LINGUIST List 5.466

Thu 21 Apr 1994

Sum: Numeral Classifiers

Editor for this issue: <>

Directory

Message 1: SUM: NUMERAL CLASSIFIERS

Date: Wed, 13 Apr 94 12:34:09 SSSUM: NUMERAL CLASSIFIERS
From: David Gil <ELLGILDNUSVM.bitnet>
Subject: SUM: NUMERAL CLASSIFIERS

Summary: Numeral Classifiers (VERY LONG!)

Following (better late than never) is a summary of three discussions that took place during 1993 and 1994 on the e-mail LINGUIST list, on the topic of numeral classifiers.

The first discussion examines the internal syntactic constituency of constructions of the form NUMERAL CLASSIFIER NOUN. The second conversation deals with constructions of the form CLASSIFIER NOUN, where the numeral is absent. And the third conversation is concerned with constructions of the form NUMERAL NOUN, in which the classifier is absent.

For each discussion, I have presented the original query, and then an edited summary of the responses.

THE FIRST QUERY:

I am fishing for constituency tests for numeral classifier constructions both in "numeral classifier languages" and in languages like English.

Consider the following phrases in Mandarin and their equivalents in English:

shi bang rou ten pound meat "ten pounds of meat"

yi guo fan one pot rice "one pot of rice"

san ge ren three unit man "three men" [no English equivalent with classifier]

My gut feeling is that the constituent structure for such phrases is different in "numeral classifier languages" than in English- type languages; specifically, that it is [NUM CL] NOUN in languages like Mandarin, but NUM [CL NOUN] in languages like English. However, when I started looking for solid evidence in support of this claim, I found it surprisingly difficult to come up with good constituency tests in either English or Mandarin.

I would therefore appreciate any suggestions and/or references pertaining to the constituent structure of numeral classifier constructions in any language, of either the English or the Mandarin types. Explicit constituency tests would be most welcome.

(Also, I wonder whether there might be language-internal variation between different classes of classifiers, for example "measure" classifiers, eg. "pound"; "container" classifiers, eg. "pot", and others.)

A further note: In numeral classifier languages, the sort of evidence I am familiar with in support of [NUM CL] NOUN constituency derives from tone sandhi (in Mandarin), the coalescence of NUM and CL into a single word (in Japanese), the ability of the [NUM CL] constituent to "float" (in Japanese), and the occurrence of NOUN NUM CL word order (in Thai). In contrast, in Vietnamese, Thompson's reference grammar seems to suggest that the constituency is NUM [CL NOUN], and the Vietnamese linguists who I've had occasion to consult would appear to agree.

*******************************************************

SUMMARY OF RESPONSES:

The responses were rather surprising (to me at least). With regard to "numeral classifier languages" such as Mandarin, there was a consensus that [NUM CL] NOUN is the appropriate constituency -- and nobody had anything new to contribute with respect to Vietnamese. However, with regard to English, there was also general agreement that [NUM CL] NOUN is the correct constituency -- contrary to my initial assumption. Finally, two correspondents provided valuable general comments without specifically supporting a particular constituency.

In the meantime, I have applied several of the proposed constituency tests, and well as others, to Mandarin, and the case for [NUM CL] NOUN constituency seems about as clear cut as can be. However, in English, the facts still seem to me to be inconclusive.

Following is a slightly edited version of the comments, arranged according to language, with the more general comments at the end.

*******************************************************

JAPANESE (Alan Hyun-Oak Kim)

I am very much interesting in the subject you posted. I have just finished my final draft of a paper in which I dealt with something relevant to the subject. The title of the paper is: Word Order at the Noun Phrase in Japanese: Quantifier Constructions and Discourse Functions. It is an extensively revised version of my paper originally presented at The Milwaukee Linguistics Symposium on "Word Order in Discourse" a couple of years ago. The paper will be included in a John Benjamins book edited by Noon and Downing: "Word Order in Discourse." For this study, I collected a little over 1,000 samples of Quantifier constructions from Japanese written texts (in MJ and OJ), examined them to determine the taxonomy and functions in discourse. Although I did not make any specific arguments for the Japanese Q-constructions as being in the Num-CL:N structure, one will find ample evidence for such a view.

*******************************************************

CHINESE

******************** (Guy Modica)

But as to your question, it seemed to me that you provided some constituency tests in your posting. Maybe this was too trivial for you to mention, but a (num-cl) constituent can stand alone, as can a (N) constituent. Therefore, a test for Vietnamese would be to check the grammaticality of the (cl-N) and (NUM) constituents in isolation, neh?

******************** (Gary Palmer)

Must Mandarin numerals always occur with classifiers? If so, that would be evidence in favor of your interpretation.

******************** (Shu-ing Shyu)

Maybe the following sentences can help. The N can be topicalized, and focalized.

(1) Rou ta mai-le shi bang. meat he buy-ASP ten pound

(2) Lian rou ta ye mai-le shi bang. even meat he also buy-ASP ten pound 'He even bought ten pounds of meat.'

(3) Ta rou zhi mai yi bang. he meat only buy one pound 'He only bought one pound of meat.'

******************** (Sze-wing Tang)

Have you read C.-C. Jane Tang's (1990) PhD dissertation 'Chinese Phrase Structure and the Extended X'-Theory'? She supposes that numeral and classifier both generate in a Cl(assifier) node and they function as a compound in Mandarin Chinese.

*******************************************************

EASTERN AUSTRONESIAN (Malcolm Ross)

I would be very interested to know whether you have got any further with this problem, as I face a similar one in trying to deal with possessive classifiers in Oceanic Austronesian languages.

Incidentally, the only evidence I have on the constituency of numeral classifiers (in the few Oceanic Austronesian languages that have them) runs exactly parallel to Japanese: the numeral and the classifier coalesce into a single word. Would you like the references?

There are two groups of Austronesian languages in Melanesia with numeral classifiers, the Kilivila group and the Eastern Admiralties (= Manus) group. The first has several member varieties, of which by far the best described is Kilivila (= Kiriwina) itself. Two published descriptions are:

Senft, Gunter, 1986. Kilivila: the language of the Trobriand islanders. Berlin: Mouton de Gruyter.

Lawton, Ralph, 1993. Topics in the description of Kiriwina (ed. Malcolm Ross & Janet Ezard). Pacific Linguistics D-84. Canberra: Australian National University.

The latter has more on classifiers, as it includes Lawton's Master's thesis on the subject.

Senft has spent a lot of time studying classifiers in Kilivila, esp. in relation to children's learning of them, I think. He has numerous papers on the subject, but I have not kept up with his work. If he has not responded to your original message, you could contact him via e-mail (Gunter Senft, cogantmpi.kun.nl) at the Max Planck Institute for Psycholinguistics in Nijmegen.

The only description of an Eastern Admiralties language that comes to mind is:

Hamel, Patricia J., 1985. A grammar of Loniu. Ph. D. dissertation, University of Kansas,

soon to be published in Pacific Linguistics (the proofs are currently with the author).

Hope this helps. These refs are descriptive, not theoretical.

*******************************************************

GERMAN (Frans Plank)

German arguably has numeral classifiers (Plank, Morphologische (Ir)Regularitaeten, Tuebingen: Narr, 1981: 142-8); they look like nouns but they differ from them in failing to inflect, with the numeral-classified noun or rather an adjective accompanying it taking the appropriate case inflection - e.g. nach drei Glas Bier, lit. after three glass beer (also possible, though subtly differing in meaning: nach drei Glaesern Bier, with Glaesern, dative plural as required by the preposition, a straight noun), nach drei Glas warmem Bier "after three glass warm-DatSg beer" (but you probably also get: nach drei Glaesern warmem Bier "after three glasses-DatPl warm-DatSg beer"). Dissociation/fronting may suggest that the classifier forms a constituent with the numeral:

(1) Er hat drei Glas (warmes) Bier getrunken he has three glass (warm-AccSg) beer drunk (2) (Warmes) Bier hat er drei Glas getrunken (warm) beer has he three glass drunk (3) *Glas (warmes) Bier hat er drei getrunken, Flaschen nur eine glass (warm) beer has he three drunk, bottles only one

(3) may not be ungrammatical, but sounds to me much worse than (2). My judgment needs checking, though. (Added in proof: I have asked two natives now: neither accepts (3) as even marginally possible.) (Nouns ending in -en in the plural, such as Flasche/Flaschen, do not form classifiers.) There is probably no such difference with the inflected- noun alternative (but one would have to check this too with other natives):

(4) Er hat drei Glaeser (warmes) Bier getrunken He has three glasses (warm-AccSg) beer drunk (5) (Warmes) Bier hat er drei Glaeser getrunken (6) ?Glaeser (warmes) Bier hat er drei getrunken, Flaschen nur eine

If the numeral-classified noun is accompanied by an adjective or case-inflecting determiner, you get a further alternative construction, viz. a plain genitive. What about constituency here? Hard to tell, because any dissociations are odd.

(7) Er hat drei Glaeser warmen/dieses Bieres getrunken he has three glasses warm-GenSg/this-GenSg beer-GenSg drunk (8) *Warmen/*Dieses Bieres hat er drei Glaeser getrunken (9) ?Glaeser warmen/dieses Bieres hat er drei getrunken

*******************************************************

ENGLISH

******************** (Jim Hurford)

Question: How much rice did he eat? Answer: Three pots

Conventional way of itemizing things in a shopping list: Sugar, 3 pounds Bread, 2 loaves Wine, 4 bottles

The following is more artificial than the first above question/answer sequence:

Question: What did he eat three of? Answer: Pots of rice.

******************** (Paul Kershaw)

An obvious substitution:

I got ten pounds of butter and several (more) of cheese.

This suggests that English is [Num Cl] Nn, not the other way.

******************** (John Koontz)

For English, as a constituency test, how about the possibility of:

one battered pot of rice (vs. one fragrant pot of rice) one rather generous pound of meat (vs. one smelly pound of meat) poss., two well-horned head of cattle (vs. two emaciated head of cattle) etc.

where the inserted modifier applies to the classifier/measure, not to the number phrase as a whole and not to the noun.

> Since my previous message, I've been pondering some more over your examples, and it seems to me that the crucial ones are those like

one fragrant pot of rice

where the modifier seems to take in its scope a [CL N] constituent; in > the above example [pot of rice]. Is this what you had in mind?

As far as I can recall, yes. I don't have a copy of the letter in hand. I think I supplied a series of examples of the form Numeral Modifier Classifier/Measure [of] Noun, in some of which the modifier applied to Measure, while in others a different modifier applied to the measured Noun instead. However, I suppose that the latter sort of example can be handled with whatever technique is used for forms like `a polite cigar told me that she'd just left' or `an elegant suit was just taking my seat', etc.

Incidentally, I notice that `a battered ricepot' is possible, but *`a fragrant ricepot' isn't. I think the problem is that in the latter fragrant has to apply to pot, whereas in `a fragrant pot of rice' it can apply to either pot or rice (and has to apply to rice, logically).

*******************************************************

******************** (Stavros Macrakis)

I'd think you could do tests along the lines of:

I bought ten pounds of meat. I bought ten pounds and three ounces of meat. How much did you buy? Ten pounds. How many pounds did you buy? Ten. How many pounds of meat did you buy? Ten. What did you buy? Meat. What did you buy? *Pounds of meat. (acceptable in the sense "a lot") Meat is sold by the pound.

Same pattern for "head of cattle", "cases/bottles of beer", "loaves of bread", "lengths of rope", "lengths/skeins of yarn".

There is marginal acceptability when the classifier distinguishes kinds of material, e.g.

What did you buy? ?Lengths of yarn. vs. What did you buy? Skeins of yarn.

I have no theoretical proposals, but perhaps this kind of construction will be helpful.

*******************************************************

UNIVERSAL

******************** (Pamela Downing)

In response to your query on LINGUIST, I'd say that the kinds of tests you're looking at seem to make sense to me and apply in many, many classifier languages (NUMERAL classifier languages, that is). I'd also add the fact that numeral + classifier phrases can appear without benefit of a co-occurring noun in most of these languages as evidence in favor of your intuition that the split is between the noun and the numeral + classifier phrase in most cases. Maybe you're getting different responses from your Vietnamese informants because of the murkiness of the noun - classifier line, and of the classifier construction - compound noun line in Vietnamese, something that may not be such a problem in many other classifier languages. As for references, I suppose you're already aware of Greenberg's papers: "Numeral classifiers and substantival number: problems in the genesis of a linguistic type" Working Papers on Language Universals 9:1-39 (1972). and "Dynamic aspects of word order in the numeral classifier" in Charles Li, ed. Word Order and Word Order Change. Austin: University of Texas Press. 1975. Pp 27-45. On Vietnamese, there's also Karen Adams 1982 dissertation, Systems of Classification in the Mon-Khmer, Nicobarese, and Aslian Subfamilies of Austroasiatic. (Ann Arbor: University of Michigan) UMI order # 8304433 (I think) Shigeru Miyagawa discusses some of the constituency issues in Structure and Case Marking in Japanese (Syntax and Semantics, Vol. 22) Academic Press, 1989 and cites some other papers by Mamoru Saito that you might be interested in. I have a paper on Q-Float in Japanese (from a functionalist perspective) which makes a fair number of references to other languages as well which was supposed to have come out in the Journal of Linguistics in March but which has still not appeared. Akio Kamio has also written some stuff on the constituent structure of classifier phrases in Japanese, if you can read it or get someone who can, e.g., Kihon Koozoo. Tokyo: Sanseidoo. Pp. 77- 126. 1983.

******************** (Jim Hurford)

As far as I know, in numeral classifier languages, the classifier ALWAYS come adjacent to the numeral. That is, of the 6 possible orders:

NUM CL NOUN CL NUM NOUN NOUN NUM CL NOUN CL NUM occur, but

NUM NOUN CL CL NOUN NUM don't occur.

Don't ask me for chapter and verse on this right now.

*******************************************************

GENERAL COMMENTS

******************** (David Stampe)

If you consider things like Engl. `stick of candy' or `head of beef' or `piece of ass', it seems to me that one good reason for taking the so-called classifier to go with the noun, rather than a numeral, is that its role is to furnish an individuation of a so- called "mass" concept. But in a language like Chinese where there happens to be no special class of count nouns, probably because there is no plural to demarcate them, so all nouns are treated like mass nouns, it would not be surprising that since numerals always take classifiers, while nouns don't (unless they're being counted), the constituency might go the other way.

There are some Munda languages (India) in which only cattle and people take classifiers, and they tend to coalesce with the numbers, e.g. two-man men, two-head cows. (Munda languages have plurals even on these nouns.) But coalescence can't be taken uncritically. Unaccented elements tend to attach rather permanently to preceding accented elements, and not to following ones, for exactly the same reason that unaccented syllables can shorten a preceding accented syllable, but not a following one: rhythmic measures in speech as in music are isochronous groupings including everything from the accent up to the next accent, unless there is a clear morphosyntactic (or harmonic) boundary that prevents it. So for example Norwegian det hund `the dog' remains two words while the younger hundet (also `the dog') has coalesced as one word. More generally, postpositions readily become case suffixes while prepositions rarely if ever become case prefixes (to answer a question Donegan and I left open in our 1983 paper cited below). In fact often they attach to anything accented that precedes them: two-a those. So, if NUM CL NOUN -> NUMCL NOUN, as in Munda, that may not be conclusive evidence of the (original) constituency. And for the synchronic construction, it seems to me that NUMCL is simply an inflected numeral, not a construction at all. Speakers don't identify the meaning of the suffix even though it's phonetically similar to the words for head, man, etc. It has become part of the morphology; the question of syntactic constituency has become moot.

I don't have any novel syntactic tests to suggest, but recent student papers and dissertations I've read on classifiers in mainland SE Asian languages have made me suspect, in languages that differ from the areally usual head - modifier pattern only in their classifier/noun order, that whatever element is ordered last is being interpreted as a modifier. Otherwise how are we to understand such apparent deviations from their uniform head - modifier order? Maybe the usual tests about what can be omitted (under what circumstances) could be tried to test the head vs modifier status of the elements, but I haven't done this systematically.

The floating of the [NUM CL] phrase in Japanese, which I hadn't heard about, might furnish a special test, if it depends on what is assumed and what is new in the particular utterance. Novel information tends to be treated as modifier, according to a general hypothesis put forth in a paper by Patricia Donegan and me on word order, drift, accent, etc., in the Chicago Linguistic Society Parasession on the Interplay of Phonology, Morphology, and Syntax (1983), that the modifier (which can be questioned relative to the head) is accented relative to the head, and word order typology and drift correlate with accentual typology and drift, would suggest some tests involving possible interrogatives and possible accentuations. It is not hard to imagine that `two bushels of rice' could answer a question about `what kind of bushels?' (as indeed it does in grain market quotes, where NUM CL i.e. one bushel is omitted, and only the NOUNs are given: corn, oats, etc.) as well a question about `how much rice?'. Note that we CAN ask in English `How many sticks of candy?' or `How many head of cattle?', taking the [CL NOUN] as given and asking for a numeral modifier. (We can't very well ask (with normal interrogative structure) `six what of candy?', `six sticks of what?', etc.

It's interesting that by various tests, demonstratives often vary from being modifiers to being heads in various languages. Most grammarians bridle at the idea of noun modifying a demonstrative, but by the test of omissibility it is common enough to be able to say just `that' for `that hat'. One doesn't even have to know that `that' refers to a hat: we can ask `What's that on your head?'.

Regardless of whether the language requires numeral classifiers, the order NUM NOUN is found in many languages which normally put modifiers last. I would guess that this pattern comes from the order of NUM modifiers of other NUMs. In most languages, NUM NUM puts a modifying (multiplier) number before the head (multiplicand, if I don't have the math terms backwards), e.g. three hundred = 300, in small large order, while conjoined (added) numbers are in the order large small, e.g. hundred three = 103. I explained the large+small order as due to new-information-last in counting (twenty-EIGHT, twenty-NINE, THIRTY, thirty-ONE) in a paper in the 1976 CLS, and then noted that combining this with small*large order in multiplying makes any number self-parsing. (Exceptions to large+small come mainly in compounds in languages or earlier stages of languages where accent was initial, e.g. TWO-and-twenty, THREE-and-twenty, etc., or frozen in THIRteen, FOURteen, etc. The law that new information goes on the accent is even less breakable than the new-information-last law.) Whatever the explanation, the pattern of NUM-modifier NUM-head is so common that it could influence the pattern of NUM-modifier NOUN-head.

It is interesting to speculate whether this pattern could extend to MODIFIER NOUN phrases in general, and explain the use of MODIFIER NOUN in Chinese and ADJ NOUN in English, where normally modifiers are after heads. But Donegan and I think (I don't think we got this into our1983 paper) that these oddities are due rather to the influence of the MODIFIER-NOUN pattern in the thousands of nominal compounds in English (and other Germanic languages) and in Chinese. Latin lacked many such compounds, and Romance had nothing to prevent Latin's ADJ NOUN pattern from becoming (except in fixed phrases) NOUN ADJ. There is a mirror image oddity in Tibetan: normally modifier first, but NOUN MODIFIER, and with thousands of NOUN-MODIFIER nominal compounds. Ironically if we assume that the compounds came from noun phrases, this means that prehistoric Chinese must have been modifier-first while prehistoric Tibetan must have been modifier-last. (We'll let the Sino-Tibetanists solve that. :-)

******************** (Edith Moravcsik)

This is re constituency within numeral classifier phrases. Some evidence might come from coordination. If, in saying 'two pots of rice and four pots of rice make six pots of rice', you can say "two pots and four pots of rice...", this would support "two pots" being a single phrase. If you have to say "two and four pots of rice...", this would go against this analysis. Another source of evidence may be ellipsis in answers. If, to the question 'How many pots of rice do you have?', you can say " Three pots.", this supports NUM-CLS as a phrasal unit; if you have to say "Three.", this would go against this claim. Ellipsis in coordination may also help. If in a language you can say "I ate three pots of rice and he ate four pots.", this would support NUM-CLS phrasehood; if you say "I ate three pots of rice and he ate four.", this is against the claim. A further source of evidence might be interruptability. Can anything go between NUM and CLS; or between CLS and N? Non-interruptability of either sequence would support phrase-hood. Also, pronouns: do question and relative pronouns stand for NUM only or for NUM-CLS?

It seems very likely to me, too, that different classifiers would participate in different constituency relations relative to the Num and the N. Perhaps different NUM-s and different N-s do, too.

THE SECOND QUERY

In languages with numeral classifiers, the most common construction is of the form (abstracting away from word order)

(1) [NUM CL N]

ie. a tripartite construction consisting of numeral, classifier and noun in some order.

However, in some languages with numeral classifiers, there exists an alternative "bare classifier" construction of the form

(2) [CL N]

ie. a bipartite construction consisting only of classifier and noun.

The purpose of this query is to solicit any information, data, thoughts or references that anybody out there may have on the bare classifier construction in (2).

So far, I have come across bare classifier constructions in two languages.

In Vietnamese they appear to be quite common; in all the examples I have seen cited, the interpretation of the resulting NP is singular, and in most or all of the examples it is also definite. For example "con cho" "CL dog" "the dog".

In Mandarin they are much less frequent, and seem to be constrained in their distribution in various ways; from informant work here in Singapore, the interpretation of the resulting NP is singular (like Vietnamese) but invariably indefinite (unlike Vietnamese).

(Note: for a construction to qualify as a bare-classifier construction, it is necessary that the numeral classifier not double as a member of some other category, eg. noun. For example, in Vietnamese, many classifiers are also nouns; however, some classifiers aren't, and it's the latter that provide the true examples of bare-classifier constructions.)

So here are some more specific questions:

(a) To speakers of Vietnamese: is it really the case that bare- classifier NPs are always singular and definite?

(b) To speakers of (or linguists familiar with) numeral-classifier languages other than Vietnamese and Chinese: does your language have bare-classifier constructions? Please note that I would be equally appreciative of negative evidence (especially such that I could quote) to the effect that language X DOESN'T have bare-classifier constructions. So what about Japanese? Korean? Khmer? Thai? Any other numeral-classifier languages?

(c) To speakers of languages with bare-classifier constructions: any examples, information, references, thoughts, etc. would be extremely welcome.

*******************************************************

SUMMARY OF RESPONSES

With regard to Vietnamese (question (a)), Helge Dyvik provided a very detailed and specific response, showing that in Vietnamese, bare classifier constructions may be interpreted as either definite or indefinite, and as either specific or non-specific -- though, like in Mandarin, they are always singular.

More generally (questions (b) and (c)), the pattern of responses was too spotty to draw any firm cross-linguistic conclusions. In particular, different respondents often differ as to whether bare classifier constructions occur in the same language. (This variability is consistent with my own informant work on Mandarin, in which different speakers frequently offer very different judgements with regard to such constructions.)

Geographically, the occurrence of the bare classifier construction appears to focus on the Chinese-Vietnamese area. According to the responses, it is marginally attested in Japanese and Korean, and sporadically in various Tai languages; however, it is said to be absent from one Karen dialect, and from Samoan. (Unfortunately, I still have no information on many linguistic areas with numeral classifiers, such as North-East India, Melanesia and Central America.)

Semantically, the bare classifier construction usually interpreted as singular, exceptions being provided by the marginal constructions in Japanese and Korean, and perhaps also by the Tai language Lungming. As for (in)definiteness, here the facts appear to be of considerable interest. Whereas in Mandarin, Shanghainese and Teochew the construction may only be interpreted as indefinite, in Cantonese and in Vietnamese, definite interpretations are also available. This seems to suggest a north-south pattern, with the availability of definite interpretations increasing from north to south; however, Thai appears to pattern with the northern Chinese languages, with only indefinite interpretations.

Following is a slightly edited version of the comments, arranged according to language.

*******************************************************

SAMOAN (Arnfinn M. Vonen)

A lot of Polynesian languages have a classifier (in Samoan to'a-) occurring with numerals (which are often relative clause predicates as in (i)) when the counted entities are human beings. Samoan:

(i) le tagata e to'a-tasi SPEC.SG person NONPAST HUMAN-one 'one person'

\ tagata e to'a-lua SPEC.PL person NONPAST HUMAN-two 'two persons'

This classifier is not 100% obligatory; e.g., it is rarely used when one is actually counting people aloud.

The occurrence of the classifier is totally dependent on the presence of the numeral:

(ii) * le to'a-tagata

Some Polynesian languages, e.g., Samoan, have a number of other classifiers as well, mainly restricted to use with nouns denoting various food items. Their syntactic properties are largely the same as those of to'a. Quite a few of them are either suffixed or prefixed to the numeral; in the latter case they denote ten times as many items, as in lua-lau (two-CL) 'two (large fish)' vs. lau-lua (CL-two) 'twenty (large fish)'. Many of the food classifiers are becoming obsolete and are little used, as the way of life in Samoa is changing. The following list is somewhat simplified; e.g., a ligative element sometimes intrudes between the numeral and the classifier:

-aea 'scores (20) of coconuts' afii- 'tens of bundles of fish' (cf. afii 'bundle of fish') -'au/'au- '(tens of) bananas, yams, etc.' (cf. 'au 'team' or 'au 'stalk') -'aui/'aui- 'tens of skipjack' (certain dialects) -fua/fua- '(tens of) breadfruit, coconuts, fowls, or certain shellfish' (cf. fua 'fruit, flower, egg') -lau/lau- '(tens of) fish large enough for cooking wrapped in a leaf' (cf. lau 'leaf') -mata/mata- '(tens of) taro' (perhaps cf. mata 'eye' or mata 'raw'?) -'ofu/'ofu- '(tens of) bundles of food (except fish)' (cf. 'ofu 'dress, garment, clothes') -oa 'pairs of coconuts or young pigs' pootoi- 'tens of balls of food' (cf. pootoi 'ball of food') sautua- 'layers (of paper, ropes etc.)' tau- 'things in bunches or clusters (e.g., coconuts)' -tino/tino- '(tens of) skipjack' (certain dialects) tua- 'rows, layers, thicknesses etc.' tu'e- 'tens of crabs or lobsters' (cf. tu'e 'edible part of crab or lobster')

In cases not covered by any of the classifiers, no classifiers are used. Source: Mosel, U. & E. Hovdhaugen. 1992. _Samoan Reference Grammar_. Oslo: Scandinavian University Press. Pp. 190, 246-250.

Thought you might be interested in a classifier system apparently more based upon function (food value) than upon such standard formal parameters as shape. But the classifiers never occur without a numeral.

*******************************************************

JAPANESE

******************** (Francis Bond)

Japanese is a numeral-classifier language, but doesn't have bare classifier constructions. Thus while (1) and (2) below are fine, (3) and (4) are impossible.

(1) inu ippiki `dog one small-animal' (2) ippiki-no inu `one small-animal of dog' (3) * hiki inu `small-animal dog' (4) * inu hiki `dog small-animal'

where small-animal is the classifier used for dogs.

I confirmed this with the native speaker seated on my right.

******************** (Peter Hendriks)

For counters in Japanese, a good place to start is S. E. Martin (1975) A reference grammar of Japanese, pp. 766-782, and references contained therein.

One possible example of a bare-classifier in Japanese might be - bai "times", as in:

Ooki-sa wa bai gurai aru deshoo 'size' TOP. "times" approx. 'have' 'probably' "As for size, it is probably about twice as big"

"bai" by itself (the numeral would precede it, for ex: san-bai 'three times' ni-bai 'twice') is the ony one in Japanese that I can think of at the moment which can occur without a numeral. When it does, it means 'twice'. "bai" is not used as a noun or as any other part of speech.

*******************************************************

KOREAN

******************** (Chung-Min Lee)

Not easy to describe.

Ton phwun -ina iss -nun saram -i way kuren cis -ul hayss-ul- kka? money cl as much as Rel man Nom why such behavior-Accdid- Presum-Q 'Why did a man who has at least some amount of money do such a thing?'

N Cl -ina (or -irato) became almost frozen (idiomatic).

******************** (Hyunoo Lee)

As far as I know of Korean, there exist no such bare classifier construction. In this language, classifiers must be accompanied by numerals or some quantifiers.

*******************************************************

SHANGHAINESE (Bao Zhi Ming)

Shanghainese has a bare numeral classifier construction with a singular indefinite interpretation similar to Mandarin.

*******************************************************

TEOCHEW [SINGAPOREAN] (Geraint Wong)

Teochew has a bare numeral classifier construction with a singular indefinite interpretation similar to Mandarin.

*******************************************************

CANTONESE

******************** (Marjorie K. M. Chan)

Mandarin Chinese does not permit it, but Cantonese Chinese does. Hence, Cantonese Chinese allows both: [Num CL N] and [CL N].

******************** (Hilary Chappell)

Re Chinese classifiers - Cantonese works similarly to Vietnamese from your description. Bare classifiers code definiteness of the noun referent. Cantonese is quite unlike Mandarin in this respect.

******************** (Ho Chee Lick)

Cantonese has a bare numeral classifier construction with a singular interpretation, but unlike Mandarin, it may be interpreted as either indefinite or definite.

******************** (Dan Jurafsky)

Cantonese looks like Vietnamese, not like Mandarin. (In fact, this is exciting for me, since I've been collecting various evidence for the (well-known) position that Cantonese has a strong south- east asian leaning/substratum.)

So bare-classifier NP's are singular and DEFINITE.

Definite example: (note no numeral)

Q: Where did i write that address? A: you wrote it on that sheet of paper gloss of answer: WROTE AT FLAT-CLASSIFIER PAPER LOCATIVE- PARTICLE

Indefinite example: (Note required numeral)

Do you have a sheet of paper? YOU HAVE NOT HAVE ONE FLAT-CLASSIFIER PAPER?

Another definite example: (again, no numeral)

Where's the book? BOOK-CLASSIFIER BOOK AT WHERE?

Incidentally, since Cantonese is usually thought to have a Thai substrate, i would look at Thai if i were you, i bet it acts like this too.

*******************************************************

VIETNAMESE

******************** (Helge Dyvik)

I am definitely not a speaker of Vietnamese, but I studied Vietnamese classifiers at one point and tried to find an answer to the same question that you ask. Hence I would be grateful if you would share the answers you get with me (or perhaps you intend to post a summary to the list?).

Writers on Vietnamese generally claim that "con cho", "cai ban" express definiteness, which ought to mean unique reference: the presupposition on the part of the speaker that the hearer is able to identify the referent uniquely. I consulted four Vietnamese informants, asking them whether a sentence like "Toi thay cai ban" 'I see the/a table' would be acceptable in a situation where the hearer has no previous knowledge of a particular table, the table is not visible to him, and the speaker knows it. The informants were unanimous in voting 'yes': this would be possible. Assuming that they really understood the kind of situation I sketched, this indicates that definiteness is not (necessarily) expressed by the construction, although it often *translates* definite constructions in Western languages. On the other hand, singular reference seems to be clearly expressed by the construction - not surprisingly, given the individuating function of classifiers. One might speculate that the construction expresses *specific* reference - but it can occur in opaque contexts without such reference. The informants tell me that "Toi tinh mua con cho" 'I plan to buy the/a dog' can be used when the speaker has no specific dog in mind. Still, there is some kind of specificity or scope-related effect involved - cf. the following pattern of grammaticality from Phong: "Le probleme des classificateurs en Vietnamien" (1975), p. 11:

Meo so chuot cat fear mouse 'Cats feat mice'

Con meo so chuot 'The cat fears mice'

Con meo so con chuot 'The cat fears the mouse'

*Meo so con chuot

If the "con chuot" construction expresses something like relatively wide scope of the existential quantifier, this pattern is explained given that topic position also expresses relatively wide scope, which is standardly assumed. For then we get a conflict of scope indicators in the last example. (These are my reflections, not Phong's.)

Cantonese has the same construction, by the way.

> You also say that it's quite natural for these constructions to be > interpreted as singular, given the individuating function of > classifiers. But isn't it still puzzling that plural is ruled out: > after all, strictly speaking, classifiers force *count* > interpretations, which can be either singular or plural.

I agree: I put it a little too simply. But assuming that individuation/countability is the marked option in Vietnamese (our plurals usually translating as nouns with a non-discrete collective interpretation), it seems justified to ask: why individuate if you don't give us the cardinality of the set by means of a numeral? It seems difficult to imagine any reason except the intention to denote a single individual.

******************** (Stephen H. Houchen)

I have been learning Vietnamese for about a year now, and thought I might respond to your query. Such "bare classifier" constructions are very common in Vietnamese and in many cases grammatically mandatory. Ex:

cai' muo^~ng = spoon cai = inanimate object classifier chie^'c du` = umbrella chie^'c = " " " con heo = pig con = animal or child ngu+o+`i cha = father ngu+o+`i = person etc...

It is never correct to say "muo^~ng" without the classifier, for example. With other words, the classifier is optional. To answer your question, as far as I have seen, [CL N] is always definite and singular. Usually to indicate an indefinite noun, the form is [mo^.t CL N] where "mo^.t" is "one", "a", or "an".

Note that there is also a pluralizer "ca'c" that can replace the usual classifier to emphasize plurality. Ex: ngu+o+`i ba.n = friend, but ca'c ba.n = friends. This type of construct seems to correspond to both definite and indefinite in English. Is this what you're looking for?

*******************************************************

THAI (Wirote Aroonmanakun)

- Thai do have a bare-classifier construction as N-CL. Example: paak1kaa0 daam2 pen CL

- This construction is not widely used. We may view it as a deletion of indefinite "one" in a construction N-CL-"one", which is not the usual form of numerative NP (N-Num-CL). So, it's singular but indefinite. It'll be definite if it is followed by demonstrative markers, "this", "that".

- This construction could be intervened by Aspectual marker. pai0 yip1 paak1kaa0 maa0 daam2 (nung0) go pick pen (COME) CL one

*******************************************************

LAO [BUA] (Connor Ferris)

This dialect of Lao (spoken in Nan, Thailand) has no bare numeral classifier construction.

*******************************************************

EASTERN KAYAH LI [RED KAREN] & TAI (David Solnit)

I can tell you a bit about Eastern Kayah Li (Red Karen), which I have write a grammar of (presently seeking a publisher). I'm pretty sure I know what you mean by bare clf constructions, and I think Kayah Li doesn't really have them. We may need to refine the terminology and definitions before finally deciding though. Briefly, Kayah Li clf's, which I consider a special type of Noun, include a fairly large subclass that I call Configurational, since it includes most of the clf's having to do with shape. Configurational clf's, besides functioning in the typical |N Num/Clf| frame (numerals and classifiers appear in both orders with respect to each other, depending on the lexical items filling the slots), also are Bound Nouns appearing as heads of compound Nouns. E.g. (impressionistic transcription) bo is clf for long flexible things like snakes, rivers, ropes; it also is the head of compound N's tare bo 'candle' (tare 'wax'), kheh bo 'ladder' (kheh perhaps 'leg'). Many compounds formed with Conf.Clf's take the (same) clf for counting, but some take a different clf. I take it that you would consider these to be clfs that can also be nouns, hence not examples of bare clf constructions.

My impression is that Standard Thai doesn't have bare clf constructions, or not extensively, but that many (maybe most) other Tai languages do. They may not always be singular and definite, though. For instance Lungming, a Central Tai language, has cang4 'clf for doors or stories of a building', also appearing in cang4 lyyn4 'an apartment, flat' (lyyn4 'house', /y/ here is the high back unround vowel). On the other had there is kaa5 'a clf of very vague meaning, sometimes roughly equivalent to English "the, that, those"; frequently its function seems to be to mark the following word as a noun)'. This Lungming material is from

Wm. J. Gedney, The Tai dialect of Lunging: glossary, texts and translations, T. J. Hudak (ed.), Michigan Papers on South and Southeast Asia *39

I believe there are many other Tai languages with similar examples.

*******************************************************

TZELTAL & WESTERN AMAZONIAN (Mike Maxwell)

As for classifiers, I'm familiar (more or less) with two cases. One is Tzeltal (a Mayan language), which has numeral classifiers: the classifiers go on the numerals. I think that is the "standard" sort of classifier, what people mean when they talk about classifiers.

The other case occurs in numerous unrelated languages of western Amazonia (and perhaps elsewhere in Amazonia, for all I know). I've looked at languages of the Tucanoan and Boran families, as well as Waorani (Auca), the latter a language of eastern Ecuador with no known relatives. Classifiers function remarkably the same in all of these languages, the only major variation being how many there are. But I would not call them *numeral* classifiers, for they have nothing particular to do with numerals. Why call them classifiers, then? Because they are affixes (for the most part) that represent shapes (or other salient characteristics, but generally shape-related). Unlike Mayan numeral classifiers, these classifiers go on all nouns and adjectives (to the extent that these languages can be said to have adjectives); some of them can also function as nominalizers on verbs. Nearly all inanimate, nonabstract nouns take classifiers, as do all adjective-like words that refer to those nouns. One exception (found in all three language groups I listed above) is plants, which can either take classifiers (in which case they refer to parts of the plants, such as leaves or seeds) or not take classifiers (in which case they refer to the plant in the abstract). There seems to be a gradation between classifiers which are clearly suffixes, and those which are more or less independent nouns but enter into a compound noun construction when they serve as classifiers. Finally, there may not be a clear distinction between classifiers, which go on inanimate nouns, and gender markers, which (in these languages) go on animate nouns. Putting it differently, classifiers in these languages serve as an extended gender system. What distinguishes the extended gender systems in these languages from so-called gender systems in languages like Swahili (I don't really know anything about Swahili, so I could be wrong) is that in Swahili, "gender" classes are largely arbitrary, whereas in these Amazonian languages the classes are semantically based. (The semantic basis can sometimes be a bit obscure, for instance when words for "canoe" and "machete" take the same classifier--apparently referring to the way both curve to a point.)

Well, I've rambled enough, and I don't even know if this is what you're looking for. I think you were interested in the syntax: in these Amazonian languages, NPs composed of more than a single word (apart from compounding) are quite rare. Usually if you want to modify a nominal concept with an adjective-like word, you just use the adjective + classifier and leave out the noun, since it's redundant given the classifier. But occasionally you hear modifier + noun constructions, and so far as I know you always get the classifier on both words.

I think the reason both Mayan-type classifiers and numeral-type classifiers are called "classifiers" is that both are affixes that denote shapes, and both go on modifiers (although the Amazonian ones go on nouns as well). I would say that the Mayan ones are simply more restricted morphologically, although one could argue that there's more to it than that.

As for whether numeral classifiers go on anything other than numerals in Mayan languages, to my knowledge the answer is "no". But I'm not really qualified to say--there are lots of "real" Mayanists out there; I hope some of them will answer!

THE THIRD QUERY:

In "numeral-classifier languages" such as Vietnamese and Mandarin, it is often claimed that the use of a numeral classifier is "obligatory" when a noun occurs in construction with a numeral. I am interested in knowing whether -- contrary to the above claim -- there may be certain contexts in which the classifier is in fact optional.

One likely context is that of restaurants and cafes. In Vietnamese, when a waiter takes an order from a table and shouts it back to the kitchen, s/he will typically omit the classifier, eg. (diacritics omitted) [hai ga ba bo] "two chicken three beef" (at a noodle stall), or [hai ca-phe den mot nuoc cam] "two coffee black one water orange" (at a drinks stall). In contrast, in (the Singaporean dialect of) Mandarin, in similar contexts, it is my impression that the classifier is usually or always present, though, quite often, the numeral-plus-classifier expression will occur after the noun, rather than in its "canonical" position before it, eg. [kafei yi bei] "coffee one cup".

My query is thus addressed to speakers of "numeral-classifier languages", or linguists living in communities of "numeral- classifier language" speakers. I would therefore like to hear from speakers of or persons familiar with Japanese, Korean, the various Chinese languages spoken in the PRC, ROC and Hong Kong, Khmer, Thai, Burmese, Nepali, or any other "numeral-classifier language".

Specifically, I would like to know whether, in your language, there are contexts (such as restaurants and cafes) in which the numeral classifier can be omitted, as would appear to be the case in Vietnamese, or whether the numeral classifier must indeed always be present, as is perhaps the case in Singaporean Mandarin.

I would also appreciate any theoretical comments on this issue, and speculation on what factors might underlie the apparent difference between Vietnamese and Mandarin (eg. maybe the numeral-plus-classifier constituent constitutes two words in Vietnamese but a single word in Mandarin).

*******************************************************

SUMMARY OF RESPONSES:

The main result to emerge from the responses is that numeral classifiers may be omitted in certain specific contexts, as in Vietnamese, in two additional languages -- Thai and Japanese.

The Thai data would appear to be the most robust. With regard to Japanese, some of the putative examples of numerals without classifiers involve the "native Japanese" numbers (hitotsu, futatsu, etc.), which actually do contain a classifier "-tsu". However, the responses do contain additional unquestionable examples of numerals without classifiers.

Another possible example of numerals without classifiers -- though apparently more restricted -- comes from the Taibei dialect of Mandarin. However, a purported example from the Singaporean dialect of Hokkien contains no classifier but also no numeral, and so does not satisfy the criteria put forward in the query.

>From the responses that came in, the only example of a language in which the classifier can never be omitted is Beijing Mandarin -- which would be consistent with my own impressions vis a vis the Singaporean dialect of Mandarin.

Thus, the facts would seem to belie the parenthetical speculation at the end of the query, to the effect that wordhood is the relevant factor: in Japanese, the numeral-plus-classifier constituent would appear, if anything, to be more closely bound than in Mandarin; nevertheless, the classifier is omissible in Japanese but not in Mandarin.

Finally, a couple of responses pointed to the omission of the classifier in similar contexts in English, as in "four coffee". One respondent wonders "what general conclusions can one draw from all this?" None so far. But given the presumably equal functional/communicative factors at play in all languages, it would seem to me as though the observed cross-linguistic variation in classifier omissibility could only reflect some (as yet hidden) structural property distinguishing between numeral classifier constructions in different languages.

JAPANESE

******************** (Emily Bender)

I am an exchange student from UC Berkeley living in Japan for one year. I saw your query on the LINGUIST list, and brought it in to my colleagues in the linguistics department at Tohoku University.

As you probably already know, Japanese has a very developed system of numeral classifiers, which is said to be starting to simplify itself (for example, many people don't know that rabbits are "supposed" to be counted like birds). Basically, there are Sino-Japanese numbers (ichi, ni, san...) and native Japanese numbers (hitotsu, futatsu, mittsu...) but the Japanese numbers only go up to 10. Most things are counted by suffixing the classifier to the Sino-Japanese numeral (or rarely to the Japanese number). There are two default counting systems, to be used when the appropriate classifier is not known, or, because it is rare, it would be strange (showing-off?) to use it. These are a) the Japanese numbers and b) Sino-Japanese numbers plus the classifier 'ko'. Everyday objects with well-known classifiers can not usually be counted in this way (books, pens, cups of coffee, etc.)

However, as you mention in your query, there are exceptions. When ordering coffee (or anything else in a restaurant) one is *required* to use the Japanese numerals. Thus, "koohii futatsu" would be "two coffees please" while "koohii nihai" ('hai/bai/pai' is the classifier for drinks served in low cups) would imply that the same person was going to drink both cups of coffee.

When I first asked the question, to make sure that the classifier is generally required, I used books for an example. Books are usually counted with 'satsu'. If one were to use 'hitotsu, futatsu' it could only be a context where one was referring to the content of the book ("it took him 10 years to write that one book") and not actual objects. If one were to use 'hitotsu, futatsu' for the actual objects, it would "sound like something a three-year-old would say." 'satsu' can be used in either context, and is unmarked in either case.

One friend commented that pizzas are ordered with 'hitotsu, futatsu' when one is in the restaurant, but are more likely to be counted with the classifier ('mai' in this case, the classifier for thin, flat things) when ordered over the phone. Other participants in the conversation promptly disagreed.

Tickets into museums, movie theaters, etc., can be ordered with 'mai' (see above) or 'ri/nin', the classifier for people. For example, 'gakusei ichimai' "one student (ticket)" or 'gakusei hitori' "one student." It was noted that the classifier may be dropped for numbers over ten, so 'gakusei juugo' "fifteen students" sounds fairly natural. (I didn't ask at the time, but from my own knowledge of the language, I would add that in other cases where one is counting people, the classifier is needed no matter how big the number is. For example, 'kono machi no jinkoo wa gosen-nin desu' "This town's population is 5,000.")

Another friend said that software is counted with 'hon/bon/pon' (this classifier originally meant long and skinny object, and is used for pens and pencils, beers, cassette or video tapes, and train lines, among other things). Everyone agreed that one would certainly use 'hon/bon/pon' when requesting a piece of software in a store. Then the person who brought the example up went on to say something else about a particular piece of software that he had two of, and said so using "futatsu" (ie, without the classifier). He was aware of it after he said it, and brought everyone's attention to the fact.

So, as usual, the situation is less than clear-cut, and introspection would seem to turn up different results than observation. My impression was that the differences between "koohii hitotsu" (one serving of coffee) and "koohii ippai" (one cup of coffee) and "hon hitotsu" (the content of one book) and "hon issatsu" (one book) had to do with abstractness. The classifiers seem to make the phrase refer to an object while the classifier- less phrases seem to allow for a more abstract interpretation: ie the unit of 'usefulness' of the object.

******************** (Osamu Fujimura)

In response to your Linguist query, I am a native speaker of Japanese.

In Japanese, the child language often does not specify the classifier, but uses the indigenous Japanese form (Yamato kotoba) like hitotu, hutatu, etc. Also, when the number is high, the classifier is often skipped. The restaurant situation is just as you say; in this context, it does not sound childish. Incidentally, in Japanese, the numeral+classifier as well as other quantifiers come after the noun and the postposition (or in formal or technical contexts between them). There are syntactic discussions available, but I don't have them handy now.

******************** (Mayumi Masuko)

In Japanese also, it seems perfectly fine to omit classifiers in restaurant/cafe contexts. I personally think it would be more acceptable if the number is in English; as you may know, there are lots of loan words from English in Japanese and simple numerals like one, two, three (with necessary phonological/phonetic adjustments) are commonly used: 'coffee one, tea two'. In case of the Japanese numerals, I think it more likely that the general classifier would be attached in this context. However, in other shop/ordering contexts, omission of classifiers seem OK. For instance, in a bookshop, when they sort out ordering, the shop clerks may simply say "_Intro to Sociology_ 2, _Advanced Sociology_ 3" etc.

*******************************************************

MANDARIN [BEIJING]

******************** (Wally Frick)

this is a fascinating topic you got there. Classifiers in Modern Standard Chinese (MSC = Mandarin) are always used when numbers are involved. As you cited "yi bei cafei" or if a waiter orders, obviously postpositioned as would also occur quite frequently in the Beijing dialect of Mandarin, thus "liang bei cafei" or if ordering in a restaurant: "cafei liang bei" is the same. The MSC can almost always substitute the standard classifier "ge" for other classifiers. Thus when going on the tube or subway in Beijing only foreigners shout to the ticketseller behind the glass window " liang zhang " whilst most local Chinese call out "liang ge" leaving out the "piao" for ticket. To simply say liang without the classifier is not possible in this context, although I have heard people say " liang ren " vice grammatically correct "liang ge ren". It is the same with "mei ren" vice " mei ge ren or mei yi ge ren". The subtle difference here certainly exists in the translation: everyone, everybody, every single person. Classifiers are also omitted in simple noun phrases such as: Zhe shi shu (this is book, meaning, this is a book). Question: Zhe shi shu ma? Answer: Bu shi, zhe shi huabao. But when buying dictionaries, e.g. one for myself and one for my friend, I would simply say to the shop assistant : " mai liang ben " (buy two + classifier). Here the classifier cannot be omitted as it is closely linked to the object. And as objects at a stationery counter in a Chinese department store are placed closely together and often in no seeming order, the classifier will make all the difference whether I am handed two pens or two books or two rulers, etc. There are a number of good books out on Chinese classifiers and their use. Should you need their titles, please send me an email message and I'll dig them out of my library for you.

1. Federal Publications (S) Pte Ltd. (1988)."Times Dictionary of Chinese Classifiers". F.P., Singapore. ISBN981 01 3900 4 2. Sorry, oops, the second one refers to radicals rather than to measure words of classifiers. There is a good selection of classifiers listed against the appropriate nouns in Wu Zhaoyi et al. (1988) "The English Chinese Pinyin Pocket Dictionary". New World Press Beijing. ISBN 7-80003-044-0.

Apart from that it is difficult to find much literature on Chinese Classifiers, although you'll find some honours, MA and PhD theses that occasionally feature chapters on Chinese measure words. That exhausts my present advice.

******************** (Xiaobo Ren)

In Mandarin, theoretically, numeral classifiers cannot be omitted when nouns are quantified by a numeral. Even though the noun is topicalized or omitted, his numeral and classifier are still there. In your example [kafei yi bei], kafei is topicalized. Numeral classifiers can only be omitted in expressions (chengyu) or in the archaic style. In ancient Chinese, numeral classifiers didn't exist. The numeral can be omitted when it is [yi] = one.

MANDARIN [TAIBEI] (Hao-yang Wang)

In a restaurant in Taipei, once I ordered one niu-rou mian (beef noodle) and one cao-gu tang (mushroom soup). The waiter wrote down on the bill as "1 niu" and "1 cao" and then shouted back to the kitchen exactly the same things as he had just written. Here the waiter omitted the numeral classifiers, but he also reduced "cao-gu tang" into "cao". I think this omission is just a part of their jargon, and not to be found in the daily usage.

When the number is used as a part of a proper name, the numeral classifier may be omitted, for example, "bai-xue gong-zhu ji qi ai-ren" (the Snow White and the seven dwarfs). In the Southern Min dialects of Chinese, where one character usually can have two different pronunciations (one "oral", the other "reading"), you can see clearly that whenever a numeral classifier is omitted, the phrase associated with it will be pronounced in the reading form, as an indication of the phrase being a proper name.

> (a) concerning the restaurant example that you cite: > what language/dialect is it in? (By asking this I reveal > my ignorance of Chinese.) Is it the Taiwanese dialect > of Mandarin, the Taiwanese dialect of Min, or what?

the Taiwanese dialect of Mandarin

> (b) In a context where you were sitting at a table with > your friends, and passing on a collective order to the > waiter, would *you* ever omit the classifier, and say > something like two tea one coffee?

No.

*******************************************************

HOKKIEN [SINGAPOREAN] (Ruby Chua)

Being from Singapore, I agree that in Mandarin, the numeral classifier normally follows the noun. However, I have heard numeral classifiers omitted in coffee-shops when there is only one of the item being ordered.

Also, I only remember this occurring in Hokkien, a Chinese dialect. For example, the waiter would generally yell back to the kitchen: [kopi o] -black coffee, or [teh see] - tea with milk. If more than one was ordered, the numeral classifier would generally be added.

*******************************************************

THAI

******************** (Bob Hvitfeldt)

According to The Fundamentals of the Thai Language (5th ed.), Campbell, S. & Shaweevongs ,C.,

"1. the classifier is always placed after the noun to which it refers and after the adjective if there is one 2. a classifier is always used with the singular form of the noun where an individual thing is specifically referred to unless the noun is qualified by a personal pronoun (when the clsf is omitted) 3. a few adjectives and ordinal numbers come after the clsf 4. a different few adjectives and cardinal numbers usually come between the noun and clsf 5. a clsf is NOT used when the noun is qualified by the adjs 'these', 'those', 'many', 'few', or 'how many' 6. in a large number of cases, where there is no suitable classifier, the noun is repeated and becomes its own classifier"

(In reality, spoken Thai omits the classifiers very frequently in, as far as I can tell, almost all environments. Having lived there for several years, I consider myself a "fair' speaker, and once got very interested in these things.

******************** (Peansiri Vongvipanond)

Here is a response to your query in Linguist List.

Pattern for regular quantified NP's in Thai is <N (adj, etc) Num CLS> However, a pattern <Num. N> is found sporadically in these situations: in road-side restaurants by owners and staffs, in newspapers headlines.

Language teachers, being prescriptive of course, frown upon these"deviant" phrases.

******************** (Gwyn Williams)

In colloquial spoken Thai numerals are often omitted in such situations, eg.,

bia iik s>>ng (khuat) beer more two (bottle) "two more beers"

where khuat is the classifier (and word) for "bottle"

I think such omission depends on the formality of the situation.

*******************************************************

INDONESIAN (Roberta L. Mitchell)

I believe Indonesian uses numeral-classifiers in a similar way to what you described in your posting. Since I am not a native speaker or a linguist, I would suggest that you might seek some higher authority on the subject, but I will give you as much as I can. Ordinarily classifiers must be used any time one refers to a certain number of anything, for instance:

Saya beli empat buah jeruk (or jeruk, empat buah) I bought four [fruits] oranges. (or oranges, four fruits.)

Different classifiers are used for all kinds of nouns, depending mainly on shape. One interesting intentional misuse I have heard was:

Ada cewek, dua ekor

[there are] women, two tails (instead of two people, which would be orang.) Some nouns seem to use classifiers most of the time regardless of number, such as: orang manusia, meaning a member of human kind, as opposed to all human-kind. To say "an orange", one would use the prefix se- (which is short for satu, one) before the classifier, thus: sebuah jeruk. or "a person" is: seorang Interestingly, to say "somebody" it is: se-seorang.

I think there may be times when classifiers are not used, for instance I distinctly remember someone saying to me:

saya harus baca buku lima. I have to read book[s] five.

He could have said something like lima buah buku, but I am positive that he could not have said lima buku. Therefore I think the use of classifiers might have something to do with word order.

*******************************************************

NEWARI (Peri Bhaskararao)

Your question (on Linguist list) about optionality of classifier usage in certain contexts is interesting. I worked on the classifiers of Newari (with Sunder Krishna Joshi, currently professor of Newari in Kathmandu, Nepal). There are certain cases where classifiers are optional in this language. I am mailing you an offprint of our paper, 'A Study of Newari classifiers' that was published in 1985. You might not have access to this publication. One more interesting point about Newari is that they have verbal classifiers!. This is also explained in the paper.

*******************************************************

ENGLISH

******************** (John Cowan)

This is not directly relevant, but I have heard waiters, speaking American English natively, omit the "mandatory" plural marker in the same situation. "Two double beef, three cheese, four coffee", to make up an example. Perhaps there is a tendency to pidginize one's own language in such situations, for the purposes of clarity in a noisy environment?

******************** (Stavros Macrakis)

Isn't restaurant slang notoriously telegraphic and idiosyncratic everywhere? Not just vocabulary, but also syntax, e.g. "double over easy on a brick" = "an order of two fried eggs flipped on the grill and served on toast" or for that matter "two halibuts with fries" = "two _portions of_ halibut with fried potatoes". Note that this case actually parallels the numeral-classifier case you give, since in standard English, two halibuts means two entire fish, and a classifier is obligatory if you want to talk about any other units, e.g. pounds of halibut, filets of halibut, portions of halibut, orders of cooked halibut, boxes (the wholesale unit) of halibut, etc.

An example I recall from a Turkish textbook is "shish kebab" = "sword meat" which according to normal rules for attributive nouns should be "shish kebabi" = "sword meat-of-it".

So what general conclusions can one draw from all this?