Ensaio teórico SYNTACTIC ERGATIVITY AND ARGUMENT HIERARCHY IN KADIWÉU* SANDALO Filomena revistadaabralin@gmail.com UNICAMP 3 1/2 SYNTACTIC ERGATIVITY AND ARGUMENT HIERARCHY IN KADIWÉU 177 194 177-194 http://creativecommons.org/licenses/by/4.0/

This paper analyses data of Kadinéu and shows, via constituent order and adverb placement, that, assuming Jehtiek & Camie's proposal, it is possible to appreciate clear evidence for syntactic ergativity in Kadinéu. The data for lists paper come from Sandalo (1995, 1997), from field notes (1993-1999), and from more specific field research in this topic.

 Resumo

Erie artigo analisa dados do Kadiwéu e, com base na ordem dos constituintes eposição do advérbio, mostra que, assumindo-se a proposta de Jtiinek e Camie (2003), é possível identificar claras evidencias de criatividade sintática no Kadinéu. Os dados deste artigo derivam de Sandalo (1995, 1997), de notas de campo (1993*1999) e de pesquisa de campo mais especifica sobre esse tema.

 Ergativity Estrutura Argumental Kadiwéu 1. Introduction

Person hierarchy has been a topic of concern in the typological literature since the seventies (cf. Dixon 1994). Only recently person hierarchy has been approached in formal linguistics. Aissen (1999, 2ÜO0) has formalized the phenomenon in terms of ranked constraints built out of relational hierarchies by means of Functional Optimality Theory. This is not, however, the only approach to person hierarchy in formal linguistics. Jelinek (1993, 2000), Isaak (2000), and Jelinek & Carnic (2003) have approached this topic in a very different perspective, and the papers by Aissen brought about an interesting quarrel about this phenomenon (cf. Carnic 2002). Jelinek & Carnic (2003) attempted to show that the phenomena of ergative splits, object shift, differential object marking, dative/accusative alternations, clitic placement, and voice alternations driven by argument hierarchies are sensitive to prcsuppositionality and they claim (following the work of Jelinek 1993, 2000 and of Isaak 2000) that all phenomena driven by argument hierarchies arc better explained from the perspective of Diesing’s (1992) mapping hypothesis syntactically encoded.

According to Dicsing (1992), there is a direct mapping between syntactic constituent structure and semantic structures at some level of representation. In diis hypothesis, the clause is divided into a nuclear scope (VP), that asserts the truth of the entities and provides the new information of the clause, and a restrictor, that asserts the presupposed information. Only non-presupposed material is allowed to stay in the nuclear scope. Presupposed material must leave the nuclear scope and be placed in the restrictive part of the clause.

This paper analyzes data of Kadiwcu and shows, via constituent order and adverb placement, that, assuming Jelinek & Camie's proposal, it is possible to appreciate clear evidence for syntactic ergativ- ity in Kadiwcu. The data for this paper come from Sandalo (1995, 1997), from field notes (1993-1999), and from more specific field research in this topic in January of 2U04 (during the Field Methods Class in EVELIN), and in November of 2003 and February of 2004 in rhe Kadiwcu territory.1

 2. Agreement and ergativity

Kadiwcu does not have case marking morphemes on nouns, but its agreement morphology indicates an ergative case system, as discussed below.

Kadiwcu has agreement prefixes for internal and external arguments. But direct arguments are in complementary distribution. There is a person hierarchy, 2 > 1 >3, that defines the argument that is morphologically marked. If die object is diird-person, a transitive verb agrees with the external argument regardless of the person of die subject:2

.

1 jema:

j-ema:n

1SUBJ-want/love

‘I love him/her.’

.

2 jema:naGa

j-ema:n-Ga

1SUBJ-want/love-pl

‘We love him/her.’

.

3 ema:ni

a-ema:n-i

2SUBJ-want/love-pl

‘You love him/her.'

.

4 yema:

y-ema:n

3SUBJ-wan t/love

'He/she loves him/her.’

.

But the verb agrees with the internal argument if the external argument is third person and the internal argument is first or second person. Tn this case, the morpheme d:- ‘inverse’ must be present:3

.

5 id:ema:

i-d:-ema:n

1OBJ-inverse-want/love

‘He/she loves me’

.

6 God:ema:

Go-d:-eman

1plOBJ-inverse-want/love

‘He/she loves us’

.

7 Gad:ema:ni

Ga-d:-eman-I

2OBJ-inverse-want/love-pl

'He/she loves you’

.

When there is no third person involved (that is, the direct arguments are first and second persons), the second person argument is marked. The inverse morpheme must be present.

.

8 Gad:ema:ni

Ga-d:-eman-i

2OBJ-inverse-want/love-pl

‘I love you’

.

9 ad:ema:ni

a-d:-eman-i

2SUBJ-inverse-want/love-pl

'You love me’

.

Intransitive verbs (i.e. unaccusatives, reflexives, and verbs that contain an incorporated noun) are marked by subject prefixes that differ from the subjects of transitive sentences.4 Below are some examples to illustrate the intransitive agreement pattern.

.

10 id.acotaGa

i-d:-acodi-Ga

3SUBJ-inverse-go.down-pl

'We go down.’

.

11 d:apiqo

Ø-d:-apiqo

3SUBJ-inverse-warm

‘It is warm.’

.

12 id:aqakGa

i-d:-aqag-Ga

1SUBJ-inverse-squat-pl

‘We squat.’

.

To sum up, Kadiwéu has a tripartite agreement system. Figure 1 below presents the three sets of agreement markers. Figure 1 attests that the intransitive set of agreement markers is different from the set of transitive subject agreement markers. Note that the fact that intransitive verbs are marked by a set of subject markers that differs from the set that marks the subject of a transitive verb indicates an ergative system.

Figure 1: Agreement

Subject Subject Object
1sg j- i- i-
2sg a-...-i a-...-i Ga-
3sg y- -w- Ø- ------
1pl j-...-Ga i-...-Ga Go-
2pl a-...-i a-...-i Ga-...-i
3pl t-...Ga o -Ø ------

One could, obviously, question whether the Kadiwéu agreement patterns indeed indicate an ergative system or whether its agreement system is merely a morphological idiosyncrasy. Next section shows by adverb positioning and other constituent order that Kadiwéu is indeed a syntactic ergative language that shows a split driven by person.

 3. The mapping hypothesis and ergativity in Kadiwéu

There are syntactic facts that show that Kadiwcu is indeed a syntactic ergative language. The first of these facts concerns focalization of subjects. As in syntactic ergative languages (see Bittner & Halle 1996), one can focalize an internal argument of a relative clause in Kadiwéu. Bur it is necessary to antipassivize a transitive verb to focalize its external argument in the same situation:

Other traditional tests for syntactic ergativity do not work consistently, however. Thus, if we coordinate sentences, the internal argument of a transitive sentence is preferred to be co-indexed with an intransitive subject but this is not obligatory. Also, it is better to antipassivize a sentence to relativizc its transitive subject but this is not obligatory. Thus, other tests must be developed to clarify the analysis of Kadiwéu case system. This is what follows.

As mentioned above, Jelinek & Carnie (2003) argue that all phenomena driven by argument hierarchies are a reflex of Diesing’s (1992) mapping hypothesis syntactically encoded. Recall that, in Diesing’s hypothesis, the clause is divided into a nuclear scope (VP), diat asserts the truth of the entities and provides the new information of the clause, and a restrictor, that asserts the presupposed information. Only non-quantificational/non-presuppositional material, (like nonspecific indefinites) is allowed to stay in the nuclear scope. Presupposed material (like definite NPs) must leave the nuclear scope and be placed in the restrictive part of the clause.

Jelinek (1993) notes that split case systems driven by person tend to occur in languages that do not have determiners. In these languages, third-person arguments are non-specific indefinites and diereforc they are allowed to stay in the nuclear scope. First and second person arguments arc intrinsically definites and diereforc they must leave the nuclear scope and be placed in the restrictive part of the clause.

In Kadiwéu, argument hierarchy affects agreement, as seen above, as well as constituent order, as it can be noticed in die data below. First/second person internal arguments must precede die verb (OV order) but third person internal arguments follow it (VO order):

The fact that first and second person, but not third person, internal arguments must precede the verb in Kadiwéu suggests that Jelinck & Carnie's proposal is on the right track. That is, this fact suggests that a definite internal argument must leave the VP.

Constituent order by itself, however, is not conclusive evidence. It is tradition in generative syntax to take adverbs as diagnostics for movement of oilier constituents. Assuming that adverbs arc adjoined to VP, all elements preceding the adverb may be argued to be outside VP, and all elements following the adverb are inside VP.

Kadiwéu does not have many adverbs. Most of the notions expressed via adverbs in better known languages are expressed via predicates in this language, but there are at least three real adverbs: efime ‘perhaps’, jaG ‘already*, and eG ‘still’.s These adverbs have exactly die same behavior concerning all the data discussed below.

3.1. Adverbs and external arguments:

Note that there is an interesting contrast in the pattern of adverb placement according to the person of the external argument:

.

20. jeG Ecabigo yema: Ekode

21. jeG Ecabigo aqa:mi Gadema:ni

22. *jeG ee aqa:mi Gadema:ni

23. *jeG ee jema: Ecabigo

.

A pronoun subject (first and second persons) cannot ever be preceded by an adverb, what shows, assuming that the adverb is adjoined to VP, that a subject pronoun must be in a high subject position (probably SPEC, TP) like the subjects of better known accusative languages.

As the examples above show, however, an adverb, in this case jaG ‘already’, can precede a third person external argument The fact that an adverb precedes a third person external argument suggests that this subject is not in SPEC, TP. It must be in a lower position, inside i-'P. Many authors have claimed that an ergative subject is licensed in situ (e.g. Nash 1996, Bitter & Hale 1996). This is the case of third person external arguments in Kadiwéu. They are ergative.

 3.2. External arguments in embedded clauses

A third person external argument can (optionally) occupy a pre-complementizer position in embedded clauses in Kadiwéu. An internal argument cannot, regardless of its person if the external argument is third person. In the data below the complementizer is me and it is underlined.

Note, however, that a first/second person external argument cannot be placed in the pre-complementizer position. The object moves (optionally) instead regardless of'its person:

The facts concerning the position of subjects in embedded clauses arc further evidence for the claim that third person external arguments and first/second person external arguments do not occupy the same syntactic position.

 3.3. Adverbs and internal arguments

The adverbs ejime ‘perhaps’, jaG ‘already’, and eG ‘stil’ can occupy any of die positions marked by (X) in the data below. The sentence is ungrammatical, however, if the adverb is placed in the position marked by (*).

39. Ecabigo ja yema: Ekode

40. *Ecabigo yema: jeG Ekode

.

41. Ecabigo jaG aqa:mi Gadema:ni

42. Ecabigo aqa:mi jaG adema:ni

.

The data show that an adverb cannot interfere between the verb and a post-verbal object. Recall that an object is post-verbal (VO order) when it is third person. The facts concerning adverb placement show that a post-verbal internal argument is internal to VP.

First and second person direct internal arguments cannot ever be post-verbal, however. They must be placed before the verb (OV order) and die inverse morpheme appears obligatorily. An adverb can occur between die verb and the object if die object is preverbal, and it shows that a prcvcrbal internal argument has moved to outside of VP. Preverbal internal position arguments are first and second persons. I believe that the inverse morpheme is the head of a functional projection diat receives an internal argument dislocated out of die VP. Note that die fact that an adverb can intervene between an ergative subject and an absolutive object show's that this object occupies the specifier position of an independent projection rather than a second specifier of jP.

Note thar although many theories of ergativity postulate diat die internal argument of an ergadve language occupies die SPEC posi- don of'IP (cf. Nash 1996 and Bittner & Halle 1995 for die languages that they label syntactically ergative), Kadiwéu docs not favor this hypothesis. Although a definite (first and second persons) infernal argument leaves the VP, it is lower than any subject. There arc wo pieces of evidence that it is lower in syntax than the subject: (i) it linearly follows the subject and (b) in embedded clauses a third person subject (die one diat is licensed in situ) has priority over an internal argument to move, as seen in section 3.2.

The Kadiwéu facts concerning internal arguments resemble some facts of North American languages. The inverse voice (morphologically marked here by the inverse d:-) is used when the internal argument is presupposed. Like in a passive, the internal object is fronted, but unlike the passive, there is no intransitivization and no argument is demoted.

Carnie & Jelinek adopt a particular view of phases that is different from die one proposed by Chomsky (2001). They state (2003:8):

.

“Chomsky proposes that phases are, in essence, propositional; they consist of a predicate and its arguments (fP), or a temporal and force operator (TP or CP). Camic offers an alternative view of phase. In this approach, phases minimally consist, of (a) a predicative element {v or V), (b) a single argument (NP), (c) a temporal operator that locates the predicate and argument in time and space (Asp or T). For a single transitive clause, then, the first phase of a sentence consists of a lexical predicate which expresses an event or state (V), any internal arguments, and the Asp head.”

.

The authors propose that definite objects shift from inside the VP (possibly at T.F in some languages) to the specifier of a projection above VP to get out of the nuclear scope defined by VP in better known languages. In Kadiwéu a presupposed internal argument (ab- solutive object) moves overtly. In other words, I believe that the inverse morpheme is a morphological realization of such projection.

4. Contrastive focus of objects

It is important to mention that while a first/second direct internal argument cannot ever be final, a third person internal argument can occupy a prcvcrbal position. Note, however, that, in this case, it is interpreted as in contrastive focus. A preverbal focused object docs not trigger agreement and the inverse morpheme does not appear:

.

43. Ecabigo Ekode yema:

Ecabigo Ekode 3SUBJ-love/want

Ecabigo loves Ecode (not somebody else)

.

One could question whether a third person in contrastive focus occupies the same position as the internal argument of the inverse voice (first/second person). Negation placement indicates that they do not occupy the same position:6

As it can be noticed above, a verb can be modified by aG ‘not* when there is a pre-verbal object pronoun (inverse voice) but the same is not true if the object is a focused noun. The negative morpheme must precede the object

Additional evidence for the claim that a focused object and an internal argument pronoun are not in the same position comes from the fact that there is no complementary distribution between a pre- verbal pronoun and a focused third person:

Note that the regular position of a noun object is final if not focused, even in a double object construction:

.

48. Ecabigo aqa:mi yolaGataGadomi libole.

Ecabigo gives you the meat.

 5. Intransitives adverb placement

Finally, it is important to mention that the behavior of intransitive sentences (unergativc or unaccusativc verbs) is different concerning adverb placement, what constitutes further evidence that transitive subjects arc special. Thus, an adverb can precede or follow a subject, regardless of its person, in an intransitive clause:

That is, the position occupied by an absolutivc argument is not the same as the one occupied by a transitive subject, regardless on whether die subject is nominative or ergative.

As a final remark, note that antipassives do not work as intransitives regarding adverb placement, they have the same behavior as transitive clauses:

The difference between antipassives and intransitives is not the presence/absence of an oblique argument Note that one can add an oblique argumcat in an intransitive clause, and still the behavior regarding adverbs positioning docs not change:

6. Conclusion

This paper attempts to show diat Kadiwéu is a syntactic ergative language that presents strong arguments to believe that there are different positions for nominative and ergative subjects as well as different positions for accusative and absolutive arguments. Furthermore, it shows that absolutivc objects do not fill the specifier of TP. It occupies a position outside of VP that is lower than the position of an ergative subject.

* This research Is funded by die SOAS/Endangered Languors Dommntation Program, FAPESP, and CNPq. I thank Norvin Richards for the opportunity to share a course during the KVKLIN I and 11 Summer School and for important suggestions regarding the analysis of Kadiwéu presented here. 1 also thank Andrew Nevins. Obviously the usual disclaim applies.

Kadiwcu is a Waikuriian language spoken by about 1,000 Indians distributed over an area of 538,000 hectares in the State of Mato Grosso do Sul, Brazil. The Waikuman language family has two branches: (a) the Waikurúan Branch, which includes Mbayá and its descendem Kadiwéu; and (b) the Southern Branch, which comprises four other languages: Toba, Pilagá, Mocovi, and Apibrtn (Ccria & Sandalo 1995). The Kadiwcus arc tlte only surviving descendants of the Mbayá people, who in the 18th eenrury dominated a large extension of the Brazilian and Paraguayan Chaco area (23,5° to 19° degrees of Latitude South, Sanchez Labrador, 1760). A short sketch in a 1760 grammar and dictionary hy Sanchez Labrador (published in Susnik 1971) is the only material available on Mbayá. Sanchc?. 1.abrador collected his data near Asuncion, Paraguay, so his data represent a dialect that presumably already differed from the immediate ancestor of Kadiwcu. Some aspects of the Kadiwcu grammar have been discussed by Griffdis & Griffiths (1976), Braggio (1981), and Griffiths (1973, 1987, 1991). A grammar and dictionary of Kadiweu arc in Sandalo (1995,1997).

1 The following abbreviations are used in Kadiweu examples in this work: 1 = first person, 2 = second person, 3 = third person, COMP = complementizer, NRG = negative, SUBJ = subject, OBJ = object, OBL = oblique argument, pi = plural, sg = singular. Note that Kadiwcu docs not allow hiatus; this language lias [t] epenthesis to avoid hiatus. An epenthetic (tj is glossed EPN.

1 Typologically similar morphemes have been called rtlaiional\i\ die literature of Brazi

lian Indian languages. I label di- invtnt, rather than relational, following a suggestion of Spike Gildea since it resembles some aspects of die inverse of North American languages. This paper attempts to show that die morheme d:- occurs when the object lias been dislocated to a higher specifier position. It is unknown, however, whether die morphemes of other Brazilian languages that arc labeled relational arc indeed similar to the kadiwcu inverse in its grammatical propcrrics.

1 Unergative verbs are not marked like intransitive verbs; they are marked like transitive

verbs whose internal arguments arc third person. This pattern is also attested in languages like Basque and Georgian. Note dial unergarive verbs are analyzed as lexically transitive by Hale & Keyscr (1993). Anti passive verbs can be marked by cither die transitive or intransitive subject marker.

1 The adverb jaG ‘already’ undergoes vowel harmony. Note also that the phoneme /

G/ drops before a consonant or glide and that the vowel harmony rule does not apply in diis ease.

1 Kadiwcu negative marker is aG, recall that there is a phonological rule that deletes / G/ before a consonant.

 637 711 17 1999 AISSEN J Natural Language and Linguistic Theory Markedness and subject choice in optimality theory Ms 2000 AISSEN J Differential Object Marking: iconicity vs. economy 1 68 27 1996 BITNER M HALE K Linguistic Inquiry The structural determination of case and agreement Brazil Universidade Estadual de Campinas 1981 BRAGGIO S Aspectos fonológicos e morfológicos do kadiwéu 169 191 37 1995 CERIA V SANDALO F Anthropological Linguistics A Preliminary Reconstruction of Proto-Waikurúan with Special Reference to Pronominais and Demonstratives MS 2002 CARNIE A Phase Geometric Hierarchies MIT Ms 2001 CARNIE A Derivation by Phase Cambridge, MA MIT Press 1992 DIESING M Indefinites Cambridge Cambridge University Press 1994 DIXON R. M. W Ergativity 63 67 Rio de Janeiro Instituto de Antropologia Prof. Souza Marques 1973 GRIFFITHS G Arquivos de Anatomia e Antropologia I Numerals and demonstratives in Kadiwéu England Reading University 1987 GRIFFITHS G Relative Clause Formation and other Word Parameters in Kadiwéu England Reading University 1991 GRIFFITHS G Wh-movement in Kadiwéu Série Linguística 6 Summer Institute of Linguistics 1976 GRIFFITHS G GRIFFTHS C Aspectos da língua Kadindu. Brasília University of Massachusetts 2000 ISAAK A Split Case Marking and Prominence Relations 15 42 18 1993 JELINEK E MIT Working Papers in Linguistics Ergative ‘Splits’ and Argument Type 51 70 1 2000 JALINEK E Papers in Honor of Ken Hale. Working Papers in Endangered and less familiar languages Datives and Argument Hierarchies Ms 2003 JALINEK E CARNIE A Argument Hierarchies and the Mapping Principle Université Paris-8 1995 NASH L Portée argumentale et marquage casuel dans Its langues SOV et dans its langues ergatives: l'exemple du géorgien 1 166 Asunción, Paraguay Musco Etnográfico «Andrés Barbero». (Lenguas Chaqueñas 1.) 1760 SANCHEZ LABRADOR J. S. I SUSNIK B Família Guaycuru Gramática Eyiguayegi-Mbayá. Según cl manuscrito del siglo XVIIT University of Pittsburgh 1995 SANDALO F A Grammar of Kadiwéu 11 1997 SANDALO F MIT Occasional Papers in Linguistics A Grammar of Kadiwéu with Special Reference to the Polysynthesis Parameter