An overview of Sino-Tibetan morphosyntax

Overview of Sino-Tibetan Morphosyntax Randy J. LaPolla City University of Hong Kong 1. Sino-Tibetan At the earliest reconstructable stage of the development of the Sino-Tibetan (ST) language family, possibly as much as eight thousand years ago (Thurgood 1994), the proto-language was monosyllabic. Matisoff (1991a:490) reconstructs the syllable canon as below: *(P) (P) Ci (G) V (:) (Cf) (s)1 It is not clear whether the prefixes in some or all cases entailed a vocalic element. If so, the structure might have been sesquisyllabic (that is, a syllable and a half (Matisoff 1973a); e.g. as in the name t«˙ r\ung ‘T’rung/Dulong’, the vocalic element of the t«˙ - prefix is very slight). There was no relational morphology (LaPolla 1990, 1992a, 1992b, 1993b, 1994b, 1995a, 1995b), but there was derivational morphology in the form of prefixes, suffixes, and voicing alternations of the initial consonants (Wolfenden 1929; Benedict 1972; Pulleyblank 1962-3, 1972, 1973a, 1973b, 1977-78, 1991, 2000; Bodman 1980; Mei 1980, 1988, 1989; LaPolla 1994c; Gong 2000). Following are examples of several types of derivational morphology.2 1P = prefix, Ci = initial consonant, G = glide, : = vowel length, Cf = final consonant, s = suffixal *-s; parentheses mark that the item does not appear in all syllables. 2This list is probably not exhaustive, and the necessarily brief discussion glosses over many controversies and details. As is always the case in attempting to find Sino-Tibetan correspondences, the lack of a single standard for the reconstruction of Old Chinese (ideally based mainly on the comparative method) makes comparative work difficult and more conjectural than would otherwise be the case. What constitutes a cognate set using one reconstruction system might not be seen as cognate using another system. I have here used the system of Baxter 1992, as this is the best system I have found to date, though even this system is in flux (see Matisoff 1995, footnote 1 for discussion of some of the recent changes). 2 *s- prefix. The *s- prefix in most cases had a causativizing, denominative, or ‘intensive’ (change of state) function (Wolfenden 1929; Pulleyblank 1973a, 2000; Bodman 1980; Mei 1989). Mei (1989) argues all of these functions are manifestations of a more general directive function. E.g. Old Chinese (OC) *mjang (亡) ‘be gone’ : *smangs (喪) ‘to lose’; OC *m˙k (墨) ‘ink’ : *sm˙k (黑) ‘black’ :: Written Tibetan (WT) smag ‘dark’; OC *C-rj˙s (吏) ‘clerk, minor official’ : *srj˙÷ (使) ‘to cause (someone to be an emissary), to send’; *tju÷ (帚) ‘broom’ : *stu÷ (掃) ‘to sweep’; *ljek (易) ‘to exchange’ *sljeks (賜) ‘to give, gift’; WT: grib ‘shade, shadow’ : sgrib-pa ‘to shade, to darken’; gril ‘a roll’ : sgril-ba ‘to roll together, to form into a roll’. *Voicing alternation. In both OC and Tibeto-Burman (TB) we find pairs of cognate lexical items which differ phonetically only in terms of the voicing or aspiration of the initial, and differ semantically in terms of transitivity, where the item with the voiced initial is intransitive, and the item with the voiceless initial is transitive. Benedict (1972:124) discussed this for TB, but argued that in Chinese no consistent pattern of morphological alternation could be recognized. Most scholars now would see the Chinese forms as parallel to the TB forms, and part of a cognate phenomenon. Pulleyblank (1973a, 2000) argues these variant forms should be the result of an intransitivizing prefix *«a- (a non-syllabic pharyngeal glide) which voiced the initial of the original transitive roots. Mei (1989) includes this prefix in a paradigm with the *s- directive prefix and the *-s direction of action changing suffix (below).3 Both Pulleyblank and Mei base the idea for the prefix mainly on the Written Tibetan a-chung (‘small a’) prefix (here marked with an apostrophe). Pulleyblank also equates this prefix with the a- nominalizing prefix found in Burmese. Baxter (1992) adopts this view in reconstructing Chinese forms, and uses *¿- for the form of the prefix,4 e.g. *kens (見) ‘see’ : *¿kens (> 3In a slightly earlier paper, Mei (1988) argues for reconstructing a voiced initial rather than a prefix. 4Baxter (1992:221; following Chang & Chang 1976, 1977) also associates his *N- prefix (posited to account for characters with phonetic elements that appear in syllables with both stop and homorganic nasal initials) with Tibetan a-chung. Gong (2000; also following Chang & Chang 1976, 1977) associates Tibetan a-chung with a nasal prefix, but uses it to explain the 3 *gens) (現) ‘appear/be visible’. While this analysis is attractive from a systemic point of view, Benedict (ibid.) points out that the prefixing and the voicing alternation in Tibetan are two different phenomena that interact in the specialization of different forms as ‘present’, ‘perfect’, ‘future’ and ‘imperative’, such that the present and future forms have the voiced initial and are intransitive or durative, while the perfect and imperative forms have the voiceless initial and are transitive or active. As an example, for the verb ‘put off, pull off, take off’, we have present 'bud-pa and future dbud, which derive from an intransitive stem *bud, and perfect and imperative phud, which derives from a transitive stem *pud. Evidence that it is not the a-chung prefix that is involved in the contrast in Tibetan is the fact that in many cases both forms of a pair of contrasting forms have the prefix, e.g. Tibetan 'gril-ba ‘to be twisted or wrapped round’ ; 'khril-ba ‘wind or coil round, embrace’. Bodman (1980:54) also mentions that he did not find any Tibetan-Chinese cognates where prefixation or lack of it in Tibetan corresponds with the voicing distinction in Chinese. We also find the voicing alternation in TB languages independent of prefixation, e.g. *kh(r)jok (曲) ‘bend, bent’ : *¿kh(r)jok (*g(r)jok) (局) ‘compressed, bent, curved (body)’ :: Bahing kuk ‘make bent’ : guk ‘to be bent’ (TB *kuk ~ *guk; Benedict 1972:125). Pulleyblank’s association of the voicing distinction in Chinese with the a- prefix in Burmese also is problematic, as the latter is a nominalizer, not an intransitivizer, and is independent of the voicing distinction, e.g. Burmese phai ‘break off a small piece from a larger, crumble’ : pai ‘to be broken off, chipped’; (cf. also Qiang ¿e-phe ‘tear (clothes)’ : de-pe ‘ be torn’; TB *pe ~ *be; Benedict 1972:59) (:: OC *phajs (破) ‘to break’ : *paj÷ (跛) ‘lame’?). Other examples: OC *prats (敗) ‘to defeat’ : *¿prats (*brats) (敗) ‘to be defeated’; *krujs (壞) ‘to destroy, ruin’ : *¿krujs (*grujs) (壞) ‘to be ruined’; *trjang÷ (長) ‘grow tall, increase; elder’ : *¿trjang (*drjang) (長) ‘long’; Bodo be≥ ‘to be straight’ : phe≥ ‘to make straight’ (TB *ble≥ ~ *ple≥; OC *bre≥ (平) ‘level’?). It seems there were intransitivizing (and nominalizing) prefixes in PTB and possibly STC, but these are represented by WT m- (e.g. development of Middle Chinese *d-, items that Baxter now reconstructs with *ml- clusters (e.g. Gong: *N-lj˙k (食), Baxter *mlj˙k (see Matisoff 1995, footnote 1; originally *Ljïk in Baxter 1992). 4 mkho-ba ‘desirable, to be wished for’ : 'kho-ba ‘to wish, to want’; Wolfenden 1929:27—notice the a-chung in the active form), and possibly *b- and/or *g-, e.g. T’rung r∑t ‘to tear down (a house)’ : br∑t ‘to collapse (of a house)’; la ‘to throw (down) : gl\a ‘to fall (down)’ (there is also a separate intransitivizing/stativizing ˙- prefix in T’rung as well: t—al ‘roll (vt.)’ > ˙t—al ‘roll (vi.)’ (LaPolla, 1995c; see also LaPolla 2000a)). These are independent of the voicing alternation. Quite a few scholars have assumed that the *s- causative suffix was responsible for all of the voicing distinctions now found in the family (e.g. Dai 2001), but, while this is true for some languages, particularly within Lolo-Burmese, the examples given in the discussion of this and the previous section show that the two are separate phenomena. *-t suffix.5 The *-t suffix most often had the function of transitivizing an intransitive verb, as in WT ¿bye-ba ‘open, separate’ (vi.) : ¿byed-pa ‘open, separate’ (vt.), Rawang ≥—∑ ‘weep’ : ≥∑t ‘mourn, cry for someone (vt.)’, but in some cases seems to nominalize intransitive verbs, as in WT ≥u-mo ‘weep’ : ≥ud-mo ‘a sob’ (see also Benedict 1991), and in still other cases seems not to have had any affect on the valency, e.g. WT g«ci-ba, g«cid ‘to urinate’; bka ‘word, speech’, skad ‘speech’ (for other examples and discussion, see Benedict 1972:98-102, Dai & Xu 1992, Michailovsky 1985, van Driem 1988, Jin 1998a). In Chinese we find pairs of related forms that differ only in the final consonant, but no clear derivational pattern can be determined, e.g. *nji (尼) ‘near, close’ : *njit (昵) ‘intimate, familiar; glue’ (from Pulleyblank 1972:11; this set is cognate with WT nye ‘near’, nyen ‘relative’). *-n suffix. The *-n suffix generally had a nominalizing function, e.g. WT rku ‘steal’: rkun-po ‘thief’; nye ‘near’, nyen ‘relative’, but in some cases seems to have had a collective sense (Benedict 1972:99ff), e.g. Proto-Tibeto-Burman (PTB) *r-mi ‘person’ : OC *mjin (民) ‘the people’. Pulleyblank (1991, 2000) also suggests that Proto-Sino-Tibetan (PST) had a morphological *-n suffix (as well as a *-t suffix), which could explain the correspondences among pairs such as *≥ja (語) ‘speak’ ± *≥jan (言) ‘say; word’ (see also Jin 1998a for more 5Although not often explicitly mentioned, except by Jin (1998a, b), the idea is that some of the finals we find on words are etymological, while others are due to affixation. Here we are only talking about affixation. 5 examples). Following Graham (1983), Pulleyblank argues that the *-n suffix marks durative or continuative aspect, and *-t marks punctual or perfective aspect. Norman (1988:86) argues that the forms *njan (然) and *w(r)jan (焉) are fusions of *nja (如) and *w(r)ja (于) and an *n- initial pronoun, possibly *nj˙j÷ (爾) or *njak (若). While a demonstrative may have been the ultimate origin of the *-n suffix, it seems this *-n could have been a more general suffix, and not the result of a chance fusion of isolated lexical items. Especially when we see the patterns of variants, it is hard not to assume there was some systematicity to it, e.g. *nja (如) ‘like’: *njan (然) ‘like this’ : *njak (若) ‘like; that’. There is also *÷a (烏) : *÷an (安) both ‘interrogative pronoun (‘where’)’, and possibly *÷ak (惡) ‘interrogative pronoun’.6 This is not to say there were no fusions. Some variation within word families may be due to a coalescence of two forms, as suggested for Tibetan by Walter Simon (1941, 1942, 1957). Simon’s idea was that many of the finals in Tibetan, such as -g, -n, -l, -r, -s were from the coalescence of two syllables, the second of which originally also had lexical content, such as -s < sa/so ‘place’. We find synchronic variation in Tibetan that points to this kind of development, such as da-ra ~ dar-ba ‘type of buttermilk’, |za-la ± |zal ‘clay’, bu-ga ± bug ‘hole’, lco-ga ± lcog ‘lark’, nya-ga ± nyag ‘steelyard’, and yi-ge ± yig ‘letter’. Norman (1988:85ff.) gives the following as examples of fusion words in OC: *tja (諸) from *tj˙ ÷ja (之於) ‘3rd person object pronoun’ + adposition ‘in, at, to’; *nj˙÷ (耳) from *nj˙ lj˙÷ (而已); *lja (歟) from *le hwa (也呼); *gap (盍) ‘negative question (‘why not’) particle’ from *gaj p˙ (何不). *-s suffix. The *-s suffix generally had a nominalizing function (Pulleyblank 1973b, Mei 1980, 1989), where the derived noun is the patient of the action represented by the verb, but also had a function that Mei (1980, 1989) and Schuessler (1985) have characterized as ‘change of direction’ or ‘inversion of attention flow’ respectively. Mei (1980) suggests these two functions derive from two different homophonous suffixes, which he equates with the Tibetan 6The usual reading of this last character when used as an interrogative pronoun is *÷a, but it is written using a character that is in other contexts pronounced *÷ak. If it is the same pronunciation as the one otherwise written (烏), it seems odd to use a character that normally is read with a stop final. 6 nominalizing and ablative suffixes respectively. In modern Chinese this suffix is now reflected in the ‘departing’ tone. In some cases the addition of the suffix resulted in the creation of a new Chinese character, but in many cases there are simply two pronunciations for the same character. E.g., OC *C-rjang (量) ‘measure’ : *C-rjangs ‘an amount’ :: WT ‘grang-ba ‘to number, to count’ : grangs ‘a number’; OC *tj˙k (織) ‘weave’ : *tj˙ks ‘thing woven’ :: WT ‘thag-pa ‘to weave’ : thags ‘texture, web’; OC *nup (納) ‘bring in’ : *nups (內) ‘inside’; *mre÷ (買) ‘buy’ : *mres (賣) ‘sell’; *dju÷ (受) ‘to receive’ : *djus (授) ‘to give’. *-j suffix. Matisoff (1989, 1995) discusses etyma that show palatal-final and non-palatal- final variants, and posits three different sources for variants with morphological differences: PST *s-way & *s-yay ‘go; motion away’, for transitive motion/motion away from the deictic center or emergent quality in stative verbs; PST *ya (& *za & *tsa & *dza) ‘child, son’ for a diminutive or affective sense; and PST *way & *ray for nominalization, subordination or other grammatical functions. The clearest examples are in the system of pronouns, where for the 1st person pronoun we get PTB *≥a : ≥aj :: OC *≥a (吾) : *≥aj (我). *----÷÷÷÷ suffix. Chinese seems to have had a glottal stop suffix which developed into the rising tone category, e.g. *trjang (張) ‘to make long, stretch’ : *trjang÷ (長) ‘grow tall, increase; elder’; *w˙k (或) ‘someone’ : *wj˙k÷ (有) ‘there is’; *kak (各) ‘each’ : *k(r/j)ak÷ (舉) ‘all’. In these last two examples I am assuming that the suffix caused the loss of the root final consonant, just as is assumed to have happened with the *-s suffix (Baxter 1992:323ff.; cf. also Bodman 1980:132), but this assumption is not widely accepted. An alternative possibility, discussed immediately below, is that there was a *-k suffix. Glottalized forms do appear in some TB languages (e.g. rGyalrong, T’rung), but it is not clear that there is any relation between the forms in these languages and those in Chinese. *-k suffix. There may have been a *-k suffix as well, as we find a large number of lexical items in both TB and Chinese that have open final and *-k final variants, e.g. TB *yu(w) : *yuk ‘descend’ (Benedict 1972:101); OC *m(r)ja (無) ‘there is not’ : *mak (莫) ‘no one’; *djuj (誰) ‘who’ : *djuk (孰) ‘which one’. This possibility was suggested by Pulleyblank (1972:13, 1973a:122) as an explanation for some of the pairs given above as examples of the glottal stop suffix: *wj˙÷ (有) ‘there is’ : *w˙k (或) ‘someone’; *k(r/j)a÷ (舉) ‘all; lift’ : *kak (各) 7 ‘each’. Pulleyblank only discusses this in relation to pronominal forms, and says the suffix marks a distributive sense. There is also the set *nja (如) ‘like’ : *njak (若) ‘like; that’ mentioned above. As the largest number of variants involve the difference between an open final and a *-k final (63 out of 99 rhymes in the Book of Poetry where the finals differed, as marked in Wang 1980; see LaPolla 1994c for discussion), it may be that there is more than one explanation; some velar stop finals may have dropped due to the influence of the glottal stop suffix, and some may have been the result of a *-k formative suffix (see also Jin 1998b).7 If PST had a particle similar to Tibetan -ga, which Das (1902:203) says ‘is sometimes used as an affixed particle of a word to complete it’, then this would be at least one explanation for the large number of *-Ø ~ *-k variants. It has long been known that within Sino-Tibetan we must deal with word families rather than isolated words (Karlgren 1933, 1956; Wolfenden 1936, 1937, 1939). Given what we now know about these derivational processes, we can see clearly how the word families are created. These forms seem to have formed paradigms (sets of choices), but of derivational possibilities rather than inflectional possibilities. Following are two examples (from Baxter 1992, p. 317 and p. 324 respectively; see also Mei 1989): *kat (割) ‘to injure, to harm’ (vt.) : *¿kat (*gat) (害) ‘to suffer harm or injury’ (vi.) : ¿kats (*gats) (害) ‘harm, injury’ (n.); *trjang (張) ‘to make long, stretch’ (vt.) : *trjang÷ (長) ‘grow tall, increase; elder’ (intransitive active verb) : *¿trjang (*drjang) (長) ‘long’ (stative verb). Aside from the suffixes mentioned above, Mei Tsu-lin (personal communication, November 1994) has suggested some of the frequent variations found in Chinese between homorganic stop and nasal final might be due to Chinese having had suffixes similar to WT -ma and -pa (which have both gender marking and formative functions). The nasal-initial suffix would cause a final stop to nasalize, while the stop-initial suffix would denasalize a final nasal. We see this sort of development with the diminutive in some dialects of Chinese, where the diminutive suffix 7Our answer to this question will affect our understanding of certain word families. For example, Pulleyblank (1991:30) suggests that *k(r/j)a÷ (舉)‘all; lift' is cognate with *kjat (揭) 'lift' (and so the latter would involve a *-t suffix). This set would stand only if we assume the root did not originally have a *-k final. 8 reduces to a nasal element (e.g. in Wenzhou, and some areas of Anhui, Zhejiang, Guangxi, and Guangdong), and in some cases nasalizes final stops, e.g. in Xinyi of Guangdong, the nasal suffix -n causes final -p, -t, and -k to become -m, -n, and -≥ respectively, as in ap££ ‘duck’ > am£∞ ‘duckling’. Certainly the use of reflexes of PTB *pa (and to a lesser extent *ma) as a gender marker and as a nominalizer (usually producing an agentive noun) is wide-spread throughout TB, though there is the possibility that many of these were independent parallel developments, such as in the case of the frequent development of diminutives from a word meaning ‘son’ or ‘child’ (Matisoff 1995), and of causatives from a word meaning ‘make’, ‘cause’, or ‘send’ (LaPolla 1994b). In Chinese the form *p(r)ja(÷) (夫/父/甫) was used as an extra-syllabic suffix for creating agentive nouns, just as in TB (e.g. *din p(r)ja (田夫) ‘farmer’), and this may be the cognate of PTB *-pa. In terms of clausal morphology, there may have been a clause-final question particle *la, as there is evidence for such a particle in several languages across the family: OC *lja (歟), Newai l— a, Burmese l»a, Meithei la (Matisoff 1995:73-74). As mentioned above, though, the Chinese form has been said to be a fusion form, from *le hwa (也呼) (Norman 1988:95). Unmarked clausal negation in PST took the form of a preverbal particle *ma-. For PTB we can also reconstruct a prohibitive (negative imperative) particle *ta- (see ex. (1) below for a Lahu example), but this is not found in Chinese. Chinese instead had two negative imperative particles *mja (毋), which was homophonous with the unmarked negator but written with a different character, and *mj˙t (or *mjut) (勿), which is often assumed to be due to fusion of the negative *mja with another particle (assumed to be the demonstrative pronoun *tj˙ (之)).8 Most languages in the family have not grammaticalized grammatical relations, but many have grammaticalized semantic marking.9 For detailed arguments against the existence of 8One might conjecture that the mysterious *-t final of the OC negative imperative *mj˙t is actually the prohibitive *ta, but we do not find *ma- and *ta- occurring together in TB. 9By grammatical relations is meant the grammatical singling out of a particular NP (the pivot of a construction) for special grammatical treatment in a construction, such that a restricted neutralization of semantic roles occurs (has conventionalized/grammaticalized) in that position in the