. These two functions apply to all the classifier languages to be analysed in the present paper, i.e., they are the core functions of classifiers. The question of why these functions form the general basis of classifiers and the question of why classifiers expand to referentialization and relationalization in some languages can be tentatively answered by looking at the cognitive background of these functions. I start my provisional sketch of the cognitive operations at work in classifiers by discussing classification and by asking what it is good for. As was pointed out by Seiler (1986: 95), "classification is a mental operation that causes an object or a multitude of objects to fall under a concept X". At the basis of classification, we find a set of properties which constitute a cer-tain class. For our further discussion, we shall only look at inherent proper-ties, since they are relevant for classifiers. External properties will be briefly discussed in §2.1. The operation of classification, i.e. the operation of sub-suming a given entity under a certain class according to its properties, can be used for two purposes. Classification can be employed to compare one par-ticular sensory perception and its properties to the properties of other sensory perceptions in order to identify that particular perception by subsuming it under a certain concept or it can be employed to establish a sensory percep-tion as an individuum by actualizing the inherent properties which constitute its conceptual unity.1 The former operation is called identification, the latter individualization. Identification can take place without referring to individu-alization. It seems possible to identify a sensual perception as belonging to

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

116 Walter Bisang

the concept of say 'apple' without explicitly referring to its inherent bounda-ries, i.e., without referring to it as an individual. On the other hand, it seems difficult to individualize a sensual perception without identifying it. Depart-ing from classification, which is "a prerequisite to identification" (Croft 1994: 161), we may tentatively establish the following hierarchy: Classification > identification > individualization.

The fact that classifiers are used in the context of counting can be ex-plained by their function of classification and individualization (for further details cf. §2.3.1). One should however keep in mind that individualization does not necessarily imply that classifiers must be used obligatorily with count-ing. This is shown by the example of Vietnamese. In this language, the indi-vidualizing function of the classifier is primarily used to make a NP syntacti-cally referential (Löbel 1996, to appear; cf. §2.3.2.2). The link between indi-vidualization and quantification is thus not as absolute as one might infer from the general literature (cf. e.g. Greenberg 1974). Vietnamese classifiers are obligatory only if humans are counted. Nevertheless, they individualize nouns. A construction of the structure 'CL-N' refers to an individual item.

The functional expansion of classifiers into the marking of reference or the marking of possession and relative clauses can be seen in the context of identification. If we define reference as an act of identifying some entity that the speaker intends to talk about (cf. Searle 1969, Croft 1991: 110), we may say that the identificational function of the classifier forms the point of depar-ture for classifiers to take on the function of referentialization. As for relationalization, one might argue in a similar way that it is linked to the previous identification of the head noun before it can be modified by a pos-sessor or a relative clause. The functional expansion of classifiers may thus be schematically represented by the following figure:

Classification > Identification > Individualization

Figure 1. Functional expansion of classifiers

As pointed out above, classification and individualization (including identifi-cation) are present in all the classifier languages of East and Southeast Asia to be discussed, even though individualization does not necessarily entail the obligatory use of classifiers in the context of counting. Furthermore, classifi-

I 1 Referentialization/ Relationalization

Counting

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 117

ers are also used for referentialization and/or relationalization. If we combine one or both of these functions to the functions of individualization and classi-fication, we get four different types of classifier languages. The following typology is an improvement to the typology suggested in Bisang (1993,1996):

1. classification — individualization

Japanese (classifiers occur only with numerals; §2.2.2.4, §2.3) Chinese (classifiers occur with numerals and demonstratives; §2.2.2.3., §2.3.2.1.) Vietnamese (individualization, but not necessarily in the context of count-ing; §2.2.2.5, §2.3.2.2)

2. classification - individualization - referentialization

Thai (secondary function in combination with adjectives [stative verbs] in the sequence N-CL-ADJ; §2.2.2.1, §2.4.3.1)

3. classification — individualization - relationalization

Cantonese (classifiers can be used in possessive and relative constructions; §2.5.2)

4. classification - individualization - referentialization - relationalization

Hmong (with referentialization being a secondary function; §2.2.2.2, §2.4.3.2, §2.5.1), Weining Miao (§2.4.3.3, §2.5.1)

Although the classification of nouns is an important function of classifier systems, they must be clearly distinguished from noun class systems or gen-der systems. As pointed out by Corbett (1991), what determines gender is agreement:

While nouns may be classified in various ways, only one type of classification counts as a gender system; it is one which is reflected beyond the nouns them-selves in modifications required of 'associated words'. ... All this means that the determining criterion of gender is agreement; this is the way in which the genders are 'reflected in the behavior of associated words'. (Corbett 1991: 4)

Croft (1994) presents four different hierarchies of semantic distinctions asso-ciated with four types of noun classification, viz. noun classes, numeral clas-

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

118 Walter Bisang

sifiers, possessive classifiers, and classifiers of spatial predication. If there are some languages like Cantonese or Hmong and some other Miao languages in which possession is expressed by numeral classifiers the two different se-mantic hierarchies for numeral classifiers and possessive classifiers as sug-gested by Croft cannot be universal. In his typology of noun classification, Croft characterizes noun classes by reference: "Noun classes are used in de-termination, which fixes reference" (Croft 1994: 161). If classifier systems can at least express some aspects of reference in some languages, reference cannot be crucial to the distinction of noun classes and numeral classifiers. Agreement, as suggested by Corbett (1991), seems to be a more distinctive candidate.

1.3. Some remarks on word order2

As shown by Greenberg (1974: 31), of the six possible word orders among the three elements NUM (numeral), CL (Classifier) and Ν (noun), only four seem to be realized. The impossible word orders are those in which the clas-sifier is separated from the numeral by the noun:3

(2) List of possible word orders with NUM, CL, and Ν

a. [NUM-CL]-N b. N-[NUM-CL] c. [CL-NUM]-N d. N-[CL-NUM]

The first two word orders will be by far the most relevant for the East and Southeast Asian languages to be analysed in the present paper. Word order (2a) is realized in Chinese, Vietnamese, Hmong and the Miao language of Weining, word order (2b) occurs in Thai and Khmer. Word order (2d) is a variant to (2b) with the numeral 'one' in Thai.4 As I pointed out in Bisang (1996), word order within the NP follows an areal pattern. In the North, rep-resented by Chinese, all the modifiers precede the head noun, whereas in the South, represented by Thai and Khmer, all the modifiers follow the head noun. The modifier which is most likely to break this rather consistent rule is the classifier/quantifier constituent. This constituent occurs in the position in front of the head noun in languages such as Vietnamese, Hmong and Weining Miao, in which all the other modifiers consistently follow the noun. In Yao, the

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 119

demonstrative also precedes the noun, while adjectives still follow it. Thus, a NP in Yao has the structure 4DEM-[NUM-CL]-N-ADJ' (Caron 1987). In the Tai language family, the postnominal position of the classifier/quantifier con-stituent seems to be limited to Southwestern Tai, which includes Thai. In the languages of the other two branches, i.e. Central Tai and Northern Tai, the classifier/quantifier constituent precedes the noun.

In Japanese, which is the only verb-final language to be discussed in this paper, there are two possible word orders. The classifier/quantifier constitu-ent occurs either in the prenominal position followed by the genitive marker no as in (3a), or it follows the noun and its case particle as in (3b):

(3) Japanese a. ni-dai no kuruma ο kai-mashi-ta.

two-CL GEN car ACC buy-HON-PST '[S/He] bought two cars.'

b. kuruma ο ni-dai kai-mashi-ta. car ACC two-CL buy-HON-PST '[S/He] bought two cars.'

Finally, there are three different word orders in Korean: [1.] The classifier phrase precedes the noun and is marked by the genitive, [2.] the classifier phrase immediately follows the noun and takes case marking, and [3.] the classifier follows the head noun and both, classifier and head noun, are case marked (Sohn 1994: 272).

2. The functional range of numeral classifiers

This section presents the four functions of classifiers in detail. In §2.1, indi-vidualization is discussed and related to the processes of counting and meas-uring. Furthermore, classifiers and quantifiers are arranged in a system of coordinates that allows to distinguish classifiers from quantifiers on the one hand, and different subcategories of quantifiers on the other. In §2.2, we shall see that classification follows a hierarchy which is generally realized in clas-sifier languages. In addition to the structure of this hierarchy and its cognitive motivation, the classifier system of the following five languages will be pre-sented: Thai, Hmong, Modern Standard Chinese, Japanese, and Vietnamese. The interaction of the core functions, i.e. classification and individualization

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

120 Walter Bisang

(including identification), and its cognitive motivation will be discussed in §2.3, with some data on languages whose classifiers are limited to the core functions. §2.4 will be dedicated to various discourse phenomena connected with classifiers. The anaphoric function of classifiers, the selection of normal vs. general classifier and the function of referentialization will be the topics of that section. Finally, §2.5 will present the function of relationalization.

2.1. Individualization, counting vs. measuring

There are two different operations of quantification, i.e. counting and meas-uring. The operation of measuring puts a given entity into a unit of measure which makes it countable or enumerable, i.e. discrete. The process of measur-ing must be applied for the quantification of semantically unbounded mass nouns such as water, wine, etc. Although boundedness as defined by Langacker (1987: 189-213) with its properties of homogeneity, expansibility, and replicability is a category inherent to a given entity which cannot be derived exclusively from linguistic context, it is possible to override it by discourse pragmatics. For that reason, count nouns, which are semantically bounded, such as man, stone, etc., can also be quantified by the process of measuring (a group of men, a cartload of stones).

The operation of counting can be applied almost exclusively to semanti-cally bounded entities. This is due to the fact that bounded entities have a delimitation and are thereby replicable. Mass nouns or unbounded entities, on the other hand, tend to be unenumerable because of their continuousness and expansibility. If mass nouns are used in the context of counting at all, this is limited to highly special situations. Thus, if we order two mineral waters in a restaurant, we refer to a particular man-made product which is semantically bounded. Utterances such as We've been drinking two nice wines yesterday imply that we are talking about two different sorts of wine.

Classifiers occur with counting or enumeration, never with measuring. In languages like those of East and Southeast Asia, where nouns show a high degree of indeterminateness (§ 1.2), they individualize a given count noun by designating its semantic boundaries or, as Croft (1994: 163) puts it, by desig-nating its 'natural unit'. Linguistic items which are used for measuring are called quantifiers. The cover term for classifier and quantifier is numerative (for this terminology cf. Hla Pe 1965). Quantifiers can be found in all the languages of the world, but classifiers are limited to languages with a high degree of indeterminateness. In languages such as English, there are quantifi-

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 121

ers like lump or ball in a lump of sugar or a ball of thread, but there are no classifiers.

Although both, classifiers and quantifiers, have the function of individu-alization, there is a significant semantic difference. The primary purpose of quantifiers is to 'create' the unit to be counted (Croft 1994:162) whereas clas-sifiers merely actualize the semantic boundaries which already belong to the concept of a given noun. The former type of individualization may be called creative individualization, the latter actualizing individualization. Creative individualization is based on properties which can only be recognized by applying external scales, whereas actualizing individualization is based on inherent properties. The creative vs. actualizing distinction will turn out to be important for explaining the particular semantic hierarchy underlying classi-fier systems, since only actualizing individualization can have recourse to inherent properties which make the individuality of a given noun (§2.2.3). Apart from its relevance for explaining the particular semantic hierarchy rep-resented in classifier systems, however, the distinction between actualizing and creative individualization seems to be of secondary importance. It does not seem to seriously affect referentialization and relationalization. If classi-fiers can express one or both of these functions in a given language, quantifi-ers seem to be able to do more or less the same. If we take identification as the basis of these two functions, it does not seem to matter whether the properties for establishing the individuality of an item are inherent or external. The dif-ference matters, however, within the context of individualization and quanti-fying, where the distinction of inherent vs. external yields the two processes of counting and measuring.

As suggested by the title of the present paper, I shall confine myself to the presentation of classifiers. Before doing this, I would like to use the rest of this section to elaborate more clearly on the position of classifiers within numeratives. For that purpose, I shall start from the two parameters of [± exact] and [± entity] as introduced by T'sou (1976), which I tried to arrange within a system of coordinates in Bisang (1993: 14). As we shall see below, these two parameters yield four types of numeratives to which I shall add a fifth type, which will be situated in the center of my system of coordinates.

The parameter of [± exact] indicates whether the quantity of the object to be determined by the numerative is exactly defined or not, and the parameter of [± entity] refers to the quality of the object, i.e. its countability or discrete-ness. By combining these two parameters we get four types of numeratives, the first type corresponding to classifiers, the other three types corresponding to different subtypes of quantifiers. I shall first present the definitions of these

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

122 Walter Βisang

types with quotations from T'sou (1976: 1217-1218) followed by some ex-amples from Hmong. T'sou (1976) also presents a variety of tests to show the syntactic consequences of these types which I shall not repeat in the present paper.

1. Classifiers

[+ exact], [+ entity] "... refers to an exact quantity and involves discrete physical entities."

2. Measures

[+exact], [- entity] "the measure is exact, but it refers to no discrete physical entity. Pounds, gallons, and feet for example are commonly known as measure words."

3. Collectives

[- exact], [+ entity] "there is a definite sense of well defined discrete entity or entities, but the quantity is not exact either by design or by convention"

4. Kind

[- exact], [- entity] "which characterizes mainly abstract nouns, the measure is neither exact nor does it refer to a discrete physical entity."

(4)Hmong

a. Classifier ib tug neeg one CL man/person 'one man/person'

b. Measures ib thoob dej one Q:bucket water 'one bucket of water'

ib rab riam one CL knife 'one knife'

ib mev hlua maj one Q: meter hemp_rope 'one meter of hemp rope'

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 123

c. Collectives ib pab qaib one Q: group/ chicken

ib xub5 nas one Q:nest rat

brood 'a brood of chickens' 'a nest of rats'

d. Kind ntau yam tsiaj txhu ib hom neeg many Q:kind animal one Q:kind man 'many species of animals' 'a species/kind/race of men'

The fifth type of numeratives can be developed from looking more closely at the feature [+exact] in type II (Measures). Measure constructions take the norm, which yields the feature [+ exact], from outside of the entity to be quantified, and therefore represent a measuring norm extrinsic to it. An alter-native possibility to an extrinsic norm is an intrinsic norm, which is taken from the entity to be measured itself. If a given mass noun such as sugar or a given count noun such as wire appear in the shape of a lump or a coil, respec-tively, these particular intrinsic structures are taken to measure these nouns: a lump of sugar, a coil of wire. In Bisang (1993: 11-13), I called this particular type of numeratives intrinsic quantifiers (intQ). Measures and intrinsic quan-tifiers can be distinguished by the additional features of (extrinsic) and (in-trinsic) which specify the feature [+ exact]. Thus, we get [+ exact(extrinsic)] for measures and [+ exact (intrinsic)] for intrinsic quantifiers. As I tried to argue in Bisang (1993: 13-14), intrinsic quantifiers belong to the center of a system of coordinates, where [± exact] forms the x-axis and [± entity] forms the y-axis:

[+] III. Collectives I. Classifiers

Ε [- exact] Ν Τ I Τ IV. Kind Y [- entity]

[+ entity] [+ entity] [+ exact]

Intrinsic Quantifier [- entity] [+ exact(intrinsic)]

I [- exact] [-] [-]

E X A C T

II. Measures [- entity] [+exact(extrinsic)]

W

Figure 2. System of coordinates for the five types of numeratives

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

124 Walter Bisang

2.2. Classification

In §2.2.11 shall first introduce the hierarchy of semantic distinctions underly-ing classifier systems in general. The next section, i.e. §2.2.2, is to illustrate how this hierarchy is realized in individual languages. Finally, I shall try to present some cognitive explanations for the semantic hierarchies underlying classifier systems in §2.2.3.

2.2.1. General remarks

The lists of classifiers as presented in traditional grammars or dictionaries are usually rather impressive. One such example is Emeneau (1951: 110-113), who presents no less than 125 classifiers. Another example is McFarland's (1944: xiv-xvi) Thai-English dictionary, which presents a list of 82 'numeri-cal designatory particles'. If one looks more closely at these lists by using the criteria introduced in §2.1, the majority of its items turn out to be quantifiers. What is left over is a fairly small set of true classifiers, whose semantic crite-ria of classification converge into a system of considerable cross-linguistic similarity and uniformity which can be described hierarchically. Neverthe-less, one must always reckon with additional criteria. In this sense, the se-mantic hierarchy of classification to be presented below is a pattern which is appropriate at least for the great majority of languages, but which emerges in different degrees of consistency in individual languages.

The properties determining classification are described by Denny (1976) and Allan (1977). Since these properties are well-known, I shall not discuss them in this paper. I shall rather more concentrate on the hierarchies of se-mantic distinctions at work in classifier languages, which include many of the categories discussed by Allan and Denny. As was pointed out by Adams — Conklin (1973), the first distinctions made in classifier systems are almost universally between animates and inanimates and humans and non-humans. A secondary criterion to [+ human] is status, i.e. the social status of the per-son denoted by the noun to be classified. Status is very important in some languages of mainland Southeast Asia such as Thai (§2.2.2.1), Khmer, Bur-mese (Becker 1975; cf. below), and Vietnamese (cf. (21), §2.2.2.5). In Chi-nese, it is only marginally represented by the classifier wei (cf. (14b), §2.2.2.3). In contrast to languages with noun classes, the criterion of sexus is an ex-tremely marginal criterion in classifier languages, if it exists at all.6 The sec-ond consistent primary criterion of classification discovered by Adams —

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 125

Conklin is shape, i.e. the extension of objects in the three spatial dimensions. Thus, shape distinguishes long (one-dimensional) from flat (two-dimensional) and round (three-dimensional) objects. Secondary to the criterion of shape are physical parameters such as rigidity vs. flexibility, relative size, empty vs. full (hollowness), irregularity vs. regularity, part vs. whole, horizontal vs. vertical (orientation), and edginess. These secondary parameters can never form an independent basis for individual classes. A last criterion is based on the nature or the function of a noun. This criterion seems to be secondary to shape as well. Thus, the criteria of [± human]/[± animate] and [shape] form the basic semantic grid at work in classifier systems.7 It can be represented within a hierarchy such as the following, which slightly deviates from the one presented by Croft (1994: 152):

(5) Animate/ < Status Animacy {Human

Inanimate/ < Shape < Orientation/ < Nature/ Nonhuman Rigidity Function

In many languages, the criterion of shape is closely linked to trees, plants, and their parts. As observed by Adams — Conklin (173: 5), 'trunk' is con-nected to one-dimensionality, 'leaf' to two-dimensionality and 'fruits' to three-dimensionality. This phenomenon is attested in Chinese, in many Austroasiatic languages (Adams 1982, 1986) and in Tai and Austronesian (Conklin 1981). Conklin presents the following explanation:

This [i.e. the development of shape-based classes out of 'related plant parts', W.B.] can be conceived of in two separate, but complementary ways. Classi-fiers are expanded generics and the physical base that lends itself most obvi-ously to extension beyond generic reference (at least, in these languages) is plant parts. Or, plant parts present a metaphorical source that is exploited lexically to create abstract shape classifications, using the plant part generics for lexemes. No doubt both of these forces operate simultaneously. (Conklin 1981: 342)

Erbaugh (1986) presents an excellent explanation for Chinese, a language whose classifier system is based on the item-oriented process (§3.1):

Trees and wood are both the source of an enormous variety of artifacts which people handle in a variety of critical ways. These action schema may have reinforced the proliferation of classifiers from several words for trees and tree parts. In addition, trees and wood are solid, discrete, countable objects, unlike

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

126 Walter Bisang

the massy quality of clay, ore, geodes, and fibres, the other early raw materials for valued objects. (Erbaugh 1986: 430)

As pointed out above, there are also other criteria of classification at work in at least some classifier languages. Unlike the above criteria, which are based on properties of the items themselves, it is also possible - as pointed out by Becker (1975) for Burmese - to classify items according to relationships be-tween them. Becker postulates a 'self-other' continuum which underlies the whole Burmese classifier system without replacing the other system based on [± human]/[± animate] and [shape]. This deictic relation as discovered by Leach (1964, also cf. Becker 1975: 120-121, footnote 14) arranges people, objects, and ideas on a near-to-far continuum, where near and far can be un-derstood physically or psychologically. Becker distinguishes four orbits of distance from the center. With regard to the classification of humans, the con-tinuum starts from Buddha, who represents the center, and ends with the fourth orbit, i.e. with "animals, ghosts, and base, depraved people" (Becker 1975: 115):

(6) Burmese (Becker 1975: 116) shu Buddhas, relics, images, the Law (secondarily: nets, staircases,

gardens) 'pa deities, saints, monks, royalty, (gems) 'u people of status, teachers, scholars yau? ordinary humans kauq animals, ghosts, dead bodies, depraved people, children.

For the classification of inanimates, the center is 'person'. In the first orbit we find items which inalienably belong to the center, e.g. hair, leaves, twigs, etc. Clothes, which belong to the second orbit for inalienable objects 'on the se l f , are understood as being nearer to 'person' than chairs (third orbit for objects 'nearby sel f ) , which in turn are nearer to 'person' than vehicles (fourth orbit for objects 'far from self'). Finally, the 'self-other' contiuum is combined to a top vs. bottom opposition in the following way:

The structure underlying classification starts with the self at the center, di-vides the self into head and body, and then ranges objects at four distances from the self, associating them either with the head (metaphorically top, round) or with the body (metaphorically bottom, straight). (Becker 1975: 118)

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 127

For each orbit, there is a classifier representing 'top', and a classifier repre-senting 'bottom'. I shall illustrate this opposition by the third orbit opposition expressed by the classifiers louq (top) and cha? (bottom). A noun such as 'cup' is classified by the 'top'-classifier /ow/7 and saucer by the 'bottom'-classifier cha?. In a similar relation we find chair and mat, and post and floor.

Finally, I would like to point out that some classifier languages have par-ticular genres in which the set of classifiers is employed according to differ-ent criteria of classification than under normal conditions. One such language is Bahnar (Austroasiatic: Bahnaric):

... speakers in the Bahnar culture of Vietnam employ a special genre when planning ceremonies such as sacrifices. This speech genre is designed to dep-recate the speakers and other objects so that the spirits do not recognize who and what is being discussed. The function of the deprecating forms is to pro-tect the speakers in case they must call off the sacrifice. If the normal Bahnar language forms, which spirits understand, were used and the promised ritual were cancelled, the spirits would know this, become angry and punish the person who had proposed the ceremony in the first place.... When using this genre, speakers can still employ the same set of classifiers ... The difference is that the objects being counted are placed in classes they would not normally belong in. (Adams 1986: 243)

2.2.2. Some data on East and mainland Southeast Asian languages

In this section, I shall present Thai (§2.2.2.1), Hmong (§2.2.2.2), Modern Standard Chinese (§2.2.2.3), Japanese (§2.2.2.4), and Vietnamese (§2.2.2.5). The languages are arranged according to the degree of lexicalization involved in the selection of classifiers. In Thai and Hmong, the classifier is to a high degree lexically selected by the noun. In Chinese and Japanese, the selection of classifiers also depends on discourse pragmatics (§2.4.2). In Vietnamese, the lexicon is of minor importance for selecting the classifier (§3.2.3).

2.2.2.1. Thai

A very straightforward illustration to the basicness of the hierarchical criteria as presented above is the classifier system of Proto-Tai as reconstructed by DeLancey (1986: 447). The first three classifiers presented in (7) are at-tested throughout the family of Tai languages, the last two classifiers are only

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

128 Walter Bisang

attested in Central and Southwestern Tai, not in Northern Tai. The classifica-tional criteria at work in this system are [± human], [± animate] and shape [one- and two-dimensional]:

(7) [- human], [+ animate]: *tueA7 (for animals) [- animate]: *?anA (for inanimates in general) [shape(one-dimensional)]: *senc

[+human]: *οηΑ

[shape(two-dimensional)] *phi'9nA

This basic system is also at work in modern Thai. The classifier system of this language is particularly consistent inasmuch as there is usually one and only one classifier for a particular noun (Haas 1942), if it is not classified by the general classifier ?an, which is nowadays used "loosely as a substitute for almost any other classifier" (Haas 1964: 614). Some of the very few excep-tions are presented below. Since Thai classifiers are extensively described by Conklin (1981) and Hundius — Kölver (1983), I shall only mention some aspects of that system.

Thai distinguishes several degrees of status within the class of [+ human]: khon for humans in general, naaj and ηααη for people who have a certain official position (naaj for men, ηααη for women), rüup for monks and ascets, ?οη and phrä-Ϋοη for the king and members of the royal family.

As we can see from the above list, there is a particular classifier for ani-mal, which is realized as tua in Thai. This classifier is used for all animals with the exception of the elephant, for whom there is the special classifier chyak whose meaning of 'rope' refers to the elephant's trunk.

Of particular interest are the Thai classifiers for shape. There is a consid-erable number of classifiers for one- and three-dimensionality, which are fur-ther distinguished by secondary criteria such as rigidity, hollowness and size. The following list presents the classifiers for one-dimensional objects:

(8) tön 'stem, stalk': [one-dimensional], [+ rigid] Classifier for trees and plants: saw 'pole, pillar', swg 'log'.

lem [one-dimensional], [+ rigid] (only occurs in the function of a classifier) Classifier for long, pointed objects such as: khem 'needle', thian 'candle', takray 'scissors'. Morover, lem is used for classifying books.

kham 'handle' [one-dimensional], [+ rigid] Classifier for long objects with a handle: ch6on 'spoon', s3m

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 129

'fork'. Classifier for vehicles: rotyon 'car', cäkrayaan 'bicycle'. daam 'handle' [one-dimensional], [+ rigid]

Classifier for long objects with a handle: päakkaa 'fountain-pen'. lam [one-dimensional], [+ hollow] (also occurs as a class noun)

mdayphay 'bamboo stem', rya 'boat', khryaij-bin 'aeroplane'. säay/sen [one-dimensional], [- rigid(flexible)], [± big]

Both classifiers are used with one-dimensional, flexible or bent objects. The classifier säay is rather used for big objects, sen rather for small objects.

For three-dimensional classifiers, there is a similar list:

(9) lüuk 'child, offspring, fruit' [three-dimensional] A rather productive classifier for fruits and three-dimensional objects in general.

phön 'result, outcome' [three-dimensional] (derived from Pali phalam 'fruit') This classifier is an elegant variety for lüuk in combination with fruits.

bay 'leaf [three-dimensional/two-dimensional] This classifier is used with leaf-like two-dimensional nouns such as thdndbat 'bank note' as well as with three-dimensional, hol-low objects, i.e. containers. Finally, it overlaps with lüuk for fruits.

duαη [three-dimensional] [luminescent] daaw/daaraa 'star', ?aathit 'sun', khoomfay 'lamp'.

met 'seed, grain' [three-dimensional] [small] khaymuk 'pearl', kradum 'button'.

koon [three-dimensional] [bulky/irregular] krüat 'stone, pepple', meek 'cloud'.

As we can see from the above list, the classifiers lüuk, bay, and phön overlap with fruits. The classifiers lüuk and bay can be used alternatively for fruits without any change of meaning, whereas phön is used in contexts of elegant style - it may even sound a bit snobbish. Another instance of overlapping is linked to the nouns päakkaa 'fountain-pen' and miit 'knife' which both can be classified either by daam or by the semantically more general classifier lem. In some cases, different classifiers are used with homophonic nouns, with polysemic nouns and with nouns whose meaning is metaphorically ex-tended (Hundius — Kölver 1983:188). The noun daaw 'star' is an example of

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

130 Walter Bisang

the last category of nouns. If daaw refers to a celestial body it takes the clas-sifier duaij, if it refers to a person, e.g. a movie star, it is classified by khon. As pointed out by Gandour et al. (1987), classifiers denoting configuration (i.e. shape, consistency, or size) are acquired with the greatest difficulty by children. Children first learn classifiers denoting animacy, followed by clas-sifiers marking arrangement or quanta, followed finally by classifiers refer-ring to configuration. The most problematic classifiers are those denoting one-dimensionality. According to Gandour et al., this is due to the high de-gree of arbitrariness of these classifiers; they are the category most fossilized by linguistic convention. A child has basically to learn for each noun with which of the different one-dimensional classifiers it occurs (for an explana-tion based on the organization of the human visual system, cf. §2.2.3).

In Thai, every count noun has to be classified. If a noun cannot take a classifier, there is an alternative technique, i.e. the repeater construction. This construction 'repeats' the noun to be individualized in the position of the classifier as shown in the following example:

(10)/:3? säam ko? N:island NUM:three CL:island 'three islands'

In complex nouns consisting of a class noun in the head position, only the head is repeated. This type of construction is called semi-repeater construc-tion:

(11) raan-?aahaan säam räan shop-meal/food three CL:shop 'three restaurants'

Repeater and semi-repeater constructions are widely used in Thai. There seem to be several hundred nouns classified in this way. As we shall see below, repeater and semi-repeater constructions are important for reconstructing the development of classifier systems (§3.2.1).

2.2.2.2. Hmong

In Hmong, there is a very limited number of seven classifiers which are used for the vast majority of nouns (cf. (12)) and four special classifiers which

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 131

occur only with a few specific words listed in (13) (Bisang 1993). As in Thai, a noun usually takes only one classifier. The seven widespread classifiers are:

(12)[+ human], [+ animate]: leej, tus [- human], [+ animate]: tus [- animate]: lub [one-dimensional]: txoj, tus [two-dimensional]: daim [three-dimensional]: lub [function(tool, instrument)]: rab [one item of a pair]: txhais

Some nouns can take one of the four special classifiers:

(13)£wv txhiaj 'song', npau suav 'dream': zaj chaw 'land' qhov (as an alternative to the

classifier lub) siab 'liver' thooj (as an alternative to the

classifier lub) ntawv8 'letter', xov 'news, message' tsab

Some particularities presented in the above list of Hmong classifiers need some more explanation. The classifier leej can only occur with humans. There-fore, it is more explicit than tus, which is used with animates in general. Both classifiers can be used for humans alternatively without much difference in meaning.

The classifier tus combines the properties of [± animate] and [shape]. It refers to animates (neeg 'man, human being', tsiaj 'animal'), to things closely related to humans or animals (plig 'soul', duab 'shadow', mob 'disease, ill-ness' , dab ntub 'sleep'), to long body parts and to long objects in general (ntiv ' f inger' ,pob txha 'bone'; mem 'pen', ntsia 'nail', ncej 'upright post, pillar'), and finally to some abstracts (cai 'law', nqe 'price'). The classifier lub is possibly the most common classifier. A great number of new words in Hmong take this classifier. It combines the functions of [- animate] and [shape(three-dimensional)]. In this function, it refers to round or hollow objects (khaub hnab 'ball', qhov 'hole'), body and body parts (cev 'body', taub hau 'head'), and to vehicles and machines (tsheb 'car', dav hlau 'airplaine'), i.e., things which are usually roughly three-dimensional and hollow in the sense that

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

132 Walter Bisang

there is some technical mechanism within them. Moreover, it also classifies garments (tsho 'shirt', ris 'trousers'), i.e. things which are hollow in the sense that they can be 'filled' by a human body, and places or buildings (teb chaws 'country, area', tsev 'house'). Like tus, lub also classifies abstracts (caij 'time', npe 'name', tswvyim 'intelligence', hwj huam 'power', neej 'life').

There are two classifiers marking one-dimensionality, i.e. tus and txoj. As was pointed out by Jaisser (1987: 175), both classifiers can occur with body parts - tus "with body parts that come in 'short' lengths", and txoj "with body parts that come in 'great' lengths, and are thin and flexible". Both distinc-tions, i.e. short length vs. great length and rigid vs. flexible, also seem to hold if these classifiers are applied to other inanimate entities.

2.2.2.3. Modern Standard Chinese

The Chinese classifier system differs from that of the majority of the other classifier languages, since there is no classifier for marking humans specifi-cally. The normal classifier for humans is the general classifier ge, which basically can replace all the other classifiers (for further discussion, cf. §2.4.2). The use of the classifier wei for humans of higher rank is limited to formal contexts. In informal contexts, we just find the general classifier:

(14) a. sän ge jiäoshöu three CL professor 'three professors [unmarked]'

b. sän wei jiäoshöu three CL:HON professor 'three professors [polite]'

The following list of Chinese classifiers, which is a selection out of the 22 most widely used classifiers (Erbaugh 1986: 405), also shows that the ani-mate/inanimate distinction and the distinction based on shape is always ex-pressed by a variety of classifiers. In this sense, the semantic hierarchy pre-sented in (5) is less straightforward in Chinese than in languages such as Thai. As I shall try to show in §3.3, this may be connected to the item-ori-ented process of development which generated Chinese classifiers.

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 133

(15) [+animal] zh.ll9 niao 'bird', yäzi 'duck', göu 'dog', hudie 'butter-fly', cängying 'fly', erduo 'ear', shöu 'hand'; shöubiäo 'wrist-watch', xinxiäng ' letter-box', chudn 'ship', etc.

köu zhü 'pig'Jmg 'well ' , shulgäng 'water vat ' , etc. 'mouth' tou niu. 'ox', zhü 'pig', ydng 'sheep', Ιύ 'donkey', 'head' xiäng 'elephant'; suän 'garlic'. wei 'tail' yu 'fish'.

[+plant] ke shu. 'tree', cäo 'grass', bäicäi 'Chinese cabbage', xiangrikui 'sun-flower' , etc.

shu yishu tdo 'a peach tree', ' tree' yi shu mei' a plum tree', etc.

[+dimensional] [1-dim.] tido xian 'thread, string, wire', sheng 'rope', xiängyän

'branch' 'cigarette', he/jiäng 'river', däolü ' w a y \ j i e 'street', hong 'rainbow', bizi 'nose', shetou 'tongue', gebei 'arm', etc. (for more on the seman-tics of this classifier cf. Tai 1990)

zhl2 mdobi 'writing brush', jian 'arrow', di 'bamboo 'twig' flute', kuäizi 'chopsticks', qiäng 'rifle, gun';

jündui 'army', ge 'song'. [2-dim.] zhäng zhi 'paper', huär 'picture', chepiäo 'ticket',

'spread, zhuözi 'table', chudng 'bed' , liän ' face ' , li stretch' ' plough', etc.

[3-dim.] kuäi shttou 'stone', xlguä 'watermelon', gutou 'bone', 'piece, lump, chunk',etc.

[instru- bä chuzi 'hoe', däo 'knife ' ,fütou 'axe, hatchet', kuäizi ments] 'hold, 'chopsticks', shüzi 'comb', yäoshi 'key', etc.

grasp' jian ylfu 'clothing', day! 'overcoat, topcoat', paozi

'gown, robe'; döngxi 'thing', huowii 'goods, mer-chandise', wüqi 'weapon, arms', liwu 'present', etc.

[places, suö fangzi 'room, house', loufang 'building of two or buildings] 'place' more storeys', xuexido 'school', däxue 'university',

ylyuän 'hospital', siyuän 'temple, monastery', yinhdng 'bank'.

jiän fdngzi 'room, house', woshi 'bedroom', 'interval' dating 'hall', etc.

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

134 Walter Bisang

[Special classifiers] ben 'root' for books, shöu 'head' for poems, ßng 'seal' for letters, liäng for vehicles, sou for ships, pi for horses and for cloth (a bolt of cloth), zun for statues, temples, pieces of artillery, etc.

As we can see from the above list, there is no classifier for animates in gen-eral. There are only classifiers such as zhll, kou, tou, and wei for animals and a special set of classifiers for plants/trees such as ke and shit. Furthermore, we find a special set of classifiers for places and buildings ('location' according to Allan 1977), for instruments, i.e. function in terms of Adams and Conklin (1973). Finally, there is a set of special classifiers denoting particular nouns.

The Chinese system of classifiers can be generally characterized by two different types of flexibility. For describing this flexiblity, I distinguish nor-mal classifiers from general classifiers. Normal classifiers such as those pre-sented in the above list individualize a noun on the basis of semantic criteria. General classifiers can replace normal classifiers. They individualize a large number, maybe even all the count nouns of a language. The most common general classifier in Chinese is ge. It can replace almost any classifier.

The first type of variety I would like to present is variety among normal classifiers. Some nouns can occur with two or more classifiers without any significant difference in meaning (some reasons for this phenomenon are given at the end of §2.2.3).10 The criteria according to which a noun can occur with only one or with more than one normal classifier seem to be rather arbitrary. Thus, the noun ydng 'sheep' can take the classifiers zhll, tou, or tido, whereas göu 'dog' can occur only with zhJl or with the stylistically less appropriate classifier tou. The noun mä 'horse' has only one normal classifier, i.e. pi, whereas luozi 'mule' can take either pi or tou. The noun chudn 'ship' can take the classifiers tido, zhll, or sou. Facts like these are already mentioned by Herrfurth (1964) and they are described in more detail by Coyaud (1973). Coyaud (1973: 63-72) presents a survey of overlapping domains of classifi-ers. According to his data, the classificatory domain of zhll extends over the classificatory domains of no less than eighteen other classifiers (e.g. tido, bd, etc.), the classificatory domain of tido overlaps with no less than eleven other classifiers (e.g. zhll, wei, etc.). A short look at the table in Lü Shuxiang (1980: 631-636) or at the Chinese dictionary of classifiers by Guo (1987) just con-firms these findings. As we learn from a look into the history of Chinese classifiers, this flexibility is not new. The boundaries between the individual noun classes have never been absolutely strict (§3.1). Thus, this picture may

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 135

simply be a reflection of the way the Chinese classifier gradually came into existence.

The second type of variety refers to the selection of normal vs. general classifier. If one reads teaching grammars of Chinese or if one asks native speakers of Chinese about their use of classifiers, one gets the impression that it is fix and invariant, i.e. that each noun takes its one particular classifier. Nouns denoting humans take ge, and nouns denoting other entities take a normal classifier. As is shown by Erbaugh (1986), a look at what is really spoken presents quite a different picture:

Special classifiers [i.e. normal classifiers in my terminology, W.B.] are quite rare; the general classifier is hundreds of times more frequent.... The general classifier is often used where a special classifier is reported as obligatory, even by highly educated, conservative teachers of classical Chinese. (Erbaugh 1986: 406).

There are several factors determining the selection of normal vs. general clas-sifier. The semantic factor will be described below, the factor of discourse will be discussed in §2.4.2. The semantic explanation I shall briefly discuss is developed by Ahrens (1995). She presents a prototype model of categoriza-tion, combined with a competition model. The prototype model allows her to make two generalizations.

The first generalization is that the farther an object deviates from the proto-type of the classifier that categorizes it, the more likely it is to neutralize. (Ahrens 1995: 216)

... the second (and related) finding in the data is that if the noun exhibits the exact same properties as the classifier with which it is supposed to occur, the classifier is likely to remain. (Ahrens 1995: 223)

The prototypical classifier zhäng is characterized by four features, i.e. [flat], [rectangular], [two-dimensional], [horizontal]. A noun like 'paper' satisfies all four of these features. Nouns like 'bed' and 'table' do not satisfy the fea-ture [two-dimensional]. Finally, the noun 'sofa' only satisfies the feature [rec-tangular]. According to the above generalizations, the classifier zhäng is least likely to be neutralized by the general classifier ge if it is used with the noun 'paper', since this noun shares all the prototypical features with its classifier. In fact, there is no instance of neutralization for 'paper' in the data presented by Ahrens (1995: 218 -220). The nouns 'bed' and 'table' are more likely to be

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

136 Walter Bisang

neutralized, because they deviate in one feature from the prototype. The noun with which neutralization is most likely to take place is 'sofa', which deviates in three features. The percentage of neutralization is 30.6% for 'bed', 45.9% for 'table' and 62.0% for 'sofa'. The difference between the percentages be-tween 'bed' and 'table' is due to the fact that "'table' is even less 'two-dimen-sional like' than beds, since beds are lower to the ground than tables, and usually have no legs" (Ahrens 1995: 218).

In addition to the prototype model, Ahrens postulates the following prin-ciple of competition:

... two or more classifiers competing for occurrence with a noun give the neu-tral classifier a greater opportunity of co-occurring with the same noun. (Ahrens 1995: 224-225)

Since there are three classifiers that can occur with 'car' (bü, Hang, tdi), the likelihood of their neutralization by the general classifier is relatively high.

2.2.2.4. Japanese

The most frequent classifiers of Japanese confirm the findings on the seman-tic criteria of Adams and Conklin (1973) (cf. Sanches 1977, Downing 1986). Furthermore, we can see from the following list, that there are some special classifiers in Japanese. The classifier -tsu for inanimates tends to become a general classifier.

(16)[+ human]:

[+ animate]: [- animate]: [shape]:

[shape], [size]: [function]:

-nin (for human beings), -mei (formal for people of higher ranks) -hiki (for animals) -tsu (for inanimates in general) -hon (for one-dimensional objects) -mai (for two-dimensional objects) -ko (for small, three-dimensional objects) -dai (for vehicles and machines of big size)11

[special classifiers]: -satsu (for books) -tsuu (for letters and documents) -kyaku (for objects with four legs such as chairs

and tables) -mon (for cannons)

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 137

The six classifiers which are first managed by children seem to be those which mirror the above semantic hierarchy (Sanches 1977: 59).

(17a) [+animate]: -nin [+ human], -hiki [+ animal] [- animate], [shape]:-hon [one-dimensional], -mai [two-dimensional] [- animate], [size]: -dai [big, instruments], -ko [three-dimensional,

small]

These classifiers are learned in the following sequence (17b) (Clancy 1985: 414—415):

(17b) -nin >-hiki > -ko >-mai/-hon > -dai

Apart from these and a few other classifiers which are applicable to a wide range of nouns, there is a considerable number of special classifiers which children learn in relation to the item to be individualized. Thus, the former group of classifiers is learned by semantic rules, whereas the latter is learned by a listing procedure. Due to the change in material and social culture in the last hundred years in Japan many of the items classified by listing no longer exist. Of course, the loss of these items also implies the loss of their classifier. For that reason, Sanches conludes that "we can only expect the system to atrophy" (Sanches 1977: 61). In this context, it comes as no surprise that younger people tend to have significantly smaller repertoires of classifiers than older people (Sanches 1977: 55).

2.2.2.5. Vietnamese

Since Emeneau's (1951) Studies in Vietnamese, this language is supposed to have a wealth of classifiers and it seems to make a distinction between nouns that can take a classifier and 'non-classified' nouns that can never take a clas-sifier (Emeneau 1951: 85). As is shown by Löbel (1996, to appear), both of these statements are subject to modifications. The first statement will be criti-cally discussed in this section, the second statement will be rejected in §3.2.2.

Löbel (1996:170-173, to appear) developed a test to establish a core group of Vietnamese classifiers. This test is based on postnominal attributes of the structure 'numeral+(classifier)+noun' as presented in the following example:

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

138 Walter Bisang

(18) Vietnamese (Löbel 1996: 171) a. nhä [ba phöng]

house NUM:three N:room 'a three room house'

b. nhä [vo'i ba cai phöng] house with three CL room 'a house with three rooms'

In example (18a) the noun phöng 'room' is not individualized. What matters is the fact that the house has three rooms, the rooms as such are of no interest. Thus, nhä ba phöng has to be translated into English as some kind of com-pound. In (18b), however, phöng is individualized, the three rooms with their individual characteristics are relevant. The classifier used in a construction like (18b) refers exclusively to the cognitive structure of the noun it individu-alizes. Thus, the construction in (18b) can be taken as a test case to establish the core group of the classifiers denoting inherent properties of the noun.12

Löbel's elicitation of a large corpus of classifiers and nouns revealed that there are only ten classifiers allowing the contrast presented in (18a,b). These core classifiers can be described fairly well by the properties of [± animate]/ [± human], [shape], and some more special characteristics such as [plantltree], [volumes], and [paper] (for [+human] see below):

(19) Core classifiers of Vietnamese (Löbel 1996: 172) cai 'a nonliving thing' cay 'a tree or plant, a stickshaped or plantlike object' chiec 'individual item of floating objects, vehicles' con 'a living thing that is not human' hon 'a stone or stonelike object' qud 'a fruit, a round, globular object' quyen 'a volume' so'i 'a hair, thread, cord, etc.' tam 'a flat piece of material' to' 'a sheet of paper, document'

The concept of inherent vs. temporary classification as discussed by Berlin and Romney (1964) and by Serzisko (1980, 1982) is based on the question whether a classifier only refers to properties inherent to the concept of a given noun or whether it provides new, i.e. temporary, information not present in

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 139

that noun. In its temporary function, classifiers can be compared to adjectivals in English (cf. Berlin and Romney 1964:79). The ten Vietnamese core classi-fiers of the above list particularly refer to inherent properties of the noun. In this sense, they actualize the semantic boundaries which already belong to the concept of a given noun as pointed out in §2.1. This kind of actualization also takes place with more specific classifiers. In such a case, the noun plus its modifier(s) are taken as one individual concept whose boundary is actual-ized. In the following example from Löbel (to appear), the classifier cdi is used with ηύϊ 'mountain' (20a). In a simple context of enumeration, where the noun is not specified by any additional information, another classifier such as qud is impossible (20b). However, if a mountain is specified by a further modifier as in (20c), its cognitive appearance and its boundaries are different and it must take another classifier, which is qud in this example. If the classifier cdi would be used in the context of (20c), it would be inter-preted automatically as 'extra cai' (Emeneau 1951: 97), a special marker for discourse phenomena such as contrast.

(20) Vietnamese (Löbel to appear: 30) a. hai (cai) ηύϊ

two CL mountain 'two mountains'

b. *hai qud ηύϊ two CL mountain

c. hai qud nui ό gän biä rung two CL mountain to_be_atnear border forest 'the two mountains near the forest'

Since what is individualized in (20a) and (20c) always is the concept as a whole, be it expressed by a simple noun or a noun-modifier construction, classification is based on inherent properties in all the above Vietnamese ex-amples. However, this does not mean that there is no temporary classifica-tion. The classificational system referring to humans seems to be a good ex-ample of temporary classification in Vietnamese. In this language, the classi-fication of humans is always linked to the marking of social status. Thus, the classifier yields additional, temporary information to a noun denoting hu-mans:13

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

140 Walter Bisang

(21) Vietnamese (Löbel 1996: 174)

a. hai ngu 'o 'i öong nghiep two CL:man colleague 'two colleagues (neutral)'

b. hai ban öong nghiep two CL: friend colleague 'two colleagues (good friends)'

c. hai vi öong nghiep14

two CL:rank colleague 'two colleagues (very respectful)'

d. hai öng öong nghiep two CL:grandfather colleague 'two colleagues (respectful for two colleagues of higher age)'

The classifier ngu'o'i 'man' (21a) seems to be the semantically most neutral classifier. However, it is not unproblematic to describe its classificational meaning, because what it tacitly refers to in the unmarked case in an androcentric society are men, not women. For that reason, it cannot be quite adequate to label ngu'o'i as [+ human].

From the example of Vietnamese as presented by Löbel we can learn that one should look at the distinction of inherent vs. temporary more carefully. Some classifiers which look like temporary classifiers at first glance may turn out to be inherent. In this sense, I think one should recheck the famous example of the eight classifiers for 'river' in Burmese as presented by Becker (1975: 113). At least many of these classifiers simply look at the concept of 'river' from a different point of view. Therefore, they refer to different prop-erties inherent to that particular concept which the speaker wants to commu-nicate to the hearer. Thus, the speaker will have to make a choice to which of these properties he wants to refer to in a given situation.

(22) Burmese (Becker 1975: 113)

myi? to ya? 'river one place' (e.g., destination for a picnic) myi? to tan 'river one line' (e.g., on a map) myi? to hmwa 'river one section' (e.g., a fishing area)

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 141

myi? to 'sin 'river one distant arc' (e.g., a path to the sea) myi? to Owe 'river one connection' (e.g., tying two villages) myi? to 'pa 'river one sacred object' (e.g., in mythology) myi? to khu' 'river one conceptual unit' (e.g., in a discussion of rivers

in general) myi? to myi? 'river one river' (the unmarked case)

2.2.3. Cognitive explanations for the semantic hierarchies underlying classi-fier systems

In §2.1, I have tried to explain that only classifiers can produce a system based on inherent properties of a given object, because only they are based on actualizing individualization, i.e. on selecting structures inherent to a given entity. The next question to be answered is why classifier systems cross-lin-guistically tend to follow a hierarchy whose primary distinctions are [± ani-mate]/^ human] and [shape] as presented in §2.2.1. The first distinction based on animacy or humanness can basically be explained along the same line of argumentation as discussed with the animacy hierarchy. Objects are classi-fied according to their categorial closeness in kind to the speaker or the hearer. This explanation is further corroborated by the fact that numeral-classifier constructions can always be used anaphorically, i.e. in the function of refer-ence tracking (§2.4.1). In this sense, the self-other distinction suggested for Burmese by Becker (1975) may be of more general importance for classifier systems. Less straightforward is the explanation of the primary status of the second distinction, i.e. shape. I agree with Croft (1994: 163-164) that the internal structure and the spatial distribution of objects is crucial for enu-meration. One can typically identify objects by their spatial organization. But this cannot be the whole story. Another explanation may be given by Marr's (1982) theory of vision.15 According to this theory, low-level visual identifi-cation is characterized by four stages based on dimensionality. At the first stage, peaks and valleys of light intensities are calculated, from which 'edges' are constructed which, in turn, lead into a 'one-dimensional sketch'. At a second stage, whose details do not seem to be understood clearly, one-dimen-sional information is rearranged into two-dimensional surfaces. At the third stage, surfaces are further developed into curved surfaces, i.e. they are con-verted into round, curved, cylinderlike entities. This is what Marr calls the 'two and one-half—dimensional sketch'. Finally, the curved surfaces are trans-formed into a 'three-dimensional sketch'. If dimensionality turns out to be

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

142 Walter Bisang

such a basic organizing pattern of vision this may possibly explain why it is primary for classifier systems. Furthermore, we have seen that the visual com-putation of entities develops in a sequence of stages in which one-dimensionality comes first. This may explain why we relatively often find most secondary distinctions with classifiers for one-dimensionality. Further-more, it may explain, why one-dimensional classifiers are acquired with great-est difficulties by Thai children, since this particular category of classifiers is not only very highly developed but also highly fossilized, i.e. arbitrary and conventionalized (§2.2.2.1). And finally, it may explain, why one-dimensional classifiers are established before two-dimensional classifiers in the diachrony of Chinese (§3.1) and why there are only classifiers denoting one- and two-dimensionality in the Proto-Tai reconstructed by DeLancey (1986; §2.2.2.1).

In the process of actualizing individualization there may be a competition between different properties. The choice of one of these properties depends on the perspective from which the speaker looks at the object to be individu-alized and from the discourse pragmatic situation in which he wants to present it to the hearer. Thus, a noun like 'snake' can be looked at from the perspec-tive of its one-dimensionality as a long and flexible object, but it can also be looked at from the perspective of its animacy. The Chinese noun for 'snake' can either be classified by tido (one-dimensional) or by zhll (animal). Ac-cording to the statistics presented by Ahrens (1994: 226-227), the classifier tido (one-dimensional) is replaced - or 'neutralized' in her terms - in 19 out of 35 cases by zhll (animals). Many of the overlapping domains discussed above (§2.2.2.3) can be explained by the fact that a given noun can be looked at from different perspectives. However, the freedom of the speaker to choose her/his perspective differs cross-linguistically. In a language like Thai, there are strict lexical rules assigning one particular classifier to a given noun. Some of the very few exceptions are presented in §2.2.2.1. Other languages allow more freedom, they even allow the classifier to introduce some additional information not present in the noun in the case of temporary classification. As I shall discuss in §2.4.2, in some languages such as Chinese, the choice of perspective is also governed by discourse pragmatics.

2.3. Classification and individualization

The first section sketches the cognitive relation between individualization and classification (§2.3.1). In the second section (§2.3.2), I shall present lan-guages in which the classifier is limited to these two core functions. The lan-

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 143

guages to be discussed are Chinese (§2.3.2.1) and Vietnamese (§2.3.2.2). Other languages not presented here are Japanese, Khmer, and Bahasa Indonesian.

2.3.1. Classification, individualization and counting: sketch of a cognitive explanation

In mathematical psychology, what is called counting in the present paper is usually subsumed under the term of measuring (Suppes and Zinnes 1963). Measuring is understood in a very broad sense as the mapping of objects onto numbers. In the context of numeric quantification, the numeral assigned to the items to be quantified indicates their number. This is illustrated by the following figure, which is inspired by Wiese's (1995: 186) stimulating paper on semantic and conceptual structures of numeral constructions in German:

Figure 3. Schema of numeric quantification

As pointed out by Wiese (1995: 219), the set a of elements to be counted consists of two components, a notional aspect, which defines the element condition for α (e.g. 'cattle'), and an individualizing aspect (e.g. 'piece'), which makes the elements of set α countable.

From the above approach of mathematical psychology, we can distinguish an operation of individualization and an operation of mapping objects onto the set of natural numbers. Furthermore, there is the notional aspect which defines what elements are to be counted.

In my view, this approach does not cover all the cognitive factors which are relevant for counting. To illustrate this, I shall start my argumentation from the following quotation by Seiler (1986: 94-95), which is based on Piaget (1970/1973:46):

4 *

items to be numerically quantified sequence of natural numbers

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

144 Walter Bisang

... J. Piaget ..., on the basis of psychological experiments, suggests that the concept of number is based on a synthesis of two different operations: One is certainly classification. It has to do with class inclusion in that 2 is included in 3, 3 in 4, etc. The other operation is ordering: Elements are ordered in space, one after the other, or they are counted consecutively, i.e. ordered in time. Ordering must supplement classification, for otherwise it would be impossi-ble to distinguish the elements from one another. (Seiler 1986: 95)

What is called operation of ordering by Seiler (Piaget) can be compared to the operation of mapping objects onto the set of natural numbers. However, the operation of ordering is more concrete, since it explicitly refers to ar-rangement in space or in time. Furthermore, Seiler (Piaget) discusses the op-eration of classification. This operation may be compared to the notional as-pect as discussed above. It is certainly true that this notional aspect defines what elements are to be counted. The question is, how exactly do we find out whether an element qualifies as a member of the set of items to be counted. If we look at figure III again, we have to find the criteria which lead us to distin-guish three different sets to be counted separately in the above perceptional environment, i.e. the set of the squares ( • ) , the set of the stars and the set of the crosses (+). The approach of mathematical psychology offers no solu-tion to this problem, because it neglects the operation of classification and its two related operations of identification and individualization ( §1.2). Croft (1994), for whom counting only consists of identification and individualiza-tion, presents the following scenario:

One must identify multiple units of the same kind in order to count them. The two cognitive processes involve[d] are individualization of units, and identifi-cation of them as being of the same kind (Croft 1994: 162).

Without the classificational criteria for identifying an element as belonging to the set of items to be counted it would be impossible to decide whether to count that element or not. Once a given item is established as relevant for counting, it will be individualized. There is general agreement on the rel-evance of individualization for counting.

In the above discussion, I have presented four operations involved in count-ing, i.e. classification, identification, individualization and ordering. The first three operations correspond to the core functions of classifiers. This may ex-plain why classifiers are compulsory in the majority of classifier languages (but not with all nouns in Vietnamese, cf. §.2.2.2.5, §2.3.2.2). As pointed out above, the fourth operation of ordering is based on the arrangement of items

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 145

in space and in time. The spatial aspect of ordering may be relevant for the question why dimensionality is a primary criterion of classification in classi-fier languages. The fact that spatial ordering is externally applied to indi-vidual items in such an important context as that of counting might as well support its relevance as an inherent property for classifying objects.

2.3.2. Some language data

2.3.2.1. Chinese

In Chinese, the situation is quite straightforward. Classifiers are used with numerals and with demonstratives. In the above example (1), I have illus-trated the use of classifiers in the context of counting. In the following exam-ple, the classifier is used with a demonstrative (23a) and with both a demon-strative and a numeral (23b):

(23) Modern Standard Chinese a. zhei feng xin

this CL letter 'this letter'

b. zhei sdn feng xin this three CL letter 'these three letters'

2.3.2.2. Vietnamese

This section will deal with syntactic referentialization and with the functional limitation of Vietnamese classifiers to individualization and classification.

As pointed out in §1.2, Vietnamese classifiers are not obligatory in the context of counting. Their main function is to create syntactic referentialization (Löbel to appear: 27-29). The above example (18a) with nominal attributes of the structure 'numeral-(classifier)-noun' shows that classifiers are not com-pulsory in the context of counting. Another example is the following:

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

146 Walter Bisang

(24) Vietnamese (Löbel to appear: 27-28) a. Trong nhä hdt kia co mu 'ο 'i bdy ghe.

in cinema this have ten seven chair 'There are seventeen chairs in this cinema.' lit.: 'This cinema is seventeen-chaired.'

b. Trong nhä hdt kia co 17 cdi ghe läm bang cäy tot. in cinema this have 17 CL chair make out_of wood good 'There are seventeen wooden chairs in this cinema.' lit.: There are seventeen chairs in this cinema which are made of pre-cious wood.'

In both examples, (24) and the above (18), the classifier occurs if the noun has to be syntactically referentialized, i.e. if it has to be individualized for being further modified. There is, however, one instance, where classifiers seem to be obligatory in the context of counting. This is with nouns denoting humans.

The evidence presented by Löbel clearly shows that the functional range of Vietnamese classifiers is limited to individualization and classification. Constructions of the type 'classifier-noun' are systematically ambiguous with regard to definiteness vs. indefiniteness.

(25) Vietnamese (Bisang 1996: 541, discussed by Löbel 1996: 228-229) toi mua quä cam I buy CL orange Ί buy the orange./1 buy an orange.'

The interpretation of the above example corrects my statement (Bisang 1993, 1996) that Vietnamese classifiers are marking definiteness. Löbel's findings clearly argue against such a statement. As is shown by Löbel, the above con-struction can be analysed in two ways:

(26) Vietnamese a. töι mua (mot) quä cam.

I buy one CL orange Ί buy an orange.'

b. toi mua [quä cam]. I buy fruit orange Ί buy the orange.'

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 147

In (26a), we have a normal classifier construction with the numeral 'one' followed by a classifier. Such a construction can be used to mark indefinite-ness. The numeral one can also be omitted, since the individualizing function of the classifier seems to imply it.16 In this case, the construction in its written form looks identical to (26b), which can be interpreted as definite. The 'clas-sifier-noun' construction of (26b) is a special type of a nominal compound which is used in discourse with aforementioned entities which cannot be re-ferred to by pronouns. A change of frame is an example of such a situation in which the nominal compound form is used (Löbel to appear: 37). The differ-ence between the two constructions cannot be seen in writing, but it can be heard because of different stress patterns. In (26a), the classifier has weak stress, in (26b) it bears the main stress (Löbel to appear: 37). From these findings, we can say that the classifier itself is not immediately involved in marking definiteness/specificity vs. indefiniteness in Vietnamese. Nominal compounds with this particular function seem to be a speciality of Vietnam-ese (§3.3).

2.4. Classifiers in discourse and referentialization

Classifiers can be relevant to discourse in several ways. Section 2.4.1 presents classifiers in their anaphoric function. In languages with one or maybe more general classifiers, one of the factors determining the selection of the normal classifier vs. the general classifier is discourse. This will be illustrated with Chinese in §2.4.2. Finally, §2.4.3 shows how reference marking can be a secondary or even a primary function of classifiers.

2.4.1. The anaphoric function of classifier constituents

In classifier languages, the noun can be omitted from the classifier construc-tion if it is known from what is said before. In such a case, the classifier anaphorically refers to the noun. In most, maybe even in all classifier languages, the classifier can be com-bined with numerals and/or demonstratives as is shown in the following ex-amples from Chinese (27a) and Vietnamese (28b):

(27) a. zhei shi shä. Zhängsän mäi-le yi ben, wo mäi-le zhei liäng ben. this is book Zhangsan buy-PFV one CL I buy-PFV this two CL

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

148 Walter Bisang

'These are books. Zhangsan bought one (of them), I bought these two (of them).'

Λ S Ζ Α. Ζ b. öäy la säch. Lan mua mot quyen, toi mua hai quyen näy.

this be book Lan buy one CL I buy two CL this 'These are books. Lan bought one (of them), I bought these two (of them).'

In most languages, the classifier does not occur alone in its anaphoric func-tion. Although there is more research to be done, Cantonese seems to be an exception:17

(28) ni1 gan' fang4 tai2 guei3, ting'yed6 wun6 guo3 gan1

this CL room too expensive tomorrow exchange ΤΑΜ CL deg' m4 deg' a3? can NEG can PART 'This room is too expensive, can [we] change it [for another room]?'

In Thai, the classifier can also occur with stative verbs (adjectives) (§2.4.3.1). For that reason, the head noun can also be omitted in 'noun-classifier-adjec-tive' constructions:

(29) Thai (Conklin 1981: 76) mäakruk sämräp nyq mii mäa sii tua, sii-dam sooq tua, chess Q:set one have horse four CL colour-black two CL sii-deeq sooq tua. colour-white two CL Ά chess set has four knights, two black and two red.'

In languages which also make use of classifiers in possessive constructions such as Hmong (30) or in relative constructions such as Cantonese (31), the classifier may also serve as a substitute of the head noun. Even though rela-tive clauses are not marked with classifiers in Hmong, classifiers can take a relative clause introduced by the relative marker uas (32):

(30) Hmong (Mottin 1978: 57) kuv lub tsev txawv koj lub. I CL house be different you CL 'My house is different from yours.'

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 149

(31) Cantonese (Matthews-Yip 1994:112) gaau leih täahn kähm go go? teach you play piano that CL 'The one who teaches you piano?' [from a Cantonese film]

(32) Hmong (Bertrais-Charrier 1979: 524) tus [uas puag tag kis no tuaj] nws puas yuav mustsev? CL REL from afar this_morning come he QUEST FUT go home 'Will the one who came this morning go back home?'

In the Miao language of Weining, stative verbs and relative clauses can occur with the classifier alone (for the different forms of the classifier, cf.§ 2.4.3.3). Possessors, however, are followed by the particle bie53, which is used espe-cially for headless possessive constructions (33c):

(33) Weining Miao (Wang Fushi 1957: 93; 1972: 138-139) a. Classifier + stative verb

fae55 ku11 sie55

CL:DEF ATTR big 'the big one [i.e. tree;/au/5 being the classifier for plants and trees]'

b. Classifier + Relative clause fa55 [ku" m'a35 a33ndlau35 ntsa55] v'ae35ku" i55 fa35

CL:DEF ATTR have leaf green that be one CL:INDEF hi" pau". cypress 'That one having green leaves is a cypress.'

c. Possessor + bie53

ku55 a5531 m'au" bie53

I elder brother POSS 'the one of my elder brother'

2.4.2. General classifiers and their discourse function

As we have seen in §2.2.2.3 in our discussion of Ahrens (1995), the selection of normal vs. general classifier depends on semantics. However, semantics are only one of many different interacting factors. Another very important

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

150 Walter Bisang

factor is discourse. New, indefinite entities are most likely to be marked by the normal classifier:

[B]oth children and adults used special classifiers to specify 1) the first men-tion of a 2) non-present object which was 3) unfamiliar or unclear to the hearer, especially in reference to a 4) new creation or as part of a 5) narrative, 6) pretend play scheme, or a 7) request. Special classifiers are particularly likely where more than one of these triggers is present. (Erbaugh 1986: 425)

Style and age are further factors influencing the selection of classifiers. Erbaugh presents the following summary of factors determining classifier use:

Level of formality, discourse type, especially narrative and poetry; length of speech turn, presence of the referent, familiarity of the referent, and age of the hearer are all critical, synergistic, and variable determinants of special classi-fier use. (Erbaugh 1986: 413)

Chinese is not the only language having the above phenomenon. Japanese (Downing 1986), Yao (Caron 1987: 160), and at least some varieties of Malay (Hopper 1986) are other languages, in which the selection of normal vs. gen-eral classifier or no-classifier vs. classifier seems to be determined by more or less the same range of factors as presented here.

2.4.3. Referentialization

In some languages, classifiers can also display the function of reference. In Thai, the referential function of the classifier is limited to one particular type of NP (§2.4.3.1), whereas it is more general in Hmong (§2.4.3.2). In both languages, however, the referential function is secondary to individualiza-tion. In the Miao language spoken in Weining (Guizhou Province, China), referentialization seems to be a primary function of the classifier (§2.4.3.3).

2.4.3.1. Thai

The referential character of the classifier in Thai is only a secondary phenom-enon limited to the 'noun-classifier-adjective' construction, in which the ad-jective is a Stative verb in attributive function. The reasons for the specific or definite interpretation of the classifier in this context is excellently described by Hundius — Kölver (1983), who refer to examples such as (34) and (35):

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 151

(34) Thai (Hundius — Kölver 1983: 172) a. rom ntt

umbrella DEM 'this/these umbrella(s)'

b. rom khan nil umbrella CL DEM 'this umbrella'

(35) a. röm sii-khiaw umbrella colour-green 'green umbrella'

b. rom khan sii-khiaw umbrella CL colour-green 'the green umbrella(s)'

As a further difference between our minimal pairs (34) and (35) we noted referential connotations: classifiers in the context of adjectives signal definite reference to specific objects or they stress contrastive properties as conveyed by the co-occurring adjectives. In order to account for this fact, we must revert to the pronounced difference between demonstratives and adjectives. For demonstratives by themselves always serve the very function of referential identification, thus there is no need to employ classifiers in any referential function in phrases like (34b). Adjectives, on the contrary, by themselves have no referential purposes at all, as (35a) shows. Yet, they are frequently needed for referential purposes, i.e. for identifying subsets of objects in terms of contrastive properties they may happen to have: specific green umbrellas vs. red ones etc. This is where clas-sifiers are made use of for referential purposes: since adjectives are referentially neutral, their combination with a classifier may serve to compensate this defi-ciency and thus bring about a definite interpretation of the phrase. This sec-ondary function is readily compatible with the one noted so far: classifiers, as we found, always relate to the quantification of objects. Therefore, in the ab-sence of other referential clues, they may be utilized to referentially identify these objects. [I have adapted the number of the examples to my numeration, W.B.] (Hundius — Kölver 1983: 176-177)

Thai classifiers in 'noun-adjective-classifier' constructions cannot only have referential function, they can also mark contrast. The next example (Hundius — Kölver 1983: 173) is to demonstrate this:

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

152 Walter Bisang

(36) röm khanl sii-khiaw khan2 nii umbrella CL green CL this 'this green umbrella'

The second classifier (khan2) marks the NP for singular. This is due to the fact that in Thai, the numeral one is implied in constructions of the type 'noun-classifier-demonstrative' as illustrated in example (34b). Hundius — Kölver (1983: 174) convincingly analyse constructions of this type as 'noun-[0-clas-sifier]-demonstrative', in which the zero of the classifier constituent is auto-matically interpreted as the numeral 'one'. The first classifier (khan j) occur-ring with the adjective is used to mark contrast. Thus, the NP in (36) refers to green umbrellas in contrast to e.g. yellow umbrellas which may be part of the speech act situation. If a classifier occurs with more than one adjective, all these adjectives are understood in a contrastive sense (Hundius — Kölver 1983: 173):

(37) rdm khan, sii-khiaw khan2 yay khan3 nii umbrella CL green CL big CL this 'this big green umbrella'.

2.4.3.2. Hmong

Hmong can have 'classifier-noun' constructions. Unlike the 'classifier-noun' constructions of Vietnamese (§2.3.2.2) and Cantonese (§2.5.2), 'classifier-noun' constructions in Hmong allow only one referential interpretation, i.e. the interpretation of definite or specific. Indefiniteness is linked to the nu-meral ib 'one' which is obligatory in this context. Nevertheless, the referen-tial function of classifiers remains a secondary phenomenon as in Thai. This can be seen from the context of counting, where the classifier does not give any information on reference:

(38) peb tug dev three CL dog a. 'three dogs (indefinite)' b. 'the three dogs (definite)'

Since I have discussed the referential function of Hmong classifiers else-where (Bisang 1993: 24-29), I shall only present an outline of the way classi-fiers are used in texts. A new entity is usually introduced by the sequence 'ib

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 153

'one'-classifier-noun' which sometimes occurs with the verb muaj 'there is' at the beginning of a sentence. If it is mentioned next, it can be taken up by the sequence 'classifier-noun-demonstrative' or 'classifier-noun'. In the fur-ther development of the discourse, the main characters of the story who are assumed to be known can even remain unmarked. The following example, which concludes this section, illustrates some of the above sequences in dis-course. Since it is quite long, I shall not present the whole text between the relevant classifier constructions:

(39) Hmong (Bisang 1993: 26-27, also cf. Bisang 1988: 59) Tus ntawd tau mus thawj rau ib tug poj ntsuag lawm. ... CL DEM ΤΑΜ go be_reborn in one CL widow ΤΑΜ Koj mus nrhiav tus poj ntsuag ntawd, ces... you go find CL widow DEM then Huab Tais nrhiav tau tus poj ntsuag muaj ib plab emperor find get CL widow have one Q: stomach me nyuam mas txawm coj poj ntsuag ntawd los. son then then take widow DEM come 'He went into [the womb] of a widow in order to be reborn. . . . Go back and find the widow, then. . . When the emperor was able to find the widow with her child, he took the widow with him.'

2.4.3.3. The Miao language of Weining

In the Miao language of Weining, the classifier seems to display referentiality as its primary function indeed (Bisang 1993: 46-47, Bisang 1996; with data from Wang Fushi 1957/1972, Wang Deguang 1986, 1987). Weining classifi-ers (and quantifiers) occur in five different forms which are presented in the figure below with the classifier lui55(cf. Wang Fushi 1957/1972; on this clas-sifier also cf. footnote 6). The basic form of a classifier consists of an initial consonant (C), a vowel (V), and a tone (T). Form I is characterized by a change of the vowel quality into [a] or [ae] with the tone being the same as in the basic form. In Form II, we can observe two changes: first the vowel change of Form I followed by a change of the tone into a high rising tone (35):

Basic form (CVT) lui55

Form I (CAET) lae55 (CAT) la55

Form II (CAE35) lae35 (CA35) la35

Figure 4. The forms of classifiers in Weining Miao

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

154 Walter Bisang

At the beginning of a text, the basic form may represent nouns of a normal size. In the further development of a discourse, however, it is used for augmentatives. The forms in [a] express diminutives while the forms in [ae] are used for nouns of normal size in the rest of the text. Form I occurs with definite nouns, while Form II occurs with indefinite nouns. With Form I clas-sifiers, referentiality is thus a primary function. The following examples will illustrate some aspects of classifier use in Weining Miao (another example is presented in Bisang 1996).

In the title of a story, the protagonists are introduced by a classifier of Form II, i.e. the form for indefinite nouns. The only exception to this rule are proper names which have no numerative. The following example from Wang Fushi (1957: 106; 1972: 161) presents the use of classifiers in titles:

(40) da35 li55a55 ndo31 dae35 a33dy33

CL: INDEF crow and CL:INDEF fox 'The crow and the fox'

The next example from 'The crow and the fox', a culturally adapted version of the well-known fable, is to illustrate how Form I and Form II classifiers interact in discourse:

(41) Weining Miao (Wang Fushi 1957: 106-107; 1972: 161-162) t'au33 i55 m'a35 i5S dae35 a33dy33 d'cey31 d'a35. time that there_is one CL:INDEF fox exit come tae33 a33dy33 qi5S la" ae55 t^'ae55 darn", i55vie33 p'i13

CL:DEF fox this also very hungry PF but he nfie55 hi33 tau33qui55 qa5$?i33 n'au35. p'i13 b'o31 ts'ae33

look_for not get food anything eat he see CL:DEF nG'ae35 ku" po55 v'ae31 ta33 li5Sa55 la55 a33nd?'au3S i55, ... meat REL at place CL:DEF crow CL:DEF mouth that 'At that time a fox came out. He too became very hungry, but he had been unable to find anything to eat. When he saw the piece of meat in the crow's mouth, ...'

In the above example (41), the fox is introduced as a new protagonist by the numeral 'one' followed by a Form II classifier in [ae], i.e. by an indefinite form denoting items of normal size. When he is mentioned again in the next sentence, he is taken up by a Form I classifier followed by the demonstrative. The indefinite noun is introduced additionally by the existential verb m 'a35.

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 155

This verb has to occur obligatorily with newly introduced items in preverbal position in Chinese; in Miao as well as in Hmong it is not always used and thus most probably borrowed from Chinese. The definite form of the classi-fier is followed by a demonstrative which does not need to occur in this situ-ation of discourse. The next classifier occurs with a piece of meat which is already mentioned in the text before. For that reason, the classifier is in Form I for definite items.18 The crow is the other protagonist of the story and he is already known. Consequently, we find the classifier lui55 in its definite form la55. With the crow, the diminutive form in [a] is used throughout the story. Even in the title we find this form. This seems to be due to the fact that the crow is much smaller than the fox.

Since there is no augmentative form of the classifier to introduce big items, the only way of marking them is by a stative verb meaning '(be) big'. In the following example, the wild boar is introduced into the story by a form II classifier for indefinite items of normal size. Then it is modified by the stative verb '(be) big' (42a). Later in the text, the wild boar, whose massive shape and terrible voice are frightening the protagonist of our folk-tale, is marked by the normal form of the classifier, i.e. by its augmentative form ((42b) and (42c)). In (42b) the forest is also marked by the normal form of the classifier. According to Wang Deguang (1987: 37), this is due to the fact that the forest is also supposed to be big as we would possibly expect it to be in a folk-tale.

(42) Weining Miao (Wang Deguang: 1986: 70) a. ßßi" nßa31 dzi" dai35 a55" lfiaui" mpa33 zau55...

s/he see PART CL:INDEF:NORM big wild_boar 'She saw a big wild boar ...'

b. dzfio35 tu33 mpa33zau55 Ifio" dzo31 vßai31 lu55

follow CL:DEF:AUG w i l d _ b o a r come arrive place/at CL ndu53 ko33a33zau33. side forest 'to follow the big wild boar into a big forest.'

c. ßßi11 hi33 mßa35 α55sie55 ku" dzßo35 tu33 mpa33zauss

s/he NEG have heart REL follow CL:DEF:AUG wild_boar a55qui55 daui". behind PF 'She didn't feel like following behind the big wild boar.'

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

156 Walter Bisang

2.5. Relationalization

2.5.1. Hmong and Weining Miao

In Hmong and in the Miao language of Weining, possession is expressed by the following construction:

(43) Possessive Construction in Hmong and Weining Miao

Possessor (Numeral) CL Possessee

(44) Hmong a. kuv lub rooj

I CL table 'my table'

b. kuv ib lub rooj I one CL table 'one of my tables'

c. nws tus txiv ntxawm tus ntxhais he CL uncle CL daughter 'the daughter of his uncle'

d. tus tub tus dev tus tw no CL boy CL dog CL tail this 'this tail of the dog of the boy' (Mottin 1978: 47)

(45) Weining Miao (Wang Fushi 1957: 88 and 95; 1972: 130 and 142) a. gi31 [o55 b'aui35

you CL flower 'your flower'

b. ku55 a55 na35 ntoey55 ku11 lie55 pi55

I two CL book ATTR red DEM 'these two red books of mine'

With inalienable possession or with constructions in which possession is suf-ficiently clear by context, the classifier can also be omitted:

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 157

(46) Hmong kuv txiv 'my father', kuv niam 'my mother', kuv npe 'my name', kuv tsev 'my house', kuv teb chaws 'my piece of land', kuv pej xeem 'my subjects [said by a king]'.

2.5.2. Cantonese

Cantonese classifiers have no referential function, but they can be used in possessive constructions. This will be illustrated in the present section. In Cantonese, 'classifier-noun' constructions are ambiguous with regard to definiteness. Thus, utterances such as the following are quite common in spo-ken colloquial Cantonese:19

(47) Cantonese ngo5 yeo5 ji1 bed', ji1 bed1 hou2 guei3 ga3. I have CL pen CL pen very expensive PART Ί have a pen, the pen is very expensive.'

As in Vietnamese, the sequence ji1 bed1 [CL-pen] can be either interpreted as indefinite, as is the case when it is mentioned first in the above example, or it can be interpreted as definite, as is the case when it is mentioned for the second time. As in Vietnamese, the numeral 'one' is facultative, since it is already implied by the classifier. However, Cantonese seems to be different from Vietnamese inasmuch as definiteness is assigned by word order. This comes as no surprise if we think of analyses of Modern Standard Chinese such as Li (1971), who shows that old information tends to occur before new information, or Li — Thompson (1975), who discuss word order and its im-plications for reference. In Cantonese, a 'classifier-noun' construction tends to be interpreted as definite (maybe specific) in the subject position, which is in front of the verb, but as indefinite in the postverbal object position (Chao 1968: 396, footnote 22). I doubt whether it is possible to analyse the 'classi-fier-noun' construction in its definite or specific use as a compound analo-gous to Vietnamese. Löbel (1996: 230), who suggests such an analysis, seems to disregard the fact that word order within NPs in Cantonese is different from Vietnamese. Since modifiers exclusively occur in front of their head noun in Cantonese, the word order to be expected in a nominal compound analoguous to the Vietnamese compound would be 'noun-classifier'. How-ever, what we find in Cantonese is the reversed order, i.e. 'classifier-noun'.

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

158 Walter Bisang

Although the situation in Cantonese is different from the situation in Viet-namese, Cantonese classifiers are still not involved in reference marking. Unlike Vietnamese classifiers, however, Cantonese classifiers can be used in possessive constructions and in relative constructions. Both constructions can occur either with the attributive marker ge , which is similar to the attributive marker de in Modern Standard Chinese, or with a classifier:

(48) Possessive construction a. Possessor ge Possessee b. Possessor CL Possessee

(49) Relative construction a. Relative Clause ge Head Noun b. Relative Clause (go 'that') CL Head Noun

The construction with the attributive marker ge "belongs to more formal regis-ters" (Matthews — Yip 1994: 108), whereas the constructions with the classi-fier are more colloquial. The following examples illustrate possessive construc-tions with a classifier (50) and relative constructions with a classifier (51):

(50) Cantonese (Matthews—Yip 1994: 107-108) a. ngoh gäan ük

I CL house 'my house'

b. louhbdan ga che boss CL car 'the boss's car'

(51) Cantonese (Matthews — Yip 1994: 111) [Ngohdeih hdi Faatgwok sihk dJ] yeh gei hou-sihk ga. we in France eat CL food quite good-eat PART 'The food we ate in France was pretty good.'

3. Two processes of classifier development and their areal distribution

Classifier systems seem to emerge from two different points of view. One point of view starts from the context of counting individual items which are of particular cultural importance. The result is a new construction containing

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 159

a new and clearly defined position for the classifier. The construction applied to the relatively few nouns successively spreads over to a wider range of nouns until all the count nouns of a given language become part of a classifier system. This is the item-oriented process of development which is predomi-nant in Chinese and in Japanese. It will be illustrated in §3.1 on the example of Chinese. The other point of view starts from a categorial system already existing in the language. This system, which is part of word formation, is based on taxonomy and meronomy. In both cases, we have two positions filled by nouns. In a process of reanalysis, one of these positions, i.e. the first position in the languages described here, is reanalysed as a classifier. This category-oriented process will be presented in §3.2 on Vietnamese, Thai and Hmong. Finally, the areal relevance of the two processes of classifier devel-opment will be discussed in §3.3.

3.1. Item-oriented development: Chinese

In the history of the Chinese language, classifiers first occur with single ob-jects of particular cultural value. For that reason, many classifiers have a very concrete basic meaning. In the course of time, however, they extend their sphere of influence from their highly specific context to the context of sys-tematic classification. Thus, the development of Chinese classifiers starts from a single noun and its concrete semantics rather than from the notion of class or category membership. This item-oriented development of classifiers seems to be still reflected in Modern Standard Chinese by the flexibility of classifier use and by the fact that the classifier system does not strictly mirror the se-mantic hierarchy based on human/animate vs. inanimate and dimensionality (§2.2.2.3).

Although the first flowering period of classifiers starts at the end of Han dynasty (220 A.D.), some potential classifiers can be postulated for the earli-est written sources, i.e. the oracle-bone inscriptions (14th—11th century B.C.). In these records, we find the nouns ren 'man', qiang 'name of a tribe' and quan 'dog' in the function of repeater classifiers (Takashima 1984/1985)20. It should be pointed out, however, that these classifiers are far from being ob-ligatory. We can just see from the texts that they sometimes occur.

(52) Chinese of oracle-bone inscriptions (Takashima 1984/1985: 235) ren shi you wu ren N:man ten and/have five CL:man 'fifteen people of the men; the men, fifteen of them'

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

160 Walter Bisang

The classifier status of the word bing occurring with horses is controversial, because it may imply a team of two horses and thus display the function of a quantifier. Similarly, the word peng is not used for counting individual items of cowries, but to measure a certain quantity of cowries. What we can see from these classifiers is that humans are the first to be counted by classifiers followed by socially particularly valuable animals such as dogs and horses, if we take bing as an item that is at least very close to a classifier.

In my subsequent sketch of the development of classifiers, I shall first discuss the change in word order from 'N-NUM-CL/Q' to 'NUM-CL/Q-N'. After word order, I shall look at the semantics of the classifiers. At the end of this section, I shall shortly compare the historical development of classifiers with the acquisition of classifiers by children as described by Erbaugh (1986).

In the above example (52), word order is 'N-NUM-CL/Q'. An alternative to this word order is 'NUM-N', in which we never find a numerative. Ac-cording to Takashima (1984/1985: 246-255), the new, i.e. relevant, informa-tion in the context of the consultation of the oracle is only the numeral in the sequence 'NUM-N', whereas the sequence 'N-NUM-CL/Q' treats both, the noun and the numeral, as new information. In my view, the fact that classifi-ers are possible only in the postnominal construction can be derived from Takashima's description. Similar to the use of normal vs. general classifier in Modern Standard Chinese (§2.4.2), classifiers as such seem to be used in circumstances where the noun was maximally unknown in Classical and Pre-Classical Chinese. Since the position of the numeral with indefinite nouns was after the noun, this position was the one to select for 'NUM-CL' constitu-ents. This situation seems to be more or less stable until the beginning of the Han dynasty (206 B.C.). For that whole period, Wang Li (1958: 234-235) lists three constructions:

(53) a. NUM Ν b. Ν NUM c. Ν NUM CL/Q

The first construction is by far the most common and will not be illustrated by any examples. Construction (53b) is illustrated by (54), construction (53c) by (55):

(54) Archaic Chinese (Shujing, Zhaogao, 40; 9th-6th century B.C.) näi she yu χ In yi, niu yl then sacrifice_to_the_spirit_of_earth in new city bull one

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 161

yang yi shl yj. sheep one pig one Than he sacrificed at the altar to the spirit of the earth in the new city a bull, a sheep, and a pig.'

(55) Classical Chinese (Zuo, Zhuanggong 18; compilated in the 5th and 4th century B.C.) jie ci yii wü gu, me sän pi. each give jade five pair horse three CL 'to each of them [two] he [i.e. the king] g a \ e f i v e pairs of jade orna-ments and three horses.'

Up to the Han dynasty, we also sporadically find the sequence 'CL/Q-NUM' instead of 'NUM-CL/Q' (Liu 1965: 67). Later, this alternative became defi-nitely impossible. Thus, the sequence of 'NUM-CL/Q' as we find it in Mod-ern Standard Chinese was established at a comparatively early time in the history of Chinese. The change from 'N-NUM-CL/Q' to 'NUM-CL/Q-N' took place later. Some first instances of a prenominal 'NUM-CL/Q' constituent turn up in the 'Analects of Confucius' (Lunyu, 5th century B.C.) and in Mencius (second half of the 4th century B.C.). All of these instances are limited to quantifiers and they are used to emphasize the numeral and the measure:

(56) Classical Chinese (Mencius, 6A18) jin zhl wei ren zhe, you yi yi bei shut, today ATTR practice humanity NOM like take one Q:cup water jiu yi che xin zhl hud ye. help one cart(load) firewood ATTR fire COPULA 'Those who practice humanity nowadays are like those who [try to] put out a whole cartload of burning firewood with one single cup of water.'

The fact that the prenominal position of the 'NUM-Q' constituent implies emphasis on the numeral and the measure seems to be linked to the above observation by Takashima (1984/1985) that what is relevant in 'NUM-N' con-structions is only the numeral or, to be more general, quantity. This relevant information can then be emphasized. For a similar reason, the numeral 'one' very often implies emphasis on the singularity of the noun in Classical Chi-nese if it occurs in prenominal position:

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

162 Walter Bisang

(57) Classical Chinese (Mencius 1B3): ci pifu zhl yöng, di yi ren this common_fellow ATTR valour oppose_an_adversary one man zhe ye. NOM COPULA 'This is the valour of a common fellow which is of use only against a single adversary.'

The prenominal use of the constituent 'NUM-CL/Q' increases parallel to the rise of classifiers in the periods of Wei-Jing and Nanbeichao (220-581 A.D.) as described extensively by Liu (1965). However, even at that time, construc-tions without classifier are by far more widely used than constructions with a classifier. This is pointed out by Gurevic (1966: 530), who criticizes Liu for never relating the number of instances with a classifier to those with no clas-sifier. If one does this, the period analysed by Liu may still be called the first flowering period of classifiers simply because of the wealth of new classifi-ers, but classifiers still seem to be much less relevant and compulsory than in contemporary Chinese. In more conservative Chinese texts such as the one analysed by Schäfer (1948) from the 9th century A.D., the above construc-tions (53a) to (53c) as introduced by Wang Li are still predominant.

The constraints which finally led to the prenominal position for the 'NUM-CL/Q' constituent are far from being understood. To conclude the discussion of word order, I would simply like to present some interacting factors which might have brought about this change of word order.

One important reason for the 'NUM-CL/Q' sequence to change into the prenominal position may be seen in the fact that all the other modifiers of the NP (genitives, relative clauses, stative verbs [adjectives], demonstratives) are in this position. This change can thus be seen as an instance of regularization. However, regularization alone cannot be a sufficient explanation. In my view, the potential linking of the preverbal position to definiteness/specificity and the postverbal position to indefiniteness is another important factor that led to the 'NUM-CL/Q-N' word order of Modern Standard Chinese as well as in Cantonese (§2.5.2). In Classical Chinese, word order does not seem to yield the above consequences for reference marking. Furthermore, the use of the numeral yJ 'one' as a marker of indefiniteness seems to come up definitely after the Han dynasty (206 B.C. to 220 A.D.). In Classical Chinese, the nu-meral 'one' was only used to count items. In prenominal position, it often even emphasized the singularity of the noun (cf. example (57)). Later, we find the numeral 'one' in combination with the existential verb you 'there is',

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 163

which is generally used to introduce new items. The following example from the 9th century A.D. (Schäfer 1948: 413) is to illustrate this:

(58) Medieval Chinese (Schäfer 1948: 413): you yi läo-ren... there is one old-man 'There was an old person ...'

Because of this new construction, the postnominal word order 'N-NUM-CL/ Q \ in which the noun and the numeral are both to be understood as relevant information according to Takashima (1984/1985, cf. above), is no longer nec-essary. This new construction replaces the older constructions of 'N-NUM' and 'N-NUM-CL/Q'. In the text analysed by Schäfer (1948), this replace-ment seems to be limited to the numeral 'one'. In the following example (59a) with the numeral er 'two', we still find the word order 'N-NUM-CL' with an indefinite noun introduced by the existential verb you 'there is'. How-ever, if the very same noun is taken up in the ongoing text we shall find the sequence 'NUM-CL' in the position for definite/specific nouns, i.e. in the preverbal position without the existential verb (59b):

(59) Medieval Chinese (Schäfer 1948: 413): a. you dä-jiäng er rin

there_is great-general two CL:man 'There were two great generals ...'

b. er jiäng... two general 'The two generals [already mentioned earlier]...'

Since the preverbal position in (59b) sufficiently marks definiteness, any fur-ther marking by the position of the numeral relative to the noun becomes superfluous. Thus, the fact that the marking of definiteness/specificity vs. indefiniteness was transferred from the position of the numeral relative to its head noun to the position of the noun relative to the verb allowed the position of the numeral constituent to its head noun to become fixed. During this proc-ess of fixation, the prenominal position was selected in conformity with the rest of Chinese modifiers which also occur in that position. However, even in Modern Standard Chinese, the 'N-NUM CL/Q' has not completely disap-peared. We still find it in lists where different items are enumerated.

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

164 Walter Bisang

In addition to the classifiers already mentioned above in the context of the oracle-bone inscriptions some more classifiers denoting highly valued and culturally significant objects came into use up to the Qin dynasty (221 B.C.): sheng2' and Hang22 for chariots (especially war chariots), mei 'tree trunk' for wooden objects, ge for arrows and pi for horses. The following example presents ge 'bamboo tree' which forms the historical basis of the general clas-sifier ge of Modern Standard Chinese:

(60) Classical Chinese (Xunzi, Yibing; 298-235 B.C.; also cf. Wang Li 1958: 238) shl wü shi ge. arrow five ten CL 'fifty arrows.'

A very special method for counting animals is attested in the Shiji (Historical Records by Sima Qian, 1357-93? B.C.)23, where body parts such as ti 'hoof', zu 'foot', shou 'hand', zhi 'finger', and jiäo 'horn' are used to count animals in the postnominal classifier position. According to the comments to these instances of counting, it was always necessary to divide the numeral occur-ring with these classifiers by two or four, if an individual animal had two or four items of that body part. Thus, niu qiän ζύ [ox-1000-foot] means '250 oxen', since one ox has four feet. Similarly, mä er bäi ti [horse-2-100-hoof] means '50 horses'. This rather complicated method did not have any conse-quences for the further development of classifiers in Chinese. A very special instance, whose translation I shall leave to the reader, is the following:24

(61) Chinese of the Shiji (Houzhizhuan; 1st century B.C.) niu ti-jiäo qiän oxen hoof-horn thousand

At the period of time described by Liu (1965), i.e. 220-581 A.D., the vast majority of the classifiers presented in (15) are already attested. Some of these classifiers are listed here: zhJl (for animals), kou 'mouth' (for animals), tou 'head' (for animals), shü 'tree', zhl2 'twig' (for one-dimensional objects), tiao 'branch' (for one-dimensional objects), zhäng 'to spread, stretch' (for two-dimensional objects), y'/an (for garments), suö/jiän (for places); special classifiers: ben 'root' (for books),ßng 'seal' (for letters), sou (for ships).

Since it is impossible to present the development of all the classifiers in detail in a paper like this (for more information cf. Liu 1965), I shall concen-trate on presenting a short outline of the development of some of the most

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 165

widely used classifiers. As I pointed out several times above, classifiers were first used with culturally valued items. Later, their use extended to other ob-jects with similar properties. At that period of extension, classifiers devel-oped prototypical features. This kind of generalization started at some time during the Han dynasty (206 B.C. to 220 A.D.), when classifiers referred to flowers, fruits, birds and animals. In the process of generalization nouns de-noting trees or parts of trees became of great importance. The general classi-fier ge of Modern Standard Chinese as well as its predecessor mei both go back to nouns denoting trees, ge meaning 'bamboo tree', mei meaning first 'tree trunk', then 'tree' in general (§2.2.1).25

Although the basic meaning of mei is 'tree trunk', it was very often used with fine, subtle varieties of stems or stalks. Furthermore, a particular type of small stalks denoted by mei was used with a particular instrument for count-ing called chöumä and with the interpretation of oracles.26 If this type of mei-stalks was generally used in the context of counting it comes as no surprise that mei was later used as a classifier. As is proved by the numerous examples quoted by Liu (1965: 77-82), the classifier mei was able to refer to almost any noun by the time of the periods of Wei-Jing and Nanbeichao (220-581 A.D.). Thus, it is well justified to look at mei as a general classifier at that time, although one should always bear in mind that classifiers did not have the same status of obligatoriness as in Modern Standard Chinese. The noun ge was first used to classify bamboo trees and arrows (cf example (60)). At the period of time analysed by Liu (1965), ge was also used to refer to other objects and even to humans in alternation with the specific classifiers denot-ing these items.27 By the Tang dynasty (618-907 A.D.) ge started slowly to get its way against mei. Other widespread classifiers based on trees are tido 'branch' and zhl2 'twig' with the former referring to big items and the latter denoting small, slender items.

As is pointed out by Erbaugh (1986: 430), the classifier zhäng for two-dimensional items developed later than tido for one-dimensional items. This is in conformity with the findings on the organization of vision as described by Marr (1982) (§2.2.3). The classifier zhäng is remarkable, because it de-rives from a verb meaning 'stretch' used e.g. in the context of stretching a bow. For that reason, it seems quite natural that this classifier was first used with bows and stringed instruments such as the qin 'Chinese zither' by the Han dynasty. By the Song dynasty (960-1279 A.D.), its extension to two-dimensional items in general was well established. Other classifiers such as those referring to buildings and places (suö, jiän) came up still later than zhäng, i.e. in the Tang dynasty.

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

166 Walte r Bisang

The diachronic development of classifiers started from a specific lexical use to the formation of prototypes which formed the basis for further generaliza-tions. In a very similar way, classifiers are acquired by children. To conclude this section, I shall describe this parallelism by quoting Erbaugh (1986):

Early use [of classifiers by children, W.B.] was lexical, specific to a single referent. Later use marked a prototypical member of a noun class. After this, the child began generalizing from the prototype, often in plausible but incor-rect ways. Shape, especially vertical extension and small size, were the most frequently generalized features. The child acquisition pattern for classifier strik-ingly parallels: 1) general semantic acquisition patterns, both in Mandarin and cross-linguistically; 2) individual and adult variation in classifier choice; 3) the historical pattern of broadening classifier scope from single to class refer-ence. (Erbaugh 1986: 415)

3.2. Category-oriented development: mainland Southeast Asia

The categories which trigger the development of classifiers in the mainland Southeast Asian languages to be presented below are based on taxonomy and meronomy. Since taxonomy seems to be more generally involved in the de-velopment of classifiers, it will be presented first in §3.2.1. §3.2.2 will be dedicated to the interaction of taxonomy and meronomy in Vietnamese as described by Löbel (1996, to appear). Finally, the situation in Vietnamese will be compared to Thai and Hmong §3.2.3. As I try to show, Vietnamese seems to be the language with the highest degree of freedom for nouns to be reinterpreted as classifiers. Furthermore, Vietnamese seems to be the language in which meronomy is most generally integrated into the formation of classi-fiers. However, meronomy is also involved in classifiers in Thai and maybe even in Hmong.

3.2.1. Taxonomy, class nouns and their role in the development of classifiers

In Bisang (1993) on Hmong, I presented a process of development based on class nouns (CN)28. Class nouns form the first step on the pathway of grammaticalization from noun to classifier:

(62) Ν > CN > Q > intQ > CL

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 167

Class nouns are based on taxonomic classification. They generally represent a rather high level of abstraction from which more concrete subcategories can be derived by further determination. An English example would be tree from which we can derive apple tree. In Hmong, there are class nouns such as ntoo 'tree', txiv 'fruit', noog 'bird', ntses 'fish', tub 'son' (for agentive nouns), kws 'expert' (for people with a certain profession), zaj (for sayings, speeches), etc. In (63), the class nouns txiv 'fruit' and ntoo 'tree' are further determined to designate special subcategories. Example (64) presents the class noun zaj :

(63) txiv kab ntxwv 'orange' ntoo kab ntxwv 'orange tree' txiv ntseej 'chestnut' ntoo ntseej 'chestnut tree'

(64) zaj dab neeg [CN-spirit-human] 'legend' zaj teev ntuj [CN-pray-heaven] 'prayer' zaj tshoob [CN-marriage/wedding] 'wedding song'

If we take the five levels of taxonomic abstraction as presented by Berlin — Breedlove — Raven (1968, 1973), class nouns generally belong to the levels [2] and [3], whereas level [1] already tends to belong to the realm of classifi-ers.

(65) Five levels of abstraction (Berlin — Breedlove — Raven 1968, 1973)

[1] unique beginner: plant, animal [2] life form: tree, bush, flower [3] generic name: pine, oak, maple, elm [4] specific name: Ponderosa pine, white pine, jack pine [5] varietal name: northern Ponderosa pine, western Ponderosa pine

Similar to Hmong, we also find a wealth of class nouns in Thai. I shall only present three of them:

(66) khon 'person': khon-khäay-chäat '[CN-sell-nation] traitor', khon-khay '[CN-sick] sick person, patient', khon-Ηααη '[CN-center/middle] middleman, mediator', khon-khäaw '[CN-news] reporter', khon-nook '[CN-outside] outsider', etc.

phäu 'elegant for "person"' (forms mainly nomina agentis): phüu-kamkap '[CN-direct/control] director', phüu-khon-khwäa '[CN-search] researcher, investigator', etc.

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

168 Walter Bisang

dook 'CN for flowers, blossoms' dook-mäay 'flower', dook-kuläap 'rose', dook-bua 'lotus flower', dook-

fin 'poppy', dook-het 'mushroom', etc.

The relation between class nouns and classifiers in Thai is particularly evi-dent with semi-repeater classifiers (§2.2.2.1), in which the class noun is 're-peated' in the classifier position:

(67) khon-khäy säam khon CN:man-sick three CL:man 'three patients'

(68) dook-kuläap cet dook CN:flower-rose seven CL:flower 'seven roses'

The above examples do not imply that every class noun can be grammaticalized into a classifier in Thai. Only a restricted number of class nouns can also display the function of a classifier. This is the case with khon and dook, but it is not the case with phüu , which mostly takes the classifier khon :

(69) phüu-khon-khwäa sii khon researcher four CL 'four researchers'

Finally, at a next level of grammaticalization, classifiers derived from class nouns are also used with nouns in which they do not occur:

(70) kuncee söoq dook key two CL 'two keys'.

3.2.2. Taxonomy and meronomy in the Vietnamese classifier system

As pointed out by Löbel (1996, to appear), taxonomy and meronomy interact in a very interesting way in Vietnamese. The first example from Tru'o'ng (1970: 256) refers to taxonomy by means of class nouns which can be reinter-preted as classifiers.

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 169

a. cdi cäy CL tree/plant

b. cäy rau CL vegetable

c. rau can CL celery

In the above example, "a noun is a classifier relative to the classified noun, i.e. if it stands in a taxonomic relation to it. This means that the classifier denotes a class, whereas the classified noun denotes the corresponding sub-class." (Löbel to appear: 18). Thus, the noun cay 'tree/plant' is subsumed under the general classifier for nonliving things, i.e. cdi. (71a). Similarly, the noun rau 'vegetable' can be seen as a subclass of trees or plants and is thus classified by cäy (71b). The same also applies to can 'celery' in (71c), which is marked as a subclass of vegetables by rau. As is pointed out by Löbel, taxonomic classification is not even limited to nouns denoting a subclass. Subclasses may also be formed by restrictive attributes as in the following example from Löbel (to appear: 19):

(72) a. hai chiec do Öam two CL dress female 'two dresses of European (in contrast to local) style'

b. hai do öam may ό Viet-Nam two CL:female dress sew in Vietnam 'two dresses of European style made in Vietnam'

On the basis of these findings, Emeneau's (1951) distinction between 'classi-fied nouns' and 'non-classified nouns' is no longer justified. Whether a noun takes a classifier or not is certainly not a matter of the lexicon, but rather a matter of use. Vietnamese has a pattern of two positions, in which the first noun is reinterpreted as a classifier with the function of individualizing (and classifying) the item of the second position. This kind of reinterpretation can take place if the item in the second position is in the relation of a subclass to the item of the first position. In this case, we have the situation referred to in the above pathway of grammaticalization (62), in which a class noun is re-interpreted as a classifier. The second situation, under which this kind of re-interpretation can be realized is if the items occurring in the two positions are

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

170 Walter Bisang

in a part-whole relation. This is shown in the following example (Löbel to appear: 13):

(73) a. mot cü'a xeho'i one CL:door car 'a door of a car'

b. mot kien xeho'i one CL: mirror car 'a mirror of a car'

If there is no part-whole relation (74a) or if the part-whole reading is not intended (74b) a core classifier (cf. (19), §2.2.2.5) has to be used (Löbel to appear: 13-14):

(74) a. mot chiec xe ho 'i one CL car 'a car'

b. mot cdi kien one CL mirror 'a mirror'

As is pointed out by Löbel (to appear: 14), there is no contextual presupposi-tion with regard to 'mirror' in example (74b). If the core classifier does not occur in that position, i.e. if kien occurs in the classifier position (73c), this implies that it is "a part of a contextually given whole":

(73) c. mot kien one CL: mirror 'a mirror (of X)'

The findings on Vietnamese point at the semantic closeness of taxonomy and meronomy as described by Cruse (1995; also quoted by Löbel to appear: 21):

Any taxonomy can be thought of in part-whole terms (although the converse is not true): a class can be looked on as a whole whose parts are its subclasses (Cruse 1995: 179).

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 171

However, to conclude from these data on Vietnamese that taxonomy and meronomy are expressed in the same way is not quite adequate. This can be shown by the following example from Löbel (1996: 152):

(75) a. mot cai ban one CL:thing table

[class subclass]

b. mot mat ban one CL:surface table

[part whole]

(75a) is an example of taxonomy. The item representing class is in the first position, the item representing subclass is in the second position. Example (75b), which illustrates meronomy, is a mirror image to (75a) inasmuch as the part which corresponds to subclass occurs in the first position followed by the item corresponding to class in the second position. Thus, the item to which the function of individualization is ascribed by its first position is the seman-tically higher noun in the case of taxonomy, whereas in the case of meronomy, it is the semantically lower noun. Thus, taxonomy and meronomy are not reflected in the same way in Vietnamese as postulated by Löbel (1996, to appear). For the same reason, the above pathway of grammaticalization (62) only covers taxonomy, where the semantically higher noun is in the class noun position and thus free to be reinterpreted as a classifier. To make our pathway of grammaticalization more general, we should thus add the notion of 'part of whole' (PW):

(76) Ν > CN/PW > Q > intrQ > CL

Even if it is not true that taxonomy and meronomy are treated in the same way in Vietnamese, what remains true for this language is that taxonomy and meronomy are two basic relations of categorization and that both of them are used for syntactic purposes (Löbel 1996:153) (§2.3.2.2). The use of semantic categories such as taxonomy and meronomy for syntactic purposes is a char-acteristic of Vietnamese by which it clearly differs from languages such as English, French or German.

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

172 Walter Bisang

3.2.3. The category-oriented development in comparison: Vietnamese, Thai, Hmong

The three languages will be compared with reference to the freedom for nouns to be reinterpreted as classifiers and with reference to the importance of meronomy.

In Vietnamese, the pattern of reinterpretation can be freely applied to any sequence of two nouns which are in the relation of taxonomy or meronomy. For that reason, given the semantic preconditions, almost any noun can po-tentially be interpreted as a classifier with its functions of individualization and classification. On this basis, Vietnamese classifiers represent a poten-tially open grammatical category, even if there is a set of highly gramma-ticalized core classifiers based on nouns of a high level of abstraction (cf. (19)). In Thai and in Hmong, the freedom of nouns to be reinterpreted as individualizers seems to be more restricted by lexical rules than in Vietnam-ese. In these languages, a noun basically has its classifier which does not change even if that noun is further specified. In Vietnamese, further specifi-cation leads to a change of the classifier (cf. (20)). Thus, Vietnamese seems to be the language in which the category-oriented development can be observed most clearly in its productive state. This productivity equally applies to tax-onomy as well as meronomy.

As was pointed out in §2.2.2.1, the degree of classifier variation is ex-tremely restricted in Thai. A great number of nouns have just one classifier which is lexically assigned to them. For these lexical reasons, the use of clas-sifiers with given nouns is highly fossilized by linguistic conventions. Even if repeater and semi-repeater constructions make sure that classifiers are a relatively large and maybe even restrictedly open grammatical category, the freedom of reinterpreting any noun as a classifier is extremely limited, since there is only a certain set of nouns which can occur with these constructions. At least for those nouns which belong to the established Thai lexicon, there is no choice. The only instance where there is some choice are new words.

The majority of classifiers seem to be based rather on taxonomy than meronomy. Nevertheless, there are some nouns denoting parts of wholes such as päak 'mouth' (for some open hollow objects), chyak 'rope' (classifier for elephants based on their trunk), khan 'handle' (for long objects with a han-dle), däam 'handle, holder' (for long objects with a handle), and tua (for animals), which seems to go back to Chinese tou 'head' (Prapin 1975: 345).

In the case of repeater classifiers, the position of the noun in 'N-NUM-CL' can be left empty, if that noun is in a part-whole relation to some other object

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 173

mentioned in the context. This is shown by the following two examples from Conklin (1981: 74):

(77) εΐιάαη mii (rjaa) sooq ηαα. elephant have tusk two CL:tusk 'Elephants have two tusks.'

(78) a. naayphraan mii ηαα sooq ηαα. hunter have tusk two CL:tusk 'The hunter has two tusks.'

b. *naayphraan mii sooq ηαα. hunter have two tusk 'The hunter has two fangs/tusks.'

In (77), the noun ηαα 'tusk' can be deleted. The 'CL-NUM' constituent, which implies a part-whole relation, is possible, because the 'tusks' are a body part of the elephant. Example (78b) without the noun is not possible, because this "would imply that the 'tusks' belong to the hunter in the sense of being part of his body rather than part of his trophy collection" (Conklin 1981:74). Thus, the situation with Thai repeater classifiers can be identical to that situation in Viet-namese, in which the occurrence of only one noun after the numeral implies a part-whole relation given by the context (§3.2.2). The difference is, however, that the implication of a part-whole situation is generally valid in Vietnamese, whereas in Thai it is limited to some instances of repeater classifiers.

In Hmong, the set of items belonging to the category of classifiers seems to be limited to the list presented in (12) and (13) (§2.2.2.2). As in Thai, a noun seems basically to have its strictly assigned classifier. In spite of the limited number of classifiers in Hmong, there are some special classifiers such as zaj , which can occur in the function of a class noun (cf. (64), §3.2.1) as well as in the function of a classifier (cf. (13) in §2.2.2.2). Instances such as these show the connection between class nouns and classifiers in Hmong. Nevertheless, the category of classifiers seems to be established as an inde-pendent category with its own position. This can be seen from the fact that items such as zaj can occur twice in a sequence as in the following example from Bertrais-Charrier (1979: 567):

(79) ob zag29 zaj tshoob two CL CN wedding 'two wedding songs'

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

174 Walter Bisang

In the position after the numeral, zaj has the function of a classifier. In the next position, it occurs again in the function of a class noun. In Vietnamese, this would be impossible. Vietnamese has no position exclusively for classi-fiers. In Thai, the special position for classifiers/quantifiers is maybe due to the fact that it occurs after the head noun.

The items within the small list of Hmong classifiers which refer to meronomy are txhais (for one item of a pair) and maybe tus, if it can be derived from Chinese tou 'head'.

From the above observations we may thus conclude that the category-oriented development is more visible in Thai than in Hmong. Of course, Viet-namese remains the clearest instance of that development.

3.3. Item-oriented vs. category-oriented development from an areal point of view30

The item-oriented process of classifier development seems to be located in Chinese, where its first traces can be followed back to the very beginning of written documents. The context in which classifiers emerged was clearly that of counting. The use of classifiers with demonstratives came up much later. The item-oriented origin of classifiers may be reflected in Modern Standard Chinese by its semantic pecularities such as the lack of a specific classifier for humans/animates and the relatively large number of nouns which can oc-cur with different classifiers without any change of meaning (§2.2.2.3). The fact that a noun had to be of particular cultural value for being eligible for classifiers at the beginning of their emergence may still be reflected by the criteria determining the selection of normal vs. general classifiers. The items the speaker wants to treat with particular distinction, because s/he thinks they are unknown to the hearer, are marked by a normal (special) classifier, while items that are not particularly noteworthy, i.e. items that are already known, are marked by the general classifier (§2.2.2.3, §2.4.2).

Chinese does not lack the category of class nouns. Their list would be about as long as for the languages of Southeast Asia. In the following exam-ple, I shall only present three class nouns. As we can see from this example, class nouns occur at the end of the noun. This is consistent with the general prenominal modifier position in the NP syntax of Chinese.

(80) tou 'head' shitou 'stone', zhuäntou 'brick', zhentou 'pillow', quäntou 'can, pot', gutou 'bone', shetou 'tongue', shitou 'snake, serpent', quäntou 'fist',

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 175

mütou 'wood', xiängyäntou 'cigarette butt' (xiängyän - 'cigarette'), etc. shuötou 'topic of conversation' from shuö 'speak, say, tell\päntou 'some thing one looks forward to' from pan 'hope for, long for', etc.

shu 'tree' xingshü 'apricot tree', sängshü 'mulberry', ydngshu 'poplar', liüshü 'willow', meishii 'plum tree', täoshü 'peach tree', etc.

suö 'place, locality, institution' suo occurs with many nouns denoting places, institutions, buildings: göngsuö 'public office', zhusuo 'dwelling place', zhensuö 'clinic', yüsuö 'residence, abode', chängsuö 'place, arena', etc.

The above list is not representative, because all its items can also take the function of a classifier. The noun tou 'head' for instance is used as a classifier for counting animals (cf. (15)) and it is used as a very productive element of noun formation (cf. its loss of the tone in (80) above). The vast majority of Chinese class nouns, however, is not used in the function of classifiers. Those class nouns which can be classifiers as well mostly denote categories that developed rather late in the history of Chinese such as places or buildings (the above suo and some others, cf. §3.1). This situation may indicate a later contact with classifier systems based on the category-oriented development in the South. Another instance which may point into that direction is the fact that there is a rather small set of Chinese nouns whose last item is a classifier (for a more complete list cf. Unterbeck 1993: 155-157):

(81) chudn-zhl [ship-CL] 'ships, vessels' fäng-jiän [room-CL] 'rooms' shü-ben [book-CL] 'books' mä-ρϊ [horse-CL] 'horses'

This position of the classifier may be a late reflection of what stood at the beginning of the category-oriented classifier development, in which a class noun was reinterpreted as a classifier. In Chinese, it comes up at a time when the classifiers involved are long established. The function of the postnominal classifier is to generalize the meaning of the noun (Unterbeck 1993: 155). However, this type of word formation is quite different from class nouns, since it is unproductive. It is limited to one, maybe two nouns, which maximally meet the prototypical properties of the classifier.

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

176 Walter Bisang

The category-oriented process of classifier development seems to be located in mainland Southeast Asia. It can be observed in its synchronically most productive form in Vietnamese (§3.2.2). Since this system does not come from a context of counting, it does not necessarily imply counting as a pri-mary function of classifiers. Even if counting is involved in individualization and classification (§2.3.1), this does not mean that these functions are auto-matically linked to counting. In Vietnamese, their main function is syntactic referentialization (§2.3.2.2). For that reason, the context of counting is sec-ondary for Vietnamese. In Chinese, where the classifier grew out of the cul-tural necessity to count valuable things such as tributes presented to the ruler, counting remained of primary importance up till now. The fact that the cat-egory-oriented development is open to alternative primary functions such as syntactic referentialization, smoothes the way for an expansion of the func-tional range of classifiers into referentialization and relationalization. If Can-tonese classifiers are also used in the function of relationalization (§2.5.2), this is most likely due to contacts with Southeast Asian languages.

As pointed out at the beginning of §3, the item-oriented process implies the emergence of a new construction containing a new and clearly defined position for the classifier. In a classifier system based on the category-ori-ented process, this is not necessary. A pattern with two nominal positions of which the first noun can be reanalysed as a classifier is enough. In my view, this applies to Vietnamese. I think it is for that reason that Vietnamese can have nominal compounds to be used in discourse with aforementioned enti-ties (§2.3.2.2). The other two Southeast Asian languages described in this paper, Thai and Hmong, both have developed a special position for classifiers (§3.2.3). For that reason, it is not possible for every class noun or for every noun denoting part of whole to be automatically reanalysed as a classifier. Maybe the emergence of an independent classifier position in Thai and Hmong was supported by the pattern from the North, i.e. by a system developed from an item-oriented process. In Thai, the situation is particularly interesting be-cause the postnominal position of the classifier precludes the reanalysis of a class noun as a classifier, since class nouns occur at the beginning of the head noun. If we assume that the prenominal position of the classifier/quantifier constituent is primary in Tai, since its postnominal position is limited to South-western Tai (§1.3), we may conclude that the movement of the classifier/ quantifier constituent to the postnominal position was possible only after an independent prenominal classifier position was established.

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 177

Notes

1. In §2.1,1 shall further specify this particular type of individualization based on inherent properties as actualizing individualization. Individualization based on external properties such as measures etc. will be called creative individualiza-tion, since the parameter of comparison has to be established, i.e. created.

2. Jones (1970) is still one of the most informative papers on word order and its areal distribution in East and Southeast Asia.

3. At first glance, word order in the context of the numeral 'one' seems to be an exception in Nung, a Central Tai language (Saul 1964: 261, Saul and Wilson 1980: 25-30). In this language, word order with the numeral 'one' is 'CL-N-NUM', whereas all the other numerals follow the pattern of 'NUM-CL-N'. (i)au ähn tdhng nu'hng ma.

take CL chair one come 'Bring a chair.' (Saul — Wilson 1980: 27).

Since nu 'hng occurs in the position of the demonstrative, its categorial status is not quite sure. If it is a demonstrative element whose function is to specify in-definiteness as in Thai (cf. Haas 1942: 204, Hundius — Kölver 1983: 180) ex-ample (i) does not disprove Greenberg's universal.

4. In Thai, word order 'N-CL-one' is used with indefinites, whereas word order 'N-one-CL' occurs in the context of counting.

5. The quantifier xub unifies small animals, insects and ants into larger groups. 6. Maybe the clearest case in which sexus is connected with classifiers is the Miao

language of Weining as described by Wang Deguang (1987). In this language, the classifier Iiu55 seems to be linked with females, whereas tsi55 occurs with males. In the following example, I present both classifiers in their normal, defi-nite form, i.e. in the form of lai55 and tsai55 (for the formal distinction of classifi-ers according to normal/diminutive/augmentative and definite/indefinite cf. §2.4.3.3.): (i) Weining Miao (Wang Deguang 1987: 36)

lai55 a33bfio35 qi55

CL:FEM woman this 'this woman'

(ii) tsai55 a55zey53 3' qi55

CL:MASC man this 'this man'

7. A potential counter-example to this hierarchical system is Yao as described by Caron (1987: 161): "There are no discernible componential features which oper-ate over the entire set of Yao CLs in any non-trivial way. Any set of descriptive semantic features ... will not operate over the entire set of Yao CLs in a hierarchi-cal fashion."

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

178 Walter Bisang

8. ntawv means 'paper'. The classifier tsab, which is specialized for classifying messages, refers to paper in its shape of a letter. With the intrinsic quantifier phau 'pile', the noun ntawv 'paper' is looked at in its form of a book, i.e. as a pile of paper: ib phau ntawv [one-intQ-book] 'a book'.

9. There are two different classifiers with the same pronunciation but with different writing. They are distinguished in this paper by the numbers 1 and 2.

10. In Cantonese, we can find a similar degree of flexibility (Matthews — Yip 1994: 105-107).

11. Small machines are counted by -ko (classifier for small, three-dimensional ob-jects).

12. This test is very useful for Vietnamese. But it is not possible to apply it to classi-fier languages in general. In Modern Standard Chinese, 'classifier-noun' con-structions are defined by the fact that they cannot occur as a modifier of the noun (T'sou 1976). In Thai, where we have the sequence 'N-NUM-CL', the 'NUM-CL' constituent cannot be interpreted in the sense of a qualitative modification of the noun.

13. Adams (1992) presents an excellent description of how rich the classificational system of humans is in Vietnamese.

14. The classifier vi is from Chinese wei presented in §2.2.2.3, example (14b). 15. A third potential explanation will be presented at the end of §2.3.1. 16. The phenomenon that classifiers already imply the numeral 'one' can be ob-

served in quite a few classifier languages. In Cantonese, we find the same impli-cation, cf. §2.5.2. In Modern Standard Chinese, the numeral 'one' can only be omitted in the postverbal object position, not in the subject position. Moreover, 'DEM-CL-N' constructions imply singular reference. For expressing plurality in Modern Standard Chinese, the NUM position in front of the classifier has to be filled by xie 'some' or by a numeral equal to or larger than 1. In Thai, the classi-fier implies the numeral 'one' in 'noun-classifier-demonstrative' constructions (§2.4.3.1).

17. I owe this example to Mrs Li Liuming, a native speaker of Cantonese, who is presently writing her doctoral dissertation on Cantonese grammar.

18. This form is actually ambiguous, since the normal form of this classifier is also ts 'ae33.

19. I owe this example to Li Liuming. 20. Wang Li (1958: 236) only speaks of two natural numeratives (tiänrän dänwei).

He does not mention quan 'dog'. 21. Today, the character representing this word is pronounced sheng in its function

of a classifier and cheng in its function of a verb meaning 'ride, go on horseback; travel'. In its function as a numerative, it is first used to individualize war chari-ots with four horses. War chariots were of great importance to measure the mili-tary power of a country.

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 179

22. The classifier liäng is closely linked to the numeral Häng ' two'. The numeral 'two' is particularly relevant for chariots which typically had two wheels at the time of Classical Chinese. The classifier liäng is thus an example of classifica-tion on the basis of a part-whole relationship. Paleographically, the phonetic part of the character of liäng can be compared to the classifier bing in the oracle bone inscriptions.

23. The majority of these instances of counting is found in the "Houzhizhuan"-chap-ter, pp. 3272-3274 in the Zhonghua Shuju edition.

24. The comment presented in the Zhonghua Shuju edition (p. 3272) translates Ί 6 6 oxen' since each ox has two horns and four hooves.

25. mei has nothing to do with bamboo as claimed by Erbaugh (1986: 428). In the Shuowen, mei is interpreted as gän 'tree trunk'. In the Shijing (Mao 10; 9th to 6th B.C.) we find the sequence tido-mei which is translated as 'branches and slender stems' by Legge (1871: 17) without any reference to bamboo (cf. Legge, James. 1871. The Chinese Classics. Volume IV: The She King). Finally, mei is written with the tree-radical, not with the bamboo radical.

26. For some examples cf. Zuo, Zhaogong 12 and Zuo, Xianggong 21 (compilated in the 5th and 4th century B.C.).

27. Thus, ge was maybe more widespread even before the Tang dynasty (618-907 A.D.) than Wang Li (1958: 238) assumes.

28. The term 'class noun' corresponds to 'class term' suggested by Haas (1942) and DeLancey (1986).

29. Since there are no final consonants in Hmong syllables, the syllable final posi-tion is used to mark tones by using some consonants. In this example, zaj changes its tone into zag because of the tone of ob 'two', which may yield certain phe-nomena of tone sandhi.

30. For a more extensive description of East and mainland Southeast Asia as a lin-guistic area cf. Bisang (1996).

References

Adams, Karen L. 1982 Systems of numeral classification in the Mon-Khmer, Nicobarese and

Aslian subfamilies of Austroasiatic. Ph.D. dissertation, University of Michigan.

1986 Numeral classifiers in Austroasiatic. In: Colette Craig (ed.), 241-262. 1992 A comparison of the numeral classification of humans in Mon-Khmer.

Mon-Khmer Studies 21: 107-129. Adams, Karen L. — Nancy F. Conklin

1973 Toward a theory of natural classification. Papers from the 9th Regional Meeting, Chicago Linguistic Society, 1-10.

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

180 Walter Bisang

Ahrens, Kathleen 1994 Classifier production in normals and aphasics. Journal of Chinese Lin-

guistics 22.2: 202-247. Allan, Keith

1977 Classifiers. Language 53: 285-311. Becker, Alton L.

1975 A linguistic image of nature: The Burmese numerative classifier sys-tem, Linguistics 165: 109-121.

Berlin, Brent — Dennis E. Breedlove — P. R. Raven 1968 Covert categories and folk taxonomies. American Anthropologist 70,

290-299. 1973 General principles of classification and nomenclature in folk biology.

American Anthropologist 75: 214—242. Berlin, Brent — A. Kimball Romney

1964 Descriptive semantics of Tzeltal numeral classifiers. American Anthro-pologist 66.2: 79-98.

Bertrais-Charrier, Yves Fr. 1979 Dictionnaire Hmong (Meo Blanc) - Frangais. Bangkok: Assumption

Press. Bisang, Walter

1988 Hmong Texte. Eine Auswahl mit Interlinearübersetzung aus Jean Mottin, Contes et legendes Hmong Blanc [Bangkok, Don Bosco 1980]. (Arbeiten des Seminars für Allgemeine Sprachwissenschaft der Universität Zürich Nr. 8.) Zürich: University of Zürich, Seminar of General Linguistics.

1993 Classifiers, quantifiers and class nouns in Hmong. Studies in Lan-guage 17.1: 1-51.

1996 Areal typology and grammaticalization: Processes of grammaticalization based on nouns and verbs in East and mainland South East Asian lan-guages. Studies in Language 20.3: 519-597.

Caron, Bruce R. 1987 A comparative look at Yao numeral classifiers. Linguistics of the Tibeto-

BurmanArea 10.2: 151-168. Chao, Yüan Ren

1968 A Grammar of Spoken Chinese. Berkeley / Los Angeles: University of California Press.

Clancy, Patricia M. 1985 The acquisition of Japanese. In: Dan Isaac Slobin (ed.), The Crosslingu-

istic Study of Language Acquisition, Vol. 1: The data, 373-524. Hillsdale, New Jersey: Erlbaum.

Conklin, Nancy Faires 1981 The semantics and syntax of numeral classifiers in Tai andAustronesian.

Ph.D dissertation, University of Michigan (DAI 42/06A-2652, UM 81-25890).

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 181

Corbett, Greville G. 1991 Gender. Cambridge: Cambridge University Press.

Coyaud, Maurice 1973 Classification nominale en Chinois (les particules numerales). The

Hague: Mouton. Craig, Colette (ed.)

1986 Noun classes and categorization. Amsterdam / Philadelphia: Benjamins. Croft, William

1991 Syntactic categories and grammatical relations. Chicago / London: The University of Chicago Press.

1994 Semantic universale in classifier systems. Word 45.2: 145-171. Cruse, D.A. 1995 [=1986] Lexical semantics. Cambridge: Cambridge University Press. DeLancey, Scott

1986 Towards a history of Tai classifier systems. In: Colette Craig (ed.), 437-452.

Denny, Peter J. 1976 What are noun classifiers good for?. In: Papers from the 12th Interna-

tional Meeting, Chicago Linguistic Society, 122-132. 1986 The semantic role of noun classifiers. In: Colette Craig (ed.), 297-308.

Downing, Pamela 1986 The anaphoric use of classifiers in Japanese. In: Colette Craig (ed.), 345-

375. Emeneau, Murray B.

1951 Studies in Vietnamese (Annamese) grammar. (University of California Publications in Linguistics 8.) Berkeley / Los Angeles: University of California Press.

Erbaugh, Mary S. 1986 Taking stock: The development of Chinese noun classifiers historically

and in young children. In: Colette Craig (ed.), 399^436. Gandour, Jack — Soranee Holasuit Petty — Rochana Dardarananda — Sumalee Dechongkit — Sunee Mukngoen

1984 The acquisition of numeral classifiers in Thai. Linguistics 22: 455—479. Greenberg, Joseph.

1974 Numeral classifiers and substantival number: Problems in the genesis of a linguistic type. Proceedings of the 11th International Congress of Lin guistics, Bologna / Florence, Aug-Sept 1972. Bologna: 1974: 17-37.

Guo, Xianzhen 1987 Xiandai Hanyu liangci shouce [Handbook of Modern Chinese numera-

tives]. Beijing: Zhongguo Heping Chubanshe. Gurevic, Ilja Semenovic

1966 Numeration constructions in the Chinese language of the III-V centu-ries A.D. Archiv Orientdlni 34: 524-534.

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

182 Walter Bisang

Haas, Mary 1942 The use of numeral classifiers in Thai. Language 18: 201-205. 1964 Thai-English student's dictionary. Stanford: Stanford University Press.

Herrfurth, Hans 1964 Die Klassifikationsaffixe im Chinesischen. In: Paul Ratchnevsky (ed.),

Ostasiatische Forschungen. Schriften der Sektion für Sinologie bei der Deutschen Akademie der Wissenschaften zu Berlin, Heft 3: Beiträge zum Problem des Wortes im Chinesischen II. Berlin: Akademie-Verlag.

Hla Pe 1965 A re-examination of Burmese 'classifiers'. Lingua 15: 163-185.

Hopper, Paul J. 1986 Some discourse function of classifiers in Malay. In: Colette Craig (ed.),

309-325. Horn, Lawrence R.

1984 Toward a new taxonomy for pragmatic reference: Q-based and R-based implicature. In: Deborah Schiffrin (ed.), Meaning, form, and use in con text: linguistic applications, 11-42. Washington, DC: Georgetown Uni-versity Press.

1988 Pragmatic theory. In: Frederick J. Newmeyer (ed.), Linguistics. The Cam-bridge survey. Volume 1, 113- 145. Cambridge: Cambridge University Press.

Hundius, Harald — Ulrike Kölver 1983 Syntax and semantics of numeral classifiers in Thai. Studies in Lan-

guage 7.2: 165-214. Jaisser, Annie

1987 Hmong classifiers: A problem set. Linguistics of the Tibeto-Burman Area 10.2: 169-176.

Jones, Robert B. 1970 Classifier constructions in Southeast Asia. Journal of the American Ori-

ental Society 90: 1-12. Langacker, Ronald W.

1987 Foundations of cognitive grammar. Volume 1. Theoretical prerequisites. Stanford, California: Stanford University Press.

Leach, Edmund 1964 Anthropological aspects of language: Animal categories and verbal abuse.

In: Eric H. Lenneberg (ed.), New directions in the study of language, 23-63. Cambridge: MIT Press.

Li, Charles N. — Sandra A. Thompson 1975 The semantic function of word order: A case study in Mandarin. In:

Charles N. Li (ed.), Word order and word order change, 163-195. Aus-tin / London: University of Texas Press.

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 183

Li, Frances 1971 Case and communicative function in the use of ba in Mandarin. Ph.D

dissertation. Cornell University. Liu, Shiru

1965 Wei-Jin Nanbeichao liangci yanjiu [A study on classifiers in the Wei-Jin and in the Nanbeichao periods], Beijing: Zhonghua shuju chuban.

Löbel, Elisabeth 1996 Klassifikatoren. Eine Fallstudie am Beispiel des Vietnamesischen.

Universität Köln: Habilitationsschrift. Ms. to appear Classifier systems vs. noun class systems: A case study in Vietnamese.

In: Barbara Unterbeck (ed.) Approaches to gender. Berlin: Mouton de Gruyter.

Lü, Shuxiang et al. 1980 Xiandai Hanyu babai ci [800 (functional) words of Modern Chinese].

Beijing: Shangwu Yinshuguan. Marr, David

1982 Vision. San Francisco: Freeman. Matthews, Stephen — Virginia Yip

1994 Cantonese. A comprehensive grammar. London / New York: Routledge. McFarland, George Bradley

1944 Thai-English dictionary. Stanford, California: Stanford University Press. Mottin, Jean

1978 Elements de grammaire Hmong Blanc. Bangkok: Don Bosco Press. Piaget, Jean

1970/73 Einführung in die genetische Erkenntnistheorie. (Suhrkamp Taschenbuch Wissenschaft 6.) Frankfurt: Suhrkamp.

Prapin, Manomaivibool 1975 Α study ofSino-Tai lexical correspondences. Ph.D dissertation. Univer-

sity of Washington. Sanches, Mary

1977 Language acquisition and language change: Japanese numeral classifi-ers. In: Ben G. Blount — Mary Sanches (eds.), Sociocultural dimen-sions of language change, 51-62. New York: Academic Press.

Saul, Janice E. 1965 Classifiers in Nüng. Lingua 13: 278-290.

Saul, Janice E. — Nancy Freiberger Wilson 1980 Nung grammar. Dallas, TX: The University of Texas at Arlington, Sum-

mer Institute of Linguistics. Schäfer, Edward H., Jr.

1948 Noun classifiers in Classical Chinese. Language 24: 408-413.

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

184 Walter Bisang

Searle, John 1969 Speech acts: An essay in the philosophy of language. Cambridge: Cam-

bridge University Press. Seiler, Hansjakob

1986 Apprehension, language, object and order. Part III: The universal di-mension of apprehension. Tübingen: Narr.

Serzisko, Fritz 1980 Sprachen mit Zahlklassifikatoren: Analyse und Vergleich. (Arbeiten des

Kölner Universalien-Projekts, akup, 37.) Köln: University of Köln, In-stitute of Linguistics.

1982 Temporäre Klassifikation. Ihre Variationsbreite in Sprachen mit Zahlen-klassifikatoren. In: Hansjakob Seiler — Christian Lehmann (eds.), Ap prehension. Das sprachliche Erfassen von Gegenständen I: Bereich und Ordnung der Phänomene, 147-159. Tübingen: Narr.

Sohn, Ho-min 1994 Korean. London / New York: Routledge.

Suppes, Patrick — Jospeh L. Zinnes 1963 Basic measurement theory. In: R. Luce et al. (eds.), Handbook of Math-

ematical Psychology, Volume 1, 3-76. New York / London: Wiley. Tai, James H.-Y.

1990 A semantic study of the classifier TIAO. Journal of the Chinese Lan-guage Teacher's Association 25.1: 35-56.

Takashima, Taiji 1984/85 Noun phrases in the oracle-bone inscriptions. Monumenta Serica 36:

229-301. Tru'o'ng, Van Chinh

1970 Structure de la langue vietnamienne. Paris: Paul Geuthner. Unterbeck, Barbara

1993 Kollektion, Numeralklassifikation und Transnumerus. Frankfurt / Ber-lin / Bern / New York / Paris / Wien: Lang.

Wang, Deguang 1986 Weining Miaoyu huayu cailiao [Language material on the Miao lan-

guage of Weining]. Minzu yuwen 1986.3: 69-80. 1987 Guizhou Weining Miaoyu liangci shiyi [Additional remarks on the clas-

sifiers of the Miao language in Weining, Guizhou province]. Minzu yuwen 1987.5: 36-39.

Wang, Fushi 1957 Guizhou Weining Miaoyu liangci [The classifier in the Weining dialect

of the Miao language]. Yuyan Yanjiu 1957: 75-121. 1972 The classifier in thei Wei Ning dialect of the Miao language in Kweichou.

In: Herbert C. Purnell (ed.): Miao and Yao linguistic studies: selected articles in Chinese, translated by Chang Yii-Hung and Chu Kwo-ray,

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Classifiers in East and Southeast Asian languages 185

111-185. Ithaca, NY: Cornell University Southeast Asia Program, xiii. [Data Paper no. 88, Linguistic Series VII]. [Translation of Wang Fushi (1957)].

Wang, Li 1958 Hanyu shigao [A draft history of Chinese]. Beijing.

Wiese, Heike 1996 Semantische und konzeptuelle Strukturen von Numeralkonstruktionen

(Untersuchung zu Kardinal-, Ordinal- und Nummer-Konstruktionen). Zeitschrift für Sprachwissenschaft 14.2: 180-235.

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM

Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois LibraryAuthenticated | 172.16.1.226

Download Date | 8/8/12 2:08 AM