The syntax of hybrid verb/affix lexemes and clause fusion in Hiaki (Yaqui)

e syntax of hybrid verb/affix lexemes and clause fusion in Hiaki (Yaqui) Mercedes Tubino Blanco, Heidi Harley, and Jason Haugen University of Arizona* Abstract We argue that verb/affix hybrids in Hiaki (Yaqui) are subject to the same conditions on clause fusion (Rude 1996) as ‘pure’ affixal verbs in spite of their different distributional behavior. We show that all verbs involved in V-V affixation under clause fusion undergo VP embedding, rather than TP embedding, whether they also have a morphologically free use or are obligatorily bound. is results in one case domain, but two binding domains, whi shows clause binding sensitivity to VP and nominative case assignment at TP. e ability of these hybrids to occur in affixal ‘clause fusion’ structures, as well as to appear as free main verbs embedding an independent clause, is unusual, and enables us to investigate the conditions on V-V incorporation. As Guerrero Valenzuela (2004) points out, su verbs may show incipient grammaticalization. Keywords: clause fusion, affixal embedding, grammaticalization 1 Introduction Hiaki¹ displays an interesting (but not uncommon) type of complex predicate formation, in whi a derivational affix is aaed to an embedded verb. e affix introduces its own arguments to the construction, and takes the content of the embedded verb phrase as its semantic complement, resulting in a ‘clause-fusion’ construction (Rude 1996). In this construction, the arguments introduced by the affix and those introduced by the embedded verb are case-marked in a monoclausal paern, despite their selectional relationships with distinct subcomponents of the complex verb. Typical examples are shown in (1a)–(1c): (1) a. Anselmo Anselmo- [uusi-ta [ild- vuiti]-sae-k run-]- ‘Anselmo told the ild to run.’² b. Carlos Carlos- [uusi-ta [ild- vuiti-]’ii’aa-k run--] ‘Carlos wanted the ild to run.’ *Many thanks to our consultants Maria Amarillas, Santos Leyva, Guillermo Gutierrez, Dora Gary, and Modesto Bule for all their help in this resear. is resear is partially supported by the NSF Award No. BCS-0131761. Unaributed data in this paper were collected during the course of this project. ¹Hiaki (Yaqui) is a Uto-Aztecan language spoken in Sonora, Mexico and Arizona, U.S. 79 c. Carlos Carlos- [uusi-ta [ild- vuiti-]tua-k run-]- ‘Carlos made the ild run.’ In the examples above, the suffixes -sae ‘to tell’ (1a), -‘ii’aa ‘to want’ (1b), and the causative -tua ‘to make/have’ (1c) combine with the complement verb vuite ‘to run’. is produces what Rude terms ‘clause fusion’, referring to the notion that two semantically distinct clauses ‘fuse’ into a single syntactic clause. is type of construction is quite productive in Hiaki, whi contains many su derivational verbal affixes (e.g. -ii’aa, ‘desiderative,’ -sae, ‘directive’, -tua, ‘causative,’ -pea ‘to feel like’). ese affixes are bound, and may never occur as main verbs on their own, either embedding an independent complement clause or taking an NP object, as the ungrammaticality of the examples in (2) indicates: (2) a. *Jason Jason- [uusi-ta [ild- koowi-m pig- am-sua-ne-veti’ivo] 3.-care.for--] ii’aa want ‘Jason wants the ild to take care of pigs.’ b. *Jason Jason tahkai-m tortilla- ii’aa want ‘Jason wants a tortilla.’ In contrast, the Hiaki verb mahta ‘to tea’ exhibits hybrid behavior, as it appears both as an affix and as an independent main verb, with the same meaning in both structures in (3a) and (3b). In (3a), we see the complement verb sua- ‘care for’ is the stem hosting the matrix verbal affix -mahta. In (3b), the same form occurs as a free verb, embedding a separate complement clause introduced by the complementizer/preposition veti’ivo ‘for’: (3) a. Jason Jason- [uusi-ta [ild- koowi-m pig- sua-]mahta-k care.for-]tea- ‘Jason taught the ild to take care of pigs.’ b. Jason Jason- [uusi-ta [ild- koowi-m pig- sua-ne-veti’ivo] care.for--] a=mahta-k 3=tea- ‘Jason taught the ild to take care of pigs.’ Affixal and free mahta in (3a) and (3b) embed different types of complement clauses. Because the two kinds of complement are clearly distinct syntactic objects, we conclude that this is not a simple case of cliticization under adjacency, as with clitic/free morpheme alternations su as I’m vs. I am or isn’t vs. is not. In (3a), the affixal case, no tense or aspect morphemes can intervene between the suffixed -mahta and the complement verb sua-. is is shown in (4), in contrast to the the clausal complement case in (3b), whi allows an embedded future tense marker -ne: (4) *Jason Jason- [uusi-ta [ild- koowi-m pig- sua-ne]-mahta-k care.for-]-tea- ‘Jason taught the ild to keep pigs.’ ²Gloss: : nominative; : accusative; : directive; : perfective; : preterite; : future; : negation; : desiderative; : causative; : plural; : imperfective; : inclination; : irrealis; : complementizer; 3: 3rd person singular clitic; 3: 3rd person plural clitic; 3: 3rd person; : reflexive; : transitivizer; : determiner; : reduplication; : interrogative; : instrumental Rice Working Papers in Linguistics 80 vol. 1, February 2009 Despite the structural contrast displayed in (3a) and (3b), both constructions share a crucial aracteristic with respect to their case and binding domains, whi presumably accounts for the general behavior that non-finite complement verbs exhibit in Hiaki, an Exceptional Case Marking (ECM) language like English. As in English, non-finite clause embedding is aracterized by a single case domain, in whi no nominative argument is licensed in the embedded clause. In contrast, two binding domains are licensed in su structures, whi confirms the presence of two different clauses despite the complement clause’s la of a second nominative argument. e English equivalent of (3a) and (3b), in (5), shows this parallel: (5) a. Jason taught the ild to take care of pigs. b. Jasonitaught the ildkto take care of himself∗i/k. c. Jasoni taught him∗i/kto take care of pigs. d. Jasoni taught him∗i/kto take care of himselfk. e. Jasoni taught himselfito take care of pigs. As (5) shows, the embedded subject him/the ild may bind the embedded object himself ((5b) and (5d)), and the embedded subject may itself be bound by the matrix subject Jason (5e). Crucially, the matrix subject may not bind the embedded object (5b). Unlike verbs like ‘want’, the possibility of PRO as the embedded subject is not available in English with verbs su as ‘tea’ as the ungrammaticality of (6b) shows: (6) a. Jason PROito take care of himselfi. b. *Jasonitaught PROito take care of himselfi. In (6a), the matrix subject Jason binds all the way into the embedded object himself by virtue of binding the embedded subject PRO. is is not possible in (6b) where the embedded subject is obligatorily distinct from the matrix subject. A similar paern of variation is also the case in Hiaki, in that some affixal predicates require control of the embedded subject and some require a distinct embedded subject: (7) a. Peo Pete. [Maria-ta [Maria- bwiik]-‘ii’aa sing]- ‘Pete wants Maria to sing.’ (Escalante 1990:13(26a)) b. inepo 1s. bwiik-pea sing- ‘I feel like singing.’ (Escalante 1990:13(26b)) c. *inepo 1s. [Maria-ta [Maria- bwiik]-pea sing]- ‘I’d like Maria to sing.’ (Escalante 1990:13(26c)) e derivational affix -‘ii’aa ‘to want’ in (7a) mandatorily requires that the subject of the embedded verb be distinct from the matrix subject; in contrast, other affixes su as -pea ‘to feel like’ (7b), exhibit mandatory control of the embedded subject by the matrix subject.³ Given the distribution of data described above, this paper explores the following issues: ³In other words, Hiaki -ii’aa, ‘want’, is like English ‘tea’ or English ECM ‘want’ in requiring a non-control construction with a distinct embedded subject NP, while Hiaki -pea is like English ‘remember’ or English control ‘want’ in requiring a controlled embedded subject. Rice Working Papers in Linguistics 81 vol. 1, February 2009 (8) a. e syntax of Hiaki affixal predicates (i.e., -sae ‘tell’, -ii’aa ‘want’, and others) as instances of complex predication in this language; b. e ‘hybrid’ behavior of verbs su as mahta ‘tea’ in their roles as both free verbs and affixes c. e syntactic behavior of non-finite complement constructions in Hiaki as obligatorily exhibiting ECM; and d. the structural consequences of these affixal ECM constructions in Hiaki in terms of binding and case. We claim a syntactic consequence in terms of both case and binding for the ‘hybrid’ verb phenomenon. Case-wise, all clause fusion configurations (i.e., affixal predicates and ‘hybrids’) show a single case domain, licensing just one nominative DP. Binding-wise, all clause fusion configurations also exhibit identical behavior: they all contain two binding domains, overlapping at the embedded subject as a result of ECM. e paper is organized as follows: in section 2, we offer an analysis where clause fusion of both ‘hybrid’ and purely affixal verbs undergoes VP (not TP) embedding. Section 2.1 illustrates the consequences of this analysis in terms of case and binding domains, whi suggest the ‘hybrid’ origins of clause fusion structures. Section 3 is an application of the analysis to other verbs of the ‘hybrid’ type, followed by a conclusion in section 4. 2 e analysis: VP embedding We claim that ‘hybrid’ verbs su as mahta ‘tea’ exhibit clause fusion exactly like pure affixes su as -sae ‘tell’. All affixal predicates in Hiaki do clause fusion by embedding a VP, instead of embedding a TP, regardless of the availability of a non-affixal construction for the affix involved. 2.1 Case Both affixes and ‘hybrid’ verbs in Hiaki subcategorize for a VP, rather than TP, complement clause. is results in one case domain, where just the matrix subject is in the nominative. Any other DP appearing in the structure is in the accusative. For ease of exposition, the examples with -sae ‘tell’ (1a) and mahta ‘tea’ (3a) are repeated in (9) and (11) respectively. (9) Anselmo Anselmo- [uusi-ta [ild- vuiti]-sae-k run-]- ‘Anselmo told the ild to run.’ In (10), the obligatorily affixal -sae combines with the complement verb vuite ‘run’ forming a complex predicate in whi both verbs share the same TP, resulting in a single case domain.⁴ e embedded subject uusi-ta, ‘ild-’, receives its accusative case via ECM from the matrix V. (See below for more details concerning the ECM structure.) ⁴We assume the VP-internal subject hypothesis (Koopman and Sportie 1986) according to whi the subject is basegenerated in the specifier of its theta-assigning VP, indicated by the shaded copy of the subject in this position in the structures above. Although we do not explicitly indicate it in the structures here, the analysis is compatible with the split-VP hypothesis as well. Rice Working Papers in Linguistics 82 vol. 1, February 2009 (10) TP l l l l l l l l l l R R R R R R R R R R Anselmo Tˈ l l l l l l l l l l R R R R R R R R R R VP l l l l l l l l l l R R R R R R R R R RT ti 66 Vˈ l l l l l l l l l l R R R R R R R R R R VP l l l l l l l l l l R R R R R R R R R RV uusi-ta ‘ild-’ V sae ‘tell’ vuiti ‘run’ II By the same token, in (11), the ‘hybrid’ verb mahta ‘tea’ fuses with the complement verb sua ‘care for’ to share the same TP, and consequently, case domain, in whi only the matrix subject Jason is in the nominative: (11) Jason Jason- [uusi-ta [ild- koowi-m pig- sua-]mahta-k care.for-]tea- ‘Jason taught the ild to take care of pigs.’ e structure involving a ‘hybrid’ matrix verb su as mahta in (12) shows the parallel between the analyses in (10) and (12): (12) TP l l l l l l l l l l R R R R R R R R R R Jason Tˈ l l l l l l l l l l R R R R R R R R R R VP l l l l l l l l l l R R R R R R R R R RT ti 66 Vˈ l l l l l l l l l l R R R R R R R R R R VP l l l l l l l l l l R R R R R R R R R RV uusi-ta ‘ild-’ Vˈ l l l l l l l l l l R R R R R R R R R Rmahta ‘tea’ koowim ‘pigs-’ sua ‘care for’ hh e structures in (10) and (12) show two VPs ea, the matrix VPs, headed by -sae (10) and mahta (12), and the complement VPs, headed by vuite ‘run’ (10) and sua ‘care for’ (12). Both VPs in (10) and (12) share a single TP. is results in one case domain, in whi just one nominative DP (i.e., Anselmo,Jason) is licensed as the subject of the matrix verb, -sae (10) or mahta (12). e complement Rice Working Papers in Linguistics 83 vol. 1, February 2009 subjects (i.e., uusita, ‘ild-’) appear in the accusative via ECM from the matrix predicate, since a second finite TP, needed to license a second nominative case, is laing. is behavior is typical of simplex, rather than complex, clausal configurations. Additionally, the verb mahta may also occur as a free verb with clausal embedding, by taking a complement clause introduced by the preposition/complementizer veti’ivo⁵ in (3b), repeated as (13): (13) Jason Jason- [uusi-ta [ild- koowi-m pig- sua-ne-veti’ivo] care.for--] a=mahta-k 3=tea- ‘Jason taught the ild to take care of pigs.’ In contrast with the clause fusion structures analyzed in (10) and (12), the configuration in (14) must include a second TP, whose head is the irrealis particle -ne. (14) TP l l l l l l l l l l R R R R R R R R R R Jason Tˈ l l l l l l l l l l R R R R R R R R R R VP l l l l l l l l l l R R R R R R R R R RT ti 66 V l l l l l l l l l l R R R R R R R R R R CP l l l l l l l l l l R R R R R R R R R Rmahta ‘tea’ uusi-ta ‘ild-’ Cˈ l l l l l l l l l l R R R R R R R R R R TP l l l l l l l l l l R R R R R R R R R RC veti’ivo (uusi-ta) 66 Tˈ l l l l l l l l l l R R R R R R R R R R VP l l l l l l l l l l R R R R R R R R R RT ne  (uusi-ta) Vˈ l l l l l l l l l l R R R R R R R R R R koowim ‘pig-’ sua ‘care for’ gg e presence of a second TP, however, does not guarantee the licensing of a second nominative DP, evidenced by the accusative uusi-ta ‘ild-’ as the complement subject. is is a consequence of the non-finite status of the complement clause, as only finite tense heads have the ability to license nominative case in languages su as Hiaki (or English). e accusative case on the embedded subject comes from the preposition/complementizer veti’ivo itself, whi assigns accusative to the ⁵e postposition veti’ivo ‘for’ (as well as the postposition -po, ‘in’) seem to have complementizer-like uses, as here, whi are apparently syntactically and semantically distinct from their postpositional uses. We will investigate this in future resear. Rice Working Papers in Linguistics 84 vol. 1, February 2009 complement subject in its specifier, in the same way as the English preposition/complementizer for may assign accusative case to the subject of the complement clause introduced by it (15): (15) Jason hopes [for the ild/him to take care of the pigs] In (15), the English preposition/complementizer for assigns accusative case to the complement subject the ild/him, since the non-finite verb to take care of las the ability to do so. In other words, for is required as a case assigner to the complement subject in order to prevent a violation of the Case Filter (Chomsky 1981:49). In the same way, the Hiaki preposition/complementizer veti’ivo ((13) and (14)) assigns accusative case to the complement subject uusi-ta ‘ild-’. (We assume a Spec-Head configuration for this assignment, but this is not necessary.) is DP would remain caseless otherwise, given the la of an appropriate case assigner in the complement TP.⁶ Escalante (1990) distinguishes between complex and compound verbs in Hiaki. Affixes su as -sae are classified by this author as complex, whereas mahta falls into the compound class. Among the properties shared by the verbs falling into either class is the obligatoriness of independent overt subjects of the matrix and complement clause,⁷ regardless of the embedding configuration. e requirement of the ‘inner’ subject to be in the accusative case, as exemplified in the analysis above, is another aracteristic shared by both verb classes in either of their configurations. is behavior is contrasted with conjoined clauses, where both verbs are tensed and ea verb takes a nominative subject as in (16): (16) empo 2. ye’eka, sing- aapo 3. into and bwiika-k dance-⁸ ‘You sang and he danced.’ (Escalante 1990:12(6)) In (16), two separate matrix clauses are conjoined while keeping their separate case domains (i.e., both finite verbs take a nominative subject). In terms of case domain, however, it is not always true in Escalante’s examples that apparently tensed verbs take nominative arguments as their subjects, as the existence of sentences su as (17) shows: (17) aapo 3. [apo’ik [3. siika’]-tia-n leave.]-say- ‘Heisaid hejwas leaving.’ (Escalante 1990:12(12)) In this sentence, the perfective embedded stem verb siika ‘le’ takes an accusative subject (e.g. apo’ik ‘him’). e phonology of the stem and the syntactic behavior of the complex predicate indicate affixation.⁹ Syntactically, it is impossible to separate the compound form with adverbial or other intervening material, again clearly indicating affixation. e accusative case exhibited by the complement ⁶We assume that the head of the complement TP, the irrealis particle -ne, is tenseless and hence las the ability to assign case to its subject. ⁷As noted above, other verbs su as -pea require that the matrix subject and (null) embedded subject are coreferential. According to Escalante (1990:13), a complex sentence is not produced in this case. We have not yet observed any cases of optionality between subject-control and ECM structures in Hiaki. ⁸Relation of glosses used in Escalante (1990) and their equivalents in the present paper:  = ;  = ;  = / ;  = ;  = . ⁹In Hiaki, long vowels in stems are shortened when a derivational affix or compounding element is added, and occasionally a gloal stop may be inserted at the end of the first member of the verbal complex to introduce a hiatus. Rice Working Papers in Linguistics 85 vol. 1, February 2009 subject is evidence of a single case domain regardless of whether the verb heading the complement clause is tensed (e.g. the perfective siika-).¹⁰ Despite the apparently doubly tensed complex predicate just shown in (17), combinations of ‘hybrid’ verbs in their affixal form directly composing with embedded heads do not normally allow any tense or aspect morphemes between the two members of the compound, as (4) [repeated here as (18)] shows: (18) Jason Jason- [uusi-ta [ild- koowi-m pig- sua-(*ne-)]mahta-k care.for-(*-)]tea- ‘Jason taught the ild to keep pigs.’ e fact that it is impossible for the irrealis morpheme -ne to appear within the verbal compound is evidence of VP embedding instead of TP embedding. is may suggest that in cases su as (17), we are not likely to be dealing with tense/aspect morphemes intervening between the two members of the verbal compound. Rather, we are dealing with a complex predicate in whi the verbal stem is in the participial form, and hence tenseless despite its homophony with the suppletive past form. is would entail that for all the suppletive past tense forms in whi the verbal participle is indistinguishable from the past tense, we would see the suppletive form here. For any suppletive past tense forms in whi the participle and past tense are distinct (if any su exist), the participle should appear in these constructions, rather than the past tense form.¹¹ Furthermore, in most cases of V-V affixation where reduplication is applied to indicate habitual tense, only the ‘outer’, affixal verb reduplicates as shown below, whether it is a mandatorily affixal one like -sae (19a) or the ‘hybrid’ mahta (19b):¹² (19) a. inepo 1. [a=nok-]sas-sae [3.=talk]-- ‘I tell him to speak up.’ (Escalante 1990:14(41)) b. Maria Maria [yee [people hiak-nok]-mah-mahta Hiaki-speak]--tea ‘Maria teaes people to speak Hiaki.’ Finally, further evidence that the constructions at hand do not involve separate TPs comes from negation, since only the ‘outer’ verb can be negated, as in (20): ¹⁰Escalante admits that this construction in whi a ‘hybrid’ matrix verb combines with a perfective complement verb is limited to combinations involving tia ‘to say’. All su examples in this paper exhibit the ‘hybrid’ lexeme -tia affixed to a complement verb whi suppletes, rather than inflects, for tense. is fact may be an indication that, in this case, the head of the complement TP may be participial instead of finite, whi would in turn explain the absence of a second nominative DP as the complement subject, as we speculate below. e example in (i), however, poses a problem to this hypothesis, as the embedded verb appears inflected for the future tense by means of the future inflectional affix -ne: (i) aapo 3. [au [3. siim-ne’]-tia-n leave-]-say- ‘Heisaid heiwill leave.’ (Escalante 1990:13(30)) In this case, we suggest that the verb -tia may subcategorize for TP embedding, as opposed to the rest of verbs studied here, whi subcategorize for VP embedding. is issue is still under investigation. ¹¹We will investigate this in future work. See footnote 10 for problems related to this point. ¹²In fact, there are a few embedding purely affixal verbs (e.g. -tua,-pea, among others) that may not themselves reduplicate, the only reduplication possible being on the embedded verb stem. is class may involve ‘light’ verbs rather than full verb roots, whi might be further indication of the grammaticalization process already suggested for these constructions. Rice Working Papers in Linguistics 86 vol. 1, February 2009 (20) inepo 1. kaa  [eni [2. siim]-sae-n leave-]- ‘I didn’t tell you to leave.’ (Escalante 1990:13(29a)) In (20), the negation kaa can only take scope over the verbal complex siim sae ‘tell to leave’ as a whole, negating only the tensed matrix predicate -sae ‘tell’. It cannot mean ’I told you not to leave’. In order to negate the ‘inner’ verb siim- (bound form of siime ‘leave’), Escalante argues that the clause containing the compound must be made explicitly subordinate, introduced by the complementizing lexeme -kai, the ‘outer’ verb becoming semantically redundant (21): (21) nee 1. eni 1. tehhoa-kan, tell- [kaa [ eni 2. siim-sae-kai] leave--] ‘I told you not to go.’ (Escalante 1990:13(29b)) In (21), the semantic content of -sae is contributed to the structure by the matrix full verb tehwa ‘to tell’. As (20) shows, the absence of this additional full verb in the structure would result in the impossibility of negating the embedded member of the complex predicate, whi in turn is further evidence of the single TP-multiple VP analysis of clause fusion in Hiaki put forward in this paper. us far, this section has argued in favor of the simplex status of clause union structures as they result in just one, rather than two, tense, case and negation domains. is is because su structures involve VP rather than TP embedding, and have just one [Spec,TP] position licensing just one nominative DP (the matrix subject) and one tense marking. e fact that just one sentential negation is possible in these structures has been offered as further evidence. Section 2.2 explores the consequences clause fusion has in terms of the correct aracterization of binding domains. e conclusions we draw from the binding facts further suggest the grammaticalization process undergone by the verbs participating in this structure. 2.2 Binding Unlike conventional simplex structures, configurations involving clause fusion display two, rather than one, binding domains. is fact, in combination with the simplex behavior in terms of case, reflects both the peculiar ‘hybrid’ syntactic status of the structure and its ‘transitional’ typological status. ese structures are aracterized by the fact that the nominative subject of the matrix verbal affix cannot bind a reflexive in the object position of the embedded verb domain. is behavior is typical of complex rather than simplex structures. is binding paern is illustrated in the examples below. e complement subject Art-ta ‘Art- ’ binds the third person reflexive anaphor au from the embedded subject position (22a), in accordance with Principle A of the Binding eory (Chomsky 1981:6), whi states that anaphors must be bound in their binding domain. e matrix subject Heidi may not bind this reflexive. In contrast, in (22b), the matrix subject Heidi binds the embedded object pronominal clitic a= (22b). e embedded subject Art-ta may not bind this clitic as pronouns must be free within their binding domain, according to Principle B of this theory. (22) a. Heidi Heidi- [Art-ta [Art- au 3. sua]-mahta care.for]-tea ‘Heidi teaes Art to take care of himself/*her.’ Rice Working Papers in Linguistics 87 vol. 1, February 2009