Optional or Syntactic Ergativity in Shawi?

Distribution and Possible Origins*

 

Luis Miguel Rojas-Berscia†    Corentin Bourdeau‡

 

In this article we provide a preliminary description and analysis of the most common ergative constructions in Shawi, a Kawapanan language spoken in Northwestern Amazonia. We offer a comparison with its sister language, Shiwilu, for which an optional ergativity-marking pattern has been claimed (Valenzuela, 2008, 2011). There is not enough evidence, however, to claim the exact same for Shawi. Ergativity in the language is driven by mere syntactic motivations. One of the most common constituent orders in the language where the ergative marker is obligatory is OAV. We close the article with a tentative proposal on the passive origins of OAV ergative constructions in the language, via a by-phrase-like incorporation, and eventual grammaticalisation, resorting to the formal syntactic theory known as Semantic Syntax (Seuren, 1996).

 

1. Introduction

 

The area in the triangle formed by the Escalera mountain range, and the Marañón and Huallaga rivers can be considered a crucible of linguistic diversity. This piece of land, no larger than modern Flanders, congregates speakers of at least seven different language families: Muniche (isolate), Aguaruna (Chicham), Kukama-Kukamiria (Tupi), Chamicuro (Arawak), Candoshi (isolate), Upper Amazonian Quechua (Quechua), and Shawi and Shiwilu (Kawapanan). To date, most of these languages remain poorly investigated. Muniche and Chamicuro are barely used, and only by some elders in the villages where they were originally spoken. Although it is impossible nowadays to provide a thorough description of both languages, there have been previous attempts at dealing with general aspects of the grammar and the lexicon (cf. Gibson 1996; Michael et al. 2009; Parker 1994). As for Upper Amazonian Quechua, in spite of its great diffusion in the area, there is no comprehensive reference grammar. Candoshi, notwithstanding its vitality, remains under-documented (cf. Cox 1957); however, there are recent efforts by Simon E. Overall from the University of Otago, NZ, to provide a thorough description of the language. The exceptions are the hot-from-the-oven descriptions for both Aguaruna (cf. Overall 2017) and Kukama-Kukamiria (cf. Vallejos 2016). For the languages we deal with in the present article, namely Shawi and Shiwilu, there is still no single comprehensive grammar, despite the serious descriptions of grammatical aspects of both languages (cf. Hart 1988; Barraza 2005a, 2005b; Rojas-Berscia 2013; Valenzuela 2011, 2016). This current state of the art is not foreign to Amazonian linguistics in general. Only when all these languages are carefully studied, will we be able to understand the dynamics behind this great linguistic diversity, as well as other poorly understood grammatical features in the region.

 

For this article, we focus on a specific understudied area of Shawi grammar: the so far problematic suffix -ri, which displays an ergative distribution but does not occur on every agentive subject of transitive clauses, as described in Bourdeau (2015).[1] Our elicited data indicate that ergative-marking is optional in Shawi when the O-NP of the sentence was a third person. We, therefore, carried out a small corpus investigation, based on narratives and speech data to understand the underlying principles conditioning the use of the ergative marker. The most important findings are the following:

 

1.	A-NPs bear the ergative marker in the following cases:
		(a) when the canonical SOV/SVO order is violated for topicalisation.
		(b) when the O-NP is omitted because it is foregrounded in the discourse.
		(c) emphasising contrast.
2.	These findings are similar to the ones presented by Valenzuela (2008, 2011) on Shiwilu, the sister language of Shawi. However, we claim that unexpectedness of the A-NP does not account for the use of the ergative marker since, in Shawi, overtly expressing the new actor suffices to make up for its unexpectedness.
3.	We eventually concluded that ergative marking in Shawi is not optional at all, but conditioned by strict syntactic rules. The factors at play are the following: word order, distinguishability of A-NPs and O-NPs, and information structure.

 

In Section 2, we provide a grammar sketch of the language. In Section 3, we first start with a short introduction on the same phenomenon in Shawi’s sister language Shiwilu, since ergativity in this language has already been studied. Section 4 provides explanations for each of the cases where Shawi’s ergative case-marker -ri is used. Section 5 sketches a hypothetical explanation for the development of ergativity in Shawi following a syntactico-semantic explanation (Seuren 1996), enriched by previous accounts on other languages from various frameworks (cf. Coghill and Deutscher 2002; van de Visser 2006, 2006; Levinson, sub.).

 

2. Grammar Sketch

 

The Shawi, or Chayahuita language, a descendant of Southern Mayna or lengua de Cerros de Maynas (Rojas-Berscia 2015), is, as mentioned above, a member of the Kawapanan language family, together with Shiwilu, with which it shares 60% of its lexicon (Valenzuela, 2012:1). There are approximately 21,000 speakers of Shawi (INEI 2007), although this number refers mostly to people identified as Shawi and not to actual speakers of the language. As is evident from the regular use of the language, monolingualism in women and children, and language attitude, Shawi is still a very vital language compared to other adjacent Andean or Amazonian languages in Peru.

 

In typological terms, Shawi is a language with a strong tendency to agglutination and synthesis, meaning that morphemes are easily distinguishable from each other and that there is a large number of morphemes per word. There are some instances, however, in which we can find phonologically independent morphemes, such as the negative marker ku or the first person pronoun ka, as in (1). Below we provide an example where several suffixes are added to the main verb tepa’to kill’, as well as a prefix, the indirect causative a- (Rojas-Berscia 2013):

 

(1)	Nunu	ka	a-tepa-ra-we-su	ku	ka’-na-we.
	monkey	1	caus-kill-non.fut-1-nmlz	neg	eat-non.fut-1
	‘I do not eat the monkey which I made someone kill.’

 

Both suffixation and prefixation are found in the language. The most common prefixes in the language are the prefix a-, for indirect causation, ichi-, for sociative causation,[2] the reciprocal ni- and the prefix shi-, which means ‘almost’. There is a valency-changing affix te- which can increase or diminish valency, as well as verbalise non-verbal roots. Its precise functions still have to be investigated. It seems to have to do with valency change in ways that are not entirely perspicacious at the moment.

 

The most frequent word order in the language is AOV for transitive clauses. However, there are some instances in which the AVO order seems to be preferred, perhaps due to intense contact with Amazonian Spanish. The OAV order is also sometimes preferred, mostly when used in 3>3 transitive clauses in which the ergatively marked argument comes after the object (Bourdeau 2015). For intransitive clauses, the preferred word order is SV, which mirrors the predicative clause word order, S(Pred). Both subject and object are marked on the verb, as in (2):

 

(2)	Ka-ri	nanian-te-ra-u-nke.
	1-erg	forget-appl-non.fut-1.A-2.O
	‘I forgot you.’

 

Nouns and verbs are open classes in the language, both occurring with rich morphological derivation and inflection paradigms. There is a closed class of adjectives with very few members. The other ‘adjectival’ predicates are verbs. Modifier precedes modified and there cannot be more than one modifier in a single noun phrase. For example, (3) shows a sentence in which two adjectives are preceding a noun. This construction is ungrammatical in the language. If two adjectival predicates are needed, a copula construction is preferred, as in in (4):[3]

 

(3)	a.	*[panka		[atari		kayu]]
		big		chicken		egg
		‘the big chicken egg’
	b.	[Panka	kayu]		[atari		kayu]-ø]
		big	egg		chicken		egg-cop.3sg
		‘The big egg is a chicken egg.’

 

In addition, the language has pronouns, deictics, adverbs, classifiers, conjunctions and interjections as closed word classes. There is a first person singular exclusive ka ‘only I’ and a first person singular inclusive kanpu’ ‘you and I’. Their pluralised forms are kiya ‘we, but not you’ and kanpua’ ‘you and we’ respectively. The second person is kema, and the third, ina. Both can be pluralised by means of the pluraliser -pita. The third person in the language is also a third person deictic. As in many other languages, all deictics in the language can be treated as third person free pronouns.

 

In terms of grammatical relations, Shawi exhibits different alignment types. It obligatorily case marks some NPs on an ergative/absolutive basis.[4] The ergative case is marked by means of suffix -ri, while the absolutive is unmarked.[5] As for verbal concordance, Shawi shows a nominative-accusative alignment in transitive clauses,[6] and a quite obscure active-stative system where indexed verbal objects and nominal predicates are marked in the same way. For example, (5) shows that the O in the sentence is marked by means of a special first person singular suffix. Interestingly, in (4), the predicate shawi is marked by means of –ku suffixation as well.

 

(4)	Ka	shawi-ku.
	1	Shawi-1.o
	‘I am a Shawi man’.

 

(5)	Kema	tepa-r-an-ku.
	2	kill-non.fut-2.s-1.o
	‘You kill me’.

 

As can be seen from the previous examples, nominal predicates and indexed verbal objects are marked in the same way in the language. It is not the case that unaccusative/unergative intransitive clauses are marked in a particular way. There is no marking split in intransitive clauses. For example:

 

(6)	Ka	chimin-a-we.
	1	die-non.fut-1.s
	‘I die’.

 

(7)	Ka	yun-a-we.
	1	swim-non.fut-1.s
	‘I swim’.

 

What can be inferred from (4), (5) and (6) is that there is a particular split which, on the one hand, groups together nominal predicates and transitive objects (see the -ku 1.o suffix in (4) and (5)), and, on the other hand, intransitive and transitive subjects (see the -we 1.s suffix in (6) and (7)).[7]

 

3. Optional ergativity in Shiwilu

 

Shiwilu, the sister language of Shawi, displays optional ergativity marking (Valenzuela 2008, 2011). Suffix -ler follows an ergative distribution as it only occurs on subjects of transitive verbs (A-NP), but still can be omitted without resulting in ungrammaticality:

 

(8)	a.	Kisu	ka’-l-i	nana	isha
		Jesus	eat-non.fut-3	that	paujil
		‘Jesus ate the paujil.’(Valenzuela 2008: 215)

 

	b.	Kisu-ler	ka’-l-i	nana	isha
		Jesus-erg	eat-non.fut-3	that	paujil
		‘Jesus ate the paujil.’ (Valenzuela, 2008: 215)[8]

 

Valenzuela (2008, 2011) accounts for the use of this ergative marker attributing to it two well-defined sorts of function, namely discriminatory functions and discourse-pragmatic functions. The discriminatory functions concern three specific morphosyntactic contexts:

 

1.	The canonical word order AOV is altered and the O-NP precedes the A-NP (see (9) and (10)). This is the only case in which the use of -ler is reported to be absolutely obligatory.
2.	When both the A-NP and the O-NP are third persons (see (11) and (12)).
3.	In the presence of valence increasing morphemes such as the causative and the applicative affixes (see (13), (14) for the applicative[9] and (15) for the causative).

 

(9)	Ipulitu	de’-tu-l-i	Pulu-ler.
	Hippolyte	kill-prfv-non.fut-3	Paul-erg
	‘Paul killed Hippolyte.’

 

(10)	Kisu	ek-lansa’-pa-l-i	Pulu-ler.
	Jesus	soc.caus-dance-prog-non.fut-3	Paul-erg
	‘It is Jesus that Paul is making dance.’

 

(11)	Nini’-wawa-ler	kite’-l-i	nana	wila-wawa.
	dog-baby-erg	bite-non.fut-3	that	child-baby
	‘The puppy bit the child.’

 

(12)	Ipulitu(-ler)	de’-tu-l-i	Pulu.
	Hippolyte(-erg)	kill-appl-non.fut-3	Paul
	Paul ‘Hippolyte killed Paul.’

 

(13)	Walinti	saka’-tu-l-i	Kishu-kin.
	Valentin	work-appl-non.fut-3	Jesus-for
	‘Valentin worked for Jesus.’

 

(14)	Walinti-ler	saka’-tu-l-i	Kishu.
	Valentin-erg	work-ben-appl-non.fut-3	Jesus
	‘Valentin worked for Jesus.’

 

(15)	Kanuta-ler	a‘-tuluner-erchu	Kulushe’.
	Charlotte-erg	caus-sing-fut;3	Cruz
	‘Charlotte will make Cruz sing.’/’Cruz will be made to sing by Charlotte.’

 

As for the discourse-pragmatic functions, they correspond to the uses of the ergative marker -ler triggered by:

 

1.	The first mention of an entity in the discourse
2.	A contrastive focus
3.	The reactivation of a participant after an absence of four or more clauses (Valenzuela 2011).

 

By grouping ‘first mention’ and ‘reactivation of a discourse participant’ together into a category labelled unexpected agent, Valenzuela (2011) accounts for 48.8% of the uses of the ergative marker, see (16) for an example of this type. Moreover, she attributes the other 30.4% of its occurrences to a contrastive focus function,[10] see (17) for an example of this type.

 

(16)

Napi’

ala’sa’

iyali’

Shiwilu=k

amana’=le’

i(n)-denma-l-i.

long.ago

one

man

Jeberos=loc

jaguar=com

rec-fight-ind;3

Nana

iyali’

itu-le’,

Luis Inuma.

Pampate’-lu’-dunsa=k

that

man

speak.of-ind;1

Luis Inuma

pasture-land-edge-loc

tampu’-nen

ni-a’pa-sik,

nane’=la’

amana’=ler

shelter-pos.3

exist-prog-dur;3;ds

there-abl

jaguar=erg

pili-tu-nta-l-i

kuser-nen.

seize-appl-come/go-non.fut;3

pig-pos;3

‘Long ago, a man in Jeberos fought with a jaguar. That man I’m talking about is Luis Inuma. On the edge of the pasture…, where his shelter was located, from there a jaguar seized his pig.’ (Valenzuela 2011: 112)

 

(17)	‘Enta’ina		a-pida-t-(k)er.		Innich-impu’-pachen		kwa=ler
	let’s.see		caus-take.off-appl-imp		can-neg-sub		1-erg
	a’ka	a’-pida-t-erch-e’			tu-l-i-ma		Ke’ki
	indeed	caus-take.off-appl-fut-1			say-non.fut-3-hsy		sun
	‘Prove it, make him take it off. And if you can’t I will make him take it off, said the Sun.’ (Valenzuela 2011: 113)

 

Valenzuela (2011) thus justifies 79.2% of the occurrences of -ler by the appearance of an unexpected agent, which allows her to claim that the Shiwilu ergative is mainly pragmatically driven. In the following section, we will account for the use of the optional ergative marker in Shawi in these same contexts.

 

4. “Optional” Ergativity in Shawi

 

In Rojas-Berscia & Bourdeau (n.d.), we show that Shawi split-ergativity is constrained by strict syntactic rules which violate the widely accepted Nominal Hierarchy. Yet, during our elicitation sessions, Shawi speakers all agreed that the use of the ergative marker is optional if the O-NP is a third person pronoun or NP (no matter the nature of the A-NP). ‘Optional’ here does not imply ‘random’. Thus, in the following, we show what principles underlie the occurrence of ergative -ri in these sentence type. To support this analysis, we collected narrative and natural speech data in the field and got hold of a copy of the New Testament translated by Hart (1978).

 

First, note that in Shawi, the subject, whether an A or an S, is omitted when easily identifiable in context,[11] see (18), following McGregor’s (1998) expected actor principle which states:

 

The episode protagonist is —once it has been established— the expected (and unmarked) Actor of each foregrounded narrative clause of the episode; any other Actor is unexpected. (McGregor 1998: 516).

 

(18)

A’na

piyapi-rawa

we’e-sa-r-in

yuki

tashinan-ke-chachin

one

boy-dim

sleep-prog-non.fut-3

moon

night-loc-precisely

ni-anta-r-in.

(Ina)

sapate

ma‘shu-ne-n

aku-r-in.

be-reit-non.fut-3

3

shoe

old-pos.alien:3

put-non.fut-3

‘A little boy is sleeping, it is moonlight. He put his old shoes (aside).’

 

Now, in case an unexpected actor pops up in the narration, the mere fact of overtly expressing it makes it clear to the listener who the actor is. As such, we think the unexpectedness of the A-NP does not suffice to account for the use of the ergative marker in Shawi. It is true that all the ergatively marked A-NPs one finds in the corpus refer to unexpected actors since the A-NP has to be overtly expressed for the ergative marker to occur. Unexpectedness, however, is not the cause for the occurrence of ergative marking. This can be summarized resorting to the following rules:

 

            • If there is an ergative marker on A, it is unexpected.

            • If there is an ergative maker on A, A is overtly expressed.

            • Conclusion: A may still be unexpected and overtly expressed without an ergative marker.

 

As we will show next, ergative marking in Shawi is conditioned by syntactic rules, the scope of which is the sentence and not the discourse.

 

4.1 Word Order violation

 

As argued by Rojas Berscia (2013), the Shawi canonical word order in transitive sentences is AOV. Therefore, in the absence of case-marking, Pitru ‘Peter’ will undoubtedly be perceived as the A-argument in (19).

 

(19)	Pitru	ni’ni	tepa-r-in.
	Peter	jaguar	kill-non.fut-3
	‘Peter killed a jaguar.’

 

Yet, AVO order is also frequent in speech due to Spanish influence, and in this case too, no ergative marking is needed so as to distinguish A from O. In (20), Pitru ‘Peter’ will thus invariably be regarded as the A-NP.

 

(20)	Pitru	tepa-r-in	ni‘ni.
	Peter	kill-non.fut-3	jaguar
	‘Peter killed a jaguar.’

 

In brief, the first NP in a transitive sentence is more prominent and is, by way of default, analysed as the A-NP, the O-NP always being right next to the verb: either before (AOV) or after it (AVO). Sometimes though, the speaker might want to move the O-NP to the front of the clause (O-fronting), see (21) or (22), for topicalisation purposes. Because of O-fronting, the first argument is the O-NP and the A-NP is the one occurring right next to the verb: either before (OAV)[12] or after it (OVA). For the listener to understand who acts upon whom, the ergative marker becomes obligatory. In (22), the first sentence of the narrative text reported by Rojas-Berscia (2013) and entitled ‘Conflict with the Aguaruna’ illustrates topicalisation by O-fronting.

 

(21)	Napuapunawe,	ku	a‘na-ya	teranta	Yuse-ri	naniante-r-in-we.
	?	neg	one-dim	even	God-erg	forget-non.fut-3-neg
	‘God does not forget anybody, NOT EVEN A SINGLE LITTLE PERSON.’

 

(22)	Iraka	shawi	kema-rusa-ri	tiki-r-in.
	formerly	shawi	Aguaruna-pl-erg	slaughter-non.fut-3
	‘THE SHAWI, the Aguaruna used to slaughter.’

 

Basically, all the possible word orders are likely to occur depending on what the speaker wants to focus on. As such, we found examples of OVA and VAO (24) sentences in our corpus.

 

(23)	Iseke	Kankan	nu‘wi-ra-r-in	ni‘ni-ra-ri.
	Here	wasp	yap-prog-non.fut-3	dog-dim-erg
	‘Then, IT IS YAPPING AT THE WASPS, the little dog.’[13]

 

(24)	A‘china-pi	mayistru-sa-ri-nta	kanpunan.
	teach-3;pl	teacher-pl-erg-addit	shawi.language
	‘And TEACH THEM Shawi, the teachers do.’[14]

 

4.2 Omission of the O-NP

 

In a transitive clause, the A and the O arguments are, by way of default, distinguished by means of word order. This distinguishing device, however, no longer helps once one of the arguments remains implicit, due to pro-drop. As a matter of fact, Shawi being a pro-drop language, the A-NP is likely to be omitted, in which case the unique NP preceding the verb is the O-NP. On the other hand, when the NP occupying the slot of the O-NP has been introduced earlier, it is also likely to be omitted and referred to on the verb only with the 3rd person object ending, which is -ø. In the latter case, the unique NP preceding the verb is thus the A-NP.[15] For the sake of comprehension, when the only NP overtly expressed is the A argument, Shawi speakers seem to always use the ergative marker, see (25) and (26),[16] even in cases where the context or world knowledge would have sufficed to figure out which one of the two arguments has been omitted, as in (26).

 

(25)	Ni’ni-ri	nuku-ra-r-in-ø.
	dog-erg	look.at-prog-non.fut-3-3.o
	‘The dog is looking at it (the toad).’

 

(26)	Kankan-ni	peya-ra-r-in-ø.
	wasp-erg	sting-prog-non.fut-3-3.o
	‘The wasps are stinging it (the dog).’

 

On the other hand, whenever the unique NP overtly expressed is in the absolutive, this is the O-NP of the clause (27).

 

(27)	Tururu	nuku-ra-r-in-ø.
	toad	look.at-prog-non.fut-3-3.o
	‘He (the boy) is looking at the toad.’

 

If the O-NP remains implicit, it is because it was introduced earlier with a full NP and because it is part of the [known] topic. As such, a sentence with a transitive verb and a unique NP overtly expressed and ergatively-marked occurs when the topic protagonist moves from the subject function (S/A) to the O function, the new actor being introduced with the ergative marker, see (28).

 

(28)

Kankan

nutuaru

tumunte-r-in,

ya’were

wa’an-ne-n-su

wasp

a.lot

rise.dust-non.fut-3

but?

his.leader-alien-3-def

nara-epa

nanpe-r-in.

Nara-e-ran

nuku-ra-r-in.

Iseke-wachin

tree-top.of

climb-non.fut-3

tree-loc-abl

look-prog-non.fut-3

here-seq;3

kankan-ni

peya-ra-r-in-ø.

wasp-erg

sting-prog-non.fut-3-3.o

‘A cloud of wasps rises but its owner (of the dog) climbs to the top of a tree. He is looking from the tree. Then, the wasps sting him.’

 

4.3 Contrastive focus

 

The ergative marker also seems to serve for the expression of contrastive focus, “evoking a contrast with other entities that might fill the same position” (Gundel and Fretheim 2006). This function of the ergative marker makes it recurrent in complex sentences in which the actors of the main and the subordinate clauses are different. In (29), the second clause of the sentence displays an AVO word order and thus does not require the use of the ergative. The ergative marker is there to mark a contrast between ‘the wasps’ and ‘the dog’, whose actions are simultaneous and in direct relation to each other. In this case, the new subject, ‘the dog’, is the new protagonist of the event.

 

(29)	Inara	iseke	kankan-i		peya-ra-r-in,			ni’ni-ri
	then	here	wasp-erg		sting-prog-non.fut-3			dog-erg
	nu‘wi-ra-r-in			kankan.
	shout-prog-non.fut-3			wasp
	‘Then, the wasps are stinging it (the dog), the dog is shouting at the wasps.’

 

In (30),[17] the ergatively marked A arguments are constrasted with the NP referring to the known discourse topic, that is to say with Sutumarusa, Kumurarusa, ‘the Sodomites and the Gomorrheans’. In this case, supai ‘devil’ is by no means a new protagonist.

 

(30)

Iraka

Sutuma-rusa,

Kumura-rusa,

ina-pita

pa’pi

ku

nuya-we

formerly

Sodomite-pl

Gomorrhean-pl

3-pl

very

neg

good-neg

kankan-tu-pi

nitun,

Yuse-ri

chiniken

a-na’in-te-r-in.

feel-appl-3.pl

because

God-erg

strong

caus-be.culprit-appl-non.fut-3

Ana-pita-nta,

supai-rusa-ri

yakuan

kan-tu-pi

nitun,

one-pl-addit

devil-pl-erg

door

arrive-appl-3;pl

because

papi

wa’yan-tu-pi.

very

evil.spirit-appl-3.pl

‘A long time ago, the Sodomites and the Gomorrheans, because they were very bad people, God punished them harshly. These ones turned completely crazy because the devils arrived at their doors.’

 

5. A hypothetical historical development: from Passive to Ergative

 

The idea that ergative languages developed through the loss of an active voice, leading to the pervasive use of a passive which eventually becomes reanalysed as an active is not new. It goes all the way back to Schuchardt (1896) and finds many follow-up explanations, although not in the same vein, in the generative tradition. Dixon (1972: 137) for Dyirbal, for example, presented structures in which the O-NP is immediately dominated by the sentence node, and the A-NP occurs as sister to the verb. This meant that nominative-accusative languages had different underlying structures for semantically equivalent transitive sentences. A mirror version of this idea is presented in Larsen (1987, cited in Levinson (sub.)), for which O/S-NPs share a position as sisters to the verb, while the A-NPs occupy a higher node.

 

The initial Passive-to-Ergative path may not explain the full picture, but it has found good cross-linguistic support.[18] Sumerian passives, for example, became reanalysed as active-transitive, leading to the development of the Hamṭu construction (Coghill and Deutscher 2002). More recently, Van de Visser (2006: 185) introduced the ergative as passive hypothesis, backed-up by data from typologically different languages. For the author, every language is basically nominative-accusative and, in languages like Basque, where independent pronouns and full DPs, in Van de Visser’s terminology, are frequently omitted and where ergative case marking is applied on pronominal arguments, transitives “are always ‘passive’ in the sense that transitive subjects are realised by incorporated pronouns. The pronouns differ from the empty nominal category in English passive constructions in having specific reference [...]” (ibid: 188). The author claims that passive clauses in some languages of the world take over the function of their active counterparts, triggering an assignment of a higher topical status to O compared to A, that forces a by -phrase to be attached to the IP constituent, where it has to C-command over an O-NP (ibid: 215).

 

It is difficult to make final statements about the origins of the ergative in Shawi. Unfortunately, there was almost no documentation of the language prior to the 1980s. There are some cases in Southern Mayna, however, that seem to contain a morpheme phonologically similar to ergative –ri. The direct link to modern suffix –ri is not very probable. See (31) and (32).

 

(31)	Kema	lowan-tu-ra-n-su’	lel-ø-in-su-ni.
	2	want-appl-non.fut-2-nmlz	do-non.fut-3-nmlz-erg?
	‘Let your will be done.’ (Rojas-Berscia 2015: 398)

 

(32)	Kema	apuri-ne-n	mucha-r-in-su-ni.
	2	name-alien-3	revere-non.fut-3-nmlz-erg?
	‘Be your name revered.’ (Rojas-Berscia, 2015, p. 398)

 

In (31)-(32) we observe that -ni forms some sort of participial construction, i.e. ‘be done’, ‘be revered’. In other languages of the world, it has been observed that passives generate the appearance of participials in the VP. Although we cannot say much more about this construction in the language, it is of special interest since it is formally very similar to ergative -ri. Functionally, however, it is quite different and a direct link is not straightforward. We prefer to think that this construction was a different one. Backed-up by cross-linguistic evidence and some structural parallelisms, we tentatively hypothesise that Shawi ergative constructions originate in passives.

 

In the framework known as Semantic Syntax (Seuren 1996),  there are two possible underlying word orders, VSO and SOV (McCawley 1972). SVO and OVS orders would be triggered by the process of subject raising in both orders respectively. SVO is triggered via leftward subject raising, since VSO languages are right-branching, while OVS is triggered via rightward subject raising, since SOV languages are left-branching. VOS and OSV would be triggered by a process of ergativisation of VSO and SOV, respectively (Seuren p.c.).[19] The latter operation, which leads towards an ergative alignment, is expressed the way it is in other transformational frameworks: ergatives originate in passives. Shawi, as mentioned in §2 is primarily an SOV language. The ergative, although restricted by the cases discussed in §4, is obligatory on the A-NP if it comes after the O-NP. Any case of OAV ordering will trigger the obligatory marking of A in the sentence. We reckon this to be a remnant of the possible origin of ergatives in the language. It is possible that -ri -NPs were once by-like-phrases which underwent the following processes as most passives in the world languages:

 

Stage 1: -ri -NPs as by -phrases

 

The following tree structure just shows the position of -ri in the sentence in the semantic analysis. As all by-phrases, it triggers both object incorporation and lowering (Seuren, 1996: 130). Ni’ni wa’an-ni teparin literally means ‘The apu killed the dog’ or ‘The dog was killed by the apu’. In this case, ergative -ri becomes -ni in contact with the previous nasal consonant.

 

 

 

Stage 2: Object Incorporation takes place

 

In this case the postposition -ri takes its argument, wa’an, via object incorporation.

 

Stage 3: Lowering takes place

 

Finally, the -ri-phrase is lowered to its final position. The landing site is between the O-NP and the VP. This is determined by what we see synchronically as obligatory marking when A-NPs occur after O-NPs. In other languages of the world with passives, this is also the landing site of by-phrases (Seuren 1996: 347).

 

 

Through the process of object incorporation and lowering, the Shawi -ri-phrase ends up inserted between the Subject NP and the VP, the canonical landing site for adverbial phrases. This would trigger the OSV order in the language after regrammatalisation. Once regrammaticalisation has taken place, the original -ri-phrases became more subject-like and subject agreement has been reassigned to them, hence its reaccommodation to SOV. Today’s pattern of ergative marking may be evidence of this latter grammaticalisation process since ergatively-marked NPs never took over entirely.

 

This explanation, unfortunately, awaits further investigation. This by-phrase to ergative-marker path will only make sense if and when we find a direct diachronic link between an oblique case marker, an instrumental for example, and today’s ergative. Still, frequency-based studies need to be carried out in order to determine what the prototypical position of adjuncts is in Shawi, such as by-phrases in other languages of the world. This would not only support our hypothesis of a post Subject landing site of -ri-phrases in previous stages of the language, but also provide a better understanding of its inter-clausal grammatical relations. Current studies on Shiwilu grammar, backed up by philological analyses of 18th century Shiwilu (Alexander-Bakkerus 2016), will hopefully clear up the historical panorama. Only when both languages of the family are more fully studied there will be more evidence to confirm or belie the aforementioned hypotheses.

 

6. Conclusion

 

From the previous sections, we can infer the following points:

 

1.	Shawi exhibits an ergative pattern, since suffix -ri attaches only to NPs in A function.
2.	Subject NPs in Shawi are generally omitted when easily identified via contextual inferences.
3.	This general omission mechanism makes it clear who the actor is in case the speaker overtly expressed A/S. Therefore, the unexpectedness of A-NPs is not an accurate diagnose for the understanding of the principles behind the occurrence of -ri.
4.	The occurrence of this suffix is not ‘optional,’ but triggered by clear syntactic factors: when the canonical word-order, namely AOV(AVO), is violated, when there is an omission of O-NPs (A-NPs and O-NPs distinguishability), and for emphasising contrast.
5.	Southern Mayna, or colonial Shawi, displays a marker similar to suffix -ri. Its function, however, remains obscure and it is hard to determine whether both were historically related or not.
6.	We tentatively proposed a historical development of ergatively-marked transitive clauses through passives in the language. object incorporation by a by-like phrase might have occurred (cf. Seuren 1996), triggering the modern OAV order, and, via a subsequent grammaticalisation, the development of an ergative marker. Although these ideas have been widely discussed in structural accounts on ergativity, more historical data is missing to confirm this hypothesis.

 

Ergativity in Northwestern Amazonia is quite rare and poorly understood. It must not be forgotten that larger corpora, provided only via scrupulous documentation, is necessary in order to enrich our understanding of this phenomena in Kawapanan (Valenzuela 2011: 118) Frequency-based studies, for example, will help clear out the panorama regarding the occurrence of certain constituent orders in Shawi. In addition, variational studies are necessary to determine the presence or absence of possible social variables triggering the occurrence or prohibition of use of the ergative. Finally, only a better comprehension of sound-change patterns in Kawapanan, coupled with a solid study of Kawapanan comparative morphosyntax will allow us to trace back the origins of ergative -ri in the language.

 

List of Abbreviations

 

1	first person pronoun	loc	locative
2	second person pronoun	neg	negative
3	third person pronoun	nmlz	nominaliser
abl	ablative	non.fut	non-future
addit	additive	NP	Noun phrase
alien	alienable	o	object of a transitive clause
appl	applicative	PostP	postpositional phrase
ben	benefactive	PostP	postposition
com	comitative	prog	progressive
def	definite	rec	reciprocal
des	desiderative	refl	reflexive
dub	dubitative	reit	reiterative
erg	ergative	seq	sequential action
excl	exclusive	soc.caus	sociative causative
gen	genitive	S	propositional structure (in tree structure)
imp	imperative	s	subject of an instransitive clause
incl	inclusive	VP	verb phrase
instr	instrumental	V	verb

 

References

 

Alexander-Bakkerus, A. (2016). Eighteenth Century Xebero: Mss. Add. 25,323 and 25,324 of The British Library, London (UK). Mu¨nchen: LINCOM.

Barraza, Y. (2005a). El sistema verbal en la lengua shawi (Doctoral dissertation, Universidade de Pernambuco).

Barraza, Y. (2005b). ¿Es la lengua shawi una lengua activa? Memorias del Congreso de Idiomas Indígenas de Latinoam´erica-II:1–8.

Bourdeau, C. (2015). Ergativity in shawi. MA Thesis. Nijmegen.

Coghill, E. & Deutscher, G. (2002). The origin of ergativity in Sumerian, and the ‘inversion’ in pronominal agreement: A historical explanation based on neo-aramaic parallels. Orientalia, NOVA SERIES, 71 (3):267–290.

Cox, D. (1957). Candoshi Verb Inflection. International Journal of American Linguistics, 23 (3):129–140.

Dixon, R. M. W. (1972). The Dyirbal language of North Queensland. Cambridge: Cambridge University Press.

Gibson, M. L. (1996). El Munichi: Un idioma que se extingue. Yarinacocha, Pucallpa: Summer Institute of Linguistics.

Gundel, J. & Fretheim, T. (2006). Topic and Focus. In The Handbook of Pragmatics, 175–196. Oxford: Blackwell Publishing Ltd.

Hart, H. (1978). Yosë nanamën. Orlando: Wycliffe Bible Translators, Inc.

Hart, H. (1988). Diccionario chayahuita-castellano castellano chayahuita: Canponanquë nisha nisha nonacaso’. Lima: SIL International.

INEI. (2007). Resumen Ejecutivo Resultados Definitivos de los Censos en Comunidades Indígenas de la Amazonía Peruana.

Larsen, T. W. (1987). The syntactic status of ergativity in Quiche. Lingua, 71 (1-4), 33–59.

Levinson, S. C. (sub.). Syntactic Ergativity in Yéli Dnye, the Papuan language of Rossel Island, and its implications for linguistic typology. Manuscript.

McCawley, J. (1972). English as a VSO language. Language, 46 (2), 286–299.

McGregor, W. (1998). ’Optional’ ergative marking in Gooniyadi revisited: Implications to the theory of marking. Leuvens Constributions in Linguistics, 87 (3-4):491–571.

Michael, L., Beier, C., Sullón Acosta, K., Farmer, S., Finley, G., Roswell, M., Sinti Saita, M. (2009). Una breve descripcíon del idioma muniche. Berkeley: Cabeceras Aid Project.

Overall, S. E. (2017). A Grammar of Aguaruna (Iinía Chicham). Berlin/Boston: Mouton de Gruyter.

Parker, S. G. (1994). Datos adicionales del idioma Chamicuro. Pucallpa: Instituto Lingüística de Verano

Rojas-Berscia, L. M. (2013). La sintaxis y semántica de las construcciones causativas en el chayahuita de Balsapuerto. Lima: Pontificia Universidad Católica del Perú.

Rojas-Berscia, L. M. (2015). Mayna, the lost kawapanan language. LIAMES, Línguas Indígenas Americanas, 15, 393–407. Retrieved from http : // pubman . mpdl . mpg . de / pubman / faces / viewItemOverviewPage . jsp ? itemId=escidoc:2237461

Rojas-Berscia, L. M. & Bourdeau, C. (n.d.). The (Anti)-Nominal Hierarchy in Shawi’s Split Ergativity. Manuscript.

Schuchardt, H. (1896). Über den passiven Charakter des Transitivs in denkaukasischen sprachen. Sitzungsberichte der philosophisch-historischen Classe der Akademie der Wissenschaften zu Wien, 133, 1–191.

Seuren, P. A. M. (1993). The Question of Predicate Clefting in the Indian Ocean Creoles. In F. Byrne & D. Winford (Eds.), Focus and Grammatical Relations in Creole Languages (pp. 53-64). Amersterdam: Blackwell.

Seuren, P. A. M. (1996). Semantic Syntax. Oxford: Blackwell Publishers Ltd.

Shibatani, M. & Pardeshi, P. (2008). The causative continuum. In M. Shibatani (Ed.), The Grammar of Causation and Interpersonal Manipulation (pp. 85-126). Amerterdam/Philadelphia: John Benjamins.

Valenzuela, P. (2008). Acusatividad y ergatividad “opcional” en Shiwilu (Kawapana). Amerindia, 32, 205–221.

Valenzuela, P. (2011). Argument Encoding and Pragmatic marking of the Transitive Subject in Shiwilu (Kawapanan). International Journal of American Linguistics, 77 (1):91–120.

Valenzuela, P. (2012). La lengua shawi, 1–3. Retrieved from https://www. chapman.edu/wilkinson/ files/pdf/Shawintro.pdf

Valenzuela, P. (2016). “Simple” and “double” applicatives in Shiwilu (Kawapanan). Studies in Language, 40 (3):513–550.

Vallejos, R. (2016). A Grammar of Kukama-Kukamiria: A language from the Amazon. Leiden/Boston: Brill.

van de Visser, M. A. (2006). The marked status of ergativity (Doctoral dissertation, Universiteit Utrecht).