Tangut, Gyalrongic, Kiranti and the nature of person indexation in Sino-Tibetan/ Trans-Himalayan* Guillaume Jacques May 6, 2016 Published ahead of print in Linguistic Vanguard: http://www.degruyter.com/view/j/lingvan.ahead-of-print/lingvan-2015-0033/ lingvan-2015-0033.xml Abstract: The diachronic analysis of person indexation systems in SinoTibetan (Trans-Himalayan) languages is currently a topical issue. Factual errors have occasionally crept in, detracting somewhat from the quality of the linguistic discussion about these systems. Evidence from Tangut, Gyalrongic and Kiranti is so central to the debates that it appeared useful to provide a few clarifications about their person indexation systems, adducing evidence from a body of texts that has been considerably enriched in the past decade. The main points made in this paper can be summarized as follows. First, the view that personal affixes derive diachronically from pronouns is by no means as self-evident as it may seem. Second, person indexation in Tangut, the oldest Trans-Himalayan language with person indexation, is not optional, as has sometimes been stated in the literature. Third, person indexation in Gyalrongic and Kiranti is sensitive to grammatical relations, a finding which calls into question its analysis as marking speech act participant involvement. Keywords: Gyalrongic, Kiranti, Tangut, Person indexation, Agreement, Grammaticalization, Sino-Tibetan * I would like to thank Scott DeLancey, Stephen Dodson, Nathan Hill, Boyd Michailovsky, Alexis Michaud, Jackson T.S. Sun and two anonymous reviewers for useful comments on previous versions of this paper. I am responsible for any error that remain in this work. The Japhug examples are taken from a corpus that is progressively being made available in the Pangloss Collection (Michailovsky et al. 2014). 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). Glosses follow the Leipzig glosses rules, to which the following are added: antierg antiergative, dir direct, hon honorific, irr irrealis, inv inverse, pot potentialis, sens sensory. 1 1 Introduction Sino-Tibetan (Trans-Himalayan) is probably one of the typologically most diverse language families in the world: it encompasses prototypically isolating languages (such as Lolo-Burmese or Chinese) as well as polysynthetic languages (such as Gyalrongic and Kiranti). There is intense debate as to how to interpret this diversity. A line of argument is that the complex verbal morphology of Gyalrongic and Kiranti should be reconstructed back (at least in part) to proto-Trans-Himalayan (Bauman 1975, DeLancey 1989, 2010, van Driem 1993, Jacques 2012). Another line of argument (LaPolla 1992, LaPolla 2003, LaPolla 2012, Zeisler 2015) is that little verbal morphology can be reconstructed back to proto-Trans-Himalayan, and that the complex person indexation systems of Gyalrongic and Kiranti were grammaticalized from the accretion of pronouns. This view had already been expressed in Hodgson’s pioneering research on these languages in the mid-nineteenth century, leading him to coin the term ‘pronominalizing languages’ (see for instance Hodgson 1857-8, still in use today). Factual errors have occasionally crept into diachronic analyses of person indexation systems in Trans-Himalayan languages, detracting somewhat from the quality of the linguistic discussion. Evidence from Tangut, Gyalrongic and Kiranti is so central to the debates that it appeared useful to provide a few clarifications about their person indexation systems, adducing evidence from a body of texts that has been considerably enriched in the past decade. The main points made in this paper can be summarized as follows. First, I address the issue of the term ‘pronominalizing language’, taking the strong stand that this term were best avoided altogether, as it tends to convey an overly simplistic and linear view of processes of grammaticalization. Second, I discuss the person indexation system of Tangut, and show that it is neither optional nor transparent, as had been stated in various publications. Third, I present counterarguments to the proposal that person indexation in Gyalrongic and Kiranti languages is based on ‘person involvement’ rather than syntactic relations. 2 Pronominalization: a descriptive label, or an explanation? ‘Pronominalization’ was used by Hodgson (1857-8) as a label to describe the presence of person indexation in some Trans-Himalayan languages, such as the Kiranti languages, of which he had first-hand fieldwork experience. This pioneering insight attracted attention to an important characteristic of these languages. On the other hand, the analysis encapsulated in this term (that person indexation is the result of the accretion of pronouns on 2 the verb complex) may need to be re-examined in greater detail, making use of newly available evidence. To preview the result of the re-analysis set out below, it would seem best to recognize that the term ‘pronominalized’ is now outdated and potentially misleading: it should be retired, for at least three reasons. First, whatever the actual antiquity of the person indexation systems in Trans-Himalayan, it is clear that a sizeable proportion of person markers on verbs have been grammaticalized from sources other than pronouns, in particular nominalized forms (see for instance Jacques to appear on the origin of the portmanteau 2→1 and 1→2 prefixes in Gyalrong languages). Second, the situation found in language groups such as Kuki-Chin, where productive person markers are identical to (and probably derive from) possessive prefixes or pronouns, is markedly different from that found in Kiranti and Gyalrong languages. Specifically, there is no compelling evidence that dental second person prefixes in Kiranti and Gyalrong derive from pronouns (Jacques 2012, DeLancey 2010, 2011 and 2014). Finally, resemblances between pronouns, possessive markers and person indexation markers do not necessarily imply that the latter must be derived from the former. Cases of degrammaticalization of person markers into pronouns are attested, but admittedly very rare (Norde 2009, Hyman 2011); on the other hand, it is not that uncommon cross-linguistically for pronouns to be built from a conjugated verbal stem, or a possessed nominal stem. For instance, in Ainu, the 1sg pronoun kuani originates diachronically in the nominalization of the 1sg form of the existential copula (ku-an-i 1sgexist-nmlz, Shibatani 1990: 31). Likewise, in Lakota the pronouns 1sg miyé, 2sg niyé and 3sg iyé are in fact conjugated verb forms meaning ‘it is me’, ‘it is you’, ‘it is him’, respectively (Boas & Deloria 1941, Ullrich 2008: 707;754). Pronouns derived from possessed nouns are found in Algonquian (as in Ojibwe n-iin ‘I’, g-iin ‘you’ and w-iin ‘he’, see Valentine 2001). In the TransHimalayan family, Japhug is a case in point. As can be seen in Table 1 below, pronouns are built by combining possessive prefixes with the root -ʑo ‘oneself’ (the same is true of other Gyalrongic languages, such as Tshobdun, see Sun 1998: 113). 3 Table 1: Pronouns and possessive prefixes in Japhug Free pronoun Prefix Person a-ʑo nɤ-ʑo ɯ-ʑo anɤɯ- 1sg 2sg 3sg tɕi-ʑo ndʑi-ʑo tɕindʑi- 1du 2du i-ʑo nɯ-ʑo inɯ- 1pl 2pl tɯ-ʑo tɯ- generic Grammaticalization pathways exist between person indexation markers and pronouns, but (to labour the point) these pathways are not unidirectional as the use of a term such as ‘pronominalizing languages’ can suggest. 3 Person indexation in Tangut The role of Tangut data in the debate of the antiquity of person indexation systems in Trans-Himalayan is critical, as of all the languages with ancient attestation (Tangut is attested from the 11th to the 16th centuries), Tangut is the only one with a full-fledged person indexation system. LaPolla (1992) argues that, in Tangut, (i) there is a one-to-one relationship between pronouns and suffixes, (ii) person agreement is optional, and hence (iii) person indexation suffixes have been recently grammaticalized from pronouns. These three points are re-examined here on the basis of a corpus that includes major narrative texts (see Jacques 2014b: 8-9), in particular the collection of short stories called The Grove of Categories, from which most examples in this section are drawn.1 3.1 Person indexation suffixes As pointed out by Kepping (1994), while agreement suffixes in Tangut do present resemblances with pronouns, as shown in Table 2, pronouns and agreement markers cannot be equated at a synchronic level. 1 This corpus, transcribed from Kepping (1983) and Shǐ et al. (1990) is available as a supplementary file to this article, to facilitate verification. 4 Table 2: Pronouns and person suffixes in Tangut (Kepping 1975, 1985) Pronoun 噴 2098 烱 3926 玻 4028 Suffix ŋa² 1sg 噴 nja² 2sg 肅 nji² 2sg honorific or 2pl 蜉 2098 4601 4884 ŋa² 1sg nja² 2sg nji² 1pl and 2pl The first person singular pronoun 噴 ŋa² is given the same reading as the 1sg suffix,2 another 1sg pronoun 禽 mjo² is also widely attested. In books 3 to 6 of the Grove of Categories, there are 31 occurrences of the pronoun 噴 ŋa², 60 occurrences of 噴 ŋa² as a 1sg suffix, and 20 occurrences of 禽 mjo². This latter pronoun always triggers agreement with the 噴 ŋa² suffix when occurring as a core argument, as in example 1. Thus, despite the homophony and homography of the 1sg pronoun and the 1sg suffix,3 it is a fact of the synchronic grammar of Tangut that the two are distinct: otherwise, we would not expect agreement between 禽 mjo² and 噴 ŋa². (1) 禽 0261 mjo² 1sg 假 杉 幟 款 視 凾 噴 2541 1139 3104 0046 0749 2620 2098 dzjwo² .jij¹ man ljij²-phji¹ njwi²-ŋa¹ ? antierg ghost see[A]-cause[A] can-1sg ‘I can make people see ghosts.’ (The Grove of Categories, 05.21B.4) It is however with the honorific pronoun 玻 nji² and the SAP plural suffix 蜉 nji² that the difference is most telling. 玻 nji² is singular; when serving as core argument, the verb takes the 2sg 肅 -nja² suffix, as in 2. No example of 玻 nji² used with 蜉 nji² has been found in the corpus under study. (2) 玻 頃 悃 厶 肅 4028 0582 3211 2699 4601 nji² thjij²sjo² nwə¹-nja² 2sg:hon how know-2sg ‘How do you know?’ (The Grove of Categories, 04.04A.6) The suffix 蜉 nji², on the other hand, appears with first or second person plural, as in 3. 2 It should be emphasized here that both the 1sg pronoun and the suffix are not directly cognate with forms such as Tibetan ŋa, as pre-Tangut *-(j)a regularly becomes -e or -ji: see Jacques (2014b). 3 On this topic, it should be noted that it is common in Tangut to write unrelated but homophonous (or perhaps, near homophonous) morphemes with the same character, see several examples in Jacques (2011). 5 (3) 昧 赴 托 契 杉 掴 驚 攻 蜉 2248 2065 1413 5970 1139 1567 0239 0508 4884 gjɨ²mji² 1pi tej¹pie¹ Taibo .jij¹ gji²lhjɨ¹ ŋwu²-nji² gen grandchildren be-1/2pl ‘We are the descendants of Taibo.’ (The Grove of Categories, 04.33A. 4) The hypothesis that the suffix 蜉 nji² is historically derived from the ancestor of the pronoun 玻 nji² further suggests that the singular honorific 2sg originates from a 2pl (a reasonable assumption), and that in verbal indexation the first vs second person plural distinction became neutralized at the expense of the first person, a rare phenomenon, but not altogether unthinkable.4 The exact mechanism for this neutralization is an interesting question; in an initial stage, the neutralization may have occurred for the first plural inclusive, before affecting the first plural exclusive. The resemblance between person indexation suffixes and some of the pronouns in Tangut (Table 2) calls for an explanation. The most likely one is that these suffixes were ultimately grammaticalized from pronouns, or at least remade after the pronouns. On the other hand, in light of the functional difference between pronouns and the corresponding suffixes brought out in the above analyses, this grammaticalization process must be of some antiquity. It does not at all appear plausible that it occurred shortly before the time when Tangut was put to writing. 3.2 Stem alternations The most compelling piece of evidence that person indexation in Tangut is not recent does not come from the person indexation suffixes, however. Gong (2001), Jacques (2009) and Jacques (2014b) have brought out a closed class of irregular verbs with two stems, called A and B. Stem B is restricted to 2sg→3, 1sg→3 forms (as well as the reflexive 1sg and 2sg forms), as illustrated by examples 4 and 5.5 (4) 超 1542 ku¹ 朞 3508 bji² 岐 0100 lew¹ therefore subject one 嗹 孰 2798 2987 .jir² lhjɨ ̣¹ 沖 5481 bo² 羈 噴 4547 2098 dzjo¹-ŋa² hundred strike staff eat[B]-1sg ‘Then I, your subject, will take (eat) a hundred blows.’ (The Grove of Categories 06.13A.5) 4 For a discussion concerning the directionality of analogy in person indexation paradigms, see Jacques (2016). 5 The presence of stem A, and the absence of 1sg indexation suffix in example 5 is explained in section 3.3. 6 (5) 廸 禽 朝 杉 嫉 拶 款 椛 帚 3133 0261 1531 1139 0795 0676 0046 5643 3092 sjij¹ mjo² today 1sg gja¹ .jij¹ rjɨr²-.wjij¹ army antierg pfv-leave 夛 帚 椛 楕 蔟 噴 2912 3092 5643 1374 4803 2098 mjɨ¹-tɕhjɨ¹-lji²-ŋa² lhjwo¹-djij² ljij² mjɨ¹djij² see[A] but come.back-dur neg-pot-see[B]-1sg ‘Today I see the army leave, but I will not see it return.’ (The Grove of Categories, 3.16B.6-7) In other configurations with the suffixes 噴 ŋa² and 烱 nja², including 1→2sg, 2→1sg, 3→2sg and 3→1sg, stem A is used, as in 6. It is also the form which appears with SAP plural or third person arguments. (6) 卵 喋 跚 冂 纉 噴 2447 1519 5165 2590 4517 2098 ljo² ɣu¹twụ¹ brother instead wjɨ²-dzji¹-ŋa² imp-eat[A]-1sg ‘Eat me instead of my brother!’ (Newly Collected Biographies of Affection and Filial Piety 17.7, Jacques 2007: 55-6) (7) 艫 朝 4689 1531 .jwar¹ gja¹ Yue 修 0866 ɣu¹ army Wu 杉 1139 .jij¹ 融 2393 ljiij² 旒 3456 lja¹ 賛 0705 zji ̣j¹ 防 款 噴 2219 0046 2098 kjij¹-ljij²-ŋa² antierg destroy come time irr-see[A]-1sg ‘When the Yue army will come to destroy Wu, it will see me.’ (The Grove of Categories, 03.21B.4-5) Table 3 summarizes the distribution of stems and suffixes in the forms of the transitive paradigm attested in the corpus. Unattested forms are indicated by a question mark. Regardless of the historical interpretation of these alternations (on which see Jacques 2009 and Jacques 2014b), they are synchronically non-productive, securely attested for less than fifty common verbs, and stem A is not predictable from stem B or vice-versa. Even if the suffixes were recently grammaticalized (i.e. only a few centuries before the Tangut script was created), it is unclear how irregular stem alternations could come into existence in such a short time span, and what their historical origin may be. 7 Table 3: Attested forms of the ditransitive paradigm in Tangut 1sg 2sg 1/2pl 3 1sg 2sg 1/2pl 3 ? A-ŋa² A-ŋa² A-ŋa² A-nja² B-nja² ? A-nja² ? A-nji² ? ? B-ŋa² B-nja² A-nji² A Until recently, Tangut texts were not easily accessible for independent study, and stem alternations had not yet been fully described. Although Nishida (1975) first suggested their existence, this discovery was not taken into account by other scholars (even Kepping 1985 and van Driem 1991), and it was not until Gong (2001) that it became widely known. What we need most at the present moment would be an investigation of stem alternations in the whole Tangut corpus, based on an exhaustive database including all verbal forms. In particular, it seems that in addition to the A/B stem alternation described above, Tangut also had other types of stem alternations, whose morphosyntactic functions are still unclear (see Jacques 2014b). 3.3 Is person marking in Tangut optional? As part of an argument that Tangut person indexation seems recent, LaPolla (1992) argues that person indexation in Tangut is optional, and interprets this as a sign of it not being fully grammaticalized. He cites Ahrens (1990) (whose corpus was based on the Grove of Categories, like the present study), according to whom: • ‘Verb agreement only occurs in quoted speech.’ • ‘Agreement is usually with the A and S argument, not with the P argument.’ • ‘When there are two SAPs involved in a clause, agreement is not necessarily with the P argument.’ Generalizations n°2 and n°3 do not match my observations about Tangut texts. Table 4 lists the number of attestations of the suffix 噴 -ŋa² in chapters 3 to 6 of the Grove of Categories. Leaving aside intransitive verbs, which are not relevant to the present debate, we see that -ŋa² appears in 1sg→3, 3→1sg and 2→1sg (in conformity with Table 3), never in 1sg→2 forms. Stem B appears in 1sg→3 forms, and stem A in all other transitive forms. Agreement with the P argument is the rule in 3→SAP and SAP→SAP configurations; the only case where agreement with the A occurs is in SAP→3 forms. But since SAP→3 are by far more common in texts than 8 all SAP→SAP and 3→SAP forms put together, the absolute number of attestations of P argument indexation (here 10 out of 60) is lower than that of A argument indexation (36); this fact may explain how Ahrens came to propose generalization n°2. When both arguments are SAPs, the indexation suffix always refer to the P, as is the case in all Gyalrongic languages (Sun 2003, Jacques 2010, Gong 2014, Lai 2015). Table 4: Distribution of the 1sg suffix -ŋa² in chapters 3-6 of the Grove of Categories Form Nb of attestations 1sg (intransitive verb) 1sg→3 (no stem alternation) 1sg→3 (stem B) 3→1sg (no stem alternation) 3→1sg (stem A) 2→1sg (no stem alternation) 2→1sg (stem A) 1sg possessor 1sg→2 13 29 7 2 2 1 5 1 0 Total 60 As for claim n°1, third person is zero marked in Tangut, so that a verb whose core arguments are all third person cannot receive person marking. A slight bias here is that the Grove of Categories is a set of short stories translated from Chinese from a third-person point of view, and clauses other than those in quoted speech only have third person arguments. In the absence of a first-person narrator or first-person comments on the text, it is no surprise that first or second person will only occur in reported speech; this would be true in any language. A translation of this text in any language where third persons are unmarked for person would warrant the same generalization. Tangut texts contain examples where verbs with a clear SAP core argument show neither indexation suffixes nor stem alternation. It appears as a useful service to the research community to discuss these examples, which (in this author’s view) put to rest the hypothesis that Tangut person indexation is optional. Example 8 is a typical example of absence of person indexation on some verbs in Tangut. Although the S/A of all verbs in this paragraph is 2sg, the verbs 豕 .wji¹, 視 phji¹ and 琺 (pronunciation unknown) have neither the suffix 肅 nja² nor stem B as would be expected (these are indicated in red below). Only the last verbal form 笄佯肅瀘 dja²-rjɨr²-nja²-sji² has person indexation (in blue). 9 (8) 玻 粉 啼 迹 湫 豕 蚌 名 4028 2104 2780 5267 3831 5113 4861 2302 - 粐 仰 5880 4408 ɕji¹ kiow²ljɨj¹ljij² 臧 視 佶 台 梔 空 慟 嫉 豕 4663 749 2503 1402 3567 289 3266 795 5113 nji² 2sg first .wji¹ .we²-dzju² rjɨr²-.wji¹ kụ¹ xũ¹luu² 肅 瀘 棘 汗 4601 3916 3583 5688 extinguish-cause[A] after Hongnong city-lord 池 琺 仰 笄 佯 1477 4039 5880 4342 2511 le² 0 məə¹ make[A] time wind-instr fire Jiangling.Ling lha¹-phji¹ ljɨ¹-ŋwu² zjọ² ŋwu² dja²-rjɨr²-nja²-sji² tja¹ pfv-make[A] .wa² 偶 290 sju² tiger ride instr pfv-leave-2sg-ifr top what be.like ‘First, when you were Jiangling Ling, you extinguished the fire with the wind, later, when you became Taishou of Hongnong, you went away riding a tiger. How did (you do that)?’ (The Grove of Categories, 04.13B.4) On superficial examination, this example might be taken as evidence that Tangut person indexation was optional. At this juncture, it is crucial to read Tangut texts in light of this language’s modern relatives, in particular those of the Gyalrongic languages, which can be observed in vivo.6 All Gyalrongic languages described up to now have converbial forms that can carry some TAM markers, but no person indexation or evidential markers. In Japhug, these forms are marked by the nominalization prefixes kɤ- or sɤ-, as in example 9 (see Jacques 2014a). (9) tɕe ɯ-ŋgɯ nɯ tɕu paʁndza ɲɤ-raʁ tɕe, tɕendɤre lnk 3sg-inside dem loc pig.fodder evd-be.stuck lnk lnk [<dian> <guan> mɤ-kɤ-βzu] kɯ mɤ-kɤ-pa kɯ electricity turn.off neg-inf-make erg neg-inf-close erg ɯ-jaʁ lo-tsɯm 3sg.poss-hand evd:upstream-take.away ‘Some pig fodder got stuck inside (the machine) she reached her hand into it without turning it off,’ (Relatives, 372-3) In Stau, where nominalization prefixes have disappeared (only leaving fossilized traces), the converbial forms are marked with suffixes such as -dʑə as in example 10. Note that the verb form tə-pʰji-dʑə pfv-run.away-conv(1) does not take the first person marker (otherwise tə-pʰjã would be expected, 6 The Gyalrongic group includes three branches (Sun 2000a,b), Core Gyalrong (comprising Japhug, Tshobdun, Zbu and Situ), Khroskyabs and Horpa (a subgroup which includes Stau). There is evidence of lexical common innovations linking this group with Tangut and other languages of the area (Jacques 2014b), including Naish and Lolo-Burmese, which have lost person indexation (Jacques & Michaud 2011). 10 see Jacques et al. 2014) (2) keeps the perfective directional prefix tə-, though cases without directional prefixes are also attested in converbial forms. (10) ŋa tə-pʰji-dʑə arədadu ɞrɞ la 1sg pfv-run.away-conv up ʁə rə-ɕã up pass head pfv:up-go:1pl ‘I ran away and went up toward the mountain pass up there.’ (The hybrid yak, 86) Returning to example 8, we see that of the four verb forms without person marking, one takes a directional prefix, one is followed by the instrumental suffix 仰 ŋwu² (actually, a probable cognate of the ergative -w in Stau), and one is followed by the relator noun 蚌 zjọ² ‘time; when’. What we have here is a converbial chain, with four non-finite verbs lacking person markers but taking various TAM and subordination markers (although zero marked converbs are also possible), and the last verb, which has all the person and TAME markers, is the only finite one in the sentence. These observations warrant the claim that Tangut person indexation (by which I mean suffixes AND stem alternation) is not optional. This claim will be falsified if counterexamples can be found, with the following characteristics: • The sentence is complete (not from a fragment missing characters, or a sentence at the end of an isolated page). • The sentence has a SAP core argument marked by an overt pronoun. • The sentence is at the end of a quotation, and cannot be interpreted as a converb. 4 Person indexation in Gyalrong and Kiranti Zeisler (2015: 53) states that ‘It should also be noted that several pronominalising Tibeto-Burman languages do not mark syntactic relations or semantic roles but simply person or, more precisely, speech act participant (1P and/ or 2P) involvement: Tangut, Gyarong, Nocte, Muya, Dulong, Kiranti, Hayu.’ The notion of SAP involvement refers to LaPolla’s (1992: 308) statement that ‘agreement in Gyarong is with 1st person any time a 1st person is involved, regardless of its semantic or syntactic function’. In other words, a SAP referent, independently of its syntactic function (argument, adjunct or possessor of an argument), would trigger agreement. In this section, I adduce fresh first-hand data from fieldwork on Gyalrongic and Kiranti languages that strongly suggest that the above view does not reflect the full story, at least as far as Gyalrongic and Kiranti languages are concerned. 11 It is a fact that some person affixes in Gyalrongic and Kiranti languages have the same form whether they refer to S, A or P. For instance, in Japhug the same 1sg -a suffix appears to mark S (11), A (12) or P (13). (11) pɯ-nɯʑɯβ-a (12) pɯ-mto-t-a (13) pfv-sleep-1sg ‘I slept’. pfv-see-pst:tr-1sg ‘I saw him’. pɯ́ -wɣ-mto-a pfv-inv-see-1sg ‘He saw me’. However, this does not warrant the general conclusion that the system of person indexation as a whole does not mark syntactic relations. If this were the case, none of the languages with direct-inverse systems (including Algonquian, Mapuche and Movima) would mark syntactic relations. In Ojibwe, in the independent order, for instance, the same prefix ni- occurs in exactly the same contexts as the suffix -a in Japhug in examples (14) to (16). (14) ni-nibaa 1-sleep ‘I sleep.’ (15) ni-waabam-aa 1-see-dir ‘I see him.’ (16) ni-waabam-ig 1-see-inv ‘He sees me.’ There is clear evidence that Ojibwe morphosyntax is in fact very sensitive to grammatical relations. First, in the conjunct order, the 1sg is marked by entirely distinct suffixes (-yaan, -ag and -id respectively for 1sg.intr, 1sg→3 and 3→1sg, see Valentine 2001: 295). Second, strict accusative and ergative syntatic pivots are attested in Ojibwe (Rhodes 1994, Zúñiga 2006: 119-126). No author has ever concluded from examples such as (14) to (16) that agreement with 1sg occurs in Ojibwe regardless of syntactic function. When analyzing complex polypersonal systems, it is important not to focus on individual affixes, whose distribution may in some cases not even be 12 describable in functional terms, but rather to study the whole set of affixes and stem alternations as a complete system. In the following, I present evidence showing that person indexation systems in Gyalrongic and Kiranti, as in Ojibwe, do mark grammatical relations, and not simply SAP involvement. First, indexation systems in these languages present numerous affixes which are restricted to S, A or P, or portmanteau markers used in specific configurations. Second, even in the case of most ditransitive verbs in Japhug, the recipient (even if SAP) is not indexed on the verb, and instead the theme (rarely human) is indexed. Third, agreement of the verb with possessors is not attested, at least in Japhug and Khaling. 4.1 Unambiguous marking of S, A and P Several affixes or morphological processes in Gyalrongic and Kiranti only occur in specific configurations of A and P. Here is a list of the most important ones in Japhug (and the same are true of Zbu and Tshobdun, see Sun 2000a, Sun & Shidanluo 2002 and Gong 2014). 1. The past transitive -t suffix only occurs with a 1sg or 2sg A and a third person P in the perfective form. It is a portmanteau morpheme indicating both the person of the A and the P as well as TAM. 2. The portmanteau kɯ- and ta- prefixes mark 2→1 and 1→2 respectively; if the indexation system only marked SAP involvement these two forms would not be distinguished. 3. In the perfective, when A and P are both third person and no inverse prefix is present, directional prefixes take a special form. 4. Stem III is used in forms with a singular agent and a third person patient in non-past TAM categories. In Kiranti languages in general, recent work by Balthasar Bickel has brought out the existence of strict ergative and accusative alignment in the person indexation systems of various languages. Most spectacularly, Bickel (2008) (see Table 5, including only singular forms) concludes that in Puma (a Southern Kiranti language) 1sg has ergative alignment, 2sg neutral alignment, and 3sg accusative alignment, the exact opposite of what Silverstein’s Hierarchy (Silverstein 1976) predicts. 13 Table 5: Counterexample to the Silverstein A S P 1sg -ŋ (1→3) -ŋa (npst) -na (1→2) -oŋ (pst) 2sg tʌ-na (1→2) 3sg ∅ -u pʌ- (3→1) Hierarchy in Puma alignment ergative neutral accusative Other Kiranti languages, such as Khaling, abound with affixes and stem alternations that not only mark the presence of an SAP argument, but also unambiguously indicate syntactic relations: 1. The portmanteau 1sg→2 non past -nɛ and past -tɛni suffixes are not found in 2→1 forms. 2. The suffix -u combined with stem 1 (Jacques et al. 2012: 1104) specifically expresses 1sg→3, and completely differs from the corresponding inverse or intransitive forms. 3. The suffix -ʉ, combined with stem 5, indicates 3sg P in the non-past. 4.2 Ditransitive verbs In the case of ditransitive verbs, in Japhug the indexation of the T vs R is not dependent on person hierarchy, but on the configuration of each verb individually. Some ditransitive verbs, like mbi, treat the R like the P of a monotransitive verb, and it is indexed on the verb.7 (17) a-me nɯ nɤ-rʑaβ ɲɯ-ta-mbi ŋu 1sg.poss-daughter dem 2sg.poss-wife ipfv-1→2-give be:fact ‘I will give you my daughter in marriage.’ Other ditransitive verbs, which follow indirective alignment, treat the T like the P, while the R cannot be indexed on the verb and receives oblique case, as in 18. Using the 1→2 form ɲɯ-ta-kho ipfv-1→2-give would mean ‘I will give you (to him)’. (18) nɤʑɯɣ ɲɯ-kham-a ŋu 2sg:gen ipfv-give[III]-1sg be:fact ‘I will give it to you.’ The noun phrase nɤ-rʑaβ ‘(as) your wife’ is a functive phrase, syntactically an adjunct (see Creissels 2014). 7 14 The same is true of most ditransitive verbs; another example is thu ‘ask’, as in 19; the form treating the 2sg as the P, tu-ta-thu ipfv-1→2-give can only mean ‘I asked you (in marriage, to your father)’. (19) nɤ-ɕki tu-the-a 2sg-dat ipfv-ask[III]-1sg ‘I ask you (about it).’ In examples such as 18 and 19, we have a third person T and a second person R. If syntactic relations had no effect on person indexation in Japhug, and only the SAP > 3 hierarchy determined the use of a particular form, we would expect the 2sg to be indexed on the verb regardless whether it is T or R. 4.3 Agreement restrictions Verbs in languages such as Japhug and Khaling only agree with core arguments, not with possessor of arguments. Thus, in (20), the verb mŋɤm cannot take the 1sg suffix -a (such a form would be nonsensical). (20) a-xtu ɲɯ-mŋɤm 1sg.poss-belly sens-hurt ‘My belly hurts.’ (Japhug) In Khaling, although the verb |ŋet| can be used in the first person form, the third person must be used in example (21): (21) ʔʌ-mupu ŋêj 1sg.poss-belly hurt:3sg:npst ‘My belly hurts.’ (Khaling) Other Kiranti and Gyalrongic languages present similar, but slightly different systems (see for instance Khroskyabs, Lai 2015), and I would not be surprised to find languages differing from Japhug and Khaling by all three properties described above. To labour the point: it appears preferable to avoid lumping all languages with person indexation into one category on the basis of partial paradigms. A promising direction for future work consists in making use of the text corpora that are becoming available in these languages. This method holds great promise for gradual progress in unraveling the complex history of the tremendous morphosyntactic diversity of the Trans-Himalayan family. 5 Conclusion The increasing body of in-depth linguistic documentation on an ever greater number of Sino-Tibetan languages has potential for spectacular breakthroughs, 15 in the next few years, on the topic of the diachronic origins of person indexation systems in Trans-Himalayan. In particular, Kiranti and Gyalrongic languages are increasingly well-documented, allowing for refined reconstructions at the level of the proto-languages of these two groups. Simultaneously, digital corpus query technology allows the community of Trans-Himalayan linguists to build on an increasingly solid empirical basis when exploring these growing sets of data – including Tangut texts, as well as freshly documented living languages. These developments allow scientific debate to be ever more firmly grounded in text corpora, following common practice in Indo-European studies, which to this day remain an inspiring example for specialists of other language families. References Ahrens, Kathleen. 1990. Re-examining the evidence for verbal agreement in Tangut. In 23rd International Conference on Sino-Tibetan Languages and Linguistics, University of Texas at Arlington. Bauman, James John. 1975. Pronouns and Pronominal Morphology in TibetoBurman: University of California, Berkeley dissertation. Bickel, Balthasar. 2008. On the scope of the referential hierarchy in the typology of grammatical relations. In Greville G Corbett & Michael Noonan (eds.), Case and grammatical relations: papers in honor of Bernard Comrie, 191–210. Amsterdam: Benjamins. Boas, Franz & Ella Deloria. 1941. Dakota Grammar Memoirs of the National Academy of Sciences 23, Second Memoir. Washington, D.C. : Government Printing Oice. Creissels, Denis. 2014. Functive phrases in typological and diachronic perspective. Studies in Language 38(3). 605–647. DeLancey, Scott. 1989. Verb Agreement in Proto-Tibeto-Burman. Bulletin of the School of Oriental and African Studies 52.2. 315–333. DeLancey, Scott. 2010. Towards a history of verb agreement in TibetoBurman. Himalayan Linguistics 9.1. 1–39. DeLancey, Scott. 2011. Notes on verb agreement preixes in TibetoBurman. Himalayan Linguistics Journal 10.1. 1–29. DeLancey, Scott. 2014. Second person verb forms in Tibeto-Burman. Linguistics of the Tibeto-Burman Area 37:1. 3 – 33. 16 van Driem, George. 1991. Tangut verbal agreement and the patient category in Tibeto-Burman. Bulletin of the School of Oriental and African Studies 54.3. 520–534. van Driem, George. 1993. The Proto-Tibeto-Burman verbal agreement system. Bulletin of the School of Oriental and African Studies 61.2. 292–334. Gong, Hwang-cherng. 2001. Xixiayu dongci de rencheng huying yu yinyun zhuanhuan 西夏語動詞的人稱呼應與音韻轉換 (Personal Agreement and Phonological Alternations in the Tangut Verb). Language and Linguistics 2.1. 21–67. (龔煌城). Gong, Xun. 2014. Personal agreement system of Zbu rGyalrong (Ngyaltsu variety). Transactions of the Philological Society 112.1. 44–60. Hodgson, B.H. 1857-8. Sifan and Horsok vocabularies. Journal of the Asiatic Society of Bengal 22. 121–151. Hyman, Larry. 2011. The Macro-Sudan Belt and Niger-Congo Reconstruction. Language Dynamics and Linguistic Change 1.1. 1–47. Jacques, Guillaume. 2007. Textes tangoutes I, Nouveau recueil sur l’amour parental et la piété iliale Languages of the World/Text Collections 25. München: Lincom Europa. Jacques, Guillaume. 2009. The Origin of Vowel Alternations in the Tangut Verb. Language and Linguistics 10.1. 17–28. Jacques, Guillaume. 2010. The inverse in Japhug Rgyalrong. Language and Linguistics 11.1. 127–157. Jacques, Guillaume. 2011. The Structure of the Tangut Verb. Journal of Chinese Linguistics 39.2. 419–441. Jacques, Guillaume. 2012. Agreement morphology: the case of Rgyalrongic and Kiranti. Language and Linguistics 13.1. 83–116. Jacques, Guillaume. 2014a. Clause linking in Japhug Rgyalrong. Linguistics of the Tibeto-Burman Area 37.2. 263–327. Jacques, Guillaume. 2014b. Esquisse de phonologie et de morphologie historique du tangoute. Leiden: Brill. Jacques, Guillaume. 2016. On the directionality of analogy in a Dhegiha paradigm. International Journal of American Linguistics 8(2). 239–248. Jacques, Guillaume. to appear. Generic person marking in Japhug and other Rgyalrong languages. In Joana Jansen & Spike Gildea (eds.), Diachrony of hierarchical systems, Amsterdam: John Benjamins. 17 Jacques, Guillaume, Anton Antonov, Yunfan Lai & Lobsang Nima. 2014. Person marking in Stau. Himalayan Linguistics 13.1. 82–92. Jacques, Guillaume, Aimée Lahaussois, Boyd Michailovsky & Dhan Bahadur Rai. 2012. An overview of Khaling verbal morphology. Language and linguistics 13.6. 1095–1170. Jacques, Guillaume & Alexis Michaud. 2011. Approaching the historical phonology of three highly eroded Sino-Tibetan languages: Naxi, Na and Laze. Diachronica 28.4. 468–498. Kepping, Ksenija Borisovna. 1975. Subject and object agreement in the Tangut verb. Linguistics of the Tibeto-Burman Area 2:2. 219—231. Kepping, Ksenija Borisovna. 1983. Лес категорий, утраченная китайская лэйшу в тангутском переводе Памяники письменности востока 38. Москва: Наука. Kepping, Ksenija Borisovna. 1985. Москва: Наука. Тангутский язык – морфология. Kepping, Ksenija Borisovna. 1994. The conjugation of the Tangut verb. Bulletin of the School of Oriental and African studies 2. 339—346. Lai, Yunfan. 2015. The person agreement system of Wobzi Lavrung (Rgyalrongic, Tibeto-Burman). Transactions of the Philological Society 113(3). 271–285. LaPolla, Randy. 1992. On the dating and nature of the verb agreement in Tibeto-Burman. Bulletin of the School of Oriental and African Studies 55.2. 298–315. LaPolla, Randy. 2003. An overview of Sino-Tibetan morphosyntax. In Graham Thurgood & Randy J. LaPolla (eds.), The Sino-Tibetan Languages, 22–42. London & New York: Routledge. LaPolla, Randy. 2012. Comments on Methodology and Evidence in SinoTibetan Comparative Linguistics. Language and Linguistics 13.1. 117–132. Michailovsky, Boyd, Martine Mazaudon, Alexis Michaud, Séverine Guillaume, Alexandre François & Evangelia Adamou. 2014. Documenting and researching endangered languages: the Pangloss Collection. Language Documentation and Conservation 8. 119–135. Nishida, Tatsuo. 1975. 西夏文華嚴經 Seikabun Kegonkyō. Kyoto: 京都大學 文學部 Kyōto Daigaku Bungakubu. 西田龍雄. Norde, Muriel. 2009. Degrammaticalization. Oxford: Oxford University Press. 18 Rhodes, Richard. 1994. Valency, inversion and thematic alignment in Ojibwe. Berkeley Linguistic Society 431–446. Shibatani, Masayoshi. 1990. The languages of Japan. Cambridge: Cambridge University Press. Shǐ, Jīnbō, Zhènhuá Huáng & Hóngyīn Niè. 1990. Leilin yanjiu 類林研究 (Study on the Grove of Categories). Yinchuan: Ningxia renmin chubanshe. Silverstein, Michael. 1976. Hierarchy of Features and Ergativity. In Robert M.W. Dixon (ed.), Grammatical Categories in Australian Languages, 112–171. Canberra: Australian Institute of Aboriginal Studies. Sun, Jackson T.-S. 1998. Nominal Morphology in Caodeng rGyalrong. Bulletin of the Institute of History and Philology 69.1. 103–149. Sun, Jackson T.-S. 2000a. Parallelisms in the Verb Morphology of Sidaba rGyalrong and Lavrung in rGyalrongic. Language and Linguistics 1.1. 161– 190. Sun, Jackson T.-S. 2000b. Stem Alternations in Puxi Verb Inlection: Toward Validating the rGyalrongic Subgroup in Qiangic. Language and Linguistics 1.2. 211–232. Sun, Jackson T.-S. 2003. Caodeng rGyalrong. In Graham Thurgood & Randy LaPolla (eds.), The Sino-Tibetan languages, 490–502. London: Routledge. Sun, Jackson T.-S. & Shidanluo. 2002. Caodeng Jiarongyu yu rentong dengdi xiangguan de yufa xianxiang 草登嘉戎語與「認同等第」相關的語 法現象 (Empathy Hierarchy in Caodeng rGyalrong grammar). Language and Linguistics 3.1. 79–99. Ullrich, Jan. 2008. New Lakota dictionary. Bloomington: Lakota Language Consortium. Valentine, J. Randolph. 2001. Nishnaabemwin Reference Grammar. Toronto: University of Toronto Press. Zeisler, Bettina. 2015. Eat and drink – if you can! A language internal explanation for the ‘irregular’ paradigm of Tibetan za, zos, zo ‘eat’. Himalayan Linguistics 14(1). 34–62. Zúñiga, Fernando. 2006. Deixis and Alignment - Inverse systems in indigenous languages of the Americas. Amsterdam: Benjamins. 19