This paper presents potential cognates of the qusheng derivations in more conservative languages of the Trans-Himalayan family, in particular Rgyalrongic and Kiranti. It is suggested in particular that the valency-increasing uses of the qusheng could be related to the applicative -t suffix and that its valency-decreasing uses may be a trace of the reflexive -si suffix.
本文將去聲構擬為*-s或*-t，並提出去聲別義可能與漢藏語系存古語言中的不同後綴同源，去聲的使動和應動的用法可能對應基蘭梯語的應動-t後綴，而被動和反被動的功能則可能是反身後綴-si的痕跡。(This article is in English.)
* This article is the revised version of a talk presented at the conference ‘Recent Advances in Old Chinese Historical Phonology’. I gratefully acknowledge the grant ‘Beyond Boundaries: Religion, Region, Language and the State’ awarded by the erc. This research was funded by the HimalCo project (anr-12-CORP-0006) and is related to the research strand LR-4.11 “Automatic Paradigm Generation and Language Description” of the Labex efl (funded by the anr/cgi). The examples are taken from a corpus that is progressively being made available on the Pangloss archive (Michailovsky et al. 2014, http://lacito.vjf.cnrs.fr/pangloss/corpus/list_rsc.php?lg=Japhug). I would like to thank Nathan W. Hill, Laurent Sagart, Scott DeLancey, Bettina Zeisler as well as the anonymous reviewer, for providing detailed and helpful comments and corrections.
While qusheng 去聲 derivation is one of the most prominent trace of morphology in Old Chinese, it is probably also the least understood one, as it presents diverse and even contradictory functions, to the extent that Downer (1959:262), in his seminal article, argued that it was simply a way of creating new words, not a derivation with a well-defined grammatical function.1
Yet, we know thanks to the work of scholars such as Haudricourt (1954), Forrest (1960); Schuessler (1985) and Sagart (1999) that qusheng derivation comes (at least in part) from *-s suffixes. As -s suffixes with functions similar to those that have been reconstructed for Old Chinese are attested and even are still productive in more conservative languages of the Trans-Himalayan family, it is worthwhile to explore the exact opposite hypothesis to Downer’s ultrascepticism, namely that the vast array of functions of the *-s is due to the merger of many independent dental suffixes, and constitute indeed obscured traces of a former inflectional system.
In this article, I first present Downer’s work and adapt it to modern-day terminology. Second, I propose a new sound law for pre-Old Chinese. Third, drawing on first-hand data from Japhug (Gyalrongic), Khaling (Kiranti) as well as Tibetan, I show that seven out of Downer’s eight functions of the qusheng derivation have potential comparanda in the conservative languages of this family.
Old Chinese is given in Middle Chinese (in an ipa adaptation of Baxter’s 1992 system) rather than oc, because all modern Old Chinese reconstruction models agree on *-s as the origin of the qusheng, and the discussion in this paper is therefore independent of any particular system (except for the reconstruction of an applicative *-t suffix, see section 4.2).
2 Downer’s Eight Functions
Despite his suggestion that qusheng derivation was basically random, Downer (1959) provides a very useful classification of attested derivations, which are employed as the basic data for this study. It is possible that other functions for the qusheng can be found which are not included in Downer’s work, but this is beyond the scope of the present work.
Downer counts the following eight categories, here illustrated by one or two representative examples.
高 kaw ‘be high’ → 高 kaw H ‘height’
處 tɕʰjo X ‘be at’ → 處 tɕʰjo H ‘place’
家 kæ ‘family’ → 嫁 kæ H ‘marry’
枕 tɕim X ‘pillow’ → 枕 tɕim H ‘use as a pillow’
In addition to this causative function, the qusheng is also used with the tropative meaning ‘consider to be X’ (Jacques 2013), as the famous case of 好 xaw X ‘be good’ → 好 xaw H ‘like’.
飲 ʔim X ‘drink’ → 飲 ʔim H ‘give to drink’
買 mɛ X ‘buy’ → 賣 mɛ H ‘sell’
渴 kʰat ‘be thirsty’ → 愒 kʰaj H ‘long for’
5. ‘Restricted meaning’
However, some of the pairs in this category are better delegated to other categories; I will show in section 4.3 that at least some examples may have the same origin as passive and antipassive values of the qusheng derivation.
少 ɕjew X ‘be few’ → 少 ɕjew H ‘be young’
In addition, although Downer does not set up a special category, some of his examples rather attest to an antipassive (P-demoting) meaning, as shown by the following pair:
散 san X ‘scatter’ → 散 san H ‘be loose’
射 ʑjek ‘shoot at’ → 射 ʑjæ H ‘practise archery’
三 sam ‘three’ → 三 sam H ‘thrice’
8. Form in compounds
Downer notes many cases of qusheng derivation in the first or second element of compounds; however, most of his examples appear to be interpretable as special instances of denominal or deverbal derivations.
3 New Sound Law
In Old Chinese, no *-Ct final clusters exist in any system. Yet, the final *-s as reconstructed in all modern systems is much too common in compar- ison with languages of the family which allow final clusters. In Tibetan for instance, there are 7,398 words ending in -s in the Tibetan-Chinese dictionary (Zhang 1993) out of 53,922, presenting 14%. This figure includes many -s originating from *-t after grave codas). In contrast, in Baxter and Sagart’s (2014) online list of reconstructions, we find 1160 words with final -s out of 4,968 (23%). This raises the possibility that final *-s is not the only origin of the qusheng.
This sound change is similar to the rule undergone by the present tense -d suffix in Tibetan, which is realized as -s after final stops and -m, as in ɴdeb-s, btab ‘plant’ vs ɴde-d, bdas ‘chase’ (Coblin 1976:52–53).
(1)*-t → *-s /C_#
This hypothesis allows comparison of Chinese qusheng derivations with not only -s suffixes, but also -t suffixes, at least following grave final consonants. In addition, in the following comparisons, I assume that the *-s allomorph of *-t suffixes generated by this rule is then extended analogically to other contexts, in particular open syllables (this idea in particular is necessary to explain the *-s applicatives, cf section 4.2).
Very few languages in st clearly preserve final -s suffixes, and complex final clusters: only data from Tibetan, Kiranti, Gyalrongic, West-Himalayish and Dulong/Rawang can be used without reconstruction, and this article therefore focuses on these languages.4
Seven of Downer’s eight functions of the qusheng are discussed here (the causative and applicative derivations are discussed in the same subsection). In addition, the suggestion of a perfective value for the qusheng in Chinese and its possible cognates is briefly touched upon.
As early as Forrest (1960), scholars have noted that the nominalization function of the qusheng derivation could be compared with the -s nominalization found in Tibetan (Conrady 1896:43, Hill 2014c:624–5). The -s derives patient nouns, nouns of manner and nouns of characteristic, as illustrated by the following examples:
Patient: za ‘eat’ → zas ‘food’
Characteristic: zab.mo ‘deep’ → zabs ‘depth’.
Manner: ɴgro ‘go’ → ɴgros ‘gait’
Incidentally, the three same categories are attested in the West-Himalayish language Bunan (Widmer 2014:179–180).
The similarity with Old Chinese, where qusheng derivation is used exactly with the first two meanings, is striking (note for instance 深 ɕim ‘deep’ → 深 ɕim H ‘depth’).
Moreover, some nouns derived from verbs are cognate between Chinese and Tibetan, eg ɴtʰag, btags “weave” → tʰags “textile”, 織 tɕik ‘weave’ → 織 tɕi H ‘cloth’.
Nominalization by -s suffixation was already only very marginally productive in Old Tibetan. The only clear case of a neologism formed in historical times using the -s suffix is gzuŋs ‘dhâraṇî’ (a type of Buddhist magical formula). This word is derived from the root |zuŋ| of bzuŋ ‘seize’, by calque of ‘dhâraṇî-’, which derives from the root dhṛ- ‘seize, hold’ in Sanskrit. It actually uses a circumfix g-…-s, the prefixal element of which has cognates in Gyalrongic languages and beyond (Konnerth 2016, Jacques 2014d).
Outside of Tibetan and West-Himalayish, vestigial traces of *-s nominalization are found in Gyalrongic languages (Jacques 2003, Sun 2014).
The applicative and causative suffixes are most clearly preserved in Kiranti (Michailovsky 1985, Jacques 2015a), though isolated traces can be found in most branches of the family.
Limbu stands out among Kiranti languages with clearly distinct -s causative and -t applicative suffixes, as illustrated by the following triplet (Michailovsky 2002):
haːp- vi ‘weep’
haːps- vt ‘cause to weep’ (causative)
haːpt- vt ‘mourn for, weep for’ (applicative)
No clear example of causative or applicative -d or -s are found in Tibetan (Hill 2014c:630). In Gyalrongic, only isolated examples are found, mainly with motion verbs. For instance Japhug has ɣi < *wi ‘come’ → ɣɯt < *wit ‘bring’. However it should be noted that despite the dearth of examples, this pair actually has cognates in Kiranti (Khaling |pi| ‘come’ → |pit| ‘bring’), a fact which strongly supports the idea that this is an archaic feature, reconstructible to the common ancestor of these languages.
Cə.ɡˁ<r>aŋ ɢʷˁək s-rəʔ tə təʔɢʷˁək n<r>əl-t tə
The character 尼 is read here with the fanqie 女乙反, corresponding to a Middle Chinese pronunciation ɳit. This reading can be opposed to the usual pronunciation of this character ɳij ← *n<r>ɨl (fanqie 女夷切), cognate with Tibetan ɲal ‘lie down, sleep’ and Japhug nɯna ‘rest’,6 and suggests a reconstruction *n<r>əl-t in Old Chinese. This example shows that rule (1) does not apply after *-l.
‘A man’s advancement is effected, it may be, by others, and the stopping him is, it may be, from the efforts of others.’ (Legge)
Yet, examples of applicative qusheng are found in syllables with final *-l or even open syllables in Old Chinese. These examples can be accounted for by supposing that rule (1) generated *-s/-t allomorphy (depending on the final consonant of the verb root), and due to the existence of a causative *-s suffix, the *-s was reinterpreted as a mixed causative/applicative, and overgeneralized to contexts where rule (1) does not apply.
Cases of passive or antipassive values for the qusheng derivation can be accounted for as a trace of the sibilant reflexive suffix, which is still productive in Kiranti, Dulong/Rawang, Thangmi, West Himalayish and Kham (Bauman 1975:94, van Driem 1993:320, Watters 2002:240–7, LaPolla & Yang 2004, Turin 2012:372–6 and Widmer 2014:452, 466).7
In the following I use first-hand data on Khaling (Jacques et al. 2016) as representative of the functions of the -si reflexive/middle suffix in Kiranti languages.
In Khaling, the -si derivation has three common identifiable functions: reflexive (4), autobenefactive (5) and impersonal subject (6), the last of which resembles a passive or an anticausative when used with transitive verbs (use with intransitive verbs are also found, but less common). Given the fact that Khaling verbal morphology has a complex morphophonology that is not directly relevant to the present article, stem alternations will not be commented upon, and verbs are indicated according to their root, an abstract form from which the whole paradigm can be mechanically derived (Jacques et al. 2012, 2016).
1sg:erg meat cut-1sg →3‘I cut the meat.’
Reflexive: ‘I did not cut myself.’
(5) ʔʌm̄ sɵwêi-si.
Autobenefactive: ‘He cuts meat for himself.’
Impersonal subject: ‘The meat is cut (by someone)’ or ‘The meat cuts easily.’
It is this last use as impersonal subject that can be compared to the passive value of the qusheng derivation in Chinese, in such examples as 聞 mjun ‘hear, smell’ → 聞 mjun H ‘be heard, be smelt’.
(7) lokpei ghrɛ-md-u.
leech be.disgusted.by-1sg →3 ‘I am disgusted by leeches.’
be.disgusted.by-refl-1sg:S/P ‘I feel disgust.’
Finally, some of the pairs that Downer classified as examples of the ‘restricted meaning’ category appear to be compatible with the analysis of the qusheng as a fossilized reflexive/middle marker. In particular the semantics of the qusheng in 憶 ʔik ‘remember’ → 意 ʔi H ‘think’ is identical to that of -si in the pair |mimt| ‘think of, miss, remember’ → |mimt-si| ‘think (that)’. Even if the verb root is different, the meaning of the derivation is clearly identical in Chinese and Khaling.
If the comparison between Old Chinese *-s and Khaling -si is indeed valid, it suggests that syllabic suffixes may have lost their vowel in Chinese, leaving only the consonant as a coda. This raises the question whether other examples of the same type can be proposed, but I defer this question to future investigations.
1. A: ɴgag ‘be stopped, break off’
B: ɴgegs, bkag ‘hinder, prohibit’
C: kʰegs ‘be hindered, be prohibited’
2. A: gaŋ ‘fill intr.’
B: ɴgeŋs, bkaŋ ‘fill tr.’
C: kʰeŋs ‘be full’
3. A: gab ‘hide intr.’
B: ɴgebs, bkab ‘cover tr.’
C: kʰebs ‘be covered over’
4. A: grol ‘be free’
B: ɴgrol bkrol ‘liberate’
C: kʰrol ‘unravel’
5. A: dul ‘be tame’
B: ɴdul, btul ‘tame, subdue’
C: tʰul ‘be tame’
6. A: zug ‘pierce, penetrate’
B: ɴdzugs, btsugs ‘plant, establish, insert’
C: tsʰugs ‘go into, begin’
Jacques (2012) argues that verbs of type B are the base forms, and that the voicing in the Present and Future stems are due to the a nasal prefix, while type A verbs are derived by anticausative derivation (on which see Jacques 2015c, b). No explanation for type C verbs, however, has ever been proposed.
Yet, if we accept Jacques’s (2012) idea that the verb roots in these triplets have unvoiced initials, and that voicing in type A and in Present and Future stem is due to morphological alternations, a solution offers itself: type C verb could be remnants of *-si suffixed middle verbs. This hypothesis explains the three morphological features of type C verbs:
The unvoiced initial is simply the bare stem without any prefix.
The -s suffix is a direct segmental trace of the -si suffix; note that the loss of vowel is not unexpected, given the constraint of verbs stems to remain monosyllabic.
The vowel alternation a → e can be explain as Umlaut due to the lost *i in the suffix.
Thus, in this hypothesis, a verb form such as kʰegs ‘be hindered, be prohibited’ would originate from pre-Tibetan *kak-si. To confirm this idea, a detailed research on the use of type C verbs in Old Tibetan texts will be necessary, particularly how they semantically differ from type A verbs.
4.4 Denominal Verbalization
While the use of the qusheng to build verbs out of nouns in Chinese is quite common, there are barely any potential cognates in the rest of the family. In Gyalrongic languages, denominal morphology is exclusively prefixal (Jacques 2014b), and Tibetan adopts zero-derivation (adding the tam markers N- in the present and b-…-s in the past, Jacques 2014c:29).
Denominal -t derivation is most clearly attested in West-Himalayish. In Bunan, Widmer (2014:426) describes the following examples of verbalizing -t suffix:
ken ‘birth’ → ken-t ‘give birth (animal)’
kur ‘load (n)’ → kur-t ‘carry a load’
len ‘work (n)’ → len-t ‘work (v)’
sur ‘weed (n)’ → sur-t ‘weed (v)’
*ti ‘water’ (lost in Bunan but attested in closely related languages) → ti-t ‘irrigate’
A few examples of denominal -t are also found in various Kiranti languages:
Limbu thiːn and Bantawa din ‘egg’ → |thiːnt-| ‘lay an egg’ and |dint| ‘lay eggs’ respectively.
The Limbu transitive verb |khaːnt| ‘wound’ <*ʔwɑːr-t derives with a denominal -t suffix from a noun not attested in Limbu, but found in Wambule ɓari ‘wound (n)’ and Khaling koɔr̄ ‘wound (n)’
Limbu |sokt| ‘to aim, to point’, Khaling |tsukt| ‘point (with a finger)’, Limbu cok ‘toe, finger’ (Limbu c comes however from proto-Kiranti *dz, implying that the noun underwent voicing of the initial).
The Khaling |kakt| ‘hoe’ corresponds to the Japhug noun qaʁ ‘hoe’ outside of Kiranti (perhaps also to Limbu kaːŋ ‘hoe’, with unexplained voicing of initial as in cok ‘toe, finger’).
The use of the qusheng to derive adverbs from verbs or nouns may originate from a locative suffix *-s that is well attested in Tibetan and Gyalrongic languages.
In Tibetan, this suffix is not attested as such, but exists as the -s element in several case markers, including the ergative -s/-kʲis, the comparative -bas, the ablative -las and the elative -nas, the last two of which are compound cases combining the -s element with the allative -la and the locative -na (Konow (1909), DeLancey 1982, Zeisler 2011:282, Hill 2012), which are all used to build various types of subordinate clauses (Tournadre 2010). It is also found in some adverbs, such as jas ‘from above’.8
The reanalysis of a clausal linker as an adverbializer is straightforward, and thus the examples of adverbialization collected by Downer might be cases of a former locative *-s that underwent the same extension as Japhug, and further grammaticalized as a derivational morpheme.
(9) kɯki sthɯci ji-kha
1pl.poss-house neg:sens-be.good sens-be.small
sens-be.narrow loc 2sg
erg imp-go lnk irr-pfv-2-search lnk
Our house is so bad, so small, so narrow, go there and look for (the fish to ask him for something). (140430 yufu he ta de qizi, 55)
4.6 Second Member of Compounds
In Old Tibetan, Uebach & Zeisler (2008) have brought to light examples of a -s suffix in the second member of nominal compounds, as for instance lag ‘arm, hand’ + riŋ ‘long’ → lag.riŋs ‘long arms’. These examples may be analyzable as particular instances of the nominalization suffix treated in section 4.1.
Likewise, in Old Chinese, compounds with qusheng in the first or second element, are best analyzed as cases of nominalization, and these hardly constitute a distinct derivational category.
Some scholars (in particular Jīn 2006) have claimed that some examples of *-s suffixes can be accounted for as traces of the *-s perfective suffix found in Tibetan (bʲed, bʲas ‘do’) or in Gyalrong languages (Japhug -t or -z 1/2sg →3 perfective suffix, Situ third person intransitive perfective -s).9
This idea is tantalizing, as it would be the only inflectional use of a *-s suffix in Old Chinese, as opposed to the previous cases, which all concern derivational functions.
However, this idea is hardly compelling, as alternative readings are not listed systematically enough in the Jingdian shiwen to allow an easy reverification. Judgement on this matter must be deferred until the Chinese evidence has been clearly sifted through on the basis of an exhaustive analysis of available sources.
This paper presents all potential comparative evidence known to me in Trans-Himalayan languages of suffixes comparable to the qusheng derivation. While not all comparisons are equally compelling, postulating several unrelated origins for the qusheng derivation solves two unrelated issues, namely (1) the excess of *-s in all reconstructions of Old Chinese is here explained as due to the fact that *-s comes from the merger of *-s, *-t and *-sV suffixes, and perhaps also from other codas such as primary *-h. (2) the numerous and contradictory functions attributed to the qusheng derivation.
Future progress in the study of Old Chinese morphology can only result from collaboration between philologists familiar with the texts and the alternative readings therein, and linguists with a first-hand knowledge of conservative Trans-Himalayan languages where morphology is still visible without any need for reconstruction.
Schuessler Axel . 2007. abc Etymological Dictionary of Old Chinese. Honolulu: University of Hawai’i Press.
Turin Mark . 2012. A Grammar of the Thangmi Language. Leiden: Brill.Uebach Helga & Bettina Zeisler. 2008. rje-blas pha-los and other compounds with suffix -so in Old Tibetan texts. In Chomolangma Demawend und Kasbek. Festschrift für Roland Bielmeier zu seinem 65. Geburtstag Vol. 1 eds. Brigitte Huber Marianne Volkart und Paul Widmer 309–334. Halle: International Institute for Tibetan and Buddhist Studies.
Zeisler Bettina . 2004. Relative Tense and Aspectual Values in Tibetan Languages: A Comparative Study . Berlin, New York: Mouton de Gruyter.
‘The present writer holds the opinion that with our present knowledge of Classical Chinese, it is better to regard chiuhsheng derivation not as a remnant of a former inflectional system of the Indo-European type, but simply as a system of derivation and nothing more. When new words were needed, they were created by pronouncing the basic word in the chiuhsheng. The grammatical regularity found in many cases would then be in a way fortuitous, being the result not of grammatical inflection, but of the need to create new words.’
These clusters are not word-final in these languages, as they always surface before inflectional suffixes.
A similar idea was suggested by Schuessler (2007:42).
Phonologically less conservative languages may provide crucial confirmatory evidence, but at the present stage it would be premature to investigate languages where final -s is only accessible through the comparative method.
I reconstruct *-l where Baxter & Sagart (2014) have *-r in Old Chinese, as (1) this is a way to avoid syllables with medial *-r- and final *-r, a structure disallowed in monomorphemic words in both Tibetic and Gyalrongic languages (Jacques 2004) and (2) this coda corresponds to -l in Tibetan in many examples, while correspondences with -r are few and less convincing. (Hill 2014b:101–2).
Loss of final *-l after *a is a common innovation of Burmo-Gyalrongic languages (Jacques & Michaud 2011).
Gyalrongic languages have innovated reflexive prefixes (Jacques 2010, Lai 2013:89), which replaced the -si derivation.
There is some evidence that this -s element had an alternative form -se in Old Tibetan and some modern Tibetan languages (Zeisler 2011:280–284), but I defer this question to further research.
See Huáng (1997) for other potential examples, though many examples in Huang’s study come from languages that do not preserve final *-s and hence cannot be cognate to the Tibetan or Gyalrong suffix, unless degrammaticalization took place.