Complementation in Japhug Gyalrong

Complementation in Japhug Gyalrong* Guillaume Jacques November 22, 2016 Jacques, Guillaume 2016. Complementation in Japhug, Linguistics of the Tibeto-Burman Area, 39(2):222–281. Abstract: This article provides a detailed survey of complement clauses and complementation strategies in Japhug. It shows the bewildering diversity of constructions attested in these languages, which are largely unpredictable and need to be specified for each complement-taking verb. Special focus is given to typologically unusual constructions, in particular Hybrid Indirect Speech. Keywords: Japhug; Reported Speech; Complement clauses; Infinitive; Participle; Serial Verb Constructions; Causative; Motion Verbs Introduction Japhug, like other Gyalrong languages, has a complex verbal morphology and a rich array of complement clauses. Building on previous research, in particular Jacques (2008, 337-356) and Sun (2012), this article presents a detailed survey of all known types of complement clauses in Japhug and their distribution among complement-taking verbs. The framework of this article is based on Dixon (2006, 9) and adopts a terminology close to that used in Sun’s (2012) study of Tshobdun. The article comprises four main sections. First, I present background information on essive adjuncts and the difference between infinitives and participles. Second, I provide an overview of the various types of complement clauses and complement strategies in Japhug. Third, I describe some * The glosses follow the Leipzig glossing rules. Other abbreviations used here are: auto autobenefactive-spontaneous, anticaus anticausative, antipass antipassive, appl applicative, emph emphatic, fact factual, genr generic, ifr inferential, indef indefinite, inv inverse, lnk linker, poss possessor, ego.prs egophoric present, prog progressive, sens sensory. 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). 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). I would like to thank two anonymous reviewers, as well as Alec Coupe, for providing detailed and helpful comments and corrections on a previous version of this article. 1 syntactic peculiarities found in complement clauses in Japhug, for instance concerning coreference restrictions between the complement and the matrix clause. Fourth, I propose a list of complement-taking verbs classified by semantic categories, indicating for each verb which types of complements are possible. 1 Background information This section presents general information on Japhug, and focuses on two issues of morphosyntax that are relevant to properly understanding complementation in Japhug: arguments and adjuncts that occur in absolutive form (without case marking) on the one hand, and a classification of nominalized forms (including participles and infinitives) on the other. 1.1 Japhug and other Gyalrongic languages Japhug (Chinese Chabao 茶堡) is a Trans-Himalayan language spoken in ’Bar.khams county, Rnga.ba prefecture, Sichuan province, China. All Japhug data in this work come from the author’s fieldwork,1 and reflect the dialect of Kamnyu (kɤmɲɯ, 干木鸟) village. Japhug belongs to the Gyalrong branch of the Gyalrongic subgroup (Sun 2000b), alongside three other non-mutually intelligible languages, Tshobdun (Caodeng 草登), Zbu (a.k.a Showu, Ribu 日部) and Situ (四土, a.k.a. Eastern Gyalrong), which includes in particular the Cogtse (Zhuokeji 卓克基) and Somang (Suomo 梭磨) dialects on which most early research on Gyalrong languages has focused. Gyalrongic languages are argued to belong to a Burmo-Gyalrongic clade including Lolo-Burmese and Naish (Jacques & Michaud 2011), but this question is still debated (Thurgood to appear). Gyalrongic languages are among the few truly polysynthetic languages of the Trans-Himalayan family, and are known for their complex verbal morphology (Sun 2000a; Jacques 2013b; Lai 2013; DeLancey 2015), directinverse indexation system (DeLancey 1981; Sun & Shidanluo 2002; Jacques 2010; Gong 2014; Lai 2015), incorporation (Jacques 2012b; Lai 2015), and large consonant clusters (Sun 2007; Jacques 2013c). 1.2 Unmarked arguments and adjuncts Like all Gyalrong languages, Japhug is very strongly head-marking, with obligatory indexation of one or two arguments and morphological marking of transitivity.2 There are nevertheless some case markers, such as the ergative 1 My main consultant is tsʰɯndzɯn Chenzhen 陈珍, born 1950. There are several distinct convergent criteria for distinguishing between transitive and intransitive verbs in Japhug: the former have Stem 3 alternation in sg→3 non-past verb forms, the past transitive -t suffix in 1/2sg→3 forms, orientation prefixes in a- in past 2 2 / instrumental kɯ (obligatory on third person subjects3 of transitive verbs, as in 1), the comitative cʰo, the genitive ɣɯ and the dative ɯ-ɕki. (1) tɤ-tɕɯ nɯ kɯ χsɤr qaɕpa nɯ cʰɤ-mqlaʁ indef.poss-boy dem erg gold frog dem ifr-swallow ‘The boy swallowed the golden frog.’ (Nyima Wodzer.1, 131) Subjects of intransitive verbs and objects of transitive verbs are indexed on the verb morphology but receive no case marking. 1.2.1 Semi-transitive verbs A small class of verbs in Japhug are intransitive (ie, their subject does not take ergative marking, only one argument is indexed, and they do not present the morphological properties mentioned in fn. 2), but take an unmarked argument which presents object-like properties. In example (2), the verb rga ‘like’ occurs with an overt noun ndɯχu ‘flower sp.’, but verbal morphology shows that the verb is not transitive (otherwise a stem 3 alternation †rge-a instead of rga-a would be expected). (2) aʑo ndɤre ndɯχu rga-a 1sg topic flower.sp like:fact-1sg ‘As for me, I like the ndɯχu flower.’ (14-sWNgWJu, 131) The syntactic status of nouns like ndɯχu ‘flower sp.’ in (2) is better understood if relativization is taken into account. Like the objects of transitive verbs, these nouns can be relativized by using the kɤ- P-participle (on which see section 1.3.1), with an optional possessive prefix coreferential with the subject. Compare the relativized object of the transitive verb sɯz ‘know’ in (3) with the construction in (4) involving the intransitive verb rga ‘like’. (3) paχɕi [aʑo a-kɤ-sɯz] nɯ nɯra ɣɤʑu apple 1sg 1sg.poss-nmlz:P-know dem dem:pl exist:sens ‘The (types of) apples that I know about are these.’ (hist-07-paXCi, 73) (4) [pɣa ra nɯ-kɤ-rga] nɯ qaj ntsɯ ŋu bird pl 3pl.poss-nmlz:P-like topic wheat always be:fact ‘(The food) that birds like is always wheat (not barley).’ (23 pGAYaR, 29) 3→3’ direct forms, while the latter don’t. 3 Subjects and objects in Japhug can be defined on the basis of relativization patterns, in addition to case marking and indexation, see Jacques (2016b). 3 In the following, arguments selected by verbs such as rga ‘like’ that can be relativized using P-participles are referred to as semi-objects; semi objects are considered to be one subtype of core arguments. Verbs that take semiobjects are called semi-transitive. Semi-transitive verbs are few, but some of them are among the most common verbs in the language, and several of them can take complement clauses, including rga ‘like’, cʰa ‘can’ and even the adjectival stative verb mkʰɤz ‘be expert’. 1.2.2 Essive adjuncts A type of unmarked adjunct not previously described in the literature on Gyalrongic languages but crucial for understanding complement clauses and complementation strategies in Japhug is that of the essive adjunct. Essive noun phrases are not arguments of the sentence, but are used to indicate ‘the property of fulfilling the role of an N’ (Creissels 2014, 606); in English for instance, they are marked by as in a sentence such as ‘I am saying this as your friend’. In Japhug, bare noun phrases without any case marker can be interpreted as essive adjuncts, such as nɤ-rʑaβ ‘your wife’ in (5), which is neither the object (recipient) nor the theme of the verb mbi ‘give’. (5) 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. (=I will give her to you as your wife)’ Noun phrases headed by the possessed noun ɯ-spa ‘material’ are often used as an essive adjuncts as in (6), a construction that is in the process of grammaticalizing into a purposive phrase, and serves as a grammaticalization strategy with verbs of manipulation (see section 2.6.3). (6) zɣɤmbu ɯ-spa ɲɯ-nɯ-pʰɯt-nɯ broom 3sg.poss-material ipfv-auto-cut-pl ŋgrɤl be.usually.the.case:fact ‘They cut it to make brooms. (=as a material for brooms.)’ (140505 sWjno, 22) 1.3 Participles vs infinitives All Gyalrong languages, including Situ (Lín 2003), Tshobdun (Sun 2012) and Japhug, distinguish many different types of non-finite verb forms. (Sun & Lin 2007; Genetti et al. 2008; Prins 2011). In the present work, I propose to additionally distinguish between several sub-types of nominalizations. 4 I call participles the nominalized verb forms that can be used as noun modifiers and that refer to an entity that is either argument or adjunct of the nominalized verb (Creissels 2006a, 224). The main function of participles in Japhug is to build relatives (Jacques 2016b); these relatives are known as participial relatives (on which, see for instance Lehmann 1984, 50-58;735;156-7 and Creissels 2006b, 214), and include in particular adjectival stative verbs in attributive function, as in (7). (7) tɕʰeme kɯ-mpɕɯ~mpɕɤr girl nmlz:S/A-emph~be.beautiful ‘A beautiful girl’ = ‘A girl who is beautiful’ Infinitives are nominalized verb forms that refer to the action itself (not to an argument or an adjunct of the nominalized verb), can be used in complement clauses and preserve the argument structure of the verb, as in example (8), where the transitive verb pʰɯt ‘cut’ in infinitival form takes an overt object: (8) aʑo nɤ-mɤlɤjaʁ kɤ-pʰɯt mɤ-cʰa-a 1sg 2sg.poss-limbs inf-cut neg-can:fact-1sg ‘I cannot cut your limbs.’ (The fox03, 190) There are in addition a few other non-finite forms, such as action nominals (Jacques 2014b) and degree nominals (on which see Jacques 2016a) which are neither infinitives nor participles. The formal distinction between participles and infinitives in Japhug is quite subtle, as there are both infinitives and participles in kɯ- and kɤ- (or kə- and kɐ- depending on the transcription system). Since both categories are formally similar and can occur in similar contexts, it is crucial to clearly explain the distinction between the two, especially since Japhug slightly differs from the other Gyalrong languages in this regard. 1.3.1 Participles The system of core argument participles in Japhug is relatively straightforward (Jacques 2016b). The prefix kɯ- is used to build the S-participle of intransitive verbs, and the A-participle of transitive verbs. A-participles differ from S-participles in taking an additional possessive prefix coreferential with the P, as in (9). (9) a-me a-kɯ-fstɯn ŋu 1sg.poss-daughter 1sg.poss-nmlz:S/A-serve be:fact ‘My daughter is the one who takes care of me.’ (The prince, 74) 5 The prefix kɤ- on the other hand serves to build the P-participle, and can optionally take a possessive prefix coreferential with the A, as in (10). Participles are compatible with polarity and associated motion prefixes (Jacques 2016b). (10) aʑo a-mɤ-kɤ-sɯz tɤjmɤɣ nɯ kɤ-ndza 1sg 1sg-neg-nmlz:P-know mushroom dem inf-eat mɤ-naz-a neg-dare:fact-1sg ‘I do not dare to eat the mushrooms that I do not know.’ (23 mbrAZim,103) 1.3.2 Infinitives There are four types of infinitives in Japhug: kɯ-, kɤ-, tɯ- and bare infinitives. The latter two are restricted to very specific constructions (see 2.2), and only the former two types are discussed in this section. Infinitives in kɤ- are by far the most common form in Japhug. The kɯ- form is restricted to stative verbs (including adjectives, copulas and existential verbs) and impersonal auxiliaries, but even with these verbs, kɤinfinitives are used in several contexts. Aside from complement clauses (for which see section 2.1), infinitives are used in two types of constructions, a brief overview of which is provided below. First, infinitives occur as the citation forms of verbs and in metalinguistic discussion in Japhug, as in examples (11) and (12) for stative vs non-stative infinitives. (11) ɯnɯnɯ tɕe dem tɕe ɯ-tɯ-tʂɯβ lnk lnk 3sg.poss-nmlz:action-sew mɤ-kɯ-βdi tu-kɯ-ti ŋu neg-inf:stat-be.good ipfv-genr:A-say be:fact ‘People call this ‘badly sewn’.’ (12-kAtsxWb, 12) (12) pjɯ-sɯ-ʁndi tɕe pjɯ-sɯ-sat. tɕe nɯ koʁmɯz nɤ ipfv-caus-hit[III] lnk ipfv-caus-kill lnk dem only.then lnk cʰɯ-nɯtsɯm ɲɯ-ra. tɕe nɯnɯ ipfv:downstream-take.away sens-have.to lnk dem kɤ-nɤʁarphɤβ tu-kɯ-ti ŋu inf-strike.with.wings ipfv-genr:A-say be:fact ‘It strikes it and kills it (with its wings) and only then takes it away. This is called kɤ-nɤʁarphɤβ ‘strike with one’s wings’.’ (hist150819 RarphAB, 11) 6 In the topical position, the infinitive is neutralized to the kɤ- form even for stative verbs, as in (13). (13) kɤ-rʑi ri pjɤ-rʑi, inf-be.heavy also ifr.ipfv-be.heavy ‘As for being heavy, (the old man) was heavy.’ (140511 xinbada, 138) Second, infinitives are used as converbs to indicate the manner in which the action of the main clause occurs (example 14), or a background event (Jacques 2014a). The kɯ- infinitive form occurs with stative verbs (as sɤscit ‘nice (of an environment)’ in example 15) but it is also attested with a handful of dynamic verbs in lexicalized form such as mɤ-kɯ-mbrɤt ‘without stopping’ in (16).4 This latter use is the last trace of the contrast between human kɐ- and non-human kə- action nominals reported by Sun (2012, 476) and Sun 2014, which otherwise appears to have been lost in the variety of Japhug under study. (14) kɤ-ŋke jɤ-ari pɯ-ra inf-walk pfv-go[II] pst.ipfv-have.to ‘He had to go on foot.’ (elicited) (15) ɕɤr tɕe nɯtɕu kɯ-sɤ-scɯ~scit ʑo night lnk dem:loc inf:stat-deexp-emph~be.happy emph ɕ-ku-nɯ-rŋgɯ ŋu transloc-ipfv-auto-lie.down be:fact ‘In the night, he goes in there to sleep cosily.’ (26-NalitCaRmbWm, 35) (16) nɯ maka mɤ-kɯ-mbrɤt ʑo ɲɯ-rɤma dem at.all neg-inf-anticaus:break emph ipfv-work ɲɯ-ɕti tɕe sens-be:affirm lnk ‘It works without stopping.’ (hist-26-GZo.txt 67) Infinitives can take orientation, negation and even associated motion prefixes, but not possessive prefixes. 1.3.3 Japhug vs Tshobdun The inventory of participial and infinitive forms presented above for Japhug differs from other previous descriptions of Gyalrong languages. In Tshobdun, Sun (2012, 476) describes five types of verbs forms with kə- or kɐ- prefixes (Table 1), corresponding to Japhug kɯ- and kɤ- respectively. 4 The implied S of the verb mbrɤt ‘break, stop suddenly’ (the anticausative of prɤt ‘break’) in this sentence is the work of the subject. 7 Table 1: Nominalization types in Tshobdun (Sun 2012, 476) Type Scope Finiteness Argument coding Prefix purposive participant clausal clausal non-finite non-finite possessor possessor infinitive action / state clausal clausal non-finite non-finite normal normal finite clausal finite normal kəkə- (subject) kɐ- (object) kɐkɐ- [+human] kə- [-human] kə- While Japhug and Tshobdun have very close systems, the present analysis conflates some of Sun’s categories. The correspondences between the two terminologies are as follows. First, the categories ‘purposive’ and ‘participant’ in Tshobdun in Table (1) correspond to the Japhug participles. The difference here is partially terminological, but also has to do with the fact that in Japhug not just the subject participle, but also the object participle kɤ-, can appear as a purposive complementation strategy of motion verbs (see section 2.6.2) and there is little reason to distinguish between the two. Second, the categories ‘infinitive’ and ‘action / state’ in Table (1) correspond to the Japhug infinitive. Tshobdun and Japhug differ by the fact that the humanity contrast on these non-finite forms, quite prominent in Tshobdun, has become marginal in the variety of Japhug under study, making it unnecessary to differentiate between action/state nominalization and infinitive. Third, the category ‘finite’, which refers to verbs forms prefixed in kəwith reduced inflection (no person / number marking), appear not to exist under exactly the same form in Japhug. Sun (2012, 481) mentions the occurrence of these clauses with verb of pretence like nəʃpəz ‘pretend’. In Japhug, verbs of this type (like nɯɕpɯz ‘pretend’, cognate with the Tshobdun verb) do take clauses whose verb is prefixed in kɯ-, but there are reasons to analyze these as head-internal relative clauses with a verb in participial form (see section 2.6.2) rather than postulating a distinct morphological category. Little data is available on Zbu (the only publications specifically on this language are Sun 2004 and Gong 2014, with some data in Lin 1993 and Jacques 2008), but this language seems to have a system of nominalization similar to that of Japhug and Tshobdun. Situ on the other hand appears to be markedly different,5 as this language has an additional type of nominal5 Data on nominalization is available from three varieties of Situ: Cogtse (Lin 1993; Wei 2001; Sun & Lin 2007; Genetti et al. 2008), Kyomkyo (Prins 2011) and Bragdbar (personal fieldwork). 8 ized verb form with the same set of person indexation affixes as finite verb forms. Comparison between the nominalization system of Situ and those of the other Gyalrong languages is more difficult and must await later research (Jacques to appear presents preliminary ideas on this issue). 2 Complement types This section illustrates the different categories of complements attested in Japhug. Five main types of complement clauses are distinguished: kɤ-/kɯinfinitival complements, bare infinitival complements, finite complements and reported speech. In additional, Japhug has many distinct complementation strategies. 2.1 Infinitive The most common types of complement clauses in Japhug are kɤ- and kɯinfinitival complements. As seen in section 1.3.2, there are kɤ- and kɯinfinitives in Japhug, the latter found in the citation forms of stative verbs and modal impersonal auxiliary verbs. In complement clauses, stative verbs take the kɤ- infinitive like dynamic verbs in many cases. 2.1.1 Case marking While infinitives bear no person indexation markers, noun phrases receive the same case markers in infinitive clauses as in main clauses, showing that infinitives have the same argument structures as finite verb forms. When an argument is shared between the complement and the matrix clause, it often has a different syntactic function in the two clauses, as in (17), where tɤɕime ‘princess’ is A in the complement clause (stu ‘do like this’ is transitive) and S in the matrix clause (cʰa ‘can’ is semi-transitive; the complement clause is its semi-object). (17) [tɤɕime nɯ kɯ nɯra kɤ-stu] pjɤ-cʰa princess dem erg dem:pl inf-do.like.this ifr-can ‘The princess succeeded in doing it.’ (140511 alading, 252) In this sentence, the noun takes the ergative marker kɯ in accordance with the verb of the complement clause, showing that it belongs to the complement clause rather than to the matrix clause directly. This is the most commonly observed pattern in Japhug texts: in infinitival clauses, the shared arguments more often take the case marking selected by the verb of the complement clause than that of the matrix clause. 9 2.1.2 Coreference between matrix and complement clause Coreference restrictions between complements in kɤ- infinitives and their matrix clauses vary from verb to verb, and three cases can be distinguished. First, in the case of impersonal verbs such as ra ‘have to, need’ (see section 4.1.3), the complement clause is the S and there is no argument coereference between the matrix clause and the complement clause. Second, with a few transitive complement-taking verbs such as spa ‘know how to’ (see section 4.1.1) and nɤz ‘dare’ (4.5), the subjects of both clauses must be coreferential. Third, for most verbs taking infinitives (like the semi-transitive rga ‘like’ or the transitive rɲo ‘experience’), the subject of the matrix clauses can be coreferential to either the S (18), the A (19) or even the P (20 and 21) of its infinitival complement clause (Jacques 2016b). (18) tsuku tɕe kɤ-nɯrɤɣo wuma ʑo (19) aʑo qajɯ nɯ (20) maka tu-kɤ-nɤjoʁjoʁ, rga-nɯ tɕe some lnk inf-sing really emph like:fact-pl lnk ‘Some people like to sing.’ (26 kWrNukWGndZWr, 104) (S=S) ra kɤ-nɤrtoχpjɤt pɯ-rga-a tɕe 1sg bugs dem pl inf-observe pst.ipfv-like-1sg lnk ‘I liked to observe bugs.’ (26 quspunmbro, 15) (A=S) tu-kɤ-fstɤt nɯ ɲɯ-rga-nɯ at.all ipfv-inf-flatter ipfv-inf-praise dem ipfv-like-pl ‘They like to be flattered or praised.’ (140427 yuanhou, 53) (P=S) In (21), while χpɤltɕɯn is semantically a patient, it receives obligatory ergative marking due to the fact that the matrix verb rɲo ‘experience’ is transitive.6 (21) aʑo kɤ-mtsɯɣ mɯ-pɯ-rɲo-t-a ri, χpɤltɕɯn kɯ 1sg inf-bite neg-pfv-experience-pst:tr-1sg but Dpalcan erg pjɤ-rɲo ifr-experience ‘I have never been stung (by a wasp), but Dpalcan has.’ ndzWrnaR, 19) (P=A) (26- The subject of the matrix clause can even be coreferential with the possessor of the S, as in example (22), where the non-overt S should be a-xtu ‘my belly’ (1sg.poss-belly). The verb mŋɤm ‘hurt’ can only take a body part as its S – the experiencer is indicated by a possessive prefix on the body part, as in (23). 6 This example illustrates the fact that ergative marking on third person referents in Japhug is syntactically conditioned, unlike many other languages of the family with optional ergative marking (see DeLancey 2011 and other articles in the same volume of LTBA.) 10 (22) aʑo pɯ-xtɕɯ~xtɕi-a ʑo ri tɯxtɤŋɤm 1sg pst:ipfv-emph~be.small-1sg emph loc dysentery nɯ-atɯɣ-a tɕe, nɯ kɤ-mŋɤm pɯ-rɲo-t-a pfv-meet-1sg lnk dem inf-hurt pfv-experience-1sg ‘When I was very small, I had dysentery, (my belly) ached.’ (24pGArtsAG, 121) (23) a-xtu ɲɯ-mŋɤm 1sg.poss-belly sens-hurt ‘My belly hurts.’ 2.1.3 Stative infinitive Stative verbs, when occurring in a complement clause, generally take the kɤinfinitive, as in example (24) and (25). The main verb of the complement clauses in these examples have the kɤ- infinitive, even though both tu ‘exist’ and scit ‘be happy’ are stative verbs and have a citation form with the kɯprefix. (24) a-rŋɯl kɤ-tu (25) kɤ-scit pjɤ-ŋgrɯ pɯ-rɲo-t-a 1sg.poss-money inf-exist pst:ipfv-experience-pst:tr-1sg ‘I used to have money’. (elicited) ɲɯ-ŋu inf-be.happy ifr-succeed sens-be ‘She succeeded in being happy.’ (150818 muzhi guniang, 6) In complement clauses, kɯ- infinitives are uncommon, and only occur in two cases. First, the conversion to kɤ- infinitive only applies to stative verbs, not to impersonal modal verbs such as ra ‘have to, need’. When the latter occur in a complement clause, as in example (26), they always have the kɯ- prefix. (26) smɤn kɤ-ndza kɯ-ra pɯ-rɲo-t-a medicine inf-eat inf-have.to pst:ipfv-experience-pst:tr-1sg ‘I used to have to take medicine.’ (elicited) Second, complements of verb of perception or thought like sɯpa ‘consider as’ or sɯχsɤl ‘recognize, realize’ can take complement clauses with stative infinitives as objects, as in (27). (27) ɯʑo srɯnmɯ kɯ-ŋu nɯ tɤ-wa nɯ kɯ 3sg râkshasî stat.inf-be dem indef.poss-father dem erg mɯ-pjɤ-sɯχsɤl, ɯ-nmaʁ nɯ kɯ neg-ifr-recognize 3sg.poss-husband dem erg ‘That she was a râkshasî, the father did not realize it, her husband.’ (28-smAnmi, 62) 11 2.2 Bare infinitive and tɯ- infinitives Several phasal verbs, such as ʑa ‘begin’, sɤʑa ‘begin’, stʰɯt ‘finish’, jɤɣ ‘finish’, causative verbs derived from adjectives such as ɣɤtɕʰom ‘overdo, do too much’ or ɣɤβdi ‘do well’ and the aspectual verb rɲo ‘experience, have already’ are compatible with bare infinitival and tɯ- infinitival complements. The verbs sɤʑa ‘begin’ and rɲo ‘experience’ are more commonly used with kɤ- infinitives. With the exception of jɤɣ ‘finish’, these verbs are all transitive. Bare infinitives without possessive prefixes are attested in a very marginal construction involving negative existential verbs (see section 4.8). 2.2.1 Bare infinitives and transitivity Bare infinitives are formed by combining the stem 1 of the verb with a possessive prefix coreferential with the object of the complement clause, as in examples (28). Intransitive verbs do not form bare infinitives. (28) nɤʑo kɯ-fse a-ŋkʰor nɯ you nmlz:stative-be.like 1sg.poss-subject top ɯ-mto mɯ-pɯ-rɲo-t-a 3sg.poss-bare.inf:see neg-pfv-experience-pst:tr-1sg ‘I never saw anyone like you among my subjects.’ (Smanmi metog koshana1.157) Bare infinitives are in complementary distribution with tɯ- infinitives, which occur when the verb of the complement is morphologically intransitive. There is obligatory coreference between the S/A of the matrix verb and the S of the tɯ- infinitival complement. When the verb of the matrix clause is transitive and that of the complement clause intransitive, there is a conflict in case assignment on their common subject: the transitive verb requires the ergative on third person subjects, while the intransitive one precludes it. In most cases of this type, such as example (29), the common subject is in the absolutive in accordance with the intransitive verb of the complement clause. However in a few examples such as (30), the shared argument takes the ergative in accordance with the transitive matrix verb (the verb ʑa ‘begin’ is transitive). I consider that this difference is related to the boundary of the complement clause: in this first case the shared subject belongs to the complement clause, while in the second case it lies outside it. (29) [<xinbada> nɯ tɕe li tɯ-ŋke] to-ʑa Sinbad dem lnk again inf-walk ifr-begin ‘Sinbad started to walk again.’ (140511 xinbada, 217) (30) pɣɤtɕɯ nɯ bird kɯ [nɯɕɯmɯma ʑo tɯ-nɯrɤɣo] cʰɤ-ʑa dem erg immediately emph inf-sing ifr-begin 12 ‘The bird immediately started to sing.’ (140514 huishuohua de niao, 221) Bare infinitives and tɯ- infinitives are only compatible with polarity prefixes (as in example 31 below), and cannot take TAM or possessive prefixes. Crucially, semi-transitive verbs are treated like intransitive verbs: they cannot form a bare infinitive, and use tɯ- infinitives instead (example 31), although their semi-object does present some object-like syntactic properties (see section 1.2.1). (31) qaɟy ɯ-me nɯnɯ, tɕendɤre kʰro mɤ-tɯ-rga fish 3sg.poss-daughter dem lnk to-ʑa a.lot neg-inf-like ifr-start ‘He started not liking the mermaid that much.’ (hist150819 haidenver, 154) The only exceptions to this distribution are some transitive verbs used in complex predicates referring to weather phenomena, in particular lɤt ‘throw’ and βzu ‘make, do’, which take tɯ- infinitives as in (32). Note that in these complex predicates, the light verbs lɤt ‘throw’ and βzu ‘make, do’, although transitively conjugated, cannot take an overt A marked with the ergative. (32) tɯ-mɯ kɯ-wxtɯ~wxti ʑo tɯ-lɤt indef.poss-sky nmlz:S/A-emph~be.big emph inf-throw pjɤ-ʑa ifr-start ‘A big rain started.’ (hist150819 haidenver, 104) 2.2.2 Coreference restrictions Bare infinitives and kɤ- infinitives strongly differ as to their coreference restrictions. With kɤ- infinitives, the subject of the matrix clause can be coreferential with either the subject, the object or even the possessor of the intransitive subject of the complement clause (see example 22 above). This ambiguity is particularly clear with the verb nɤkʰu ‘invite to one’s home as a guest’ (see examples 33 and 34), as with this verb both arguments are equal in terms of volition and control. (33) ɯʑo kɯ kɤ-nɤkʰu pɯ-rɲo-t-a 3sg erg inf-invite pfv-experience-pst:tr-1sg ‘I have been to his house as a guest.’ (= ‘He has invited me to come to his house as a guest and I came.’) (P=A) (34) ɯʑo kɤ-nɤkʰu pɯ-rɲo-t-a 3sg inf-invite pfv-experience-pst:tr-1sg ‘He has been to my house as a guest.’ (= ‘I have invited him to come to my house as a guest and he came.’) (A=A) 13 In the case of bare infinitives, on the other hand, the subjects of the matrix and complement clause must be identical, but the object of the matrix clause can however be neutralized to third person.7 In example (35), the shared subject (referring to the host) is 3sg. The verb of the matrix clause takes the complement clause as a 3sg object (hence the verb takes the 3→3’ form without 1sg marking), while the verb of the complement clause takes a 1sg object (referring to the guest), marked by the possessive prefix a-.8 (35) a-nɤkʰu pa-rɲo (36) ɯʑo ɯ-nɤkʰu 1sg.poss-bare.inf:invite pfv:3→3’-experience ‘I have been to his house as a guest.’ (= ‘He has invited me to come to his house as a guest and I came.’) pɯ-rɲo-t-a 3sg 3sg.poss-bare.inf:invite pfv-experience-pst:tr-1sg ‘He has been to my house as a guest.’ (= ‘I have invited him to come to my house as a guest and he came.’) This generalization is observed for all transitive verbs taking bare infinitive complement clauses. The intransitive impersonal verb jɤɣ ‘finish’ takes the bare infinitive clause in S function, and remains in third person singular regardless of the subject and object of the complement clause, as in (37), where although the subject of the complement clause is third person plural, no plural marker can appear on jɤɣ. (37) nɯra ɯ-ti to-jɤɣ tɕe dem:pl 3sg.poss-bare.inf:say ifr-finish lnk ‘After having finished saying that, (they went to the park)’ (140515 congming de wusui xiaohai, 15) 2.2.3 Historical origin Bare infinitives probably derive from action nominals. There are marginal examples in Japhug of bare infinitives used in this way, as in example (38) (Jacques 2014b). (38) ndʑi-mi ɯ-tsʰoʁ ɯ-tsʰɯɣa nɯra 3du.poss-foot 3sg-bare.inf:attach.to 3sg.poss-form dem:pl wuma ʑo naχtɕɯɣ-ndʑi. very emph npst:similar-du 7 Note that apart from this syntactic difference, there is no detectable meaning difference between examples (33) and (34) on the one hand and their bare infinitive equivalents (35) and (36) on the other. 8 In the English translation, the 1sg is rendered as a subject, because translating a-nɤkʰu pa-rɲo as ‘He has invited me’ would be inexact, as this English sentence does not imply that the 1sg did attend the invitation. 14 ‘The way their feet (of fleas and crickets) touch the ground is very similar.’ (26-mYaRmtsaR, 17) As for the tɯ- infinitives, there are at least two possible ways of analyzing their origin. First, they could be related to the tɯ- action nominals (Jacques 2014b), found in light verb constructions such as (39) or (40)9 and used to build abstract nouns (si ‘die’ → tɯ-si ‘death’). This solution is attractive due to the fact that tɯ- action nominals are relatively common, but it does not account well for the complementary distribution of bare infinitives and tɯinfinitives, since both intransitive and transitive verbs can build tɯ- action nominals. (39) tɯ-rɟaʁ pɯ-βzu-t-a (40) kuxtɕo cʰondɤre kɯrtsɤɣ nɯra nmlz:action-dance pfv-do-pst:tr-1sg ‘I danced.’ ɯ-pa nɯtɕu basket comit leopard dem:pl 3sg-down dem:loc tɯ~tɯ-tʂaβ ʑo pjɤ-βzu together~nmlz:action-cause.to.roll emph ifr:down-make ‘(The rabbit) caused the leopard and the basket to roll down together.’ (The rabbit 2002, 72) Second, one could interpret the tɯ- here as the indefinite possessor prefix tɯ-, which is added to inalienably possessed nouns when no definite possessor is present (Jacques to appear). In this hypothesis, the bare infinitive takes a possessive prefix coreferential with its P with transitive verbs, but in the case of intransitive verbs, given the absence of P argument, the indefinite possessor is used instead. 2.3 Finite complements Finite subordinate10 clauses are common in Japhug and other Gyalrong languages. For instance, some specific types of relative clauses (Sun 2006; Jacques 2016b) or temporal subordinate clauses (Jacques 2014a) take a verb in finite form, in some cases without a subordinator.11 Likewise, finite complement clauses are common in Japhug and other Gyalrong languages.12 9 When the tɯ- nominalization prefix is reduplicated as in (40), it conveys the meaning of several persons/objects being subjected to the same action together. 10 In this work, ‘subordinate clauses’ is understood as a cover term for relative, complement and adverbial clauses. 11 Finite relatives, however, have some properties that distinguish them from corresponding independent clauses (Jacques 2016b, 18-21). 12 Sun (2012, 475-7) uses the term ‘S-like’ (sentence-like) clause instead of ‘finite’, which in his terminology refers to a different type of clause (see section 1.3.3). I chose ‘finite’ rather than ‘S-like’ to avoid confusion with ‘S’ as ‘intransitive subject’. 15 2.3.1 TAM forms Like finite relatives, there are constraints on the TAM forms that the main verb of a finite complement can take. Two cases can be distinguished. The most common type of finite complement have their verb in the imperfective, regardless of the TAM category of the complement-taking verb, whether in sensory or imperfective form as in (41), in irrealis form as in (42), or even in inferential or perfective forms as in (43) and (44). (41) tɤjpa kɯ-xtɕɯ~xtɕi ka-lɤt ri, snow inf:stat-emph~be.small pfv:3→3’-throw but mɯ́ j-ʁdɯɣ, pɤjkʰu tu-ndʐi ɲɯ-cʰa neg:sens-be.serious still ipfv-melt sens-can ‘There was a little snow, but it doesn’t matter, it can still melt.’ (conversation, 2015/12/17) (42) nɤʑo koŋla ʑo tu-kɯ-tɕat-a a-pɯ-tɯ-cʰa 2sg really emph ipfv-2→1-take.out-1sg irr-ipfv-2-can a-pɯ-ŋu irr-ipfv-be ‘If you can take me out of here...’ (140516 huli de baofu, 82) (43) ko-spa-nɯ qʰendɤre, nɯnɯ rɤɣo cʰɯ-tɯ-ʑa ifr-be.able-pl lnk ʑo cʰɯ-ti-nɯ qʰe, tɤrcɯrca dem song ipfv-imm-start lnk together to-cʰa-nɯ. emph ipfv-say-pl ifr-can-pl ‘They had learned (its songs) and as soon as it would start his song, they had become able to sing together with it.’ (140519 yeying, 156) (44) a-rɤɣo ɲɯ-tɯ-sɤŋo tɤ-tɯ-sɯso-t tɕe, 1sg.poss-song ipfv-2-listen pfv-2-think-tr:pst lnk a-ɣɯ-jɤ-kɯ-sɯ-ɣe-a qʰe nɯ ɕti irr-cisloc-pfv-2→1-caus-come-1sg lnk dem be.affirm:fact ‘When you want to listen to my song, just come and ask me.’ (140519 yeying, 238) Imperfective complements are the preferred construction when cʰa ‘can’ takes adjectival stative verbs as complements, as in (45). The meaning of this construction is specifically ‘be able to become X’. (45) cɤtʂʰa nɯnɯ ɯ-tɯ-mbro ɯ-fsu jamar ma nɯ tɯ-mtʰɤɣ plant.sp dem 3sg.nmlz:degree-be.high dem indef.poss-waist tu-mbro mɤ-cʰa 3sg.poss-size about apart.from ipfv-be.high neg-can:fact ‘The cɤtʂʰa plant can only grow as high as a person’ waist.’ (18-NGolo, 182) 16 Second, some complement-taking verbs, including impersonal modal verbs such as ntsʰi ‘have better’, ɬoʁ ‘have to’ or ra ‘have to, need’ or transitive verbs such as sɯso ‘think’ allow complements in the irrealis form, as (46).13 (46) ndɤre kɯ-xtɕɯ~xtɕi lnk inf:stat-emph~be.small a-mɤ-pɯ́ -wɣ-nɯ-ɕlɯɣ ɲɯ-ra ma irr-neg-pfv:down-inv-auto-drop sens-need because rca nɯ ɲɯ-ndoʁ qʰe ɕlaʁ ʑo unexpectedly dem sens-be.brittle lnk at.once emph pjɯ-ɴɢrɯ ɲɯ-ɕti ipfv-acaus:break sens-be:affirm ‘However, one should not let it drop even a little, otherwise, as it is very brittle, it would break at once.’ (30-Com, 27) Finally, complements in direct or hybrid indirect speech (in the case of verbs of speech and thought) do not have restrictions on TAM or person marking as found with other types of finite complements (see section 2.5). 2.3.2 Coreference restrictions Verbs taking finite complements can be divided into three groups depending on the coreference restrictions holding between the complement and the matrix clause. First, verbs such as cʰa ‘can’ require subject coreference between the clauses. For instance, in (47) and (48), the semi-transitive main verb cʰa ‘can’ respectively takes second and first singular person indexation, like the subject of the transitive complement verbs. Absence of person indexation on the verb of the matrix clause or indexing the object of the complement clause is impossible. (47) aʑo ɲɯ-kɯ-ɕɯɣ-mu-a mɤ-tɯ-cʰa 1sg ipfv-2→1-caus-be.afraid-1sg neg-2-can:fact ‘You cannot scare me.’ (140516 guowang halifa, 54) (48) cʰɯ-ta-ɕɯ-fka mɤ-cʰa-a ipfv-1→2-caus-be.satiated neg-can:fact-1sg ‘(If you eat me, as I am lean,) I will not be able to satiate your hunger.’ (140516 guowang halifa, 92) Second, other verbs such as rga ‘like’ allow coreference of the subject of the main clause with either the subject or the object of the complement clause (as in 49). 13 The same is also found in Tshobdun, see Sun (2012, 483). 17 (49) nɤʑo kɯ tu-kɯ-nɤjoʁjoʁ-a nɯra rga-a 2sg erg ipfv-2→1-flatter-1sg dem:pl like:fact-1sg ‘I like when you flatter me.’ (elicited) Unlike infinitival complements (as in example 22), coreference with a possessor is not possible. While it is possible to say (50), it is not grammatical to use a finite complement here, and one cannot replace the infinitive kɤ-mŋɤm (inf-hurt) with a finite form such as ɲɯ-mŋɤm (sens-hurt) for instance. (50) a-xtu kɤ-mŋɤm mɤ-rga-a 1sg.poss-belly inf-hurt neg-like:fact-1sg ‘I don’t like to have belly ache.’ (elicited) Third, in the case of complement clauses in S function, the verb of the matrix remains in third person singular form regardless of the person of the subject and object in the complement clause, as in (51) (had the matrix verb been in 1sg form like the subject of the complement clause, an (impossible) form such as †ɲɯ-ntsʰi-a would have been expected). (51) ki maka qala kɯ pjɤ́ -wɣ-nɯβlu-a tɕe dem at.all rabbit erg ifr-inv-cheat-1sg lnk z-ɲɯ-ɕar-a ɲɯ-ntsʰi transloc-ipfv-search-1sg sens-have.better ‘That rabbit has cheated me, I have to go to look for him.’ (31-qala, 39) 2.4 Complement-taking nouns Not all subordinate clauses modifying a noun should be analyzed as relative clauses: only subordinate clauses whose head noun (overt or covert) has a syntactic role in the clause (whether argument, adjunct or noun modifier) can be considered to be a relative clause (Sun 2006; Jacques 2016b).14 In example (52), the head noun ftɕaka ‘method, manner’ is neither a core argument nor an adjunct. It is not possible to transform the subordinate clause into an independent clause that would include this noun. This sentence is thus a complement clause, rather than a relative, despite having a noun rather than a verb as its head. 14 This point of view is not shared by scholars such as Matsumoto (1988) and Comrie (1998), but whatever the theoretical merits of their approach, for the purpose of detailed language description it is always better to err on the side of splitting than on that of lumping. In Japhug, there are in any case observable, though subtle, differences between relative and complement clauses modifying a noun. 18 (52) [[jɯm kɤ-ɕar ] ftɕaka] ɣɯ tɯ-rɟaʁ wife:hon inf-search method gen nmlz:action-dance sɯ-βzu-j caus-do:fact-1sg ‘Let us make dances to look for a wife (for the prince).’ (The Prince, 8) There are also examples of reported speech finite complement clauses of this type, with nouns relating to speech and information, such as the possessed nouns -fɕɤt ‘story’ or -tɕʰa ‘information’, as in examples 53 and 54. (53) tɕeri [zlawiɕɤrɤβ kɯ tɕʰoz pɯ-asɯ-zgrɯβ] but Zlaba.shesrab erg religion pst.ipfv-prog-accomplish tu ma [jɯm pɯ-asɯ-ɕar ] 3sg.poss-story exist:fact but wife:hon pst.ipfv-prog-search ɯ-fɕɤt ɯ-fɕɤt me 3sg.poss-story not.exist:fact ‘People say that Zlaba shesrab was studying religion, not that he was looking for a wife.’ (The prince, 79-80) (54) [<donggua> cʰo <qiezi> ni tɕʰi ʑo mɯ́ j-naχtɕɯɣ] gourd and eggplant du what emph neg:testim-similar ɣɯ ɯ-tɕʰa a-jɤ-tɯ-ɣɯt ra gen 3sg.poss-information irr-pfv-2-bring need:fact ‘(Then,) come to tell me in what way a gourd and an eggplant are different.’ (yici bi yici you jinbu 7) Complement-taking nouns all have corresponding complement-taking verbs, which can be either denominal verbs (-tɕʰa ‘information’ → ɣɯtɕʰa ‘answer’), verbs from which the nouns are derived (fɕɤt ‘tell (a story)’ → fɕɤt ‘story’) or complement-taking noun-verb collocations, such as ftɕaka+βzu (from ftɕaka ‘method, manner’ and βzu ‘do, make’) which can either mean ‘do by any means possible’ as in (55) or ‘prepare to X’ (as in 56; in this meaning, this collocation is homonymous with the denominal verb nɯftɕaka ‘prepare to’). (55) a-<xuetang> ɯ-tɯ-mbro <kongzhi> 1sg.poss-blood.sugar 3sg.poss-nmlz:action-be.high control tu-βze-a ŋu, tu-βze-a ŋu mɯ-tu-kɤ-mbro ftɕaka ipfv-do[III]-1sg be:fact neg-ipfv-inf-be.high manner ipfv-do[III]-1sg be:fact ‘I do whatever I can to prevent my blood sugar being too high’. (conversation, 15/12/05) 19 (56) qartsɤβ kɤ-kɤ-βzu ra kɤ-tɤβ ftɕaka ɣɯ-βzu harvest pfv-nmlz:P-make pl inf-thresh manner inv-make ra have.to:fact ‘Then one has to prepare to thresh the (cereals) that have been harvested.’ (2010.10-tArAku, 71) Complement clauses with nominal heads differ from relatives in one important respect: while most types of relative clauses can be either prenominal, postnominal or head-internal (Jacques 2016b), complement clauses are exclusively prenominal. 2.5 Reported speech Verbs of speech and cognition can take reported speech complements, in which the speaker either (exactly or partially) reproduces a sentence uttered by the person he is quoting, or verbalizes the words he assumes a person is thinking. These clauses have finite verb forms, and are thus a sub-category of finite complement clauses, but present some properties distinguishing them from the complements studied in the previous section. While Japhug allows direct speech quotation, the corpus reveals examples of mismatches between the viewpoint of the original speaker (the person whose speech or thoughts are quoted) and the current speaker (the person quoting the words of the original speaker), which following Tournadre (2008) I will refer to as Hybrid Indirect Speech (also called Semi-Indirect Speech, Aikhenvald (2008)). In Hybrid Indirect Speech, the verb morphology (in particular person indexation) invariably presents the viewpoint of the original speaker, while pronouns and adverbs follow that of the current speaker. Since grammatical relations are mainly marked by verb morphology, and overt pronouns are not common in texts in Japhug and other Gyalrongic languages (a typological feature shared with Kiranti languages, see Bickel 2001), distinguishing between Direct Speech and Hybrid Indirect Speech is only possible in a minority of cases. Mismatch between pronouns and person indexation only occurs when a pronoun or possessive prefix is overt and when at least one argument in the sentence is referred to by a different person form by the original speaker and the current speaker. Example (57) provides an example of this phenomenon. The verb nɯɣi ‘he comes/will come back (home)’ in the complement clause of the verb kɤsɯso ‘think’ is in the Factual third person singular form. In the same clause we find the second person singular pronoun nɤʑo ‘yous ’; there is no pause between the pronoun and the verb, and no indication from the prosody that nɤʑo is left-dislocated. This type of mismatch between pronouns and indexation on the verb is anomalous and never found in independent sentences. Here the verb form 20 corresponds to the point of view of the original speaker (indicated in blue in all following examples), whose original sentence would have been ɯʑo nɯɣi 3sg come.back:fact ‘he is coming back’. The pronoun reflects the point of view of the current speaker (in red), for whom the equivalent sentence would be converted to nɤʑo tɯ-nɯɣi 2sg 2-come.back:fact ‘you are coming back’, since the addressee of the current situation corresponds to the subject of the original situation. Three distinct translations are proposed here: a direct speech translation, reproducing the words used by the original speaker in his thoughts, an indirect speech translation, and an attempt at representing Japhug Hybrid Indirect Speech in English. (57) ma nɤ-wa kɯ [nɤʑo nɯɣi] kɤ-sɯso kɯ lnk 2sg.poss-father erg 2sg come.back:fact inf-think erg kʰa ɯ-rkɯ tɕe ʁmaʁ χsɯ-tɤxɯr house 3sg.poss-side lnk soldier three-circle pa-sɯ-lɤt ɕti tɕe pfv:3→3’-caus-throw be.affirm:fact lnk Direct: ‘Your father, thinking ‘He is coming back’, put three circles of soldiers around the house.’ Indirect: ‘Your father, thinking that you are coming back,’ Hybrid indirect: ‘Your father, thinking that ‘you’ is coming back,’ (qachGa 2003, 154) Examples (58) and (60) illustrate the case of possessive prefixes on nouns, which undergo the same shift towards the point of the view of the current speaker, while the verb remains in the same form that was either thought or uttered by the original speaker. (58) tɕendɤre ta-ʁi nɯ kɯ lnk indef.poss-younger.sibling dem erg ɣɯ ɯ-sci [ɯ-pi tu-nɤme-a 3sg.poss-elder.sibling gen 3sg.poss-revenge ipfv-make[III]-1sg ra] ɲɤ-sɯso tɕe, have.to:fact ifr-think lnk Direct: ‘The (younger) sister thought “I have to get revenge on my brother”.’ Indirect: ‘The (younger) sisteri wanted to get revenge on heri brother.’ Hybrid indirect: ‘The (younger) sisteri thought Ii have to get revenge on heri brother”.’ (xiong he mei, 17) The original sentence corresponding to the complement clause in (58) is presented in (59): the possessive pronoun was first person (coreferential 21 with the A of the main verb) and undergoes a shift to third person in (58) (representing the point of view of the person telling the tale). (59) a-pi ɣɯ ɯ-sci tu-nɤme-a 1sg.poss-elder.sibling gen 3sg.poss-revenge ipfv-make[III]-1sg ra have.to:fact ‘I have to get revenge on my brother.’ (elicitation based on 58) Example (60) illustrates the same phenomenon as (58), but with the verb of speech ti ‘say’ instead of sɯso ‘think’. In this example, we know from the context that the girl is the addressee, so that if the sentence were in direct speech, a second person singular prefix form nɤ-kɯmtɕʰɯ (2sg.posstoy) ‘your toy’ would be expected instead. (60) tɤɕime nɯ kɯ pjɯ-tɯ-mtsʰɤm tɕe, [nɯnɯ dem erg ipfv-conv:imm-hear lnk dem nɯ ju-ɣɯt-a ŋu] ɯ-kɯ-ti dem ipfv-bring-1sg be:fact 3sg-nmlz:S/A-say girl ndɤre, ɯ-kɯmtɕʰɯ 3sg.poss-toy pjɤ-tu ifr.ipfv-exist lnk Direct: ‘As soon as the girl heard that there was someone saying “I will bring your toy”.’ Indirect: ‘As soon as the girl heard that there was someone saying that he would bring her toy.’ Hybrid indirect: ‘As soon as the girli heard that there was someone saying “I will bring heri toy”.’ (140429 qingwa wangzi, 49) In example (61), the pronoun ɯʑo ‘he’ has no case marking, while the matrix verb sɯso ‘think’ is transitive and requires its subject to be marked with the ergative, indicating that ɯʑo ‘he’ belongs to the complement clause whose verb rɤʑi ‘remain, stay’ is intransitive. The verb form mɯ-pɯ-rɤʑi-a with first singular marking reflects the viewpoint of the original speaker, while the pronoun ɯʑo ‘he’ corresponds to that of the current speaker. (61) “ɯʑo χsɯ-sŋi χsɤ-rʑaʁ ma mɯ-pɯ-rɤʑi-a” 3sg three-day three-night apart.from neg-pst.ipfv-stay-1sg ɲɯ-nɯ-sɯsɤm pjɤ-ŋu ipfv-auto-think[III] ifr.ipfv-be Direct: ‘He was thinking “I have only stayed for three days and three nights”.’ Indirect: ‘He was thinking that he had only stayed for three days and three nights.’ 22 Hybrid Indirect: ‘He was thinking that he have only stayed for three days and three nights.’ A potentially even more confusing case occurs when the original speaker is the current speakers’ addressee, and when both the original and the current speakers are referred to in the original utterance. This is the situation observed in example (62), a sentence pronounced by a fox who helped a prince to succeed in various tasks. Here, the first singular possessive prefix a- on the possessed noun -tʂɯnlɤn ‘favour’ and the first person singular suffix -a on the verb ɲɯ-nɯ-fsɯɣ-a do not correspond to the same referent. The verb form ɲɯ-nɯ-fsɯɣ-a ‘I will pay back’ is the sentence that the fox attributes to his addressee (the prince), so that the first person here corresponds to the prince, while the possessive prefix on -tʂɯnlɤn ‘favour’ reflects the point of view of the fox and thus refers to himself. (62) a-tʂɯnlɤn ɲɯ-nɯ-fsɯɣ-a ɯ-ɲɯ-tɯ-sɯsɤm nɤ, 1sg.poss-favour ipfv-auto-pay.back-1sg q-ipfv-2-think[III] lnk nɯ tɤ-ste ti ɲɯ-ŋu dem imp-do.this.way[III] say:fact sens-be Direct: ‘If you think “I will requite the favour (which I received from you)”, do like that.’ Indirect: ‘If you want to requite the favour (which you received from me), do like that.’ Hybrid Indirect: ‘If you think “I will requite the favour (which you received from me), do like that.’ In such a situation, the referents corresponding to first and second person are exactly reversed between the point of view of the current and the original speaker, and therefore between pronouns and possessive prefixes on the one hand and verbal indexation on the other hand. The corresponding sentence in Direct speech would be (63), with a second person singular possessive prefix on the noun -tʂɯnlɤn ‘favour’ instead. (63) nɤ-tʂɯnlɤn ɲɯ-nɯ-fsɯɣ-a 2sg.poss-favour ipfv-auto-pay.back-1sg ‘I will requite the favour (which I received from you).’ Surprisingly, despite this complex shift of perspective between the original speaker and the current speaker, there is no logophoric pronoun in Japhug (Hagège 1974; Nikitina 2012). A logophoric pronoun is however attested in the closely related Stau language, which appears to have a similar system of Hybrid Indirect Speech (?). 23 2.6 Complementation strategies Dixon (2006) introduces the term ‘complementation strategy’ to refer to constructions corresponding with a meaning expressed by complement clauses in some languages, which either are not core arguments or the verb of the main clause or are not clauses with a complete argument structure (Dixon 2006, 34-40). Complementation strategies include nominalizations (when the verb sheds its argument structure as it becomes a noun), relative clauses (which are formally a modifier of a core argument), serial verb constructions and clause linking. 2.6.1 Headless relative clauses in core argument function Not all clauses occurring in subject, object or adjunct function are complement clauses. Most relative clauses in Japhug are headless (Jacques 2016b). When headless relative clauses occur in subject or object function, they may be superficially similar to complement clauses. In example (64), the main verb cʰa ‘can’ takes a finite complement clause. Note that the subject of the matrix clause and that of the complement clause are coreferential. (64) tɕe li ndʑi-pɤrtʰɤβ χpaχtsʰɤt kɯngɯt jamar lnk again 3du.poss-space.in.between yojana ɲɯ-sɯ-ɤβze pjɤ-cʰa nine about ipfv-caus-become[III] ifr-can ‘He succeeded in making (the distance) between them to be of nine yojanas.’ (hist-28-smAnmi, 237) In example (65), despite the fact that we have the same form pjɤ-cʰa in the matrix clause as in (64), the subordinate clause tɯ~ta-tɯt ‘all that he had said’ is here a headless relative clause. Evidence for this analysis is as follows. (65) tɤ-pɤtso nɯ kɯ nɯra [tɯ~ta-tɯt ] indef.poss-child dem erg dem:pl total~pfv:3→3’-say[II] nɯra pjɤ-cʰa dem:pl ifr-can ‘The child had succeeded in doing everything that (the old king) had said.’ (140428 yonggan de xiaocaifeng, 256) First, the subject of both clauses are not coreferential (this example cannot be interpreted as meaning ‘the boy succeeded in saying all these things’). If the subordinate clause in (65) were a complement clause, subject coreference would be expected with the verb cʰa ‘can’ (see section 2.3.2). 24 Second, the verb of the relative clause has totalitative reduplication, which is attested in relatives and some temporal subordinate clauses (Jacques 2014a, 295), but not in complement clauses. Third, the verb of the relative clause is in the perfective form while that of the the main clause is in the inferential; in finite complement clauses other than reported speech, the verb should be in imperfective form (see section 2.3.1). Fourth, although demonstrative and plural markers can follow a complement clause (see section 3.3), only relative clauses can take both preclausal and postclausal demonstratives as in (65), like any noun phrase. Fifth, it is possible to make a sentence with an overt head noun in (65), while this option does not exist with (64). The contrast between finite headless relative and complement clauses in object function is not always as clear as that between (65) and (64), but one of the criteria presented above, or a combination thereof, can be used to discriminate between the two. Formal near-ambiguity between headless relative and complement clauses can also exist in the case of non-finite subordinate clauses. In (66), the intransitive verb ŋgrɯ ‘succeed’ takes an infinitive complement in kɤ- as its subject, while in (67), its subject is a headless relative clause with an object participle, as is shown by the fact that it can take a possessive prefix coreferential with its A. Note also the semantic difference between a relative (‘may what you wish for come true’) and a complement clause (? ‘may you succeed to wish it.’). (66) kɤ-nɯsmɤn a-mɤ-pɯ-ŋgrɯ cʰo ɯ-nɯsmɤn inf-treat irr-neg-ipfv-succeed comit 3sg.poss-bare.inf:treat a-mɤ-tɤ-βdi tɕe pjɯ-kɯ-sat irr-neg-ipfv-be.good lnk ipfv-genr:S/P-kill ɲɯ-ŋgrɤl sens-be.usually.the.case ‘If one cannot treat it (rabies), if one does treat it well, it is fatal.’ (29-chWsYu, 29) (67) nɤ-kɤ-nɯsmɯlɤm nɯ a-pɯ-ngrɯβ 2sg.poss-nmlz:P-wish dem irr-pfv-succeed ‘May your wishes (=the things that you wish for) succeed.’ (elicited) A more delicate ambiguous case of relative vs complement clause is discussed in section (2.6.2) in the case of pretence verbs. Besides plain finite relatives and participial relatives, correlatives (with an interrogative pronoun) can also appear in subject or object function, as in (68). (68) [ɯ-wa tɕʰi tɤ-stu-t-a] nɯ 3sg.poss-father what pfv-do.like-pst:tr-1sg dem 25 tu-ste-a ɲɯ-ɬoʁ ipfv-do.like[III]-1sg sens-have.to ‘I have to deal with him in the same way as I dealt with his father. (=How I treated his father, I have to treat him like that)’ (Slob.dpon2, 159) There are, however, cases where a surface form can be analyzed either as a complement clause or as a relative, in particular with verbs of perception. In example (69), the verb form tɤ-kɤ-ta can either be analyzed as the object participle, or as the infinitive of ta ‘put’, resulting in two slightly different but semantically nearly identical translations.15 (69) nɯɕɯmɯma ʑo iɕqʰa kɯm nɯtɕu immediately emph the.aforementioned door dem:loc ɯ-ftaʁ tɤ-kɤ-ta nɯ pjɤ-mto 3sg.poss-mark pfv-inf/nmlz:P-put dem ifr-see ‘She immediately saw the mark that had been put on the door / that someone had put a mark on the door.’ (140512 alibaba, 183) Similar ambiguities can also occur with finite relatives. Example (70) can be either parsed as having a finite complement clause nɯra ta-tɯt ‘He said these words’,16 or a relative ta-tɯt ‘what he said’ (with preclausal and postclausal demonstratives nɯra ‘these’), with little semantic difference. (70) ɯ-ʁjoʁ nɯ kɯ nɯra ta-tɯt nɯra 3sg.poss-servant dem erg dem:pl pfv:3→3’-say[II] dem:pl pjɤ-mtsʰɤm ifr-hear ‘His servant heard what (the king) had said / that he had said these (words).’ (140428 yonggan de xiaocaifeng, 265) In the case of example (71), the speaker started by saying the relative clause [spjaŋkɯ kɯ sɯŋgi ɯ-ɕki tɤ-kɤ-tɯt ] ‘what the wolf had said to the lion’ and then corrected herself and said the verb form pɯ́ -wɣ-nɯmtɕʰu, which can either be interpreted as a relative (‘The slandering words that the wolf said’)17 or alternatively as a complement clause (‘that the wolf had slandered him’), showing the close proximity of these two possible analyses. 15 Genetti (2007) discusses similar cases of formally ambiguous ‘bistructural’ constructions in Dolakha Newar (in particular p. 375) and their crucial importance for understanding grammaticalization processes. 16 With verbs of cognition, speech or perception, there are no coreference or TAM restrictions between the complement and the matrix clause, see (4.3). 17 The verb nɯmtɕʰu ‘slander’ takes the person slandered as its object, but it can be construed as a secundative ditransitive verb whose third argument (the slandering words) can also be relativized with a finite relative clause, like the theme of the verb mbi ‘give’ (see Jacques 2016b, 16-17). 26 (71) spjaŋkɯ kɯ wolf sɯŋgi ɯ-ɕki tɤ-kɤ-tɯt,... pɯ́ -wɣ-nɯmtɕʰu erg lion 3sg-dat pfv-nmlz:P-say[II] pfv-inv-slander nɯnɯra pjɤ-mtsʰɤm. dem:pl ifr-hear ‘(The fox) heard what the wolf had said to the lion... (that the wolf) had slandered him.’ (140425 shizi lang huli, 16) 2.6.2 Participial clauses Japhug, like Tshobdun (Sun 2012) has a highly grammaticalized type of complementation strategy involving participial clauses. It occurs with motion verbs (ɕe ‘go’, ɣi ‘come’, but not rɟɯɣ ‘run’) and a few aspectual verbs (rɤŋgat ‘be about to’, ʑɣɤmɲo ‘prepare to’). These verbs are all morphologically intransitive. These subordinate clauses, although lexically selected by the verbs, are neither objects nor semi-objects, and therefore cannot be considered to be complement clauses in Dixon’s sense (see also Sun’s 2012 analysis of Tshobdun). Example (72) illustrates this construction, with the A-participle ɯ-kɯ-nnɤjo ‘waiting for him’. Note that the common argument shared between the participial clause (whose verb is transitive) and the matrix clause (whose verb is intransitive) takes the ergative, showing that it belongs to the participial clause, a pattern already observed with infinitival complements (see section 2.1.1). (72) [ɯ-wa nɯ kɯ kʰɤpa tɕe 3sg.poss-father dem erg downstairs lnk ɯ-kɯ-n-nɤjo] pjɤ-ɣi 3sg-nmlz:S/A-auto-wait ifr:down-come ‘His father had come downstairs to wait for him.’ (140506 loBzi, 5) S/A- vs P-participles In examples such as (72), the S or A of the participial clause is obligatorily coreferential with the S of the matrix clause. Coreference between the P of the participial clause and the S of the matrix clause is possible but requires using the P-participle. Several examples of this construction are found in the corpus with the verb nɤkʰu ‘invite to one’s home as a guest’, as in (73) (see section 2.2.2 concerning the argument structure of the verb nɤkʰu). (73) <xingqi> raŋri ʑo tɕe nɯnɯ sɤβʑɯ ɣɯ ɯ-kʰa week each emph lnk dem mouse gen 3sg.poss-house nɯtɕu kɤ-nɤkʰu] ju-ɣi pjɤ-ŋu dem:loc nmlz:P-invite ipfv-come ipfv.ifr-be ‘He would come to the mouse’s house as a guest.’ (150818 muzhi guniang, 299). 27 We know that this form is the P-participle rather than the infinitive because it is possible to optionally add a possessive prefix coreferential with the A, as in (74). (74) a-kɤ-nɤkʰu jo-ɣi 1sg.poss-nmlz:P-invite ifr-come ‘He came to my house as a guest.’ (elicited) Coreference between the P of the complement clause and the S of the matrix clause is possible only if the P has control over the action, something that is possible for only a few transitive verbs and explains the rarity of this construction. Verbs of pretence The transitive verb nɯɕpɯz ‘pretend, imitate’ and the semi-transitive ʑɣɤpa ‘pretend’ superficially appear to have the same complementation strategy as motion verbs, as could be deduced from examples such as (75). (75) tɤ-mu nɯ kɯ ɯ-ku ci indef.poss-mother dem erg 3sg.poss-head indef ɲɯ-kɯ-rɤβraʁ to-nɯɕpɯz ipfv-nmlz:S/A-scratch ifr-pretend ‘The (râkshasî)-mother pretended to scratch her head.’ (Slob.dpon1, However, unlike verbs such as ɕe ‘go’ or rɤŋgat ‘be about to’, coreference between the subject of pretence verbs and the subject of the verb in participial form is not required, as in example (76). (76) ɯ-zda nɯra, [pɣɤtɕɯ kɯ-ɣɤwu,] [kʰɯna 3sg.poss-companion dem:pl bird nmlz:S/A-cry dog kɯ-ɤndzɯt,] [lɯlu kɯ-ɣɤwu,] [qacʰɣa kɯ-mbri] nmlz:S/A-bark cat nmlz:S/A-cry fox nmlz:S/A-cry kɯ-fse, nɯra tu-nɯɕpɯz ɲɯ-spe. inf:stat-be.like dem:pl ipfv-imitate sens-be.able[III] ‘It is able to imitate other animals, cry like a bird, bark like a dog, meow like a cat or call like a fox.’ (27-kikakCi, 141) Since the verb nɯɕpɯz ‘pretend, imitate’ is transitive and can take as its P the person imitated by the A, as in (77), a different analysis of (75) and (76) offers itself: the clauses containing the S/A-participles there (indicated here between square brackets) are in fact head-internal/postnominal relatives. (77) ɣzɯ ra kɯ li ʑɤni tú-wɣ-nɯɕpɯz-ndʑi monkey pl erg again 3du ipfv-inv-imitate-du ‘The monkeys imitated the two of them (and repeatedly threw back the coconuts at them).’ (140511 xinbada, 262) 28 Examples (75) and (76) could be literally translated as ‘The mother pretended to be someone who is scratching her head’ and ‘It is able to imitate a crying bird, a barking dog, a meowing cat, a crying fox’ respectively. In Tshobdun, Sun (2012, 481-2) posits a distinct category of finite nominalized predicates in kə- to refer to the verb forms in nominalized clause in core argument position and in subordinate clauses associated with verbs of pretence. In Japhug, this analysis is not necessary for three reasons. First, transitive verbs occurring with nɯɕpɯz ‘pretend’ or ʑɣɤpa ‘pretend’ take a possessive prefix coreferential with the object, optional if TAM or polarity prefixes are present (78) and obligatory if no other prefix is found. (78) ɯ-ŋga ɯ-qʰu nɯ 3sg.poss-clothes 3sg.poss-behind dem ɯ-tu-kɯ-rɤɕi ra to-ʑɣɤpa-nɯ 3sg.poss-ipfv-nmlz:S/A-pull pl ifr-pretend-pl ‘They pretended to pull the train of his gown.’ (140521 huangdi de xinzhuang, 176) Second, verbs of pretence are compatible with object participles in kɤ-, with coreference of the subject of the main clause and the object of the participial clause (79). (79) ɕɯ-kɤ-nɤkʰu to-nɯɕpɯz transloc-nmlz:P-invite ifr-pretend ‘He pretended to go there as a guest.’ (not: ‘He pretended to invite him.’) Third, the prefix kɯ- in Japhug cannot be used to express action nominalization as the prefix kə- in Tshobdun (Sun 2012, 482). For these reasons, despite the surface similarity of the constructions in Japhug and Tshobdun, there is no need to establish a morphological category distinct from participles in Japhug. From a historical perspective, it is possible that the use of participles with motion verbs and with rɤŋgat ‘be about to’ was grammaticalized from a construction with such a head-internal relative clause in essive function (cf section 1.2) instead of P function, as in (80). Reanalysis was complete when coreference between the subjects of the matrix and of the participle became obligatory. (80) ɯ-kɯ-nɤjo pjɤ-ɣi 3sg-nmlz:S/A-wait ifr:down-come *‘He came as someone waiting for him’ ⇒ ‘He came to wait for him’. (from example 72) 29 In order to test this hypothesis however, data from other Gyalrong languages, in particular Situ, need to be taken into account, and this topic has to be deferred to future work on comparative Gyalrong morphosyntax. 2.6.3 Relative clause in essive function While motion verbs use purposive participial clauses (section 2.6.2), manipulation verbs such tsɯm ‘take away’ or ɣɯt ‘bring’ cannot. Examples such as (81), with kɤ- prefixed verb forms appearing before a manipulation verb, could appear to be an example of purposive participial clauses.18 (81) ɯ-mbro ɯ-ndʐi nɯra kɤ-ntsɣe jo-tsɯm 3sg.poss-horse 3sg.poss-skin dem:pl ???-sell ifr-take.away ‘He took the horses’ skins to (the market) to sell them.’ (150814 kelaosi, 85) However, in all examples such as (81), the noun ɯ-spa ‘material’ can be added after the verb prefixed in kɤ- without changing the meaning (kɤ-ntsɣe ɯ-spa jo-tsɯm). This indicates that the syntactic function of kɤ-ntsɣe here is in fact that of an essive adjunct (see section 1.2), meaning literally ‘He took the horses’ skins (there) as something to sell’, and that it should be analyzed not as an infinitive form, but as an object participle meaning ‘which is to be sold.’19 This construction markedly differs from the purposive participial clauses of motion verbs in not being fully grammaticalized. 2.6.4 Action nouns The transitive verb kʰɤt ‘do repeatedly, do a long time’ and its causative sɯ-kʰɤt ‘cause to do repeatedly, cause to do a long time’ can take infinitival complements, but they more commonly occur in a construction with instrumental-like noun phrases marked with the ergative / instrumental kɯ, indicating the action which is performed repeatedly or done over a long time. These noun phrases can include either an action nominal derived from a verb with the prefix tɯ- as in (82), or a action noun from which verbs are derived by denominal prefixes, as (83).20 18 It cannot be a P-participle, as this would require coreference between the subject of the matrix clause and the object of the participial clause, see section 2.6.2. 19 An anonymous reviewer suggested the alternative possibility of viewing this infinitival form as an adverbial clause. While infinitives are indeed used as converbs in Japhug (see Jacques 2014a, 321-2), they are used in manner clauses, not purposive clauses. 20 The noun ta-ma ‘work’ in (83) is the base from which several common verbs such rɤma ‘work, do work (vi)’ and nɤma ‘work on (vt)’ are derived. 30 (82) tɯ-qioʁ kɯ tó-wɣ-sɯ-kʰɤt ʑo tɕe, nmlz:action-vomit erg ifr-inv-caus-do.a.long.time emph lnk tɕe nóʁmɯz nɤ tɯɣ nɯnɯ ló-wɣ-sɯ-tɕɤt lnk only.then lnk poison dem ifr-inv-caus-take.out ‘(The medicine) caused (Gesar) to vomit a long time until he expelled the poison.’ (Gesar, 266) (83) ta-ma kɯ ta-kʰɤt ʑo indef.poss-work erg pfv:3→3’-do.a.long.time emph ‘He did a lot of work.’ Since the action nominals in tɯ-, have a neutralized valency, this construction cannot be considered to be a type of complement clause. 2.6.5 Compound action nouns Beside action nouns derived from verbs by means of a nominalization prefix, Japhug is also rich in compound action nouns (either noun-verb or verb-verb, see Jacques 2012b). Used in combination with light verbs, some compound nouns form complex predicates with very specific meanings. The most frequent construction of this type involves compounds comprising the noun kʰramba ‘lie, cheating’ as the first element and a verb root as the second element. The combination of these compounds with the light verb βzu ‘do, make’, has the meaning ‘pretend to do X’ as shown by example (84) (see sections 2.6.2 and 4.6 for other constructions with a similar meaning). (84) ʑara kɯ cʰa nɯ kʰramba-tsʰi ka-βzu-nɯ, 3pl erg alcohol dem lie-drink ntsɯ ka-znɤrko-nɯ ɲɯ-ŋu tɕʰeme nɯra pfv:3→3’-do-pl girl dem:pl always pfv:3→3’-force-pl sens-be ‘They pretended to drink alcohol, and forced the women (to drink).’ (Slobdpon05, 100) In this construction, the light verb βzu ‘do, make’ takes the orientation prefix selected by the verb included in the compound (similar phenomena are observed with complement clauses, see section 3.4). In example (84) for instance, the lexeme tsʰi ‘drink’ selects prefixes with the ‘toward east’ orientation in finite forms and some participial forms (Perfective/Irrealis kɤ-, Perfective 3→3’ ka-, Imperfective ku- etc). The compound kʰramba-tsʰi cannot take any orientation prefix (or any verb morphology, for that matter), but the verb form ka-βzu-nɯ inherits the orientation selected by tsʰi ‘drink’ and thus appears with the Perfective 3→3’ ka- corresponding to the orientation 31 ‘toward east’.21 Another similar construction involves compounds noun with the adverb kɯzɣa ‘for a long time’ as first element:22 compare examples (85) and (86) (with raising of the Inferential ‘toward west’ orientation prefix ɲɤ- on the light verb). (85) kɯzɣa ʑo ɲɤ-ɕar long.time emph ifr-search ‘He looked for him for a long time.’ (elicited) (86) kɯzɣɤ-ɕar ʑo ɲɤ-βzu long.time-search emph ifr-do ‘He looked for him for a long time.’ (elicited) 2.6.6 Serial verb constructions In Japhug, as in Tshobdun (Sun 2012, 490-1), we find a serial verb construction comprising two verbs sharing TAM category, core argument(s) (both subject and object in the case of transitive verbs) and transitivity. One of the verbs expresses the main action, and the other describes the manner in which the action is performed. Unlike Tshobdun, there is no constraint in Japhug against inserting a linker such as tɕe between the two verbs in the serial construction. This construction is most common with deideophonic verbs,23 as exemplified by (87), where nɯdrɯβ ‘gore again and again’ can only be used in this construction together with tɕʰɯ ‘gore’. The ideophonic verb can either follow (87) or precede the main verb (88), the latter construction being by far more common. Some non-ideophonic verbs expressing manner (such as Japhug nɤxɕɤt ‘do with force’, like its Tshobdun cognate nɐʃeʃet ‘exert oneself’) can also be used in this construction. (87) iɕqʰa srɯnmɯ nɯ to-tɕʰɯ to-nɯdrɯβ tɕe the.aforementioned râkshasî dem ifr-gore ifr-repeatedly.gore lnk pjɤ-sat ifr-kill ‘(The rhinoceros) gored the râkshasî repeatedly and killed her.’ (28smAnmi, 403) (88) srɯnmɯ nɯ to-nɯdrɯβ ʑo to-tɕʰɯ râkshasî dem ifr-repeatedly.gore emph ifr-gore 21 The verb βzu ‘do, make’ when used as a full verb most commonly takes the orientation ‘up’, never ‘toward east’. 22 It takes the status constructus form kɯzɣɤ- in these compounds. 23 On deideophonic verbs and their morphosyntactic properties, see Sun & Shidanluo (2004) and Jacques (2013c). 32 ‘(The rhinoceros) gored the râkshasî repeatedly and killed her.’ (elicited on the basis of 87) The second most common type of serial verb construction in Japhug involves the manner deixis verbs stu ‘do like this’ (transitive) and fse ‘be like this’ (intransitive). Examples like (89) could seem to indicate that stu ‘do like this’ and the lexical verb do not share the same object, as ki ‘this’, which obligatorily occurs before the manner deixis verbs, appears to be its object. (89) ɯ-ru nɯ ki tú-wɣ-stu pjɯ́ -wɣ-qlɯt 3sg.poss-stalk dem dem:prox ipfv-inv-do.like ipfv-inv-break ‘One breaks its stalk like this.’ (14-tasa, 81) However, when the lexical verbs takes a non-third person object, the manner deixis verb indexes it as its object too, as in (90): stu ‘do like this’ is in fact a secundative ditransitive verb, and the demonstrative is an unmarked T argument. (90) aʑo kɯki ntsɯ kú-wɣ-stu-a-nɯ tɕe, 1sg dem:prox always ipfv-inv-do.like-1sg-pl lnk kú-wɣ-znɯkʰrɯm-a-nɯ ipfv-inv-punish-1sg-pl ‘They punished me like this.’ (Gesar, 278) The verb stu ‘do like this’ cannot be used with intransitive verbs in a serial construction. Instead, its intransitive counterpart fse ‘be like this’ occurs with a demonstrative such as ki as in (91). (91) aʑo nɯ sŋiɕɤr ʑo kutɕu ki fse-a 1sg dem night.and.day emph here dem:prox be.like:fact-1sg ndzur-a ntsɯ ɲɯ-ra tɕe, stand:fact-1sg always sens-have.to like ‘I have to stand like this night and day.’ (The divination, 2002, 44) Third, some verbs which usually take complements are also compatible with serial verb constructions, including the causative form of adjectives (see section 4.10.2) as well a phasal and modal verbs like ʑa ‘begin’ and cʰa ‘can’. These serial verb constructions differ from the corresponding complement constructions in that the verbs cannot take either negative or associated motion prefixes. The phasal verb ʑa ‘begin’, though most commonly used with bare infinitive or tɯ- infinitive complements, can also appear in a serial verb construction expressing the specific meaning ‘start doing X from ... until...’, as in (92). 33 (92) tɕe βzɯr ri tɕe cʰɯ́ -wɣ-mpʰɯr cʰɯ́ -wɣ-ʑa tɕe lnk corner loc lnk ipfv-inv-wrap ipfv-inv-begin lnk mɤpɕoʁ cʰu βzɯr nɯ-ɕki mɤɕtʂa cʰɯ́ -wɣ-mpʰɯr opposite.side loc corner 3pl-dat until ipfv-inv-wrap ‘One starts to wrap it up from one corner until the opposite corner.’ (30-mboR, 20) Likewise, the modal verb cʰa ‘can’ occurs in a serial verb construction, as in example (93), which slightly differs from the previous ones in allowing an intransitive verb to combine with a transitive one (one possible factor for this exception is that cʰa ‘can’ is semi-transitive, and thus presents morphosyntactic features of both transitive and intransitive verbs). (93) tarmgɯnku ɯ-ʑɯβdaʁ nɯnɯ kɯ-mɯrkɯ Dar.mgon 3sg.poss-tutelary.spirit dem nmlz:S/A-steal cʰɤ-ɕe. tɕe tɯ-ci, nɤki tɯ-tsʰɤʁrɯ ifr:downstream-go lnk indef.poss-water dem one-goat.horn nɯ, ɯnɯnɯ lo-mɯrkɯ pjɤ-cʰa. dem dem ifr-steal ifr-can ‘The tutelary spirit of Darmgon went to steal it, and the goat horn full of water, he succeeded in stealing it.’ (02-montagnes-kamnyu-cz) Finally, there is a closed class of idiomatic serial verb constructions whose meaning is not predictable from that of the two verb roots (either because meaning is not compositional, or because one of the verb roots only appear in the serial construction, never independently). The most common example is stu+mbat ‘try hard’, as in (94). (94) tɤ-stu tɤ-mbat je imp-try.hard(1) imp-try.hard(2) hort ‘Try hard!’ 2.6.7 Coordination In Japhug, some attitudinal verbs such as ʁnɯ ‘suspect’, nɯsɯmɲiz ‘hesitate’, nɯʁlɯmbɯɣ ‘guess, estimate’ or nɯʁjɯβtsʰɤt ‘guess, estimate’ do not take complement clauses. Rather, they occur in a coordinating construction strikingly similar to that described in Tshobdun by Sun (2012, 487-8): the attitudinal verb is followed by the affirmative copula ɕti ‘be’ and an adversative linker such as ri ‘but’, as in (95). (95) nɯ tu-nɯʁlɯmbɯɣ-a ɕti nɯ sɤpe ri, ɯʑo kɯ kɤ-nɤma dem ipfv-guess-1sg be.affirm:fact lnk 3sg erg nmlz:P-work dem do.well:fact ‘I guess that he will perform this task well.’ 34 3 Morphosyntactic clauses properties of complement This section discusses various topics related to the syntax of complement clauses in Japhug, including word order, syntactic pivots (coreference restrictions and restrictive neutralization), demonstratives as possible complementizers, and case marking mismatch. 3.1 Word order and constituency Complement clauses in Japhug have the same position in the sentence as core arguments: they are strictly pre-verbal. While complement clauses are generally located directly before the verb, we find examples in which the subject appears between the complement clause and the (transitive) main verb, as in (27) above. Discontinuous complement clauses are rare in Japhug. The only clear example in my corpus is (96). In this example, the 1sg pronoun aʑo (the subject of the matrix clause, which has no syntactic role in the complement clause) appears between the A lɯlu kɯ ‘the cat’ and the P ʁnɯz ‘two’ of the complement clause. Despite the rarity of this construction, this sentence was not considered to be unusual by my consultant when listening again to the recording. (96) tɕe [lɯlu kɯ aʑo ʁnɯz ʑo ka-ndo] pɯ-mto-t-a lnk cat erg 1sg two emph pfv:3→3’-take pfv-see-pst:tr-1sg ‘I saw a cat catching two of them.’ (22-kumpGatCW, 61) 3.2 Syntactic pivots Table (2) presents a summary of coereference restrictions between matrix and complement clauses in Japhug, based on the data in section (2). Some verb classes are named by a representative example (for instance rga ‘like’) because at this stage of research, it is not yet clear to what extent the classes of all verbs with the same behaviour can be given a simple functional label. 35 Table 2: Coreference restrictions in complement clauses in Japhug Verb class Complement type motion verb kɤ- participle motion verb transitive verb transitive verb spa ‘be able’ cʰa ‘can’ sɯxcʰa ‘(cause to) be able’ kɯ- participle bare infinitive tɯ- infinitive infinitive, finite infinitive, finite infinitive rga ‘like’ rga infinitive finite cognition, perception impersonal impersonal impersonal finite bare infinitive tɯ- infinitive finite, infinitive Coreference complement = main clause {P}={S} {S,A}={S} {A}={A} {S}={A} {S,A}={A} {S,A}={S} {S,A}={P} {S,A,P,P’}={S} {S,A,P}={S} no constraint zero zero zero This table confirms the observation that although Gyalrong languages have ergative case marking, syntactic pivots mainly follow an accusative alignment, with restrictive neutralization of S and A (Van Valin & LaPolla 1997, 275, Sun 2003; Jacques 2016b). Note that the construction with motion verbs and kɤ- participial complements, despite showing obligatory coreference between the intransitive subject of the main clause and the object of the complement (S=P), cannot be considered to have an ergative-absolutive pivot, since the same construction cannot express coreference between the intransitive subject of the main clause and that of the complement clause (S=S). Some verbs such as rga ‘like’ present looser coreference restrictions, with differing rules depending on the construction. It is possible that additional subtypes will be revealed by finer examination of the behaviour of individual complement-taking verbs. 3.3 Plural and demonstrative markers Finite, infinitival or participial complements in Japhug can be optionally followed by the distal demonstrative nɯ ‘that’ (97), the plural ra (98) or a combination of the two nɯra ‘those’. In example (97), nɯ ‘that’ has a topicalizing function (‘as for growing high, it can grow high, but on the other hand it cannot grow thick’). 36 (97) tu-mbro nɯ ɲɯ-cʰa ri tu-mbro tɕe ʁnɯ-rtsɤɣ, ipfv-be.high dem sens-can lnk ipfv-be.high lnk two-stair χsɤ-rtsɤɣ jamar tu-mbro ɲɯ-cʰa ri three-stair about ipfv-be.high sens-can lnk ‘Although it can grow high, although it can grow two or three stairs high,...’ (16-CWrNgo, 151) The marker ra is an associative plural; in examples such as (98), the use of ra implies an open list of activities (‘crawl, walk etc’). (98) kɤ-nɯrtsɯ kɤ-ŋke ra tɤ-cʰa tɕe inf-crawl inf-walk pl pfv-can lnk ‘When (the baby) becomes able to crawl or to walk, ...’ (140426 tApAtso kAnWBdaR, 65) In Tshobdun, Sun (2012, 481) analyzes these demonstratives as complementizers, an analysis which would imply a grammaticalization pathway identical to that of English ‘that’. No attempt will be made to solve this complex question in the present article. An argument for the special status of demonstratives with complement clauses is that they can only be postclausal, whereas in the case of relatives (or any noun phrase) demonstratives can be pre-clausal or circum-clausal (see section 2.6.1). Demonstrative and plural markers are not the only grammatical elements that can follow a complement clause. For instance, to express a restriction (‘only’) having scope over a complement clause, the postposition ma ‘apart from’ is used after the complement, sometimes with the postposition repeated two times [X ma nɯ ma] ‘apart from X, apart from it’ as in example (99). (99) kɤ-mtsʰɤm ma nɯ ma inf-hear apart.from dem apart.from mɯ-pɯ-rɲo-t-a neg-pfv-experience-pst:tr-1sg ‘I only heard about it.’ (I did not see it and even do not claim that it exists, of a mythological animal) (20-RmbroN, 118) 3.4 Orientation prefixes In Japhug, all finite verb forms except the Factual take an orientation prefix.24 For some categories (egophoric present, sensory, past imperfective), 24 See for instance Jacques (2014a, 265-9) for a brief description of the system. In Japhug as in all Gyalrong languages, six orientations are possible: up’, down’, ‘upstream’, ‘downstream’, ‘east’ and ‘west’, to which can be added the ‘unspecified orientation’ prefix used with motion verbs. Only three verbs have defective conjugations and lack orientation prefixes (Jacques 2012b). Similar systems of orientation prefixes are widespread in the area (Sun 1983; Lin 2002; Shirai 2009; Lin 2011). 37 all verbs take the same orientation prefix (respectively the directions for ‘east’, ‘west’ and ‘down’). In the rest of the conjugations, each verb selects one or more orientations, and consistently uses it to build all TAM forms. Table 3 illustrates some examples of lexically selected orientation prefixes. The forms indicated are third person singular (intransitive verbs) and 3→3’ (transitive verbs). Table 3: Examples of lexically selected orientation prefixes in Japhug Base form Orientation Perfective Inferential Imperfective ndza ‘eat’ (tr) fɕɤt ‘tell’(tr) fsoʁ ‘be day’ (intr) nɯrɤɣo ‘sing’ (intr) tsʰi ‘drink’ (tr) ɕar ‘search’ (tr) Up Down Upstream Downstream Toward east Toward west ta-ndza pa-fɕɤt lɤ-fsoʁ thɯ-nɯrɤɣo ka-tsʰi na-ɕar to-ndza pjɤ-fɕɤt lo-fsoʁ cʰɤ-nɯrɤɣo ko-tsʰi ɲɤ-ɕar tu-ndze pjɯ-fɕɤt lu-fsoʁ cʰɯ-nɯrɤɣo ku-tsʰi ɲɯ-ɕar In non-finite complement clauses, verbs rarely take orientation prefixes (they are even impossible in the case of bare infinitives and participial complements). Most complement-taking verbs select one or two orientation prefixes and use them regardless of the verb in the complement. This is the case of rɲo ‘experience’, which takes the ‘down’ series of prefixes (Inferential pjɤand Perfective pɯ-) irrespective of the verb complement. Thus, in (100), while the verb tsʰi ‘drink’ in the complement clause selects the ‘towards east’ orientation, the main verb rɲo ‘experience’ takes the ‘down’ prefix. (100) ɯ-se kɤ-tsʰi pɯ-rɲo-t-a 3sg.poss-blood inf-drink pfv-experience-pst:tr-1sg ‘I have drunk its blood.’ (27-qartshAz, 106) A minority of complement-taking verbs, including phasal verbs and causative verbs, systematically inherit the lexical orientation of the verb of the complement clause. The following examples show how the verb ʑa ‘start’ takes the orientation prefix of its complements’ verbs, respectively ‘up’ (example 101), ‘down’ (102), ‘upstream’ (103), ‘downstream’ (104) and ‘towards west’ (105). The Inferential forms of these verbs are provided in Table 3 above. (101) ɯ-ndza pɯ-kɤ-ta nɯra 3sg.poss-food pfv:down-nmlz:P-put dem:pl ɯ-ndza to-ʑa 3sg.poss-bare.inf-eat ifr-start ‘(The horse) started eating the food that had been put (there for him).’ (140507 jinniao, 384) 38 (102) tɤtɕɯpɯ nɯ boy kɯ li ɯ-χpi ɯ-fɕɤt dem erg again 3sg.poss-story 3sg.poss-bare.inf:tell pjɤ-ʑa ifr-start ‘The boy told her again a story.’ (140517 buaishuohua, 69) (103) tɯ-fsoʁ lo-ʑa tɕe, inf-be.clear ifr-start lnk ‘The light of day started to appear.’ (140511 1001 yinzi, 39à (104) pɣɤtɕɯ nɯ bird kɯ nɯɕɯmɯma ʑo tɯ-nɯrɤɣo cʰɤ-ʑa dem erg immediately emph inf-sing ifr-start ‘The bird immediately started singing.’ (140514 huishuohua de niao, 221) (105) nɯnɯ ɯ-ɕar ɲɤ-ʑa-nɯ dem 3sg.poss-bare.inf-search ifr-start ‘They started searching for it.’ (140518 jinyin chengbao, 59) The raising of the orientation prefix of the verb in the infinitival clause is observed with verbs such as ɣɤβdi ‘do well’ or ɣɤtɕʰom ‘do too much’ (derived from the adjectives βdi ‘be good’ and tɕʰom ‘be too much’ respectively; see Jacques 2015, 184). (106) cʰa kɤ-tsʰi ko-ɣɤ-tɕʰom alcohol inf-drink ifr-caus-be.too.much ‘He drunk too much alcohol.’ (elicited) A related phenomenon occurs in light verb constructions with verbs such as βzu ‘do, make’, lɤt ‘throw’ or ti ‘say’. When these verbs form complex predicates with their objects, they take the same orientation as the corresponding verb derived by denominal derivation (Jacques 2012b, 1220). For instance, both the denominal verb nɯrɤɣo ‘sing’ derived from rɤɣo ‘song’ and the corresponding collocation rɤɣo,βzu ‘sing’ (‘make a song’) select the ‘downstream’ orientation (Inferential cʰɤ-, Imperfective cʰɯ- etc), as shown by examples (107), (108), (109). (107) nɯnɯ tɕʰeme nɯ dem girl kɯ li rɤɣo cʰɤ-βzu dem erg again song ifr-make tɕe, ‘The girl sung again.’ (140428 mu e guniang, 167) (108) nɯ-rɤɣo cʰɯ-lat-a tɕe, nɯʑora pɯ-rɟaʁ-nɯ 3pl.poss-song ipfv-throw-1sg lnk 2pl imp-dance-pl ‘I will play a song for you, dance!’ (140513 mutong de disheng, 100) 39 (109) rɤɣo kɯ-mpɕɯ~mpɕɤr ʑo cʰɤ-ti. song nmlz:S/A-emph~be.beautiful emph ifr-say ‘It sung a beautiful song.’ (140519 yeying, 78) An extension of this phenomenon occurs in the compound action noun complementation strategy (section 2.6.5). 3.5 Case marking When the verb of the matrix and the complement clauses sharing the same subject have distinct transitivity values, the subject noun phrase can either take absolutive or ergative marking. Examples (110) and (111) provide a minimal pair illustrating this optional treatment. In both examples, the matrix verb rga ‘like’ is semitransitive (and its subject cannot take ergative marking), while ndza ‘eat’ is transitive (and requires a subject with the ergative). In example (110), the common subject paʁ ra ‘pigs’ takes the ergative kɯ selected by the verb ndza ‘eat’ in the complement clause, suggesting that it should be analyzed as belonging to the complement clause. (110) [paʁ ra kɯ kɤ-ndza] wuma ʑo rga-nɯ pig pl erg inf-eat very emph like:fact-pl ‘Pigs like to eat it.’ (12 ndZiNgri, 149) In (111), the subject fsapaʁ ra ‘domestic animals’ has no ergative marking, a difference which can be accounted for by assuming that the complement clause in this example is restricted to the sole infinitive verb form kɤ-ndza to eat’. (111) fsapaʁ ra [kɤ-ndza] wuma rga-nɯ animals pl inf-eat very like:fact-pl ‘Domestic animals like to eat it.’ (hist-19-qachGa mWntoR, 116) 3.6 Complements of participles When a complement-taking verb is itself in the S/A-participle form, it is possible for the complement either to be in the expected form (infinitive or finite), or to be in S/A-participle form itself, as in example (112). (112) [rŋɯl kɯ ndʑi-kɯ-sɤndu] kɯ-cʰa silver erg 3du-nmlz:S/A-exchange nmlz:S/A-can kɯ-fse pɯ~pɯ-tu nɤ nmlz:S/A-be.like cond~pst.ipfv-exist if ‘If there was someone who could redeem (the life of two brothers) with money, ...’ (140507 jinniao, 339) 40 Such examples are rare, but not considered to be mistakes by consultants when listening again to the recordings. 4 A classification of complement-taking verbs 4.1 Modal verbs This section includes complement-taking verbs expressing deontic and epistemic modality, to the exclusion of verbs of volition (want, wish etc) and attitudinal verbs, which are classified among verbs of cognition and speech. Table 4 presents the list of all the auxiliary verbs in this category. They can take both finite and infinitival complement clauses – since all verbs in this category are relatively uniform with regards to the type of complements they can take, complement types are not indicated in this table. The impersonal verbs are not compatible with person or number marking, and take the complement clause as their subject. A detailed study of the fine semantic differences between these verbs goes beyond the topic of the present article. More detailed data is provided for three of them, spa ‘be able, know how to’ (section 4.1.1), sɯxcʰa ‘be able’ (4.1.2) and ra ‘have to’ (4.1.3). Table 4: Inventory of modal verbs in Japhug Meaning Transitivity Coreference cʰa spa sɯxcʰa sna kʰɯ can know how to (cause to) be able be worthy of, be usable to be possible semi-tr tr inverse stative stative S/A=S S/A=A S/A=P S/A/P=S S/A/P=S, zero ɬoʁ ra ʁzi ntshi mna ŋgrɯ zgɤt jɤɣ have to have to need be preferable be preferable succeed should be possible, be authorized imp. imp. imp. imp. imp. imp. imp. imp. zero zero zero zero zero zero zero zero 4.1.1 spa ‘be able to, know how to’ and cʰa ‘can’ The verb spa ‘be able to, know how to’ originates from the abilitative form of the verb pa ‘do’ (Jacques 2015) and has a cognate in Tangut (Jacques 41 2014c, 255-6), showing that its lexicalization occurred even earlier than proto-Gyalrongic. The verb spa ‘be able to’ can have a noun phrase as its object, as in (113). (113) ɕoŋβzu ko-spa carpentry ifr-be.able ‘He learned carpentry.’ (elicited) As a complement-taking verb, spa ‘be able to, know how to’ takes both infinitival (examples 114 and 115) or finite (116) complements, and its A is coreferential with the S or the A of the complement clause. The subject coreference is required, and no example of coreference with objects, adjuncts or other elements of the complement clause have been observed with this verb. (114) nɯ ɯ-mdoʁ nɯ aj kɤ-ti mɯ́ j-spe-a dem 3sg.poss-colour dem 1sg inf-say neg:sens-be.able[III]-1sg ‘I am not able to name its colour.’ (06-qaZmbri, 57) (115) kɤ-nɤre ɯ-tá-spa? inf-laugh q-pfv:3→3’-be.able.to ‘Is he now able to laugh?’ (conversation, 2014, of a three month old infant) (116) ɯ-tʰoʁ nɯ kɯ-fse, sɤtɕʰa nɯ 3sg.poss-ground dem inf:stat-be.like ground dem ju-rɤtɣe kɯ-fse qʰe, ipfv-measure.handspan.by.handspan inf:stat-be.like lnk tu-ŋke ma nɯ ma ɲɯ-nɯqambɯmbjom ra ipfv-walk apart.from dem apart.from ipfv-auto-fly mɯ́ j-spe pl neg:sens-be.able[III] ‘It is only able to move on the ground as if measuring it handspan by handspan, it cannot fly.’ (26-qambalWla, 79) The verb cʰa ‘can’ occurs in the same construction as spa ‘be able to, know how to’, except that it is a semi-transitive verb (section 1.2.1), whose subject is not marked with the ergative and whose semi-object is not indexed by verb morphology. 4.1.2 sɯxcʰa ‘(cause to) be able’ The verb sɯxcʰa ‘(cause to) be able’ is the causative derivation of cʰa ‘can’. It can take both an object and an infinitival complement.This verb nearly 42 exclusively occurs in inverse form with a non-overt causer (subject), and has the specific meaning of ‘(cause to) be physically able to X’, ‘(cause to) be strong enough to X’, as in examples (117) and (118). The subject is not expressed, but implicitly as the heavy object in (117) and the animal taken by the eagle in (118). (117) ɲɯ-rʑi tɕe kɤ-fkur sens-be.heavy lnk inf-carry.on.the.back mɯ́ j-tɯ-wɣ-sɯx-cʰa. neg:sens-2-inv-caus-can ‘It is heavy, you won’t be able to carry it on your back.’ (elicitation) (118) tʰɯ-wxti-nɯ tsa tɕe tɕe, ta-tsɯm tɕe pfv-be.big-pl a.little lnk lnk pfv:3→3’-take.away lnk tu-ɣɤrʁɤβjɤβ ɲɯ-ɕti tɕe ɯʑo kɤ-tsɯm ipfv-struggle sens-be.affirm lnk 3sg inf-take.away mɯ-ɲɯ́ -wɣ-sɯx-cʰa. neg-sens-inv-caus-can ‘When (piglets, lamb) have grown up, when (the eagle tries to) take away (one of them), it struggles and (the eagle) is not strong enough to take it away. (150819 RarphAB, 6) The object (causee) of sɯxcʰa ‘(cause to) be able’ is coreferential with the subject of the complement clause.25 Such a coreference restriction is extremely rare in Japhug, but it is the logical consequence of the fact that the base verb cʰa ‘can’ has subject coreference restriction, and that the causative derivation demotes the subject of the base verb (causee) to object status. Exceptionally, this verb is attested in direct forms, as in (119), but only in its use as a plain transitive verb ‘cause to be able to bear’, and does not take infinitival complements. (119) kumpɣa pʰu fowl nɯ ɲɯ-βʁa tɕe, mu nɯra male dem sens-win lnk female dem:pl mɯ-ɲɯ-sɯx-cʰe neg-sens-caus-can[III] ‘(Otherwise) the roosters are too strong, and and the hens cannot bear it. (150819 kumpGa, 9) 4.1.3 ra ‘have to’ The modal verb ra ‘have to, need’ cannot take any person or number marker. Its subject is either a noun phrase or a complement clause. If the intransitive 25 In these examples the subject of sɯxcʰa ‘(cause to) be able’ is coreferential with the object of the clause, but this is not a requirement. 43 subject of ra ‘have to, need’ is a noun phrase, the experiencer is marked with the genitive case, as in (120). In complement clauses however, the experiencer does not received special marking, as in (122). (120) aʑɯɣ ɯ-ɕa ra ma nɯ ma 1sg:gen 3sg.poss-meat have.to:fact apart.from dem apart.from kɯ-ra me nmlz:S/A-have.to not.exist:fact ‘I need its meat, I don’t need anything else.’ (02-deluge2012, 14) The verb ra ‘have to, need’ is compatible with either finite (see 46 or 58 above) or infinitive (example 121) complements. Other types of complements or complementation strategies are not attested. (121) tɤ-pɤtso nɯ, tɯ-pɤrme roro jamar tɕe tɕe indef.poss-child dem one-year.old over about lnk lnk tɯ-nɯ kɤ-sɯ-βde pjɤ-ra indef.poss-breast inf-caus-abandon ipfv.ifr-have.to ‘Children had to be weaned at about one year old.’ (140426 tApAtso kAnWBdaR, 83) The verb in the complement clause is almost always either in imperfective or irrealis form. It can be in the Perfective if ra ‘have to’ takes the past pɯprefix, as in (122). (122) tɕe iʑo ji-kʰa kɯnɤ kɤ-fstɯn-i lnk 1pl 1pl.poss-house also pfv-take.care.of-1pl pɯ-ra pst.ipfv-have.to ‘We also had to take care of him at our home.’ (14-tApitaRi, 358) The modal ra ‘have to’ is one of the few verbs that can take one or several complement clauses in the imperative, as in (123). In this construction, ra ‘have to’ appears in the factual form. (123) a-mke kɤ-rqoʁ qʰe a-rca tɤ-ɣi 1sg.poss-neck imp-hug lnk 1sg.poss-following imp:up-come ra have.to:fact ‘Hug my neck and come with me (in the heavens).’ (31-deluge, 109) 4.2 Phasal verbs and other aspectual auxiliaries Aspectual and phasal complement-taking verbs present a much greater variety of constructions than modal verbs. Table 5 summarizes the constructions 44 attested with each verb, not all of which are equally common.26 As shown in section 2.2.2, verbs in this group have different coreference restrictions depending on the complement type. With rɲo ‘experience’ for instance, when used with a bare infinitive / tɯ- infinitival complement, subject coreference is obligatory, while when it takes a kɤ- infinitival complement, coreference with the object or with a possessor of the subject is also possible. Table 5: Inventory of phasal and aspectual auxiliaries in Japhug Verb Meaning I. BI F. Strategy rɲo sɤʑa ʑa stʰɯt sɤtɕɯtʂi nɯftɕaka sɯɣjɤɣ kʰɤt jɤɣ ŋgrɤl rɤŋgat ʑɣɤmɲo mda experience begin, start begin, start finish continue prepare to finish do repeatedly tr. tr. tr. tr. tr. tr. tr. tr. be finished be usually the case be about to prepare to be time to imp. imp. intr. intr. imp. ! ! ! ! ! ! ! ! ! ! ! ! ! serial ! ! ! ! ! ! ! ! serial (2.6.4) participle participle In addition to rɲo ‘experience’, which has been treated in previous sections, data are provided for stʰɯt ‘finish’ (the verb in this group with the greatest diversity of complement types) and ŋgrɤl ‘be usually the case’ (as a representative of impersonal aspectual verbs). 4.2.1 stʰɯt ‘finish’ Of all the complement-taking verbs in Japhug, stʰɯt ‘finish’ is one of those which are compatible with the greatest diversity of complement types. It appears with kɤ- infinitives (124), bare infinitives (125) or in serial verb constructions (126), with respectively 16, 8 and 3 examples in the corpus. (124) tɯ-nɯ kɤ-jtsʰi na-stʰɯt tɕe tɕe indef.poss-breast inf-give.to.drink pfv:3→3’-finish lnk lnk tɤ-pɤtso nɯ li ɯ-sta nɯtɕu ko-ɕɯ-rŋgɯ indef.poss-child dem again 3sg.poss-bed dem:loc ifr-caus-lay 26 The abbreviations are as follows: I. (kɤ- Infinitive), BI (bare infinitive and tɯ- infinitive), F. (finite complement), tr. (transitive), imp. (intransitive impersonal). 45 ‘After she had finished breastfeeding, she put back the child on his bed.’ (140429 jiedi, 270) (125) nɯ (126) nɯra ɯ-ti ta-stʰɯt dem 3sg.poss-bare.inf:say pfv:3→3’-finish ‘When she finished saying that,’(150818 muzhi guniang, 125) pa-βzjoz pa-sthɯt tɕe ɯ-sloχpɯn dem:pl pfv:3→3’-learn pfv:3→3’-finish lnk 3sg.poss-teacher nɯ kɯ taqaβ tɯ-ldʑa ɲɤ́ -wɣ-mbi dem erg needle one-cl ifr-inv-give ‘When he finished learning this (craft), his teacher gave him a needle.’ (140508 benling gaoqiang de si xiongdi, 97) Like ʑa ‘start’ (see section 3.4), stʰɯt ‘finish’ inherits the orientation of the verb in the complement clause. In the examples (124), (125) and (126) above, stʰɯt ‘finish’ respectively takes the ‘towards west’ (Perfective na-), ‘up’ (Perfective ta-) and ‘down’ (Perfective pa-) orientations, which are the orientations lexically selected by jtsʰi ‘give to drink’,27 ti ‘say’ and βzjoz ‘learn’ respectively. 4.2.2 ŋgrɤl ‘be usually the case’ The verb ŋgrɤl ‘be usually the case’, although very common, occurs in a very restricted construction. It cannot take person/number marking, and is only used with finite complements. The verb in the complement clause is nearly always in the imperfective, as in (127), except for existential verbs such as tu ‘exist’, which appear in the factual (128). (127) aʑo kumpɣa cʰɯ-nɯ-χse-a (128) ɕɤr pɯ-ŋgrɤl 1sg chicken ipfv-auto-feed[III]-1sg pst.ipfv-be.usually.the.case ‘I used to raise chickens /(for my own sake).’ (150819 kumpGa, 69) tɕe tu ŋgrɤl, tu-mbri night lnk exist:fact be.usually.the.case:fact ipfv-call ŋgrɤl be.usually.the.case:fact ‘(Owls) appear, howl during the night.’ (22-pGAkhW, 19) 4.3 Verbs of cognition and speech Verbs and complex predicates expressing cognition, perception and speech can be divided into three categories depending on the type of complements they can take. 27 This verb is the causative of tsʰi ‘drink’ which take the ‘towards east’ orientation: despite being etymologically related, these two verbs select distinct orientations. 46 First, some verbs accept both infinitival and finite (including reported speech, see 2.5) complements; this category mainly includes modal verbs of volition, verbs of thought and attitudinal verbs (‘fear’, ‘hate’). Second, some verbs of perception and cognition accept finite complements, but cannot take infinitival complements. Third, some attitudinal verbs and intentional perception verbs (‘look at’, ‘listen to’) cannot take complement clauses, and can only use the coordination complementation strategy (section 2.6.7). Table 6: Inventory of verbs of cognition, perception and speech in Japhug Verb Meaning I. BI F. Strategy sɯso think, want forget like fear hate intend to want to obtain want wish tr tr semi-tr tr tr tr tr colloc colloc ti realize that see hear remember know, understand know hope say tr tr tr tr semi-tr tr tr tr ru sɤŋo nɯsɯmʁɲiz nɯsɯmŋɤn ʁnɯ nɯʁlɯmbɯɣ look at listen to hesitate suspect suspect estimate it lab it it it it jmɯt rga nɯɣmu qʰa ʁmɯɣ nɯmga ɯ-sɯm+ɕe ɯ-ʁjiz+ɣi sɯχsɤl mto mtsʰɤm ɕɯftaʁ tso sɯz nɯmgro ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! coordination coordination coordination coordination coordination coordination 4.4 Motion verbs Two motion verb ɕe ‘go’ and ɣi ‘come’ take participial purposive clauses (see section 2.6.2, example 129 below). Other motion verbs (such as ŋke ‘walk’, rɟɯɣ ‘run’, nɯqambɯmbjom ‘fly’) and verbs of manipulation (ɣɯt ‘bring’ 47 and tsɯm ‘take away’) cannot take purposive clauses, though the latter use relative clauses in essive function as a complementation strategy (section 2.6.3). Japhug has both motion verbs with purposive clauses and associated motion prefixes (venitive ɣɯ- and andative ɕɯ- grammaticalized from ɣi ‘come’ and ɕe ‘go’ respectively). As shown in Jacques (2013b, 203), these two constructions present an obvious semantic contrast when used in the perfective. In the purposive clause construction, when the motion verb is in the perfective, only the motion is implied to have taken place, nothing is said about the action described by the complement. For instance, example (129) makes sense only if the action of the purposive clause has not taken place yet. (129) tɕʰi ɯ-kɯ-nɤma jɤ-tɯ-ɣe? what 3sg-nmlz:S/A-do pfv-2-come[II] ‘What have you come to do?’ By contrast, the associated motion construction implies that both the motion event and the action following it have been completed, as in example (130). (130) ɕɯ ra nɯ-kʰa tɕe ɕ-pɯ-tɯ-rɤ-tʂɯβ tɤ-ti who pl 3pl.poss-house lnk transloc-pfv-2-apass-sew imp-say ra have.to:fact ‘Tell us in whose house you have gone to do sewing.’ (140512 alibaba, 162) 4.5 Attempt The verbs tsʰɤt ‘try’ and rɤtsʰɤt ‘try’ (no discernible semantic difference between the two could be ascertained) are not attested with complement clauses in the corpus. They can occur with a coordinated reported speech (section 2.6.7), as in example (131). (131) tɕe ɲɯ-βʑoʁ-nɯ ɯkʰɯkʰa qraʁ lnk ipfv-carve-pl while ɯ-ɲɯ-ɤ́ tʰoʁmpʰrɤt kɯ nɯnɯ tu-tsʰɤt-nɯ qu-sens-fit nɯnɯ tu-tsʰɤt-nɯ tɕe ploughshare dem ipfv-try-pl lnk ntsɯ ra qu dem ipfv-try-pl always have.to:fact ‘While they shape (the hole to insert the ploughshare), they try the ploughshare, they try it to determine whether or not it fits in.’ (24mbGo, 31) The verbs nɤz ‘dare’ and pʰot ‘dare’ (the second is barely used in the Kamnyu dialect) can be used with both infinitival and finite complements. They are among the few verbs whose infinitival complements require subject 48 coreference. In example (132) with the verb nɤkʰu ‘invite to one’s home as a guest’ (one of the few transitive verbs implying a volitional action of both subject and object, see section 2.2.2), the interpretation ‘I do not dare to go to his house as a guest’ (with coreference of the object of the complement clause and the subject of the main clause) is not possible. (132) ɕɯ-kɤ-nɤkʰu mɤ-naz-a transloc-inf-invite neg-dare:fact-1sg ‘I do not dare to invite him.’ (elicited) 4.6 Pretence The verbs of pretence nɯɕpɯz ‘pretend, imitate’ and ʑɣɤpa ‘pretend’ do not accept complement clauses, they only take participial relatives as their objects or semi-object, as described in section 2.6.2. An alternative construction for expressing pretence involves complex predicates with the auxiliary verb βzu ‘do’ and a compound noun with kʰramba ‘lie, cheating’ as the first element (see 2.6.5). 4.7 Deixis There are two manner deixis verbs in Japhug, the transitive verb stu ‘do like this’ (transitive) and the intransitive fse ‘be like this, look like’. Both can take an essive noun phrase not indexed in the verb morphology (often restricted to a demonstrative). They can be combined with other verbs in a serial verb construction to express the meaning ‘do X like Y’ (2.6.6). (133) ɯ-xɕɤt cʰɯ-lɤt-nɯ ki cʰɯ-stu-nɯ tɕe 3sg.poss-power ipfv-throw-pl this ipfv-do.like.this-pl lnk ‘People insert (bullets in the muzzle of the gun) forcefully like this.’ (28-CAmWGdW, 61) Alternatively, the same meaning can be expressed with an infinitive complement clause, as in (134). (134) tɕe ɕɤmɯɣdɯ kɤ-lɤt nɯ tu-stu-nɯ ɲɯ-ŋu lnk gun inf-throw dem ipfv-do.like.this sens-be ‘They shoot it like that.’ (28-CAmWGdW, 96) 4.8 Negative existential verbs Although Japhug, like other Gyalrongic languages, has a series of negative prefixes (whose form vary depending on the TAM category), negation can also be expressed by post verbal negative auxiliaries like maʁ ‘not be’, me ‘not exist’ or maŋe ‘not exist (sensory)’. Example (135) illustrates both negative forms (negative prefix mɯ-pɯ-mto-t-a vs negative auxiliary pɯ-mto-t-a me). 49 (135) tɕe ɯ-mdoʁ tɕʰi ʑo fse mɤxsi, lnk 3sg.poss-colour what emph be.like:fact neg:genr:know a-kɤ-ti me ma 1sg.poss-nmlz:P-say not.exist:fact lnk mɯ-pɯ-mto-t-a. ɯ-ndʐi kɯnɤ pɯ-mto-t-a neg-pfv-see-pst:tr-1sg 3sg.poss-skin also pfv-see-pst:tr-1sg me not.exist:fact ‘I don’t know what its colour is, I cannot say as I have not seen it. I never even saw its hide.’ (27-kikakCi, 21-22) This situation is not unusual in Sino-Tibetan, and has led in some languages to the complete replacement of pre-verbal by post-verbal negation (see Post 2015). The negative copula maʁ ‘not be’ is used as a negative auxiliary in compound tenses (on which see Jacques 2014a, 268-9), to express focus (example 136) or polar questions (137). (136) nɤʑo nɤ-pʰe tu-ti-a tu-ti-a ɕti maʁ, pɤnma wombɤr ɯ-pʰe 2sg 2sg-dat ipfv-say-1sg not.be:fact p.n. p.n. 3sg-dat ipfv-say-1sg be.affirm:fact ‘I am not talking to you, I am talking to Padma ’Od.’bar.’ (Slobdpon, 190) (137) ‘ɯ-kɤrme tɯ-ldʑa ci nɤ a-mɤ-jɤ-tɯ-ɣɯt ’ 3sg.poss-hair one-cl one even irr-neg-pfv-2-bring tɤ-tɯt-a maʁ ɯ́ -ŋu pfv-say[II]-1sg not.be:fact qu-befact ‘Didn’t I say ‘Don’t bring her back, not even one hair?” (2014kWlAG, 686) The negative existential verbs me ‘not exist’ and maŋe ‘not exist (sensory)’ differ from negative prefixes in that they express emphatic negation, their meaning being translatable with an indefinite negative pronoun or adverb such as ‘nothing’ or ‘never’, as in (135) and (138). (138) kɯmaʁ tu-nɤme-a nɯ-sɯso-t-a me other ipfv-work[III]-1sg pfv-think-pst:tr-1sg not.exist:fact ‘I never thought of doing anything else.’ (150819 woniu, 39) With transitive verbs, this construction is often used to express negation with an indefinite object, but it is not necessarily the case, as in (139). 50 (139) kɤntɕʰaʁ kʰro ju-tsɯm-a street me, kʰa tɕe a.lot ipfv-take.away-1sg not.exist:fact house lnk tu-nɯ-ŋge-a ŋu ipfv-auto-wear[III]-1sg be:fact ‘I don’t take it much on the streets, I wear it at home.’ (conversation 150418) Negative prefixes can be combined with negative auxiliaries to build a double negation construction, as in (140). (140) mɯ-pjɯ-mtam-a me, a-kɤ-tso neg-ipfv-see[III]-1sg not.exist:fact 1sg.poss-nmlz:P-know wuma dɤn really be.many:fact ‘I have seen everything (there is nothing I do not see), I know really a lot.’ (2002qajdoskat, 36) The nature of the finite clauses occurring in this construction is in question. It could seem at first glance that these are relative clauses. Indeed, one of the most common ways of expressing a negative indefinite in Japhug is by using participial relative clauses with a negative existential verb, as in examples (141) and (142). (141) nɤʑo nɯ-nɯ-ɣɤwu ma, nɤ-kɯ-nɯɣ-mu you imp-auto-cry because 2sg-nmlz:S-appl-be.afraid me ma mɤ-ta-mbi fact:not.exist because neg-1→2-fact:give ‘Cry as you wish, nobody is afraid of you, I will not give her to you.’ (The frog, 38) (142) [smɤɣ-ri nɯ ra ɯ-kɯ-ntɕhoz] maka maŋe wool-thread top pl 3sg-nmlz:A-use at.all not.exist:sensory ‘Nobody uses woollen threads.’ (Coloured belts, 89) Since Japhug also has finite relatives, the same analysis could be applicable to examples such as (140) (‘the things I do not see do not exist’). However, in Japhug only objects can be relativized with finite verb forms (Jacques 2016b, 12-3). If this analysis were correct, we would thus not expect to find intransitive verbs negated by negative existential verbs, and negated transitive verbs should only have an indefinite object reading – thus pɯ-mto-t-a me (pfv-see-pst:tr-1sg not.exist:fact) in (135) should only be interpretable as ‘I did not see anything’, not ‘I never saw it.’ Thus, while this construction probably originates from a negated finite relative, in the synchrony of Kamnyu Japhug it is better analyzed as a complement clause. 51 The negative existential verbs me ‘not exist’ and maŋe ‘not exist (sensory)’ also appear in a very unusual construction, [X mɤ-X negation] (where X stands for the bare stem of the verb) meaning ‘whether or not X, it amounts to the same’. This form is not even the bare infinitive, since no possessive prefix can be added. In this construction, contrary to all previous ones, the negative auxiliaries can take person marking, and are obligatorily coreferential with the object if the verb in the complement clause is transitive, as in (143). With intransitive verbs, no person marking appears on the negative verb, as in (144). In this construction, the subject (whether of transitive or intransitive verbs) cannot be expressed. (143) ndza mɤ-ndza me-a eat neg-eat not.exist:fact-1sg ‘Whether (you ) eat me or not, it amounts to the same.’ (sentence obtained as the correction of a sentence I produced to translate a story in Japhug) (144) tɕe ɯ-qiɯ ɲɯ-mtsʰam-a, ɯ-qiɯ lnk 3sg.poss-half sens-hear-1sg 3sg.poss-half mɯ́ j-mtsʰam-a qʰe, ɕe mɤ-ɕe maŋe neg:sens-hear-1sg lnk go neg-go not.exist:sens ‘I can hear half of it, can’t hear the other half, whether or not (I) go it amounts to the same.’ (conversation 140510) 4.9 Analytic causative constructions While Japhug has two productive causative prefixes, ɣɤ- restricted to (some) stative verbs and the general causative (sɯ-/ sɯɣ-/ z- (see Jacques 2015 and Sun 2014; Lai 2014 for a comparative perspective on Tshobdun and Khroskyabs), it also presents no less than two main analytic causative constructions. The first type of analytic causative construction involves an auxiliary (including βzu ‘do, make’, sɤβzu ‘cause to become, make’, sɤpa ‘cause to become, consider’ and tɕɤt ‘take out’) and a subordinate clause with its main verb in subject participle form. This construction is especially common with adjectives, as in (145), and (146). In example (146), both the synthetic causative verb ɣɤwxti ‘make bigger’ and the analytic causative with βzu are attested. (145) cʰɯ-ndɯl-nɯ tɕe kɯ-ndɯ~ndɯβ ʑo cʰɯ-βzu-nɯ ipfv-grind-pl lnk nmlz:S/A-emph~fine emph ipfv-make-pl ‘They grind (tobacco) and make it very fine-grained.’ (30-CnAto, 38) 52 (146) ɯ-pʰɯ ɲɯ-wxti tɕe, nɯra tʰamtɕɤt 3sg.poss-price sens-be.big lnk dem:pl all ma-tɤ-tɯ-ɣɤ-wxti, aʑo nɯ tu-nɯ-χti-a nɯra neg-pfv-2-caus-be.big 1sg dem ipfv-auto-buy[III]-1sg dem:pl stʰɯci kɯ-wxti ɯ-pʰɯ ma-tɤ-tɯ-βze as.much nmlz:S/A-be.big 3sg.poss-price neg-pfv-2-make[III] kɯ-tʂaŋ ci tɤ-βze nmlz:S/A-be.fair a.little imp-make[III] ‘It is too expensive, don’t make it that expensive, I will buy it, don’t make its price that expensive, give it for a fair price.’ (Bargaining 12, 12) With dynamic verbs, examples of this construction are not attested in the corpus. Rather, the preferred construction is to use an impersonal modal verb such as kʰɯ ‘be possible’ or ra ‘have to’ in participial form taking a complement verb, as in (147) and 148 (note that these complements can be either in finite or infinitive form). (147) la-rɤɕi-nɯ tɕe, lu-nɯ-ɬoʁ pfv:3→3’-pull-pl lnk ipfv:upstream-auto-come.out mɤ-kɯ-kʰɯ tu-βzu-nɯ neg-nmlz:S/A-be.possible ipfv-make-pl ‘They pull (on the thread to close the opening) and prevent it from coming out.’ (30-CnAto, 42) (148) ɕɯ-kɤ-βde mɤ-kɯ-ra nɯ ndʑiʑo kɯ transloc-inf-throw neg-nmlz:S/A-have.to dem 2du erg nɯ-tɯ-sɤβzu-ndʑi ŋu pfv-2-cause.to.become-du be:fact ‘Thanks to both of you, there is no need to throw (people in the lake) anymore.’ (2011-05-nyima, 191) In this construction, the causative auxiliary verb takes the participial clause as its object. When the participial clause clause has a first or second person object, as in (149) and (150), the auxiliary remains in third person object form (a form such as †tu-kɯ-sɤβzu-j ipfv-2→1-cause.to.become-1pl ‘you caused us to become X’ would not be possible). (149) iʑora kɤ-nɯʑɯβ mɤ-kɯ-kʰɯ 1pl inf-sleep neg-nmlz:S/A-be.possible tu-tɯ-sɤβze ɲɯ-ŋu ipfv-2-cause.to.become[III] sens-be ‘You prevent us from sleeping.’ (elicitation) 53 (150) a-tɯ-ci ɲɯ-tɯ-s-qarndɯm tɕe aʑo 1sg.poss-indef.poss-water sens-2-caus-be.muddy lnk 1sg tɯ-ci kɯ-ɤmgri kɤ-tsʰi indef.poss-water nmlz:S/A-be.clear inf-drink mɤ-kɯ-khɯ ɲɯ-tɯ-sɤpe ɲɯ-ŋu neg-nmlz:S/A-be.possible sens-2-cause.to.become[III] sens-be ‘You have spoiled my water, you caused me to be unable drink clear water.’ (lang he yang, 26) As shown by examples (148) and (150) in particular, the use of causative auxiliaries with dynamic verbs is common to express indirect causation, the (voluntary or involuntary) indirect result of the action performed by the causer. The verb tɕɤt ‘take out’ is very rare as a causative auxiliary, and only occurs with negative participial forms, as in (151). (151) kɤ-ɤlɯlɤt mɤ-kɯ-ra ɲɤ-tɯ-tɕɤt inf-fight neg-nmlz:S/A-have.to ifr-2-take.out ‘You prevented them from fighting.’ (elicitation, based on the story Nyima ’Odzer) An alternative synthetic causative uses the causative form ɣɤkʰɯ of the verb kʰɯ ‘be possible’. This verb takes infinitive or finite complements as in (152). Like the construction above, the causative auxiliary ɣɤkʰɯ ‘cause to be possible’ takes a third person object form regardless of the person of the complement verb. In example (152) it would not be possible to use the 2sg→1sg form †mɯ-nɯ-kɯ-ɣɤ-kʰɯ-a (ipfv-2→1-cause.to.become-1sg). (152) kɤ-sci mɯ-nɯ-tɯ-ɣɤ-kʰɯ-t inf-be.born neg-ipfv-2-caus-be.possible-pst:tr ‘You prevented me from being born.’ (Gesar, 61) Future research will be necessary to ascertain the precise semantic difference between all causative constructions, a task made difficult by the dearth of examples of the constructions illustrated by examples (151) and (152) in the corpus. 4.10 Complements of adjectives Adjectives in Japhug can be formally defined as the subclass of stative verbs allowing the tropative nɤ- derivation, a derivation that turns a stative verb into a transitive verb meaning meaning ‘find/consider X’ (eg ‘beautiful’ → ‘find/consider to be beautiful’, see Jacques 2013a).28 28 This definition excludes some noun-like property words. 54 4.10.1 Infinitival and finite complements A few adjectives are semi-transitive, like mkʰɤz ‘be expert, be knowledgeable’ and optionally take either a noun (153) or a complement clause (154) addition to their S. The complement clause can be either infinitival or finite, with a verb in the imperfective. (153) ɯ-nmaʁ jɤ-kɯ-ɣe nɯ ɕoŋβzu 3sg.poss-husband pfv-nmlz:S/A-come[II] dem carpentry mkʰɤz tɕe be.expert:fact lnk ‘Her husband (who came to live in her family) is very good at carpentry.’ (14-tApitaRi, 273) (154) tɕiʑo rcanɯ, kɤ-taʁ wuma ʑo mkʰɤz-tɕi 1du unexpected inf-weave really emph be.expert:fact-1du ‘We are very good at weaving.’ (140521, huangdi de xinzhuang, 20) 4.10.2 Causative forms of adjectives The causative forms of adjectives, in addition to their base causative meaning ‘cause to become X’, can also be used with complement clauses to express manner (Jacques 2015, 184). In this construction, both kɤ- infinitives and bare infinitives are possible. For instance, the verb ɣɤ-βdi derived from βdi ‘be good, be well’ can either mean ‘repair, treat’, as in (155) or ‘do well, do properly’ with a complement clause as in (156). (155) a-ʁi kɯ nɯ ma spe 1sg.poss-younger.sibling erg dem apart.from be.able[III]:fact me ri, <tuolaji> kɯ-fse, mkʰɯrlu not.exist:fact but tractor nmlz:S/A-be.like.this machine nɯra tu-ɣɤβdi spe dem:pl ipfv-caus-be.well be.able[III]:fact ‘My brother is only able to do one thing, repair tractors and cars.’ (14-tApitaRi, 166) (156) kʰa ɯ-ʁɤri nɯtɕu ɯ-fkrɤm house 3sg.poss-front.of dem:loc 3sg.poss-bare.inf:place a-kɤ-tɯ-ɣɤ-βdi tɕe, ɕ-pɯ-sɤtse irr-pfv-2-caus-be.good lnk transloc-imp-stick.into[III] ‘Place these in front of your house in orderly fashion and stick them (into the ground).’ (Smanmi 2003, 129) The complement-taking causative verbs inherit the orientation prefix and person marking of the complement verb (see 3.4). 55 Alternatively, causative verbs derived from adjectives can occur in a serial verb construction similar to the one found with manner deixis verbs (see sections 2.6.6 and 4.7), as in example (157) where the causative form of the verb asɯɣ ‘be tight’ shares the same TAM, person and orientation marking as the other member of the serial construction xtɕɤr ‘tie’. (157) tɕe tɯmbri ɲɤ-ɕar lnk rope tɕe ɯ-mŋu nɯra koŋla ʑo ifr-search lnk 3sg.poss-opening dem:pl really emph ko-xtɕɤr ko-sɯ-ɤsɯɣ ʑo ifr-tie ifr-caus-be.tight emph ‘He looked for a rope, and tied the opening (of the bag) very tight.’ (kelaosi, 288) Conclusion This article is the first step towards a comprehensive description of complementation in Japhug, but much remains to be done, in particular in ascertaining fine-grained semantic differences between related constructions. It has three main contributions. First, it confirms the observation from previous research on subordinate clauses in Gyalrong languages (Jacques 2016b, Sun 2003) that while Japhug has ergative case marking, in complex clauses only neutral or accusative pivots are attested, and not a single ergative pivot is found – the only ergatively-aligned construction being generic person marking on the verb (Jacques 2012a). Second, it illustrates how the reanalysis of unmarked relative clauses in essive function (1.2) is a widespread mechanism to create complement clauses (2.6.2, 2.6.3). Third, this article documents several typologically unusual constructions, in particular Hybrid Indirect Speech (section 2.5), the verb sɯxcʰa ‘(cause to) be able’ which only appears in inverse forms (4.1.2) and the puzzling construction [X mɤ-X negation] meaning ‘whether or not X, it amounts to the same’ (4.8) with exclusive indexation of the object of the complement clause on the negative auxiliary. References Aikhenvald, Alexandra. 2008. Semi-direct speech: Manambu and beyond. Language Sciences 30. 383–422. Bickel, Balthasar. 2001. Deictic Transposition and Referential Practice in Belhare. Journal of Linguistic Anthropology 10.2. 224–247. 56 Comrie, Bernard. 1998. Rethinking the typology of relative clauses. Language Design 1. 59–86. Creissels, Denis. 2006a. Syntaxe générale, une introduction typologique, tome 1, Catégories et constructions. Paris: Hermes Science Publications. Creissels, Denis. 2006b. Syntaxe générale, une introduction typologique, tome 2, la phrase. Paris: Hermes Science Publications. Creissels, Denis. 2014. Functive phrases in typological and diachronic perspective. Studies in Language 38(3). 605–647. DeLancey, Scott. 1981. The category of direction in Tibeto-Burman. Linguistics of the Tibeto-Burman Area 6(1). 83–101. DeLancey, Scott. 2011. “Optional” “ergativity” in Tibeto-Burman languages. Linguistics of the Tibeto-Burman Area 34(2). 9–20. DeLancey, Scott. 2015. The Historical Dynamics of Morphological Complexity in Trans-Himalayan. Linguistic Discovery 13(2). 37–56. Dixon, Robert M.W. 2006. Complement Clauses and Complementation Strategies in Typological Perspective. In Robert M.W. Dixon & Alexandra Aikhenvald (eds.), Complementation, 1–48. Oxford: Oxford University Press. Genetti, Carol. 2007. A Grammar of Dolakha Newar. Berlin: Mouton de Gruyter. Genetti, Carol, Ellen Bartee, Alec Coupe, Kristine Hildebrandt & Youjing Lin. 2008. Syntactic aspects of nominalization in five Tibeto-Burman languages of the Himalayan area. Linguistics of the Tibeto-Burman Area 31.2. 97–144. Gong, Xun. 2014. Personal agreement system of Zbu rGyalrong (Ngyaltsu variety). Transactions of the Philological Society 112(1). 44–60. Hagège, Claude. 1974. Les pronoms logophoriques. Bulletin de la Société de Linguistique de Paris 69. 287–310. Jacques, Guillaume. 2008. 嘉絨語研究 Jiāróngyǔ yánjiū (Study on the Rgyalrong language). Beijing: Minzu chubanshe. Jacques, Guillaume. 2010. The inverse in Japhug Rgyalrong. Language and Linguistics 11(1). 127–157. Jacques, Guillaume. 2012a. Argument demotion in Japhug Rgyalrong. In Katharina Haude & Gilles Authier (eds.), Ergativity, Valency and Voice, 199–226. Berlin: Mouton De Gruyter. 57 Jacques, Guillaume. 2012b. From denominal derivation to incorporation. Lingua 122(11). 1207–1231. Jacques, Guillaume. 2013a. Applicative and tropative derivations in Japhug Rgyalrong. Linguistics of the Tibeto-Burman Area 36(2). 1–13. Jacques, Guillaume. 2013b. Harmonization and disharmonization of affix ordering and basic word order. Linguistic Typology 17(2). 187–217. Jacques, Guillaume. 2013c. Ideophones in Japhug Rgyalrong. Anthropological Linguistics 55(3). 256–287. Jacques, Guillaume. 2014a. Clause linking in Japhug Rgyalrong. Linguistics of the Tibeto-Burman Area 37(2). 263–327. Jacques, Guillaume. 2014b. Denominal affixes as sources of antipassive markers in Japhug Rgyalrong. Lingua 138. 1–22. Jacques, Guillaume. 2014c. Esquisse de phonologie et de morphologie historique du tangoute. Leiden: Brill. Jacques, Guillaume. 2015. The origin of the causative prefix in Rgyalrong languages and its implication for proto-Sino-Tibetan reconstruction. Folia Linguistica Historica 36(1). 165–198. Jacques, Guillaume. 2016a. From ergative to comparee marker: multiple reanalyses and polyfunctionality. Diachronica 33(1). 1–30. Jacques, Guillaume. 2016b. Subjects, objects and relativization in Japhug. Journal of Chinese Linguistics 44(1). 1–28. 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. Jacques, Guillaume, Anton Antonov, Yunfan Lai & Lobsang Nima. to appear. Stau. In Graham Thurgood & Randy LaPolla (eds.), The SinoTibetan Languages (2nd edition), London: Routledge. 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. Lai, Yunfan. 2013. La morphologie affixale du lavrung wobzi. Université Paris III MA thesis. Lai, Yunfan. 2014. Causative Constructions in Wobzi. In 14th International Symposium on Chinese Languages and Linguistics, 319–336. Taipei: Academia Sinica. http://iscll-14.ling.sinica.edu.tw/files-pdf/ Papers/Session7/Lai.pdf. 58 Lai, Yunfan. 2015. The person agreement system of Wobzi Lavrung (Rgyalrongic, Tibeto-Burman). Transactions of the Philological Society 113(3). 271–285. Lehmann, Christian. 1984. Der Relativsatz. Typologie seiner Strukturen Theorie seiner Funktionen - Kompendium seiner Grammatik. Tübingen: G. Narr. Lin, Xiangrong. 1993. Jiarongyu yanjiu 嘉戎語研究 (A study on the Rgyalrong language. Chengdu: Sichuan minzu chubanshe. Lin, Youjing. 2002. A Dimension Missed: East and West in Situ rGyalrong Orientation Marking. Language and Linguistics 3.1. 27–42. Lin, Youjing. 2011. Perfective and imperfective from the same source: directional “down” in rGyalrong. Diachronica 28.1. 54–81. Lín, Yòujīng. 2003. Tense and Aspect Morphology in the Zhuokeji rGyalrong Verb. Cahiers de Linguistique - Asie Orientale 32.2. 245–286. Matsumoto, Yoshiko. 1988. Grammar and semantics of adnominal clauses in Japanese: University of California, Berkeley. dissertation. 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. Nikitina, Tatiana. 2012. Logophoric discourse and first person reporting in Wan (West Africa). Anthropological Linguistics 54(3). 280–301. Post, Mark W. 2015. Sino-Tibetan negation and the case of Galo: Explaining a distributional oddity in diachronic terms. Language and Linguistics 16(3). 431–464. Prins, Marielle. 2011. A web of relations: A grammar of rGyalrong Jiaomuzu (Kyom-kyo) dialects: Leiden University dissertation. Shirai, Satoko. 2009. Directional prefixes in nDrapa and neighboring languages: An areal feature of Western Sichuan. In Yasuhiko Nagano (ed.), Issues in Tibeto-Burman historical linguistics (Senri Ethnological Studies 75), 7–20. Osaka: National Museum of Ethnology. Sun, Hongkai. 1983. Liujiangliuyu de minzu yuyan jiqi xishu fenlei 六江流 域的民族语言及其系属分类 (Minority languages of the Six River Valley and their genetic classification). Minzu xuebao 3. 99–273. 59 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 Inflection: 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. 2004. Verb-Stem Variations in Showu rGyalrong. In Ying-chin Lin, Fang-min Hsu Hsu, Chun-chih Lee, Jackson T.-S. Sun, Hsiu-fang Yang Yang & Dah-an Ho (eds.), Studies on Sino-Tibetan Languages: Papers in Honor of Professor Hwang-cherng Gong on His Seventieth Birthday, 269–296. Language and Linguistics Monograph Series W4. Sun, Jackson T.-S. 2006. 草登嘉戎語的關係句 Caodeng Jiarongyu de guanxiju (Relative clauses in the Tshobdun language). Language & Linguistics 7.4. 905–933. Sun, Jackson T.-S. 2007. Morphological causative formation in Shangzhai Horpa. Bulletin of Chinese Linguistics 1(2). 207–227. Sun, Jackson T.-S. 2012. Complementation in Caodeng rGyalrong. Language and Linguistics 13(3). 471–498. Sun, Jackson T.-S. 2014. Sino-Tibetan: Rgyalrong. In Rochelle Lieber & Pavol Štekauer (eds.), The Oxford Handbook of Derivational Morphology, 630–650. Oxford: Oxford University Press. Sun, Jackson T.-S. & Youjing Lin. 2007. Constructional Variation in rGyalrong Relativization: How To Make a Choice? In Pre-Conference Proceedings of the International Workshop on Relative Clauses, 205–226. Taipei: Institute of Linguistics, Academia Sinica. 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. Sun, Jackson T.-S. & Shidanluo. 2004. 草登嘉戎語的狀貌詞 Cǎodēng Jiāróngyǔ de zhuàngmàocí (The ideophones in Tshobdun Rgyalrong). Minzu yuwen 民族語文 5. 1–11. 60 Thurgood, Graham. to appear. Sino-Tibetan: Genetic and areal subgroups. In Graham Thurgood & Randy LaPolla (eds.), The Sino-Tibetan Languages, 2nd Edition, London: Routledge. Tournadre, Nicolas. 2008. Arguments against the Concept of ‘Conjunct’ / ‘Disjunct’ in Tibetan. In Brigitte Huber, Marianne Volkart & Paul Widmer (eds.), Festscrift für Roland Bielmeier zu seinem 65. Geburtstag, 281–308. Halle: International Institute for Tibetan and Buddhist Studies. Van Valin, Robert & Randy LaPolla. 1997. Syntax: Structure, Meaning, and Function. Cambridge: Cambridge University Press. Wei, Jie-Wu. 2001. Verbal prefixation in Zhuokeji Rgyalrong: the problem of kə-/ka- alternation. National Tsing Hua University, Hsinchu MA thesis. 61