Robbins Burling

University of Michigan

October 1999

The language that is known to everyone except its own speakers as "Garo" is spoken by about 700,000 people in northeastern India and in Bangladesh. Most of these Garos live in a hilly district in the western part of the Indian state of Meghalaya, but about 100,000 live across the border in Bangladesh, most of them just south of the Garo Hills. Smaller settlements are found in several locations in Assam, in the Khasi Hills of Meghalaya, in Tripura state, and near Modhupur in Bangladesh. Most of these people prefer to call themselves "A'chik" or "Mande" but neither of these terms has gained general acceptance, and until one does, I have little choice but to call them "Garo". Two small enclaves, one with people known as "A'tong", the other with "Ruga", are found within the Garo Hills. Each of these groups has its own language but they consider themselves to be Garos and are accepted as such by all other Garos. (See the article on Northeast Indian Languages, this volume). Except for A'tong and Ruga all dialects spoken by Garos are mutually intelligible, although speakers who are unfamiliar with a dialect that is spoken far from their own home may need some patience and an occasional explanation in order to understand.

A written form of Garo was developed by American Baptist missionaries during the last decades of the 19th century. The missionaries based the orthography upon the dialect of the northeastern corner of the Garo Hills. This was the area with the first substantial number of educated and literate Garos, and their dialect has influenced the speech of educated Garos everywhere. The northeastern dialect on which the written language is based is sometimes called "A'we". The dialect that covers the western part of the Garo Hills and that is spoken in Bangladesh is known as "A'beng" or "Am'beng". Matchi, Chisak and Dual are found in smaller areas in the central and southern part of the district, but all of these dialects grade into one another without sharp breaks. Garos compare their dialects with curiosity and amusement, but they do not correspond to important social divisions within the larger Garo community. The examples in this paper are from the dialect that has become the de facto standard, originally that of the northeastern corner of the Garo hills.

Brief word lists were collected about 1800 by British officials (Eliot 1794, Hamilton 1940 [1820]) but more extensive descriptions of the language had to wait almost a century before American Baptist Missionaries produced the first grammars and dictionaries. The romanization introduced by the missionaries is now well established, and the language has been used as a medium of elementary education in the Garo Hills for many decades. A few collections of Garo stories, have been published, and a few thin weekly newspapers appear, but apart from school books, the most important publications are religious, and since most Garos are now Christians these include the bible which has long been available in Garo translation, hymn books, and various other Christian texts. The written language is used for private correspondence and a few signs are posted in the language, but a fluent Garo reader would certainly require no more than a few months to read the entire corpus of Garo printed literature.

By the standards of northeastern India, the Garo community is well served by Garo to English dictionaries. The most widely available is the modestly titled but reliable The School Dictionary, Garo to English (Nengminza 1946 and later). It can be supplemented with two others: Marak 1975, and Holbrook 1998 [1940]. An early English to Garo dictionary (Mason 1905) is reprinted periodically, and dictionaries based closely upon it but under the names of other authors have been published as well. The English to Garo dictionaries all consist of Garo definitions of English words rather than Garo equivalents for English words. This is useful for Garos, for whom the dictionaries were written, but it is awkward for someone who does not know Garo. Garo grammars are less satisfactory than the dictionaries. Keith (1873) and Phillips (1904) were early sketches of the grammar written by missionaries. After a period of ethnographic field work, I wrote a somewhat amateurish grammar (Burling 1961) and hope to complete a fuller description of the language in the near future.

Garo does not have tones, but its syllable structure is very much like that of an east or southeast Asian tone language. Syllable boundaries are phonologically sharp, and except in borrowed words (of which there are a great many), 90% of syllable boundaries probably correspond to morpheme boundaries. A single syllable virtually never includes more than a single morpheme, but some two-syllable morphemes are found, even among native words.

. p . t . k . '

. m . n . ng

. m'. n'. ng'

. l . l'.(r) (s)

Syllable initial consonants and consonant clusters are shown in Table 1, and finals in Table 2. Initial /p, t/ and /k/ are aspirated and /b, d/ and /g/ are voiced, very much as in English. Syllable final stops are unvoiced but also unaspirated and unreleased. The nasals are all very much as in English, even to the extent that /ng/ does not occur initially. /s, ch/ and /j/ are all more palatalized than English /s/ but less so than English /ch/ or /j/. /s, ch,/ and /j/ are homorganic. /r/ is a flap. Except in borrowed words, /l/ does not occur as a syllable initial (hence the parentheses in the chart) and /r/ does not occur as a syllable final, so they are in complementary distribution and could be transcribed with the same symbol. Even ignoring the problem of borrowed words, however, a system of writing that does not mark syllable boundaries is made clearer by writing them differently. Mol-a 'tobacco mixture' and mo-ra 'round basketry stool' are pronounced very differently. /s/ occurs as syllable final only in borrowed words.

Glottal stops can occur syllable finally, either alone or in combination with a nasal or an /l/. Minimal pairs for the presence and absence of a glottal stop are plentiful: cha-a 'grow', cha'-a 'eat'; ring-a 'drink', ring'-a 'sing'. When used with a nasal or lateral, the glottal stop is pronounced right in the middle of the other phone, but Garos conventionally write it second. This has the advantage of marking the end of the syllable and so avoiding a few ambiguities, but it has the disadvantage of making a rather simple morphophonemic process seem more complex than it really is. A glottal stop never occurs in the final syllable of a Garo word, and whenever a glottal threatens to appear as word final, an echo vowel is inserted that protects it. For example, the combining form do'- 'bird' (do'-ni 'bird's'; do'-tip 'nest') becomes do'-o when no other bound morpheme follows. Syllables that end with both a glottal and another consonant undergo a similar change: gol'- 'stick' (gol'-chok 'pointed stick, stake'; gol'-ko 'stick' accusative) becomes go'-ol when used without a suffix. The rule that inserts the echo vowel suggests that the glottal stop is rather insecurely joined to the other consonant of its "cluster", and I have even suggested that the glottal stop is a rather tone-like constituent of the syllable (Burling 1992. For a contrary opinion, see Duanmu 1994)

Orthographic Garo has five simple vowels. This is phonologically appropriate and the only serious complication is that /i/ embraces both high front and high back unrounded vowels. Since, in native Garo words, high front vowels are found only in open syllables and high back unrounded vowels only in closed syllables, they are in perfect complimentary distribution. Writing them with the same vowel would be entirely appropriate if syllable boundaries were consistently marked. Since they are not, ambiguities occasionally arise. In fact, similar though less salient ambiguities arise with all the vowels, since they are all shorter in closed than in open syllables, but even native speakers find the phonetic difference between open and closed syllable /i/ to be highly salient while the rather modest length variation shown by other vowels is hardly noticed.

The glottal stop never occurs in the second syllable of a word. The loss of the glottal stop is apparent in many pairs such as pil'-a 'return' andkat-pil-a 'run back' (kat- 'run'). The glottal stop reappears in third syllables, as in kat-ba-pil'-a 'run back here' (-ba- 'in this direction'). It is difficult to construct fully convincing examples where a syllable with an underlying glottal stop appears as the fourth syllable in a word, but it seems to disappear in that position just as it does in second syllables.

Any two vowels can be adjacent if they occur in successive syllables without intervening consonants. It is difficult to find clear criteria by which to consider some vowel sequences but not others to be diphthongs, but /ai/ and /ao/ occur regularly enough with no morpheme break between them to suggest that they should be counted as diphthongs and thus to constitute a single syllable: ai-ao 'wow, my gosh!'

At its simplest, a Garo sentence requires nothing except a verb base and a tense suffix. Optionally, one or more nouns, noun phrases, pronouns and adverbs can precede the verb, and with the help of additional affixes, the verb itself can be made very complex. Here is a very ordinary Garo verb that can act as a complete sentence.

speak -answer-again-NEG-FUT-probably

'[He] will probably not answer again'.

The only obligatory parts of this verb are the verb base a-gan- 'speak' (a two-syllable morpheme) and -wa, a tense marker for 'future'. The three morphemes that occur between the verb base and the tense marker are examples of an extensive class that I will call "adverbial affixes". A much smaller number of suffixes can follow the tense marker. I will call these "post-tense suffixes".

The future is shown by -wa only when it follows the negative -ja-. In positive sentences the future marker is -gen. -gen and -wa form the only fully suppletive pair in the language and their alternation is one of the very few genuine morphological irregularities. Usually, morphemes follow each other with almost no phonological modifications. In addition to -gen/-wa, the tense markers include -a 'present, neutral', -a-ha 'past', -gin-ok or -na-jok 'immediate or intentional future', -bo 'imperative', -na-be 'negative imperative', and -jok a suffix that indicates a change of state. -jok- can often be translated by a perfect tense: cha'-jok 'has eaten', but a more literal translation would be 'has changed from the state of not having eaten to a state of having eaten'. The literal meaning becomes clear in the negative: cha'-ja-jok 'not eat any more', or, more literally, 'has changed from a state of eating to a state of not eating'

Post-tense suffixes include -chim a kind of perfect or irrealis marker. It shows that the proposition is untrue now but that it was true once or might be true at some other time: re'-ang-g en chim 'would go' (-gen 'future'); re'-ang-a-ha-chim 'had gone' (-a-ha 'past'); re'-ang-gin-ok-chim 'would like to go' (-gin-ok 'intentional future').

Other post-tense suffixes include -kon 'probably', -ma 'question marker for yes-no questions'; -mo 'question marker used when expecting agreement' (i.e. tag questions), -na 'it is said' (quotative), -ne 'please' (used to soften imperatives).

In addition to the suffixes that can follow a tense marker, Garo has scores of affixes that can be placed between the verb base and the tense marker. These include nonproductive affixes that can only be added to a limited number of verb bases and that sometimes confer quite idiosyncratic meanings. For example, -chak- generally indicates some action directed toward another person: a-gan-chak-a 'answer' (a-gan-a 'speak'), dak-chak-a 'help' (dak-a 'do, make'), ra'-chak-a 'borrow' (ra'-a take, bring), ka-sa-chak-a 'feel pity' (ka-sa-a 'love'). Several affixes that show direction of motion can be used freely with verbs that describe motion, but not with others: -ba- 'in this direction'; -ang- 'away'; -on- downward. Still others are fully productive: rong- 'habitually', -be- 'very', -tai- 'again', -grik- 'each other, reciprocal', -at- 'causative', -tok- 'all',-ku -'still, yet', -ja- 'negative', -eng- 'progressive', and many others. Several of these affixes can be used together with the same verb and their order is almost completely fixed. The least productive affixes are always closest to the verb base while the increasingly productive ones come later. The following examples illustrate a few of the most productive of these affixes:

a-song-tai -ku -ja -eng -a

sit -again-yet -NEG -PROG -PRES

'is not yet sitting again'

help -reciprocal -CAUSE-PRES

'make [them] help each other'

fly-down-habitually-PAST

'regularly flew down'

The pieces that are glued together to form verbs can be chosen with great freedom. Except for the least productive of these affixes, their meanings are transparent and consistent. They might almost be regarded as separate words rather than as bound morphemes, but there is an intonational unity to the entire set of morphemes that form a verb, and even when they widely separated, the two obligatory parts, the verb base and the tense marker, pull the whole set together. Some, though by no means all, of the affixes that go between the verb base and the tense marker are transparently derived from independent verbs: pil'-a 'return'; re'-ba-pil'-gen 'will come back.'

Word Order. Garo sentences require nothing except a verb. Even though Garo has no hint of verb agreement, neither agent, patient, nor any other actor needs to be explicitly mentioned so long as the larger verbal and nonverbal context makes the intended meaning clear. Nouns, noun phrases, pronouns, and adverbs are, of course, frequently used to flesh out the meaning of a sentence, and noun phrases can have great internal complexity. Nevertheless, they are not essential.

Like most of its Tibeto-Burman cousins, Garo is an SOV language. When asked to provide linguistic examples, speakers almost always place the verb last. In running speech, however, it is not uncommon for a pronoun, or more rarely a noun, to be moved to the post-verbal position. Occasionally, they are even copied to the post-verbal position.

eat -yet -not I -EMPH

'[I] haven't eaten yet, not me'.

he/she different village-to run-away-PAST he/she

'He/she ran away to a different village'.

Post-verbal pronouns such as this are clearly set apart from the rest of the sentence by their intonation, and most pronouns and noun phrases come before the verb. The order of preverbal pronouns and noun phrases relative to each other is quite free. The subject more often precedes than follows the object, but the role of each pronoun and noun phrase is so clearly shown by its case marker that the order can easily be changed.

Pronouns. Among the simplest noun phrases are the personal pronouns, but unlike nouns, several pronouns have a free (or nominative) form that differs from the form to which other case markers are attached.

Free/Nominative, Combining

I ang-a, ang-

you, sg. na'-a, nang'-

he, she bi-a, bi-

we, exclusive ching-a, ching-

we, inclusive an'-ching, an'-ching

you, pl. na'-sim-ang na'-sim-ang

they, human bi-sim-ang, bi-si-mang

Notice that four of these pronouns end with -a. When any (other) case marker is used, it replaces the -a, and the -a can be regarded as a nominative suffix that is used only with monosyllabic pronouns. The -a forms of the pronouns are also used as the free or citation forms. Polysyllabic pronouns, like nouns, lack any overt mark for the nominative. Na'-a 'you singular', has an irregular combining form, nang'-, another of the handful of morphological irregularities in Garo. All the case markers can be added to pronouns as easily as to nouns: ang-o 'with/by me' (locative), nang'-ni 'your' (genitive), ching-ko 'us' (accusative), na'-sim-ang-na 'to you all' (dative).

Garo pronouns do not have phonologically reduced forms. If a pronoun is pronounced at all, it is fully stressed, and where English might use a reduced form, Garo simply leaves out the pronoun entirely, relying on the context to provide the sense.

To anyone familiar with the stability of pronouns in many languages, a surprising fact about Garo pronouns is their dialectal variability. Many speakers of the A'beng dialect use the following pronouns in place of those given above:

'we inclusive' na'-ching

'you plural' na'-song, no'-ong

'they' bi-song

In addition, orthographic Garo uses u-a (less often i-a) for 'he, she' rather than bi-a., and it uses u-a-mang for 'they'. In all varieties of Garo, u-a and i-a are the demonstrative pronouns and they are also used where English has 'it'. U-a-mang is a plural form of u-a. The use of i-a, u-a, and u-a-mang as personal pronouns in written Garo may be based upon genuine oral usage in the dialect of the northeastern part of the Garo Hills, although in my experience, even educated and literate speakers more often use bi-a and either bi-si-mang or bi-song in their colloquial speech. Ang-a, na'-a, and ching-a are, as far as I know, used without variation by all Garo speakers.

Noun phrases that are more complex than pronouns can have some or all of the following constituents although none is obligatory, not even the noun: 1. Demonstrative; 2. Genitive; 3. Classifier phrase; 4. Modifier (deverbal adjective, relative clause); 5. Noun; 6. Case Marker and Postposition. Demonstratives and genitives always come first, while case makers and postpositions always come last so, when present, these two constituents frame the noun phrase. Case markers are suffixed to the final constituent of the noun phrase (except for any following postposition), whatever that may be, so they are more accurately called "clitics" than "suffixes". Classifier phrases and simple modifiers more often follow the noun than precede it, but they can, and often do, precede instead. Relative clauses always precede. When both a modifier and a classifier phrase are used in the same noun phrase, there is some tendency to put one, often the modifier, before the noun and the other after it, although any order is possible. If both the modifier and classifier phrase are placed on the same side of the noun, the modifier is always closer to the noun. A noun is not a required constituent of a noun phrase. A demonstrative, a classifier phrase, or a modifier can be used with no noun at all, but they will still take a case marker like any other noun phrase.

1. Demonstratives. The most important demonstratives are i-a 'this' and u-a 'that'. They can be used as either adjectives or pronouns. In a language without obligatory articles u-a, and less often i-a, is often used when a definite meaning is essential: u-a man-de 'that person, the person'. As pronouns, i-a and u-a are the nearest equivalents to English 'it' and, like other pronouns, they can take case markers. Like other monosyllabic pronouns, they drop their final -a when another case marker is added.

I DEM-ACC see -PAST

'I saw it, I saw that'.

2. Genitives. Genitives are formed by suffixing the case marker -ni to a noun or pronoun. A genitive always precedes the name of the thing possessed: ang-ni jak 'my hand'; nang'-ni ma'-gip-a 'your mother'; bi-ni nok 'his/her house'; nok-ni bol-gru 'the ridge pole of the house'. A genitive can be used without mentioning the thing possessed and can even be followed by another case marker:

I -GEN-ACC look-IMP

'Look at mine'

3. Classifier phrases. Garo has a rich set of numeral classifiers that are used with numbers and chosen according to the nature of the thing being counted: people, animals, roundish things, thin flat things, long thin things, poles, posts, slices, portions, parts, teams, groups, kinds, number of times, abstract things such as stories or ideas, and many others. Even unsophisticated speakers are sufficiently aware of the classifier system to advise learners to use -ge, when a more specific classifier is not known. -ge covers a residual category of mostly physical objects.

In addition to these core classifiers, three other sets of words are used in very much the same way as classifiers, though each has its own special characteristics. 1. Containers. The name of any container can be used to count units of the amount it can hold. Borrowed words pose no problem. gil-es-gin-i 'two glasses of', nok-git-tam 'three houses of' (i.e. 'three families'). 2. Time words. Units of time can be used with numbers just as classifiers can, but unlike ordinary classifiers the resulting phrase cannot be used with a noun: sal-sa 'one day'; wal-gin-i 'two nights'; ja-git-tam 'three months'. 3. Measures. All units of weight and size can be used with numbers to indicate the amount of some material that is being counted. Again, borrowed words pose no problem: ba'-ra mik-sa 'one cubit' (the length from finger tip to elbow) of cloth', mail-bri 'four miles', gong-bri 'four rupees.'

Classifiers are never used without a number, and numbers only rarely without a classifier. Classifiers are sometimes, though not always, omitted when counting. It is quite possible to count sak-sa, sak-gin-i, sak-git-tam. . . 'one person, two people, three people. . .', but sa, gin-i, git-tam. . . 'one, two, three. . .' will also do. Classifiers with numbers often modify nouns. Typically they follow the noun but occasionally they precede: Me'-chik sak-git-tam, sak-git-tam me'-chik 'three women' (sak- 'classifier for people'); meng-go mang-bong-a 'five cats' (mang- 'classifier for animals'). Classifiers with numbers are often used with no noun at all. Their semantic specificity lets them convey considerable information about what is being counted, and often this is all the information that is needed in the context. It would be entirely normal to report having seen three people without using any noun, as long as it is not necessary to say what sort of people they are:

Sak -git-tam-ko nik-a-ha.

people-three -ACC see-PAST

'[I] saw three people'.

It is often possible to choose among several alternative classifiers for the same noun: te'-rik rong-sa 'two bananas'; te'-rik pang-sa 'two banana trees'; te'-rik gal-sa 'one small bunch (hand) of bananas'; te'-rik ol-sa 'one large bunch (arm) of bananas.' This, and the ease with which classifiers can be used with no noun at all makes it impossible to regard them as constituting some elaborate system of gender in which the choice of classifier is governed by the noun.

A few other morphemes than just the numerals can be used with classifiers: prak 'each', sak-prak 'each person', mang-prak 'each animal'; gim-ik 'whole, entire', mang-gim-ik 'whole animal, whole body (of an animal)', sal-gim-ik 'all day'; gip-in 'other, another', mang-gip-in 'another animal'; -san 'alone, only' (cf. sa 'one'), sak-san 'alone (of a person)'. Garo lacks obligatory articles, but a classifier with -sa 'one' is often used where an indefinite article would be used in English: mat-cha mang-sa 'one tiger, a tiger.'

4, Modifiers. Most meanings that, in English, are conveyed by adjectives are, in Garo, conveyed by words that act syntactically like intransitive verbs. Only a handful of core adjectives have distinct syntactic characteristics and these are quite idiosyncratic. Both intransitive and transitive verbs can be used to modify a noun.

A verb is put into a form that can modify a noun by means of a "nominalizing suffix" that is placed in the position that would otherwise be occupied by a tense marker: -a is either the same as the present-neutral tense marker or homophonous to it; -gip-a is used in essentially the same circumstances as -a, but its fuller form makes it more explicit and it is more likely to be used in a long and complex constructions -gip-a is also more likely to be used when a modifier precedes the noun while -a is often used when it follows. The longer form is appropriate to the less common (or "marked") position of the modifier, but either suffix is possible in either position: mat-chu dal'-a, dal'-gip-a mat-chu 'big cow'.

I cow big -NOM-ACC see-PAST

'I saw the big cow'.

Transitive verbs can modify nouns just as intransitives can, but they are often most naturally translated into English as relative clauses.

eat -PROG-NOM person satisfied-FUT

'The man who is eating will be satisfied (full)'.

Satellites of the verb can be drawn into a Garo modifier just as they can be drawn into an English relative clause. In the following example, most of the first sentence is turned into a relative clause in the second sentence, where it modifies man-de 'person'.

I that person -ACC yesterday-LOC see-past

'I saw that person yesterday'.

my yesterday-LOC see-NOM person today -LOC move-away-PAST

'The person I saw yesterday went today'.

As the examples suggest, the line between modification by an adjective and modification by a relative clause is less sharp in Garo than in English. However, as soon as satellites are pulled into a modifier the resulting clause must precede the noun rather than follow. Single-word modifiers more often follow. As befits its relative complexity and its position before the noun, a modifier with many constituents is also more likely to be marked by -gip-a rather than -a. Like demonstratives and classifier phrases, modifiers can be used without a noun, or, perhaps, such words should themselves be regarded as nouns: Dal'-a-ko nik-a-ha-ma? 'Did [you] see the big [one]? When suffixed to a verb base, -gip-a means, approximately, 'the one who' and the resulting word can be used either alone or to modify a noun: cha'-eng-gip-a 'the one who is eating, the eater', cha'-eng-gip-a me'-chik 'the woman who is eating'; dal'-gip-a 'the big one', dal'-gip-a man-de 'the big person.'

5. Nouns. Demonstratives, classifier phrases, and modifiers can all be used without a noun, but when a noun is present it acts as the head of its phrase. Together with verbs, nouns form one of the two largest Garo word classes, but unlike verbs, nouns are frequently used with no suffix at all. Many nouns are both monomorphemic and monosyllabic, but Garo probably has more bi- and tri-syllabic nouns (many of them also bi- and tri-morphemic) than some other Tibeto-Burman languages. The largest number of polysyllabic nouns are compounds, at least one part of which has a transparent meaning, even if that part never occurs alone. Mik- the Garo reflex of the widespread Tibeto-Burman word for 'eye', never occurs alone in Garo, but it does occur in many compounds where it clearly means 'eye': mik-ron 'eye'; mik-chi 'tear' (chi 'water, liquid'); mik-gil 'eyelid' (cf. bi-gil 'skin'); mik-sim-ang 'eyebrow' (cf. pak-sim-ang 'underarm hair', re-sim-ang 'male pubic hair') etc.

Many compounds begin with a classificatory morpheme, almost always a single syllable. For example, dozens of names for birds have do'- as their first syllable: do'-til-eng 'woodpecker', do'-po 'owl' etc. The second part of such words gives the word its specific meaning but many of these second parts are never be used otherwise. As far as I am aware, -til-eng and -po never occur except in the words for 'woodpecker' and 'owl'. Many names for fish start with na- and many names for varieties of trees start with bol-. Many body part terms are constructed in the same way. Unlike mik- 'eye', jak 'hand, arm' and ja'-a 'leg, foot' can be used alone, but they also enter into many compounds: jak-pa 'palm', ja'-pa 'sole of the foot'; jak-sku 'elbow', ja'-sku 'knee'; jak-si 'finger', ja'-si 'toe', and many others. Many plant parts have bi- as their first syllable: bi-gil 'bark, skin, peel', bi-bal 'flower', bi-gron 'pit, large seed of a fruit'. These classifying first syllables of Garo nouns have readily identifiable meanings, so they are quite different from the more generalized "prefixes" that occur in some Tibeto-Burman languages.

Nouns can also be formed from verbs by means of the nominalizing suffix, -a-ni. This yields a noun with an abstract meaning something like 'that which is . . .': cha'-a-ni 'that which is eaten, food' (cha'-a 'eat'), chan-chi-a-ni 'that which is thought, thoughts' (chan-chi-a 'think').

A number of suffixes can be added to nouns, but none is obligatory. -rang 'plural', precedes any case marker that may be suffixed to the noun. The absence of -rang does not necessarily imply 'singular', and to make singularity explicit, a classifier phrase with -sa 'one' must be used. Several other noun suffixes can follow case markers: -de 'emphatic', -sa 'only', -ba 'also'. ang-ko-sa 'only me' (accusative); te-bil-o-ba 'on the table also' (locative).

6. Case Markers and Postpositions. The final constituent of a noun phrase is the case marker, sometimes with a following postposition. It is difficult to draw an absolutely clear line between case markers and postpositions, but the following certainly count as case markers.

Zero/-a . Nominative. Only monosyllabic pronouns and demonstratives have -a in the nominative. The -wa of sa-wa 'who?' and ja-wa 'someone else' can be regarded as an irregular nominative suffix that, like the -a of monosyllabic pronouns, is lost when some other case marker is added. Nouns and longer pronouns lack an overt marker for the nominative.

-ko. Accusative. Garo is a straightforward nominative-accusative language and most objects are marked with -ko. The -ko can be omitted when its noun immediately precedes the verb and when the meaning is not definite, however. Adding -ko gives the noun phrase a definite sense: bol den'-a 'chop wood', bol-ko den'-a 'chop the wood'; mi-ko cha'-a 'eat the rice'; mi cha'-a 'eat a meal'.

-na . Dative, 'for, to'. Ang-a nang'-na ki-tap-ko on'-a 'I give you the book'. A number of postpositions regularly follow datives: ang-na skang 'before me'.

-ni. Genitive. -ni is a straightforward genitive case marker that can be used to show possession of body parts, kinsmen, and physical objects. Many postpositions follow -ni. Some of them are transparently derived from nouns: ang-ni jang-gil-o 'behind me, at my back', (jang-gil 'back of the body', -o 'locative'). Other postpositions have no obvious etymology, although they are parallel in form: bi'-sa-ni gim-in 'because of [the] children, (bi'-sa 'child/children', gim-in 'because of').

-o. Locative, either temporal or spatial, 'in, at, on'. nok-o 'at the house, at home', kin-al-o 'tomorrow', wal-o 'at night'.

-o-na. 'Toward, in the direction of'; -o-ni, from, in the direction away from. Tu-ra-o-ni Reng-sang-gri-o-na 'From Tura to Rengsanggri'; pring-o-ni wal-o-na 'from morning to night.'

-chi. Locative, spatial only. -chi sometimes indicates movement with respect to a place rather than simply a position at a location. Tu-ra-chi 'in Tura, to Tura'.

-chi. Instrumental, with, by means of. This is homophonous with the spatial locative but the meanings seem quite distinct. Ru-a-chi 'with an ax'.

-ming. 'Along with, accompanying'. ang-ming 'with me'.

Large numbers of postpositions can follow one or another of the case markers, the genitive -ni taking the most. It is not always clear whether an ending should count as a case marker or a postposition. A postposition that follows the nominative case (which is generally not marked) would usually be indistinguishable from a case marker, except when following a monosyllabic pronoun where the nominative is distinguished by a final -a. Unfortunately for those who like their grammar unambiguous there is some vacillation. Either ang-a gri or ang-gri can be used to mean 'without me'. The first gri seems like a postposition, since it follows a distinctive nominative form of the pronoun. The second is like a case marker, since it is attached directly to the pronoun's combining form. Even the -na of -o-na 'toward' and the -ni of -o-ni 'from' might conceivably be considered postpositions that follow the locative -o.

Other words that consistently follow one of the case markers are less ambiguously postpositions: gim-in 'because of', a-chak-ni gim-in 'because of the dog'; king-king 'until', kin-al-o-na king-king 'until tomorrow'; pal 'instead of', bi-ni pal 'instead of him'; skang 'before', ang-na skang 'before me', and many others.

Adverbs and Reduplication.

In addition to its nouns, pronouns, verb bases and classifiers, all of which join with other morphemes to form words, Garo has a large class of adverbs, many of them reduplicative or partially reduplicative, that take no affixes at all: pang-nan 'always'; bak-bak 'quickly'; sruk-sruk 'quietly, secretly'; jol-jol 'directly, systematically' pil-ap-pil-ap 'in a flapping manner', pil-eng-pil-eng 'rocking back and forth', gu-rung-ga-rang 'aimlessly' (of wandering about). Many reduplicative, or partially reduplicative, adverbs are transparently derived from verbs: ring-reng-ga-reng 'in a back and forth swinging manner', from ring-reng-a 'to swing back and forth'; rip-ong-rip-ong 'flying around' from rip-ong-a 'to fly around'; srot-srot 'in a sliding manner', from srot-a 'slip, slide'. Such adverbs, however, cannot be productively created from any verb at all. Adverbs are often placed directly before the verb and thus after any noun phrases that the sentence may have, but they can come earlier in a sentence as well.

Some adverbial affixes that are used within verbs also have reduplicated forms, many of them conveying a sense of repetition or continuous action: chot-tip-tip-a 'break (string) into bits', from chot-a 'break'; ru-kring-krang-a 'pour all around', from ru-a 'pour'; sel-gol-gol-a 'leak a lot, rapidly', from sel-a 'leak'. A few verb bases, again often conveying repetitive actions, are reduplicated in form: jok-jok-a 'bounce (as when riding in a bus)', deng-deng-a 'squirm, wiggle'.

Finally, reduplication of numbers to convey a distributive sense is fully productive: gong-gin-i-gin-i 'two rupees each' (gong- 'classifier for rupees', gin-i 'two'); le-ka king-git-tam-git-tam-ko on'-bo 'give three sheets of paper to each' (le-ka 'paper', king 'classifier for thin flat things', git-tam 'three').

The simple Garo sentences that I have considered so far can include various noun phrases and adverbs and a verb. Complex sentences are built from two or more simple sentences of this kind. Only a few of the most common types of complex sentences can be illustrated here.

Very often, one sentence is turned into a subordinate clause that precedes the main clause. The subordination is shown by a verb suffix that fills the position that would be occupied by the tense suffix in a sentence-final verb. The most common of these subordinating suffixes are -e and -e-ming:

I house-to arrive away-SUB eat -PAST

'Having arrived at the house, I ate'.

Ang-ni a-bi-tang nam-en neng'-be -e-ming, nok -chi nap -ang -a-ha

I -POS sister very tired -very-SUB, house-LOC enter-away-PAST

'My sister, being very tired, went into the house'.

Subordinating constructions of this sort are exceedingly common in Garo and they are used to tie together what would otherwise be separate sentences. This gives unity to the discourse by creating what amount to long run-on sentences. They give an impression like an English monologue in which the sentences are linked with phrases such as "and then" or "so".

Notice that we have now considered three types of verb suffixes, which occur in the same position: 1. Tense suffixes, one of which completes each sentence; 2. Nominalizing suffixes such as -gip-a that put a verb into a form that can act as a noun or modify a noun; 3. Subordinating suffixes such as -e and -e-ming. Every verb requires a suffix in this position, and the type of suffix is determined by the syntactic role of the verb: main, nominalized, or subordinate.

A new sentence can also begin with a subordinate form of the same verb that completed the previous sentence. This, too, ties the successive sentences together. In the following example, the verb sok- completes one sentence and introduces the next. The second sentence has a short introductory -e clause followed by a somewhat longer main clause,

Dos ba-ji -o ang-a song -chi sok -ang -a-ha. Sok -ang -e ang-a mi cha'-e tu-si-a-ha

ten o'clock-LOC I village-to arrive-there-past arrive-there-SUB I rice eat -SUB sleep-PAST.

'At ten o'clock I arrived at the village. Having arrived, I ate rice and slept.'

An important set of subordinating suffixes includes -o-de 'if'; -o-ba 'although'; -o-sa 'only if.'

he come-here-not-if I angry -very-FUT

'If he doesn't come I will be very angry.'

The infinitive suffix, -na, also allows two verbs to be used together.

we (inclusive) go-away-INF need-PRES

`We need to go.'

Verbs can also be made subordinating by means of a conjunction such as gim-in 'because of' or ja'-man-o 'after' that follows a verb that has been nominalized by -a-ni:.

sick-ness because I go -away-INF can -NEG

'Because of sickness, I cannot go, Because I am sick, I cannot go'.

Rice cook-ing after -LOC we eat -ables -ACC give -FUT

'After cooking rice, we will serve the food'.

In addition to questions made with question words and simple yes-no questions formed with the verb suffix -ma, Garo speakers frequently use balanced questions in which a yes-no question is immediately followed by a corresponding negative clause which may, but need not, have a question marker as well:

Eat -FUT-QUEST, eat -NEG-FUT-QUEST

'Will you eat or not?

Elephant -ACC see-have-QUEST, see -yet-NEG

'Have you seen elephants or not yet?'

Finally, Garo has a considerable number of conjunctions that can be used to tie successive clauses together. Some of these have specific meanings: u-ni-gim-in 'therefore', literally 'because of that'; un-bak-sa-ba 'in addition to that'; in-di-ba 'but, however'; ong'-ja-o-de 'or' literally 'if [it] is not'. Others mean little more than 'and then, so': u-ni-ko, u-non, in-di-de and others. Speakers of other languages, however, may find the language curiously lacking in a simple equivalent for 'and' that can be used to conjoin not only entire clauses but simple nouns or simple verbs. Instead of conjoining two verbs with a word that means 'and', the first verb is more often subordinated to the second. Two nouns may be used beside each other with no overt conjunction at all, or both may be suffixed with -ba 'also': na'-a-ba ang-a-ba 'you and I, both you and I'.

On their northern, western, and southern borders most of the neighbors of the Garos speak Bengali or a closely related dialect of Assamese. Garos always call this language 'Bengali' and they have probably been borrowing from it for many centuries. The impact of Bengali is particularly strong among the Garos of Bangladesh where all primary education is conducted in Bengali, and where it is needed for everyday dealings with peddlers, shop keepers, government officials, and church leaders. All adult Garos in Bangladesh are able to use Bengali for practical purposes and many are fluent. Bengali presses less insistently upon Garos living in India, for it is not the language of education or government, and even Bengali traders learn enough Garo to deal with their customers in their own language. Nevertheless, even the dialects of the most remote areas of the Garo Hills have absorbed large numbers of borrowed words. The influence of English is more recent, but it now competes with Bengali as a source of borrowings. The influence of English is stronger among the Indian Garos than among Bangladeshi Garos in direct proportion to the relative weakness of Bengali.

The most obvious impact of Bengali and English comes with borrowed words. Many Bengali words are thoroughly assimilated into Garo, but educated and bilingual Garos borrow freely and on the fly. It would be impossible to speak about education, politics, Christianity or modern technology, without calling on borrowed words, and Garos feel free to use any word from Bengali or English that they believe their listeners will understand. These words bring some innovations to the phonology, though mostly by placing familiar sounds in new positions rather than by introducing entirely new sounds

The impact of borrowed words is great enough, particularly in Bangladesh, to worry some Garo. A few despair at the flood of Bengali words that they feel are corrupting their language, but they feel powerless to avoid them. Garos in Bangladesh receive all their education in Bengali and even in the Garo Hills, Garo medium education stops after elementary school. All high school subjects come with their foreign vocabulary. Too little has been printed in Garo to sustain a richly literate community, and well-educated Garos must rely upon English or Bengali for many literate purposes. Nevertheless, with 700,000 speakers, Garo is not yet on the list of endangered languages. Even in outlying areas like Bangladesh, most children of Garo parents still learn Garo as their first language. Whether they will still be doing so a century from now is by no means certain.