The Language of the Modhupur Mandi (Garo), Volume 1

Introduction

(A)

The Garo words that I will call "numerals" consist of two parts: first a "classifier" (or "numeral classifier") that shows the kind of units that are being counted, and second the number. In phrases such as three heads of lettuce and two sheets of paper English has something very much like classifiers ( heads, sheets ), but Garo differs from English in requiring a classifier with virtually every number. Numbers occur as one part of a numeral, rarely alone.

In a sense, classifiers and numbers, define each other. Anything that is used right after a classifier can be considered to be a number, and anything that is used immediately before a number can be considered a classifier. When animals are counted, for example, mang- , must always be included in the phrase: meng-gong mang-gin-i 'two cats', mat-ma mang-git-tam 'three buffaloes'. When people are counted, sak- or its Mandi variant ak- is needed: man-de sak-sa 'one person', me'-chik sak-bri 'four women', chau'-kok ak-gin-i 'two thieves'. Long thin things such as string are classified with ding- : do-ri ding-sa 'one piece of string'. Whatever is counted must have a classifier so the classifiers give an implicit classification to the countable phenomena of the world—-people, animals, abstract phenomena, and objects of various sizes, shapes and uses. Since it is probably impossible to construct a fully logical classification system that encompasses every countable thing, it may be inevitable that languages with a rich classifier systems such as Garo need a residual category classifier that can be used when nothing more specific quite fits. In A'chik that classifier is usually ge- ; in Mandi it is more often kan- : nok ge-gin-i or nok kan-gin-i 'two houses'.

Numerals share a number of features with nouns. In particular, they occur as constituents of noun phrases, and under the right circumstances case markers can be suffixed to them. Nevertheless numerals also differ from ordinary nouns in important ways. Not only does their internal structure differ from that of nouns, but they are unable to take the plural suffixes that nouns can take. Nor do they enter into compounds as nouns so often do. Numerals are different enough from ordinary nouns to deserve to be considered a different part of speech. Each of the two parts from which numerals are formed will be described in this chapter.

Numbers

(A)

The A'chik dialect of Garo has a good supply of numbers. It is possible to count up to 999 with numbers that are purely Garo. The word for 'thousand' and the words for the higher numerical units are borrowed from Bengali but they are well assimilated into the language, and, when used together with the native Garo words for the smaller numbers, they allow A'chik speakers to use numbers with complete freedom. The dialects of Mandi spoken in Bangladesh have been so heavily influenced by Bengali that speakers frequently use borrowed Bengali numbers instead of the older Garo numbers. Schools and markets are the places where numbers are most often needed, and they are places where the Bengali language is dominant. As a result, Bengali numbers become so familiar that they are used by Mandis even when they speak their own language. Most Mandi speakers in Bangladesh do not even know the older Garo numbers above the word for 'five'.

The numbers from 'one' to 'five' are used without classifiers in only two circumstances: First, classifiers can be omitted when counting: sa, gin-i, git-tam , but speakers often include classifier even then: sak-sa, sak-gin-i, sak-git-tam 'one person, two people, three people'. Since Bengali numbers are so often used in Modhupur, counting with Mandi numbers is not common, and counting for very long is impossible, but A'chik speakers count easily with A'chik numbers, with or without classifiers. The second use of numbers without classifiers is as ordinals meaning 'first, second, third' and so forth, or 'first one, second one, third one'. These are constructed by adding gip-a to the cardinal number: sa-gip-a 'first, first one', gin-i-gip-a 'second, second one' etc. These are nouns and they can be used in all nominal constructions.

In view of the limited use of the older Garo numbers, it is surprising that the system of classifiers has survived in such a healthy state. The numbers between 'one' and 'five' do continue to be freely used, even byPage  245speakers who do not know the native Garo word for 'six', and when any number at all is used, it requires a classifier. The numbers in A'chik and in Mandi are given in Table 10—1. They are given in the table without classifiers, but in use they are almost joined to a classifier.

A few older people in Modhupur still remember an alternative counting system that is based on cycles of twenty rather than ten. This, and similar systems, are heard occasionally in the Garo Hills, but they have not found their way into orthographic Garo or into the speech of most educated people, so this way of counting is probably on its way to extinction. Through nineteen, the numbers are the same as in decimal counting, but for higherPage  246numbers they are grouped by 20's instead of by 10's. These numbers are rarely used, and the only ones at all likely to be heard are the even multiples of 20. This dying system is shown in Table 10—2.

In addition to the numbers, a handful of other morphemes can be used with classifiers in the position where a number would otherwise go.

-gip-in 'another, a different', ak-gip-in or sak-gip-in 'someone else', mang-gip-in 'a different animal':

-rik-kit 'every, each'. This can describe events happening jointly to everyone or to events that effect each one separately, but in the same way. It is likely to refer to a larger group of individuals than -prak : ak-rik-kit 'each person', do'-mok mang-rik-kit 'each goat'; a'-ba dam-rik-kit 'each field'.

-prak 'each'. This is similar in meaning to -rik-kit but it is likely to imply a smaller number, and it is more focused on individual acts and events: sal-prak 'each day'; kan-prak-ko ron'-bo 'give one of each. -prak can be reduplicated, emphasizing the distributive sense: sak-prak-prak-na kan-sa-kan-sa ron'-bo 'give one to each person'.

Gip-in, prak and rik-kit can be considered to be numbers when they are used with classifiers. Unlike the numbers from 'one' to 'five', however, these three words are used in other ways as well, so they are not always found with a classifier.

Classifiers

(A)

Classifiers specify both the nature and the size of the unit being counted. This is obvious in the case of measures such as ke-ji- 'KG, kilogram', mail- 'mile', and kap- 'the amount held in a cup', as in cha kap-sa 'one cup of tea'. It not quite so obvious for the classifiers that are used for people, animals and physical objects.

The natural translation for a-chak mang-sa is 'one dog' but on the analogy of cha kap-sa 'one cup of tea', we might translate the phrase more literally as 'one mang of dog'. This is made more plausible by such phrases as ma-su ju-ra-sa 'one team of cattle, one pair of cattle' in which the classifier ju-ra- specifies two animals rather than just one. If ju-ra- is a measure for two animals it is reasonable to consider mang- to be a measure for one animal—-the amount of animal stuff contained in one animal. Similarly, Man-de sak-sa has the literal meaning of 'one sak of humanity' or 'the amount of humanity that constitutes one person'.

One way to understand the role of classifiers is to realize that languages like Garo do not make such a clear distinction between "mass nouns" and "count nouns" as English does. In English, mass nouns such as water, milk, meat, iron, and sand cannot be counted without the help of some measure word: one puddle of water, two pounds of meat, three piles of sand . Puddle, pound and pile specify, with varying degrees of precision, the size of the unit that is to be counted, as well as enough about its nature to know whetherPage  248it can be piled or weighed. Count nouns, on the other hand, require no classifier in English, but can be counted directly: one person, five animals, three stars, four songs . In Garo all nouns act like English mass nouns, and all require some unit of measurement whenever their amounts are counted.

Classifiers are not simply assigned to nouns in an arbitrary way. Rather, classifiers convey important meanings. It is often the case that a single noun can be used with more than one classifier, and each classifier contributes a different meaning: te'-rik rong-sa 'one banana, one rong- of banana', te'-rik pang-sa 'one banana tree', te'-rik gal-sa 'one hand (small bunch) of bananas', te'-rik ol-sa 'one arm (large bunch) of bananas', te'-rik ke-ji-sa 'one kilogram of bananas'. When all these are considered together, it is not unreasonable to understand rong- as specifying the amount of banana stuff held by a single fruit, just as ke-ji- specifies the amount of banana stuff held in a single kilogram.

The meanings that classifiers convey make it easy, when the context is clear, to use a numeral (i.e. the combination of a classifier and number) with no noun at all. In answer to a question about how many bananas are wanted, it would be entirely natural to say ke-ji-gin-i 'two kilograms' or rong-bong-a 'five fruits'. If the particular type of person is not known or is known so well that it need not be repeated, it is entirely appropriate to say sak-gin-i sok-ba-jok 'two people arrived', where a classifier is used, but no noun. Since sak- is used only for people, there is no possibility that this will be misunderstood as meaning two animals. Indeed, man-de sak-gin-i is completely redundant since man-de , the noun meaning 'people', conveys no more information than sak- . Of course, if it is important to specify a particular kind of person one is talking about, it is necessary to use a noun: me'-chik sak-gin-i 'two women'.

I will define the class of numeral classifiers to include all morphemes that can be prefixed to the numbers, and any combination of such a classifier with a number will be considered a "numeral". This is a somewhat broader definition for classifiers than has sometimes been used for other languages, but it is convenient for Garo. Defining classifiers this way does mean that the class includes several somewhat different kinds of morphemes. Most typical of what have generally been called classifiers are those that are used when counting discrete physical objects that have enduring shapes. These are the sorts of things that are named by count nouns in English, and it is the classifiers used for these objects that are most foreign to languages like English. In addition to the classifiers for people and animals, many of the most common and central classifiers in Garo indicate something about the shape of the object counted—-whether it is more or less round, thin and flat, long and thin, and so forth. A few classifiers indicate the materialPage  249from which an object is made, or the use to which it is put. There are also classifiers for nonmaterial but countable things such as songs and ideas, and there is a residual category classifier for things not otherwise provided for.

In addition to the classifiers that indicate a single object, there are a very large number of classifiers that are used for parts of objects and for various kinds of collections: bunches, clumps, slices, chunks, etc. Classifiers that refer to parts and collections may seem less exotic to English speakers than those for single discrete objects, because chunks and clumps need to be specified in English just as they do in Garo. We are used to such 'classifiers' from our own language.

Words that refer to things without consistent shapes or boundaries, (i.e. the words for things that are named by mass nouns in English) require different classifiers, and these will also seem less exotic to English speakers, because even English needs measure words when counting units of sand, water, or meat. The Garo classifiers that specify amounts of unbounded objects like dirt, grain, or sand fall into two types. First, there are containers. Any word for any container can be used as a classifier that specifies the amount held in the container. As my earlier example with kap 'cup' suggests, borrowed words for containers are easily used as classifiers. Second, there are weights and measures: mile, kilogram, yard, and so on. Most of the weights and measures now used in Garo are borrowed from either Bengali or English, but they can be considered classifiers because they are used with numbers. Garo also has some older measure words based on lengths of parts of the human body, and these continue to be used for some purposes.

Finally, there are units of time: 'day, week, year', and so on. These words can be used in Garo without another intervening classifier, so by the definition I am using, they are, themselves, classifiers. To English speakers they will seem the least exotic of all classifiers, but they are really very much like weights and measures. Instead of specifying units of weight, length, area, or volume, they specify units of time. Unlike other weights and measures, all of which can measure a wide variety of things (water, milk, meat, sand etc.), the time classifiers can measure only one thing: time. Uniquely, therefore, time classifiers are never used with a noun. sal-sa 'one day' says all that needs to be said, and no noun for 'time' can be added to it in the way that be'-en 'meat' can be used in be'-en ke-ji-sa 'one kilo of meat'.

Some morphemes can be used as both nouns and classifiers. This is true of all words for containers, all the time words, and a handful of others. Song 'village' is very unusual in being neither a time word or a containerPage  250but still being usable as both a noun and a classifier. This makes possible a phrase such as song song-gin-i 'two villages' in which the classifier echoes the noun. Echo classifiers are common in some languages but this is the only perfect echo that I have identified in Garo. A few other classifiers are obviously related to nouns or to verbs and some allow partial echoes. Tom- is a classifier for round things such as stomachs and -tom appears as a component of a few compounds: a-tom 'stomach' and jak-tom 'fist'. A-tom tom-sa 'one stomach', a partial echo, is possible. Dol'-a is a verb meaning 'wrap up'. Dol'- is a classifier for 'wrapped up bundles'. Many classifiers, including majority of the most common ones, however, are used for nothing else than classifiers and suggest no derivation from any other part of speech.

Each of these types of classifiers will now be described, along with a number of examples.

Core Classifiers (A) . This first group of classifiers include those that are most common and most essential. They include the classifiers for human beings and for animals, and these, together with the more transparent classifiers for containers, time, and weights and measures, are enough to allow even early learners to make themselves understood.

Ak-, sak- 'people', and occasionally 'ghosts, spirits, gods'. A'chik allows only sak- but ak- is at least as common in Modhupur: me'-chik ak-bri 'four women'.

Mang- 'animals'. This is used for all sorts of animals, including mammals, birds, insects etc. I was once treated to a microscopic view of one of my own malarial parasites, and heard it referred to as mang-sa 'animal-one'. Occasionally, mang- is also used for ghosts or spirits. A homophonous syllable occurs in bi-mang 'body' and it must carry a bit of the sense of 'body' into its use as a classifier: do'-bak mang-git-tam 'three bats', gang-gu mang-gin-i 'two mosquitoes'.

Kan- 'residual category'. A language with an elaborate classifier system like that of Garo needs to have one classifier that can be used for a residual category of objects for which no more specialized classifier is readily available. This, at least, is necessary where the choice of classifier depends on meaning rather than on the arbitrary assignment of nouns to classes, as in a gender system. In Modhupur the residual category classifier is kan- . It is used most characteristically for miscellaneous man-made objects, but it may be used for anything that is sufficiently unusual not to have acquired a conventional classifier. Mandis easily advise non-native speakers, "when in doubt, use kan- ", but to use kan- for a person or animal would be bizarre. Kan- can be used for such things as shoes, glasses, hats, ears, houses, doors,Page  251lights, letters, books, baskets etc., though for many of these, it would be more elegant to use a more specialized classifier.

Ge'-, gi'- . The residual category classifier most often used in the A'chik dialect is ge'- and it would be understood anywhere. Like the Mandi kan- , it can be used for a wide variety of constructed artifacts.

Ro-kom- 'kinds, varieties'. Ro-kom- was originally borrowed from Bengali but it has been thoroughly assimilated into Garo. It can be used as both a classifier and a noun: meng-gong ro-kom gin-ing 'two kinds of cats'.

Ming- 'abstract concepts, incorporeal phenomena, diseases, songs,colors, spirits': gan ming-git-tam 'three songs', chan-chi-a-ni ming-sa-han 'only one thought'.

Rang'-, chang- 'times'. Rang'- is more common inMandi while chang- seems to predominate among A'chikspeakers. These are used for the number of times that something hashappened or been done. Do-ri rang'-git-tam bi'-a-ming 'thestring broke three times'. Do-ri ding-git-tam bi'-a-ming , with a different classifier, means 'Three strings broke'.

Shapes, Materials, Places (B) . Classifiers for physical objects, especially constructed artifacts, are very often selected on the basis of the shape of the object. Somewhat less common are those that specify the material from which an object is made or the location where it is found.

Rong- 'round' and, especially, 'globular objects, pots, fruit, eyes, coins': ku-mil-a rong-sa 'one orange', me-dik rong-gin-ing 'two rice pots'.

Ding- 'long thin things: hairs, cords, strings, pieces of cane, wire, roads etc.': do-ri ding-sa 'one piece of string'.

King- 'thin flat things: ears, leaves and, especially, paper and things made from paper, including books': bi-jak king-gin-i 'two leaves'. king- is not used for cloth or for objects made from cloth.

Kol- 'holes: windows, doorways, holes in the ground or in a tree'.

Ku'- ' mouths, words, bits of language' (which come from the mouth), 'mouthsful'; also for 'solid somewhat flattish things: doors (as opposed to doorways), gongs, single split sticks of firewood': a'-bol ku'-gin-ing 'two sticks of firewood.

Sam- 'any bilateral body part, hands, eyes, etc.': jak sam-gin-i 'two hands', mik-on sam-sa gan-ing-a 'blind in one eyes'.

Dam- 'places and things in fixed places, villages, fields, houses': ha'-ba dam-sa 'one field', dam-git-tam mat-a 'wounded in three places', ku'-gi-sep dam-gin-ing 'two corners of the mouth'.

Song- 'villages'. Song is the ordinary Mandi noun for 'village' and this allows the atypical reduplication of song song-gin-i 'two villages' in which the classifier echoes the noun.

Dot- 'living or once living objects, most often those that are vertical and round in cross section: posts, lengths of bamboo, stalks, unsplit pieces of firewood': a'-bol dot-sa 'one unsplit piece of firewood'.

Bol- 'axes, bamboo knives ( at-te ), drums'. As a noun, bol means 'tree, wood' and as a classifier it is used for things that are long and wooden. Axes and knives have wooden handles and drums are made from a length of trunk.

Pat- 'things made of cloth, shirts, saris, quilts', also 'units of paper money, taka' (the Bangladeshi unit of currency).

Pieces, Parts, Groups, Bundles, Loads (C) . Mandi has a very large number of classifiers that denote parts or groups of objects. Many of these are highly specific, referring to a particular kind of piece, or a particular kind of bundle. A few typical examples are given here:

Chom'-bak- 'bunches, such as bunches of onion plants or beans'.

Gal-wang- 'small bunches of bananas, 8 or so fruit'.

Pek- 'packs of cigarettes', an obvious borrowing.

Sreng- 'single bunches of thatch'.

Chong- 'banana clumps that typically have several diverging stalks'.

Gip-ak- 'loads, as firewood'.

Ma'- 'groups of chicks with a mother hen, a clutch'.

Gil- 'crosswise slices, such as loops of pineapple etc.'

Sri- 'lengthwise slices, of papaya, pineapple etc.'

Tong'- 'lengths, for pieces cut crosswise, of bamboo, wood, firewood, slices of a banana, the parts of a broken pencil'.

Pak- 'halves, half an areca nut, half a fruit, half of a piece of bamboo split lengthwise, one side of a piece of paper or of a coin: sal pak-sa 'half a day'. Pak- can sometimes mean 'the other side' as in ra-ma pak-sa-cha 'the other side of the road'.

Containers (B) . The name of any container can be used as a classifier for the amount of material that it can hold. Since many new kinds of containers have found their way to the Mandis in the last century and brought their names with them, this is an area where borrowed classifiersPage  253are common: cha kap-sa 'one cup of tea', chi gil-es-sa 'one glass of water'. It is impossible to list all the objects that might be used as containers, since almost any physical object might be pressed into service to hold something, and any of their names could become a classifier. The name of a container is used to count amounts held in the container. When containers themselves need to be counted, some other classifier must be used: gil-es kan-sa 'one glass'. kap kan-sa 'one cup'. Even nok- 'house' can be used as a classifier, as in nok-git-tam 'three housesful, three families'. When it is physical houses that need counting, kan- the residual category classifier can be used: nok kan-git-tam 'three houses (buildings)'.

Weights and Measures (C) . Marketing has become so important that Mandis have fully absorbed modern units of weights and measures into their language. These measures always come with their English or Bengali names, and they continue to be adopted today. When I first knew the Garos, in the 1950's, the kilogram was unknown, but with the adoption of the metric system in both Bangladesh and India, the ke-ji- has become an important measure and a common classifier. Older Mandi words that come closest in meaning to the modern weights and measures all seem to be based upon parts of the human body. They continue to be used in situations where modern industrial and market precision is unnecessary.

Kru- 'finger span'. A unit of length measured from end of the end of the thumb to end of middle finger or little finger, when they are stretched as far apart as possible.

Mik- 'cubit'. A unit of length measured from the elbow to the tip of the extended middle finger, and counted as half a yard.

Mik-tom- . The length from the elbow to the end of the clenched fist.

Cho'-om- 'handful'. The amount that can be held in an open hand.

Man-de- . As a noun, of course, this is the word for 'person, human being'. As a measure, it is the height of a human being. It is used most often for heights, such as the height of a tree, rather than for lengths. Man-de-git-tam 'height equal to three times the height of a person', man-de-sa ong'-jok 'has grown to the height of a person', which might be said of a young tree.

Among the many weights and measures now used by Mandis and other Garos are the following:

Ke-ji 'kilogram'.

Ser, sel 'seer', an old standard Indian unit of weight equal to a bit more than two pounds. It is close enough to a kilogram to be rapidly retreating in favor of the more modern ke-ji .

Mon 'maund'. A old standard Indian unit of weight equal to 40 seers or about 82 pounds. It is still known, but probably dropping from use in both countries as the metric system becomes predominant.

Iin-chi- 'inch'.

Put- 'foot, 12 inches'. My impression is that iin-chi and put survive a bit better than ser and mon .

Time Words (B) . Since words for time units are followed directly by numbers they are, by the definition I have adopted, "classifiers". It may seem odd to call units of time "classifiers", but they really are measures for time, just as kap and ke-ji are measures of volume and weight. Time words do differ from other classifiers in at least two respects, however. First, alone of all classifiers, they are never used with a noun. No noun for 'time' forms a phrase with a numeral, in the way that the noun meaning 'water' for example, can form a phrase with measure of volume. Garos say the equivalent of 'cloth one-yard' but they do not say the equivalent of 'time one-hour'. This is reasonable. Nothing except time can be divided into time units, so no information would be added if it were explicitly named. A measure of weight, such as a ke-ji could measure rice, meat, milk, or any number of other substances. A measure such as sal 'day' can measure only time.

The second way in which time classifiers are distinguished from most other classifiers (though not from containers) is that the same morphemes can also be used as nouns. This, too, follows from the nature of the meanings conveyed. Speakers often need to specify a number of days or years, and that requires a classifier, but speakers must also talk about 'this year' or 'next year' or an unusually 'hot year', where no number is called for. This requires time words to be usable as nouns, as well as classifiers. ke-ji rarely needs to be used as a noun and sak 'classifier for people', never does.

Sal- As a noun, sal- can mean either 'sun' or 'day'. It frequently takes a locative case marker to become sal-o 'during the day'. Sal-bri 'four days', da'-sal 'today', sal-a-rik-kit 'every day'.

Ja- 'month' is related to ja-jong , a noun meaning 'moon', but only the single syllable is used as a classifier.

In addition to ja- and -sal Mandis use wal 'night' with numbers so that it can be used as a classifier, as well as a verb or a noun: Wal-bri 'four nights' (classifier), wal-jok 'it has become night' (verb), wal-o cha'-no-a 'will eat at night' (noun). In spite of being derived from a noun meaning 'sun' the classifier sal- can measure 24 hour periods, as well as 'day time', just as the word day can in English. Wal- , on the other hand, refers only to the dark part of the day. Wal measures the intermittent and boundedPage  255times of darkness rather than a continuous flow of time in the way that sal- often does. I have been able to persuade a speaker to accept an'-tam-gin-i 'two evenings', in which an'-tam is used as a classifier, though it is certainly unusual, but * wal-ni-gin-i 'two mornings' seemed to be beyond the pale and was decisively rejected.

All the other time classifiers may be borrowed. Even bil-si- 'year' may ultimately be derived from Bengali bo-chor . Except for modest phonological adaptations, gon-ta 'hour' and sop-ta 'week' are straight from Bengali, while mi-niit , of course, is from English. In A'chik the word for 'week' is an'-ti rather than the borrowed sop-ta that is used by Mandis. An'-ti otherwise means 'market', in A'chik, and since markets in the Garo Hills usually convene every seven days, it is an appropriate word for 'week' as well.

Bengali and English Classifiers and Numbers (C) . Bengali is unusual among Indo-European languages, and even among Indic languages, in having a system of numeral classifiers, although by comparison with Mandi its system is restricted to the point of impoverishment. Bengali has only two common morphemes that are unambiguously classifiers: -jon for 'people' and -ta for almost any other physical object. In addition to these, it uses words for weights, measures, time, and containers in ways that are reminiscent of Mandi classifiers, or for that matter, English measures and containers. Mandis regularly use numerals constructed from a Bengali number and one of the two principal Bengali classifiers, and they even embed these numerals in sentences that are otherwise Mandi. Indeed, since so few Mandis know any numbers above 'five' except those from Bengali, they have no choice but to use Bengali numbers, and when they do so, they use Bengali classifiers as well.

Although Bengali numerals, including their classifiers, are widely used in Mandi (much less so in A'chik), the order of the morphemes is different in the two languages. Classifiers always come before the number in Mandi, but they always follow the number in Bengali. The difference in word order probably inhibits the borrowing of either a number or a classifier by itself. Both need to be borrowed together so that their relative order can be preserved. Garos do not borrow a Bengali number to use with a Mandi classifier, or a Bengali classifier to use with a Mandi number. Nevertheless, the numerals as a whole are borrowed, so that one can hear both riik-sa kan-gin-ing and dui-ta riik-sa 'two rickshaws'. With one apparent (but not genuine) exception, classifier and number come together or not at all.

The apparent exception stems from the fact that a good many words for containers, weights, measures, and units of time have been borrowed into Mandi, some from Bengali, some from English, and some from EnglishPage  256via Bengali. This means that many of these words exist in both Bengali and Mandi. Such words may not be borrowed as classifiers, but once borrowed, they can easily be used as classifiers with either Bengali or Mandi numbers. They can be used in the order that is appropriate for the numerals of either language. Mandis can say either chi gles-bri or char-gles chi 'four glasses of water'. Ro-kom-sa man-de or ek-ro-kom man-de 'one kind of person' would both do the job. To say 'six bananas' they would have to say choi-ta te'-rik because they have no word for 'six' except choi , which is borrowed from Bengali. It must have the Bengali classifier -ta and the numeral always comes before the noun as it does in Bengali.

While borrowing is much more widespread in Mandi than in A'chik, borrowing is not negligible even in the hills. Even in the 1950's, people in the most remote villages occasionally used clock time, and when they did so they always used Bengali numbers: tiin-ba-ji 'three o'clock'. -ba-ji is a Bengali measure of time that is used for clock time after numbers such as tiin- 'three'. Mandis always refer to school classes with English numbers. Schools in Bangladesh use Bengali as the medium of instruction but children attend kles wan, kles tu, kles tri, kles por, and kles paip . A'chik speakers, like Mandi speakers, always use the Bengali word order for borrowed numbers and classifiers. Telephones have come to the Garo Hills since the fifties, more recently than clocks and clock times, but today, A'chik speakers who use the telephone always use use English when giving telephone numbers, but they use Bengali numbers for clock time, and their own Mandi numbers for most other purposes.

Numeral Suffixes and Reduplication

(C)

Numerals are used with fewer suffixes than either verbs or nouns, but numerals are not always completely bereft of suffixes. First, numerals can take the full complement of case markers, but since case markers are really suffixed to noun phrases rather than to nouns, the numerals simply pick up the case markers in their capacity as constituents of noun phrases. The same can be said for the four final noun suffixes, -ba 'also', -de 'but', -ha or -sa 'only', and -in 'emphatic'. These are also clitics and they follow case markers. The plural markers, which form a distinctive set of noun suffixes, cannot be suffixed to numerals. As was pointed out early in this chapter, -gip-a can be suffixed to numbers to form ordinals ( gin-i-gip-a 'second' etc.), but -gip-a cannot be suffixed to full numerals (i.e. it cannot be used in the same word with a classifier).

This leaves only one real candidate for a distinctive numeral suffix. -sru 'each' can follow numerals, but it is not entirely clear whether it shouldPage  257be considered a suffix or a separate word. It is rarely used except after numerals, however, and this suggests that it should be considered a suffix: tang-ka pat-sa-sru 'one taka (the Bangladeshi unit of currency) each', bol king-gin-ing-sru 'two books each', Mi rang-tal-sa-sru ron'-bo 'give one plate of rice to each'.

Unlike nouns and verbs, numerals do not become paired with echo words, but they are subject to partial reduplication, and this also gives a distributive sense. Reduplication can be accomplished in three ways, without it seems, much difference in meaning. First, the entire numeral can be reduplicated:

Second, and perhaps most commonly, the number alone can be reduplicated: Sak-gin-ing-gin-ing 'two each'.

Finally, reduplication can be limited to the first syllable of the number. It can even be limited to the initial consonant and the vowel of the number, a process that has more in common with the phonological anticipations that are considered in the next chapter than with most reduplication. Such partial syllable reduplication occasionally puts a high back unrounded vowel (symbolized here with [ɨ]) in an open syllable, an exceedingly strange place for that vowel in Garo.

[sak-gɨ-gin-ing dak-e] 'doing it in pairs of people, two by two'.

Two numerals can be used one right after the other with the implicit meaning of 'or': Mandi sak-bri sak-git-tam 'one or two people'. It is startling to an English speaker to hear the higher number occasionally placed first: kan-bong-a kan-bri 'five or four things'.

Conclusions

(C)

The examples given in this chapter are only a sample of the rich variety of classifiers found in Mandi and in other dialects of Garo. Many additional examples will be found in the lexicon in Volume II. It should be clear that the class of classifiers, as defined here, is an open one in the sense that when new needs arise, new classifiers can be added to the language. BorrowingsPage  258are common, and any speaker can create a classifier from a word for a container.

Some of the morphemes that are used as classifiers are used in no other way. Sak- (or ak- ), the classifier for people, for example, has no use except as a classifier. Other morphemes that are used as core classifiers and as classifiers for shape are occasionally found as constituents of noun compounds, but hardly ever as independent words. Mang- , the classifier for animals has no independent existence in the present language, but the morpheme does appear as part the compound bi-mang 'body'. Rong- 'classifier for globular objects' is also found in bi-rong 'pit of a fruit'. Ding- 'classifier for long thing things' is found also as a part of the compound kil-ding 'thread'. Such double use of morphemes, as both classifiers and as constituents of noun compounds is fairly common, but quite irregular and by no means productive.

Mandi does have one construction, in addition to numerals, in which a large number of classifiers participate. King-dal'-a 'big (of flat things)' and king-chon-a 'small (of flat things)' are constructed from king- , the classifier for 'flat things', and either dal'-a 'big' or chon-a 'small'. A considerable number of other classifiers for shape can be used like king- and coupled with dal'-a or chon-a to describe the size of something that has the appropriate shape. Similarly, the derivational suffix -ma 'big' can be suffixed to a number of classifiers to form words meaning 'big' of something that could be counted with that classifier: king-ma 'big of flat things'. This is a rather eccentric construction and it is not fully productive, being used only with the classifiers for shape, not others.

Classifier systems have sometimes been compared to gender systems, as if each classifier defines a gender to which all nouns that are used with that classifier belong. Seen in this way, a language like Garo would have a very large number of genders. Where familiar European languages may have two or three genders, scores would be needed for Garo. If classifiers define genders, it might seem reasonable to include a note with each noun in the dictionary that specifies just which classifier or classifiers that noun can be used with. Perhaps this would be reasonable for some classifier languages, but it would be impossibly complex for Garo, where a single noun can often be used with several classifiers. Moreover, numerals are often used with no noun at all, so there would be no noun to "govern" which classifier to choose. Hearing sak-gin-i i'-ba-ing-a 'two sak are coming' it is perfectly clear that it describes people of some sort, but no antecedent noun is needed to force the choice of sak- . Rather, the sak- is chosen because of its meaning, just as any ordinary noun is chosen because of its meaning.

While Garo does not make as sharp a distinction between count nounsPage  259and mass nouns as English, even in Garo, different classifiers tend to be used when counting the kinds of bounded and shaped objects that are named by count nouns in English than when counting the kinds of unbounded materials for which English uses mass nouns. Sak 'people', mang- 'animals', kan- 'residual category', and the classifiers that indicate shapes are almost always used with bounded objects that would be named by count nouns in English. Measures and containers are more often used for counting things that would be named by mass nouns in English.

Garo dialects differ in the particular classifiers that they use, but they differ little in the kinds of distinctions that classifiers make. Sak is used in A'chik, ak- is often substituted in Mandi; ge- is more common in A'chik, kan- in Mandi. Still, these dialectal differences are not sharp, and most classifiers would be understood by Garo speakers anywhere.