Professional Documents
Culture Documents
1983.syntax and Semantics of Numeral
1983.syntax and Semantics of Numeral
1983.syntax and Semantics of Numeral
VOLUME: 7
ISSUE: 2
MONTH:
YEAR: 1983
ISSN: 0378-4177
OCLC #:
This copy has been made for use solely by a student, staff member, faculty member, or library patron for the purpose of
research, private study, review, criticism, news reporting, education, parody or satire, and may only be used by the recipient
for these purposes. When using copies for the purpose of news reporting, criticism or review, the source must be mentioned,
and, if given in the source, the name of the author or creator of the work.
Note that further copying, scanning, faxing, transmitting or otherwise making or distributing copies of this item except as
permitted by law, may be an infringement of copyright if done without proper license or the consent of the copyright owner.
Failure to abide by these conditions may constitute violation of Canadian Copyright Law, and/or relevant licensing
agreements, including agreements with Access Copyright.
STUDIES IN
LANGUAGE
INTERNATIONAL JOURNAL
“ FOUNDATIONS OF LANGUAGE ”
JO H N B E N JA M IN S R.V. P U B L IS H E R AM STERDAM
S T U D I E S IN L A N G U A G E
International Journal sponsored by the Foundation
“Foundations of Language”
EDITORIAL BOARD
John W. M. Verhaar, Managing Editor Werner Abraham, Review Editor
Gonzaga University, Spokane, WA Univ. of Groningen, Netherlands
Elsa Whitfield
16 Queensway,
Ponteland,
NEWCASTLE-UPON-TYNE
England NE20 9RZ
CONSULTING EDITORS
'Mu/,
'■*<% Je ,
: . " V; :
SYNTAX AND SEMANTICS OF NUMERAL ; Clip, ?v\
CLASSIFIERS IN THAI ,^
0. Introduction
In the paper presented here we try both to analyze the syntactic function, and
to document the semantic diversity, of Thai numeral classifiers. The study was
motivated by current comparative research of the Cologne Language Univer
sal Project; its general background is a series of enquiries (the results of
which are forthcoming presently1) into different techniques employed across
languages to relate nominal notions to linguistic categories of quantification.
In this framework, the various classificatory techniques encountered in
different groups of languages most strikingly invite language comparison. For
there are conspicuous affinities between such classificatory phenomena as,
for instance, numeral classification in East Asia and the various classificatory
principles in Amerindian languages.2
Therefore, our article is meant as a case study which illustrates one such
classificatory technique within the system of an individual language. Thai
seems most suitable for this purpose, for it is in the core group of East Asian
numeral classifier languages, (so much so that Jones (1970:11) hypothetically
suggests the Thai languages as the original source of this areal phenomenon, —
anissue which can be definitely settled only on more detailed and reliable dia
chronic evidence than is so far at our disposal).
In comparison to other numeral classifier languages of the area, Thai
employs numeral classifiers a) in the widest range of distribution in NP con
structions, and b) in terms of an extremely diversified network of separate
classes. Therefore, Thai (and its immediate cognates) offer the most differen
tiated evidence as to both the syntactic and semantic properties of this type of
category.
166 HARALD HUNDIUS & ULRIKE KOLVER
The article deals with both these aspects in the following ways: in §1, clas
sifiers are discussed in their categorial affiliations. It is shown that they forma
separate word class, as against nouns on the one hand, as against measuring
numeratives on the other. In §2, the various possible distributions are ana
lyzed as to their syntactic function. This consists in establishing immediate
reference to individual objects. In this function classifiers serve to compensate
the most specific property of Thai nouns: these purely denote concepts and,
for this reason, are incompatible with direct quantification. In §3, some gener
al semantic regularities are discussed which relate to the principle of classifica
tion by inherent features of the corresponding nouns. Finally, in §4, a list of
classifiers is presented which tries to document as fully as possible the network
of lexical classes for the present state of the language. The more general
results of this lexical survey are commented on at the end of §4.: the distribu
tion of lexical classes shows a most characteristic unevenness. It reveals a se
mantic scalarity which is regulated in terms of more of less easy visual appre
hension of objects.
The lexical materials for this study have been assembled from the avail
able dictionaries of Standard Thai. The chief source was that of Haas (1964)
which lists appropriate classifiers along with most noun entries. This has been
checked through and, additionally, other dictionaries have been consulted for
further data and more detailed glosses, (see references).
For both the syntactic and the semantic parts additional evidence beyond
the published sources has been obtained from native informants. We should
here like to express our gratitude to all those who have so kindly and coopera
tively helped in our work, in particular to Mr. and Mrs. Achintaarangkorn,
Mrs. ManidaaHuebner, Dr. R. Thiraporn (Univ. of Kiel), Mr. S. Cangkacit,
Mr. P. Na Nakhorn (Bangkok), Acharn Sudaporn and Dr. P. Sornhiranya
(Chulalangkorn Univ. of Bangkok). Equally we should like to thank Mr. R.
Dolling (Univ. of Kiel) for his substantial help in collecting lexical materials.
any particular language). No doubt there will be measuring units specific to li
mited areas and to certain varieties of objects (as e.g. ihanaan denoting a hol
lowcoconut traditionally used for measuring rice), which thus convey some
cultural and linguistic information. In any case, our point is that measures, by
their very nature, carry lexical information of their own which is handled in
dependently of the information conveyed by a potential head noun.
Words meaning ‘type, kind’etc. have been included here though they ob
viously do not denote quantities. For one thing, like measures, they introduce
notions extrinsic to their head nouns. Moreover, they are subject to the same
restrictions as measures (for which see below). Nor do they enter into
specific relationships with particular subsets of nouns. Thus, they have no
bearing on nominal subcategorization or on the classification of count
nouns.14
As against this, classifiers proper always and in principle reflect intrinsic
semantic features of those sets of nouns that they are systematically related to,
a principle which culminates in complete identity of head noun and classifier
as exemplified above in (4), (see 2.2.3,3.2. a., 4.4. below).
1.3. Syntactic restrictions: Mass nouns vs. Count nouns
As introduced, this distinction is a semantic one. It is, however, well borne out
insyntax, which becomes apparent as soon as adjectives are introduced into a
NP. .' ■ , ;- .
Besides obligatorily combining with numerals, classifiers proper op
tionally combine with adjectives (in a way more fully analyzed in 2 .1.2 .
below):
(5) a. nok tua jaj -
bird (clf:body) big i v ; ;'
“the big bird”
(5) b. nok tua sii-khiaw tua jaj
bird (elf) green (elf) big
..“the big green bird”
Inthese examples the classifier, tua, does not contribute any independent lex
ical meaning to the NP. Therefore it combines with any adjective in such a way
that the meaning of the adjective is mapped onto the head noun itself. .
No analogous construction, however, is possible with numeratives of the
measuring variety, precisely because these carry independent lexical informa
tion.
170 HARALD HUNDIUS & ULRIKE KOLVER
Here, too, the adjective is not mapped onto the head noun, but modifies the
numerative itself. Adjectives meant to modify the head noun must not appear
in this position:
(8) b. *ndk fuuij stt-khiaw
bird swarm green.
Instead the adjective must immediately follow its head:
(8) c. ndk jajfuurj jaj
“a large swarm of big birds”
d. ndk sti-khiawfuuij jaj
“a large swarm of green birds”.
Thus a numerative like fuurj when combined with a count noun like ndk dis
plays exactly the same structural properties as the measure in (6) co-occurring
with a mass noun.
That means, the numeratives in (6) — (8) do not differ syntactically, for
which reason they have been included in the same subcategory, as opposed to
genuine classifiers. The difference is merely a semantic one: khanaan and
fuinj are both measure terms exemplifying the dissociative and the associative
variety respectively.17In other words, in Thai there is no syntactically distinct
type of collective construction, it is merely a lexical variant of the measure
construction.
Thus, as we saw, measure constructions in themselves do not warrant a
distinction between mass and count nouns. The distinction, however, does
emerge from adjectival expansions as in (5) when contrasted with (6) b.: only
count nouns, not mass nouns, admit of phrases like ndk tuajaj “the big bird”.
To sum up: the view that Thai, or numeral classifier languages generally,
do not distinguish mass from count nouns is not supported by the evidence in
Thai. Yet, Thai nouns have properties which nouns e.g. in German do not
share. In the attempt to define these properties we shall concentrate upon the
type of construction most conspicuously contrasting with the enquirers’
mother tongue. This is count nouns and the methods employed to quantify
them. And it is these which are syntactically and semantically analyzed in the
following paragraphs.
2. Syntacticfunction of classifiers
2.1. Distribution
The most frequent use of classifiers is their obligatory occur
rence in counting constructions as above in
172 HARALD HUNDIUS & ULRIKE KOLVER
overtly expressed numeral, it has to co-occur with some other nominal deter
miner, — a demonstrative in our present case. This must be preceded by the
classifier for the reasons a) that the position of the implied numeral “one” has
to be left empty and b) that demonstratives invariably occur as terminal ele
ments of a NP.
In this way a surface structure like (9) b., rather than constituting an en
tirely different type of classifier construction, can be traced to a basic pattern of
counting construction. This can be formed in two alternative ways: either
there is an overt numeral construction as in (1) a., or there is a covert, implicit
counting construction. In this latter case the unambiguous interpretation of
the implied numeral position triggers the singular interpretation of the phrase
as in (9) b..
2.1.3. Variation of order among determiners
In the light of our analysis so far let us now turn to the variants permissible
when adjectives and numeral phrases simultaneously figure in a NP, as illus
trated in (14). Of these, variant a. rom sti-khiaw sdam khan “three green um
brellas” presents a straightforward expansion of (10) a. rom sti-khiaw. In nor
mal NP order, the noun is first expanded by an adjective, so as to bring about a
modified nominal concept, which in turn is determined by a numeral phrase
containing the obligatory classifier.
Against that, b. rom khan sti-khiaw sdam khan “the three green umbrel
las” presents an expansion of (10) b. and thus contains two occurrences of the
classifier. Finally, c. rom sdam khan sti-khiaw “the three green umbrellas”,
changes the order of numeral phrase and adjective and thus is structured
exactly like (13). No tangible difference in meaning could be ascertained be
tween b. and c. except for emphasis: informants agreed on b. more emphati
cally referring to specific, identifiable objects.
The analysis of (13) above thus obviously applies to (14) c. as well: the
classifier may appear once only, and that is, as the obligatory co-constituent of
the numeral, so that the phrase divides into N + (Num+Clf) + Adj. Evi
dently, on this analogy between (13) and (14) c., the phrase rom khan sti-
khiaw can be analyzed in the same way as rom khan nti: here again the clas
sifier reflects an implied quantitative notion due to an implied numeral posi
tion, which usually and most likely is interpreted as “one”. For, as we noted
above, phrases like ( 10) b. preferably have a singular reading, too.
However, in contrast to (9) b., such an interpretation is not compulsory
inthis case, and this is where we have to take the differing orders of (14) b. and
176 HARALD HUNDIUS & ULRIKE KOLVER
c. into account.
Unlike demonstratives, which are restricted to the terminal position of a
NP, adjectives allow for much more positional variation. The relative order
Clf- Adj,however, may not vary, precisely because the classifier may combine
with an adjective only, if it is not bound to a preceding numeral, as it is in (14)
c. Yet the phrase rom khan sii-khiaw with its implied numerical connotation
may still be followed by an overt numeral phrase, which of course determines
the actual quantitative interpretation of the phrase. For this reason rom khan
sii-khiaw, which maybe related to either order variant (14) b. ore. respective
ly, has quantitative connotations different from rom khan mi.
A demonstrative necessarily marks the end of a NP, so that no further
specification, numerical or otherwise, can possibly follow and interfere with
the numerical interpretation of the classifier. Therefore this interpretation in
variably is “one”, being the most natural reading in the absence of an overt
specification.
The order variants permissible with adjectives, however, open up a wider
scope: when no numeral is actually specified, the most ready and natural in
terpretation here, too, is “one”. Yet, in principle there is the possibility of fol
lowing up the implicit counting construction Clf - Adj by an overt numeral
phrase within the same NP, and this is why the numerical interpretation is not
absolutely restricted to a singular reading.
2.1.4. Classifiers and reference
As a further difference between our minimal pairs (9) and (10) we noted refe
rential connotations: classifiers in the context of adjectives signal definite re
ference to specific objects or they stress contrastive properties as conveyed by
the co-occurring adjectives.
In order to account for this fact, we must revert to the pronounced differ
ence between demonstratives and adjectives.21 For demonstratives by them
selves always serve the very function of referential identification, thus there is'
no need to employ classifiers in any referential function in phrases like (9) b.
Adjectives, on the contrary, by themselves have no referential properties
at all, as (10) a. shows. Yet, they are frequently needed for referential pur
poses, i.e. for identifying subsets of objects in terms of contrastive properties
they may happen to have: specific green umbrellas vs. red ones etc. This is:
where classifiers are made use of for referential purposes: since adjectives are
referentially neutral, their combination with a classifier may serve to compen
sate this deficiency and thus bring about a definite interpretation of the
SYNTAX AND SEMANTICS OF NUMERAL CLASSIFIERS IN THAI 177
phrase. This secondary function is readily compatible with the one noted so
far: classifiers, as we found, always relate to the quantification of objects.
Therefore, in the absence of other referential clues, they may be utilized to re-
ferentially identify these objects.
In much the same way classifiers may acquire a fairly strong contrastive
referential value. That is, in phrases like (11) rom khan sii-khiaw nii “this
green umbrella”, reference is sufficiently specified by the demonstrative. The
classifier preceding the adjective in this case simply stresses the contrastive
value of the adjective as an essential identifying property of the object refer
red to. Similarly, repetition with several adjectives may serve such contrastive
purposes, as in
(15) a. rom khan sii-khiaw khan jaj
“the big green umbrella”.
Or the classifier may be used to emphasize the one contrastive property of an
object among its other merely descriptive ones:
{■' ) (15) b. rom sii-khiaw khan jaj
“the big green umbrella”.
In this way classifiers may be utilized in quite a number of variants of NPs.22
It must be kept in mind, however, that no referential or contrastive func
tions are inherent in classifiers by themselves: for their obligatory use in overt
counting constructions does not imply referential properties. Phrases like (1)
a. are perfectly neutral in this respect.
The essential function of classifiers, as far as could be shown from the
analysis of our minimal pairs, consists in relating phrases like (9) b. and (10) b.
to quantitatively specifiable objects, while their unmarked counterparts are
absolutely neutral as to quantity. They can be shown to form implicit counting
constructions, which, for their actual quantitative interpretation, depend on
the kind of nominal determiner they occur with.
2.1.5. Further environments
Though not in principle different from the constructions discussed so far, the
remaining distributions mentioned above should be illustrated briefly.
1) Classifiers may precede ordinal numbers, in which case they function as
they do with adjectives.
Ordinal numbers consist of the morpheme thii (basically a noun meaning
“place where”23) and a following numeral (for ordinal use of simple numerals
178 HARALD HUNDIUS & ULRIKE KOLVER
struction.
2) Like numerals, some non-numerical quantifiers have to be followed by a
classifier, while others do not.
The first type comprises lexemes such as baarj “some, a few”, Ifiaj
“many”, thuk “each”, kii “how many” etc., e.g.:
(17) rom laaj khan
“many umbrellas”.
The other type is exemplified by lexemes like baarj “some, a little”, maak
“much”, thag “all”, thawraj “how much”, etc.
The difference, as suggested by some of the English glosses, once more
reflects the distinction between counting and measuring constructions discus
sed above in 1. Indefinite quantifiers of the first type quantify individual ob
jects as such, while the other type denotes quantities not conceptualized as
consisting of discrete items. Thus, mass nouns freely combine with the second
type, as in
(18) mil naamman maak/bdag
“there is much oil/some oil”
while count nouns usually occur with the first type.
Under certain conditions though, count nouns may take the measuring
type of lexeme, too , viz. with reference to quantities too large to be readily es
timated in figures. Thus one can say
(19) a. mii nakrian maak
have student much
“there are lots of students”.
Against that, a phrase
(19) b. mii nakrian laaj khon
have student many (clf:pers)
“there are many students”
implies that those referred to are still perceived individually. A quantifier like
Wap “some, a little” will hardly occur with count nouns.
3) Furthermore, there is a distributive variant of classifier construction:
(20) khdw syy sdm liiuk Id nyg bdat
he buy orange (clfrfruit) per one baht
“he bought oranges a baht per piece”.
180 HARALD HUNDIUS & ULRIKE KOLVER
naturally as “one”.
This means, instead of contrasting an “obligatory” use with an “optional”
one (as the literature usually describes it), we can now say more precisely that
classifiers occur to indicate that a NP applies to specified or specifiable
numbers of individual objects. In this way they serve to introduce the very no
tion of quantity the NP is otherwise devoid of. Thus, when a NP is meant to
refer to individual objects in terms of overt lexical means, such as numerals
and certain indefinite quantifiers, the classifier is their necessary correlate en
suring the quantitative meaning of the phrase, — a point which is going to be
more fully illustrated presently. In other contexts, involving determiners which
do not lexically relate to quantity, the choice is between phrases neutral as to
quantity (and, in the case of adjectives, neutral as to reference as well) and
phrases implying a quantitative interpretation.
Classifiers thus provide the specific technique Thai employs to refer to in
dividual objects in terms of quantity. The central question then, of course, is
why this particular technique is necessary. Why is it that quantitative mean
ings cannot be simply expressed by combining nouns with lexemes denoting
quantitative notions?
To answer this question we have to consider a) the status of nouns and b) the
status of numerals in Thai.
2.2.1. Nouns
As is the typical case in isolating languages, Thai nouns do not in themselves
contain any numerical or referential indications. They always have to occur in
their one unalterable form, irrespective of their context in a sentence. This
means, in themselves nouns are in no way modifiable so as to reflect the vary
ingsituational settings of the corresponding objects.
When a noun appears as the sole constituent of a NP in a sentence like
(22) plaa waaj jiiu naj mee ndam
fish swim remain in river
it allows for a certain range of different interpretations, roughly shown in the
following glosses:
a. fish swim in rivers (generic)
b. the fish(es) swim(s)/swam in a river/rivers
c. a fish/fishes swim(s)/swam in the river/a river ... etc.
Unless more context will narrow down the choice, such sentences, which
are perfectly customary, in principle have quite a number of possible read
182 HARALD HUNDIUS & ULRIKE KOLVER
the syntactic function of individuation. That is, the categorial distinction be
tween concepts and objects manifests itself in terms of a limited set of lexical
features, so that counting objects is achieved with reference to certain salient
features only, while the full complexity of the conceptual universe embodied
in nouns remains outside the scope of countability. This network of classifi-
catory features selected from the entire grid of those theoretically conceivable
will be discussed presently.
The essential, functional point, however, is the categorical distinctionas
such, rather than a consistent systematization of nouns in terms of lexical clas
ses.. i
This point, which the syntactic investigation has made evident, is also
borne out in the lexical inventory of noun classes itself.
For, as is the case in most numeral classifier languages, the inventory of
pre-established lexical classes in practice fails to fit for a good many nominal
notions. Thai copes with this deficiency by using two distinct methods which
are semantically opposed, yet logically cognate. One might be called hyper
generalization . It consists of one particular classifier being used irrespective of
any specific semantic feature of the word so classified. The clearest instance of
this method is the classifier ?an, which may be applied to practically any noun
short of those denoting living beings.
The other tendency might be termed over-specialization, which is a most
prominent phenomenon in Thai. In its most extreme form, it has led to the
emergence of repeater constructions which, from a semantic point of vieware
peculiar, indeed, and which have gone unexplained until now.28 These are
cases where a noun recurs as its own classifier, as in
(25) thaleesaap sdam thaleesdap
“three lakes”
or as in (5), above. Evidently, a notion of class does not make sense with re
petitions of this type ( — and we shall revert to this lexical deficiency: see
4.4.; 4.5. below). Yet, far from being the odd occasional exception, as one
might perhaps surmise, such complete semantic redundancy recurs with hun
dreds of Thai nouns. This testifies to the productivity of the pattern (which, in
cidentally, is well attested in some other numeral classifier languages29).
The classifiers in (5) and (25) are devoid of semantic content, are seman
tically redundant. By this very property they reveal the categorial distinction
in its purest form. They prove that, while the semantic, classificatory capacity
of the category may well be reduced to zero, the categorial distinction be
SYNTAX AND SEMANTICS OF NUMERAL CLASSIFIERS IN THAI 185
The remaining two paragraphs will deal with the semantic aspects of the clas
sifier network. We shall first, in this paragraph, discuss the general question of
meaningfulness and variablility of classifers in relation to the notion of intrin
sic feature.
Paragraph 4 will then document in detail the diverse semantic parameters
and their mutual affiliations. Finally, we shall try to outline the more general
results emerging from this survey.
3.1. Intrinsicfeatures and qualification
Classifiers, as we said in 1.2., always express intrinsic features of their respec
tive head nouns and thus do not contribute any independent lexical meaning to
a NP (and this is a fortiori true for the lexically fully redundant pattern just
mentioned).
This point is related to the question of whether classifiers qualify the
meanings of their head nouns, — an assumption which has been repeatedly
expressed for numeral classifier languages. Thus Adams and Conklin put
down the difference between measuring and counting constructions (see 1.2 .)
asone of quantification vs. qualification: “.. .numeral classifiers qualify rather
than quantify the head noun, and as such require the presence of some par
ticular inherent feature” (1973:2). Similarly Plam says with respect to Thai
classifiers that they “precise la signification du mot mediat (i.e. the head
noun) correspondant (ainsi que le determinant precise le determine)”
(1972:197). In particular the fact that nouns do not infrequently allow for a
choice among several possible classifiers has often led to the conclusion that
suchvariation may cause changes of meaning in the head noun.30
We cannot here discuss this assumption for numeral classifier languages
generally. However, Thai, unlike e.g. Vietnamese with its different order reg
ularities, provides unambiguous evidence that classifiers here are not in
volved in altering the meanings of their associated nouns.
Before we illustrate this fact, a few general considerations may be useful.
For, evidently, the notion of qualification is hardly compatible with that
of intrinsic features: a word simply is not qualifiable in terms of features in
cluded in its own semantic structure.
A comparison with adjectives will readily show this: adjectives, indeed,
186 HARALD HUNDIUS & ULRIKE KOLVER
conceptually qualify head nouns which is the very reason why tautological
combinations like *“unmarried bachelor” are usually precluded. That is,
applicability of an adjective to a noun presupposes a genuine choice of values
within a given semantic parameter suitable to the semantic structure of that
noun. Violation of this general rule either leads to tautologies (which may or
may not be thought acceptable in a given case) or else to downright contradic
tions of terms as in *“married bachelor”.
By contrast, classifiers are not chosen among several values of the same
parameter. Variation of classifier is tantamount to variation of semantic stan
dard. As a principle rule, classifiers display the very kind of tautological re
lationship to their head nouns which is ruled out for adjectives in most cases.
While adjectives require certain scopes of choice, classifiers more often than
not may not vary at all, and, if they may, they vary in terms of different seman
tic criteria.
This, in our opinion, rules out qualification as an adequate description of
the semantic relationship between nouns and classifiers.
3.2. Variation of Classifiers
Apart from these general observations, in Thai there is clear evidence to illus
trate this point. For one thing, in most cases classifiers do not vary for a given
noun. For another thing, evidence comes forth from the very cases where
nouns do admit of a choice of classifier.
This can be shown, for example, from one of the stock items of classifier
variation, viz. plant denotations in their various designations as stem, fruit,
leaf etc. respectively31; compare
(26) a. mamuag saam luuk
mango three (clfrfruit)
“three mangoes” (i.e. fruits)
b. manual] saam baj
(elf: leaf)
“id.”
c. mamuagsaam phon
(clf:fruit)
“id.” •
According to the most unhestitating assertion of native speakers, these varia
tions do not coincide with a difference in meaning: a. and b. are fully equiva
lent in meaning, while c., with the Pali loanword for “fruit” as its classifier,
SYNTAX AND SEMANTICS OF NUMERAL CLASSIFIERS IN THAI 187
connotes a more elegant stylistic level. In all three cases mamuai7preserves its
meaning of “mango fruit”. This is noteworthy in viewof b. where the classifier
has the basic nominal meaning “leaf” (and, by associative extension, is
applied to a variety of concepts, see 4.2.1. and 4.5.1.). Nonetheless b. unam
biguously denotes mango fruits, not mango leaves. This means that the clas
sifier baj, in spite of its original nominal meaning “leaf”, does not map this
meaning on to the meaning of the head noun so as to convey the meaning of
“three mango leaves”.
In order to express this latter notion, there is no other way than first to
form the corresponding nominal compound bajmdmuarj “mango leaf” which
then, in its turn, is appropriately classified under baj (as a semi-repeater, see
3.3.):
(27) bajmamuag sdam baj
“three mango leaves”.32
Analogous formations are encountered for other denotations of parts of
plants, e.g.:
(28) a. tonmamuat) sdam ton
“three mango trees”
b. dookmamuaq sdam dook
“three mango blossoms”
but not *mamuai7sdam ton or *mamuai] sdam dook.
In any of these cases, application of a given classifier presupposes
adequate nominal word formation, while the classifier cannot map different
meanings on to the head noun. It does not participate in the processes of con
ceptual modification, it always presupposes these processes.
This seems plausible and quite in keeping with the syntactic function clas
sifiers have been shown to fulfil. For, if they were “used up”, as it were, in a
qualifying function on the conceptual side, they could not at the same time
serve to refer to individual objects on the other side.
In (26) above, variation of classifier is not matched by a change of mean
ing. Against that, compare the following examples where different classifiers
do coincide with a change of meaning of the phrase:
(29) a. posaamluuk
bamboo-fishtrap three (clf:fruit/round shape)
“three fishtraps” (of a special kind)
188 HARALD HUNDIUS & ULRIKE KOLVER
b. posaam baj
lampshade three (clf:leaf)
“three lampshades”
(30) a. kuncEE saam dook
key three (clf:flower-shaped obj)
b. kuncEE saam ?an
wrench three (clfdong obj)
“three wrenches”
(31) a. daaw saam duarj
star three (clf:round obj)
“three stars, planets”
b. daaw saam khon
star three (clf:person)
“three stars (i.e. stage or film stars)”.
These cases do not invalidate our previous evidence. For, as the interlinear
versions show, it is not the classifier that changes the meaning of the noun; the
case must rather be stated the other way round: the classifier changes in con
sequence of prior change of the meaning of the noun. This, in (29), simply con
sists in homophony of two altogether different lexemes: while po in a. is a
genuine Thai word,po in b. is a loanword from Chinese, and each is assigned
independently to its appropriate class.
(30) is a case of polysemy of the noun kuncEEwhich, no matter whether it
occurs in a classifier construction or not, has to be disambiguated by context as
to its possible readings “key” and “wrench” respectively. This may, in actual
speech, come about by the classifier providing the first conspicuous clue as to
which reading is intended in cases where the context has left room for misun
derstanding. Such a pragmatic disambiguating value, however, is entirely dif
ferent from the function of conceptual qualification.
Next, (31) is a clear case of metaphorical extension of a basic nominal
meaning where the classifier in each reading is chosen in accordance with the
intended meaning.
• Finally, let us consider an example which appears as a borderline case
with respect to change of meaning:
(32) a. ryysii saam ?or)
hermit three (clf:sacred being)
“three hermits”
SYNTAX AND SEMANTICS OF NUMERAL CLASSIFIERS IN THAI 189
4. Noun Classes
In the following pages the inventory of genuine classifiers in the sense defined
ini. above is documented for the present state of Standard Thai as completely
as we could ascertain from our sources, i.e. available dictionaries and
supplementary information from native speakers (see Introduction and Re
ferences).
Deviating from other lists, as e.g. that in McFarland (1941:xiv), which,
running to about eighty lexical items, contains both classifiers in our sense and
quite a number of measuring terms, we do not list measure terms here at all,
for the reasons indicated in 1.: measures usually do not enter lexicalized re
lationships with specific sets of nouns and they form an open category readily
available to innovation, as e.g. is shown by the recently adopted Western deci
mal measures.34
Against this, classifiers are lexicalized in the code itself according to in
trinsic features of the corresponding nouns, which the following lists attempt
to systematize.
We have not aimed at quantifying our data, for obvious reasons: nouns
are an open category so that any statistical figure on correlations between
nouns and classifiers would just arbitrarily reflect the input of the data used.
And, still more obviously, statistics of linguistic categories is futile in principle:
it does not give us any insights into how a language organizes its structural and
semantic oppositions.
However, in order to illustrate the internal organization of the classifier
192 HARALD HUNDIUS & ULRIKE KOLVER
as kings, priests and deities, at the one end and beings of somewhat sinister or
at least dubious supernatural faculties, such as demons and ascetics, at the
other.
By contrast, animals are represented in one class only (with one culture-
bound exception); again there are some overlappings with nouns denoting
human beings.
4.1.1. Human
(1) khon l.n. “person”, 2 . general clf for nouns denoting persons,
except sacred and royal ones. Productive clf, substitutable
for any of the more specific ones.
More specialized classes relating to social status:
(2) caw l.n.(a) “prince, ruler”, (b) “owner, master”, 2 . clfrper-
sons engaged in commercial and monetary transactions, e.g.
hdapree “peddler”, luuknii “debtor”, cawnii “creditor”. Re
stricted to very few nouns.
(3) naaj l.n. “master, employer”, 2 . address with proper names of
men, “Mr.”, 3.pers. pron. 2nd ps. (in address to sons and
younger brothers)37, 4.clf: men of some social standing, pub
lic or official position (implying politeness of the speaker),
e.g. tamruat “policeman”, tidaakaan “judge”,pheet“doctor,
physician”. Fairly productive class.
(4) naarj equivalent to naaj with reference to women.
(5) than l.pers pron. 2nd, 3rd ps. (implying respect), 2. title with
nouns indicating rank and with proper names, 3.clf:persons
of high social rank, official position, lesser nobility, e.g. ?aa-
caan “teacher, professor”, naajdk “prime official”,
rathamontrii “minister”, cawphrajaa (highest degree of con
ferred nobility). Fairly extensive class.
4.1.2. Human —Superhuman
(6) ruup l.n. “form, shape, picture, idol”, 2 . clf:priests and idols (of
deities), e.g; phraphlksu “Buddhist priest, monk”, chii
“ascetic”. Restricted to very few nouns. Also with some
nouns denoting inanimate things, cf. 4.2.1.2.
194 HARALD HUNDIUS & ULRIKE KOLVER
(12) saaj l.n. “line, channel, route”, 2.clf: line-like objs. of larger
size, e.g. khhorj “canal”, meendam “river”, rayooij “rope (of
sailing ships)”, thanon “street”. Fairly extensive class.
(13) sen 1.numerative “line, strand”, 2. clf: line-like (or curved)
objs. of smaller size, e.g. phom “hair”, jaj “fibre, web”, soo
“chain”, laajmyy “line on the palm of the hand”. Frequent
overlappings with saaj. Fairly extensive class.
(14) thaaq l.n. “way, path, road”, 2. clf: roads and paths, e.g.
honthaar) “road, way”. Very few nouns, mostly compounds
with thaar) as first element, the clf saaj being preferred with
nouns denoting ways and roads.
(15) khabuan 1. n. “procession, train”, 2. clf: trains. The only examples
are rotfaj “train” and rotduan “express train”.
4.2.1.2. Saliendy flat shape
(16) phen Clf: flat shapes generally, e.g. chanuan “slate”, kradaan
“board”, paaj “poster, label”, caanslag “phonograph re
cord”, bkepfoom “blank form (to be filled in)” (from Engl.),
saket “flake, chip”. Very extensive class.
More specialized classes:
(a) FLAT AND INFLEXIBLE SHAPE:
(17) baan 1. bound nominal morpheme in compounds such as baan-
pratuu “wing of a door”, 2. clf: flat, mostly rectangular objs.,
e.g. pratuu “door”, kracdkpaw “mirror”, ndataap “win
dow”. Few nouns.
(18) duarj 1. bound nominal morpheme in compounds such as
duapcan “moon”, duarj?aathlt “sun”, 2. clf: round, radiant
objs., e.g. daaraa “star”,?aathit “sun”, khoomfaj “lamp”,
takiap “lamp, lantern”, satlem “stamp” (from Engl.),
duarjtraa “seal”. The classification seems to reflect a disk-like
198 HARALD HUNDIUS & ULRIKE KOLVER
(20) baj l.n. “leaf’, 2. elf: leaves and leaf-like objs., e.g. thanabat
“bank note”, chal&ak “label, lottery lot”, tua “ticket”,
beepphim “blank form”. Further, baj classifies things made
of leaves and stalks, especially containers. The class thus in
tersects with the field of round, three-dimensional objs., see
4.2.1.3.(b). Very productive class.
(21) phyyn 1.numerative “piece, strip” (of cloth or land), 2. elf: gar
ments of flat pieces of cloth, flat textile products, e.g.
phaanutj (lower garment, a cloth tied round the waist),
phaacettua “towel”, maan “curtain”,phrom “carpet”'. Fairly
extensive class.
(24) met: l.n. “seed, grain” (short form of malet “seed, grain”), 2.
numerative “grain”, 3. clf: small globular objs., e.g. khajmuk
“pearl”,phlooj “gem, jewel”, kradum “button”,/?/ “pustule,
boil”. Very few nouns.
(25) koorv. 1.numerative “lump, piece”, 2. clf: bulky shapes, e.g.
krtiat “stone, pebble”, ntw “gall stone”, thaanfajch&aj “bat
tery, dry cell”, meek “cloud”. Very few nouns, koort in most
cases occurs as a measuring numerative.
(b ) THREE-DIMENSIONAL AND HOLLOW SHAPE:
(20) baj cf. ab. 4.2.1.2., baj also classifies containers, e.g. kraddp
“bamboo tray”, takraa “basket”, krathop “tray” (made of
banana leaves), turn “earthen jar”. Further, round, hollow
objs. e.g. kloop “drum”.
Finally, baj overlaps with luuk for fruit names, an application
which may be based on the observation that some fruits such
as coconuts contain a hollow space or grow round kernels.
From this point of connexion with the feature /hollow/ it may
have been generalized to any fruit name, (cf. below 4.5.1.).
Very productive in this use too.
(26) foop l.n. “bubble”, 2. clf: bubbles and eggs, only examples:
foopn&am “water bubble”, foop?aakaat “air bubble”, khaj
“egg” (and its compounds).
(27) paak l.n. “mouth”, 2.clf: some open hollow objs., e.g. Me
“fishing net”, sawlp “small fishing net”, plee “hammock,
cradle”. Very few nouns.
(28) ryan l.n. “house, housing”, 2.clf: clocks, the only example is
naallkaa “clock, watch”.
There are three more classifiers for special irregular shapes, viz.:
(29) ddok l.n. “flower”, 2.clf: flowers and some flower-resembling
shapes, e.g. kuncee “key”, phlu “firecracker”, luukthanuu
“arrow”, het “mushroom”. Very few nouns in the latter
usage.
200 H A R A L D H U N D IU S & U L R IK E K O LV ER
A few classifiers do not denote any specificlexical feature, but simply classify
things as “discrete items”. These apply when none of the more special ones
seem suitable: that is, if no typical shape is envisaged, as e.g. with nouns like
khbopkhw&n “gift”, khaawkhoop “belonging, equipment” which denote a
functional destination rather than a specific article. Classifiers available in
such cases are:
(1) ?an (cf.4.2.1.1.) general residue clf.
(37) chin 1. numerative “piece”, 2. clf: thing.
(38) sip clf: thing, item.
(39) beep l.n. “model, type”, 2. clf: model, item.
(40) khdap clf: single item of a structurally connected pair.
(The last two lexemes border on numeratives of type 1: measures, grouping
units and types, cf. 1.2. above)!
4.2.4. Function
Two classifiers based on function are set apart from the other ones; these are:
(41) chabap clf: written documents, e.g. cdtmaaj “letter”, chanoot
“title deed for pieces of \and.”, phanthabat “financial bond”.
chabap (noteably a loan word from Khmer) is separated here
from other clfs relating to thought and language (4.3.), for
the reason that it specifically takes into account the medium
of script, while the other clfs are not sensitive as to the actual
medium of representation.
(42) khryap l.n. “apparatus, instrument, machine”, 2.clf: technical
contrivances, machines. With nouns denoting traditional im
plements it only occurs as a semi-repeater. It has turned pro
ductive with nouns denoting modem machinery, e.g.
thoorasap “telephone”, thoorbphaap “television set”, kon-
lacak “engine”.
202 HARALD HUNDIUS & ULRIKE KOLVER
Here there are two groups, viz. (a) mental or verbal concepts and (b) events.
4.3.1. Mental or verbal concepts:
(1) kham l.n. “word”, 2.clf: words and expressions, verbal units
generally, mostly compounds of the noun kham itself, e.g.
sap “word”, krijaa “verb”, khunnasap “adjective”, khamdaa
“abuse”.
(2) khoo l.n. “joint, node”, 2. clf: mental units (including their re
sults in spoken or written language), e.g. panhda “problem”,
khati “moral, motto”, khawq$an “clue”; many again com
pounds of khoo itself, e.g. khooklut “idea”, khootdojeeg “ar
gument, point of dispute”.
(3) prakaan 1.numerative “kind of (mental or verbal unit)” 2.clf: men
tal and verbal units. More literary alternative for khoo in
many cases.
(4) r$aq l.n. “story, subject matter”, 2.clf: literary products, topics
of speech, e.g. tamnaan “legend”, bdtkhwaam “article” (in a
newspaper), lakhoon “drama”, khaawlyy “rumour”.
(5) bdt clf: religious texts, precepts, literary products of some re
ligious flavour, lessons, e.g. kham?athitthdan “prayer, wish”,
mon “sacred word, incantation”, khloog (a kind of Thai
verse form), khampraphan “poetry, writing”,.bdtrian “les
son”.
(6) maattraa l.n. “measure, standard”, 2. clf: laws, juridical texts,
clauses, e.g. kotmaaj “law”, banjat “law, regulation”. Very
few nouns, more commonly khoo is employed.
(7) cop l.v. “to be finished”, 2.clf: the only example is kaanteet
“sermon”.
4.3.2. Events
4.4. Repeaters
At the end of our survey it seems necessary briefly to revert to the highly pro
ductive device of simply repeating a noun as its own classifier which, as we said
above, occurs with hundreds of nouns so that it seems well worth asking which
kinds of nouns defy semantic generalization within the network of genuine
classifiers.
From checking up a sample of about five hundred repeater nouns roughly
the following semantic ranges seem to emerge:
(1) Non-physical or non-visible objects, abstract notions
(2) Objects ill-accessible to immediate visual perception, such as geographi
cal units
(3) Objects of irregular or variable appearance
(4) Body part terms
(5) Tabu-concepts or concepts of ill-omen
, The groups (1) — (3) are closely akin to each other with obvious possible
transitions between them. Some few examples out of the vast number are:
(1) khwdti “psyche, spirit, morale”, chiiwlt “life”, wtchaa “subject of study”,
wlthii “method”, ch^achdat “nationality, race”, ch$y “name”, pooramaanuu
“atom”, klin “odour”.
(2) keen “axis”, thaan “base, foundation”, khiit “limit”,khua “pole (of earth,
magnet)”, kheet “area, zone, boundary”, look “world”, pratheet “country”,
myap “town”, kd “island”, khaapsamut “peninsula”, byp “large swamp”
(3) khaneen “mark”, traa “brand, seal, trademark”, kradaj “stair, staircase,
ladder”, paw “target”, laaj “pattern, design”.
204 HARALD HUNDIUS & ULRIKE KOLVER
By contrast, group (4) contains body part terms, especially those denot
ing external, visible parts (while some nouns denoting internal organs are clas
sified by shape features). The reason is obvious enough: a human being’s own
visible, tangible body parts will naturally be conceived of as unique, vitally in
tegrated to the self and hardly associated with the idea of quantifying objects.
Against that, counting of internal organs under normal conditions will typical
ly refer to those of animals, not of men.
The last group contained only few examples in our sample, though possi
bly there are many more, as e.g. saak “corpse”, loop “coffin” (which, inciden
tally, may alternatively be classified simply as a container under baj, cf.
4.2.1.3.). These obiously denote concepts set apart as having some emotional
impact on the tabu side of life.
4.5. Conclusions
It has become apparent from the list above that the network of classifiers by no
means provides anything comparable to a folk taxonomy: for one thing there
are neither definite hierarchical layers41nor clear-cut boundaries between the
classes. For another thing, the most striking impression after surveying these
lexical classes is the evident unevenness of distribution: while some few sem
antic domains are differentiated fairly thoroughly, vast areas of nominal con
cepts are not captured in the network at all.
4.5.1. Complex features
Before reverting to this conspicuous fact, a few additional remarks on the
manifold intersections of the complex features pervading the core of the net
work seem necessary.
For, on the one hand, in some respects the criteria chosen are so similar to
each other that the possible correspondences between nouns and classifiers
leave considerable room for variation. Thus, there are abundant examples
where classifiers may be substituted for each other and the classes as a whole
almost merge into one another. E.g. those for long objects lim, khan and
ddam (long, pointed and handled objects) form overlapping classes which
moreover typically intersect with the general classifier ?an. Even less clear-cut
is the distinction between sdaj and sen (larger and smaller, long, flexible ob
jects). Similarly,phen (flat objects) overlaps with hay (flat, flexible and hollow
objects) j luuk (three-dimensional objects) with baj and met (small globular
objects).
Apart from the evidence for classifier variation cited in 3. above, these
SYNTAX AND SEMANTICS OF NUMERAL CLASSIFIERS IN THAI 205
cases amply illustrate the fact that classifier variation does not trigger concep
tual modifications of the corresponding nouns.
Conversely, the very productive classifiers often cover strikingly wide
and diversified ranges of nominal notions.
These productive classifiers altogether amount to only about twenty lexi
cal items, for convenience summarized here: ■!.
(a) Animate: (1) khon (person), (10) tua (animals)
(b) Inanimate: (1) ?an (long objs. and residue clf) (2) ton (plants), (3) lem
(long pointed objs.), (4) khan (long handled objs.), (12) saaj and (13) sen
(long flexible/curved objs.), (16)phen (flat objs.) (19) duag (round, radiant
objs.), (20) baj (flat, flexible and hollow objs.), (29) luuk (three-dimensional
objs.), (30) heeg (localities), (38) chabap (written documents), (39) khryag
(machines)
(c) Immaterial: (1) kham (words), (2) khoo (mental and verbal units), (3) ryag
(literary products, topics), (7) khrag (events)
While some of these, like khon, actually do denote simple, ultimate fea
tures, most items of this core inventory establish very complex diversified
classes.
Cases in point are e.g. baj and Iduk. What emerges here is a kind of asso
ciative technique of gradual expansion and shifting of features which accounts
for their complexity and productivity. Thus baj, from its nominal meaning
“leaf’, is adopted as a classifier for leaves, then, due to similarity of appear
ance, for flat flexible shapes generally. At the same time the basic meaning
“leaf’accounts for the inclusion of leaf-made obj ects in this class, such as trays
and baskets.
This in turn serves as a starting point for associating baj with container
nouns in general (many containers traditionally being made out of leaves any
way), irrespective of their actual material quality. And, in turn again, this ex
tension leads to associating the feature of hollowness with baj, as well as that
of flatness and flexibility. By yet another associative step the shift /container
hollowness/ may have led to linking baj with certain typically hollow fruits,
such as coconuts (which furnished common traditional containers). And this
latter association once more seems to have been carried a step further by ex
tension to fruit names generally. In this way there seems to be a chain of asso
ciative links such that the classifier first applies to some special set of nouns
from which a generalized feature is extracted, repetitions of this associative
process leading to gradual interconnexion of ultimately widely diverging con
cepts. ' ; i>:
206 HARALD HUNDIUS & ULRIKE KOLVER
Since we have not gone into diachronic investigations here, the proposed
chain cannot be definitely verified at present, yet it would seem that the intri
cate intertwining of features (and the obvious connexions between the corres
ponding objects in the language community) carry their own suggestive force
and plausibility.
A similar, more easily transparent chain seems to hold for luuk as well:-
starting (as it were, where baj ends up) from the nominal meaning “fruit,” it
applies, as a classifier, to nouns denoting fruit and, by visual similarity once
again, to globular and bulky shapes generally. This in turn leads to associating
the notion of three-dimensionality (or else, of undifferentiated “bulkiness”)
so that luuk is extended to any kind of three-dimensional shape, irrespective
of size or actual specific contour and comes to be applied to things as diverse as
mountains, waves, drums, dice etc.
Rather than neatly pigeonholing the diversity of concepts, classification,
as it appears from its lexical distribution, operates most productively by asso
ciative interconnexion of nominal concepts.
Now, while examples like baj and luuk provide evidence suggestive
enough, other classifiers like lem (4.2.1.1.), which are not linked synchroni-
cally to a source lexeme, do not afford similar plausible interconnexions in con
temporary Thai. Thus, naturally, quite a few of the peculiarities noted
stand a chance of explanation on diachronic evidence only, which, in the pre
sent state of Thai historical linguistics, for obvious practical reasons has not
been aimed at here. We did, however, occasionlly question native speakers of
traditional erudition and of different ages.
The fact that native speakers usually handle even those synchronically ill-
transparent classifiers unhesitatingly and consistently seems to testify to the
productivity of this associative, interconnective technique.
4.5.2. Form and Function
Two diversified groups of classifiers relate to man himself in his social and re
ligious world and to his most distinct capacities for thought and speech.
The majority of classifiers, however, serve to group the diversity of con
cepts of the inanimate world in terms of shape features, with a marked de
crease of differentiation from one-dimensionality to three-dimensionality:
flat shapes being much less differentiated than long ones and three-dimen
sional ones hardly at all.
Now, while motivation by shape is perfectly obvious for most of the pro
ductive classifiers, the question may arise, however, whether this is also true
for the different, more specialized classes within this general pattern. What of
SYNTAX AND SEMANTICS OF NUMERAL CLASSIFIERS IN THAI 207
such classes as e.g. law (for wind instruments), thaap (for roads),phyyn (for
textile pieces), ryan (for clocks)? Are they genuinely classified by shape or
rather by their inherent functional properties as instruments, roads, clocks
etc. respectively?
The answer to this question, it seems to us, ties in with our observations
on complex features so far: the dominant features are those that are based on
shape, while features such as rigidity/flexibility, hollowness, size and those
even more specialized ones pertaining to just some such closed groups as the
above are merely concomitant.
That is, while the more general shape features often may be substituted
for the more special classifiers (as e.g. saaj (long flexible/curved) for thaap
(roads)), this does not apply the other way round.42
Moreover, all of these specialized groups single out objects of some strik
ing unmistakeable shape from the more comprehensive shape classes. A con
clusive example is e.g. law for wind instruments: if function were the motive to
group them apart, one would expect either to find a classifier for musical in
struments in general of else, separate ones for diverse types of such. Yet, what
we do find is that stringed instruments are classified under khan (on account of
being handled) and drums etc. under luuk or baj (on account of being three-
dimensional and hollow at that).
In view of this we would think that the general observation43that numeral
classification is based most prominently on shape as the most distinctive cate
gory of visual perception is fully supported by the Thai evidence.
Such visual features, however, may be perceived by observers in differ
ent communities in somewhat different ways: they may appear as conspicu
ously different in just one dimension, while differentiation along several di
mensions may increasingly appear as too complex for representation. Thus,
certain shape features may be summarized as one overall impression (as luuk
for any kind of three-dimensional shape), while others may be isolated as typi
cal of just a specific, limited set of objects (such as those special classifiers just
cited).
The exceptional classifiers of 4.2.2. and 4.2.4. support this view: for (a)
classifiers relating to localities and buildings either overlap with repeaters
(that is, nouns denoting localities ill accessible to direct visual surveying). Or,
as far as classifiers for buildings are concerned, they might as well have been
subsumed under the group of three-dimensional hollow objects where there
are suggestive links between e.g. ryan (nominal meaning “house, housing”,
the clf for clocks) and lap (clf for houses and house-like structures). However,
208 HARALD HUNDIUS & ULRIKE KOLVER
since the evidence here is not conclusive enough, it seemed best to retain these
as a group of their own.
(b) Clearly those of 4.2.4., exceptionally, are motivated by function. They
isolate notions which represent very important and most spectacular cultural
innovations from the entire network: chabap (a loan word from Khmer, Cam
bodia having been of vital influence on the historical development of Siam)
marks off script, and khryap, in modern times, sets apart nouns denoting tech
nical instruments and machinery transcending ordinary man and animal
power. This, as in earlier times the introduction of writing most likely did, may
have struck the speakers more by what these innovations achieve than by what
they look like.44
4.5.3. Scalarity in Classification
We are now led back to the observation that, while classification is well
worked out for concepts belonging to the immediate sphere of man and for in
animate things readily discernible by shape, wide semantic fields are not
genuinely classified at all. This means that nouns are either lumped together
in broad general classes (as e.g. hep for localities) or treated as unclassifiable,
i.e. repeaters.
Among these latter the conspicuous group of body part terms (see 4.4.
above) forms part of the semantic field of notions immediately related to man
himself.
As for the vast diversity of inanimate objects, the distribution of the clas
ses clearly reflects a scale of decreasing accessibility to easy visual perception,
corresponding to a converse scale of increasing uniqueness of concepts.
This is first from salient one-dimensionality to salient three-dimensional
ity where there is a marked decrease of differentiation in terms of classes.
Next down the scale there are:
(a) irregular shapes, for which there are very few special classifiers or the res
idue classifier?an, or, finally, repeaters.
(b) visible objects without a definite contour; events: variable objects (such
as presents, equipment articles etc., which are represented simply as “items”
by ?an, chin (piece) and sin (piece); secondly, events, simply represented as
“instances” by khrap.
(c) visible objects too large or too complex and diversified to be apprehended
and identified as to typical contour, such as geographical units, which are rep
resented by either hep (localities in general) or as repeaters. Further, such
SYNTAX AND SEMANTICS OF NUMERAL CLASSIFIERS IN THAI 209
things as designs, patterns and the like, represented either by ?an or as repeat
ers. ■! ■: ,
Yet further down the scale there are:
(d) visible qualities other than shape, like colour (sfi “colour” is a repeater)
and invisible objects which are apprehensible by other senses, represented
mostly as repeaters, e.g. stag “sound, voice”.
(e) immaterial objects, abstract notions, emotional values etc., represented
as repeaters.
It hardly needs pointing out that this scale perfectly matches the syntactic
function of classifiers as analyzed above: for one thing shape, a readily discer
nible contour, is the most suitable feature for apprehending objects as indi
vidual items. For another, the scale outlined corresponds to a decrease in
countability or, conversely, to an increasing scale of immateriality or unique
notions. This ends up with nominal concepts which will hardly require the syn
tactic marking of individuation, concepts which have no corresponding, as it
were, “objective” counterparts.
NOTES
1) For this general framework see Seiler and Lehm ann (eds) 1982.
2) Cf. e.g. G reenberg 1974; Seiler and Lehm ann (eds) 1982.
3) Cf. e.g. H aas 1942; E m eneau 1956:16; Noss 1964:104ff; Jones 1970:1; G reenberg 1974:29;
Goral 1978:23; Lehman 1979:155.
4) The term is now most commonly used, while quite a num ber of other term s have been
employed in the less recent literature, cf. G reenberg 1974:29. The term numerative introduced
below in this paper takes up this more neutral terminology, in accordance with current terminology
in the Cologne Language U n iv e rsa l Project.
5) Num erals in Thai may occur as first members of certain nominal compounds such as siamlio
‘three-wheel’, “three-wheeled pedicab” w here the num eral forms part of the process of word for
mation.
6) Noss 1964:104ff. bases his definition of classifiers (numeratives in our sense) on stress fea
tures: he claims that they are distinguished by ‘weak stress’. This criterion, however, seems less reli
able than word order.
. 7) The term is adopted from Hla Pe 1965:181
8) Cf. Noss 1964:104ff., H aas 1942 and 1964. H aas assigns the symbol C (= classifier) to any lexi
cal entry that may occur in the position following a num eral, without any further subcategorization.
Other lists of Thai classifiers, as e.g. in McFarland 1941, do not m ake a distinction between m ea
sures and classifiers either, though M cFarland at least uses the more comprehensive term ‘num eri
210 HARALD HUNDIUS & ULRIKE KOLVER
tween adjectives and relative constructions. It would go far beyond the scope of the present paper,
however, to describe in detail the conditions of such attributive use of verbs, since this would in
volve an analysis of the valency of Thai verbs in general. It must suffice here to use the term adjec
tive with reference to attributive use of verbs in a general sense.
19) O n determ ination as a continuum between the extrem es of maximum conceptual characteri
zation and maximum referential specification cf. Seiler 1977, esp. 1Iff.
20) Lehm an 1979 argues on the basis of Thai examples that numeral classifiers serve the function
of set-partitioning. H e puts down the reverse order among determ iners to a difference between
cardinal and ordinal sets.
W ithout going into his wider fram e as to various other num eral classifier languages, it must be
noted here that he does not fully analyze the Thai examples into their ultim ate constituents. Thus,
he assigns the function o f set-partioning to classifier phrases like khan n(i as a whole (155) without
isolating the ultim ate functions of khan and nti respectively. Now, set-partioning is one of the main
functions involved in nominal determ ination generally. T herefore, this designation does not speci
fically explain the function of classifiers. If we are to arrive at a precise functional designation, the
minimal pairs such as (9) and (10) and their expansions in (13) and (14), which Thai fortunately puts
at our disposal, must be taken into account. For these enable us to isolate the functional values of
the ultim ate elem ents and to account for the different order regularities.
21) Cf. Seiler 1977: Ilf . a n d 23.
22) Jones 1970:7 claims that each adjective must be preceded by the classifier. This is not in ac
cordance with the actual facts: adjectives do not require classifiers obligatorily anyway, con
sequently, combinations of several adjectives do not either.
23) Cf. H aas 1964:s.v.
24) Cf. Noss 1964:112.
25) W e owe this observation to Lehm an 1979: 162f. How ever, his argum ent, that there are
more severe restrictions for combinations of cardinal and ordinal num bers within the same N P in
Thai than there are in English does not seem valid. F or phrases like “the fourth five m en” etc. sound
as odd in English as they doubtless are in Thai. The restriction on combining different types of
counting to num erals not higher than “two” or “three” probably is not due to language specific con
straints, but rather to limits of easy perception.
26) The same view has been expressed for another num eral classifier language o f the isolating
type, viz., Vietnam ese, by E m eneau 1951:94 and the note therein.
27) For the notion of individuation as a functional continuum realized in a range of cognate
techniques across languages cf. Seiler and Lehm ann (eds) 1982.
28) Cf. G reenberg 1974:29, and more specifically, Allen 1977:295 who declares himself some
what mystified by this particular phenom enon.
29) Cf. H la P e 1965:180ff., also Okell 1969,1:213.
30) Cf. e.g. Chao 1968:507; Nguyen D inh Hoa 1957:127
31) Cf. e.g. Em eneau 1951:96.
32) Noss 1964:64f., 105 and Blagonravova 1971 regard compounds such as bajmdmuaij as con
sisting of classifier + noun, apparently because baj usually occurs as first m em ber of compounds
rather than as an independent noun. This analysis would seem to state the case the wrong way
212 HARALD HUNDIUS & ULRIKE KOLVER
round: it implies that original classifier lexemes are adopted into the noun category. The general
syntactic and semantic evidence, however, clearly points to the reverse pattern: original nouns are
adopted as classifiers, not the other way round. T hat boy usually occurs as a bound morpheme is no
objection: bound nominal m orphem es which occur as heads of compounds, but never as classifiers,
are by no m eans rare in Thai. Thus there seems to be no need to postulate an exceptional type of
compound for bajmamuap and the like,
33) Chao 1968:507 points out for Chinese classifiers that “the association of specific classifiers
with nouns is primarily the association of words and only roughly in term s of m eaning.” This is to a
certain extent true for Thai as well, since the classes are fixed in the code itself so that conceivable
alternatives of standard are subject to fixed lexical rules. However, these standards are in principle
based on meanings of words rather than on the words themselves.
34) Traditional m easures certainly are an interesting field of study, yet relatively more relevant
anthropologically than linguistically.
35) Cf. Becker 1975:109
36) Cf. Loyns 1968:470ff.
37) Against the fairly rough social distinctions captured in classification independently of rela
tive position of the individual speaker, Thai paired personal pronouns supply a most elaborate sys
tem of distinctions which mirrors minute variations of the mutual relations between persons in pri
vate and public life. Cf. Cooke 1968.
38) The noun phra “the B uddha” might be expected to be treated as a unique concept remaining
outside the scope of countablity. While this is ultimately true, Thai religious and mythological no
tions, essentially influenced by Indian traditions, also reflect the typical Indian idea of different
manifestations of what is ultimately the same deity.
39) lem etymologically corresponds to hem “pointed, sharp” .
40) khan in its other meaning “dike of a rice field” probably has a different etymology. In North
ern Thai the lexemes are still differentiated by different tones.
41) Obviously there are hierarchical traits to the network acknowledged here in terms of domin
ant vs. concomitant features. Yet the tendency towards hierarchy is in principle counteracted by
the cross-classificatory tendencies pervading the entire structure.
42) Cf. Adam s and Conklin 1973:6: “no secondary [i.e. concomitant] feature can be the sole
basis for a classification” .
43) Cf. e.g. G reenberg 1974:34, Adam s and Conklin 1973:5f. Friedrich 1970:382ff.
44) This holds equally for the functional connation noted for khan with reference to modern
vehicles.
REFERENCES