University of Oregon
The purpose of this paper is to offer a framework for organizing the three primary types of ergative main clauses in the Cariban language family, and to briefly characterize what is known about the morphological and syntactic properties of each. I will also discuss briefly the main point of typological interest, which is that the tense-aspect based ergative splits in individual northern Cariban languages do not conform with expected universal patterns (cf. §4). While I will offer brief examples from individual languages to illustrate the claims I make here, in most cases, many more examples and much more detailed argumentation can be found in prior publications (cf. specific citations in various subsections). I encourage the interested (and especially the skeptical) reader to use this brief synopsis as a guide to points of interest that can be found in those works.
The first important point to clarify is what I intend by the term "northern Cariban languages": this is not a genetic unit — e.g. a branch — of the family, but is rather a loose geographic category that subsumes all Cariban languages spoken to the north of the Amazon River, in Brazil, French Guiana, Surinam, Guyana, Venezuela and Colombia. Figure 1 presents my most recent hypotheses about the classification of the Cariban family, taking as a foundation the proposal found in Kaufman (1994), as modified subsequently by Meira (2000a), Gildea (to appear), and Mattéi-Muller (2002). While the "Northern Cariban languages" do not form a coherent category in this classification, it is nevertheless the case that the only languages specifically excluded from this synopsis all belong to the proposed Southern Branch P-Q. In anticipation of the subject of the subsequent sections, I have highlighted languages that have ergative main clause constructions as follows: languages presenting the Full Set II clause type occur in boldface; languages that present the Partial Set II clause type occur in italic; and languages that present the *t-V-ce clause type occur with with an underline.
The most important fact to note when discussing ergativity in the Cariban family is that every language that has ergative clause types presents a split with other clause types in the language that are non-ergative. My explanation for this fact is that the ergative clause types are always innovative, and so in most cases the split is between innovative clause types that present one of the three ergative patterns, and conservative clause types that maintain non-ergative patterns (Gildea 1998, ch. 4). In two cases, that of Kuikuro and Makushi, the innovative ergative patterns have become the dominant pattern and (nearly ) all traces of the conservative non-ergative clause type have been lost. However, these languages still present split ergativity in that they each also present a new non-ergative clause type, for Makushi the progressive (Gildea 1998, ch 12), and for Kuikuro the de-ergative in interactive modes (Franchetto 1990, this volume; Gildea 1998, ch. 11). I return to the question of the splits in section 4.
A. Opón-Karare language
B. Yukpa group: Yukpa, Japreria, †Koyama
C. Kari'nya language
D. Tiriyó group
D1. Tiriyó subgroup: Akuriyó, Tiriyó
D2. Karihona subgroup: Hianákoto, Karihona
Parukotoan Branch E-F
E. Kashuyana group: Katxúyana-Warikyana, Xikuyana
F. Waiwai group: Waiwai, Hixkaryana
North Amazonian branch: G-H
G. Jawaperi group: †Bonarí, Jawaperi (Waimirí-Atroarí)
H. Paravilyana group
H2. Paravilyana subgroup: Pawishiana, †Paravilyana
Venezuelan Branch: [[I-T]-N]-J-O
I. Pemong group
I1. Pemong proper subgroup: Makushí, Pemóng (Taurepang, Kamarakotó, Arekuna), Kapóng (Akawaio, Patamona, Ingarikó).
T. Panare language
N. Mapoyo-Yavarana language (†Tamanaku)
J. †Kumaná language (†Chayma, †Cumanagoto)
O. Makiritare group: Makiritare (De'kwana), Wajumará
Southern Branch: P-Q
P. Bakarí group: Bakairí, Amonap (Kuikúro, Kalapalo)
Q. Arara group: Arara-Pirirí, †Apiaká-Apingi, †Juma,
†Yarumá, Chikaon (Txikão)
K. †Yao group: †Tiverikoto, †Yao
L. Wayana group: Wayana, †Arakajú
R. †Palmella language
S. †Pimenteira language
Figure 1. Classification of the Cariban language family
Modern reflexes of Full Set II ergative main clauses are attested in seven languages from five groups: C - Cariña; I -Makushi, Pemón and Kapón; P - Kuikúro and Kalapalo; and T - Panare (Gildea 1998, ch. 9). More recent work suggests Full Set II systems also in N - Mapoyo/Yabarana (Mattéi-Muller 2000) and †Tamanaku (Meira and Gildea in progress). The Partial Set II verbal system is clearly attested in Panare, and perhaps occurs also in Kaxuyana and Yukpa (Gildea 1998, ch. 10). The *t-V-ce system is clear in Wayana, Tiriyó, and Kaxuyana, antecedents (i.e. somewhere between passive voice and active ergative clause) can be seen in Apalaí and Carib of Surinam, and hints of the system are found in Bakairi (Gildea 1998, ch. 13). More recent work on Tamanaku (Meira and Gildea in Progress) brings together all examples of Gilij's (1965. III:268) "aorist tense", which is clearly another instance of the *t-V-ce system, and recent work on Kuikuru (Franchetto this volume) shows the *t-V-ce system in that language as well.
The three categories of ergative clause types have two distinct origins: Full Set II and Partial Set II come from nominalized clauses (sometimes as objects of postpositional phrases) functioning as predicates (often with a copula, which then becomes an auxiliary); *t-V-ce comes from an adverbialization that began to be used in a passive construction, and from there was reanalyzed as main clause ergative. While each system is different (and hence must be treated separately), they do arise from the same pool of source morphology and syntax, so there are overlaps in their synchronic morphology and syntax. Also, as each system occurs in multiple languages, each system presents at least some unique properties in each language. I turn now to a description of the morphology of each system.
The full Set II and Partial Set II systems both present the same series of absolutive prefixes or proclitics, derived historically from the absolutive possessor of the nominalized verb. Both systems also present an absolutive number marker as a suffix or enclitic. The reconstructed forms are given in Table 1 (from Gildea 1998.114) followed by illustrative examples from Akawaio (1) and Panare (2).
Absolutive prefixes / proclitics: Absolutive collective suffixes / enclitics:
*u(y)- ‘1Abs’ *komo 'Collective Absolutive'
*a(y)- ‘2Abs’ (Proto-Carib)
*y(ï)-‘3Abs’ -'nogoN 'Collective Absolutive'
*k(ï)-‘1+2Abs’ (Pemong Group)
Table 1. Reconstructed absolutive person and number markers
(1) Akawaio (Kapóng) illustrating Full Set II absolutive proclitics and the number enclitic
S V s-V
a. kone'o e' -pödï -'pï b. y- e' -podï -'pï
rabbit be-FREQUENTATIVE-PAST 3-be-FREQUENTATIVE-PAST
‘Kone’o used to be...’ ‘S/he used to be.’
s-V-s O V A
c. a-dö-bödï-'pï-gong d. t- are'na pïrï'sï'ka-'pï i-ya
2-go-PAST-PL.ABS 3REFL-tail detach-PAST 3-ERG
‘you-all used to go.’ ‘He pulled his own tail out.’
o-V A o-V A=o
e. i-pïrï'sï'ka-Ø u-ya f. a-wönö-'pï i- ya ='nogong
3-detach-PRESENT 1-ERG 2- hit -PAST 3-ERG =COLLECTIVE
‘I pulled it out (with one swipe).' ‘He hit you-all.’
(2) Panare as illustration of Partial Set II absolutive proclitics, nominative Aux agreement
s-V (s:Aux) S S V S V
a. y-u-të-n (këh) e'ñapa b. Ake wë- të -n mënkae wi wë-të -n tïna kawa
3-SA-go-PRES 3.be Panare snake SA-go-PRES like boa SA-go-T/A water
‘The Panare goes’ ‘The venomous snake goes like the boa goes,
through the water.’
O V A o-V (a:Aux) A
c. amën pétyuma-ñe yu d. a-petyúma-ñe (këh) kën
2SG hit -PRES 1SG 2- hit -PRES 3.be 3SG.DISTAL
‘I hit (castigate) you.’ ‘S/he hits (castigates) you.’
There is some evidence that the postverbal ergative pronoun in Makushi, Pemóng and Kapóng cliticizes to the verb, creating what some have described as ergative suffixes. This is treated in the next section, on nominal case-marking.
Whereas the Proto-Carib verbal system had no case-marking on A, S or O NPs, and the Partial Set II system continues to present no case-marking on these arguments, both the Full Set II and *t-V-ce main clauses in Northern Cariban languages mark the A with an ergative suffix/postposition, a modern reflex of the postposition *wïya 'Allative/Dative/Causee/Agent'. (Note: southern Cariban languages Kuikuro and Kalapalo mark the ergative with -peke/-heke, apparently most recently derived from an ablative postposition). This morpheme, used in so many different constructions with so many different meanings, has undergone idiosyncratic phonological reduction across the family, with only the final vowel attested in all synchronic reflexes (Table 2).
Ergative nouns and free pronouns are always free forms followed by the ergative postposition/enclitic (degree of phonological dependency on the (pro)noun varies depending on the language & analyst). Full pronouns are treated just like nouns. Ergative pronouns are (at least etymologically) the same person prefixes as the absolutive verbal prefixes, except in this case prefixed to the ergative postposition. In some languages the 1+2 prefix has been lost, including as a formative for the ergative pronoun, so the modern 1+2 ergative pronoun is the full free pronoun followed by the ergative postposition/suffix (Table 3).
Panare, Tamanaku: uya
Pemong Group: -wya / -ya / -θa / -a
Kari'na: 'wa / :wa
Mapoyo: -wa, -wah, -ya, -a
Table 2. The ergative case-marker as it occurs on nouns and pronouns
The idiosyncratic changes in the first and second person ergative pronouns in the Pemong group correlate with their reanalysis as ergative verbal enclitics. Syntactically, these reduced-size ergative pronouns only occur immediately after the inflected verb. Interestingly, although it modifies the absolutive argument (bound to the very front of the verb), the absolutive collective number enclitic follows the ergative enclitic (cf. 1f), further encouraging the analysis that the ergative suffix is becoming part of a single complex of verbal morphology. Additionally, when the ergative suffix follows the Perfect/Past inflection –sa' (Makushi) / -sak (Akawaio), there is metathesis between the final segment of the inflection (realized as a glottal stop in both languages) and the initial vowel (which coincides with the person-marking portion of the ergative pronoun). This is illustrated with Akawaio data (3).
1Erg 2Erg 1+2Erg 3Erg
Proto-Carib *u-wïya *a-wïya *kï-wïya *i-wïya
Makushi -u-ya -Ø-ya -Ø i-ya
Pemón -Ø-ya -u-ya -Ø-kon (*-ya) i-ya
Kapón u-ya a-wya urö'nogong-ya i-ya
Panare y-úya oya (< a-uya) yuto uya ty-úya
Ø-wïya yutakon uya
Cariña u-'wa a-'wa kï-'wa i-'ya
Mapoyo/ ih-a, wïh-a ër-a ——— ehnë-ya, ti-a, tëy-a
Yabarana wïrë-ya, urë-ya mëre-ya tëwï-ya, tawï-ya
Tamanaku Ø-uya a-uya kiwe uya ite-uya
Tiriyó Ø-wïya ëë-ya kïï -ya ii-ya
Apalaí ï-a o-a kï-a e-a
Table 3. Variation in ergative pronouns/verbal enclitics
(3) Akawaio ergative enclitics
a. /a-kö'ma-sak u-ya/ [agə'mazau'ya] ‘I have called you.’
b. /u-kö'ma-sak i-ya/ [gwə'mazai'ya] ‘S/he has called me.’
c. /i-kö’ma-sak au-ya/ [gyə'maza' awya] ‘You have called him/her/it’
[gyə'mazaw'ya] (homophonous w/ 3a)
The immediate etymology of the ergative postposition/case-marker is from the optional agent-phrase as marked in nominalized clauses, and as innovated in adverbialized clauses during the evolution from simple stative forms to eventive passives (cf. Gildea 1997.185ff). However, case syncretism is extreme with this form, which is also attested to mark a range of oblique functions, including allative, recipient, addressee, and causee (these last three might be argued to be Indirect Objects rather than obliques, but no convincing morphosyntactic evidence has been put forward in a single Cariban language to date that distinguishes any proposed IO from the rest of the obliques). A typologically informed internal reconstruction would suggest the following stages in the evolution that gives rise to such polysemy (all but the third step in the chain, Causee > Agent-Phrase, well-attested in the literature):
Allative > Dative > Causee > Agent-phrase in Nominalizations > Agent-Phrase in Passives
In this section I distinguish between the more concrete syntax of constituency (§3.1) and the more covert subject properties of coreference control (§3.2).
Phrase-structure constituency is generally straightforward to describe in Cariban languages, with multiple grammatical patterns all converging on a single analysis (cf. Franchetto 1990.408; Payne 1993.124-126; for a historical perspective, cf. Gildea 2000.72-81; for some less clear cases, cf. also Payne 1994, Meira 1999.493-499). The conclusions offered here are basically summaries distilled from previously published comparative work (Gildea 1998, ch.9-14, 2000 inter-alia); rather than repeat the many examples here, I direct the interested reader to those works, especially Gildea 2000.
Full Set II and Partial Set II clauses both present a clear [OV] verb phrase in transitive clauses and a clear [SV] verb phrase in intransitive clauses. In the transitive Full Set II clause, the ergative-marked A noun may occur either before or after the verb, with no evidence for any syntactic difference between the two; the person-inflected ergative markers (called ergative pronouns in the preceding section) are generally postverbal. In the partial Set II clause, the most common order for a full noun A and S (almost rigidly required, cf. Gildea 1993) is following the verb and its auxiliary.
Constituency does not group either A or O with S in the *t-V-ce clause type, as no constituency patterns have been identified for any of the three arguments. The syntax of the *t-V-ce clause type has been best documented for Tiriyó (Meira 1999) and Carib of Surinam (Hoff 1995), neither of which show any evidence for a VP, or for any clause-level syntactic asymmetry between A, S, and O with relation to V. The only order restriction attested is that, in Carib of Surinam, the copular auxiliary must follow the inflected verb. In a brief survey of one Wayana narrative text collected and analyzed by Tavares (pc), I found no examples of transitive verbs with both arguments explicit (and in fact many clauses consisted of only the verb and particles); A, S and O as single explicit arguments occurred freely both pre- and postverbally.
As Berend Hoff explained so clearly for Carib, the Cariban family does not present many of the syntactic properties that we would like to explore in search of ergative versus accusative patterns, which we could then use to argue for categories like "subject" (e.g. Givón's 1997.25, 2002.178 lists of "behavior and control" properties).
Other observations [subject tests — SG] are precluded by the peculiar properties of Carib grammar: there are no relative pronouns and no conjunctions, and dependent (nonfinite) verbs contain personal prefixes too, including two third-person prefixes distinguishing identity or non-identity with a participant referent of the main verb. Also, question words can easily reconcile their location in front position with any role. Hoff 1995.362, note 19.
To Hoff’s list I would add that analytical constructions involving changes of valence, such as passive, antipassive and applicative, generally do not exist, and as such we cannot readily identify categories of grammatical relations that are "promoted" or "demoted" in syntax. Different causativization constructions exist for transitive and intransitive verbs, preventing a relationship between A and S as "causees". Relativization is generally done by means of nominalizations (e.g. instead of "the man who left" we find "the man, the leaver", and instead of "the knife that I bought", we find "the knife, my buyee/bought one", etc.) so a hierarchy of accessibility to relativization does not differentiate between nominative or ergative patterns. Topicalization appears to be done by simple changes in word order or by clefts that utilize these same nominalizations, and as such, "extraction" appears not to be a relevant concept in the grammar of most Cariban languages. That said, control of coreference does appear to be relevant in nearly every Cariban language, as documented in §4.2.1-4.2.3.
The one test that is available in almost every Cariban language is the control of coreference with a third person logophoric possessive prefix, *t- '3Refl'. This prefix can occur on any noun or postposition in the clause except the head noun of the S/A noun phrase. It is always understood as being coreferential with A and S in Carib, Apalaí, Wayana, Katxúyana, Tiriyó, Makushi, Akawaio (Kapóng) — i.e., everywhere it has been tested (Gildea 1997, 1998). I illustrate the ergative A controlling coreference with t- in Makushi (4) and Carib of Surinam (5).
(4) Logophoric prefix in Full Set II clause type in Makushi (Abbott 1990.44, 111)
[ 3i -Loc ] o-V-ai
a. t- ette yo’ko i- tîrî-‘pî -i -ya
3Refl-hammock under 3Abs-put-Past-3-Erg
‘He put it under his hammock.’
[ ti -Loc ] si -V
b. t- akon pia attî -‘pî
3Refl-brother to 3:go-Past
‘He went to his own brother.’
(5) Logophoric prefix in *t-V-ce clause type in Carib of Surinam (Hoff 1995.364)
a. ti-O V Ai o:AUX
[tïi -saano ]j t- otaawa-Ø ii-/wa man
[3Refli-mother]j Ad-visit-Prtcpl 3i-Agt 3j.Cop
‘Hisi own motherj wasj visited by himi.’
b. * V [ t-A ] o:AUX
*t- otaawa-Ø [[ t -saano ] /wa] man
Ad-visit-Prtcpl [[3Reflj-mother ]i Agt] 3j.Cop
((hei) wasi visited by hisi own motherj)
In most Cariban languages, there is either no evidence regarding coreference conditions between coordinate clauses or else the evidence suggests that coreference is flexible, and will be interpreted according to pragmatic likelihood rather than being determined by grammatical rule. This is not particularly surprising, since verbal morphology distinguishes all participants except in the case where the first verb contains third person singular acting on third person singular and the second contains only third persons again. In two languages, rules have been proposed, and in both cases, the *t-V-ce clause type presents an absolutive controller of coreference (6-7). Hoff (1995) uses this as his primary criterion for deciding that the Carib version of *t-V-ce is best analyzed as a passive construction, whereas Meira (1999) merely notes that this is the sole ergative syntactic property that is found with the Tiriyó version of *t-V-ce clauses, which he analyzes as a Remote Past with ergative alignment.
(6) Absolutive control of coreference in Carib of Surinam (Hoff 1995.362)
Oi V o:AUX [ A ] // Si si-V
a. seeri t- owa/ma -Ø n -eei Baaku /wa indombo Ø n -oreh -ko -i
Silviaj Ad-embrace-Prtcpl 3j-be Baaku Agt then 3j-angry-Inch-TAM
‘Silviaj was embraced by Baakui, then (shej) became angry.’
*‘...then (hei) became angry.’
Ai O V // Si si -V
b. Baaku Seeri ewa/ma -i, indombo Ø n-oreh -ko -i
Baku Sylvia embrace-TAM then 3-Angry-become-TAM
‘Baku embraced Sylvia, and then (he) became angry.’
(7) Control of coreference in Tiriyó (Meira 1999.507)
a. Absolutive control with coordinated *t-V-se clauses
V Si // Oi V A
ma, irë_mao _rëken tï- w- ëe -se ariweimë, Ø t- ëne -Ø ii- ja
well, only.then RM.PST-SA-come-RM.PST cayman RM.PST-see-RM.PST 3-ERG
'Well, only then did the caimani come, and hej saw himi.'
b. Nominative control with coordinated Set I (non-ergative) verbs
Si si-V // Ai Oj a/o-V
wei waraarë Ø n- ee -ja -n, Ø Ø ni- tuuk -Ø -n
day every 3SA-come-PRES.IMPRF-DOUBT 3A3O-hit-PRES.IMPRF-DOUBT
'Every day hei comes and then (Øi) hits (Øj)'
Gildea (1997.168-174) presents the evidence against such coreference rules for the *t-V-ce clause types found in Apalaí and Wayana. To my knowledge, coreference patterns in interclausal coordination have never been documented in any language possessing either the Full Set II or the Partial Set II clause types.
Most subordinate clauses across the Cariban family are nominalizations (cf. Gildea 1998, ch. 7), which avoid questions of coreference by simply taking possessive prefixes that leave all participants clearly identified. However, coreference relations have been identified for two types of subordinate clause, the Purpose of Motion (supine) clause found in almost every Cariban language (§126.96.36.199) and the innovative ahtao 'if/when' clause in Tiriyó (§188.8.131.52).
The purpose of motion verb (called 'Supino' by Mosonyi 1986 and 'Supine' by Meira 1999), is formed by adding to the verb a modern reflex of *-ce 'Purpose' (identical to the second half of the *t-V-ce inflection), which then heads a dependent clause that contains no mention of either A or S. As implied by the name, the main clause always entails motion, and either the S of an intransitive verb of motion or (in a few languages) the O of a causative verb of motion (like ‘send’) controls coreference to the A/S in the purpose of motion clause. It makes no difference whether the matrix clause is ergative or not, this pattern does not change (8-9).
(8) Nominative control of A/S in Purpose of motion clauses in Makushi (Abbott 1990.76-77)
Si V [ O V Ai]
a. tiaron-kon wîtî-'pî, moro' yapî' -se Ø
other-Coll go-Past fish catch-Purp
‘Others went to catch fish.’
[Si V ] Si V [ Loc ]
b. moropai Ø su'minan-se to' epa'ka-pîtî-'pî poro pona
and play -Purp they go.out-Iter-Past outside to
‘And they went outside to play.’
(9) Nominative control of A/S in Purpose of Motion clauses in Tiriyó (Meira 1999.506)
[Ai o-V ] si-V [Si V ] si-V
a. Ø ë- eta -e wï- të -Ø -e b. Ø etainka-e wï- të -Ø -e
2-hear-Supine 1Sa-go-Pres.Imprf-Cert run-Supine 1Sa-go-Pres.Imprf-Cert
‘I'll go there to listen to you.’ ‘I'll go there to run.’
Tiriyó has innovated two new subordinators that follow any finite clause, making it into an adverbial clause. Relations between the main clause and the adverbial clause are usually marked by morphology, but in those rare cases where morphology leaves ambiguity, the nominative argument of the main clause controls coreference with the nominative argument of the adverbial clause. Meira's (1999) otherwise comprehensive grammar does not mention these patterns, so I cite them here from Meira (1995), as previously published in Gildea (1997.177). The documented pattern is for the main clause controller of coreference to be S/A, and for the adverbial clause controlled referent to be S; the examples given here illustrate only A controlling S, whether A is nominative (10a) or ergative (10b).
(10) Nominative control of S of Adverbial clause in Tiriyó
[Si si-V Sbrd ] Ai a/o-V Oj
a. [Ø n- urakanun -ya -n ahtao] pahko n-ene-Ø yi-pawana
3i-walk:around-TAM-Evid while 1:fatheri 3i-see-TAM 1- friendj
‘While (hei was) walking around, my fatheri saw my friendj’
*‘While (hej was) walking around, my fatheri saw my friendj’
[Si si-V Sbrd ] Oj V Ai
b. [Ø n- urakanun -ya -n ahtao] yi-pawana t- ëne -Ø pahko ya
3i-walk:around-TAM-Evid while 1-friendj Ad-see-Prtcpl 1:fatheri Agt
‘While (hei was) walking around, my fatheri saw my friendj’
*‘While (hej was) walking around, my friendj was seen by my fatheri.
We can hope that as a part of Meira's continuing work with Tiriyó, he will further document coreference patterns between main clauses and adverbial clauses.
This is actually the domain in which uniquely Cariban ergativity has the most to offer. There are no Cariban languages with exclusively ergative main clauses: all northern Cariban languages with ergative main clauses present tense-aspect based splits of some sort, and each of the three types of ergativity presents an apparently counter-universal pattern. I begin with the finite-nonfinite split (§4.1), found in virtually all Cariban languages, then examine splits involving the *t-V-ce (§4.2), Partial Set II (§4.3), and finally Full Set II (§4.4).
All subordinate clauses are either nominalized (Gildea 1998, ch. 7) or adverbialized (Gildea 1998, ch. 8), hence the majority contain ergative orientation. (note: a few innovative finite subordinate clauses have been attested — these exceptions are found only in a few languages, and they constitute a small minority of the subordinate clauses in the languages where they are attested). Thus, subordinate clauses throughout the family present ergative patterns, with the result that every Cariban language has a split between ergative subordinate clauses and some non-ergative main clause type (in several languages, there being no ergative main clauses, and in all languages, there being at least some non-ergative main clause type).
Given the origin of the *t-V-ce inflection from a verbal adjective cum passive participle, the expected evolution would be verbal adjective > passive participle > perfect inflection > perfective/past inflection. As expected, modern *t-V-ce clauses are generally past-perfective (cf. Carib, Apalaí, Tiriyó, and Katxúyana, Gildea 1997, and Gilij's 1965 label of ‘Aorist’ in †Tamanaku). However, both Petronila Tavares (pc) and Eliane Camargo (pc) have overheard conversational examples of Wayana t-V-he to refer to past, present and future situations, certainly for perfective, and apparently also for imperfective aspects. Tavares' work in progress documents many instances, and will (hopefully) find a clear pattern that correlates with (and thus could be said to condition) selection of the ergative clause type. I do not give examples here, as neither has published on this topic yet; I did not encounter such examples during my brief stays with the Wayana, and I do not understand this aspect of Wayana grammar myself.
Panare shows counter-universal splits, as laid out in Gildea (1993; 1998, ch 2, 4). The most salient split is that between the clearly ergative Partial Set II clause type, instantiated in at least 5 inflections, and the nonergative Set I clause type, instantiated in 4 inflections. As can be seen in (11), the non-ergative inflections are all past tense (and also perfective), whereas the ergative inflections include nonpast, future, imperfective, and irrealis. This is precisely the opposite of the expected direction for a tense-aspect based split.
Partial Set II inflections Set I inflections
-n/-ne ‘Nonpast’ -yah ‘Immediate Past’
-sehpa ‘Future’ -i ‘Medial Past’
-tëh ‘Desiderative’ -yake ‘Remote Past’
-poi ‘Abilitative’ -i ‘Past Interrogative’
Table 4. The Ergative – Nonergative split in Panare
Three different patterns of splits are attested involving Full Set II clauses, each the result of independent innovation. The most conservative of the three is found in Kari'nya (Carib of Venezuela), which presents a single Full Set II tense, the future (§4.4.1); the most innovative language in this group, Makushi, presents almost exclusively Full Set II tenses and aspects (§4.3.2). But the most confusing split is attested in both Kapóng and Pemóng, which are halfway through a changeover from the prior nonergative Set I system of inflections to the innovative Full Set II inflections (§4.3.2).
As documented in Mosonyi (1986) and reiterated in Gildea (1992, 1993, 1998, ch. 15), the sole main clause that gives an ergative pattern in Cariña is the Full Set II future tense, again, precisely the opposite of the expected direction for such a split. As this is documented in several publications, I do not illustrate it further here.
Notwithstanding Abbot's (1990) assertion that Makushi is a completely ergative language, I believe it is clear that (at least) the innovative progressive clause type in Makushi is nonergative (Gildea 1992, ch. 9; 1998, ch. 12). Since all tenses and aspects are ergative except progressive, which is nominative, the split is in the expected universal direction (cf. Gildea 1993).
In this section I will summarize work in progress between myself and my colleague, Desrey (Fox) Caesar, a native speaker of Akawaio (Kapóng). Some examples from the work of José Álvarez (1998, plus addition examples given in personal communication) have shown parallel phenomena for Pemóng (viz. past, present and future in both ergative and non-ergative clause types). In examining main clauses in Akawaio, Caesar and I have encountered the following tense-aspect based split:
-'pï ‘Past-Perfective’ -i ‘Immediate Past-Perfective’
-sak ‘Perfect’ -dai ‘Distant Past-Perfective’
-Ø ‘Nonpast-Imperfective’ -(')aik ‘Nonpast-Imperfective (certain)’
-do'odong ‘Future’ -(')ang ‘Nonpast-Imperfective (uncertain)’
Table 5. The ergative – nonergative split in Akawaio (Kapóng)
In considering what might condition this split, we are immediately forced to discard all but one of the categories proposed in Dixon (1994): this is clearly not conditioned by person, finiteness, or verbal semantics — Tense-Aspect-Modality is the only possibility. But Akawaio presents both past and nonpast tenses, and both perfective and imperfective aspects, on each side of the split. Clearly, this is a language in the midst of a major renovation of the main clause tense-aspect-modality system, and speakers are apparently presented with a choice, for almost any tense-aspect of utterance they would like to make, between the new (ergative) clause type and a conservative (non-ergative) clause type. A major focus of our research in the coming years will be trying to understand in maximum semantic detail what conditions the choice of each individual inflection, after which we will try to find some pattern that holds across all inflections of one type or the other. At this point, we are not hopeful that such a pattern will be motivated by any “ergative” function that has been proposed thus far in the typological or theoretical literature.
ÁLVAREZ, José. 1998. Split ergativity and complementary distribution of NP’s and pronominal affixes in Pemón (Cariban). Opción 25.69-74.
CAESAR (Fox), DESREY. In Progress. zauro’nöndok Agawaio yau: Variants of Akawaio Spoken at Waramadong. Rice University Ph.D. dissertation.
DIXON, R. M. W. 1994. Ergativity. Cambridge: Cambridge University Press.
GILDEA, Spike. 1992. Comparative Cariban Morphosyntax: On the Genesis of Ergativity in Independent Clauses. Ph.D. Dissertation, University of Oregon.
1993. On the evolution of a counter-universal pattern of split ergativity. In Proceedings of the Ninth Eastern States Conference on Linguistics. ed. by Michael Bernstein, 103-16. Ithaca: Cornell University, Department of Modern Languages and Linguistics.
1997. Evolution of grammatical relations in Cariban: How functional motivation precedes syntactic change. Grammatical Relations: A Functionalist Perspective, Typological Studies in Language, v. 35, ed. by T. Givón, 155-198. Amsterdam: John Benjamins Press.
1998. On Reconstructing Grammar: Comparative Cariban Morphosyntax. Oxford: Oxford University Press.
2000. On the genesis of the verb phrase in Cariban languages: Diversity through reanalysis. Reconstructing Grammar: Comparative Linguistics and Grammaticalization Theory, Typological Studies in Language, v. 43, ed. by Spike Gildea, 65-106. Amsterdam: John Benjamins.
To appear. The Venezuelan Branch of the Cariban Language Family. Amérindia
GILIJ, Filippo Salvatore. 1780-84. Saggio di Storia Americana, vols. 1-3. Rome. (1965. Ensayo de Historia Americana, vols. 1-3. [Trans. Antonio Tovar]. Caracas: Italgráfica).
GIVON, T. 1997. Introduction. Grammatical Relations: A Functional Perspective. Typological Studies in Language 35, ed. by T. Givón, 1-84. Amsterdam: John Benjamins.
2002. Syntax: An Introduction, v. 1. Amsterdam: John Benjamins.
HOFF, Berend. 1995. Configurationality and non-configurationality in the Carib language of Surinam. International Journal of American Linguistics 61.347-77
KAUFMAN, Terrence K. 1994. The native languages of South America. Atlas of the World’s Languages. Ed. Christopher Mosely and R. E. Asher, 46-76 (pp. 72-74). New York: Routledge.
MATTEI-MULLER, Marie-Claude. 2000. Mapoyo, Yawarana, two variants of a same Carib language of Venezuela: New outlook on ergativity and classification of Cariban languages. ms.
2002. En busca de criterios clasificatorios para la Familia Caribe. Línguas Indígenas Brasileira: Fonologia, Gramática e História. Atas do I Encontro Internacional do Grupo de Trabalho sobre Línguas Indígenas da ANPOLL, pp. 37-53. Belém, Brazil: Editora Universitária Universidade Federal do Pará.
MEIRA, Sérgio. 1995. From adverbs to verbs in Tiriyó (Cariban). Presented at the Summer Meeting of SSILA, Albuquerque, New Mexico.
1999. A Grammar of Tiriyó. Ph.D. dissertation, Rice Uiversity.
2000a. A Reconstruction of Proto-Taranoan: Phonology and Morphology, LINCOM Studies in Native American Linguistics 30, Munich: Lincom Europa
2000b. The accidental intransitive split in the Cariban family. Reconstructing Grammar: Comparative Linguistics and Grammaticalization, ed by Spike Gildea, pp. 201-230. Amsterdam: John Benjamins.
MEIRA, Sérgio and SPIKE Gildea. In Progress. Tamanaku. In the series Materials. München: Lincom Europa.
MOSONYI, Jorge. 1982. Morfología Verbal del Idioma Cariña. Universidad Central de Venezuela, Caracas, thesis for the title of Magister en Lingüística.
PAYNE, Doris L. 1993. Nonconfigurationality and discontinuous expressions in Panare. Berkeley Linguistics Society 19. 121-138.
1994. OVSu versus VSuO in Panare (Cariban): Do syntax and discourse match?. Text 14.581-610.
 I use the terms A, S and O to label morphosyntactic categories of verbal arguments: S is the sole participant of intransitive verbs, and A and O are the two distinct arguments of transitive verbs. A contains the agents of prototypical transitive verbs, along with all other participants (regardless of semantic role) that pattern morphosyntactically with agents; O is the other participant. While they are grammatically coherent, the categories of A, S and O are clearly not semantically coherent in Cariban —I have encountered no evidence in any Cariban language for case-marking, verb agreement or any other argumental variation based on semantic category of verb or type of participant in the event/state coded by the verb (cf. Meira 2000b on the lack of semantic basis for the split-S morphology observed in some Cariban languages).
 For discussion of how the S can be both preverbal, as part of the Abs-V verb phrase, and postverbal, as part of the Aux-Nom bond, cf. Gildea 1998, ch. 10; 2000.76-81.
 This not to say that functional equivalents of such constructions are lacking, but rather that such functional equivalents do not offer us the syntactic evidence we are seeking to identify categories of grammatical relations.
 All publications in Cariban linguistics, including my own, refer to this as the third person reflexive possessive prefix; the term ‘logophoric’ is generally accepted in Linguistics for this function, and allows us to avoid confusion between this nominal / postpositional prefix and the verbal reflexive prefix (some allomorphs of which have similar form).
 The final velar nasal in the language names Kapóŋ and Pemóŋ is a predictable word-final allophone of either /m/ or /n/. Hence, these names are written as Kapón and Pemón in many publications. Following traditional spelling in Guyana, I use Pemóng and Kapóng here.