Printer Friendly

Instruments as verb classifiers in Kam (Dong).

Abstract

This article is a contribution to an emerging crosslinguistic typology of verb classifier systems. It presents an in-depth study of one particular type predominant in East Asian and Southeast Asian languages and epitomized through Kam (Dong), a Kam-Tai language spoken in Southwest China. In Kam and related languages there are basically two types of verb classifiers: sortal and mensural. Both kinds are involved with event-phase counting and temporal measuring and hence mirror the function of their twins in the nominal realm. Sortal verb classifiers can be characterized as deriving lexically from nouns that occupy an instrumental participation role. Kam exhibits ca. 40-50 of them.

Although the term of noun and verb classifier has been around for many years, it is noteworthy that almost no scholar has ever come forward with an authoritative definition of what exactly constitutes a linguistic classificatory phenomenon. McGregor (2002) is an exception and his definition is based on distributional-probabilistic properties. I adopt, augment and partially rewrite his criteria and demonstrate that the system of sortal verb classifiers in Kam constitutes indeed a phenomenon by which verbs are categorized.

The basic semantic requirement for classifiable verbs (or events) in Kam is to incorporate one minimal part. A minimal part of an event is defined as a TOUCH-type event by which two objects meet in various ways. The prototypical verb referring to a TOUCH-event is 'hit'. Further subdivisions can be made for classifiable verbs. Verbs are categorized in terms of the instruments by which objects are caused to collide: by a physical instrument-object or by a trigger medium.

1. Introduction

1.1. Verb classifiers in Kam (Dong) and across the region

Kam (Dong) (1) exhibits an array of morphemes (ca. 40-50) that select verbs in syntactic constructions such as verb-phase counting or quantifying. In combination with numerals the morpheme articulates the number of times an event occurs within an implied timeframe. (2)
(1) [mau.sup.33]   [keu.sup.35]   [jau.sup.11]   [ja.sup.11]
    [cui.sup.11].
    3P SG           beat          IP SG          NUM:2
    VCL:fist
                    Verb                         Numeral
    Verb classifier
    'He beat me twice with his fist(s).'

(2) [kun.sup.323]   [i.sup.55]   so33
    sound           NUM: 1       VCL:voice, noise
    Verb            Numeral      Verb classifier
    '(it) sounded with one (loud) noise.'


The verbs that may be picked by these morphemes form a restricted albeit prominent section of the Kam lexicon. They basically coincide with verbs that admit the semantic role of instrument. Actually, the presumed classifier morphemes are derived from nouns that elsewhere occupy the role of instrument role of the verbs they also modify as classifier morpheme. Consider example (3) with the instrumental coverb (preposition) (3) [au.sup.55] 'take' contrasting with (1):
(3) [mau.sup.33]   [au.sup.55]   [cui.sup.11]   [keu.sup.35]
    [jau.sup.11].
    3P SG          COV:take      fist            beat
    1P SG
                                 Instrument
    'He beat me with his fist(s).'


Linguists working on languages of East and Southeast Asia employ the term "verbal classifier" because of the structural resemblance these morphemes bear with nominal classifiers. Both constructions (1) and (4) display a head (= verb or noun), a (noun or verb) classifier and a numeral, though the order of constituents differs slightly in each case.
(4) [sam.sup.35]   [mun.sup.31]     [nen.sup.11]
    NUM:3          NCL              person
    Numeral        Nounclassifier   Noun
    'three people'


The objective of this contribution is to corroborate the claim that the system of verbal morphemes is classificatory by presenting solid theoretical and empirical arguments. The Kam data in this article originate mainly from first-hand elicited sessions with Wu Shihua, a native from Sanjiang County, representing the Southern Dialect spoken in Guangxi Province. (4) The occurrence of verb classifiers in narratives is not infrequent. In a collection of unpublished folk stories I counted ca. 12 cases in a total of 8,500 words.

The knowledge of constructions such as (1) in sinospheric languages has existed for many years, even though not all constructions derive from instrumental noun phrases. It appears that these constructions exist in at least five language families of East Asia and Southeast Asia: Tibeto-Burman, Sinitic, Kadai, Miao-Yao and Mon-Khmer. For more details see Bhaskararao and Joshi (1985: 17) for Newari (Tibeto-Burman); Dai and Cui (1985: 43) for Achang (Tibeto-Burman); Xu and Zhao (1984: 31-32) for Bai (Tibeto-Burman); He and Jiang (1985: 66-67) for Naxi (Tibeto-Burman); Sun (1981: 95-96) for Qiang (Tibeto-Burman); Xu and Xu (1984: 64-65) for Zaiwa (Jingpo) (Tibeto-Burman); Chen et al. (1985: 117) for Nuosu (Yi) (Tibeto-Burman); Xu et al. (1986: 55-56) for Lisu (Tibeto-Burman); Matisoff (1973: 308-309) and Chang (1986: 36-37) for Lahu (Tibeto-Burman); Li and Wang (1986: 73-74) for Hani (Tibeto-Burman); Chao (1968: 312-314), Greenberg (1972: 30), Li and Thompson (1981: 352-354), Paris (1989: 4-7), Zhou and McGregor (1999), Yang (2001: 140-161) for Mandarin; Hashimoto (1972: 25), Killingley (1983: 99-104), Matthews and Yip (1999) for Cantonese; Schafer (1948: 413) for Classical Chinese; Yu and Luo (1980: 37-38) for Dai (Kadai); Haas (1942: 205) and Noss (1964: 108-109) for Thai (Kadai); Matthews and Leung (2001) for a comparative view of this phenomenon in Cantonese and Thai; Wei and Tan (1980: 32) for Northern Zhuang (Kadai); Yu (1980: 31-32) for Buyei (Kadai); Wang and Zheng (1993: 48) for Mulao (Kadai); Liang (1980b: 40-41) for Maonan (Kadai); Zhang (1980: 33-35) for Shui (Kadai); Long and Zheng (1998: 95-96) and Liang (1980a: 38) for Kam (Kadai); Ouyang and Zheng (1980: 30-32) for Li (Kadai); He (1983: 35-36) for Gelao (Kadai); Wang (1985: 57) for Miao (Miao-Yao); Court (1987: 147-148) for Iu Mien (Miao-Yao); Mao et al. (1982: 35, 91, 144-145) for Mien, Bunu and Lajia (Miao-Yao); Zhou and Yan (1984: 39-40) for Wa (Mon-Khmer); Gorai (1978: 16) and Nguyen (1997) for Vietnamese (Mon-Khmer). In most of these sources the issue of verb classifiers is only dealt with in a cursory and superficial fashion.

1.2. Other uses of "verb classifier'" in the literature

Before being preoccupied with the bulk of this article, we have to clarify two other uses of "verb classifier" in the literature somewhat differing from the phenomenon considered in this article. The first type of verbal classifiers actually concerns noun categorization devices and is termed as such not because the classifiers in question categorize verbs, but because their syntactic locus is within the verbal paradigm from where they categorize nominal arguments in terms of generic-specific, animacy or shape. Verbal classifiers of this type generally categorize nominal arguments that are in the S- or O-roles (cf. Dixon 1994 for the protoroles S and O) and come in three varieties: (i) as a generic incorporated noun (mainly in North American Indian and North Australian languages, see Mithun 1984: her Type 4; Evans 1996; Sands 1995), (ii) as verbal affix providing information about the animacy or shape of a nominal argument (in North American Indian and in Papuan languages, see Seiler 1985: 120-125), and (iii) as suppletive verbal stem (in North American and Tibeto-Burman languages, see Carter 1976; LaPolla 1994). (5)

The second type of "verbal classifiers" constitutes a verb classification phenomenon proper and is in circulation mainly among Australianists (cf. Capell 1979; Silverstein 1986; McGregor 2002: 25-29; among others). In many Northern Australian languages, a small set of inflected generic verbs co-occurs with an open set of uninflected specific verbs in so-called "compound verb constructions". The set of inflected verbs hereby categorizes the set of uninflected verb stems. This classification technique bears some similarity with numeral classification systems in East and Southeast Asian languages where a set of specific nouns is categorized by a set of generic lexemes.

1.3. The lack of a crosslinguistically valid definition of classification

The topic of classification has been exceedingly fashionable with linguists around the world. In the last 20 years a flurry of works has appeared focusing on all kinds of aspects: descriptions of classification phenomena in individual languages, syntheses on various ranges of languages, typologies of classification techniques and so forth. In spite of this mass of literature, it is striking that there is no commonly accepted definition of what exactly constitutes a phenomenon of classification in natural languages. In most of the works no attempt has even been made to work towards a characterization. Instead, it is taken for granted in most publications that a specific phenomenon under consideration is classificatory. The interest has focused more on various classification techniques than on a conceptual approach to the question. McGregor (2002: 18-19) is a remarkable exception. He proposes a definition of classification phenomena based on mathematical sets and remains fairly comprehensible in the presentation. His definition is basically correct and contains all crucial ingredients. In Section 2, I intend to further develop the notion of collocation of classifieds (= items to be classified) with classifiers (= item that produces a classification) and propose an amended and augmented characterization of classificatory phenomena that uses simple mathematical notions to smooth the representation.

The characterization offered in Section 2 establishes a comprehensive portrait of all the aspects associated with linguistic categorization: morphosyntactic constructions that constitute the backbone of any classificatory system, distributional features of the scheme, and semantic properties. The Kam verb classifier system indeed exhibits all the relevant characteristics: morphosyntactic (Section 3), distributional (Section 4), and semantic (Section 5).

2. General definition of classification phenomena in natural languages

Classification is a concept with wide ramifications in all kinds of scientific and nonscientific activities. As a concept, it has a long and distinct history that traces back to the Greek philosopher Protagoras (485-414 BC), cf. Royen (1929: 1) and Dixon (1982: 159). In its most general version, classification (linguistic or not) can be understood as a process in which a set of classifieds D is divided into a family of subsets through the operation of a set of categorization devices M. There is a very simple mathematical representation of this situation. A classification of D by M is defined as a function

(5) F: M [right arrow] p(D)

where p(D) is the power set of D, i.e., the set of all subsets of D. Every device m [member of] M generates a set of classifieds F(m) [subset or equal to] D. Without any further restriction, this definition will overgenerate plenty of phenomena in linguistics (and other sciences) not interesting enough to build theories upon. For instance the 26 letters of the Latin alphabet may function as M and classify thousands of entries in a dictionary. There are endless lists of such linguistic and paralinguistic examples. In Section 2.1, I review some relevant literature and explore especially their way of depicting the concept of classification. In Section 2.2 1 will present my own definition of classification phenomena.

2.1. The approach of some major works to linguistic categorization

As alluded above, we can scarcely find any hard criterion in the literature helping differentiate between linguistic systems that are classificatory and those that are not. This issue is far from trivial. Consider for instance the set of (verbal) predicates. Every predicate selects nouns from the lexicon and incorporates them into argument structure. Why is it not commonplace in linguistics to relate this selection process to classification, as a process by which the set of nouns is classified by the set of predicates (see Lucy 2000)? Linguists do not offer many clues helping acknowledge or reject such systems. Instead, the starting point of many publications is what Grinevald describes in the following statement

... one of the first issues to be resolved is whether one is faced with a classifier system or another kind of nominal classification system. And if one identifies a classifier system as such, the next issue is to push the analysis further and to determine its actual type. (Grinevald 2000: 54)

It is presumed in this assertion that the system under scrutiny is of course classificatory. The work of scholars (or rather their ideas) can be "classified" into three basic approaches with respect to the contribution they make to a general definition of classification phenomena: the extensional, the cognitive and the distributional approach.

2.1.1. The extensional approach to linguistic categorization. The extensional approach is characteristic of many recent typological works, e.g., Adams and Conklin (1973), Denny (1976), Allan (1977: 286), Aikhenvald (2000: 4), Senft (2000:11), Grinevald (2000). This stance is basically observational inductive in nature and is based on the study of individual grammars of a representative range of languages. A number of morphosyntactic constructions in which categorization devices collocate with major word categories (i.e., nouns or verbs) have been identified. Morphosyntactic as well as semantic properties emanated from these surveys that served to put forward typologies of categorization devices. In this perspective, Allan (1977: 286) was the first to explicitly connect four hitherto disparate noun classifier types: (i) numeral classifiers, (ii) concordial (e.g., gender) classifiers, (iii) predicate classifiers, and (iv) intralocative classifiers. A recent state-of-the-art composition is Aikhenvald's massive book Classifiers, which distinguishes six crosslinguistic classifier types, all of which are categorization devices of nouns: (i) noun class systems and gender systems, (ii) noun classifiers, (iii) numeral classifiers, (iv) possessive classifiers, (v) verbal classifiers, and (vi) locative and deictic classifiers. Similar lists are presented in Senft (2000) and Grinevald (2000). The nature of each extensional method is to provide characterization through explicit enumeration of exponents. Of course, the proposed typologies are not purely extensional as they try to capture morphosyntactic as well as semantic features of categorization devices. Nonetheless, they often do not go beyond stating a correlation between morphosyntactic collocations and semantic taxonomies in this or that language. An essentially extensional line is unsuitable for the treatment of atypical or exceptional exponents when no additional metalinguistic considerations are applied. This was particularly pertinent for verb classification, which scholars were unsure whether to consider it classificatory or not (cf. McGregor 2002: 17). In contrast to the extensional method, the other two approaches mentioned below, i.e., cognitive and distributional, are intensional.

2.1.2. The cognitive approach to linguistic categorization. Burling (1965: 259-260) and Greenberg (1972) were probably the first to view individualization as the prime cognitive function of sortal and also mensural classifiers in East Asian and Southeast Asian languages. Croft (1994: 162-163) and Bisang (1993, 1996, 1999: 121) have further split up the individualization functions of sortal and mensural classifiers. Bisang labels the function of sortal classifiers by actualizing individualization (they "actualize the semantic boundaries which already belong to the concept of the noun") and that of mensural classifiers by creative individualization (they create the semantic boundaries of the concept of a noun "by applying external scales").

Individualization, however, is not the only and not even the most basic (albeit prominent) cognitive function of classifiers. Seiler (1986: 95) regards classification as the basic cognitive function of morphemes that are termed as classifiers. Relying on Frege's distinction between concept (Begriff) and object (Gegenstand), Seiler characterizes the cognitive operation of classification as "a mental operation that causes an object or a multitude of objects to fall under a concept X". It has been pointed out, however, by a number of linguists that numerous categorization devices across languages would not count as classifiers if classifiers were taken as equivalent to denotations of the more general category to which the denotata of classifieds belong. (6) In this context, Lucy (2000: 326) proposes to distinguish between classification of experience (cf. categorization of linguistic forms that correlates with a grouping of referents) and classification of linguistic form (cf. categorization of linguistic forms that does not correlate with a classification of referents). The approach of this article is not to elevate classification of experience into a prerequisite of linguistic categorization devices. The main reason is that even in undisputed instances of classified experience (e.g., classifier of one-dimensional noun referents, or classifier of noun referents with a handle), the grouping only happens with respect to certain perceived features that the referents share in common. The classifier will rarely play the role of denotatum for the class itself. (7) Of course, this topic is reminiscent of a medieval debate, the quarrel of universals about the real existence of words and things between nominalists and realists in which we do not want to get trapped.

In the frameworks of Croft and Bisang, classification is taken as a precondition of two other cognitive functions: identification and (the above) individualization. Bisang (1999: 115) defines the cognitive operation of identification as a comparison of "one particular 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". In this scheme, identification would then build on classification ("classification is a prerequisite of identification", Croft 1994: 161), but would not be required for the cognitive operation of individualization, as identification happens even without reference to entities as individuals. On the other hand, identification is a prerequisite of the operation of individualization (Bisang). These three cognitive functions can be arranged as an application hierarchy (">" means "is applied subsequently to"):

(6) classification > identification > individualization (Bisang 1999: 116)

Silverstein (1986) introduces two slightly different cognitive functions he formulates sufficiently abstract to embrace nominal and verbal classificatory phenomena: individuation and characterizability (1986: 509-511). The former roughly corresponds to the above individualization, the latter to identification. For Silverstein, a classifier (nominal or verbal) is then featured by its ability to give either "a characterizability (= identification) condition presupposed to be so of the denotatum [of the classified]" or "an individuating/quantifiability condition presupposable to be so of the denotatum [of the classified]". These two functions give rise to two types of nominal and two types of verbal classifiers (Silverstein 1986: 511):
                         Nominal               Verbal

Identification           sortal classifier     predicate-perspective
  (characterizability)                           classifier
Individua(liza)tion      mensural classifier   aspectual classifier


The verbal classifiers, which have inspired this rapprochement with nominal classifiers, are Australian-style ones (briefly mentioned above, at the end of Section 1.2). In summary, the cognitive approach succeeds in mapping global properties of classificatory devices. One open question is whether other linguistic forms that are usually not associated with classification may also manifest any of the cognitive functions discussed above. As mentioned above, there is ground for debate, for example, to regard the set of verbal predicates (or a subset of it) exhibiting the above cognitive functions in their selection process of nominal arguments. In other words, while cognitive functions constitute a necessary condition for classification, it is not clear whether they are also sufficient (cf. McGregor 2002: 17). Therefore, I will not look at cognitive functions for a tool of justifying the status of the Kam verb classifiers. However, cognitive functions play a role in subdividing different types of verb classifiers (see Section 3).

2.1.3. The distributional approach to linguistic categorization. Finally, a number of authors have chosen a distributional approach to classificatory phenomena. Few scholars have gone this path. Dixon (1982: 160, 1986: 106) distinguishes noun class systems from classifier systems by the number of noun categories each system manages to generate. For Dixon, noun class systems involve a grouping of all the nouns in a language into a smallish number of categories, generally lying between 2 to 40. Noun classifier systems exhibit a largish number of noun categories, usually being situated between 50 and 200. McGregor (2002: 16-22) is otherwise the only scholar who presents a definition of classification phenomena based on their distributional-probabilistic properties. He views a set of classifiers or categorization devices M classifying a set of lexemes D when

(7) i. members of M are of the same order of reality (i.e., a mixture of phonemes or morphemes is excluded as are allomorphs and morphemes)

ii. members of M collocate with members of D in well-defined constructions of the language such that the following properties hold:

--every member of M can collocate with some member of D;

--every item that can collocate with a member of D in these constructions must be a member of M;

iii. the cardinality of M must be greater than one but significantly smaller than that of D;

iv. there must be at least two members m and m' of M such that if [m] and [m'] denote the sets of lexemes of D that can collocate with m and m', then [m] and [m'] must be significantly different.

This grid constitutes the first algorithmlike definition of linguistic classificatory phenomena allowing one to check specific cases. As mentioned above, there are many classifierlike operations in linguistics and there is a need to separate the wheat from the chaff. McGregor offers a method that allows putting verb classification at an emancipated stage with noun classification. In my opinion, this definition can be improved in two regards. The notion "collocation of classifier and classifieds" needs a sharper portrayal. Moreover, the idea expressed in (7iv) that lexeme classes should be "significantly" different from each other is left relatively vague.

2.2. Definition of linguistic classification phenomena (revised)

The roadmap for this section is to proceed with the definition of classification phenomena in three steps: firstly, some pointed comments on the nature of the sets M and D (cf. [7i] above), secondly, a portrayal of the notion of construction in which items of M and D may collocate (cf. [7ii] above) and thirdly, a depiction of cardinality restrictions imposed on M and D (cf. [7iii]-[7iv] above).

The essence of the set M of categorization devices is variable. In the isolating East-Asian languages, it may consist of a number of (classifier) morphemes. In gender languages of the inflectional type, M may be composed of a number of morpheme vectors. Each morpheme vector would stand for one gender and consist of the inflectional paradigm for this gender, i.e., of a definite number of morphemes that reveal the idiosyncratic forms of this gender (e.g., le, ce, un, mien,... for French masculine gender). From a generalizing perspective, we define M as a set of morpheme vectors, i.e., a set of k-uples (with k a natural number). In the isolating languages of East Asia, the natural number k has the value ONE which is to say that each morpheme vector consists of ONE classifier morpheme, whereas in inflectional gender languages we have k > 1 pointing to a definite number of forms (morphemes) per gender. The set D is a set of lexemes typically corresponding to a major word category such as nouns or verbs.

The well-defined constructions mentioned by McGregor (see [7ii]) can be written as strings of word category variables tons = [X.sub.1] ... [X.sub.m] such that each [X.sub.i] is either Y or Y-Phrase with Y [member of] {noun, verb, adjective, preposition, adverb, determiner,...}. For two category variables X, Y we mark cons(X,Y) as a morphosyntactic construction in which X and Y occur as substrings. Particularly, for two given constants M = ([M.sub.1],..., [M.sub.k]) [member of] M and D [member of] D we write cons(M, D) to imply that one [M.sub.i] collocates with D in cons, i.e., cons([M.sub.i], D).

For a set of classifier devices M and a set of classified lexemes D, we suppose that there are n constructions [cons.sub.1](M, D),..., [cons.sub.n](M, D) in which M and D collocate in an exhaustive and nonredundant way. Items of D are categorized (or classified) by items of M through a pattern of collocations. Mandarin Chinese, for instance, involves nominal classifiers exactly in the context of numeral constructions, demonstrative pronouns and quantifier expressions, as given as in (8). The categorization of nouns by classifiers in Mandarin is thus established through these constructions.

(8) i. [cons.sub.1] = Num CL N;

ii. [cons.sub.2] = Dem CL N;

iii. [cons.sub.3] = Qua CL N.

As mentioned above, the n constructions must constitute exhaustive and nonredundant collocations of items in M with those in D. The n constructions have to be exhaustive in three respects. Firstly, there is no other construction, in addition to the n constructions, into which M and D have a projection (in other words, any construction cons(M, D) in which a classifier vector M collocates with a lexeme D must contain one of the [cons.sub.i]: [cons.sub.i](M, D) being a substring of cons(M, D)). Secondly, the n constructions exhaust the set M (i.e., for every classifier vector M there is actually a lexeme D that can co-occur with M in some construction cone). Thirdly, the n constructions also exhaust the set of lexemes D (i.e., for every lexeme D we have a morpheme vector M that appears with D in some construction cone). These three exhaustiveness properties find their expression in (9i)-(9iii). Moreover, the n constructions should be optimal and nonredundant in the sense that they constitute a minimal system of strings in which no sequence is included in another. This property is echoed in (9iv).

(9) Constraints on the n construction types:

i. [for all]M [member of] M, [for all] D [member] D, [for all] cons(M, D), [there exists]i [member of] {1, ..., n}: [cons.sub.i](M, D) [subset or equal to] cons(M, D),

ii. [for all]M [member of] M, [there exists]D [member of] D, [there exists]i [member of] {1, ..., n}: [cons.sub.i](M, D),

iii. [for all]D [member of] D, [there exists]M [member of] M, [there exists]i [member of] {1, ..., n}: [cons.sub.i](M, D),

iv. [for all]i, j [member of] {1, ..., n}: [cons.sub.i](M, D) [not subset] [cons.sub.j](M,D) and [cons.sub.j](M, D) [not subset] [cons.sub.i](M, D).

Explanation: [there exists] 'there exist' (existential quantifier), [for all] 'for all' (universal quantifier), [subset or equal to] (inclusion symbol).

Next, McGregor (2002: 19) poses several cardinality conditions on the sets M and D. First, the cardinality of M must be greater than one, but significantly less than the cardinality of D. Both conditions are meant to exclude frontier cases, which intuitively go against the idea of classification. If card(M) were singleton, then every lexeme would have to belong to the unique class generated by M, which is not an interesting case of classification. Yet, it is difficult to capture the subjective notion of significant. In fact, as we'll witness below, the Kam verb classifiers comply with all the formal requirements of a classification phenomenon with the exception of the cardinality of M being significantly less than D. Instead of dismissing the Kam system, we will say that its authenticity is slightly lessened (see Section 4 for more details). Even though "significantly less" appears vague, we will represent this relation by the symbol << and evaluate then concrete situations on a case-for-case basis. Hence, we may embody this cardinality property as card(M) << card(D).

For depicting McGregor's other cardinality condition, let's introduce [M], the class of lexemes generated by (or collocating with) a classifier vector M [member of] M in one of the n constructions:

(10) Convention:

For every M [member of] M let us write

[M] = {D [member of] D | [there exists]i [member of] {1,..., n}: [cons.sub.i](M, D)}.

McGregor states that for a genuine system of classifiers there should be at least two classifier vectors M, M' [member of] M generating two "significantly different" lexeme classes [M] and [M']. Significantly different means that [M] and [M'] may share some items but not many. This condition intends to exclude the following scenario. Suppose that in English we have M = {the, this}. The class [the] of all the nouns that are compatible with the greatly overlaps with [this], the class of all the nouns that can concatenate with this. In a further effort, we wish indeed to exclude such situations. Now, the two sets [M] and [M'] are significantly different, if the cardinality of [M]\[M'] (i.e., the set [m] without the set [M']) and the cardinality of [M']\[M] are significantly greater than zero. We will write this as card([M]\[M']) >> 0 and card([M']\[M]) >> 0. Let's put the various parts together and devise the following definition.

(11) General Definition of classification phenomenon:

In a natural language, a set M of classifier vectors categorizes a set D of lexemes if and only if

i. there are n constructions [cons.sub.1] (M, D),..., [cons.sub.n](M, D) such that these data satisfy the following properties:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

We may link a classification phenomenon such as given by (11) to the other general characterization at the start of Section 2 by posing the following definition:

(12) F:M [right arrow] p(D)

M [member of] M [right arrow] [M] [member of] p(D).

In the subsequent sections, we will authenticate the claim that the set M of instrumental nouns briefly sketched in Section 1.1 constitutes in fact a ser of classifiers for the set D of verbs (or more precisely for a subset D of verbs). In this procedure of corroboration, the roles of morphosyntax, semantics and distributional properties may be laid out as follows: Morphosyntactic constructions will be our point of departure (Section 3). They precisely delimit the nature and size of M and of D. As a next step (Section 4), we will demonstrate that M and D satisfy the distributional properties spelled out in (1 l) which in turn will yield classifier status to the elements of M. Only after placing the Kam VCLs on firm classificatory grounds (achieved in Section 4), will we elaborate on the semantics of verb classes (in Section 5). As explained above (in Section 2.1.2), semantic considerations alone cannot be taken as definitional. The fact that verb classes can be assigned a monadic semantic concept will be interpreted as a byproduct of classification rather than its essence or nature.

3. Morphosyntactic constructions of verb classifiers in Kam

In Kam, let D be the set of all the verbs and M the class of all the words X that may occur in one of the following two morphosyntactic constructions. These constructions are supposed to form one semantic unit relating to the meaning of counting and quantification of events: (8)

(13) Two protoconstructions of verbal classifiers in Kam

i. V Num/Qua X (postverbal)

ii. Num/Qua X V (preverbal)

In the course of Section 3, we will portray and delimit those V of D and especially those X of M that may occur in the above constructions (13i) or (13ii). By anticipation, we will call the items X of M appearing in (13i) or (13ii) verb classifiers (VCLs), even though strictly speaking this appellation will only be justified at the end of the exposition in Section 4.

A quick tour of the VCLs that may fill in the X slot in (13i) or (13ii) shows that VCLs in Kam count the number of times a denoted event occurred (respectively measure the amount of time a referenced situation persisted). With numerals or quantifiers, VCLs form verb classifier phrases which technically function as quantity adverbial phrases. (9) The two constructions (13i) and (13ii) can be further translated into the five structures shown in (14). VCLs are thus characterized by their ability to occur in one of the three quantity adverbial phrases displayed in (14).

(14) Defining morphosyntactic construction types of verbal classifiers in Kam:
i. V           (No)      Num/Qua   VCL   (Sortal VCLs, Collective
                                           VCLs, Measure VCLs,
                                           V = to 323)
ii. V          Num/Qua   VCL       BN    (Collective VCLs, Measure
                                           VCLs)
iii. Num/Qua   VCL       V         No    (Collective VCLs, Measure
                                           VCLs)
iv. V          Num       V               (Verbal Autoclassifiers)
v. to 323      Num       V               (VCLs of to 323)


V = Verb; No = Noun O-argument (cf. Dixon 1994); Num = numeral; Qua = quantifier; VCL = verb classifier; BN = bare common noun (e.g., 'table', 'water'); to 323 = versatile coverb for Patient, Beneficiary and Instrument roles.

Even though not every VCL can come out in all of these construction types, every VCL (a quite representative list is given in Section 4) may be projected in at least one of them. The rules (14i)-(14v) feature the construction types highlighted in the definition of classification above (cf. [11iii]-[11vii]). VCLs may be further subdivided into sortal VCLs (Section 3.1), mensural VCLs (Section 3.2), double NCLs/VCLs (Section 3.3) and verbal autoclassifiers (Section 3.4). In Section 3.5, I establish the versatile coverb (= verb plus preposition) to 323 that admits a wide range of sortal VCLs.

3.1. Instrumental nouns as sortal verb classifiers

Sortal verb classifiers in Kam can be portrayed as participants of (14i) and as functioning elsewhere as nouns that are typically assigned the semantic role of instrument in appropriate predications. In East Asian and Southeast Asian languages, clauses with an overt instrumental noun phrase mark this role with one or several coverbs. Coverbs are grammaticalized verbs that function as prepositions. In Kam, three coverbs are linked to instrumental NPs: au 55 'take', jon 33 'use' and to 323 'put, apply'. The semantics of the verb determines the choice of the coverb. Verbs may select one to three admissible coverbs; in (15a)-(15c) all three may be drawn in, with slightly different semantic colorations in each case. Native speakers would understand (15a)-(15c) as denoting one and not two events.
(15) a. mau 33     au 55      ljim 31      (k)an 53   nan 13.
        3P SG      COV:take   sickle       cut, saw   grass
                              Instrument
        'He cut the grass with a sickle [lit. he took a sickle and cut
        the grass].'

     b. mau 33   jon 33    ljim 31      (k)an 53.   nan 13.
        3P SG    COV:use   sickle       cut, saw    grass
                           Instrument

        'He cut the grass with a sickle [lit. he cut the grass by
        using a sickle].'

     c. mau 33   to 33     ljim 31      (k)an 53.   nan 13.
        3P SG    COV:put   sickle       cut, saw    grass
                           Instrument

        'He cut the grass with a sickle [lit. he cut the grass by
        applying a sickle].'


Moreover, the instrumental noun ljim 31 'sickle' is projected in postverbal position as VCL. When no O-argument (i.e., direct object) is specified, then it occurs with the numeral immediately after the verb, as in (16a). On the other hand, the VCL is posed after an overt No, as illustrated in (16b). In this case, it has scope over the whole verb phrase and not merely the verb. Instead of a numeral, it is possible to involve a noun quantifier such as m j en 11 'several' or oi 55 'many', as in (16c).
(16) a. mau 33   (k)an 53   ja 11     ljim 31.
        3P SG    cut, saw   NUM:2     VCL:sickle
                 Verb       Numeral   Verb classifier

        'He cut twice with a sickle.'

     b. mau 33   (k)an 53   nan 13   ja 11     ljim 31.
        3P SG    cut, saw   grass    NUM:2     VCL:sickle
                 Verb                Numeral   Verb classifier

       'He cut the grass twice with a sickle.'

     c. mau 33   (k)an 53   nan 13   oi 55        ljim 31.
        3P SG    cut, saw   grass    QUA:many     VCL:sickle
                 Verb                QUANTIFIER   Verb classifie

        'He cut the grass many times with a sickle.'


The prototypical verb that can collocate with a VCL describes an event in which two objects are brought into touch contact (see Section 5). From a phonological perspective, the verbs in Kam that can be categorized by VCLs are monosyllabic without exception. They greatly overlap with physical core-activity verbs, which are monosyllabic too. Similarly, most VCLs count only one syllable as well. Though, there are two disyllabic exceptions:
(17) mau 33   non 53   ja 11   con 3 wan 35.
     3P SG    hack     NUM:2   punching pin

     'He hacked twice with the punching pin.'

(18) mau 33   tap 11   ja 11   ta 11 k w iu 53
     3P SG    hack     NUM:2   weight of steelyard

     'He pressed twice with the weight of steelyard.'


Most verbs admitting VCLs are monotransitive, (10) though a few (perhaps 4-5) exceptions of intransitive verbs involving VCLs exist, e.g., the verbs tam 31 'stamp' in (18) and sui 53 'sit' in (19). The latter bars numerals other than 'one' from forming VCL-phrases.
(19) mau33   tam 31   ja 11   tin 55.
     3P SG   stamp    NUM:2   foot

     'He stamped twice with his foot.'

(20) mau33   sui 53   i 55    sen 31.
     3P SG   sit      NUM:1   buttocks

     'He sat once on his buttocks.'


Next, the question of whether, for a given verb, instrumental nouns stand in a one-to-one correspondence with VCLs has a negative answer. Even though every sortal VCL in Kam is derived from an instrumental noun, not every instrumental noun can also serve as a VCL. Two rules of thumb indicate the usage of an instrumental noun as VCL (see also Section 5): an instrumental noun may not emerge as VCL when the instrumental relationship to the verb is highly abstract or indirect as in (21), or when the verb itself is a complex activity in which the instrument only assists during one of several phases, cf. (22). In (21), a written dispatch appears as the instrument of a communication (lit. tell someone something through mail). As the instrumental relationship is relatively abstract, the instrumental noun sen 31 'letter' is prohibited to occur as VCL of lep 11 'tell'. Concerning (22), the verb tan 55 'eat' can project m j a 11 'hand' as an instrument argument, but since eating also consists of chewing and swallowing for which a hand does not offer assistance, the Kam noun for 'hand' cannot function as a VCL here. (Note, however, that san 31 'letter' in (21) and m i a 11 'hand' in (22) may occur with one of the coverbs given above.)
(21) * mau 33   lep 11   i 55     sen 31.
     3P SG      tell     NUM: 1   written letter

     'He told one written letter.'

(22) * mau 33   tan 55   i 55    mja 11.
     3P SG      eat      NUM:I   hand

     'He ate one hand.'


3.2. Mensural verb classifiers

Sortal noun classifiers have been put into contrast with mensural NCLs through the cognitive function of individualization: Sortal NCLs actual ize shape boundaries inherently present in the noun concept, while mensural NCLs create shape boundaries that are not intrinsic to the noun concept (cf. Croft 1994, Bisang 1999). This difference is mirrored in the verbal realm and was first pioneered by Matthews and Yip (1999) who discriminated between sortal and mensural VCLs in Cantonese. Sortal VCLs actualize temporal or phasal boundaries, which are implicit in the verb concept, while mensural VCLs create temporal boundaries which are not inherent to the verb. The verb 'beat', for example, has a smallest phase in which two objects enter in collision. Sortal VCLs of 'beat' actualize this smallest phase. Conversely, the verb 'wait' has no smallest phase. A mensural VCL like 'two hours' creates a temporal boundary for 'wait'. I will advance a more detailed exposition of these features in Section 5.

Morphosyntactically, mensural VCLs are not allied with a particular construction type (14i)-(14v). For each construction, there are a number of VCLs that may be involved with it. Like mensural NCLs, mensural VCLs come in two varieties: as collective VCLs and as measure VCLs (cf. Bisang 1999: 122). In the nominal realm, collective classifiers create artificial boundaries for entities that have inherent minimal parts: 'a group of people', 'a collection of roses', 'a herd of oxen'. Entities without minimal parts do not admit collective classifiers: * 'a group of water', * 'a collection of air'. Contrastively, measure NCLs collocate with nouns referring to entities without minimal parts: 'a bottle of water', 'a cubic meter of air'. In Section 5, I will develop the concept of minimal parts of events. The order in this section is to expose for Kam collective VCLs in Subsection 3.2.1, and measure VCLs in Subsection 3.2.2.

3.2.1. Collective verb classifiers. Kam exhibits two collective VCLs: tau 53 'time, occasion' (11) and tan 53 'time, round'. The VCL tau 53 interacts with verbs possessing individuable phases or so-called minimal parts (cf. 5) and establishes a grouping of these phases. For instance, 'beat on three instances' does not necessarily entail that agent and patient entered into collision exactly three times, but rather that a beating event came to pass on exactly three occasions. The collective VCL tau 53 permits expressions of distributive beating in which on each occurrence agent and patient crash a certain number of times. Sentence (23) below epitomizes this kind of grouping that mirrors closely the condition of nominal phrases such as 'three groups of five people'.

Morphosyntactically, the VCL tau 53 can be projected in the three construction types (14i)-(14iii) above. It appears after the verb in two cases. When no O-argument is specified as in (23a), or when the O-argument is definite (e.g., a personal pronoun) as in (23b), then tau 53 appears in sentence-final position and is associated with structure (14i).
(23) a. mau 33   heu 35   ja 11     tau 53.
        3P SG    beat     NUM:2     VCL:time
                 Verb     Numeral   Verb classifier

     'He beat on two occasions.'

(23) b. mau 33   heu 35   jau 11   ja 11     tau 53.
        3P SG    beat     1P SG    NUM:2     VCL:time
                 Verb              Numeral   Verb classifier

'He beat me on two occasions.'


However, when the O-argument consists of a bare noun (which is understood to be nonreferring and indefinite), then the VCL is positioned between the verb and the O-argument, as in (24), expounding the construction type (14ii).
(24) mau 33   sak 55   ja 11     tau 53            uk 323.
     3P SG    wash     NUM:2     VCL:time          garment
              Verb     Numeral   Verb classifier

     'He washed clothes on two occasions.'


The VCL tau 53 occurs before the verb, when the verb phrase is structurally rich containing for instance a verb serialization (cf. example [25]), or a sortal VCL (cf. example [26]). This use is linked to construction type (14iii).
(25) ke 35        kun 11       tau 53            t h a 453
     3P PL        QUA:many     VCL:time          ascend
                  Quantifier   Verb classifier   Verb Serialization

     au 31        ta 53        tan 55   tum 55
     LOC:at, in   LOC: there   RECL     gather

     'They went up to that place on many occasions for a meeting.'

(26) mau 33   ja 11     tau 53            cek 13
     3P SG    NUM:2     VCL:time          all
              Numeral   Verb classifier

     heu 35   jau 11    sam 35    cui 11
     beat     1P SG     NUM:3     VCL:fist
     Verb               Numeral   Verb classifier

    'On each of two occasions, he beat me three times with his
    fist(s).'


The other collective VCL ten 53 'time, round' is semantically more restricted. Like the VCL tau 53, it expresses a grouping of individuable phases of the verb. The class of compatible verbs is restricted to a few verbs of affect such as heu 35 'beat', it 31 'bite', wa 33 'criticize', etc.

Measure VCLs, the other subcategory of mensural VCLs, can be coupled with almost every verb (either with or without minimal parts), provided that it refers to an activity or state without inherent endpoint (see next section).

3.2.2. Measure verb classifiers. Measure VCLs are time units, either natural or man made, indicating the duration of an event or state. Vendler (1967) established tests for ascertaining the situation type (Aktionsart) of a clause. One of the tests assesses the compatibility of clauses with FOR-adverbials (e.g., 'for two hours'). It transpires that activities and states admit FOR-adverbials, while achievements and accomplishments reject them. The Kam measure VCL-phrases echo the function of FOR-adverbials in English. They comply with underlying clauses that are activities or states (e.g., examples [27] and [28] below), but decline accomplishment clauses referring to events with an inherent endpoint, as (29). Example (29) without the VCL phrase would correspond to an accomplishment clause.
(27) sai 35   mau 33   wa 33   i 55      ha 35.
     let      3P SG    say     NUM: 1    VCL: instant
                       Verb    Numeral   Verb classifier

     'Let him say something for an instant.'

(28) mau 33     nau 33       au 31     sem  31
     3P SG      COV: be at   LOC: at   room

     nai 33     nak 35       ja 11     man 55.
     DEM:PROX   sleep        NUM:2     VCL: day
                Verb         Numeral   Verb classifier

     'He slept in this room for two days.'

(29) * mau 33   tan 55   sam 35   nan 55   tui 55   i 55
       3P SG    eat      NUM:3    NCL      fruit    NUM:I
                Verb                                Numeral

     cen 33.
     VCL: instant
     Verb classifier

* 'He ate three fruits for one period of time.'


Morphosyntactically, measure VCLs always occur after but never before the verb. Like the VCL tau 53 'time' (see previous section), measure VCLs are projected in clause-final position after the O-argument, when the O-argument is definite, exhibiting thus the construction type (14i). Measure VCLs appear between the verb and the O-argument, when the O-argument is an indefinite bare noun. This usage stands for the construction type (14ii).
(30) mau 33   heu35   jauH    i 55      ha 35.
     3P SG    beat    1P SG   NUM:I     VCL:instant
              Verb            Numeral   Verb classifier

'He beat me for an instant.'

(31) mau 33   sak 55   i 55      man 55            uk 323.
     3P SG    wash     NUM: 1    VCL:day           garment
              Verb     Numeral   Verb classifier

'He washed clothes for one day.'


There exist ca. 10 measure VCLs (time units) in the Kam lexicon. They vary in length from 'instant' to 'lifespan', see the following table:

3.3. Double classifiers of nouns and of verbs

While parallels between NCLs and VCLs in East Asian languages are undeniable, several scholars also explored the question of whether NCLs and VCLs partially overlap, especially whether there are morphemes with double involvement as NCL and as VCL (see Paris 1989: 4-5; Matthews and Yip 1999: 11-12; Matthews and Leung 2001, Yang 2001: 129-137). None of the sortal VCLs actually emerges as NCL, but scholars have detected an overlap potential for both mensural VCLs and sortal/mensural NCLs that occur in O-argument position of the clause. This relationship between NCLs and VCLs is also reflected in Kam.

3.3.1. Sortal verb classifiers do not function as noun classifiers. Sortal VCLs (cf. Section 3.1) lexically stem from nouns that denote physical instruments. They have no involvement as categorization devices of other nouns in constructions such as 'Num VCL N'.

3.3.2. Most mensural verb classifiers do function as noun classifiers. Concerning mensural VCLs, the issue of overlap has been raised for Mandarin's VCL ci 'time' (corresponding to Cantonese's VCL ci and Kam's VCL tau 53 'time'). Yang (2001: 129-137) suggests that Mandarin's mensural VCL ci divides the class of nouns into three subclasses: class-1, class-2 and class-3. Class-1 groups nouns that collocate with NCLs, in both preverbal and postverbal sentence position. Moreover, class-1 nouns may compound with VCLs in postverbal but not in preverbal position (cf. example [32]). Semantically, they do not denote events but only physical entities such as 'table', 'book', etc. Class-2 consists of nouns that may collocate with NCLs and the VCL ci in both syntactic environments: before and after the verb. They may refer alternatively to physical objects or to events. Consider, for example, 'film', 'rain(fall)', etc. Class-3 nouns cannot co-occur with NCLs, but may only collocate with VCLs in both preverbal and postverbal position. These nouns do not refer to physical objects but only to events or states such as 'life need', 'attack', etc. Kam's VCL tau 53 'time' maps nouns into a similar grid. Witness the following examples:
(32) a. Class-1         sam 35   mun 31   mun 31   tan 55   k hWau 13.
        Preverbal NCL   NUM:3    NCL      person   drink    wine

                        'Three people drink wine.'

     b. Class-1          jau 11   hem 53   sam 35   mun 31   nen 11.
        Postverbal NCL   1P SG    call     NUM:3    NCL      person

                         'I call three people.'

     c. Class-1          * sam 35   tau 53     k hwau 13   cek 13
        Preverbal VCL    NUM:3      VCL:time   wine        all

                         lai 55   tan 55.
                         good     drink

                         'Three times wine is delicious.'

     d. Class-1          mau 33   tan 55   sam 35   tau 53
        Postverbal VCL   3P SG    drink    NUM:3    VCL:time

                         k hW au 13.
                         wine

                         'He drinks wine on three occasions.'

(33) a. Class-2          sam 35   nat 55   nui 55   tok 55   lui 33
        Preverbal NCL    NUM:3    NCL      snow     fall     go down

                         ma 35.
                         come

                         'Three snowflakes fell down.'

     b. Class-2          tok 55   sam 35   nat 55   nui 55.
        Postverbal NCL   fell     NUM:3    NCL      snow

                         'It snowed three flakes.'

     c. Class-2          sam 35   tau 53     nui 55   cek 13   tok 55
        Preverbal VCL    NUM:3    VCL:time   snow     all      fall

                         li 323   na 55.
                         have     thick

                         'All three snowfalls were abundant.'

     d. Class-2          tok 55   sam 35   tau 53     nui 55.
        Postverbal VCL   fell     NUM:3    VCL:time   snow

                         'There were three snowfalls.'

(34) a. Class-3          * i 55   nan 55   wen 453    nan 11
        Preverbal NCL    NUM:1    NCL      business   difficult

                         pau 31
                         guarantee

                         'No guarantee for a business.'

     b. Class-3          * mau 33   li 323   i 55    nan 55
        Postverbal NCL   3P SG      have     NUM:1   NCL

                         wen 453
                         business

                         'He has one business.'

c. Class-3               ja 11   tau 53     wen 453    lai 55
   Preverbal VCL         NUM:2   VCL:time   business   good

                         we 31.
                         do

                         'The business was successful on two
                         occasions.'

d. Class-3               jau 11   nim 35   mau 33   we 31   i 55
   Postverbal VCL        1P SG    with     3P SG    do      NUM:1

                         tau 53     wen 453.
                         VCL:time   business

                         'I did business with him on one occasion.'


The quintessence of the above examples is that the VCL tau 53 'time' can co-occur with certain nouns and thus may qualify for the status of NCL. Though, it is questionable whether in (32d) the VCL really modifies and individualizes the noun [k.sup.hw][au.sup.13] 'wine'. Since (32d) is semantically equivalent to (32e) below, it is more plausible to regard tau53 in (32d) as a backtracking VCL rather than as a forward-tracking NCL.
(32) e.                  mau 33   sam 35   tau 53     tan 55
                         3PSG     NUM:3    VCL:time   drink

                         k hw au 13.
                         wine

                         'He drinks wine on three occasions.'


However, examples (33c)-(33d) and (34c)-(34d) show genuine involvement of tau 53 'time' as a classifier of nouns. The sort of nouns that tau 53 may categorize concern abstract entities referring to nominalized events or states. The above syntactic constructions were tested with the informant on a wide range of nouns. The table below displays the distilled results.

The other measure verb classifiers displayed in the Table 1 (especially the macro time units such as 'day', 'year', etc) behave as the VCL tau 53 and can be involved with class-3 nouns that refer to individualized events such as 'labor'.
(35) a. Class-3         ja 11    man 55        on 55  nai 53
        Preverbal VCL   NUM:2    VCL:day       work   DEM:PROX
                        nan 11  sen 13 hu 13.
                        really  tired
                        'Working two days causes a great fatigue.'

     b. Class-3         mau 33  we 31         ja 11    man 55
        Postverbal VCL  3P SG   do, make      NUM:2    VCL:day
                        on 55.
                        work
                        'He works for two days.'


The VCL ha 35 'moment' in Table 1 is an exception as it cannot occur in preverbal but only in postverbal position (hence expounding only the construction types [14i] and [14ii]). It is one of the very few mensural VCLs that cannot be involved with any type of noun.

Another classifier with a double nominal and verbal identity is ten 53 (see Section 3.2.1). As a double classifier, it significantly differs from tau 53 'time' because its nominal and verbal meanings are disconnected. As NCL, ten 53 is a mensural classifier of 'food' implying the meaning of 'meal'. When involved as a VCL, it refers to a vague period of time similar to English 'round'. 12
(36) a. mau 33  ti 55  sam 35   ten 53           eu 31.
        3P SG   eat    NUM:3    NCL: meal        food
                Verb   Numeral  Noun classifier  Noun
        'He eats three meals.'

     b. mau 33  heu 35  jau 11  sam 35   ten 53.
        3P SG   beat    1P SG   NUM:3    NCL: round
                Verb            Numeral  Verb classifier
        'He beat me in three rounds.'


3.3.3. Postverbal noun classifiers do not function as verb classifiers. One could argue that a quantized (13) noun in the O-role whose head noun is omitted should be understood as VCL, since it articulates a boundary of an event, in a similar way a measure VCLs communicates the duration of an event. Judge example (37a) where the head noun tui 55 'fruit' of the classifier nan 55 is ellipsed and example (37b) where it is spelled out:
(37) a. mau 33  tan 55  sam 35   nan 55.
        3P SG   eat     NUM:3    NCL
                Verb    Numeral  Noun classifier
        'He ate three (fruits).'

     b. mau 33  tan 55  sam 35   nan 55           tui 55.
        3P SG   eat     NUM:3    NCL              fruit
                Verb    Numeral  Noun classifier  Noun
        'He ate three fruits.'


Clauses such as (37a) are well-formed in Kam, though they need some contextual input in order to identify the referent of nan 55 unambiguously. In any case the hearer has a clue that the wide-range noun classifier nan 55 refers to an edible object, as it collocates with the verb for 'eat'. It seems inappropriate to regard the classifier in (37a) as a VCL or as a double nominal and verbal classifier, because it is the result of an eclipse. (37a) is semantically equivalent to (37b) where the NCL nan 55 is dependent on the noun tui 55 'fruit' and not on the verb.

The same line of argumentation applies to almost every sortal and mensural noun classifier that collocates with a noun in the O-role. Notably, there are a number of mensural noun classifiers that bear some surface similarity with the sortal VCLs explored in Section 3.1. Depending on the context, these mensural NCLs modify indefinite nouns in the O-role or in the Instrument role.
(38) a. mau 33  au 55     tau 55  ceu 13  mas 55.
        3P SG   COV:take  pot     cook    dish, vegetable
                          Instrument
        'He cooked vegetables with a pot.'

     b. mau 33  ceu 13  ja 11    tau 55.
        3P SG   cook    NUM:2    NCL: pot
                Verb    Numeral  Noun classifier
        'He cooked two pots (of vegetables).'

     c. mau 33  ceu 13  ja 11    tau 55           ma 55.
        3P SG   cook    NUM:2    NCL: pot         vegetables
                Verb    Numeral  Noun classifier  Noun
        'He cooked two pots (of vegetables).'


Example (38a) projects the noun tau 55 'pot' into an instrumental role. In (38b), it appears as mensural NCL whose head noun is ellipsed, while in (38c) the head noun is spelled out. Example (38b) reveals some surface similarity with sortal VCLs (cf. Section 3.1), but sortal VCLs can never appear in a structure such as (38c). It is hence more appropriate to understand tau 55 'pot' as a forward-tracking NCL rather than as a backward-tracking VCL. A number of similar instances of instrumental NCLs are represented in the following table.

3.4. Verbal autoclassifiers

The term autoclassifier was first forged by Matisoff (1973: 89) for Lahu, a Tibeto-Burman language spoken in Thailand, Myanmar and China. He alluded herewith to the ability of certain nouns to function as their own NCL, i.e., to occur in constructions such as the following:
(39) Lahu
     z[??]  21 te 54  z[??] 21
     house  NUM:1     house
     'one house'

(40) Lahu
     q h a[??] 45  ni 45  q h a[??] 45
     village       NUM:2  village
     'two villages'


Scholars noted that in many East Asian and Southeast Asian languages verbs might function as autoclassifiers as well, expounding the construction type (14iv) (see Li and Thompson 1981: 232-236; Matthews and Yip 1999: 9-10). The range of numerals and quantifiers permitted to compound with autoclassifiers varies across East Asian and Southeast Asian languages. In Mandarin, only the numeral 'one' is associated with verbal autoclassifiers. Mandarin's autoclassifiers are sometimes termed delimitative aspect (Li and Thompson 1981: 232). In Cantonese, verbal autoclassifiers appear to concatenate with the whole range of numerals, though quantifiers such as 'several', 'many' are prohibited (cf. Matthews and Yip 1999: 9-10). Kam demonstrates even greater freedom than Cantonese and admits verbal autoclassifiers with any numeral (see [41a]) or quantifier (e.g., m j en 11 'several', oi 55 'many'), as in (41b).
(41) a. mau 33  t h ik 13  jau 11  ja 11    t h ik 13.
        3P SG   kick       1P SG   NUM:2    kick
                Verb               Numeral  Verbal autoclassifier
        'He kicked twice (lit. he kicked two kicks).'

     b. mau 33  t h ik 13  jau 11  oi 55       t h ik 13.
        3PSG    kick       1PSG    Qua:many    kick
                Verb               Quantifier  Verbal autoclassifier
        'He kicked many times.'


The number of verbal autoclassifiers in Kam is limited and lies around 6-8 exponents. Now, certain strings in Kam have a double nominal and verbal meaning similar to related words in English, e.g., siu 53 'chisel' (verb) and siu 53 'chisel' (instrumental noun). Such strings can be interpreted as verbal autoclassifiers and as sortal VCLs.
(42) siu 53  i 55      siu 53
     chisel   NUM:1    VCL:chisel
     Verb     Numeral  Verbal autoclassifier/Verb classifier
     'chisel once with a chisel.'


The last three entries in Table 4 represent similar cases of alternative autoclassifiers and sortal VCLs. In each case, they relate to homophonic strings of verbal and nominal meanings. The first four records of Table 4 pertain to verb-only uses and function as autoclassifiers but not as sortal VCLs.

In a more general perspective, the emergence of autoclassifiers in both the nominal and the verbal realms may be motivated by the cognitive function of classification (see Seiler 1986: 95; Bisang 1999: 115; Section 2.1.2 of this article). The speaker wishes to communicate default categories of various lexemes. The default category of each lexeme is the conceptional class to which it belongs. 'House' belongs to the conceptional class of HOUSE-entities; 'kick' expounds the conceptional class of KICK-events. Of course, providing a default category is not a genuine instance of classification, since for classificatory phenomena we assumed in Section 2.2 (under [11vii]) that the number of categories should be significantly lower than the number of classified items. If we admitted categorization of the lexemes by themselves, there would be as many categories as classified lexemes, thereby contradicting (11vii). (14) Even though autoclassifiers do not achieve classification of items in the obvious sense, we will nonetheless continue the tradition of nomenclature since autoclassifiers share the syntactic environment of sortal VCLs.

3.5. Classifier constructions involving the versatile verb to 323

Kam exhibits a versatile morpheme to 323 which as plain verb means 'put', 'apply', 'use' and as preposition (coverb) marks nouns for their semantic roles of benefactive and instrument (see example [15c] above). Moreover, it can appear in two types of verb classifier constructions. As main verb, it collocates with a wide range of sortal VCLs (cf. construction type [14i] above), though by far not with all. It also co-occurs with a limited number of verbs (perhaps 20-30) which function as a kind of verbal classifier of to 323 (cf. unique construction type [14v]), partially mirroring the operation of autoclassifiers in the previous section. Witness example (43).
(43) a. mau 33  to 32     sam 35   tok 323.
        3PSG    put; use  NUM:3    hoe
                Verb      Numeral  VCL
        'He used (e.g., dug with) a hoe three times.'

     b. mau 33  to 323    sam 35   pai 53.
        3PSG    put; use  NUM:3    pay respect, worship
                Verb      Numeral  Classifying verb
        'He worshiped three times (to a god).'


In the table below, 15 representative examples for each construction type illustrate this idiosyncratic involvement of to 323.

4. Distributional properties of verb classifiers in Kam

The next step of the exposition is to display evidence for the claim that the set of Kam VCLs identified in (14) through a number of constructions constitutes in fact a set M of categorization devices for D the set of verbs (or more precisely a subset of verbs). We draw on our principled account of classificatory phenomena by taking distributional properties as the sole definitional features (cf. Section 2.2). While reflecting on the nature of the sets M and D, it appears that mensural VCLs cannot participate in this exercise of mapping distributional properties. Since they reveal no significant selectional restrictions on verbs, they would only blur the distributional picture yielded by sortal VCLs. The same also holds for verbal autoclassifiers. Actually, this situation is strictly parallel to noun classification where mensural NCLs only carry the classifier label because of morphosyntactic and cognitive (but not distributional) commonalities with sortal NCLs. Our considerations in this section will therefore only focus on M composed of sortal VCLs, which are characterized by their involvement in construction type (14i).

Also the set D of classified verbs needs some a priori restriction. D is the set of classified verbs. Both M and D are semi-open or open sets. Distributional features can only be given proportionally not in absolute figures. In Section 3.1, we have demonstrated that sortal VCLs are semantically derived from instrumental nouns in virtually all cases. For a delimitation of D, we would therefore have to investigate those verbs that admit the semantic role of instrument. With my informant I explored ca. 300 basic verbs in Kam as to whether they might display instrumental participation roles. Approximately 80 verbs of the sample have this ability admitting a total of 60 different instrumental nouns. Some of the 60 nouns can occur as the instrument of several verbs. A number of the 80 verbs also admit several instrumental nouns.

From the data in Section 3, it is obvious that the class of verbs which may collocate with sortal VCLs is included in the class of verbs with an instrumental participation role. Every verb admitting a sortal VCL may involve this VCL also as an instrumental noun prefixed by an appropriate coverb. In particular, verbs without instrumental participation role cannot co-occur with sortal VCLs. On the other hand, not every verb with an instrumental noun may project this noun also as a sortal VCL. My lexical survey found that about 42 of the above 60 instrumental nouns might function as sortal VCLs combining with approximately 72 of the 80 verbs. Of course, these are only indicative values, not absolute numbers albeit representative ones. As we intend to substantiate the claim that the system of sortal VCLs in Kam constitutes a classificatory phenomenon, we should depart from the set M of 42 VCLs and the set D of 72 classified verbs. Both are given in detail in the table below:

Let us check now whether this grid of VCL-classes exhibits the distributional features of a classificatory scheme as required by (11) in Section 2.2. We will therefore have to confirm that in the context of Kam VCLs there are n constructions cons 1(M,D), ..., cons n(M,D) matching the properties (11ii)-(11viii). The sole construction cons 1 (M, D) which projects verbs of D with VCLs of M is the one mentioned in Section 3:

(14) i. cons 1(M,D): V (N O) Num/Qua VCL

with D = V and M = VCL. (14i) sums up four constructions depending on the presence or absence of N O and Num/Qua. Without going into technical details, it is obvious that these four constructions constitute a representation in which items of M and D meet each other in an exhaustive and nonredundant way. Consequently, the properties (11ii)-(11v) on construction types are fulfilled.

Some of the cardinality constraints (11vi)-(11viii), though, are more problematic. To start with, the constraint (11vi) is manifestly verified, as the cardinality of M is 42, greater than 1. The constraint (11viii) necessitates that at least two VCL-classes be significantly different. Choosing m = m j a 31 'knife' and m' = tin 55 'foot' from Table 6 above, it appears that the verb classes generated by these VCLs are disjoint, thus confirming (11viii). The only delicate condition is (11vii) which states that the cardinality of M must be significantly smaller than the cardinality of D. The cardinality of M in the Table 6 is ca. 42 and that of D about 72. The set of classifiable verbs is relatively small, even though it represents a prominent segment of the basic vocabulary. Moreover, the ratio between M and D is 58%, a much higher percentage than in noun classifier systems. The number of sortal NCLs in Kam lies around 50-60 with the ability of categorizing virtually all count nouns. The ratio between M and D for NCLs is well below 10%. The high ratio of about 58% slightly hampers the system of VCLs in Kam to come out as strong and genuine classificatory system, even though there exist no crosslinguistic standard yet. As a classificatory system is defined in (11) not by a single property but by a grid of features, it appears plausible to argue that the Kam system slightly falls beside the most prototypical center, while still belonging to a reasonable radius as opposed to lying on the fuzzy fringes. Consequently, it appears justified to understand the system of Kam VCLs as classificatory.

5. The semantics of classifiable and nonclassifiable verbs in Kam

Distributional facts thus suggest that the system of sortal VCLs in Kam is classificatory. It would be interesting now to probe whether to this classification of linguistic form also corresponds a classification of experience (cf. Lucy 2000: 326; see Section 2.1.2). Again, the position held in this article is that semantic features of classes of classifieds are more like a by-product rather than the essence of classification. As demonstrated in the previous section, the verbs unable to collocate with VCLs clearly outnumber those that can. A number of reasons for verbs to disallow VCLs are outlined in Section 5.1, while in 5.2 and 5.3 the semantics of verb classes generated by VCLs is scrutinized.

An inspection of Table 6 divulges that most if not all the verbs listed there describe physical events in which two objects enter into touch contact either once (e.g., 'beat') or iteratively (e.g., 'hammer'). This touch contact is generally brought about by an instrument. The idea of touching is fundamental to the proposed analysis in this section. Many physical activity verbs (or rather their denotations) can actually be decomposed into atomic phases in which two objects meet in various ways. Consider the activity of calling.
(44) mau 33  pan 55  sam 35   so 33.
     3P SG   call    NUM:3    voice, breath
             Verb    Numeral  VCL
     'He called three times (lit. he called three voices).'


During a CALL-event, breath flows through the articulation organs and constrictions at various places create sounds. Each constriction can be deemed as an atomic TOVCH-event in which the air meets the articulation parts. After all, speaking is nothing else than an iterative constricting and widening of the articulation channel. (15) The VCL tracks and individualizes each verb phase that expresses a touch contact. The idea of touch contact is somehow underlying all the verbs of Table 6, let it be 'beat', 'call' or 'row'.

Now there are parallels in the nominal realm for which scholars have actively debated whether mass nouns refer to objects with minimal parts or without, cf. Bunt (1979: 255-256), Krifka (1989: 40). Minimal parts can be illustrated for the mass noun 'water' in the following way. While a molecule of water still passes for water, a proton of a water molecule probably does not. Minimal parts do not only constitute a controversial issue at the micro or nanolevel, but even at the macrolevel, where objects are palpable without special technological assistance, arguments should be advanced carefully (consider for example mass terms like 'furniture' or 'cattle'). The consensus now widely reached by scholars is the formula of Bunt (1979: 255-256, 1985: 45-46) who questioned the relevance of minimal parts at the linguistic level in these terms: "mass nouns provide a way of speaking about things as if they do not consist of discrete parts" (Bunt, 1985:45 [italics in the original]; see also 1979: 255).

Verbs mirror the situation of nouns by virtue of the fact that they may point to events with minimal parts. The smallest components of an event (if there are any) are TOUCH-events in which two objects enter into contact in various ways. Not every verb relates to an event with minimal parts. Stative verbs like 'wait', for example, do not invoke any minimal touch contact. The thumb rule for verbs to admit (sortal) VCLs is that (a) they must sanction the semantic role of instrument; (b) they denote events which are either TOUCH-events or multiple iterations of TOUCH-events.

In Section 5.1, we will scrutinize verbs that do not admit any VCL, because they fail on either (a) or (b). In Section 5.2, we will demonstrate that verbs that admit VCLs satisfy both (a) and (b). Finally, there are a few sortal VCLs that combine with verbs, even though neither (a) nor (b) are met. They constitute a problematic rest category (Section 5.3).

5.1. Events that are not TOUCH-events or multiple iterations of TOUCH-events

Many verbs do not point to TOUCH-events or iterations of TOUCH-events for a variety of reasons. Certain events may be broken down into a number of elementary phases such that each phasal atom is tantamount to a TOUCH-event of a different type (Section 5.1.1). Moreover, there are verbs that are indeterminate with regard to the elementary phases they (or the denoted events) are composed of. This type comprises abstract verbs such as 'make', 'hinder', etc (see Section 5.1.2). Finally, stative verbs refer to situations that do not consist of any phase resembling a TOUCH-event (see Section 5.1.3).

5.1.1. Events with diverging minimal parts. Let us regard the verb 'eat' as a prototype of this category. An EAT-event is an iteration of three subevents. One subevent comprises the insertion of food into the mouth by means of a spoon or by chopsticks. The insertion of food is a TOUCH-event. Another subevent of eating is chewing. Chewing is the iteration of a TOUCH-event whereby the upper and lower rows of teeth touch the food. The third subevent of an EAT-process is swallowing. Swallowing transfers food to the stomach and consequently forms a TOUCH-event. An EAT-event incorporates therefore three minimal parts: inserting, chewing and swallowing. Tools that are generally associated with EAT-events (e.g., 'hand', 'spoon' or 'chopsticks') concern, strictly speaking, only the first subevent of eating, i.e., inserting. A hand or a spoon does not assist the action of chewing or swallowing. Here lies the explanation of why the Kam verb tan 55 'eat' cannot collocate with VCLs, even though admitting the participation role of instrument.
(45) a. mau 33  au 55     mja 11      tan 55  keu 31.
        3PSG    COV:take  hand        eat     food
                          Instrument
        'He ate with his hands.'

     b. *mau 33  tan 55  sam 35   mja 11.
        3P SG    eat     NUM:3    hand
                 Verb    Numeral  VCL
        'He ate three hands.'


The involvement of a VCL always distills (i.e., individualizes) an event as iteration of the same TOUCH-event. Consequently, since tan 55 'eat' refers to the sum of three different TOUCH-events (i.e., inserting, chewing and swallowing), there is no VCL that can individualize an EAT-event as an iteration of the same TOUCH-type event. The involvement of the VCL m j a 11 'hand' in (50b) produces thus ungrammaticality.

Many events resemble EAT-events and consist not only of one minimal part (= one TOUCH-type event), but of multiple minimal parts. The activity of weaving with a loom is composed of an array of minimal parts resulting in an impossibility of involving VCLs, see (46a) and (46b). The same applies to the activity of buying, cf. (47a) and (47b), and to the activity of burying, cf. (48a) and (48b), even though they can be carried out with various instruments and tools.
(46) a. mau 33  au 55     tak 323  tam 323  ja 55.
        3P SG   COV:take  loom     weave    cloth
        'He weaved cloth with a loom.'

     b. *mau 33  tam 323  (ja 55)  i 55   tak 323.
        3P SG    weave    cloth    NUM:1  loom
        'He weaved cloth one loom.'

(47) a. mau 33  jon 33   sin 11  tei 323  nu 453.
        3P SG   COV:use  money   buy      pig
        'He bought a pig with money.'

     b. *mau 33  tei 323  i 55   sin 11.
        3P SG    buy      NUM:1  money
        'He bought one money.'

(48) a. k h e 35  au 55     sei 11  mok 55   mau 33
        3P PL     COV:take  coffin  bury 3P  SG
        'He buried with a coffin.'

     b. *k h e 35  mok 55  (mau 33)  i 55    sei 11.
        3P PL      bury    3P SG     NUM: 1  coffin
        'He buried one coffin.'


5.1.2. Indeterminate events. A number of verbs such as 'work', 'hinder', 'cause', etc refer to events in a vague and abstract way. Even though they relate to physical activities and admit the semantic role of instrument, they are indeterminate about whether the denoted event is composed of one unique minimal part or multiple minimal parts. In Kam, indeterminate verbs cannot take potential VCLs that derive from instrumental noun phrases, because their semantics does not make it plain whether they consist of one or multiple minimal parts.
(49) a. mau 33  au 55     m j a 11  we 31     on 55.
        3P SG   COV:take  hand      make, do  labor
                          Instrument
        'He works with his hands.'

     b. *mau 33  we 31     on 55  sam 35   m j a 11.
        3P SG    make, do  labor  NUM:3    hand
                 Verb             Numeral  VCL
        'He works three hands.'


5.1.3. States. Stative verbs point to states that do not display minimal parts (= TOUCH-type events), nor involve instruments. Consequently, stative verbs do not collocate in Kam with any of the sortal VCLs displayed in Table 6. One exception is the atypical expression 'sit one buttock' in (50b). Its extraordinary status can be seen from the observation that 'buttock' does not function as instrumental noun since it cannot combine with any of the instrumental coverbs, see (50a). Furthermore, 'buttock' can only collocate with the numeral 'one', see (50b).
(50) a. *mau 33  au 55     sen 31      sui 53.
        3P SG    COV:take  buttocks
                           Instrument
        'He sits with his buttocks.'

     b. mau 33  sui 53  i 55     sen 31.
        3P SG   sit     NUM: 1   buttocks
                Verb    Numeral  VCL
        'He sits on his buttocks for a while.'


5.2. Events that are TOUCH-events or multiple iterations of a TOUCH-event

McGregor (2002: 29-34) identifies three semantic features that outline how verbs in Australian languages are categorized: vectorial configuration (spatial orientation of the action path), Aktionsart (lexical aspect) and valency (lexical transitivity). (16) Among these categorization principles, vectorial configuration gets closest to the idea of TOUCH-contact developed above, even though minimal parts of events (i.e., TOUCH-events) has not been a topic that has captured the imagination of scholars working on event semantics.

VCLs establish a classification of verbs (cf. Sections 3 and 4) to which also a classification of experience corresponds. TOUCH-events (or multiple iterations of TOUCH-events) can be subsumed in classes generated by the instrument or object that institutes the contact. These event classes can be further grouped together according to whether contact is established through a physical object (Section 5.2.1) or through a trigger medium (Section 5.2.2). TOUCH-events may furthermore portray the connection or integration of one object into another (Section 5.2.3).

5.2.1. TOUCH-event: contact through physical object. In this category fall TOUCH-events in the most physical sense of the term, typically body part instruments or artifacts hitting another physical object. The following table exhibits a representative listing of the relevant exponents.

5.2.2. TOUCH-event: contact through an intermediate trigger medium. Events of this sort depict touch contact in a more hidden and abstract way, notably through an intermediate trigger medium. For example, it is possible to view the eye as being 'hit' by a visual stimulus causing visual apprehension of something. Human breath causes the articulatory organs to emit audible sounds: human breath is a trigger medium. The gun is a trigger medium for a bullet to hit a target. A fan is a mover (trigger medium) of the air.

5.2.3. TOUCH-event: contact through insertion, connection or association. A number of verbs point to events in which objects are brought into temporal or permanent association. In each case, the respective VCL denotes the instrument that causes an insertion, connection or association. For example, a needle patches (= associates) a thread to a piece of cloth. A rope or a rattan is an instrument of tying (= associating) two objects up. A pen spreads (= inserts) ink on a piece of paper. A tube helps air enter the lungs.

5.3. Other

Finally, there are a small number of verbs that do not support explicitly the idea of a TOUCH-event. Their collocation with VCLs can be regarded as a syntactic imitation of the cases displayed above.

6. Conclusion

This article shares the conviction that every phenomenon of classification requires a multilayered analysis: morphosyntactic, distributional and semantic. The distributional properties determine whether the presumed system is genuinely classificatory. This article departs from a general definition of classificatory systems based on distributional pattern (see Section 2) and demonstrates that the set of so-called verbal classifiers in Kam (Dong) exhibit the required distributional properties (see Section 4) through involvement in a number of morphosyntactic constructions (see Section 3). With respect to the semantic layer (see Section 5), the observation is made that classifiable verbs have minimal parts consisting of TOUCH-events or of iterations of a TOUCH-event.

In a crosslinguistic perspective, East Asian-type VCLs contrast with Australian-style verb classifiers. Australian VCLs consist of a closed set of inflected verbs that categorize an open set of other uninflected verbs, whereas in East and Southeast Asian languages verbs (that are never inflected) are categorized by a set of noun-derived morphemes that, in addition to being VCLs, also occupy the semantic role of instrument.

City University of Hong Kong

Appendix. List of abbreviations
[for all]             'for all' (universal quantifier)
[there exists]        'there exist' (existential quantifier)
[member of]           'is element of' (membership symbol)
[??]              ___ included in (inclusion symbol)
[??]                  not included in (negated inclusion symbol)
<<                    significantly smaller than
>>                    significantly greater than
\                     without (set substraction symbol)
[right arrow]         Application symbol
[ ]                   Class generated by a classifier vector
[??]                  Power set (set of subsets of a set)
1P SG                 First person singular pronoun
3P SG                 Third person singular pronoun
BN                    bare common noun (e.g., table, water)
card                  cardinality
CL                    Classifier
Cons                  Construction
COV:put               Coverb with verbal origin if any
COV:take              Coverb with verbal origin if any
COV:use               Coverb with verbal origin if any
LOC                   Location particle
D                     Set of classified lexemes
LOC:in                Location particle with gloss
M                     Set of classifier vectors
N                     Noun
NCL                   Noun classifier
NO                    Noun O-argument (cf. Dixon 1994)
Num/NUM               Numeral
NUM:9                 Numeral with its value
Qua/QUA               Quantifier
QUA:many              Quantifier with meaning
RECL                  Reciprocal Pronoun
V                     Verb
VCL                   Verb classifier
VCL:fist              Verb classifier with nominal origin
X                     Word variable


Received 23 January 2006

Revised version received

20 September 2006

References

Adams, Karen & Nancy Conklin. 1973. Towards a theory of natural classification. In Claudia Corum, T. Smith-Stark and Ann Weiser (eds.), Papers from the ninth annual regional meeting of the Chicago Linguistic Society, 1-10. Chicago: Chicago Linguistic Society.

Aikhenvald, Alexandra. 2000. Classifiers: A typology of noun categorization devices. Oxford: Oxford University Press.

Allan, Keith. 1977. Classifiers. Language 53(2). 285-311.

Benedict, Paul. 1942. Thai, Kadai, and Indonesian: A new alignment in South-Eastern Asia. American Anthropologist 44, 576-601.

Benedict, Paul. 1975. Austro-Thai: language and culture. New Haven: HRAF Press.

Bhaskararao, Peri & S. K. Joshi. 1985. A study of Newari classifiers. Bulletin of the Deccan College Research Institute 44, 17-31.

Bisang, Walter. 1993. Classifiers, quantifiers and class nouns in Hmong. Studies in Language 17(1). 1-51.

Bisang, Walter. 1996. Areal typology and grammaticalization: Processes of grammaticalization based on nouns and verbs in East and mainland South East Asian languages. Studies in Language 20(3). 519-597.

Bisang, Walter. 1999. Classifiers in East and Southeast Asian languages: Counting and beyond. In Jadranka Gvozdanovic (ed.), Numeral types and changes worldwide, 113-185. Berlin & New York: Mouton de Gruyter.

Bunt, Harry. 1979. Ensembles and the formal semantic properties of mass terms. In Francis Pelletier (ed.), Mass terms: Some philosophical problems, 279-294. Dordrecht: Reidel.

Bunt, Harry. 1985. Mass terms and model-theoretic semantics. Cambridge: Cambridge University Press.

Burling, Robbins. 1965. How to choose a Burmese numeral classifier? In Melford Spiro (ed.), Context and meaning in cultural anthropology, 243-276. New York: The Free Press & London: Collier-Macmillan.

Capell, Arthur. 1979. The classification of verbs in Australian languages. In Stephen Wurm (ed.), Australian linguistic studies (C-54), 229-322. Canberra: Pacific Linguistics.

Carter, Robin. 1976. Chipewyan classificatory verbs. International Journal of American Linguistics 42(1). 24-30.

Chamberlain, James. 1997. Tai-Kadai anthropods: A preliminary biolinguistic investigation. In Jerold Edmondson and David Solnit (eds.), Comparative Kadai: The Tai branch (Publication 124), 291-326. Dallas: Summer Institute of Linguistics & The University of Texas at Arlington.

Chang, Hongen. 1986. Lahuyu Jianzhi [An outline grammar of Lahu]. Beijing: Nationalities' Press.

Chao, Yuenren. 1968. A grammar of spoken Chinese. Berkeley & Los Angeles: University of California Press.

Chen, Shilin, Bian Shiming & Li Xiuqing. 1985. Yiyu Jianzhi [An outline grammar of Yi]. Beijing: Nationalities' Press.

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

Court, Christopher. 1987. Some classes of classifiers in Iu Mien (Yao). Linguistics of the Tibeto-Burman Area 10(2). 144-150.

Croft, William. 1994. Semantic universals in classifier systems. Word 45. 145-171.

Dai, Qingxia & Cui Zhichao. 1985. Achangyu Jianzhi [An outline grammar of Achang]. Beijing: Nationalities' Press.

Denny, Peter. 1976. What are noun classifiers good for? In Salikoko S. Mufwene, Carol A. Walker, and Sanford B. Steever (eds.), Papers from the 12th annual regional meeting of the Chicago Linguistic Society 12, 122-132. Chicago: Chicago Linguistic Society.

Dixon, Robert. 1982. Where have all the adjectives gone? And other essays in semantics and syntax. Berlin & New York: Mouton de Gruyter.

Dixon, Robert. 1986. Noun classes and noun classification in typological perspective. In Colette Craig (ed.), Noun classes and categorization, 105-112. Amsterdam: John Benjamins.

Dixon, Robert. 1994. Ergativity (Cambridge Studies in Linguistics 69). Cambridge: Cambridge University Press.

Edmondson, Jerold & David Solnit. 1988. Introduction. In Jerold Edmondson & David Solnit (eds.), Comparative Kadai: Linguistic studies beyond Tai (Publication 86), 1-26. Dallas: Summer Institute of Linguistics & The University of Texas at Arlington.

Edmondson, Jerold & David Solnit (eds.). 1997. Comparative Kadai, The Tai branch (Publication 124). Dallas: Summer Institute of Linguistics & The University of Texas at Arlington.

Evans, Nicholas. 1996. The syntax and semantics of body part incorporation in Mayali. In Hilary Chappell & William McGregor (eds.), The grammar of inalienability: A typological perspective on body part terms and the part-whole relation, 65-109. Berlin & New York: Mouton de Gruyter.

Goral, Donald. 1978. Numeral classifier systems: A Southeast Asian cross-linguistic analysis. Linguistics of the Tibeto-Burman Area 4(1). 1-72.

Greenberg, Joseph. 1972. Numeral classifiers and substantival number: Problems in the genesis of a linguistic type. Working Papers in Language Universals 9. 1-40. Stanford, CA: Stanford University Press.

Grinevald, Colette. 2000. A morphosyntactic typology of classifiers. In Gunter Senft (ed.), Systems of nominal classification, 50-92. Cambridge: Cambridge University Press.

Haas, Mary. 1942. The use of numeral classifiers in Thai. Language 18(3). 201-205.

Hashimoto, Anne. 1972. Studies in Yue dialects I: Phonology of Cantonese. Cambridge: Cambridge University Press.

He, Jiashan. 1983. Gelaoyu Jianzhi [An outline grammar of Gelao]. Beijing: Nationalities' Press.

He, Jiren & Jiang Zhuyi. 1985. Naxiyu Jian-hi [An outline grammar of Naxi]. Beijing: Nationalities' Press.

Killingley, Siew-Yue. 1983. Cantonese classifiers: Syntax and semantics. Newcastle: Grevatt & Grevatt.

Krifka, Manfred. 1989. Nominalreferenz und Zeitkonstitution: Zur Semantik von Massentermen, Pluraltermen und Aspektklassen. Munich: Fink.

Krifka, Manfred. 1992. Nominal reference, temporal constitution and thematic relations. In Ivan Sag & Anna Szabolcsi (eds.), Lexical matters, 29-53. Stanford, CA: CSLI.

LaPolla, Randy. 1994. Parallel Grammaticalizations in Tibeto-Burman languages: Evidence of Sapir's drift. Linguistics of the Tibeto-Burman Area 17. 61-80.

Lehman, Frederick. 1979. Aspects of a formal theory of noun classifiers. Studies in Language 3(2). 153-180.

Lehman, Frederick. 1990. Outline of a formal syntax of numerical expressions, with special reference to the phenomenon of numeral classifiers. Linguistics of the Tibeto-Burman Area 13(1). 89-120.

Li, Charles & Sandra Thompson. 1981. Mandarin Chinese: A functional reference grammar. Berkeley: University of California Press.

Li, Yongsui & Wang Ersong. 1986. Haniyu Jianzhi [An outline grammar of Hani]. Beijing: Nationalities' Press.

Liang, Min. 1980a. Dongyu Jianzhi [An outline grammar of Dong]. Beijing: Nationalities' Press.

Liang, Min. 1980b. Maonanyu Jianzhi [An outline grammar of Maonan]. Beijing: Nationalities' Press.

Long, Yaohong & Zheng Guoqiao. 1998. The Dong Language in Guizhou province, China (transl. Norman Geary) (Publications in Linguistics 126). Dallas: Summer Institute of Linguistics.

Lucy, John. 2000. Systems of nominal classification: A concluding discussion. In Gunter Senft (ed.), Systems ofnominal classification, 32641. Cambridge: Cambridge University Press.

Mao, Zongwu, Meng Chaoji & Zheng Zongze. 1982. Yaozuyuyan Jianzhi [An outline grammar of the Yao languages]. Beijing: Nationalities' Press.

Matisoff, James. 1973. The grammar of Lahu (Publications in Linguistics 75). Berkeley: University of California Press.

Matthews, Stephen & Tommi Tsz-Cheung Leung. 2001. Verbal vs. nominal classifier constructions in Cantonese and Thai. Paper presented at the Southeast Asian Linguistic Society (SEALS XI). Bangkok, May.

Matthews, Stephen & Virginia Yip. 1999. Verbal and nominal classification: Syntactic and semantic parallels in Cantonese and beyond. Paper presented to Symposium on verb classification (ALT-III). Amsterdam, August.

McGregor, William. 2002. Verb classification in Australian languages. Berlin & New York: Mouton de Gruyter.

Mithun, Marianne. 1984. The evolution of noun incorporation. Language 60(4). 847-894.

Nguyen, Dinh Hoa. 1997. Vietnamese. Amsterdam: John Benjamins.

Noss, Richard. 1964. Thai Reference Grammar. Washington DC: Foreign Service Institute.

Ouyang, Jueya & Zheng Yiqing. 1980. Liyu Jianzhi [An outline grammar of Li(Hlai)]. Beijing: Nationalities' Press.

Paris, Marie-Claude. 1989. Linguistique generale et linguistique chinoise: quelques exemples d'argumentation (Collection ERA 642. U.F.R.L.). Paris: Universite Paris 7.

Rijkhoff, Jan. 1991. Nominal aspect. Journal of Semantics 8. 291-309.

Rijkhoff, Jan. 2002. The noun phrase. Oxford: Oxford University Press.

Ross, Malcom. 1994. Some current issues in Austronesian linguistics. In Darrell Tryon (ed.), Comparative Austronesian dictionary: An introduction to Austronesian studies (Trends in Linguistics: Documentation 10), 45-120. Berlin & New York: Mouton de Gruyter.

Royen, Gerlach. 1929. Die nominalen Klassifikations-Systeme in den Sprachen der Erde: Historisch-kritische Studie, mit besonderer Berucksichtigung des Indogermanischen (Anthropos Linguistische Bibliothek, Vol. IV). Vienna: Anthropos.

Sands, Kristina. 1995. Nominal classification in Australia. Anthropological Linguistics 37. 247-346.

Schafer, Edward. 1948. Noun classifiers in Classical Chinese. Language 24. 408-413.

Seiler, Hansjakob. 1986. Apprehension, language, object and order, Part III: The universal dimension of apprehension. Tubingen: Narr.

Seiler, Walter. 1985. Imonda, a Papuan language. Canberra: Pacific Linguistics.

Senft, Gunter. 2000. What do we really know about nominal classification systems? In Gunter Senft (ed.), Systems of nominal classification, 11-49. Cambridge: Cambridge University Press.

Silverstein, Michael. 1986. Classifiers, verb classifiers and verbal categories. Proceedings of the 12th Annual Meeting of the Berkeley Linguistics 12. 497-514.

Sun, Hongkai. 1981. Qiangyu Jianzhi [An outline grammar of Qiang]. Beijing: Nationalities' Press.

Sun, Hongkai. 1982. Dulongyu Jianzhi [An outline grammar of Dulong]. Beijing: Nationalities' Press.

Vendler, Zeno. 1967. Linguistics in Philosophy. Ithaca: Cornell University Press.

Wang, Fushi. 1985. Miaoyu Jianzhi [An outline grammar of Miao]. Beijing: Nationalities' Press.

Wang, Jun & Zheng Guoqiao. 1993. An outline grammar of Mulao. [Transl. by Luo Yongxian of 1980. Molaoyu Jianzhi. Beijing: Nationalities' Press]. Canberra: National Thai Studies Centre, Australian National University.

Wei, Qingwen & Tan Guosheng. 1980. Zhuangyu Jianzhi [An outline grammar of Zhuang]. Beijing: Nationalities' Press.

Xu, Lin, Mu Yuzhang & Gai Xingzhi. 1986. Lisuyu Jianzhi [An outline grammar of Lisu]. Beijing: Nationalities' Press.

Xu, Lin & Zhao Yansun. 1984. Baiyu Jian-hi [An outline grammar of Bai]. Beijing: Nationalities' Press.

Xu, Xijian & Xu Guizhen. 1984. Zaiwa(Jingpo)yu Jianzhi [An outline grammar of Zaiwa(Jingpo)]. Beijing: Nationalities' Press.

Yang, Rong. 2001. Common nouns, classifiers and quantification in Chinese. New Brunswick, NJ: Rutgers, the State University of New Jersey dissertation.

Yu, Cuirong. 1980. Buyiyu Jianzhi [An outline grammar of Buyei]. Beijing: Nationalities' Press.

Yu, Cuirong & Luo Meizhen. 1980. Daiyu Jianzhi [An outline grammar of Dai]. Beijing: Nationalities' Press.

Zhang, Junru. 1980. Shuiyu Jianzhi [An outline grammar of Shui]. Beijing: Nationalities' Press.

Zhou, Xiaokang & William B. McGregor. 1999. Verbal classification in Mandarin Chinese. Paper presented to Symposium on verb classification (ALT-III). Amsterdam, August.

Zhou, Zhizhi & Yan Qixiang. 1984. An outline grammar of Wa [Wayu Jianzhi]. Beijing: Nationalities' Press.

Notes

* wish to express warm regards to Wu Shihua, a native Kam from Sanjiang county, who worked with me on the verb classifiers in Kam during my stay in Kunming (P.R. of China) in the first half of 2003. Comments of two anonymous reviewers were very helpful. My gratitude also goes to Noel Johnston for valuable comments. I gratefully acknowledge the generous support of City University of Hong Kong in the form of a Strategic Research Grant (Project No. 7001921). Correspondence address: Department of Chinese, Translation and Linguistics and Halliday Research Center, City University of Hong Kong, 83 Tat Chee Avenue, Kowloon, Hong Kong, E-mail: mgerner@cityu.edu.hk.

(1.) Kam (autonym) or Dong (Han-Chinese appellation) belongs genetically to the KamSui language group. Scholars who adhere to the "Kadai hypothesis" (cf. Edmondson and Solnit 1988: 4-5; 1997: 2) associate Kam-Sui with the Tai group into a superordinate Kam-Tai nod, which in turn is assembled with Hlai (Li) and Geyang (including Gelao) as the Kadai language family. Others wish to dissociate the Kam-Sui languages from the Tai languages and suggest four instead of three first-order nodes: Tai, Kam-Sui, Hlai and a Kadai nod (roughly equivalent to the above Geyang nod). This classification is put forward by Chamberlain (1997) and is based on zoological data. Its appellation is the "Tai-Kadai hypothesis". The "Tai-Kadai hypothesis" is also very popular in China where this language family is labeled "Zhuang-Dong". Kadai (as in the "Kadai hypothesis") is one of the components of the Austro-Tai language phylum (cf. Benedict 1975; Edmondson and Solnit 1988, 1997). There is an ongoing controversy as to whether Kadai languages should be joined with the Austronesian language family into an Austro-Tai phylum. The Austronesian-Tai connection was pioneered by Benedict (1942, 1975) and this link is now accepted by a certain number of historical linguists working on Kadai. However, Austronesian linguists remain skeptical about the reconstructions made and have characterized them as "too loose" (Ross 1994: 96).

(2.) The numbers 55, 13 etc are tone markers and indicate relative pitch on a scale from 1 (lowest) to 5 (highest). The first number represents the beginning and the second number the end of the tonal contour. In Kam, there are nine distinctive tones.

(3.) In the isolating languages of East Asia and Southeast Asia, a coverb is a grammaticalized verb that happens to function as a marker of semantic participation roles, while retaining some of its verbal meaning.

(4.) Wu Shihua is the former now retired director of the Sanjiang County Museum. He has sound linguistic education and is well versed in all kinds of oral genres of his native language.

(5.) Verbal classification through suppletive verbal stems refers to a closed set of verbs, typically verbs of location or existence, that select the S- or O-argument according to its inherent shape or animacy features. In many Tibeto-Burman languages, for example, there are a number of verbs for 'exist' or 'be at', each compatible with a different subset of nouns (S-arguments), see La Polla (1994).

(6.) For example, Lehman (1979, 1990) argues that noun classifiers in East Asian and Southeast Asian languages are not classificatory devices generating a more or less language and culture-specific, comprehensive system of semantic classification ("classification" being taken in the strict sense of a taxonomic system of categories). Rather, for him the so-called classifiers have basically to do with the way the language handles the phenomenon of quantification or reference. This judgment, however, does not correspond to my personal observations in this linguistic area and also to published data of recent years. Most languages there have partial or holistic taxonomies. Concerning gender, the picture may be more diverse. Corbett (1991) differentiates between semantic systems and formal systems. In a semantic system, nouns are classified according to their concordial behavior in a way that the denotata of nouns grouped together belong to the same semantic category.

(7.) In the languages of East and Southeast Asian languages I am aware of, the classifier is often a semantically bleached morpheme that has historically originated from a specific noun (not a class noun such as 'kind', 'animal' etc).

(8.) At end of the previous section, we said that semantic considerations do not intervene in the definition of classification (but would rather constitute an epiphenomenon). The fact that we appeal here to semantic features of the defining morphosyntactic constructions does not contradict this approach. We are referring here to semantic qualities of syntactic constructions and not to those of the word classes generated by these constructions.

(9.) Cf. description of Li and Thompson (1981: 352) for Mandarin.

(10.) "Monotransitive" refers here to two-place predicates as opposed to intransitive (one-place) or ditransitive (three-place) predicates.

(11.) This sound structure corresponds to the standard pronunciation of Rongjiang county. Through cross-regional contact, tau 53 is also in use in Sanjiang county, even though the original native form for Sanjiang is con 33 'time'.

(12.) It tends to be slightly more restricted as VCL in Sanjiang County than in Rongjiang County, but again these differences are increasingly leveled through cross-regional exchange.

(13.) To simplify the exposition, we assume that "quantized" means "quantified", even though both terms diverge. For a logical definition of "quantized" see Krifka (1992).

(14.) It would be interesting to investigate the topic of autoclassifiers from a point of view of language acquisition. I suspect nominal and verbal autoclassifiers to be also characteristic of child language, even beyond the examples given in Table 4. In the learning process of sortal NCLs and VCLs, a child might find it more straightforward to produce an autoclassifier than the abstract sortal classifier required by a specific construction type. This claim, of course, would have to be substantiated further.

(15.) The speakers, of course, do not have explicit knowledge of articulatory phonetics, but the nature of language is to often refer vaguely to physical facts.

(16.) Vectorial configuration corresponds to spatial extension of events. It is the verbal analogue of shape in nominal classification systems. Aktionsart mirrors constituency features of nominal classificatory systems (e.g., collection, mass). One author to quote in this regard is Rijkhoff (1991, 2002) who developed the notion of Seinsart in the nominal realm. His idea is that noun phrases just as verb phrases can be marked for shape or internal structure/constituency. Depending on the on/offset of these features, Rijkhoff distinguishes a taxonomy of four Seinsarten and four Aktionsarten. Finally, McGregor's third category--valency--has no obvious analogue with semantic features of nominal classification systems (cf. McGregor 2002: 29).
Table 1. Measure VCLs in Kam

                                      Kam time units
Kam time units (microdomain)          (macrodomain)

ha 35 'moment, instant,               nam 53 'evening
  short stretch of time'                and night'
cen 33 'hour, vague time
  unit between 5 and 60 minutes'      man 55 'day'
ci 11 'traditional time unit          nan 55 'month'
  of 2 hours = 120 minutes'           nin 11 'year'
                                      sam 33 'generation,
                                        lifespan'

Table 2. Three noun classes generated by NCLs and VCLs in Kam

Class-1 Nouns       Class-2 Nouns        Class-3 Nouns

nam 31 'water'      nui 55 'snow'        lu 33 'lawsuit'
k hw au 13          p j en 55            wen 453 'business'
  'wine'              'rain'
eu 31 'food'        u 31 'hail'          on 55 'labor'
nen 11 'person'     sip 323 'festival'   con 33 'rite'
to 55 'door'        (...)                cei 33 'catastrophe'
sen 55 'village'                         (...)
(...)

Table 3. NCLs that function as instrumental nouns in Kam

Instrumental        Instrumental phrase with coverb
NCL                 [au.sup.55] 'take'

                    'take'

tau 55 'pot'        au 55          tau 55 ceu 13
                                     'cook with pot'
pat 323 'bowl'      au 55          pat 323 ma 453
                                     'soak with bowl'
jon 453 'barrel'    au 55          jon 453 ma 453
                                     'soak with barrel'
pen 11 'basin'      au 55          pen 11 ma 453
                                     'soak with basin'
mjai 53 'gourd      au 55          mjai 53 ljem 11
  ladle'                             'pour with gourd ladle'
teu 53 'rice        au 55          teu 53 mei 35
  steamer'                           'stew with rice steamer'
ku 11 'vessel'      au 55          ku 11 tun 55
                                     'cook with vessel'

Instrumental
NCL                 Noun classifier phrase

tau 55 'pot'        ceu 13 i 55 ma 55 'cook
                     one pot of vegetables'
pat 323 'bowl'      ma 453 i 55 pat 323
                      keu 31 'soak one bowl of rice'
jon 453 'barrel'    ma 453 i 55 jon 453
                      keu 31 'soak one barrel of rice'
pen 11 'basin'      ma 453 i 55 pan 11
                      keu 31 'soak one basin of rice'
mjai 53 'gourd      ljem 11 i 55 mjai 53
  ladle'              nam 31 'pour one ladle of water'
teu 53 'rice        mei 35 i 55 teu 53
  steamer'            keu 31 'stew one rice steamer
                      of rice'
ku 11 'vessel'      tun 55 i 55 ku 11
                      keu 31 'cook one vessel of rice'

Table 4. Autoclassifiers in Kam

Kam verb autoclassifier              Kam autoclassifier phrase

[??]ik 13 'kick'                     [??]ik 13 sam 35 [??]ik 13 'kick'
                                     three times'

k w en 33 shake (one's head)'        k w en 33 sam 35 en 33 'shake
                                     three times (one's head)'

[??]e 53 'press (for oil), extract'  [??]e 53 sam [??]e 53 'press (for
                                     oil) three times'

t h oi 35 'dig'                      t h oi 35 sam 35 t h oi 35 'dig
                                     three times'

siu 53 'chisel (v.)'                 siu 53 sam 35 siu 53 'sting three
                                     times with a chisel'

Siu 53 'sting (v.)'                  siu 53 sam 35 siu 53 'sting three
                                     times with a sting'

keu 55 'hook (v.)'                   keu 55 sam 35 keu 55 'hook three
                                     times with a hook'

Table 5. VCLs of the versatile verb [to.sup.323]

                                Construction type (1Oi)
Sortal VCL                      to 323 + Num/Qua + VCL

tok 323 'hoe'                   to 323 i 55 tok 323
                                  'use (= dig with) a hoe once'
t h an 31 'shovel'              to 323 i 55 an 31
                                  'use (= dig with) a shovel once'
mak 323 'palm'                  to 323 i 55 mak 323
                                  'beat with one's palm once'
cui 11 'fist'                   to 323 i 55 cui 11
                                  'use (= beat with) one's fist once'
kui 53 'hammer'                 to 323 i.sup 55 kui 53
                                  'use (= beat with) a hammer once'
k w an 55 'axe'                 to 323 i 55 k w
                                  an 55 'use (= chop with) an
                                  axe once'
con 53 'gun'                    to 323 i 55 con 53
                                  'use (= shoot with) a gun once'
mja 31 'knife'                  to 323 i 55 mja 31
                                  'use (= cut with) a knife once'
tan 323 'oar'                   to 323 i 55 tan 323
                                  'use (= row with) an oar once'
tan 55 'foot'                   to 323 i 55 tin 55 tj
                                  'stamp with one's foot once'
t h em 35 'needle'              to 323 i 55 t h
                                  em 35' use (= sew with) a
                                  needle once'
ep 55 'mouth'                   to 323 i 55 ep 55
                                  'bite with one's mouth once'
siu 53 'chisel (n.)'            to 323 i 55 siu 53
                                  'use (= chisel with) a chisel once'
can 55 'spear'                  to 323 i 55 can 55
                                  'use (= jab with) a spear once'
kan 33 'pole'                   to 323 i 55 kan 33
                                  'use (= prop up with) a pole once'

                                Construction type (1Ov)

Verb                            to 323 + Num + V

pai 53 'pay respect'            to 323 i 55 pai 53
                                  'pay respect once'
peu 11 spade, dig'              to 323 i 55 peu 11
                                  'dig once'
tai 11 'draw, pull'             to 323 i 55 tai 11
                                  'draw once'
mak 323 'cleave'                to 323 i 55 mak 323
                                  'cleave once'
tap 31 'step on'                to 323 i 55 tap 31
                                  'step on once'
tun 55 'carry'                  to 323 1 55 tun 55
                                  'carry once'
th ik 13 'kick'                 to 323 i 55 t h ik l3
                                  'kick once'
lja 11 'lick'                   to 323 i 55 lja 11
                                  'lick once'
(k)it 31 'bite'                 to 323 i 55 (k)it 31
                                  'bite once'
tok 55 'pound'                  to 323 i 55 tok 55
                                  'pound once'
p h eu 453 'plane'              to 323 i 55 p h
                                  eu 453 'plane once'
te 53 'chop, dig'               to 323 i 55 te 53
                                  'chop once'
mat 323 'rub'                   to 323 i 55 mat 323
                                  'rub once'
neu 31 'pry, prize'             to 323 i 55 neu 31
                                  'pry once'
niu 13 'patch up'               to 323 i 55 niu 13
                                  'patch up once'

Table 6. List of sortal VCLs in Kam

Kam verb classifier     Cardinality

m j a 31 'knife'        10
cui 11 'fist'           6
k w an 55 'axe'         6
m j a 11 'hand'         5
tin 55 'foot'           5
t h em 35 'needle'      5
mei 31 'rod, stick'     4
kui 53 'hammer'         4
ta 55 'eye'             4
pjet 55 'pen'           4
ton 11 'tube'           4
lam 33 'rope'           3
mit 31 'small knife'    3
so 33 'voice, breath'   3
miu 11 'scissors'       3
tok 323 'hoe'           2
t h an 31 'shovel'      2
pau 55 'horn'           2
mak 323 'palm'          2
con 53 'gun'            2
keu 55 'hook (n.)'      2
tan 323 'oar'           2
wai 11 'fan (n.)'       2
siu 53 'chisel (n.),
  sting (n.)'           2
con 53 wan 35 'iron
  punching pin'         2
cap 55 'fish fork'      2
kan 33 'pole'           2
tau 55 'rattan'         1
t h oi 35 'pickaxe'     1
sen 31 'buttocks'       1
ljim 31 'sickle'        1
cik 13 'rule'           1
k w iu 53 'balance'     1
k w an 55 set 55
  'broom'               1
so 13 'lock'            1
ma 11 'tongue'          1
p j a 55 'stone'        1
ta 11 k w iu 53
  'weight of
  steelyard'            1
ne 11 'tooth'           1
ep 55 'mouth'           1
can 55 'spear'          1
p h a 453
  handkerchief'         1

Kam verb classifier     Generated verb classes

m j a 31 'knife'        (k)at323 'cut', tat 55 'chop, slice',
                          te 53 'chop, dig', las 53 'tear',
                          cep 13 'prick', nak 11 'stick in',
                          mak 55 'cut about', p h i 35 'pare,
                          peel, cok 13 / nok 13 'jab', to
                          323 'put'
cui 11 'fist'           keu 35 'beat', map 11 'thump',
                          tok 55 'pound', lem 55 strike',
                          pem 55 'attack', to 323 'put'
k w an 55 'axe'         keu 35 'beat', pam 323 'hack, fell',
                          mak 323 'cleave', tok 55 'pound',
                          te 53 'chop, dig', to 323 'put'
m j a 11 'hand'         tai 11 'draw', nam 55 'grasp', pei
                          323 'repay', pan 55 help', won
                          13 'push'
tin 55 'foot'           tam 31 'stamp on', tap 31 'step on',
                          t h ik 13 'kick', cai 13 'trample',
                          nan 13 'step on'
t h em 35 'needle'      tip 323 'sew', niu 13 'patch up',
                          t h iu 35 'embroider', to 323
                          'put', teu 53 'prick'
mei 31 'rod, stick'     keu 35 'beat', sat 13 'whip', lat
                          13 'strike', to 323 'put'
kui 53 'hammer'         keu 35 'beat', tak 323 'hit', tok
                          55 'pound', to 323 'put'
ta 55 'eye'             nu 53 'see', ljin 55 'watch', pi53
                          'peep, spy', k w in 53 'watch out'
pjet 55 'pen'           ca 323 'write', k w a 55 'record',
                          wa 35 'paint', to 323 'put'
ton 11 'tube'           tan 55 'drink', sot 13 'inhale',
                          k h eu 35 'hit', tui 323 'scoop'
lam 33 'rope'           suk 31 'bind' (only with numeral
                          '1'), tai 11 'draw', lak 11 'tie up'
mit 31 'small knife'    (k)at 323 'cut', k w et 31 'shave',
                          p h i 35 'pare, peel'
so 33 'voice, breath'   pan 55 'call, cry', t h ok 35
                          'neigh', k h eu 453 'bark'
miu 11 'scissors'       la 53 'tear open', kun 55 'cut',
                          k h ep 55 'cut'
tok 323 'hoe'           leu 55 'dig', to 323 'put'
t h an 31 'shovel'      peu 11 'spade', to 323 'put'
pau 55 'horn'           p h eu 453 'plane', neu 31 'pry'
mak 323 'palm'          keu 35 'beat', to 323 'put',
con 53 'gun'            pen 53 'shoot', to 323 'put'
keu 55 'hook (n.)'      keu 55 'hook (v.)', to 323 'put'
tan 323 'oar'           wai 11 'row', to 323 'put'
wai 11 'fan (n.)'       wai 11 'fan (v.)', to 323 'put'
siu 53 'chisel (n.),    siu 53 'chisel (v.)', siu 53
  sting (n.)'             'sting (v.)'
con 53 wan 35 'iron     non 53 'dig out', leu 55 'dig out'
  punching pin'
cap 55 'fish fork'      pen 53 'shoot', to 323 'put'
kan 33 'pole'           cen 55 'prop up', k h eu 35 'hit'
tau 55 'rattan'         tai 11 'draw'
t h oi 35 'pickaxe'     t h oi 35 'dig'
sen 31 'buttocks'       sui 53 'sit' (only with numeral 'one')
ljim 31 'sickle'        (k)an 53 'cut'
cik 13 'rule'           tak 31 'measure'
k w iu 53 'balance'     k w iu 53 'weigh'
k w an 55 set 55
  'broom'               set 55 'sweep'
so 13 'lock'            so 13 'shut'
ma 11 'tongue'          l j a 11 'lick'
p j a 55 'stone'        tap 11 'press'
ta 11 k w iu 53
  'weight of
  steelyard'            tap 11 'press'
ne 11 'tooth'           (k)it 31 'bite'
ep 55 'mouth'           (k)it 31 'bite'
can 55 'spear'          cok 13/nok 13 'jab'
p h a 453
  handkerchief'         mat 323'rub'

Table 7. Verb classes of Toucx-event verbs. contact
through physical object

Kam verb classifier              Generated verb classes

tin 55 'foot'                    tam 31 'stamp on', tap31 'step
                                   on', t h ik 13 'kick', cai 13
                                   'trample', nan 13 'step on'
m j a 11 'hand'                  ta 11 'draw', nam 55 'grasp',
                                   pei 323 'repay', won 13 'push'
cui 11 'fist'                    keu 35 'beat', map 11 'thump',
                                   tok 55 'pound', lem 55
                                   'strike', pem 55 'attack',
                                   to 323 'put'
mak 323 'palm'                   keu 35 'beat', to 323 'put'
ma 11 'tongue'                   l j a 11 'lick'
ne 11 'tooth'                    (k)it 31 'bite'
ep 55 'mouth'                    (k)it 31 'bite'
m j a 31 'knife'                 (k)at 323 'cut', tat 55 'chop,
                                   slice', te 53 'chop, dig', la 53
                                   'tear', cep 13 'prick', etc.
k w an 55 'axe'                  keu 35 'beat', pam 323 'hack,
                                   fell', mak 323 'cleave',
                                   tok 55 'pound', te 53 'chop,
                                   dig', etc.
mei 31 'rod, stick'              keu 35 'beat', sat 13 'whip',
                                   lat 13 'strike', to 323 'put'
k w i 53 'hammer'                keu 35 'beat', tak 323 'hit',
                                   tok 55 'pound', to 323 'put'
mit 31 'small knife'             (k)at 323 'cut', k w et 31
                                   'shave', p h i 35 'pare, peel'
tok 323 'hoe'                    leu 55 'dig', to 323 'put'
t h an 31 'shovel'               peu 11 'spade', to 323 'put'
tan 323 'oar'                    wai 11 'row', to 323 'put'
m j u 11 'scissors'              la 53 'tear open', kun 55 'cut'
t h oi 35 'pickaxe'              t h oi 35 'dig'
lim 31 'sickle'                  (k)an 53 'cut'
con 53 wan 35 'iron punching
  pin'                           non 53 'dig out'
p j a 55 'stone'                 tap 11 'press'
ta 11 k w iu 53 'weight of
  steelyard'                     tap 11 'press'
can 55 'spear'                   cok 13 / nok 13 'jab'
k w an 55 set 55 'broom'         set 55 'sweep'
p h a 453 'handkerchief'         mat 323 'rub'
cap 55 'fish fork'               pen 53 'shoot'

Table 8. Verb classes of roucx-even verbs.' contact through
intermediate trigger medium

Kam verb classifier     Generated verb classes

ta 55 'eye'             nu 53 'see', lig 55 'watch', pi 53
                          'peep, spy', k w in 53 'watch out'
so 33 'voice, breath'   pan 55 'call, cry', t h ok 35 'neigh',
                          t h eu 453 'bark'
con 53 'gun'            pen 53 'shoot', to 323 'put'
wai 11 'fan (n.)'       wai 11 'fan (v.)', to 323 'put'

Table 9. Verb classes of TOUCH-even verbs: contact through insertion,
connection or association

Kam verb classifier   Generated verb classes

t h em 35 'needle'    tip 323 'sew', niu 13 'patch up',
                        t h iu 35 'embroider', to 323 'put'
lam 33 'rope'         suk 31 'bind' (only with numeral '1'),
                        tai 11 'draw', lak 11 'tie up'
tau 55 'rattan'       tai 11 'draw', lak 11 'tie up'
p j et 55 'pen'       ca 323 'write', k w a 55 'record'
kan 33 'pole'         cen 55 'prop up'
ton 11 'tube'         sot 13 'inhale'

Table 10. Verb Classes of exceptional VCLs

Kam verb classifier     Generated verb classes

sen 31 'buttocks'       sui 53 'sit' (only with numeral 'one')
cik 13 'rule'           tak 31 'measure'
k w iu 53 'balance'     k w iu 53 'weigh'
COPYRIGHT 2009 Walter de Gruyter GmbH & Co. KG
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2009 Gale, Cengage Learning. All rights reserved.

 
Article Details
Printer friendly Cite/link Email Feedback
Author:Gerner, Matthias
Publication:Linguistics: an interdisciplinary journal of the language sciences
Article Type:Report
Geographic Code:9CHIN
Date:May 1, 2009
Words:18051
Previous Article:Clausal coordination and coordinate ellipsis in a model of the speaker.
Next Article:Young children's understanding of ongoing vs. completion in imperfective and perfective participles.
Topics:


Related Articles
Grammatical development in adolescent first-language learners *.
Instrument inversion in Toqabaqita *.
Culture Secretary on yet another 2008 visit - for a right old Ding Dong.
Bleeps and bangs cause a right old Ding Dong at Fact.

Terms of use | Privacy policy | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters