S/A detransitivisation in exactly the same form as in Chintang is also found in a number of other Ki- ranti languages. e languages for which syntactically informed reference grammars are available are surveyed in this section.
One interesting generalisation is that S/A detransitivisation appears in a large coherent re- gion in the southeast corner of the Kiranti area. is region is constituted by the languages (from east to west) Limbu, Yakkha, Athpare, Belhare, Chiling, Chintang, Bantawa, and Puma, which are discussed in detail further down. is statement is relativised, though, by the problem that one tra- dition of writing Kiranti grammars has it that these languages have such a complex morphology that their syntax is negligible. is is obviously wrong, as evidenced by the rich literature that is now available on the syntax of, for instance, Chintang, Belhare, and Puma (see e.g. Bickel 2004b, Bickel 2004a, 2007b, 2007a, 2010, Paudy¯al et al. 2010, Gaenszle 2011, Stoll and Bickel in press, 2012, Schikowski et al. forthcoming).
Still, many Kiranti grammars do barely or not at all provide information on syntax, so just because S/A detransitivisation is not mentioned for a language that does not necessarily mean that it does not exist there. One striking example for this is Limbu. S/A detransitivisation has been clearly described for this language as “middle conjugation” by Weidert and Subba (1985) but is nevertheless completely ignored by Driem (1987).
Languages whose status is insecure due to this problem are (again from east to west) Yamphu, Kulung, Dumi, Wambule, Jero, and Sunwar. ese are briefly discussed below.
Rutgers (1998) in his Yamphu grammar does not mention anything like S/A detransitivisation. However, the appended dictionary contains rudimentary information about syntactic verb classes in the form of the labels “v.intr.” and “v.tr.”. Interestingly, the majority of verbs which are labelled as transitive at the same time carry the label “v.intr.”. Verbs with both labels seem to be either S/A or S/O ambitransitive. e two classes can be distinguished by the placement of labels. Only for S/O ambitransitives are the labels repeated before each sense (apparently following the intuition that intransitive ‘break’ and transitive ‘break’ (with S=O) are semantically more different than intransitive and transitive ‘cut’ (with S=A)). Compared to the ubiquitous S/A ambitransitives, S/O ambitransitives seem to be somewhat rarer. Among the S/A ambitransitives are not only classical cases like uŋma ‘drink’ or cama ‘eat’ but also less typical verbs like aŋma ‘chop up’, hamma ‘ration’ or khupma ‘scratch’. I therefore consider Yamphu as a likely candidate for another language having S/A detransitivisation, which would extend the S/A detransitivisation area further to the north.
For Kulung, Tolsma (2006) does also not talk about syntactic verb classes or valency. Differently from Rutger’s Yamphu grammar, his dictionary appendix also does not contain ambitransitives. On p.137 there is one example that looks as if the verb was detransitivised (glosses adapted here and in the following citations from grammars; translation unchanged):
(267) Cʰam-ci song-ns so also ləi-ya-ke. sing-1p[iS].NPST-ASS
‘Let’s also sing songs.’ (Tolsma 2006:137)
However, cʰam ləima ‘sing a song’ might have special properties as in Chintang where it belongs to the class of conventional V/N combinations (section 2.6.5.2). In fact, several other examples with apparently non-specific objects have bipersonal agreement:
(268) Samkʰe
potato soalsolet-a-m.plant-3O-[NPST.]1p[i]A
‘We plant potatoes.’ (Tolsma 2006:151)
All in all the evidence thus is against S/A detransitivisation.
Van Driem states in his grammar of Dumi (Driem 1993:228) somewhat enigmatically that “the concept of transitivity is versatile” but then discusses only a few verbs with diverse syntactic prop- erties none of which comes close to S/A detransitivisation. His chapter on transitivity is long but unsystematic, and since he also overlooked S/A detransitivisation in Limbu (Driem 1987) it is probably a bit early to conclude that Dumi does not have S/A detransitivisation.
By contrast, it seems relatively safe to say that Wambule does not have S/A detransitivisation. Opgenort (2004:151) mentions differential agent marking conditioned by the markedness of tran- sitive scenarios and thus seems to be aware of differential marking in general. In his discussion of verb classes he mentions what he calls “middle verbs” (p.250), but these have the frame {A-ERG P-NOM V-s(A)}. ey are also functionally quite different from S/A detransitivisation as it is found in Chintang and other eastern Kiranti languages in that they mark reflexivity.
Another completely unclear candidate is Jero. Allen (1975:42) mentions that a few verbs can be used both transitively and intransitively, but only one of his examples is clearly S/A ambitransitive (hut- ‘fly’ or ‘fly to somebody’). Opgenort (2005) also does not provide more detailed information. Finally, Sunwar as described by Borchers (2008) is very unlikely to have S/A detransitivisa- tion. Sunwar occupies a special position among the Kiranti languages because the older bipersonal agreement has broken down and made way for monopersonal agreement with S or A. Borchers mentions that a few verbs such as cīcā ‘wash, bathe’ and mecā ‘vomit’ can be used with both in- transitive and transitive inflection (that is, either with the verbal affixes normally indexing S or with those normally indexing A) and says that the use depends on the presence of an object. How- ever, she also emphasises that these verbs are few compared to the rest, which are always used with either intransitive or transitive inflection, so Sunwar seems to have lexicalised S/A ambitransitiv- ity but nothing more. Genei (1988) describes an earlier stage of the language where bipersonal agreement was still in use but focusses on morphology and morphophonology and does not give any information about alternations.
2.9. S/A DETRANSITIVISATION IN OTHER KIRANTI LANGUAGES Apart from these problematic languages, there are a few more western languages which are very unlikely to have S/A detransitivisation. Nothing in the direction is mentioned for Camling by Ebert (1997a), who did recognise and describe S/A detransitivisation for Athpare in Ebert (1997b). ulung and Koyu are described in Lahaussois (2002) and Lahaussois (2009). ulung has split DAM depending on lexical factors (nouns and 3s/3d/3p/2p pronouns vs other pronouns), whereas Koyu has a fluid DAM system where the ergative is apparently used to mark the agent in scenarios where otherwise O could thought to be A. Both languages also have a borrowed case marker -lai (< Nep. -lai [DAT]) which they use for DOM in a similar fashion as Nepali (however, without any effects on O-AGR). ese independently existing paerns make it unlikely that S/A detransitivisation should exist in these languages.
For the remaining Kiranti languages either no data are available at all or only articles with a focus on a topic other than syntax exist. ese are Bahing, Chɨlɨng, Dungmali, Hayu, Khaling, Khambu, Lohorung, Mewahang, Mugali, Nachiring, Sam, Sampang, and Tilung.
2.9.2 Limbu
S/A detransitivisation in Limbu is described early by Weidert and Subba (1985) and later in a dedi- cated article by Angdembe (1998). It is ignored by Driem (1987).
Weidert takes a somewhat eccentric view on the phenomenon. He speaks of an “anti-passive transformation”, which fulfills the formal definitional criteria for S/A detransitivisation, and says that “presumably most” verbs are open to this transformation. Below are examples given by him. (269) a. Am-baˑ-rɛ
1sPOR-father-ERGpit-nuˑ-ncow-milk-DEFthuŋ-u-rɔdrink-[3sA.]3[s]O-CONJ.PTCPyak.stay[.3sS] ‘My father is drinking milk.’
b. Am-ba
1sPOR-fatherpit-nucow-milkthuŋ-lɔdrink[.3sS]-CONJ.PTCPyak.stay[.3sS]
‘My father drinks milk.’ (Weidert and Subba 1985:108) (270) a. Aŋgaʔ
1s sɔksɔkk-inbook-DEF niˑr-u-ŋ-lɔread-3[s]O-1sA-CONJ.PTCPyakk-aʔ.stay-NPST.1sS ‘I am reading the/a book.’
b. Aŋgaʔ
1s sɔksɔkbook niˑt-aʔ-rɔread-NPST.1sS-CONJ.PTCPyakk-aʔ.stay-NPST.1sS
‘I read books; I am a reader of books.’ (Weidert and Subba 1985:109) What is special about Weidert’s view is that he does not consider S/A detransitivisation in isolation. Instead, he views all predicate frames containing a nominative argument and a verb with monop- ersonal agreement as related. is does not only include S/A detransitivised transitive verbs but also normal intransitive verbs and reflexives. He calls the bipersonal paradigm the “active” and the monopersonal paradigm the “middle conjugation” and accordingly analyses S/A detransitivisation as active verbs in middle conjugation. He presents Table 2.11 to summarise the contrasts between the two conjugation types in various dimensions.
A few comments are in place here to explain Weidert’s terms:
• “Directionality” refers to the existence of a “cause-effect relationship” and is similar to Hop- per and ompson’s (1980) concept of affectedness. “Corporeal” is supposed to mean that in the active conjugation this relationship becomes visible in the form of a (oen physical) effect of the action initiated by the agent on the goal (i.e. the patient). In the middle conjugation there are “no determinable causal consequences” for the goal.
• e terms “centrifugal” and “centripetal” are not further explained. ey are introduced in the discussion of “directionality” and seem to signify whether the action performed by the agent (Weidert’s cover term for S and A) is directed towards a goal outside its origin (centrifugal) or not (centripetal).
• “Possessivity” refers to whether the goal is inalienably possessed by the agent or not. is criterion is not meant to be relevant for all cases.
active/‘transitive verbs [sic] middle/‘intransitive’ verbs
directionality corporeal Ø (for one-argument predicates)diffuse (in anti-passive construction type)
volitionality strong reduced
attentional focus ergative agent absolutive agent attention vector centrifugal centripetal actor animacy animate, preferentially human unconstrained
ergativity necessary Ø
possessivity alienable inalienable
Table 2.11: Limbu active and middle verbs (Weidert and Subba 1985:122)
Treating intransitive verbs and detransitivised transitive verbs as one category as Weidert does is problematic. If S/A detransitivisation in Limbu is indeed possible for most verbs, it is likely that is not lexically conditioned, as in Chintang. Being an intransitive verb, by contrast, is a lexically fixed property. Weidert thus tries to bring together two constructions with very different degrees of freedom. at this does not work very well can be seen in Table 2.11. Most of the dimensions of contrast listed are not confirmed by convincing elicited examples, and none are corroborated by corpus data. Some of the concepts such as directionality, aentional focus, and aentional vector are ill-defined, and for all but the formal criterion of ergativity, counterexamples can be easily found. For instance, in (269) it’s not clear why there should be a difference in directionality, volitionality, aentional focus, or aention vector between the two examples, nor why human A should be preferred in the first sentence of each pair but not in the second.
Apart from his aempt to give a generalised characterisation of all monopersonal predicate frames, Weidert also talks about the Limbu antipassive as a construction in its own right in some places. He mentions the known formal features and also that the antipassive cannot be used with personal pronouns, demonstrative pronouns, numeral expressions above 1, and the plural suffix -haʔ on the goal noun (Weidert and Subba 1985:108). As concerns function, Weidert says that the antipassive removes the argument position for the goal (P) and that as a consequence the noun becomes a part of the verb, that is, incorporated (although Weidert does not use the word). is is meant both as a formal and a functional property. Weidert does, however, not expand on this; for him, the main functional characteristic of the antipassive is its being “middle”.
One interesting difference between the Limbu antipassive and Chintang S/A detransitivisation is that the antipassive is unproblematic with possessed objects. Weidert even claims that the an- tipassive is obligatory with objects which are inalienably possessed by A with two verbs. One is shown in (271):
(271) a. Aŋgaʔ
1s a-bik1sPOR-cowkɔmm-aʔ.herd-NPST.1sS ‘I’ll herd my cow.’
b. Aŋgaʔ
1s ku-bi-in3sPOR-cow-DEFkɔm-u-ŋ.herd-3[s]O-[NPST.]1sA
‘I’ll herd his cow.’ (Weidert and Subba 1985:117)
Weidert’s work is hard to appreciate because of his special terminology and eccentric ideas. Still, he presents a solid analysis of the formal properties of the construction to which the dedicated paper by Angdembe (1998), does not add much. What’s especially interesting is that even the central claim of Angdembe’s paper that the phenomenon has to be analysed as noun incorporation has in principle been anticipated by Weidert. Although he does not use the term, he compares antipassivised objects to the nominal component of lexicalised noun-verb combinations that can no longer be used separately from the verb and form a morphological unity with it.
Angdembe mentions a few more formal characteristics en passant but does not present data: the object cannot be dropped, it cannot be modified by adjectives, numerals (without Weidert’s
2.9. S/A DETRANSITIVISATION IN OTHER KIRANTI LANGUAGES restriction “greater than 1”), and the definite article, and it cannot be a proper noun or (contradicting Weidert) an inalienably possessed noun. He also touches on ditransitive verbs (where G triggers O- AGR) and says that the verb is only detransitivised when both direct and indirect object (i.e., T and G) are “incorporated” but does not present examples for the crucial case of G being incorporated and T not.
Angdembe explicitly rejects an analysis of the Limbu antipassive in terms of definiteness. His argument, however, is not very convincing: “the situation in (3) is the same as the situation in (2)” (where (3) and (2) are the default and detransitivised versions of a sentence translated as ‘e friend killed the buffalo’; Angdembe 1998:21). If the “situation” was indeed the same in both clauses any functional explanation including noun incorporation would have to fail.
2.9.3 Yakkha
Schackow (In preparation) mentions S/A detransitivisation as “detransitivization” in her Yakkha grammar and says that “any verb in Yakkha can basically be inflected intransitively” (Schackow In preparation:61). She does not talk explicitly about the case of A and presents examples with dropped A:
(272) a. Cog-uks-u=na.
do[PST.3sA]-TEL-3O=NMLZs ‘He did it.’
b. ekdam
very cog-a-nuŋdo[3sS]-PST-whilecog-a-nuŋdo[3sS]-PST-while
‘while he worked hard/while he did a lot’ (Schackow In preparation:61) She goes on to say that the default frame is used when the object is “definite or specific”, whereas detransitivisation is used when it is “unspecific or generic” or when “the maer is rather about the structure and manner of the event”.
2.9.4 Athpare
S/A detransitivisation in Athpare is described by Ebert (1997b) as “undergoer demotion”. Ebert does not mention the connection of this phenomenon to case marking explicitly but shows with an example that the paern is just as expected, A being zero-marked in the detransitivised frame: (273) a. Un-na
he-ERGlaribobananachoŋs-u-na.sell-[3sA.]3[s]O-NMLZ (no translation provided)
b. Un he laribo banana choŋ-na. sell[.3sS]-NMLZ
‘He sells bananas.’ (Ebert 1997b:122)
Ebert only touches on the function of undergoer demotion. She mentions that it is used when “the undergoer noun does not denote a specific entity” and that the noun is “quasi-incorporated” (p. 122), all of which sounds very similar to Chintang.
As a special case she mentions “inherent objects” without defining the term but listing the combinations ‘speak a language’, ‘sing a song’, and ‘cook food’ as examples. What is intuitively inherent about these objects is that they correspond more or less to the type of referent required by the verb. For instance, ‘song’ comes close to covering all possible objects of ‘sing’. ere is some variation, though: ‘language’ is obviously an important but not the only type of object licensed by ‘speak’, and food is not the only type of thing that can be cooked. Such cases correspond to what has been called composite activities here (see section 2.6.5.2).
2.9.5 Belhare
Bickel uses the terms “object downgrading” (Bickel 2003a) and “detransitivisation” (Bickel 2004a, Bickel et al. 2010) for S/A detransitivisation in Belhare. He adds to the defining criteria that the concerned object cannot be pluralised, possessed, or specified by a demonstrative or any other at- tribute, and that it cannot be moved to the right of the verb. Detransitivisation is also impossible with “inherently specific” objects (Bickel 2004a:167).23 He presents various arguments against a
formal analysis as noun incorporation: even though the object cannot stand to the right of the verb, other elements can intervene between the two, for instance, focussed agents. e object can be dropped if “the context is clear enough” (Bickel 2004a:169), and it is accessible to information structuring processes such as topicalisation, focalisation, and questioning. He concludes that de- transitivisation leaves the argument status and role of the object untouched. Here are his examples: (274) a. (I-na)
DIST-DEMwachickenkhuʔ-yu.steal-NPST[.3sS] ‘is [guy] steals chicken.’ b. (I-na-ŋa)
DIST-DEM-ERGwachickenkhuiʔ-t-u.steal-NPST-[3sA.]3[s]O
‘is [guy] will steal a/the chicken.’ (Bickel 2003a:557) e function of the construction is described in familiar terms: “e nominal does not refer to a specific referent but to a kind of referent” (Bickel 2004b:167, emphasis by Bickel). is points to the token : type distinction being important for Belhare. However, Bickel does not go deeper into this and also mentions less specifically in his other paper that detransitivisation “partially fulfills an antipassive function” (Bickel 2003a:556). In Bickel et al. (2010:388) it is claimed that detransitivisation marks non-specificityand may imply the notion of a “general activity”.
An interesting detail has to do with the Belhare perfect. ere is a perfect marker with supple- tive allomorphs, -sa aer intransitive stems and -ŋa aer transitive stems. In detransitivised frames the transitive allomorph is used. is is different from Chintang, where both variants of the marker -hat(t) [AWAY], which has similar formal behaviour, are allowed with S/A detransitivisation.
2.9.6 Chiling
Chiling (Chɨlɨŋ) has so far not been linguistically documented at all. However, since it is spoken in Ā̃khisallā, a VDC neighbouring Chintang, I had the chance to do some preliminary elicitation work in 2012 and 2013. e data below clearly show that Chiling is part of the Eastern Kiranti area where S/A detransitivisation is common. e intransitive verb in (275a) carries the same agreement marker as the transitive verb in (275b). (275c) shows a contrasting transitive form of the same verb. e motivation for S/A detransitivisation seems to be the count/mass distinction (cf. section 2.6.3.1 on the same factor in Chintang).
(275) a. Mu-bak
DEM.DOWN-LOCyuŋ-yu-wa.be.there-IND.NPST-1sS ‘I’m down.’
b. Cama
rice ca-yu-wa.eat-IND.NPST-1sS ‘I eat rice.’
c. Sontorok
orange cay-u-ku-ŋ.eat-3[s]O-IND.NPST-1sA
‘I eat an orange.’ (elicitation RKU 2013)
How similar this paern is to S/A detransitivisation in the neighbouring languages can only be shown by further research.
23Note, however, that his example for this is the possessed noun ucha ‘his child’, so it’s not clear whether this is really
2.9. S/A DETRANSITIVISATION IN OTHER KIRANTI LANGUAGES
2.9.7 Bantawa
e first grammar of Bantawa is R¯a¯ı (1984), who does, however, not talk about S/A detransitivisa- tion. e relevant reference work is therefore Doornenbal (2009). Similarly to Weidert and Subba (1985), Doornenbal views intransitive and transitive inflection as conjugation classes and says that “many” verbs can be inflected both transitively and intransitively. He is apparently not aware of the ambiguity of this statement with respect to S/A and S/O detransitivisation but makes it clear in the following that he is talking about S/A detransitivisation by using the term “antipassive”.
He distinguishes two antipassives, an “implicit” (i.e. unmarked) and an “explicit” one that makes use of the “dummy object marker” kha. e explicit antipassive is syntactically different from S/A detransitivisation in that it requires ERG on A. is particle is likely to be cognate to Chintang kha-, which codes [1nsO] in the dialect of Sambugaũ and is used as a detransitiviser on a handful of verb roots in both dialects (e.g. copt- ‘look at’ vs khacopt- ‘look around, stare’). ere is also a corresponding prefix in Puma (see below).
e implicit antipassive, on the other hand, fulfills the formal criteria for S/A detransitivisation. Here are Doornenbal’s examples:
(276) a. Ŋa
fishlaʔ-u-ŋ.catch-3[s]O-1sA ‘I caught a fish (the fish).’ b. Ŋa
fishlaʔ-a-ci-ʔa.catch-PST-[1]d-e[S]
‘We (dual, excl) went fishing.’ (Doornenbal 2009:223) Concerning the function of the implicit antipassive, Doornenbal first says that the antipassive is the default for “verbs where the object is less specific or less obviously affected” (p. 223). Aer that he gives a couple of other conditions that are very different from what is found in Chintang: the