City University of New York. Department of Linguistics

CatWPL 412. 1995 329-343 The Antipassive in Jacaltec: A Last Resort Strategy Francisco Ordóñez City University of New York. Department of Linguistics. Graduate Center 33 West 42 Street. New York. NY 10036. USA [email protected] .EDU Received: November 1 lth 1994 Accepted: June 1st 1995 Abstract This paper gives reasons to recover the notion of cclast resort,, found in Chomsky (1991) by examining two uses of the antipassive suffix -ni in Jacaltec, an ergative Mayan language. This suffix is inserted as a last resort where UG makes the assignment of case to the object (transitive aspectless embedded clauses) and the extraction of the subject with ergative agreement (aspectual main clauses) otherwise impossible. Therefore, this notion cannot be restricted to instances of movement, as in Chomsky (1994). Rather, it must also include lexical insertion of affixal material in the derivation, as in the broader view found in Chomsky (1991). Key words: syntax, last resort strategies, antipassive, jacaltec. Resum. L'antipassiva en jacaltec: una estrat2gia d'últim recurs Aquest treball proposa de recuperar la noció d'ctúltim recurs,, que trobem a Chomsky (1991) tot inspirant-se en dos usos del sufix antipassiu -ni del jacaltec, una llengua ergativa de la família maia. Aquest sufix s'inserta com a últim recurs quan la GU fa impossible assignar cas a l'objecte (oracions subordinades transitives sense aspecte) i extreure el subjecte amb concordan~a ergativa (oracions matriu amb aspecte). Per tant, aquesta noció no es pot restringir a casos de moviment, com fa Chomsky (1994). Més aviat, ha &incloure la inserció lbxica d'afixos durant la derivació, com trobem en la visió més implia de Chomsky (1991). Paraules clau: sintaxi, estratbgies d'últim recurs, antipassivitat, jacaltec. Table of Contents 1. General characteristics 3. Conclusion 2. The distribution of References the antipassive suffix -ni 330 CatWPL 412, 1995 Francisco Ordóñez This paper gives a uniform account of the presence of the antipassive suffix -ni in Jacaltec, an ergative Mayan language.' This suffix appears in what seems to be two completely unrelated syntactic environments: 1. Transitive verbs in embedded clauses that lack aspect. 2. Transitive verbs in root clauses where the subject has been extracted.* In both cases, it is shown that -ni suffixation is a last-resort strategy in the language in order to save configurations that could not give any output otherwise. In 1) the suffixation is motivated by the lack of source of case for the object in clauses without aspect. The suffix -ni is the provider of such case. In 2) the suffixation provides a way of escaping the ban on extraction of subjects with ergative agreementlcase. The suffix -ni eliminates the ergative agreement on the subjects by ccintransitivizing,, them.3 Thus, the -ni suffix in both contexts acts as an intransitivizer for the subject and as a case assigner for the object. In order to make sense of this characterization, it is proposed that -ni is a prepositional affix, which takes the DO as its complement and assigns case to it. As a consequence, the configuration obtained is analogous to that of an intransitive verb which takes a prepositional complement. 1. General characteristics Jacaltec is a Mayan language with a strict VSO order. Sentence (1) exemplifies the canonical order in the lang~age:~ (1) xil naj ix. saw he her 'He saw her.' As with many other Mayan languages, Jacaltec has two major sets of affixes: absolutive/nominative and ergative. The absolutive/nominative set cross-refers subjects of intransitive verbs and objects of transitive verbs. Subjects of transitive verbs are cross-referred by the ergative set of affixes. Absolutive/nominative affixes are attached to an aspectual marker that precedes the verbal head.5 First and second person absolutive/nominative affixes are 1. I take the definition of ergative language given by Dixon (1994: 1): aThe term ergativity is used to describe a grammatical pattem in which the subject of an intransitive clause is treated in the same way as the object of a transitive clause, and differently from the transitive subject.), 2. Other related Mayan languages which also show the same suffix in 1) and 2) are Kanjobal and Chuj (Robertson, 1980) and Kekchi (Stewart, 1980). 3. I use the word intransitive to mean those verbs that do not assign case to their direct objects, i.e. verbs with one argument or verbs with prepositional complements. 4. All examples are extracted from Craig (1977). 5. ln many instances the aspectual head amalgamates with the verb to form one phonological unit. The Antipassive in Jacaltec: A Last Resort Strategv CatWPL 412, 1995 331 overt, but third person ones are either nul1 or there is no affix at Ergative affixes are overt for all persons and they always appear prefixed to the verb. (2) ergative set before consonant ergative set before vowel absolutive set 1Sg hin whin 2Sg ha haw hach 3Sg s Y 0 lPl cu Y hoii 2P1 hey hey hex 3Pl s Y 0 The typical shape of a verbal complex is represented in (3). Absolutive affixes always precede the ergative ones:7 (3) x-0-s-watx'e naj te' iiah. Asp-A3-E3-make clhe cltthe house 'He made the house.' First and second person affixes behave differently from the third person ones in both the ergative and the absolutive set. They are incompatible with the realization of a first or second person pronoun in the subject or object position in the sentence. Example (4) shows this incompatibility with a first person absolutive affix in an intransitive verb; example (5) shows this incompatibility with an ergative first person affix in a transitive verb: (4) ch-in-axni ("hayin). Asp-Al-bathe I 'I bathe.' (5) x-0-w-i1 ("hayin) ha-man. asp-A3-El-see I your-father 'I saw your father.' First and second person affixes incorporate, it seems, to the aspectual head or verb. However, overt third person affixes require the presence of the NP in a clausal position. This can be seen overtly in the ergative set for subjects of transitive verbs as in (6). No dropping of the NP subject is allowed. This affix can be plausibly thought of as an agreement element. (6) x-0-s-watx'e "(naj) e' iiah. asp-A3-E3-make clhe cllthe house 'He made the house.' 6. In the glosses for 3 person absolutive I use, for convenience, the ones by Craig (1977) where she assumes the existence of a 8 affix. Nevertheless, I want to remain uncommitted with respect to whether such empty affixes exist. 7. E3= ergative 3rd person; A3=absolutive 3rd person; Asp=aspect. 332 CatWPL 412, 1995 Francisco Ordóñez 2. The distribution of the antipassive suffix -ni 2.1. Transitive verbs of aspectless embedded clau,ses The -ni suffix appears in what Craig (1977) calls <<aspectless embedded clauses,, with transitive verbs. These clauses do not have overt complementizers and they lack the aspectual marker that obligatorily appears in all the other types of embedded finite clauses. Aspectless embedded clauses are also rather peculiar with respect to the distribution of the ergative affixes: subjects of intransitive verbs take the ergative set of affixes as in (7), (8), contrary to what we described in the previous section. The suffix -ni must always be present. (7) x-0-w-i [ha-caiialwi] asp-A3-El E2-dance 'I saw you dance.' (8) x-0-w-abe [ y-ok ix] Asp-A3-El-hear E3-cry she 'I heard her cry.' Subjects of transitive verbs also require the ergative agreement. Objects always take the absolutive set of afíixes. First and second person absolutive/nominative pronouns must appear to the left of the verb and to the left of the ergative marker in (9).* Third person NPs follow the verb as in (10). (9) x-0-w-ilwe [hach hin-col-ni] asp-A3-El -try A2 El-help-ni 'I tried to help you.' (10) wohtaj [ hin watx'e-n kap camixe] I know El make-ni the shirt 'I know how to make shirts.' That ergative affixes appear in subjects of intransitive verbs as well as transitive can be explained by examining possessor constructions with NPs. Possessor NPs are cross-referred by the same ergative affixes. First and second person possessors are exemplified in (1 1) and (12). The classifier determiner of the head noun appears to the left of the ergative affixes and is ~~tional:~ (1 1) (no') hin-txitam cl El-pig 'my pig' (12) haw-uxtaj E2-brother 'your brother' 8. This ordering of affixes follows the one of regular main clauses where absolutive first and second person affixes precede the ergative ones. 9. Classifiers can be deteminers and third person pronoun. The Antipassive in Jacaltec: A Last Resort Strategy CatWPL 412, 1995 333 Example (13) shows how 3rd person NP possessors appear to the right of the head noun, which bears the ergative agreement.1° (13) (no') s-txitam ix cl(anima1) E3-Pig cl(woman) 'her pig' Given these facts, it is tempting to make an analogy between the subject of aspectless embedded clauses and possessors of nominal constructions. That way, aspectless embedded clauses can be treated as a norninalization construction of the poss-ing type for English (Reuland 1983). The subject of the gerund poss-ing bears the same genitive case as the possessor in a nominal. From this point of view, it is not surprising that subjects of intransitives and transitive are equally treated as possessors in Jacaltec. There is further empirical support for this nominalization hypothesis for aspectless embedded clauses. It is possible to find certain demonstratives modifying the whole embedded clause. These demonstratives occur exclusively with nominals. In (14) the demonstrative this modifies the embedded verb I walk and in (15) it modifies the embeded verb You insult. (14) caw ya' [ hin belwi ti]. very painful El walk this 'My walking was very painful.' (15) tx'of [ ha bahwa ti'] bad E2 insult this 'This insulting of yours is bad.' The analysis proposed for these two constructions -nominals with possessors and aspectless embedded clausesis similar. For third person, there is Spec-head agreement between the head (N for nominals or V for aspectless embedded clauses) and the possessor NP. This agreement would be reflected by the ergative set of affixes. The head NIV would move past the possessor in both cases and would attach to the AgrP responsible for the ergative agreement. For first and second person possessors, there would be direct incorporation to the V or N head. 10. Other languages show the same ergative pattern for the subject of transitive verbs and possessors in possessor constructions. One example is given by Alana Johns (1992:68) for Inuktitut. 334 CatWPL 412, 1995 Francisco Ordóñez Since this construction is nominal-like, the question arises as to why should -ni only attach to transitive verbs. I take as a point of departure that the aspectual head is responsible for the absolutive/nominative case for the subjects of intransitive verbs and objects of transitive verbs is in main clauses." Therefore, any aspectless clause should have an extra way of providing case to objects of transitive and subjects of intransitive verbs. Since the construction is nominal-like, the subject intransitives act as possessors and therefore enter into ergative agreement as in (7) and (8). However, for transitive verbs, the subject takes the ergative agreement, leaving the object without that possibility. The mechanism invoked is the suffixation of a morpheme to the verb that has the capacity of assigning case to the object: the suffix -ni. This morpheme could be thought of as a counterpart of of insertion for English nominals or of the morpheme of inflected infinitives in Portuguese, which allows the exceptional assignment of nominative case. Thus, -ni is needed for the assignment of absolutive/nominative case to the object. 2.2. Transitive verbs in root clauses with extraction of third person subjects Questions in Jacaltec are formed by the extraction of wh-words to the left of the initia1 verb. Example (17) shows the extraction of a subject of an intransitive clause, while (18) shows the extraction of the object of a transitive clause. Note that the extraction of the subject of a transitive clause in (19) requires the suffix -ni, and contrary to aspectless embedded clauses, there is no ergative agreement prefixed to the verb. (17) mac, x-u1 ej ewi? who Asp-come yesterday 'Who came yesterday?' (18) macj x-y-i1 naj ej? who asp-E3-see he 'Who did he see?' (19) macj x'ilni ej ix? who Asp-see-ni she 'Who saw her?' Clefting in Jacaltec also involves movement of the focused element to the left of the verb and the use of -ni. A special particle which indicates clefting shows on the left of the sentence. Example (20) exemplifies the extraction of the subject of a transitive clause which shows the lack of ergative agreement and the obligatory suffixation of -ni: (20) ha' najj x'ilni ej ix. cleft he Asp-see-ni she 'It was him that saw her.' 11. This position follows Murasugi (1992) and Bittner and Hale (forthcoming) in that tense head is the ultimate source for nominativelabsolutive in ergative languages. This point is elaborated in section 2.2.2. The Antipassive in Jacaltec: A Last Resort Strategy CatWPL 412, 1995 335 Similar uses of -ni are found in relatives. These constructions lack an overt complementizer. Examples (21) and (22) show the relativization of the subject of an intransitive verb and the object of a transitive verb re~~ectivel~.'~ Example (23) exemplifies the extraction of the subject of a transitive verb which again shows the suffix -ni and the lack of ergative agreement on the embedded verb. (21) xwil [naj winajj 'xapni e, ewi] Asp-El-see C1 man arrived yesterday 'I saw the man that arrived yesterday.' (22) Wohtaj [ixj xil naj ej] Asp-El-know she see he 'I know the woman that she saw.' (23) Wohtaj [najj x'ilni e, ix] Asp-El-know him see-ni she 'I know the man that saw her.' 2.2.1. -ni: the prepositional uffix It is evident that the insertion of the suffix -ni and the lack of the ergative agreement must be related. Extraction of subjects which enter into ergative agreement in main clauses is impossible as shown in (24) and (25):13 (24) *macj x-y-i1 ej naj? who Asp-E3-see he 'Who saw hirn?' (25) *macj x-achi s-mak ej? who Asp-A2 E3-hit 'Who hit you?' Jacaltec is not the only ergative language which shows this ban on the extraction of elements which show ergative agreement or ergative case marking. Dixon (1994) points out that relativization of the NP with the ergative case marking in Dyrbal is not allowed. Johns (1992) also shows that the relativization of an argument with the ergative case marking in Inuktitut yields an ungrammatical result. Example (26), from Inuktitut, can only be interpreted as the relativization of the object, not the subject.14 (26) [Anguti-up nanuq kapi-ja-a] ani-juq. man-erg bear stab-pass.part-3ss go out '*The man who stabbed the bear left.' (From Johns 1992) 12. The example with the intransitive verb involves an unaccusative verb. There is no different syntactic behaviour of unaccusative and unergative in the language. Subjects of unergatives and of unaccusatives show the same absolutive/nominative pattern. 13. In 2.2.2 and 2.2.3 there are various constructions that are exceptional i11 allowing the extraction of the argument with ergative agreement. 14. The theoretical explanation for the impossibility of the extraction of the subject with the ergative agreement is dealt with in the next section. 336 CatWPL 412, 1995 Francisco Ordóñez The existence of this constraint leaves the question as to why the -ni suffix permits the extraction of subjects in what looks like transitive verbs as in (19), (20) and (23). From the data from aspectless embedded clauses, it is apparent that -ni is the element responsible for the assignment of the case to the object. The lack of aspectual head, motivated this special mechanism of insertion of -ni. However, this is not the case with the extraction cases we are discussing: the aspectual head is available. I would like to propose an answer to this puzzle that relies on the idea that the extracted subjects in (19), (20) and (23) are the ones taking the absolutivelnorninative case made available by the aspectual head. This interpretation is supported by the lack of ergative agreement in the verbal complex. From this point of view, the object of these <<transitive>> verbs is still in need of a special mechanism for the assignment of case, as with aspectless embedded clauses. However, in an ergative language, a subject can only take the absolutive/nominative case when the verb is intransitive. Therefore, it must be that contrary to all appearances all the cases of verbs containing the suffix -ni are intransitives: in other words, -ni acts as an intransitivizer, qualifying the construction as an antipassive. Note that the only difference between this antipassive in Jacaltec and other antipassives discussed by Baker (1988) is that in Jacaltec the object is not demoted. This analysis is still perfectly compatible with our previous analysis of aspectless embedded clauses with -ni: in those constructions, subjects of verbs talung -ni act like other subjects of clearly intransitive verbs -such as in (7) and (8). They all take ergative agreement since both act like possessors in this nominal type of construction. There is also morphological evidence for the intransitive nature of verbs containing -ni. There is a special future morphology in Jacaltec. The suffix -oj only appears with intransitive verbs, whereas transitive verbs take future suffix -o1. Constructions with -ni, however, take -oj not -o', as can be seen by comparing (27) with (28) where the subject is extracted, and ni is suffixed. (27) x-s-lok-o' naj no' cheh. Asp-E3-buy-fut clthe cllthe horse 'He will buy the horse.' (28) w-hojtaj naji x-lok-n-oj ei no' cheh. El-know cVhim Asp-buy-ni-fut cllthe horse 'I know the man who will buy the horse.' Example (29) shows -ni in an embedded aspectless clause, which again takes the intransitive suffix -oj: (29) mach ch-u s-to najj [ hachj y-il-oj najj] l5 Lao Asp-can E3-go he A2(you) E3-see-ni-fut he 'He cannot go to see you.' 15. The second naj is deleted later on because it is coreferent with the subject naj in the main clause. This rule is called Noun Classifier Deletion Rule by C. Craig. The Antipassive in Jacaltec: A Last Resort Strategy CatWPL 412,1995 337 In order to reconcile the two roles of-ni as an intransitivizer and as case assigner to the object, it is necessary to suppose that -ni is a prepositional kind of affix that gets inserted in the derivation and takes the object as a complement. As a prepositional element it assigns case to the object (30): (30) V S -ni + object isertion The configuration in (30) is, in essence, similar to a verb subcategorizing for a prepositional phrase as in (3 1). In both, the subject, as in other intransitive configurations, takes the nominative/absolutive case. Therefore, it can also capture the intransitivizing properties of the suffix -ni.16 (3 1) x-0-ahtoj naj y-abaii no' cheh.17 go up he Agr-P the horse 'He went up with the horse.' Later on in the derivation, -ni needs to be incorporated into the verb given its affix nature. (32) V+-nij S tj + object. In conclusion, the insertion of -ni in these cases should be taken as a last resort strategy to overcome the ban on the extraction of an element with ergative case or agreement in main clauses. The extraction is possible with the suffix -ni because it intransitivizes the verb, making the subject take the absolutive/nominative case. 16. We conclude that the assignment of ergative case for subjects is dependent of the existence of an inflectional source for the case of the object. This notion looks reminiscent of Bittner and Hale's (forthcoming) idea that the assignment of ergative case depends on the existence of a case competitor for it. 17. As we can see in the glosses of (31) non affixed prepositions agree with their objects in the language. This poses a problem to our proposal. As suggested by an anonymous reviewer it is possible to think that the incorporation of the prepositional affix -ni takes place in the lexical relational structures of Hale and Keyser (1994). The prepositional agreement in (31) takes place at the level of syntax, which could explain why incorporated -ni does not show agreement. This raises the more general question whether incorporated elements can show syntactic agreement.