Prefixed *m- as a pronominal element can profitably be compared with TB *a-, of almost universal distribution in the family. This element occurs as an independent 3rd person pronoun in Kiranti and Kuki-Naga (a-ma, a-ni), and as a pronominal prefix (ǝ-) in these same groups; cf. Aimol rǝmai ‘tail’, rul ǝrmai ‘snake's tail’; Bahing biŋ ǝta-mi ‘calf’ (‘cow its-child’), byar ǝpwaku ‘sugar-cane’ (‘cane its-juice’). Throughout the TB area in general, however, a lapsing of function can be observed, and the prefix is retained only in forms (normally kinship terms or words for parts of the body) used independently, i.e. without the customary pronominal prefixes, e.g. K wa or ǝwa ‘father’ (nwa ‘thy father’, kǝwa ‘his father’), mun∼ǝmun ‘body hair’, myi∼ǝmyi ‘eye’; Nung ǝkhö ‘uncle’, ǝna ‘ear’; B ӑbhá (ӑphá) ‘father’, ӑmí ‘mother’ (but mí-bá ‘parents’), ӑsà ‘flesh’ (but nwà-sà ‘beef’ = ‘cattle-flesh’); G apa ‘father’, ama ‘mother’; Mikir ari∼ri ‘hand’, aso∼so ‘child’; Lhota Naga okhe ‘hand’, eŋu ‘neck’, oka ‘daughter’, eŋü ‘wife’ (*a->o-∼e-), sho ǝho ‘tooth’, ǝtü ‘grandson’. Semantic specialization is sometimes encountered; cf. B swà ‘tooth’, ӑswà ‘cutting edge’; im ‘house’, ӑim ‘sheath’; myak ‘eye’, ӑmyak ‘knot in timber’; Lepcha uŋ ‘water’, ӑuŋ ‘water in which meat has been boiled’; vi ‘blood’, ӑvi ‘menses’; vyeŋ ‘door’, ӑvyeŋ ‘pass’; kuŋ ‘tree’, ӑkuŋ ‘bush’; rip ‘flower’, ӑrip ‘flower of cloth’.

Prefixed *r-, of uncertain function, appears in a number of noun roots, and must also be reconstructed for a few verb roots. It is preserved in Tibetan, Kachin, Bodo-Garo, Mikir (ar-), and occasionally elsewhere; note especially Magari ar-, as in arghan ‘wasp’, arkin ‘fingernail’, armin ‘name’, but l∂wat ‘leech’ < *r-pat. Kachin usually has l∂- for *r- in verb roots (and in l∂tsa < *r-gya ‘100’), but n-∼niŋ-∼num- in noun roots. Noun roots with prefixed *r- include the following:

1. TB *r-ka ‘earth’: Nung r∂ga, K nga (n-ga).

2. TB *r-saŋ ‘lizard’: T rtsaŋs-pa, K nsaŋ.

3. TB *r-ka·m ‘edge, precipice’: K ngam (n-gam)∼niŋgam, G rikam.

4. TB *r-gu·ŋ ‘edge; shin’: K nguŋ (n-guŋ), G rikiŋ, Dimasa ruguŋ, Mikir arkoŋ.

5. TB *r-luŋ ‘stone’: K nluŋ, Mikir arloŋ.

6. TB *r-miŋ ‘name’: Magari armin, Gyarung -rmi, Rangkhol ermiŋ.

7. TB *r-may ‘tail’: Digaro l∂mi∼l∂miŋ, Aka ∂rim < *∂rmi, K nmai, Dimasa khermai∼bermai, Aimol r∂mai, Mikir arme.

8. TB *r-ni(y)∼*r-ni(y) ‘gums’: T rnil, Dimasa ha-rni G wa-riŋ < *wa-rni (in comp. with ‘tooth’).

9. TB *r-pat ‘leech’: Magari ∂wat, B krwat < *k-rwat, G ruat, Rangkhol ervot, Angari Naga reva.

(441) K nwa∼niŋwa, G rua, Dimasa roa ‘ax’ (TB *r-wa).

TB prefixed *m- is more readily interpreted than the stop prefixes analyzed above. With verb roots this prefix has a ‘middle voice’ force, often durative, intransitive, or reflexive. Tibetan m-, as brilliantly interpreted by Wolfenden, represents a ‘neuter’ subject, as opposed to b- andʾ-representing an ‘acting’ subject; cf. mguba ‘rejoice’, mŋa-ba ‘to be, exist’, mnal-ba ‘to sleep’, mtśhi-ba ‘appear, show oneself’, mnab-pa ‘dress oneself’. Prefixed *m- in this role is retained also in Kachin, Bodo-Garo and Kuki-Naga, while Nung replaces this prefix with phǝ-<*bǝ-: phǝsin ‘liver’ <TB *m-sin, phǝlε ‘tongue’ <TB *m-lay. The contrast with TB prefixed *s- is especially clear in the following root; note that the unprefixed root may be either transitive or intransitive, whereas the prefixed *m- form is always intransitive:

(464) T mnam-pa ‘to smell, stink’ (intr.), snam-pa∼snom-pa∼snum-pa ‘to smell’ (tr.), Lepcha nom<*nam ‘to smell’ (intr.), nyom<*s-nam (tr.); Vayu nam ‘to smell’ (tr.), nam-saŋ ‘odor’; Bahing nam ‘to smell’ (tr.), nam-ba ‘having odor’; Miri nam ‘to smell’ (tr.); K nam ‘to taste or smell, as of spices’, mǝnam ‘to smell; smell, scent’ (mǝnam nam ‘to smell offensively’); Nung phǝnam ‘to smell’ (use uncertain); B nam ‘smell offensively, stink’ (intr.), nàm ‘smell, receive scent’ (tr.), ӑnám ‘odor’; Bodo manam ‘to smell’ (intr.); Dimasa maram ‘to stink’ (n<*r through dissimilation); L nam, Ao Naga menem ‘to smell’ (intr.); Tangkhul η∂nam ‘odor’, kh∂η∂nam ‘to smell’ (intr.); Mikir iηnim ‘to smell, be odorous’ (intr.), aηnim ‘odor’, nem-so ‘slight smell, stink’ (-so is diminutive), from TB *m-nam.

Two great taxonomic problems must be considered in connection with Sino-Tibetan, viz. the nature of the affiliations of the three primary groups, and the classification of the multitudinous divisions within Tibeto-Burman itself. The former of these problems has been resolved in the following manner. Tibeto-Burman and Karen are regarded as constituting a superfamily (Tibeto-Karen) standing in opposition to Chinese. The relationship between Tibeto-Karen and Chinese is a distant one, comparable with that between Semitic and Hamitic, or between Altaic and Uralic. Karen, on the other hand, stands in relation to Tibeto-Burman much as Hittite stands in relation to Indo-European, i.e. Tibeto-Karen is on the same taxonomic level as Indo-Hittite. On the negative side, Sino-Tibetan must be kept distinct from all other linguistic stocks.

The writer has recently attempted to show that Thai is related to Indonesian rather than to Chinese, and that the traditional view of a Chinese-Thai relationship must be abandoned. A number of students, including Ramstedt, Donner, Lewy, Bouda, and Findeisen, have sought to connect Yenisei-Ostyak (Ket) with Sino-Tibetan, and this view has gained some favor (Schmidt, Trombetti), yet a critical examination of the evidence strongly indicates that the two stocks have nothing in common. Sporadic attempts to connect Sino-Tibetan with Caucasic (Hodgson, Bouda), Mon-Khmer (Conrady), or other linguistic families have been equally unsuccessful.

The manuscript of this work, completed ca. 1942–3, was put aside until such time as further analysis could be attempted. It lay buried in the clutter of the author's library until unearthed in 1968 by Professor James Matisoff of Columbia University, who had it mimeographed to serve as a text for his pioneering course in Tibeto-Burman offered at that university. Its appearance in published form at this time is entirely the product of the enthusiasm of Professor Matisoff, who generously consented to edit this work, bringing the bibliographic data up to date and supplying modern material from his researches in the Burmese-Lolo branch of the stock.

The author prepared a new version of the original manuscript, rearranging some of the material and adding the minor emendations noted on the manuscript, then extensively annotated the whole, with emphasis on the Karen and Chinese sections. These annotations represent in part previously published findings (especially in Benedict, 1948 bis), in part an intensive re-analysis of all the Sino-Tibetan materials, aided by the more recent publications in the field, notably those of Haudricourt (Karen reconstructions), Forrest (Lepcha analysis), Jones (Karen), Burling (Karen, Burmese-Lolo, Bodo-Garo and Kuki-Naga), Matisoff (Burmese-Lolo), Henderson (Tiddim Chin), Stern (Siyin), Kun Chang (Gyarung and Ch'iang) and Lo Ch'ang-p'ei (Trung).

The relationships that obtain among the several units of the TB sentence are indicated (a) through the relative positions of the units, and (b) through the employment of special relating morphemes, normally prefixes or suffixes. The syntactical factor tends to be the dominant one, however, hence one can describe Tibeto-Burman as ‘relatively isolating’. Throughout the TB area the invariable syntactical rule is that the verb must be placed at the end of the sentence, followed only by suffixed elements or sentence-final particles. The object normally immediately precedes the verb and follows the subject, though no invariable rule can be stated here (in Burmese the object is somewhat emphasized when placed before the subject). The concepts of ‘subject’, ‘object’, ‘indirect object’, ‘instrumentality’, and the like are reinforced or expressed in modern TB languages by morphemes suffixed to nouns. The subject is often found standing alone, or construed as an instrumental, as in T ŋa-s kho-la rduŋ ‘by-me to-him beat’ = ‘I beat him’, ŋa-s de śes (or ŋa-la in modern dialects) ‘by-me that know’ = ‘I know that’. Subordinated elements regularly precede rather than follow, although modifying elements are often suffixed; cf. Modern B tyìdέ; khwè ‘big dog…’ or ‘dog (that is) big…’ (-dέ with ‘creaky tone’, a morpheme of subordination), khwè tyìdε ‘dog is-big’, khwèdyì ‘big-dog’ (t<d in intervocalic position).

All five vowel phonemes occur in medial position. Lushei distinguishes between short and long vowels in this position, and this distinction is reflected in Haka and other Kuki-Naga languages. Other TB languages, insofar as they have been recorded accurately, do not show this feature in any consistent way, although vowel length is sometimes marked. The Lushei distinction between a and a·, and between u and u·, is reflected in certain correspondences in Bodo-Garo and Burmese-Lolo (see below), hence we must suppose that the distinction obtained also for o and o· e and e·, i and i· although it is possible that the TB vowel system was asymmetrical. Lushei has relatively few forms with long vowels connected with general TB roots, and it would appear that TB medial vowels were ‘normally’ short (all final vowels were phonemically long). Numerous examples of roots with short medial vowel are scattered throughout the preceding pages; in the discussion below emphasis is placed on roots with long medial vowel.

Medial a is preserved before all types of finals in Tibetan, Kachin, Burmese, Garo, Lushei and most TB languages. Lepcha, which has -o for TB final *-a, normally shifts to o, e.g. ǎtsom ‘hair’ < TB *tsam, lom ‘road’ <TB *lam, but tyaŋ ‘dark’ <TB *tyaŋ.

Chinese is the third and last major division of Sino-Tibetan to be considered in this review. Three stages of the language are conveniently recognized: (a) Archaic Chinese (Ar. Ch.), ca. 1200–800 B.C., (b) Ancient Chinese (Anc. Ch.), ca. A.D. 600, and (c) the modern dialects. Ancient Chinese has been reconstructed from the modern dialects together with the material found in the Ch'ieh Yün and other lexicographical works of the first millennium A.D. Archaic Chinese represents a still farther projection into the past, achieved through the analysis of the Shih Ching rhymes and the phonetic elements of Chinese characters. A number of scholars, including Maspéro, Simon and Li Fang-kuei, have contributed to the brilliant results attained in this field, but we are indebted above all to the monumental studies by Karlgren. Our purpose here is not to review the developments within Chinese itself, but rather to study the earliest known stage of the language (Ar. Ch.) in the light of our reconstruction of Tibeto-Burman and Tibeto-Karen. The forms cited below accordingly are those of Ar. Ch., often along with the later Anc. Ch. forms, all as given in the Grammata Serica of Karlgren.

Final consonants are greatly reduced in Karen. Pwo parallels Modern Burmese in replacing final nasals by nasalization, and final stops by glottal stop. Sgaw lacks even nasalized vowels, but has glottal stop as in Pwo. Taungthu appears to retain distinctions in final nasals at least in part, but final glottal stops have not been recorded for this language. Several of the Taungthu words with final nasal are isolated in Karen; cf. prŏŋ ‘mouth’, hau poŋ ‘good’, lɔn ‘come’, lam ‘house’, kam ‘gold’ (a Thai loan-word). The regular correspondences for TB final nasals are illustrated below:

Pwo khą, Sgaw khɔ, Taungthu kaŋ-ya ‘foot, leg’; cf. T rkaŋ-pa.

Pwo kǝshą, Sgaw kǝshɔ ‘elephant’; cf. B tshaŋ.

Pwo mi mą, Sgaw mi mɔ ‘dream’ (in comp. with ‘sleep’); cf. TB *maŋ.

Pwo wą, Sgaw wɔ ‘surround; circular’; TB *hwaŋ.

Pwo ą, Sgaw ɔ, Taungthu am ‘eat’; cf. the following root:

(481) Nung am ‘eat’, Dhimal am ‘drink’ (TB *am).

Pwo phü-thą∼phu-thą, Sgaw thɔ ‘bear’; TB *d-wam.

Pwo γε, Sgaw γe ‘row’; TB *ren.

Pwo khlε ‘speak’; cf. T gleŋ-ba ‘say, talk, converse’.

Pwo khε, Sgaw ki ‘tie around, gird, bind’; cf. B khyań ‘bind, fasten’.

Pwo nǫ, Sgaw nǝ, Taungthu nuŋ ‘horn’; cf. TB *ruŋ (K nruŋ).

Pwo lǫ, Sgaw lǝ, Taungthu luŋ ‘stone’; TB *r-luŋ.

The Chinese tonal system can be interpreted in terms of three tonemes, viz. level (unmarked), rising, and falling, or pʿing shêng, shang shêng, and chʿü shêng, respectively. The so-called ju shêng of Chinese philologists is simply the level tone in syllables with final stop consonant (glottal stop in many modern dialects). The three tonemes are conditioned (in Ar. and Anc. Ch.) by the initial, being relatively high in words with surd initial, relatively low in words with sonant initial. With the general shift from sonant to surd initials shortly after the Anc. Ch. period (A.D. 600–900), the high and low varieties of each toneme tended to become phonemically distinct, so that all modern dialects have several separate tonemes. The falling toneme has generally been regarded as of late origin, by Chinese as well as Western scholars. It is undoubtedly significant that many Anc. Ch. words derived from Ar. Ch. forms in -g or -d should bear this toneme, but we find it also with words in final -m, -n, or -ng, in which loss of final cannot be postulated. We must infer, then, that all three tonemes existed in Ar. Ch.

A careful comparison of Chinese tones with the Tibeto-Karen system represented by Burmese-Lolo and Karen has yielded no positive results.

The study of TB morphology is in large measure simply the study of those prefixed and suffixed elements which can be shown to be of some antiquity. Certain of these prefixes (*g-, *b-, *l-, *d-) have already been pointed out in connection with the numerals. In many instances, as here, no function can be assigned these elements, i.e. loss of morphological utility had already occurred in proto-TB times. A few suffixed elements can be readily analyzed. They include the ‘gender’ suffixes *-ma (fem.) < *ma ‘mother’, and *-pa (masc.) < *pa ‘father’, as well as *-la (masc), used with words for animals (in Tsangla, Digaro, Nung, Kachin, Burmese-Lolo, Konyak, Garo-Bodo, Mikir, and Meithei); also the verbal noun (infinitive) suffix -pa∼-ba ‘that which is’ (in Tibetan, Bahing, Meithei, Garo-Bodo, Burmese-Lolo); cf. T khyi smyon-pa ‘mad dog’, lit. ‘a dog, one which (-pa) is mad’ (see Wolfenden, Outlines, p. 75); Lahu qai-pa mâ- ‘there is no one to go’, lit. ‘one-to-go there-is-not’. This suffix is probably connected with the masculine noun suffix -pa mentioned above; note that Meithei sometimes distinguishes between -ba (masc.) and -bi (< Kuki-Naga *pwi) (fem.) in adjectival forms, paralleling the distinction occasionally made in Tibetan, e.g. dma-mo ‘low’ but mthon-po ‘high’, rgad-po ‘old man’, rgad-mo ‘old woman’ (rgad-pa∼rgan-pa ‘old’).

The negative elements *ma and *ta precede the verb in Tibeto-Burman (*ma is often prefixed, as in Burmese).

The Arc. Ch. vowel system, as reconstructed by Karlgren, comprises some 10 vowel phonemes, half of which are distinguished quantitatively:

Front vowels: i; ĕ∼e; ε

Back vowels: ŭ∼u;

Central vowel: ǝ

Low vowels: â; ӑ∼a

It is apparent that this vowel system is far richer than anything to be found in Tibeto-Burman, and indeed serious difficulties arise in comparative analysis. In final position only the following vowels (all long) are found: a, â, å, o, and u. Final i appears only as the first member of a diphthong, while e, ε, ô, and ǝ appear only before final stop, nasal, or -r. The dissimilarity of the two systems, then, extends even into the features of distribution.

Diphthongization, as already pointed out by the writer (Benedict, 1940), is the keynote of the development of vowels in Ar. Ch. This feature is best revealed in the Ar. Ch. treatment of original medial *i before surd stops and nasals (*i>-įĕ-):

sįĕn ‘firewood’; TB *siŋ ‘tree, wood’ (Trung also ‘firewood’).

sįĕn ‘bitter’; TB *m-sin ‘livery’ <*sin ‘bitter, sour’.

mįĕn ‘order, command’ (this earlier reading for mįӑng is revealed in several Shih Ching rimes); cf. B mín, id.

įĕt ‘1’; cf. Kanauri id.

ńįĕt ‘sun, day’ (with suffixed -t); TB *niy.

TB prefixed *s- in verb roots is directive, causative, or intensive. It plays a prominent role in Tibetan (s-), Gyarung, Kachin (ś∂-∼dź∂-), and Nung (ś∂-), as well as in Lepcha (in the form of palatalization) and Burmese (in the form of aspiration or surdization of the initial); cf. T ʾkhor-ba ‘turn round’, skor-ba ‘surround’; K dam ‘stray’, ś∂dam ‘lead astray’; thum ‘to be ended’, dź∂thum ‘to end’ (dź∂- for ś∂- before surd stops); Nung ∂nem ‘to be low’, ś∂nem ‘make low, lower’; Lepcha thor ‘escape, get free’, thyor ‘let go, set free’ (T thar-ba ‘become free’); rop ‘stick, adhere’, ryop ‘affix, attach’; nak ‘to be straight’, nyak ‘make straight’; B pyauk ‘disappear, be lost’, phyauk ‘cause to be lost, destroy’; lwat ‘to be free’, hlwat ‘free, release’ (cf. the discussion in §8). Maru li ‘come’, ś∂li ‘bring’ (‘cause to come’), cited only by Abbey, lends support to our interpretation of the Burmese data, although it must be pointed out that Maru has come under direct Kachin influence. Prefixed *s- with verbs appears only sporadically elsewhere, e.g. Kanauri stam<snam ‘give forth smell’, an intransitive rather than transitive form (T snam-pa is tr.); G stu ‘spit’ (see n. 189).

The prefixes and suffixes (apart from those used with numerals) of the reconstructed TB speech are listed below. In modern TB languages the prefixes normally have reduced stress and the neutral ∂ type of vocalization. Thus, the form written *g- is to be interpreted as *g∂ (with ∂ as a separate phoneme) or as gǎ (with ǎ an allophone of the phoneme /a/ in syllables with reduced stress). The vowel of the prefix is affected by vocalic harmony in several groups, notably Bodo-Garo and Mikir; cf. Dimasa gabaŋ ‘much’, gosoŋ ‘steep’, gusum ‘blue’, gepher ‘flat’, gimin ‘ripe’.

1. Suffixed *-s: original function uncertain; often reflexive in verb roots.

2. Suffixed *-t and *-n: original function uncertain; sometimes used in deriving nouns from verb roots; also causative or directive.

3. Prefixed *s-: causative, directive, or intensive with verb roots; often stands for TB *sya ‘flesh; animal’ in noun roots.

4. Prefixed *r-: both in verb and noun roots; function unknown.

5. Prefixed *b-: perhaps pronominal in some roots, but function generally unknown.

6. Prefixed *g-: rare; function unknown.

7. Prefixed *d-: rare; function unknown.

8. Prefixed *m-: pronominal in noun roots; intransitive in verb roots.

The 1st person and 2nd person independent personal pronouns are *ŋa ‘I’ and *naŋ ‘thou’:

(406) T ŋa, Kiranti aŋ (Rai, Rungchengbung) ∼ aŋ-ka (Waling) ∼ ka-ŋa (Rodong)∼∂ŋa (Limbu), Nung ŋa, B ŋa, G aŋ ‘I’, with which must be grouped *ŋay ‘I; self’, and perhaps Dhimal ka, L (and general Kuki) ka ‘I’.

(407) Thami, Magari, Chepang naŋ, K naŋ∼na, B naŋ ‘thou’, G na?a ‘thou’, naŋ-ni ‘thy’ (cf. B ńàń<*nyìŋ ‘thou [female to female]’), L naŋ, also Dhimal and Nung na ‘thou’.

Subordination is effected simply through anteposition or prefixation to the noun, often in abbreviated form, e.g. K naŋ∼na ‘thou’, nwa (n-wa) ‘thy father’. Pronominal inflection, clearly of secondary origin, is encountered in the Himalayish group and occasionally elsewhere; cf. B ŋa ‘I’, ŋá ‘mine’; naŋ ‘thou’, náŋ ‘thine’ (see above); Dhimal ka ‘I’, na ‘thou’; kaŋ ‘my’, naŋ ‘thy’; kyel ‘we’, nyel ‘you’; kiŋ ‘our’, niŋ ‘your’. Various types of refinements, none of which can be regarded as inherited TB features, appear in random distribution. These include the distinction between exclusive and inclusive forms of the 1st person pronoun (notably in Himalayish, also in Tibetan and Mikir), the dual (Kanauri, Tibetan, Kachin), and distinctions in sex of speaker (notably in Burmese). The concept of plurality is generally expressed through suffixation (as for nouns).