Agreement in Tuyuca

The main premise of this thesis is that subject agreement morphology in Tuyuca can be isolated from the rest of the morphology. Subject agreement appears on evidentials, nominalizers, animate classifiers, gerunds, and verb stems requiring an auxiliary. This agreement is instantiated by a pervasive final vowel pattern that codes various values of gender, number, and person features. These final vowels also code the same information on nouns and pronouns. Before arguing for my analysis I provide some preliminary material on Tuyuca. Chapter 1 is a brief discussion of the sociolinguistic context of the language. Chapter 2 discusses issues relevant to Tuyuca data and surveys some of the literature related to Tuyuca; it also discusses some methodological concerns arising from the data and important to the thesis in general. Chapter 3 is a brief sketch of Tuyuca grammar important to agreement. Analysis is done in Chapters 4 and 5. In Chapter 4 I argue, in a descriptive-typological framework, that by isolating agreement a general deverbalizing function can be seen coded in the morpheme /-g-/. This morpheme has predictable interpretations in restricted morphosyntactic environments. It can be interpreted as a progressive or perfective aspect, an animate classifier, a gerund, and a nominalizer. In Chapter 5 I relate the general premise of isolating agreement in Tuyuca to theoretical issues belonging to the Minimalist Program. I show that isolating agreement morphemes from evidentials is, assuming the analysis in Chapter 4, straightforward. This has a practical advantage of making it easier to observe variation between present tense and past tense morphology of the evidentials. I take this as straightforward evidence that tense is fused with evidential. I also give evidence that supports the pro-drop status of Tuyuca, conjecturing that subject agreement is packaged with nominative case. I also argue informally that verbal inflection of tense-evidentials and subject agreement are "extensions" of the verb phrase and relate the predication of VP to some speech time and discourse situation of the verb event, relative to some specific world. This results in a model of functional hierarchy that places Evidential under Tense Phrase. I conjecture that this Evidential position is a predicational one, in contrast to the more accepted notions of Mood[evidential] or Modal[epistemic], which are known to be above Tense Phrase. I provide two detailed models, one with the conventional hierarchy and one with my hierarchy, arguing for the latter--based on general principles of syntactic economy and locality. I also provide a technical analysis of syntactic locality for the morphosyntactic fusion of tenseevidentials in a Distributed Morphology framework.

AGREEMENT IN TUYUCA by Joshua Wayne Bowles A thesis submitted to the faculty of The University of Utah in partial fulfillment of the requirements for the degree of Master of Arts Department of Linguistics The University of Utah December 2008 Copyright © Joshua Wayne Bowles 2008 All Rights Reserved T H E U N I V E R S I T Y OF U T A H G R A D U A T E S C H O OL SUPERVISORY COMMITTEE APPROVAL of a thesis submitted by Joshua Wayne Bowles This thesis has been read by each member of the following supervisory committee and by majority vote has been found to be satisfactory. THE UNIVERSITY UTAH GRADUATE SCHOOL satisfactory. T H E U N I V E R S I T Y U T A H G R A D U A T E S C H O OL F I N A L APPROVAL I have read the thesis of Joshua Wayne Bowles m l t s f m a i fo rm format, Chair/Dean THE UNIVERSITY OF UTAH GRADUATE SCHOOL FINAL READING APPROVAL To the Graduate Council of the University of Utah: in its final fonn and have found that (1) its fonnat, citations, and bibliographic style are consistent and acceptable; (2) its illustrative materials including figures, tables, and charts are in place; and (3) the final manuscript is satisfactory to the supervisory committee and is ready for submission to The Graduate School. Date Campbell air: Supervisory Committee Approved for the Major Department ChairlDean Approved for the Graduate Council David S. Chapman Dean of The Graduate School ABSTRACT The main premise of this thesis is that subject agreement morphology in Tuyuca can be isolated from the rest of the morphology. Subject agreement appears on evidentials, nominalizers, animate classifiers, gerunds, and verb stems requiring an auxiliary. This agreement is instantiated by a pervasive final vowel pattern that codes various values of gender, number, and person features. These final vowels also code the same information on nouns and pronouns. Before arguing for my analysis I provide some preliminary material on Tuyuca. Chapter 1 is a brief discussion of the sociolinguistic context of the language. Chapter 2 discusses issues relevant to Tuyuca data and surveys some of the literature related to Tuyuca; it also discusses some methodological concerns arising from the data and important to the thesis in general. Chapter 3 is a brief sketch of Tuyuca grammar important to agreement. Analysis is done in Chapters 4 and 5. In Chapter 4 I argue, in a descriptive-typological framework, that by isolating agreement a general deverbalizing function can be seen coded in the morpheme l-g-l. This morpheme has predictable interpretations in restricted morphosyntactic environments. It can be interpreted as a progressive or perfective aspect, an animate classifier, a gerund, and a nominalizer. In Chapter 5 I relate the general premise of isolating agreement in Tuyuca to theoretical issues belonging to the Minimalist Program. I show that isolating agreement morphemes from evidentials is, assuming the analysis in Chapter 4, straightforward. This r I r I-I. r has a practical advantage of making it easier to observe variation between present tense and past tense morphology of the evidentials. I take this as straightforward evidence that tense is fused with evidential. I also give evidence that supports the pro-drop status of Tuyuca, conjecturing that subject agreement is packaged with nominative case. I also argue informally that verbal inflection of tense-evidentials and subject agreement are "extensions" of the verb phrase and relate the predication of VP to some speech time and discourse situation of the verb event, relative to some specific world. This results in a model of functional hierarchy that places Evidential under Tense Phrase. I conjecture that this Evidential position is a predicational one, in contrast to the more accepted notions of Moodevidentiai or Modale pistemic, which are known to be above Tense Phrase. I provide two detailed models, one with the conventional hierarchy and one with my hierarchy, arguing for the latter-based on general principles of syntactic economy and locality. I also provide a technical analysis of syntactic locality for the morphosyntactic fusion of tense-evidentials in a Distributed Morphology framework. v drop oftense-ofthe ofYP ofthe offunctional Moodcvidential Modalepistemic, latter-based tense­evidentials framework. v I dedicate this work to my wife, who has always believed that I will be a good linguist-I hope the following does not disappoint. linguist-TABLE OF CONTENTS LIST OF TABLES ix LIST OF ABBREVIATIONS x ACKNOWLEDGMENTS xi INTRODUCTION 1 1. TUYUCA AND ITS PLACE 2 1.1. The people and the land 2 1.2. The language and the family: Sources and influence 4 2. THEORETICAL BACKGROUND 6 2.1. Introduction 6 2.2. Morphosyntactic assumptions 8 2.3. Additional comments on the organization of raw data 9 2.4. Literature overview 10 2.5. Some terminology 12 2.6. A note on glosses in the data 13 2.7. A note on data 14 3. BASIC TYPOLOGICAL PROFILE 16 3.1. Introduction 16 3.2. Word order in Tuyuca 16 3.3. Some features of agreement in Tuyuca 18 3.5. Personal pronouns 23 3.7. Auxiliary verbs 25 3.8. Evidentials 26 ABSTRACT .................. ........... .. ............................. .................. ....... ................... ............... iv ..... ...... ................ ............ ....................................... ............. .. .. ............. .. ABBREVIA TIONS ................. .. .. ......... .................... ........... ............. .. ................ .. .. ............. ....................................... .............. .......... .. ............. ............................ ........................................................................ ........... I I. ......... ........ ...... .. .......................................... .... ................ ............................... ......................................................... 1.2. The language and the family: Sources and influence ................................... ....... .4 2. THEORETICAL BACKGROUND ................................................... ...................... .. .. 6 2.1. Introduction ................................................... ..................................... .......... .. ...... 6 2.2. Morphosyntactic assumptions .................................................... ................... ....... 8 2.3. Additional comments on the organization of raw data ......... ...... .......................... 9 2.4. Literature overview .............................................. ............................................. .. 10 2.5. Some tertninology .... ..... ..... .............................................................................. .. 12 2.6. A note on glosses in the data .......................... .................................................. .. 13 2.7. A note on data .. ................... .. ............................... ............................................. .. 14 ..................................................... .............. ...... .......................... .. ................................................... ......................... ........................................................................ ................. .................. ........... ......... ......... ............... 3.4. Nouns .... ... ... ...... ...... ...... ............. .............. .. .............................................. ........... 19 ..................................................... .. ....................................... 3.6. Verbs ......... ............................ ....... .. ........... .......................................................... 23 ............................. .. .................................................................. .. .. ........................................................................................................ 4. "G" CLASS AGREEMENT 30 4.1. Introduction 4.2. The basic data 4.3. predictable environments 45 Animate possessive noun is not 46 5. AGREEMENT AND EVIDENTIALS 53 5.1. Introduction: Against conflating agreement and evidential morphemes 53 A note on auxiliary agreement 73 5.5. Tuyuca, agreement, and tense-evidentials 75 5.6. A note on reanalysis of aspect 86 5.7. Conclusion 87 6. THE END RESULT 89 Appendices A. TUKANOAN FAMILY CLASSIFICATION 94 B. TUKANOAN LEXICAL COGNATES 96 C. POTENTIAL NOUN CLASSIFIER COGNATES 98 D. CINQUE HIERARCHY 100 E. DATA SET 102 REFERENCES HO viii 4. CLASS AGREEMENT ...................................................................................... 4.1. Introduction ............................................................ ...... ...................................... 30 4.2. The basic data ............................................ .................................. .... ... .. .............. 31 4.3 . Specifying predictable environments ................................................................ .45 4.4. Animate possessive noun classifier is not a gerund .......................................... .46 4.5. The gerund ......... .. ..................... ............................................ .............................. 49 4.6. Conclusion .......................................................................................................... 50 EVIDENTTALS ....... ................. ........... .................. ................. ............ 5.2. Basic issues ................................................... ....................... ................... ............ 62 5.3. Tuyuca does not have pronominal argument suffixes ...................... .................. 64 5.4. agreement.. .......................................................................... ...... .. .. ... .... .............. .............. ............. ....................... .. ........ ... .. ............... .. .... .. ...... ... ...... 5.7. Conclusion ........................................................................ ..... ..... ........................ 87 6. THE END RESUL T ........................................................ ..................................... .. .... 89 Appendices A. TUKANOAN F AMIL Y CLASSIFTCA TTON ..................................................... ........ 94 B. TUKANOAN LEXICAL COGNATES ............. .. .......... ......... ....... ............ ... .. ... .......... 96 C. POTENTIAL NOUN CLASSIFIER COGNATES ...................................................... 98 D. CINQUE HIERARCHy ................................... ... ........ .. ................................... ........ .. 100 E. DATA SET ..................... ........ ........................................................................... ......... 102 REFERENCES ................................. ............................................................................... 110 LIST OF TABLES Table Page 1. 2. Noun agreement in East Tukanoan 20 5. Evidential paradigm for Tuyuca 22 6. Nominalizer/Deverbalizer agreement 32 7. Agreement paradigm for dependent verbs 32 8. Present tense evidentials in Tuyuca 35 10. Present tense agreement for verbs 40 11. Present tense nominalizer agreement 40 12. Animate classifiers 48 13. Predicted environments for "g" class morphemes 51 14. Basic typology for tense-evidential languages' 57 15. Revised Tuyuca paradigm 61 I. Nouns in Tuyuca ..................................... ........ .... .... ... ...... .. .................................. .... 20 ...... ... ....... .... ...................................................... 3. Personal pronouns ...... ...... .. .. .. ....... ......... ...... ........................................... .. ............... 21 4. Agreement paradigm for dependent verbs ...................................................... .... ..... 21 ......... ... .... ....... ........................ ...... .. .. ..... ................. ......... ... ............ .................... ....... ........... .... .... ........... ........... .......... .. .................. ....... .... ... .... .... ......... ........ .... ........................ .. ........... .... 9. Gerunds ..... ... .................. ................ ......... ... ........ ........ .. ... ... .... .... .................. ........ .... 40 ................................... ........ ........................... ... .40 II. agreement.. ........................................................ ........ .. .40 ................ ............ .... ....... ......... .... .... .. .......... .. .. ... ......... ............. ............................................. ..... ................... ............... .. .......... .... .... .................. ..... ....... ........................................................ LIST OF ABBREVIATIONS DAT desiderative feminine future genitive gerund IMPF imperfect(LOC locative masculine modal negative nominative nonpast NVIS nonvisual evidential OBJ object INAN inanimate INFR inferential evidential INT interrogative INTR intransitive INTS intensifier/intensive IO indirect object P transitive patient PART participle PASS passive PCLF possessive classifier PFCT perfect PREP preposition PRF perfective PRN pronoun PL plural POSS possessive POST postposition PROG progressive PRS present PRT preterite PRXT proximate (1 deictic) PST past QUOT quotative RDP reduplication REC recent (past) REL relative clause marker REP reportative evidential SBJ subject SCD second-hand evidential SG singular SPEC specific SUBJ subjunctive SUBR subordinator SUP superlative SUS sustantivador TNS tense TOP topic TR transitive VISL visual evidential VN verbal noun 1 first person INTS intensifier/intensive 2 second person 10 3 third person P ACC accusative ADVR adverbializer PASS passIve AN animate APR apparent evidential ASM assumed evidential ASP aspect PRF AUX auxiliary BEN benefactive CLF classifier pass COMPL completive CONC concessive OAT dative DEF definite DES des i derati ve I DUB dubitative EVD evidential QUaT F feminine reduplication FUT future GEN genitive marker GER gerund evidential TMPF imperfect( ive) SB1 subject LaC locative evidential M masculine SG singular MDL modal SPEC specific NEG negative SUB1 subjunctive NOM nominative subordinator NPST nonpast superlative NVIS nonvisual evidential SUS sustantivador OB1 object TNS tense TN AN inanimate TOP topic INFR inferential evidential TR transitive TNT interrogative VISL visual evidential TNTR intransitive VN verbal noun ACKNOWLEDGMENTS I would like to acknowledge my intellectual debt to Lyle Campbell, who changed my view of linguistic work, literally, on the first day of graduate classes. I thank him for all his help and for his unwavering insistence that students not only understand a theory but critically evaluate it. Much of the material of Chapters 1, 2, 3 and 4 had their beginning in his Documenting Languages and Typology and Universals classes. Thank you also to Mauricio Mixco for always being open to observations and for help on Spanish to English translations. I also want to thank Ed Rubin for being a great Department Chair and for effectively communicating his respect for the study of linguistic science. An additional thank you goes to Aniko Csirmaz for all her detailed help and observations. Her dedication to understanding linguistics, critical awareness, and observational scrutiny is an example. Much of the material in Chapter 5 had its beginning in her Morphology and Advanced Syntax classes. Thank you also to Wilson Silva, who convinced me in our American Indian Languages course with Professor Campbell to start studying Amazonian languages. He also deserves thanks for allowing a near stranger to accompany him on a field work grant we wrote together in summer 2007, and for letting me stay at his parent's house in Manaus for a few weeks-and also for putting up with my horrible Portuguese. Lastly, thanks to Thiago Chacon who was always excited to talk and share his knowledge about Tukanoan languages. This thesis could not have been written without the help of those mentioned I, trarislations. weeks-There are many linguistic issues not covered in these chapters, but in another grammaticalization, evidentiality, economy in minimalist syntax, cartographic domains, tense-evidentials, and sisterhood relations for fused morphemes in a Distributed Morphology thesis; there is a definite risk that I have bitten off more than I can chew. However, I T am dealing with necessitates touching T welcome the challenge. Anytime one comes across a sufficiently I thesis has sufficiently narrowed the focus of research to two particular aspects of agreement in Tuyuca, namely the morpheme l-g-l and the inflectional ending of verbs, and it would be regrettable if I shyed away from the complicated problems that these pieces of Tuyuca morphosyntax point towards. INTRODUCTION sense, there are too many. For example, Chapter 5 touches on issues of language change, the semantics of tense and aspect, the typology and morphology of fused tense­evidentials, framework. Chapter 5 is by far the most ambitious, abstract, and theoretical section of the T believe that an analysis of the grammatical issues I am dealing with necessitates touching on all these issues and I welcome the challenge. Anytime one comes across a sufficiently narrowed focus of research that seems to be at the crossroads of so many theoretical paths one should never shy away in fear of having to confront complications. T believe that this I-I 1. TUYUCA AND ITS PLACE 1.1 .The people and the land The Tuyuca people, as part of the larger Tukanoan language family and culture, belong to land on the border of Colombia and Brazil, which makes up part of the Amazon region. More specifically, they live along the Tiquie River and the Onca, Cabari, and Abiyu streams as well as along parts of the Papuri River (Barnes 1984, 1990, 1994)-all are areas of the Upper Rio Negro area and Vaupes River region. The language is spoken by fewer than 300 people in Colombia and 590 people in Brazil (Barnes 1984, 1990, 1994); see Figure 1. The Tuyuca have roughly 15 sibs, or tribes, and they are expert basket-weavers and canoe carvers. The Tukanoan family, to which Tuyuca belongs, consists of twenty languages separated into three subbranches (see Appendix A for classification). Tuyuca is commonly called by native speakers Dokapuara or Utapinomakaphona. Other names include Tuiuca, Tuyuka, Tuyuka-tapuyo, Doch Kafuara, Dohka-poara or Doxkapuara (Stenzel 2006, Fabre 2005). Tuyuca has no reference grammar but articles written by Janet Barnes, a missionary linguist, over a period of twenty years contain much essential data. My analysis is based on her work (see references), as well as on Karn (1976, 1979). The Tukanoan people are famous for their exogamous marriage patterns. Exogamy here is dependent on language identities. This results in a bilingual (at least) home environment where the maternal and the paternal languages are spoken. However, 1.1. Onlj:a, Abiyli 1994)-all ofthe I. Utapinomakaphomi. the child identifies with the paternal language. This language is considered the prime determinate of ethnicity. Consequently, there is no distinction between an individual's language, ethnicity, and status in Tukanoan culture. If one identifies with the Tuyuca language then, one is also ethnically Tuyuca and assumes the social status that Tuyuca ethnicity has in the Tukanoan communities. In other words, the label Tukanoan is a cultural and social umbrella term which subsumes under it the various languages and cultures that are part of the Tukanoan family. This includes the Tukano language, from which the family derives its name, as well as the twenty or so other languages. Tukanoan people have a very keen linguistic sense for recognizing affiliations to their own languages. As far as I know, there are not many instances where linguists and native speakers of Tukanoan languages disagree about which languages belong to the family; though linguists disagree among themselves about the internal classification, specifically in regards to the split between the Eastern branch and the other languages (Appendix A). Tukanoan languages are closely related and share many similarities (Appendices B and C for some comparisons). The Tukanoan social hierarchy of language/ethnicity reportedly has its origin in a Boreka-Desano creation myth communicated to me by Feliciano Pimenta Lana, a native Tukanoan who identifies with the Boreka dialect of the Desano language/ethnicity. According to his myth (personal communication) the beginning of the world consisted of all people riding inside a giant snake that swam the river waters. At some point the snake came on land and the more daring people that happened to be at the front of the snake decided to go outside into the world. The relative time at which people exited the snake 3 Tuka no an specifically Band 1,2.The language and the family: Sources and influence Little is known about how agreement systems in Tukanoan languages work. For example, interactions between stress and tone/pitch-accent (de Lacy 2002 for Kubeo and Bowles unpublished 2007b for Wanano) may help mark distinctions in person categories Gomez-Imbert (1996) and Gomez-Imbert and Kenstowicz (2000) show that a distinction between person categories in Barasano is partially coded by a tonal alternation such that 1s t and 2 n d person, [+person], are signaled by a High tone and 3 r d person, [-person], is signaled by a falling tone HL (see also Barnes 1996 and Smith 1998 for analyses of Tuyuca prosody and its interaction with morphosyntax). Needless to say this area needs much more investigation. T do not consider prosodic factors in analyzing the agreement 1 of a determined their status within the Tukanoan society. The first to emerge from the snake and explore the world were the Boreka-Desano group. Portuguese and/or Spanish are spoken by most Tukanoan young adults and children. These colonial languages are taught in school and commonly used in commerce. The Tucano language, from which the family gets its name, is well known as a lingua franca for the Tukanoan people and may possibly be considered a prestige language along with Spanish and Portuguese. Most or all of the Tukanoan languages (Appendix A) are losing ground to Tucano, Spanish, and Portuguese. The future for Tuyuca1 is the same as the other Tukanoan languages: encroaching economic and cultural forces in Brazil and Colombia are endangering the language and culture. 4 lingua Tuyuca1 1.2. fam ily: influence categories. 1st 2nd 3rd r do not consider prosodic factors in analyzing the agreement I During field work in Summer 2007 I had heard stories ofa Tuyuca school in one of the communities. It was, reportedly, doing well and had sufficient funding. system of Tuyuca because the relevant information is not fully available in written sources and the emphasis of this analysis is morphosyntactic. The Tukanoan languages exhibit other typologically interesting phenomena that, if investigated properly, may add tremendously to our understanding of the typology of languages and to language universals. For example, Silva and Bowles (2007) note some possible inconsistencies in current typologies of noun classification systems compared to the function of noun classifiers in selected Tukanoan languages. Malone (1988) uses Barnes' evidential data from Tuyuca to give an historical account for the origin of evidential morphemes in a number of Tukanoan languages. De Haan (2001) uses the same Tuyuca data in work about the cross-linguistic grammatical origin of visual evidential ity; see also Palmer (2001) and Payne (1997) for use of Barnes' data. Various theoretical accounts have used data from Tukanoan languages to support their claims: Kaye (1971) attempts theoretical explanations of nasal harmony based on his field work of Desano, and Kaye (1970) is one of the first generative analyses of evidentials. Walker (2000, 2001) uses Barnes work on Tuyuca for an optimal ity theory account of nasal harmony and transparency. Smith (1998) uses Tuyuca data found in Barnes' work to make claims about noun-faithfulness and word stress in an optimality theoretic framework. Faller (2001) also uses Barnes' work in attempting to establish a hierarchy of evidential types. Lastly, Bowles (to appear) incorporates data from Barnes' work on evidentials to support claims about phrase structure hierarchies and the typology of fused tense-evidential morphemes in a generative/minimalist program framework. 5 200 I) evidentiality; De sa no, 2000,2001) optimality framework. 2. THEORETICAL BACKGROUND 2.1 .Introduction It is commonly understood that there is no linguistic (or scientific) analysis that is completely free from theoretical persuasion or bias (Chomsky 2000, Fodor 1984, Greenberg 1970, Hanson 1969, Popper 1959, Quine 1960). For example, in compiling the of Agreement, Corbett (2003: 4) states "the notion of an atheoretical database is a chimera. We have to base ourselves on generally shared theoretical views...." Also, Newmeyer (1998) has argued that an understanding of how any particular language works in a descriptive typological-functional view can only arise of that language's formal one's commitment to the ontological2 status of linguistic elements when writing reference grammars. That is, recognition of the existence and interaction of categories such as typological phenomena such as predication, case, modification, and argument structure. Different theories have different ways of explaining these language phenomena. In providing felicitous descriptions of a language L one does not want notation and jargon to 'get in the way' of a useful description. The tension between useful description and 2 1) 1960). 2.1. Surrey Database '"views .... " from the more abstractly theoretical (generative) analysis ofthat fonnal structure. Campbell (2007 class notes) has highlighted the importance of recognizing ontological2 Noun, Verb, and Adjective presupposes some theoretical commitment to basic fel icitous 2 By "ontology" I mean: (I) the general linguistic ontology found in Farrar and Langendoen (2003) and the rest of the GOLD and E-MELD projects, as well as (2) the narrow philosophical notion of the existence of a class of certain primitive objects. For example, most linguists make at the very least an ontological commitment to the existence of nouns, verbs, and sentences/clauses (see Quine \960). theoretical explanation3 is also carried over into projects with less ambitious scope than databases or reference grammars, such as this one. 1 take issue with certain inconsistencies in the description of Tuyuca. Namely, there seem to be obvious and systematic patterns of agreement marked on verb phrases that are not reflected in the interlinear glosses of primary data. I assume a particular theoretical framework about morphology and syntax (discussed in §2.2), in which revising these glossing inconsistencies leads to conclusions about the morphosyntax of Tuyuca that go beyond mere notational differences. This thesis is about the agreement system of a language belonging to in a group of languages that is geographically situated in an area of the world not well documented. It is done explicitly in a functional-typological framework (Chapters 3, 4) but also assumes some particular insights from generative-formal theories (see §2.2) that are applied in Chapter 5. Because of the lack of general knowledge about grammatical systems of languages in the Amazonian region (compared to Indo-European languages), the morphemic glossing techniques and the information they convey take on more importance than the glossing techniques of languages with a larger body of critical literature, e.g., English or Spanish. The Tuyuca language data I am working with are entirely second-hand, coming from published articles by Janet Barnes and Gloria Jean Karn (see references). Their publications are the only materials based on first-hand field work that are widely accessible to the mainstream linguistic community; no other 3 See Woodbury (2003: 42) "... there is a dialectical relationship between corpus and apparatus - the corpus informs the analytic apparatus; but analysis - including everything you bring to the table when doing grammatical and lexical elicitation - in turn also informs the corpus." Campbell (2007 class notes) highlights the notion of the hermeneutic cycle: one's expectations or theory of a language is transformed by the primary data while at the same time the kind of primary data one looks for or collects is informed by what one's linguistic theory says is important for understanding a language and its organization. See also (1998) 7 explanation3 I Namely, differences. 3 " ... Lass (1997) and Himmelmann (\998) for discussion of hermeneutic explanation. 8 accessible documentation of the Tuyuca language exists. There are also no competing analyses about the data from either native speakers or linguists (Karn sought the advice of Barnes in her analysis and interlinear glossing of Tuyuca). To make matters worse, Barnes is not consistent in the way she glosses the primary data. For example, work published in English glosses the auxiliaries tii and wz as English 'be' (Barnes 1984, 1990, 1994, 1996, 1999), but work published in Spanish gives 'ser' for riu and 'hacer' for tii (Barnes 1977, Barnes and Malone 2000). Translated into English this should give 'be' for 'ser' and 'do/make' for 'hacer.' The negative result of this inconsistency is someone only familiar with Barnes' English publications may assume that tii and riu are morphophonemic alternates or that a syncretism in the paradigm of 'be' is left unexplained; she gives no explanation for the pair nor does she draw attention to it in her English publications. The reality is that they represent two different classes of auxiliaries that influence the behavior of the Tuyuca inflectional system (see Chapter 4 and footnote 3 for more details). 2.2. Morphosyntactic assumptions The theoretical model of morphology and syntax that I assume can be summed up nicely by Baker's (1985: 375) Mirror Principle. (I) The Mirror Principle Morphological derivations must directly reflect syntactic derivations (and vice versa). Linear relations between morphemes reflect syntactic embedding. For example, take the sequence /gi/. The difference between analyzing -gi as the agreement morpheme versus -i ofthe Kam Iii nTi nTi Iii Iii nTi assumptions Principle 9 4 See §2.1 here and Barnes (1984, 1994, 1996). Two phonemically distinct auxiliary verbs are glossed as 'be', implying they may be allomorphs. But in Barnes and Malone (2000) the same auxiliary verbs are glossed as Spanish 'hacer' and 'ser', translated into English as 'do/make' and 'be', respectively. Comparison with other Tukanoan languages shows that the auxiliaries are in fact semantically distinct and are not allomorphic, as Barnes (1984, 1994, 1996) seemed to imply. as the agreement morpheme (where g- would have to be some other morpheme) means that two different syntactic structures are being implicated (see Chapter 4). Some uses of the Mirror Principle can be found in Cinque (1999), which is a typological-generative survey of languages using the Mirror Principle to establish a universal syntactic hierarchy, and Rice (1997) which uses an approach similar to the Mirror Principle that related semantic scope and morpheme order in Athapaskan languages. A typological-functional parallel of the Mirror Principle can be found in Bybee (1985). 2.3. Additional comments on the organization of raw data The decisions of how to organize the raw linguistic data (reflected in the interlinear glosses) have already been made by Janet Barnes. It should be noted that her interlinear glosses from the period of 1976 to 2000 are not entirely consistent.4 Furthermore, some of the publications contain only a translation of the Tuyuca data into Spanish or English while other publications contain fairly detailed interlinear glossing. No other substantial work exists to test her analytic decisions against, and access to native speaker judgments about her analysis is hard to come by. While this makes things philosophically difficult, and even linguistically hard, it does not make a significant (re)analysis of Tuyuca agreement patterns impossible. I assume most of Barnes' initial description is correct and generally accept her decisions about how to organize the data. In fact, the motivation for my analysis partly comes from a suggestion in Barnes (1984: 4 significant 4 2. I 1994. 10 5 rarely glosses evidentials, instead glossing them as agreement. This is not impressive, as the evidential system in Tuyuca is one of the most complex and unique systems in the world; to not recognize the evidential system through interlinear glosses is missing an essential part of the grammar of Tuyuca and the general character of the typology of the evidentials. 258) about the need for further study in the possible separation of agreement morphemes from evidential morphemes. A large part of this thesis concerns my reanalysis of the interlinear glossing of the existent Tuyuca data. I take advantage of Barnes' various morphemic glosses and the glosses found in Karn (1976, 1979),5 as well as cross-linguistic work on other Tukanoan languages including sketch grammars, journal articles, and dictionaries. Although it may seem to some that taking liberties with the interlinear glossing of data is not productive, or that changing interlinear glosses is merely a matter of notational difference, I show in Part II that such liberties are productive, and in fact allowed by the Leipzig Glossing Rules. Also, I show that a "mere" change in notation by the placement of hyphens between morphemes can result in very dramatic differences of analysis - even if the Mirror Principle (and the entire body of generative theory that goes with it) is not assumed. In the work that follows I will always explicitly point out glosses that are the result of my analysis. 2.4. Literature overview There are four groups of publications that are relevant to the analysis of Tuyuca agreement that I attempt here. Of these groups, none actually contain raw data from the language. (For all work on Tuyuca containing raw data see the references for Janet Barnes and Gloria Jean Karn). 10 5 [ [ overview 5 Karn's (1976, 1979) interlinear glosses are problematic for many reasons but most obviously because she ofthe ofTuyuca 2 . 4 . 1 . The Summer Institute of Linguistics (SIL) There are four SIL books focusing explicitly on a Tukanoan language and published under the Studies in the Languages of Colombia series. They are all roughly 200 pages long and read like small sketch grammars. The languages included are Barasano (Jones and Jones 1991), Cubeo (Morse and Maxwell 1999), Desano (Miller 1999), and Retuara (Strom 1992). In addition to these texts, SIL also published a series titled Estudios Tucanos volumes 1-5 in the 1970s containing essential data and analysis covering phonetics/phonology, morphology, syntax, and discourse on several Tukanoan languages. 2.4.2. Other works on Tukanoan languages What follows is not a complete bibliography. Instead, I list works that I am familiar with and that have provided significant cross-linguistic help in understanding Tuyuca. These include Ball (2004), Cook and Gralow (2001), Gomez-Imbert (1996, 2001, 2007), Gomez-Imbert and Kenstowicz (2000), Kaye (1970), Malone (1988), Payne (1990), Silva and Bowles (2007), Stenzel (2004, 2007), Sorensen (1969), and Waltz and Wheeler (1972). See also Fabre (2005) for a fairly comprehensive bibliography of Tukanoan language sources. 2.4.3. Typological studies There is a lot of work related to agreement; I list what has been directly helpful: Corbett (1991, 2000, 2003, 2006), Comrie (1981), Givon (1984, 1990), Harris and Campbell (1995), Haspelmath et al. (2005), Mithun (2003) and Siewierska (2004). 2.4.1. S IL) I 970s languages 11 200 I), 2001,2007), agreement~ 2.4.4. Generative studies In the generative framework agreement has become a crucial topic. I list here work that T find stimulating and interesting; though some of the work listed below has not positively influenced me, it has had an effect on my thinking about significant aspects of agreement and what kinds of language data provide relevant information: Baker (1985, 1996, 2003a, 2003b, 2008), Bejar (2002), Bobaljik (2006, 2007), Boskovic (2007), Chomsky (1981, 1995, 2001, 2005), Cinque (1999), Csirmaz (2006), Halle and Marantz (1993, 1994), Harley and Ritter (2002), Jelinek (1984), Miyagawa (2005), Pesetsky and Torrego (2001, 2004, 2007), Rubin (2002, 2005), and Speas (2004a, 2004b, 2007). They all represent various, if not at times conflicting, analyses important to agreement in current (morpho)syntactic theory. 2.5.Some terminology There is some variation in the way certain terms relevant to agreement systems are used. Different linguists use and define terms differently. Many of these terms include: concord, agreement, government, controller, feature, value, domain, condition, target, probe, anaphoric pronoun, pronominal argument and agreement marker. If, or when, I use these terms I explicitly define them and hopefully have made their use clear in the context in which I use them. Also, in Chapter 5 there is quite a bit of technical formalism that I use in order to sketch the theoretical significance of the results of my analysis. Understanding these formalisms is not crucial to understanding my analysis in Chapter 4, but a solid-at least introductory-background in Minimalist Syntax is 12 studies I find stimulating and interesting; though some of the work listed below has not oflanguage Boskovi6 Pesetskyand morpho )terminology condition, solid-introductory-background 13 necessary to understand the potential issues relating to those sections and Chapter 5 in general. 2.6. A note on glosses in the data I highly recommend looking at Appendix E. is part of a larger project of mine (The Tuyuca Data Set) to collect all the published data on Tuyuca and apply "good practices" consistent interlinear glosses so that the data set can be formatted for digital archiving. For the purposes of this thesis, I have taken part of The Tuyuca Data Set and given two contrasting glosses for each example. The first set of glosses (with whole sequential numbers, e.g. 1, 2, 3...) represents what I have superficially modified from the work of Barnes and Karn (see references) to conform to the Leipzig Glossing Rules (LGR). The second set (numbered 1.1, 2.1, 3.1...) reflects the results of my analysis. I provide Appendix E as way to prevent possible confusion in reading various glosses of the same data this thesis. For example, Chapter 3 contains glosses superficially modified to conform to LGR and comprise the first set Appendix E. Chapter 4 contains multiple versions of glosses for the same data as a means to show how Barnes' collected work is itself inconsistent, and also to reflect different analyses and their interpretations. Chapter 5 contains data that reflect the application of my analysis and constitutes the second set Appendix E. By providing an appendix that compares "before" and "after" glosses next to each other one can quickly reference my analysis of Tuyuca without having to troll through each chapter and risk being confused by various analyses of data. It I, 3 ... ) 3.1 ... ) in superficially in inconsistent. appl ication in 14 2.7. A note on data of the (Christian) material from the Joshua Project (www.ioshuaproject.net) that need to be transcribed. Most of the material comprising the accessible data for Tuyuca comes from narratives. It is common knowledge that narratives usually have some kind of scripted form or standard structure, e.g., beginning, middle, end, as well as the use of proverbs, metaphor, and other narrative techniques. Tuyuca is no exception (Karn 1976, 1979). Barnes' work contains a lot of dialogue, but it appears to be mainly dialogue between her and native Tuyucans. In other words, none of data appears to be representative of common daily discourse between native Tuyuca speakers. This does not pose an insurmountable problem, but it does mean that certain avenues of investigation can go so far. For example, in Chapter 5 I apply numerous diagnostic techniques in order to form hypotheses about the structure of Tuyuca. Namely, I try to see if Tuyuca is a configurational or nonconfigurational language by testing for the existence of pronominal arguments, which are full arguments of the verb that look like agreement markers. One way to test for this is to find sentences where a full subject/object NP co-occurs with subject/object agreement. In languages with pronominal arguments the full subject/object NP will be treated as an extra, dislocated, element. One test for this is to look at word order-if the subject/object NP has a variable and unpredictable position in the clause it might be dislocated. Another test is the presence of an intonation pause between the full subject/object NP and the rest of the clause. The problem for Tuyuca is that most of the data only shows the canonical S(0)V order-and even in the strictest word order The data that comprise general knowledge ofthe Tuyuca language come from published, accessible sources. There are also quite lengthy recordings of religious joshuaproject.T nonconfigurationallanguage tind subject/order-ifthe SUbject/object O)order-languages one can always find variable word orders due at lest to topic-focus factors (of course, pidgins may be an exception). But variable word orders do not really show up in the Tuyuca data, though one can reason that they must be allowed by the constraints of Tuyuca grammar; see especially §5.3.2 and footnote 23. 1 will assume throughout this thesis that while the lack of a more complete set of data showing a richer set of phenomena puts limits on the depth of analysis, it does not make analysis impossible or unproductive. I take for granted that the data that exist are a fair characterization of the basic form of Tuyuca grammar and are a consistent reflection of the empirical facts. It would be nice to have a large-scale Tuyuca reference grammar, but this does not preclude a basic analysis. And a basic analysis does not limit the quality of interesting and significant results. 15 T of lim its 3. BASIC TYPOLOGICAL PROFILE 3.1 .Introduction The profile here is intended to introduce basic data from Tuyuca as presented in the published sources. The intention is to familiarize the reader with the language and to serve as the backdrop for the analyses I propose in Chapters 4 and 5. The profile here is not comprehensive and deals only with basic issues relevant to agreement, covering word order and agreement in nouns, verbs, auxiliaries, and evidentials. The interlinear glosses in the following examples are modified from Barnes' originals but do not reflect a substantial difference from her initial work; they merely conform to the Leipzig Glossing Rules (http://www.eva.mpg.de/lingua/index.html). 3.2. Word order in Tuyuca Barnes (1984, 1999) classifies Tuyuca as exhibiting a flexible SOV word order. Sentence (1) has the referential animate noun 'father,' the other sentence (2) contains a pronominally possessed subject ('her dog'). Only sentence (1) meets the strict criteria established for determining basic word order (e.g., it is indicative, has full noun phrases, the subject is definite, agentive and human, the object is a definite patient, and the verb is an action with two arguments where there is an obvious transfer of activity and an obvious effect from agent to patient). 3.1. (http://www.(999) SOY l) I) 17 (1) Pak+ yai sTa-yfg+. father jaguar kill-EVD.PST.scD.3MSG 'Father killed a jaguar.' Ko6-ya-g+ diyi y4+-re tuti-wi. 3FSG-P0SS.SG-NR.1 SG-SPEC SC0ld-EVD.PST.VIS.3MSG 'Her dog barked at me.' (adapted from Barnes 1994: 327-329) Because of the abundant use of pronominal agreement morphemes in Tuyuca, it is difficult to find data samples where the subject is named explicitly, i.e., is a bare definite noun. Tuyuca, like other Tukanoan languages is highly agglutinative with multiple suffixation of meaningful morphemes to the root. The verb phrase is the locus of most of the suffixation while the noun phrase is typically made up of smaller sets of morphemes and plays a smaller role in sentence construction than the VP. Postpositional case clitics are suffixed to NPs. The most common nominal elements consist of pronominal agreement markers, suffixed to the verb or auxiliary, signifying semantic and/or grammatical gender, number, or person. Because of the lack of example sentences that meet the strict criteria for determining basic word order, it is difficult to establish whether or not Tuyuca is consistent with SOV, OV, and SV, but the small number of examples clearly show S(0)V typology. The data also seem to suggest the following orders: AdjN, and GenN. The WALS database (Haspelmath et al. 2005) classifies Tuyuca as having postpositional clitics (NPost). These orders are confirmed by (3) AdjN, (4) GenN, and (5) NPost. (3) STa-ri-d+ka nana-ri-d+ka lNAN.NCL.(1) Pakt sia-yfgt. kill-EVD.PST.SCD.3MSG (2) Koo-g.j. dfyi yi-t-tutf-3rSG-poSS.SG-NR.MSG dog I SG-SPEC scold-EVD.PST.VIS.3MSG SOY, O)W ALS Haspel math ditics Sia-rf-d.j.rf-d.j.illuminate-NR.INAN.SG-NCL.stick be.bad-NR.lNAN.SG-NcL.stick 18 b+k+-d+ka. old.object-CLF.stick 'A terrible, old flashlight.' (4) Koo-ya-g+ k+f-ya-gi-kdro nn-T. 3MSG-P0SS.SG-MSG-3MSG 'Her animal is the same as his animal.' (5) Wese=p+. LOC 'To/at the field.' (adapted from Barnes 1994: 328-330) The linear order of elements in (3)-(5) conform to the general predictions of the order of elements in the standard SOV, OV, and SV typologies (e.g., Greenberg 1966, Comrie 1981, Givon 1984, Song 2001). This provides more evidence that in spite of the small number of prototypical word order sentences showing SOV order that it is in fact a correct generalization. Additionally, Tuyuca appears to have a very strict SOV order. Although Barnes classifies it as having "flexible" word order none of the available data actually shows examples of discourse dependent alternatives such as verb or object fronting with full NP subjects, i.e., VSO, OSV; see Appendix E. 3.3. Some features of agreement in Tuyuca review here some aspects of agreement on different categories showing how they interact on a basic level (i.e., subject-verb agreement). I do not cover the nominative (null) and accusative (/-re/) case system of Tuyuca because there is not a large degree of morphological case marking in Tuyuca that deals with subjects and direct objects (e.g., examples 2 and 5). biki-dika. cLF.flashlight.' ya-gi kH-koro nit-i. 3FSG-POSS.SG-NR.MSG 3MsG-poss.SG-NR.MsG-alike be-EVD.PRES.VIS.3MSG animaL' field=Loc Tolat SOY, 200 I) . SOY SOY Tuyuca I (I-reI) ofTuyuca 19 Tuyuca agreement features can be found marked on nouns (Tables 1 and 2), personal pronouns (Table 3), verbs (Table 4), auxiliary verbs, and agreeing tense-evidentials (Table 5). These are not all the possible elements, but for purposes of space T limit the analysis to crucial items. Since agreement is obligatory in all clauses in Tuyuca, examples of agreement can be seen in any clause or sentence given throughout this thesis. 3.4.Nouns Agreement in nouns is marked in its most general form by masculine -i or -i, feminine -o, and plural -a; although these can code more information for person and/or gender depending on the morphosyntactic environment. They all have nasal counterparts determined by phonological context that do not differ in grammatical features. A difference in marking gender on nouns appears in Table 1 between (1 .a) and (1 .b) with the additional l\d preceding the gender marker in (1 .a). In plurals, the final vowel can signal plurality (l.c). Tuyuca also, however, has lexicalized plurals (l.d) that require a singularizing suffix (1 .e). In general, agreement features of the head noun determine agreement with other clausal elements. Table 2 shows the isolated agreement morphemes in Tuyuca that are suffixed to nouns-I compare Tuyuca with other Eastern Tukanoan languages that exhibit similar patterns. In these other languages such nominal agreement markers have been analyzed as noun classes (Retuara by Strom 1992) and noun suffixes (Wanano by Stenzel 2004); there is also the potential that they are a type of pronominal noun classifier or perhaps a pronominal clitic. I tense­evidentials -0, I I.I./k/ I.Tn I.I.I.e). Tn noun nouns-T Tn 20 Table 1 Nouns in Tuyuca Tuyuca Nouns with Gender Tuyuca Nouns in Plural l.a l.b l.c l.d l.e M Pa'ki 'father' bdbi 'child' ba'i 'brother' SG jukd 'vulture' and 'viper' wa'i 'pieces' -Wi Pa'ko singular 'mother' bd'bd 'child' bai'jo 'sister' PL jukd-a 'vultures' dnd-d 'vipers' bubid 'grandmas' Table 2 Noun agreement in East Tukanoan6 Tuyuca -ki/-i -ko/-o -a Cubeo -ki -ko -wa Desano -g» -go -ra Retuara -ki -ko -ra Wanano -«/-k« -o/-ko -a/-na 6 Included -ki - ko 'feminine singular', and -wa ' plural' (Morse and Maxwell 1999: 77); Desano suffixes -gi ' masculine singular', -go 'feminine singular', -ra 'animate plural' (Miller 1999: 35); Retuara noun class suffixes: -ki masculine', -ko feminine', -ra -u -o -al-na -ku -ko feminine', al-na plural' 1 Tuy_uca I I.I.I.I.1 babi juka ana -wi F ba'b6 juka-a ana-a biibia Tukanoan6 MSG FSG PL k+/--gi tt/-ktt 61ncluded here is Cubeo, whose most common noun agreement suffixes include ' masculine singular', ­ko singular ', wil ril plural ' ' masculine' , ' feminine ', ril 'animate plural' (Strom 1992: 47); Wanano bare nominal roots ending in 'masculine,' -0 'feminine,' ai-nil 'animate plural' and Wanano nominal suffixes 'masculine', -' feminine ', and -ai-nil 'animate plural ' (Stenzel 2004: 128). Table 3 Personal pronouns in Tuyuca 1s t Person Singular y+'t Plural exclusive Tsa inclusive badt 2 n d Person Singular m Plural *b+a 3r d Person Singular masculine k?1 feminine ko'o 3 r Plural 'k+a Table 4 Agreement for dependent verbs DEPENDENT VERB SUFFIXES Animate Singu Masculine ar Feminine Plural PRESENT -g+ -ra PAST rig+ -rigo -rira REC.PAST -arigi -arigo -arira FUTURE -idi -odo -adara 21 lSI Person yi't t'sa beIdi nd bt'i 'bta ra kt'i kta Singular -gt -go -ngt nra angt anra 22 Table 5 VISUAL APPARENT SECONDHAND PAST OTHER -w+ yiro hTyu 3MSG -hTyi 3FSG -to -yo -hTyo 3 PL -ya -hTya PRES OTHER -ga -ku 3MSG -i -gi 4TTT 3FSG -hto -ko 3 PL -hlra -kua Evidential paradigm for Tuyuca V ISUAL NONVISUAL ApPARENT ASSUMED Wt -tt -yu -ylro -hiyu -wi -ti -yi -yigi hiyi -wo -yigo hiyo 3PL -wa -ta -yira hiya -a -gi -hii -ki -yo -go hio 3PL -ya -ga hira 23 3.5. Personal pronouns The final vowels on the personal pronouns group generally to mark the distinction between singular -4 and plural -d; see Table 3. Two exceptions stand out from the singular/plural pattern: first person plural has an inclusive form and the third singular has a feminine form that correlates with the pattern for agreement on nouns. Examples (6) and (7) show the use of personal pronouns. (6) K3a-re Tna-d+ga-ri-yigo. See-DES-NEG-EVD.PST.SCD.'(She) did not want to see them.' (7) Koo-ya-gi diyi yit-re tuti-wi. lSG-SC0ld-EVD.PST.VIS.3MSG 'Her dog barked at me.' (adapted from Barnes 1994: 330) 3.6. Verbs Agreement on verbs, Tables 4 and 5, is typically marked through suffixing an obligatory agreeing-evidential to the end of the verb stem. The form of the agreeing-evidential is contingent on the agreement feature values of the head noun of the clause and constitutes subject-verb agreement. In the situation where an auxiliary verb accompanies the main verb (in which case I refer to the main verb as a dependent verb, i.e., dependent on the auxiliary), the evidential is suffixed to the auxiliary while the dependent verb expresses subject-verb agreement through a different class of morphs that also agree with the head noun. In these clauses the agreement feature values of the verb and auxiliary will be virtually the same, both agreeing with the head noun of the clause. This can be seen in Table 4, which represents my organization of the morphemes as f plural-a; Kia-irla-dtga-ri-yigo. 3PL-SPEC see-DES-NEG-EVD.PST.SCD.3FSG gt yii-3FSG-POSS.SG-NR.MSG dog I SG-SPEC scold-EVD.PST.VIS.3MSG verb, 24 7 itself (mood). found in the original sources; see Barnes (1977, 1984, 1994, 1996, 1999), Barnes and Malone (2000), and Karn (1976, 1979). To summarize, verbs that depend on an auxiliary must still agree with the subject/head noun of the clause, the morphs in Table 4 are used for this purpose; agreement on lexical verbs is suffixed to the end of the evidential morph and are placed in bold in Table 5. 3 . 6 . 1 . Order of constituents in the verb phrase Barnes (1999) notes that "modality indicators are suffixes which follow the verb root and precede an evidential, imperative or interrogative ending. These mood indicators include: negative, probability/conditional, contraexpectation, desiderative, ability, and we can derive the following order form Barnes' statement: [V-Modal/Mood7]. Barnes and Malone (2000) state that "Los modos irreal, frustrativo, desiderativo y potencial... siguen al tema verbal y a los sufijos de aspecto." In other words, the desiderative, frustrative, and potential moods follow the verb root and the aspect suffix. From this statement we can derive the overall [V-Aspect-Modal/Mood] order. Typically, no more than two elements occur between the verb root and the evidential ending, which are the only obligatory parts of the verb phrase. This order of verbal suffixes in Tuyuca appears to be fixed as [V-(Asp)-(Mod/Mood)-Evd]; elements in parenthesis are optional. The evidential is a 'fused' or 'portmanteau' form (i.e., contains multiple kinds of grammatical information 3.6.1. emphatic." Assuming the universal morpheme order in Cinque (1999, see Appendix D) ModaI/Mood7 ]. irreal. V -ModallMood] ModlMood)-7 Barnes makes no semantic distinction between mood and modality. For consistency I will follow her assumption in this Chapter, though I think she is wrong. A clear distinction between mood and modality can be defined as the difference between variables ranging over the speaker's opinion of a world (modality), and variables ranging over the world itself(mood). 25 3.7.Auxiliary verbs There are two Tuyuca auxiliary verbs, tii- 'do/make' and riii- 'be.' They only occur in declarative clauses when they have an agreeing-evidential suffixed to them. In (8-11) the main verb is suffixed with an agreement morpheme from Table 4, while the auxiliary carries the agreeing-evidential (examples are as they appear in the original texts except I superficially changed the interlinear glossing to conform to the Leipzig Glossing Rules). (8) we'se tada-'ra tii-'kua hacer-'seguramente estan rozando (preparando un terreno)' 'They are clearing the field' (my translation) (9) yi'i waa-'gi tii-'w+ hacer-EVD.PST.viS.1/'yo estaba remando' 'I was rowing' (my translation) (10) waa-ri-g+ ntf-wi RESLT-3MSG EVD.PST.VIS.3MSG 'He went.' (11) dfiga ape-gi tii-i MSG do/make-EVD.PRES.vis.3MSG 'He is playing soccer.' (adapted from Barnes 1984: 259 and Barnes and Malone 2000: 442) in one morpheme), encoding person, number, gender, tense, and the speaker's source of information; see 3.8 for more details on evidentials. nii- ni chagra cortar-PL hacer-EVD.PRES.ASM.3PL terreno), yj.'t gt wt yo remar-MSG hacer-EvD.PST. VIS.! /2 n-gt oii-gO-RESL T -3MSG be-EvD.PST.VIS.3MSG II) gt soccer play-MsG do/make-EvD.PRES.Vls.3MSG 3 . 8 . 1 . Types of evidentials in Tuyuca The morphosyntactic distribution of the five types of evidentials (Table 5) is the consistent between all the subtypes of eviential. That is, all types (and subtypes) may occur in the same environment: obligatorily suffixed to the end of the verb stem or auxiliary in declarative clauses. Tense, which is fused with the evidential, is marked for the whole clause whether it is suffixed to the main verb stem or the auxiliary. In (12a-c) the evidential is suffixed to the main verb. When there is an auxiliary the evidential attaches directly to it (12d-f). Both types of constructions for Tuyuca can be seen in (12). (12) tuti-wi b. hea-wa EVD.PST.vis.arrive-EVD.PST.vis.3PL t n ey) sla-yig* kill-EVD.PST.play-do-VIS.3MSG '(he) killed' '(He) is playing' Payne (1997: 256) says "This language [Tuyuca] has one of the most complex systems of evidentiality I have seen. It has the added complexity of having evidentiality interwoven with the verbal participant reference system and the tense system." 3.8. Evidentials The Tuyuca evidential paradigm is relatively large and complex, and has been used by many (e.g., Faller 2001, Palmer 2001) as an exemplary case of an evidential system.8 The evidential paradigm consists of the general distinctions between direct and indirect and shown in Table 5. These categories encode a general distinction between 3 r d person and OTHER (1/2 person). The OTHER (1/2 person) category makes no distinction between singular, plural, feminine, or masculine. On the other hand, the 3 r d person does make this distinction. 26 200 I, 200 I) 8 (VISUAL) (NONVISUAL, APPARENT, SECONDHAND, ASSUMED), rd 3rd 3.8.1. attaches directly to it (12d-t). Both types of constructions for Tuyuca can be seen in (12). a. c. tutt-scold-EVD.PST.VIS.3MSG 'barked' sia-yigi ki II-EVD.PST.SCD.3MSG h. d. arrive-EvD.PST.VIs.3PI, '(they) arrived' ape-g+ tii-i plaY-MSG dO-EVD.PRES.V[s.3MSG 8 ofevidentiality 27 rigi nfi-wi gO-3MSG EVD.PST.VIS.3MSG '(He) went* bue-go tii-a Study-MSG do-VIS.1/2 '(I am/you are) studying' (adapted from Barnes 1 9 8 4 , 1 9 9 4 ) The order of morphological affixes for the verb is given in the formula in ( 1 3 ) and for auxiliaries in 1 4 ) ; see also 3 . 6 . 1 . Optional elements are in parentheses and obligatory elements are not. The verb root and the tense-evidential-agreement morpheme are obligatory. (The symbol indicates a strict surface ordering relation in which V always comes before and there is no intervening material). 1 3 ) [VR OOT-(ASP)-(MOD)/(( 1 4 ) [VRQOT-AGR] + [AUX -(NEG)-(REC.PST)-EVD.TENSE.AGR]9 The five evidential categories generally split between one direct form and four indirect forms, as mentioned above. The [± 3R D person] forms are grouped under PST and tense, as can be seen in Table Barnes' original work shows two evidential categories that are not found in the common literature: and These may be replaced by the more common rNFERRED and HEARSAY, respectively. There 9 I TENSE EVIDENTIAL. I for placing TENSE after EVIDENTIAL; however, it should be noted that the two categories are morphologically fused (portmanteau) and any argument for an ordering relation must rely on an abstract characterization. , e. waa-ngt nii-wi go-3MSG be-EvD.PST.VIS.3MSG went' f. study-MSG dO-EVD.PRES.VlS.1I2 1984, 1994) 13) ( 14); 3.6.1. '+' Aux ( 13) [VRooT-(ASP)-(MOD)/(MOOD)-EVD.TENSE.AGR] (14) [VRooT-AGR] + [Aux -(NEG)-(REC.PST)-EVD.TENSE.AGR]9 3rd PRES 5. APPARENT SECONDHAND. INFERRED 9 Notice [ have been placing the category as occurring after [ give no arguments for this ordering here and assume it throughout Chapters 3 and 4. Chapter 5 contains the thrust of the argument TENSE EVIDENTIAL; 28 is no evidence that Barnes' labels refer to categories independent from the more common names. I will continue to use her labels for consistency. Lastly, the final vowels on the evidentials (Table 5) correlate with the gender and plural marking patterns in Tuyuca in general (Tables 1, 2, 3 and 4); analyzing these final vowels as separate agreement morphemes results in a very pervasive pattern throughout the entire language. And in fact Barnes herself acknowledges the pattern, stating that "Since the person, number, and gender information is carried mainly by the final vowel, it would seem that the evidential information is carried by the rest of the morpheme... this area needs further study" (Barnes 1984: 258). I investigate this pattern in more depth in Part II. 3.8.2. Defective forms in the evidential paradigm In the evidential paradigm, Table 5, the NONVTSUAL category in PST.1/2 and both share the final vowel /*/. Compare this to the category l/2 and PST.3MSG forms: they alternate between III ~ /+/. Segmental phonology of Tuyuca shows that l\l ~ /+/ alternations are phonemic (Barnes and Takagi de Silzer 1976). If the difference in vowel alternation in the category marks a distinction between 3r d] and [-3r d ] persons, then the loss of alternation in the category suggests a syncretism of the agreement values in the two forms that do not show the alternation. Further problems arise when looking at the VISUAL PRES. 1/2 and VISUAL PRES.3MSG forms. They appear not to have an evidential at all but to be composed simply of the vowel agreement morpheme. This really only holds, however, in the PRES.3SG morphs; the PRES. 1/2 is deviant. A pragmatic and functional explanation arises for the [ Tables!, morpheme ... [ paradigm NONVISUAL PST.! 12 PST.3MSG Iii. VISUAL PST. 112 Iii ~ Iii. ofTuyuca Iii ~ Iii VISUAL [+ 3rd ] [_3 rd ] NONVISUAL VISUAL PRES. I 12 VISUAL 112 dropping of a form that refers to the source of information, if and only if it is assumed that during natural speech all speech participants are located in the same spatio-temporal domain and have equal access to the reference of event times and world situations. this context the source of visual information is explicit to all speech participants and need not be specified. Thus, dropping the visual evidential morpheme is economical as it need not be redundantly stated to the speech participants that the source of information is visual; (see 5.5.2 for another explanation). However, this does not explain the other present tense visual forms of 3FSG and 3PL. These issues are important but they do not directly affect the analysis of agreement and so I ignore them in this thesis. Lastly, the OTHER evidentials clearly do not fit the final vowel pattern so pervasive in the 3 r d person forms. For this reason I ignore the OTHER evidential forms. 29 In 5.5 .2 of3FSG 3pL. affect rd r AGREEMENT 4.1 .Introduction There is a systematic pattern of agreement in Tuyuca by which the final vowel of certain classes of morphemes functions as the agreement marker cross-referencing the head noun of the clause (see Chapter 3). I investigate here the consequences of analyzing of morphemes, namely what is usually analyzed as the nominalizer, dependent verb agreement, and gerund of these morphemes (i.e. the analysis of Barnes and Karn) rests on viewing them as portmanteau forms - typically fusing gender and person with a of the "g" class morpheme. The goal of this chapter is to treat these "g" class morphemes as morphologically separable from the final vowels that If this separation is valid then a consistent analysis of the segmentable parts of what is assumed to be a portmanteau form will result. In other words, the "g" class morphemes will have consistent grammatical functions in specific and predictable morphosyntactic environments, as will the agreement morphemes. If this is so, then I believe I will have made a strong argument for analyzing agreement as a separate morpheme within this limited set of Tuyuca morphology. Additionally, this separation reveals some precise and interesting functions of the small set of "g" class morphemes that have not received adequate attention. If such an argument is valid then it may necessitate a reconsideration of other Tukanoan languages in this regard. 4. "G" CLASS AGREEMENT 4.1. the final vowel agreement markers as separate from a specific set morphemes. The usual analysis specific grammatical function encode the agreement feature values. environments. 31 Nevertheless, I limit analysis to Tuyuca and make no explicit comparisons between my conclusions here and the conclusions in work on other Tukanoan 1 0 4.2. The basic data Much of the analysis of Tuyuca data in this chapter will concern the interlinear glosses originally given in Barnes' work (see references). As a prelude to the analysis, I quote from the Leipzig Glossing Rules (http://www.eva.mpg.de/lingua/index.html): "Glosses are part of the analysis, not part of the data. When citing an example from a published source, the gloss may be changed by the author if they prefer different terminology, a different style or a different analysis." I make explicit the places in which I change Barnes' original interlinear glosses, but it should be noted that by and large I follow her analyses and conclusions. A difference in analysis, however, will not change the form of the basic data. It is to these data I now turn. Barnes (1994) and Barnes and Malone (2000) (BM from here after) analyze a specific set of morphemes that function to reduce the valency of the verb or turn it into a noun, which results in a reduction (or elimination) of the number of arguments of the verb. Table 6 shows this set of morphemes, a large portion of which can also be seen functioning as agreement markers on dependent verbs (lexical verbs that are dependent on an auxiliary in the same clause; see Table 4). Table 7 shows the full paradigm of agreement morphology for depenent verbs and is partly based on Barnes (1996) and on my observations of data in the rest of Barnes' work, as well as Karn (1976, 1979). 1 0 I call "I languages. 10 J (http://www.different tum. Kam 10 See Ramirez ( 1997) for comparison of what r call" g" class morphemes. r thank Thiago Chacon for pointing out some general similarities between my analysis and Ramirez's (1997). 32 Table 6 Nom inalizer/Deverbal izer agreement NOMFNALIZERS/DEVERBALIZERS Animate Inanimate Sinj Masculine jular Feminine Plural Non-count Place PRESENT -go, no -ra, ra -re -ro, -rd PAST -rig+ -rigo -rira -rige -riro REC.PAST -arigi -arigo -arira -arige -ariro 2 -idi -odo -adara -adare -adaro Table 7 Agreement paradigm for dependent verbs DEPENDENT VERB SUFFIXES Animate Inanimate Singu Masculine ar Feminine Plural PRESENT -g+ -go -ra PAST -rig* -rigo -rira REC. PAST -arigt -arigo -arira FUTURE -+d+ -odo -adara -ro 1 1 nominalized and the agreement marker refers to the head noun of the main clause. I am not concerned here with relative clauses. 1 2 so the paradigms can be seen in their entirety. ll NOMINALIZERS/Singular Non- count -gt, 1)1 1)0 -ra rigt nra arigt anra ange FUTURE' l. ad are Singular -gt rigt PAST .. angt -tdt II These nominalizers are also used in the construction of relative clauses, in which the restricting clause is 12 The future is a different matter from the other tenses and I will not deal with it here; though I include it (15) bue-g+ estudiar-MSG 'professor' (16) w++-ri-w+ volar-sus-CLS:vehiculo fly-NR-CLF:vehicle 'ariplane' (17) yaa-'re comer-sus eat-NR 'food* (18) buegotii-a '(I) am studying.' (19) diiga apegi tif-i '(He) is playing soccer.' (20) waarigi nTf-wi '(He) went.' (21) wese soerigi nTf-wi '(He) burned his field.' (adapted from Barnes 1984: 259) Examples (15)-( 17) show the nominalizing or deverbalizing function of morphemes from Table 6. (18)-(21) show the unanalyzed interpretation of morphemes from Table 7-1 argue that these morphemes (in bold) can be decomposed and analyzed as an agreement morpheme (the final vowel) and either an aspect, classifier, gerund, or nominalizer morph (the l-l element). Tuyuca systematically uses the morphs in Table 7 in [V Aux] combinations; 33 17) 7-J I-g-I + gt MsG study-NR.MSG wtt-wt SUS-CLs:vehfculo flY-NR-cLF:SUS food' buego tif-dfiga no-s6erigi no-wi 34 In spite of the fact that some of Barnes'original data provide no morpheme-by-morpheme widely cited (1984) paper on evidential, these examples provide interesting evidence about how progressive and perfect aspect are formed in Tuyuca. Barnes notes for (18) and (19), which constitute an event the speaker is witnessing at the time of the speech act report, "a progressive construction is used" (Barnes 1984: 259). For (20) and (21) she states that "Visual evidential are used in a compound construction to describe the end result of a state or event when the state or event itself was not seen but the end result was" (Barnes 1984: 259). By this latter description Barnes is referring to a perfective aspect (Barnes and Malone 2000: 442). (18), (19) and the perfective in (20), (21); especially when the auxiliary appears to be partly responsible for these aspectual interpretations. It is true that there are many cases in which evidentials have originated from reanalyzed aspectual-like morphemes, such as participles (Harris and Campbell 1995 and Campbell 1991) in Estonian and resultatives (Csato 2000, Johanson 2000, Shroeder 2000) in Turkish. In fact, Malone (1988: 139) has paradigms appear to have developed from a progressive (or other) aspectual gerundial construction... [and] the '+/- direct' paradigms appear to have developed at some later stage from an old perfect construction plus evidential suffixes...." There is even a small set of homophonous pairs in Tuyuca that retain both an evidential and an aspectual meaning. The forms g-i, g-o, and g-a can be progressive or perfective aspect and they are part of the evidential paradigm in Table 8. Barnes' original by­morpheme gloss to examples like (18)-(21), nor to any data in her very important and evidentials, evidentials ofa itselfwas It is not clear how the visual evidential is responsible for both the progressive in Iike Tn analyzed evidentials in Tukanoan as originating from aspectual morphemes: "Nonvisual construction ... suffixes .... " a Table 8 Present tense evidentials in Tuyuca VISUAL NONVISUAL APPARENT SECONDHAND ASSUMED PRES OTHER -a -ga 3MSG -i -gi -ki 3FSG -yo -go -hto -ko 3 PL -ya -ga -htra -kua 35 ApPARENT -ku -I -gi -hii hio 3PL hira 36 However, as I will soon show, there is no environment in which g-i, g-o, and g-a have simultaneous evidential and aspectual interpretations. The meanings of the morphemes are entirely dependent on morphosyntactic environment. This situation is similar to what has occurred in both Estonian and Turkish, where the reanalyzed morpheme still retains its original meaning in specific environments, while the evidential meaning is also restricted to other specific environments. For example, in Estonian the participle interpretation is found in subordinate clauses and the reported speech evidential is found suffixed to finite verbs in main clauses and subordinate clauses in both past tense /-vat/ and present tense /-mid/. In Turkish /mis/ is an indirect past tense-evidential with finite verbs, but signals resultative aspect with nonfinite verbs. Tuyuca is very different from Turkish and Estonian in that it has a lot more evidential morpheme types and subtypes that make many direct/indirect distinctions. In Estonian and Turkish a small set of aspectual-like morphemes underwent the historical process of reanalysis and, subsequently, came to have an evidential interpretation in restricted morphosyntactic environments. But in Tuyuca the evidence is weak that only g-i, g-o, and g-a (either as aspect or nominalizer/gerund) were reanalyzed and extended to make up the entire Tuyuca paradigm (see Malone 1988 for Proto-Tukanoan aspect and Harris and Campbell 1995 for technical exposition of Preanalysis' and 'extension'). Evidentials, cross-linguistically, interact with aspect and in some cases signal an imperfective, perfective, or progressive aspect (see Johanson and Utas 2000, Stenzel 2004, Sumbatova 1999 for Turkic languages, the Eastern Tukanoan language Wanano, and Kartvelian languages, respectively). In many cases of historical reanalysis of aspect to evidential there are cases of ambiguity between aspectual or evidential interpretations. But the homophonous pairs g-spectfic I-vat! I-nud/. Iml~1 different ofaspectual-techn ical 'reanalysis' extension '). cross­linguistically, Vtas 37 of nonvisual evidentials and aspectuals in Tuyuca (g-i, g-a, g-o), while left unexplained for now, never appear in environments where they can be simultaneously or ambiguously interpreted as evidential and/or aspect. Barnes and Malone (2000) analyze progressive aspect as deriving from a gerund morpheme suffixed to the main verb-in collaboration with an auxiliary that has an evidential suffixed to it. They do not mention any aspectual role of the evidential. And while it may be true that at least some Tuyuca evidentials originated from aspectual morphemes, they do not appear to play any aspectual role. In Tuyuca, the morphosyntactic distribution for evidentials and progressive and perfective aspect is completely different. As I will show, an auxiliary is needed to construct a progressive or perfective. In this case, the aspect is suffixed directly to the verb stem. As I argue in the rest of this chapter, partly following Barnes and Malone (2000), the morphosyntactic construction of progressive and perfective aspect does not support evidence that the aspectual interpretations are in fact due to the presence of an evidential. 4 . 2 . 1 . Progressive aspect in Tuyuca A more precise interlinear glossing of (18) and (19) can be seen in (22) and (23) but are not my final gloss. (22) bue-go tif-a do-EVD.PRES.VIS.'(I am) studying.' (23) diiga ape-gi til-i play-MSG do-vis.'(He) is playing soccer.' or verb--morphemes. T 4.2.1. Tuyuca study-FSG dO-EVD.PREs.vrs.1/2 gt tii-soccer plaY-MsG dO-EVD.PRES.VIS.3MSG 38 The glosses above are consistent with other interpretations given by Barnes in various publications. They are also generally consistent with Karn's (1976, 1979) glosses of Tuyuca texts and with glosses found cross-linguistically within the Tukanoan family (e.g., Cook and Gralow 2001, Strom 1992, Stenzel 2004). I have separated what is called the nominalizer and/or gerund by Barnes from the lexical verb. These are the dependent verb agreement suffixes of Table 7, also discussed in sections 3.6, 3.7, and 3.8. They only appear on the lexical verb when it is a dependent verb: that is, when the main lexical verb is accompanied by an auxiliary the latter takes the fully specificed agreement morph (gender, person, and number) and the evidential, whereas the dependent verb takes an underspecificed agreement suffix (gender, number) shown in Table 7; both dependent verb and auxiliary agree with the head noun of the clause-but they differ in the specification of this agreement. The distribution of the morphemes in Table 7 is restricted to the specific morphosyntactic environment of a dependent verb. Glossing the auxiliary as 'do' is consistent with BM glosses of this morpheme as Spanish hacer 'to do/make.' Partly following BM, I argue that the progressive interpretation is due to the presence of the auxiliary in combination with a morpheme from Table 7 suffixed to the lexical verb. However, I will eventually argue that one can isolate the agreement markers, in the form of the final vowel, from the rest of the morphemes in Table 7. BM briefly discusses the progressive aspect and give a morphological template I translation beneath it in (25). (24) [tema verbal + sufijo de genero/numero/animado] + [/// + evidencial+tiempo] (25) [verb root + gender/number/animate suffix] + [tii + evidential+tense] Kam's r dependent 3.6,3.7, clause-tii 8M ofthis 8M, r tii r for it, which r modify by adding tense (tiempo) in (24) and providing an English [tii 39 BM state that, in general, aspect in Tuyuca is formed in three ways: (a) from a combination of the gerund (seen in Table 9) and auxiliary verb, (b) an independent aspect morpheme between verb root and evidential, or (c) subordination suffixes. El aspecto progressivo se forma por medio de un gerundio mas el verbo auxiliar tii - 'hacer' flexionado para la categoria evidencial. El gerundio esta integrado por un tema verbal mas un sufijo que indica el genero, el numero y el condicion animada o inanimada del sujeto. (Barnes and Malone 2000: 440) What is crucial here is the first strategy (a): combining the gerund and auxiliary. BM go on to say that the gerund is integrated with the verb root and the suffix that indicates gender, number, and person as seen in the template in (24). BM give a paradigm for the gerund, which T adapt in Table 9. Notice that Table 9 corresponds equivalently with Table 10 and partially with Table (alo see Tables 6 and 7 for more detailed comparison). 1 4 It also partially corresponds with tense evidentials from Table 8, which I will deal with later. With respect to the correspondence of gerunds in Table 9 with the morphs in Tables 10 and 11, it seems that one set of morphemes is responsible for the general function of making verbs more nominal; assuming one interprets gerunds, nominalized verbs, and progressive/perfective aspect as less 'verby.1 most of Barnes' work the morphemes in Tables 6, 7, 9,10 and are generally glossed as agreement morphemes, though she recognizes their aspectual, nominalizer, and gerund functions. The goal here is to see if deverbalizing functions can be isolated to precisely these grammatical functions. So far, it seems that progressive aspect, gerunds 1 3 El combination verbo o siguen al tema verbal y preceden a la termination evidencial, o a los sufijos de subordinacion" (BM: 442). 1 4 BM subordinacion"-suffixes relative clauses-can be used for constructing aspect in Tuyuca. The morphemes in Tables 6 and 11 are also employed in making relative clauses and have almost the same form as those in Tables 7, 10 and 8. fonned suffixes. 13 EI fonna EI condici6n 0 8M adapt in Table 9. Notice that Table 9 corresponds equivalently with 11 14 PRESENT NONVISUAL r II, verby.' In 9, I 0 11 particular morphemes, which may warrant an interlinear glossing that reflects more 13 "EI aspecto se indica ya sea por una combinaci6n del gerundio y un verba auxiliary, 0 por sufijos que terminaci6n a subordinaci6n" 8M: 14 Interestingly, 8M note that "a los sufijos de subordinaci6n"-suffixes of subordination used to construct clauses-II 40 Table 9 Gerunds Class Number Gender Animate Singular Masculine -go Feminine Plural -ra Inanimate -ro Table 10 Present tense agreement for verbs DEPENDENT VERB SUFFIXES Animate Inanimate Singu Masculine ar Feminine Plural -ro PRESENT -go -ra Table 11 Present tense nominalizer agreement NOMINALIZERS/DEVERBALIZERS Animate Inanimate Singular Plural Non- Place Masculine Feminine count PRESENT nT -go, rjo -ra, ra -re -ro, -ro -g+ Singular I I -g+ I-II I -g+, I)r 1)0 -ra 41 and nominalizers are easily segmentable into an isolated agreement morpheme suffixed to various aspect, classifier, gerund, and nominalizer morphemes. For example, I see no reason why -gi cannot have the four glosses in (28), one as aspect, one as classifier, one as gerund and one as nominalizer-it may be the case that gerund and nominalizer are the same, but I treat them differently to see if they have different distributional properties. (28) g4 b. g4 c. g-+ d. g4 ASP-MSG GER-MSG NR-MSG Instead of progressive aspect being built from the gerund, now /-i/ can be interpreted as an aspectual morpheme with agreement when it is in the correct environment. This is similar to English: we do not say that progressive aspect comes from the composition of an auxiliary with a gerund, or that gerund comes from deleting the auxiliary of a progressive. Instead, English -ing functions as either progressive or gerund in a particular morphosyntactic environment. The environments for the interpretations in (28) will be made more explicit in later sections. I turn now to a more precise morphological analysis of (20) and (21). 4.2.2. Perfective aspect in Tuyuca The perfective aspect is the term for an event that has had a clear 'end result.' Given the similarity of the 'end result' constructions to other aspectual constructions as described by BM and as seen in I interpret the 'end result' meaning in and suffixed nominalizer-a. g-i g-t g-i g-i CLF-MSG I-g-il correct tum (22)-(25) (20) 42 (21) as the perfective aspect. More explicit versions of (20) and (21) can be seen in (29) ri-g+ nTT-wi gO-RSLT-MSG be-EVD.PST.3MSG *He ri-g* nTT-wi field burn-RSLT-MSG be-EVD.PST.3MSG -ri is suffixed to the auxiliary verb nil." In other glosses she never shows the ri/ 21)). /-ri/ part of the deverbal agreement marker. Although she mentions that an end result nil, 23)). separated it from /-g+/, which has the aspectual interpretation here. It is reasonable that interpreted as progressive would yield a perfective reading; adding the glossing conventions in (28) to (30) gives (31). and (30) but are not my final gloss. (29) waa-n-gt nii-wi gO-RSLT-MSG be-EvD.psT.3MsG 'He went.' (30) wese soe-gi nii-wi field burn-RsL T -MSG be-EvD.psT.3MSG 'He burned his field.' Again, the glosses are consistent with glosses made in Barnes' other work and with other Tukanoan works. Barnes (1984: 259) says that "In this construction, the main verb is suffixed by the 'resultative' morpheme and gender-number morpheme. The evidential nIT." RSLT interpretation of /-ril (see Table 7 and examples (20) and (21 ». Instead, I-ril is glossed as interpretation is assigned to constructions in (29) and (30) she does not communicate this through her glosses. Lastly, there is a different auxiliary, nIT, used in these sentences (compare tii in (22) and (23». I have explicitly shown in my glossing of the data the resultative morpheme and I-gil, agglutination of a resultative morpheme followed by an aspectual morpheme that can be 43 (31) wese soe-ri-g4 nTT-wi RSLT-EVD.PST.3MSG 'He burned his field.' Here I still use ASP for the morpheme that appears to yield a consistent perfective interpretation in the correct morphosyntactic environment; I will continue to do so in order to keep the analysis from making too strong a claim. Only native speaker judgments can verify the claim that l-g-l is progressive or perfective in these constructions; but it seems clear that it is some form of aspectual marker. Under these conventions, (29) would look like (32). Barnes also shows that /-a-/ is the recent past morpheme that can agglutinate to the perfective aspect, but fails to explicitly gloss it in some instances (see Table 6) while glossing it in others. In (33) and (34) I show what some of the data looks like with my glossing conventions. (32) waa-ri-g-i nTT-wi EVD.PST.3MSG 'He went.' (33) Pade-ri-a-ri-g-o yii mako nTT-yo WOrk-NEG-RSLT-PST.1SG EVD.APR.PST.3FSG 'The one who did not want to work is my daughter.' (34) Yaa-ri-a-ri-g4 nTT-a-wT RSLT-PST-3MSG 'Evidently he did not eat (the food is still here).' Literally: 'Evidently he was a non-eating one' (adapted from Barnes 1994: 333-4) Barnes includes the extra translation in (34) to highlight the deverbalizing nature of the morphology. One difference between my gloss of (33) and (34) and Barnes' is that 3 I) ri-g-i- nit-wi field burn-RsLT-ASP-MSG be-EvD.psT.3MSG /-g-/ constructions: g-i- nit-wi gO-RSLT-ASP-MSG be-EvD.psT.3MsG yit mak6 nIl-work-NEG-REC.PST-RSL T-ASP-PST.FSG I SG daughter be-EvD.APR.PST.3FSG The g-i nii-wi eat-NEG-REC.PST-RSL T-ASP-MSG be-REC.PST -EVD.PST.VIS.3MSG of(44 1 5 share two categories in common with one phoneme: inanimate and/or plural animate appears to be coded by initial l-r-l where the other categories have initial l-g-l. The future tense in all categories has a l-l where l-g-l and l-x-l occur but it appears to have a significantly different form from the other tenses, suggesting a different historical path. I explicitly separate the resultative /-ri/, which she recognizes but rarely separates, and I separate the agreement marker from l-g-l. This latter separation results in the necessity of assigning a morphological label to l-g-l. My argument is that in this morphosyntactic environment the interpretation for l-g-l is deverbalizing and aspectual; whereas, in another environment it may function as a nominalizer or a gerund. What my analysis implies is a more detailed syntactic structure, given the assumptions of the Mirror Principle. The different functional interpretations of l-g-l in these examples can be unified by a general theme of deverbalization: making a verb less 'verby' or more nominal. The glosses in (33) and (34) can be cumbersome but I believe they are more accurate and reflect the actual function of the morphology. However, the function of l-l must be consistent and predictable in order for the more intricate glossing to be warranted. I show in the next section, 4.3, that the environments for the different functions are in fact highly predictable. Lastly, I have so far been dealing only with the l-l element in present, past, and recent past tenses. I will continue to do so even though the full paradigms of Tables 6 and 7 include morphemes that vary from l-l. If a consistent analysis for l-l can be worked out then this analysis can be extended to account for the varying forms found in Tables 6 and 7; specifically the plural animate and inanimate categories, as well as the future tense, which seems to be entirely different from the other tenses. 15 I-I-I. I-I. I-I I-I I-g-I I-g-I I-g-I. I-g-I IS 15 The direction for dealing with the "variant" forms of the paradigms should be fairly obvious: they all initial/-I initial/-I. I-d-I I-I I-r-I 45 4.3.Specifying predictable environments The most obvious predictable environments for interpreting l-g-l are the aspectual (progressive and perfective) ones. What appears to be the most highly restricted environment of the two is the perfective, which requires an obligatory resultative morpheme to directly precede it. I have found no data in which a perfective interpretation of the clause occurs without the resultative morpheme. One might assume that the resultative plus aspectual l-gl formed the perfective. But there is more. The auxiliary riii 'be,' compared to tii 'do/make,' must also occur in tandem with the resultative marked aspect. Applying the glossing conventions given in (28) to (25) we can derive (35) for the perfective formula. The progressive data show that the auxiliary co-occurring with aspectual l-g-l is tii 'do/make.' On analogy to (35) we get (36) for the progressive. (35) PERFECTIVE ASPECT: [VSTEM + l-g-l gender/number/animate suffix] [nTi evidential] PROGRESSIVE ASPECT: [VSTEM + l-g-l gender/number/animate suffix] [tii evidential] The formulas in (35) and (36) appear to hold for all the data I have seen in all of Barnes' works (see references), as well as in Karn (1976, 1979). Progressive and perfective aspect are predictable. Aspectual l-g-l has a consistent analysis as an independent morpheme-separate from resultative and subject agreement. environments /-g-/ ofthe /-g/ nTi Iii /-g-/ (35) ASPECT: [VSTEM + /-g-/ + + nit + (36) ASPECT: [VSTEM+ /-g-/ + + + evidential] T (\ 976, \ 979). /-g-/ morpheme-separate 1 6 gerund [NP + + l-g-l + gerund-suffixed case-was (1990) analysis of animate classifiers resulted in my change of mind. 4 3 ) l-g-l. Ko6-ya-g4 k+t-ya-g-t-koro nTT-T.16 3FSG-SG-CLF.AN-MSG 3MSG-GEN.CLF.AN-MSG-alike be-EVD.VIS-3MSG Ko6-ya-g4 yti-SG-ACC SC0ld-EVD.PST.VIS-3MSG mTI-ya-g4 2SG-GEN. SG-CLF.AN-MSG ya-g4 GEN.SG-CLF.AN-MSG bird,...)' 4 1 ) mw-ya-hen,...)' m+t-ya-46 4.4.Animate possessive noun classifier is not a gerund This section is a very late addition. I had originally argued that there was evidence for a possessive gerund in Tuyuca. Its environment was defined as + GEN + /-g-/ + gender/number suffix]. However, explaining how an NP, usually a pronoun, came to acquire a gerund-suffixed to a genitive case-was problematic. The combination of only recently reading Baker's (2005) paper on gerunds and taking another look at Barnes' (37) - ( 43) show a genitive case marker (underlined) directly preceding /-g-/. The genitive has the singular value for number, even in instances of referential plurality; see (42). (37) Koo-YJ!-g-i k~YJ!-g-i-k6r6 nrt-i.16 3 FSG-GEN. SG-CL F .AN -MSG 3 MSG-GEN .SG-CLF .AN -MSG-al ike be-EvD .PRES. VIs-3 MSG 'Her (animal) is the same as his (animal).' (38) Koo-YJ!-g-i diyi yit-re tuti-w-i. 3FSG-GEN.SG-CLF.AN-MSG dog 1 SG-ACC scold-EVD.PST.VIS-3MSG 'Her dog barked at me.' (39) mii-YJ!-g-i SG-CLF .AN -'your (animal)' (40) bariya-YJ!-g-i mary-GEN .SG-CLF .!\N-MSG 'Mary's male creature (dog, bird, ... )' ( 41) mti-ya-g-o 2SG-GEN.SG-CLF.AN-FSG 'your female creature (dog, hen, ... )' (42) mti-YJ!-r-a 2SG-GEN.SG-CLF.AN-PL 16 This is not an auxiliary. 47 'your creatures' (adapted from Barnes 1990, 1994) (43) ki+, poterimacara-ya-g-i, kit, Coamaci, GEN.SG-CLF.AN-MSG, Coamaci, 'he, Coamaci, belonged to us Indian people' (adapted from Karn 1976: 1) Tuyuca possessives come in three varieties: (44a) a noun/pronoun followed by the genitive singular /-ya-/ or genitive plural /-ye-/ plus a noun classifier, (44b) just the genitive singular plus noun classifier, or (44c) noun/pronoun plus noun - where the second noun is a kinship term. In all cases the first noun is the possessor and the second possessed; (glosses are modified from BM to conform with LGR and are translated into English by me). (44) paki-ya-wi ya-wi c. yii-paki CLF:SG-father 'father's house/ house of father' 'my house' 'my father' (adapted from Barnes and Malone 2000: 446) (37) - (43) most closely resemble (44a), except in (37) - (43) the classifier is a general animate classifier with subject agreement marked; see Table 12. Tukanoan languages do not typically express this kind of restricted subject agreement in the noun classifier system (see Silva and Bowles 2007; see Gomez-Imbert 2007 for arguments that classifiers do constitute agreement in Tatuyo). kit, poterimacani-~-g-i, kH, Coamad, 3MSG, indian.people-GEN.SG-CLF.AN-MSG, 3MSG, Coamact, Coamact, Kam I) a. pakt-b. father-GEN.SG-CLF:house GEN.SG-cLF:house ytt-pakt 1 SG-father classifier 12'7 Animate classifiers DEPENDENT VERB SUFFIXES Animate Singu Masculine ar Feminine Plural PRESENT -& -go -ra PAST -rig+ -rigo -rira FUTURE 4d+ -odo -adara 1 7 morphs in Table 7. 48 Table 1217 Singular -gt rigt rtgo -KIt 17 Notice Table 12 is a subset of nominalizer morphs in Table 6 and dependent verb agreement and aspect 49 There is good reason to believe that Tukanoan noun classifier systems are developing into a noun class system similar to those found in Bantu languages (Grinevald and Seifart 2004). Barnes (1990: 289) characterizes the classifier forms in (37) - (43) as tensed animate noun classifiers; -gi = present tense masculine singular, and -go presenttense feminine singular, and -ra - present tense plural. Barnes goes on to say that when these noun classifiers are suffixed to verbs they function as nominalizers. I argue that l-g-l in Table 12 and (37) - (43) functions as an animate possessive noun classifier. The distribution of the possessive forms created from this morphology is restricted by the typical distribution of possessive nouns or noun phrases they attach to. The environment for morphemes in Table 10 is (45). ANIMATE POSSESSIVE CLASSIFIER: [NP + l-g-l gender/number suffix] 4.5.The gerund There is, instead of (45), another environment for the gerund: as a modifier of a stative verb of the type glossed 'be.bad' or 'be.big.' The environment includes a copula riil and a stative verb where the gerund can function as the predicate adjective, as in (46)- (47). An interesting example showing how the predicate adjective gerund (in bold) and the animate classifier (underlined) interact is shown in (48). (46) pai-g-i be.'(is) big' gf = = singular. -ra = present tense plural. Barnes goes on to say that /-g-/ classifier. lOis (45) CLASSIFIER: + GEN.SG + /-g-/ + suffix] is. 45). niT be.big-GER-MSG (47) Nana-q-* nTT-ri-ht-T be.GER-MSG APR-kame-rf-a-y-i. NEG-REC-PST-EVD.APR-'Apparently he is not bad, apparently he did not reciprocate (wound for wound).' (48) Kff-ya-g4 mff-ya-g4 nemd-ro pai-g-i AN-MSG big-GER-MSG nTT-T. EVD.PRES.'His animal is bigger than your animal.' Here, nii-i is not an auxiliary andpai-g-4 functions as the predicate adjective (see (49)), while Ktt-ya-g-i and mtt-yd-g-4 show the possessive classifier of (45). The formula for the predicate adjective gerund is (49). also include a formula for nominalizer l-g-l in (50), for which some data was provided in (15)-(17). The nominalizers seem, initially, to have a different distribution than the gerunds, requiring no stative verb. For this reason I treat nominalizers separately from gerunds. I leave open the idea that gerund and nominalizer are the same thing. (49) PREDICA TE ADJECTIVE GERUND: [VSTATIVE + l-g-l + gender/number/animate suffix] (50) NOMINALIZER: [VROOT + l-g-l gender/number/animate suffix] 4.6.Conclusion The l-g-l class of morphemes shares the consistent and unified function of "deverbalizing" a verb (assuming progressive/perfect aspect, gerunds, and nominalizers IJ-i nTI-hi-i be .bad-GER -MSG be-NEG-EVD.PRES.APR -3MSG reciprocate-NEG-REC-PST-EVD.PST.APR-3MSG 50 Kti-g:i mti-g:i nemo-3MSG-GEN.SG-CLF.AN-MSG 2SG-GEN.SG-CLF.AN-MSG more-ADVR be.big-GER-MSG nii-i. be-EvD.PRES.VIS-3MSG n'ii-iis g-i Kli-ya-g-i m#.ya-g-i I for /-g-/ (49) PREDlCATEADJECTlVEGERUND: [VSTATIVE + I-g-/ + gender/number/animate suffix] (50) NOMINALIZER: [VROOT + /-g-/ + suffix] /-g-/ are deverbalized forms). This class receives specific interpretations based on restricted morphosyntactic environments. I have separated the agreement markers from the l-g-l class of morphemes in order to more precisely isolate their functions. But it seems clear that the gender/number/animate suffixes used to mark agreement are obligatory on all these forms. I believe I have given sufficient evidence that interlinear glosses in Tuyuca need to reflect the precise functions of the morphemes in Tables 6, 7, 8, and 9-and that isolating l-g-l results in a more detailed and consistent morphology for Tuyuca. Barnes herself recognizes virtually all of the functions dealt with here, but is inconsistent in her glossing and analysis, typically only indicating the agreement values or using the label nominalizer for the l-g-l class morphemes. I have argued that the "nominalizer" label is not precise enough. It does not capture the aspectual interpretations, nor does it indicate the possessive animate classifier, nor the predicate adjective gerund. This makes my analysis appealing because it is easy to test. Given the predicted interpretations based on the morphosyntactic environments I have proposed here, one need only find instances where the l-g-l class morpheme behaves counter to my predictions. For convenience I group the predicted environments together in Table 13. Lastly, taking into account the typological similarity between the Tukanoan languages, I believe my arguments necessitate another look at agreement in other Tukanoan languages based on the results in this chapter and in Table 13. 51 /-g-/ 9-and /-g-/ I-g-/ /-I 52 Table 13 Predicted environments for "g" class morphemes (35) PROGRESSIVE ASPECT. [VSTEM + + + + (36) PERFECTIVE ASPECT. [VSTEM RSLT + l-g-l gender/number/animate suffix] [nn evidential] ANIMATE POSSESSIVE CLASSIFIER: [NP + l-g-l gender/number suffix] (49) PREDICATE ADJECTIVE GERUND: l-g-l + suffix] (50) NOMINALIZER: [VROOT + l-g-l + suffix] ASPECT: [VSTEM + /-g-/ + gender/number/animate suffix] + [tii + evidential] P ERFECTIVE ASPECT: [VSTEM + + /-g-/ + + no + P OSSESSIVE CLASSIFIER: (45) + GEN.SG + /-g-/ + suffix] G ERUND: [VSTATIVE + /-g-/ + gender/number/animate suffix] NOMINA LlZER: [VROOT + /-g-/ + gender/number/animate suffix] 5. AGREEMENT AND EVIDENTIALS 5.1 .Against conflating agreement and evidentials The goal of this chapter is to survey some theoretical consequences of separating agreement morphology from evidential morphology in Tuyuca. I discuss what evidentials are, their grammatical category, types of evidentials, and their relation to tense, aspect, person, and subject agreement. The same general premise that worked in Chapter 4 also works here, i.e., that there is a systematic and predictable pattern of final vowels that features-generally -i masculine, -o - feminine, and -a plural-and that these agreement markers are of the have on current syntactic generative theory. I try to limit dependence on technical terminology in order to communicate the basic ideas and keep the general typological format of the thesis. But an understanding of the basic tenets of X-bar theory, Principles-and- Parameters theories, Distributed Morphology theory, and Minimalist Program attitudes is helpful; see Carnie (2002, 2008), Chomsky (1995, 2001), Epstein and Seeley (2007), Hornstein, Nunes, and Grohmann (2005), Lasnik and Uriagereka with Boeckx (2005), Pesetsky and Torrego (2002), Radford (2006), and Uriagereka (1998). By far the most technical sections are 5.4, 5.4.1 and 5.4.2. 5.I.evidentials correlates with the distribution of person, gender, and number features-most generally-= -0 = = plural-isolable from the rest ofthe morphology. The analysis of separating agreement from evidentials is straightforward and simple; it requires little argumentation. Instead, the bulk of this chapter deals with the significant impact Tuyuca evidential morphology may Principles­and- 200 I), 54 5.1.1. What are evidential morphemes? Evidential morphs are inflectional morphemes that refer to the source of evidence of the speaker's proposition. Different analyses of evidentials group them under different grammatical categories. The first kind of analysis makes use of the distinction between evidentials as Mood (Cinque 1999) or Modality (Chung 2005, Matthewson et al. 2006). Other accounts analyze the evidential as either fused with or relating to the category Aspect (Johanson and Utas 2000, Malone 1988, Stenzel 2004, Sumbatova 1999). The general typological conclusions about evidential morphology and evidentiality in 1 Q general, seem to be that it is either an independent category (Aikhenvald 2006, Speas 2004a, 2004b, 2007) or part of the Tense-Aspect-Mood system, and that individual languages express evidentiality in a variety of ways (see footnote 15). This variety is constrained and principled. Speas (2004b, 2007) argues that it is reasonable to conclude that the concept of "evidence" is itself not a grammatical primitive-in the sense that the types of evidence coded in evidential morphemes is limited to a few basic types that can be organized along a direct and indirect partition evidentiality. language; as seen in these English examples. a. I see that John washes his car. b. I see John washing his car. properties of the verb see, interact to give different interpretations of the evidential base of the proposition. In (a) the intuitive interpretation is that the speaker infers from some indirect kind of evidence - contrary to directly watching John wash his car - that John in fact washes his car. In (a), the evidence may be indirect, in that the speaker always notices that John's car is clean. Based on this, the speaker infers that John washes his car, but the proposition does not entail that the agent of washes is actually John; it may be somebody John hires. In (b), the interpretation is that the speaker has direct visual evidence that John is washing his car, which entails that John is the agent of washing. The difference between (a) and (b) can also be reduced to the difference of the verb see; in (a) it has an indirect base and can be substituted by other verbs such as understand, know, hear, and even perhaps feel. This is only possible for (b) with the verb hear, and perhaps feel. But clearly, English has no evidential morpheme equivalent to the inflectional type seen in Tuyuca (§3.8). See Drubig (2001) for an analysis of English and German epistemic modality as expressing degrees of evidentiality. morphemes? intlectional ofthe different general, 18 seem to be that it is either an independent category (Aikhenvald 2006, Speas 2004a, 2004b, 2007) or part of the Tense-Aspect-Mood system, and that individual languages express evidentiality in a variety of ways (see footnote 15). This variety is constrained and principled. Speas (2004b, 2007) argues that it is reasonable to conclude that the concept of "evidence" is itself not a grammatical primitive-in the sense that the types of evidence coded in evidential morphemes is limited to a few basic types that can be organized along a direct and indirect partition IH There is a difference between evidential morphemes and eVidentiality. The latter can be expressed in any In (a) and (b) the complex interaction between the tense and aspect, as well as the different lexical b). to the difference of the verb see; in (a) it has an indirect base and can be substituted by other verbs such as understand, knolV, hear, and even perhaps feel. This is only possible for (b) with the verb hear, and perhaps feel. But clearly, English has no evidential morpheme equivalent to the inflectional type seen in Tuyuca (§3.8). See Drubig (2001) for an analysis of English and German epistemic modality as expressing degrees of evidentiality. 55 (Faller 2001, Willet 1988). Although there are disputes about the number of types of evidential, they tend to fall into five or seven basic types: Visual, Nonvisual, Inferred, Hearsay, Secondhand, Apparent, and Assumed. Some of these terms may overlap, or other people may use different terminology, but the general typological constraints seem to define a very specific boundary. For example, one does not see any type of evidential morpheme that refers to evidence based on religious ceremony, religious text, proverbial sayings, or legal discourse; though these are reasonable pragmatic sources for basing one's evidence about propositions. The conclusion seems to be that the semantic domain of evidentials is limited to the sensory-perception. Dreams, visions, feelings, and general objects of imagination, however, are not coded in evidential morphemes - unless one counts rational inference as a feeling or a vision. Only sensory evidence based on sight (or lack of), hearing (or lack of), and possibly touching or smell; in the extreme perhaps one could argue for taste. Even in the extreme, one finds a highly constrained boundary for the types of evidence used, and thus, can conclude that "evidence" is not a grammatical primitive. The type of "evidence" coded in evidential morphemes is cross-linguistically constrained and has a principled, albeit poorly understood, distribution. 5.1.2. Evidentials with inherent 1s t person features? The interaction between evidential morphemes and 1s t person pronouns is well-attested (Aikhenvald 2006). It is clear that in cases where the speaker and grammatical subject are the same, and no overt 1s t person marker is present, the evidential still refers to the source of evidence for the speaker. In this way one might argue that evidential evidentials, cross­linguistically 15t features? 15t well­attested 15t Tn 56 morphemes do in fact have unpronounced 1s t person pronominal features, but the kinds of evidence and datum that should be used to show this is still not clear. What is without dispute is that evidential morphemes are inflectional, and because of this can inflect for subject agreement, including 1s t person. Furthermore, evidential morphemes are known to be expressed in some languages as portmanteau forms that simultaneously code source of evidence and tense. Data for portmanteau forms that simultaneously code source of evidence and subject agreement are weakly attested (see the morphs under the 1/2 and VISUAL categories in Table 5 and discussion of them in §3.8.2 and §5.5.2). 5.1.3. Fused tense-evidentials and inflection for agreement A survey of the World Atlas of Language Structures (Haspelmath et al. 2005) for languages with tense-evidentials resulted in a count of twenty-four. To this number I added five additional languages that were not in the WALS database, for which I provide the source references; results are shown in Table 14 (see also Bowles to appear). OV and VO refer to word order and <D-F refers to whether or not subject agreement features (gender, number, and person) are inflected on the tense-evidential. The typological origin (i.e., grammaticalization or reanalysis) of evidentials in general appears to be verbal in nature. For example, verbs with the meaning of'say' can be become hearsay or quotative evidentials (Campbell 2004 and references therein). The origin of tense-evidentials, on the other hand, isn't always the fusion of an evidential morpheme with a tense morpheme. For example, Fleck (2007) shows the origin of tense-evidentials in Matses to be nominalizers. As discussed in section 4.2, Harris and un pronounced I st pronom inal 1st 112 agreement W ALS database. OVand I>-intlected of 'say' oftense-tense­evidentials 57 Table 14 19 Basic typology for tense-evidential '' ov v o (D-F FAMILY 1 Abkhaz / N.West Caucasian Cinque (1999: 155) 2 Armenian / Armenian 3 Barasano Tukanoan 4 Bulgarian 0 Slavic 5 Carapana Tukanoan 6 Carib Cariban 7 Chechen Nakh-Daghestanian 8 Ekari Trans-New Guinea 9 Estonian 0 Finno-Ugric Campbell (1991) 10 Evenki Altaic Gagauz Altaic 12 ?Georgian Kartvelian Bejar (2001) 13 Godoberi Nakh-Daghestanian 14 Haidi Haida 15 Hunzib Nakh-Daghestanian 16 Ingush Nakh-Daghestanian 17 Khowar Indie 18 Kurmanji Iranian 19 Ladakhi Sino-Tibetan 20 Laz Kartvelian 21 Matses Panoan Fleck 22 Persian Iranian 23 Salar Altaic 24 Sherpa Sino-Tibetan Woodbury (1986) Tariana Arawakan 26 Tucano Tukanoan 27 Turkish Altaic 28 Tuyuca • Tukanoan 29 Yakut Altaic 1 9 VO) So far, the only explanation for why head-final languages would be more likely to fuse tense and evidential morphemes is that the highly suffixal nature of head-final languages could provide the conditions under which fusion is more likely; Lyle Campbell (p.c.) and Mauricio Mixco (p.c). This is as far as I pursue the matter here. 2 0 tense-evidential 2 1 free... SOV." Basic typology for tense-evidential 19,20 NAME OV VO <fJ-LANGUAGE F AMIL Y REFERENCE Abkhaz ../ N.West Caucasian Cinque Annenian ../ Annenian ../ YES Tukanoan Bulgarian [{] Slavic Carapana NO INFO Tukanoan ../ Cariban ../ ../ [{] to ../ 11 NO INFO ../ YES 200 I ) Godoberi NO INFO ../ Haida ../ N akh-Daghestan ian ../ ../ Indic NO INFO Iranian Ladakhi ../ Laz ../LI ../ YES (2007) Persian ../ Iranian Salar NO INFO ../ Sino-Tibetan 25 ../ Arawakan ../ YES Tukanoan ../ ../ YES Tukanoan ../ Altaic 19 Word order (OV and YO) is included because during the sampling I noticed that an overwhelming majority of tense-evidential languages were head-final. The results of this sample are by no means definitive, but they do suggest that there may be a correlation between tense-evidentiality and head-finality. c.). 20 Table 14 is not meant to be exhaustive. That is, virtually all the Tukanoan languages have tense­evidential morphemes (only 4 of the 20 are shown). The Altaic, Nakh-Daghestanian, and Kartvelian families probably have more languages that could be shown. 21 WALS has no information on Laz, but Harris and Campbell (1995: 216) state that "In Laz word order is relatively free ... but the unmarked order is SOY." 58 Campbell (1995) and Campbell (1991) show that Estonian tense-evidentials come from reanalysis of fused tense-participle endings in subordinate clauses. The Turkish resultative /mis/ appears to have been reanalyzed as an indirect past tense-evidential with finite verbs, while nonfinite verbs still yield the resultative interpretation (Csato 2000, Johanson 2000, Shroeder 2000). Finally, Malone (1988: 139) states for Tuyuca that "Nonvisual paradigms appear to have developed from a progressive (or other) aspectual gerundial construction... [and]'+/- direct' paradigms appear to have developed at some later stage from an old perfect construction" (see also §4.2). Interestingly, if one compares the present tense nonvisual evidentials (Tables 8) to my analysis of perfective and progressive forms (Table 13), as well as to the present tense forms of Barnes' nominalizers, dependent verb subject agreement markers, and gerunds (Tables 6, 7, 9, 10, and 11) there is a striking equivalence in form. My analysis, then, could corroborate Malone's claim about the aspectual and gerundial origin of evidentials (with obvious extension and phonological change in the paradigm). Aspectual morphemes from which evidentials were derived still have their aspectual interpretation in restricted morphosyntactic environments. This is similar to the Turkish or Estonian examples of reanalysis, where one form was reanalyzed but still retained its original meaning. 5.1.4. E v i d e n t i a l l y as agreement Another possibility is to consider agreement morphemes as sub-classifying evidentiality among one of its other features, i.e. gender, person, and number. But this is an unattested feature of agreement (Corbett 1991, 2000, 2003, 2006, Comrie 1981, Siewerska 2004, Song 2001). Deciding what kind of empirical evidence would support /mI~/ construction ... and] '+/- 7,9, II) Evidentiality agreement sub-59 such a claim is highly problematic. A more reasonable conjecture is to assume all evidential morphemes can agree with subjects without having to show overt subject agreement inflection. The approach that assumes an unpronounced 1s t person feature inherent to evidentials (§5.1.2) assumes a deep interaction between evidentials and agreement. It does not classify evidential morphemes as agreement morphemes themselves. But it does recognize, based on empirical data that shows well-attested interactions between evidentials and 1s t person, that the category Evidential shares some common behavior with the category Tense in terms of deixis and interactions with pronouns (see Speas 2004a, 2004b, 2007). Tuyuca evidentials do not primarily function to mark agreement with the subject of the clause, just as tense does not. Instead, evidentials code the deictic source of evidence for speaker proposition just as Tense codes the deictic source of time for an event and speech act. Technically, in this approach evidence is not a grammatical primitive, and so evidentials cannot actually "code the source of evidence." Instead, they code the source of situation for a discursive and evaluative event (Speas 2007). Here, evidentiality can be distinguished from mood and modality in the following way. If a clear distinction between mood and modality can be defined as the difference between variables ranging over the speaker's opinion/attitude of the world (modality), and variables ranging over the world itself (mood), then evidentiality can be defined as variables ranging over the situations in the world. Tense, to draw a parallel, is then defined as variables ranging over times and/or events. One conjecture is that the category Evidential may agree with other categories in terms of worlds, events, times, and situations-and perhaps 1s t person. Either way, the gender/number/person endings on tense-evidentials in Tuyuca are marking agreement I st I st function and situations-I st 60 with the subject and do not play a role in evidentiality (see §3.8 and §5.3 for detailed arguments). 5.1.5. Concluding remark: Conflating agreement and evidentials The general conclusion that can be drawn is that subject agreement (gender, number, and person) is separate from evidentiality. By factoring out the common agreement morphs one can easily see the variation in form between present and past tense, as well as between types of evidential. The focus here will be on those evidentials with 3 r d person values. Table 15 is an application of the general premise of this thesis to evidential morphology (see Chapter 4 for arguments concerning other parts of the morphology). One benefit from separating agreement from evidential is that one can easily see the variation in morphology between the two tenses. The agreement morphemes do not vary from present to past tense; but the evidential does. I take this as straightforward evidence that tense is fused with the evidential and not with agreement. Although the notational innovation for separating the agreement marker from the evidential is not very interesting, the theoretical issues arising from it are. Namely, where universal syntactic hierarchies of functional structure have been proposed the usual order is Evidential over Tense (E > T) (Cinque 1999; see Appendix D). But analysis of the interaction between agreement and evidentials in Tuyuca, assuming Minimalist constraints on agreement and clause structure as well as the basic intuition that TP should constrain tense and subject-verb agreement, results in a surprising reversal of two functional categories, placing Evidential under Tense (T > E). I also attribute this surprising result partly to the fused nature of the evidential morphology in terms of playa Conclud ing evidentials rd E> Table 15 Revised Tuyuca paradigm VIS Nvis APR SCD ASM PAST OTHER(1/2) -W-+ -t4 yir-hTy-3MSG w-i -t-i -y-i -yig4 -hTy-i 3FSG -w-o -t-o -y-o -yig-o hTy-3PL -t-a -y-a -hTy-a PRESENT OTHER(1/2) -a -g-a 3MSG -i i -k-i 3FSG -hT-o 3PL -htr-a a 61 NVIS APR SeD ASM PAST OTHER( 1/2) W-t t-t -y-u -ylr-o -hiy-u 3MSG -W-I I ylg-+ hiy-3FSG -hiy-o 3PL -w-a a -yir-a hiy-PRESENT OTHER( 1/2) -k-u I -g-I -hi-i 3FSG -y-o -g-o hi--k-o 3PL -y-a -g-a hir--ku-a defining sisterhood relations that license fusion. Tuyuca seems to provide evidence that other languages with fused tense-evidentials may also require the T E order. 5.2. Basic issues In Chapter 4 I assumed that gender/number/person morphology suffixed to verbs was subject agreement and that Tuyuca clauses do not need to express an overt subject. Reference to various types of definite or indefinite subjects is made through subject agreement marked on the verb stem (see Alexiadou and Anagnostopoulou 1998, Chomsky 1981, and Jelinek 1984 for discussion of nonovert subjects). However, it is typologically possible to express a full argument subject through what is called a pronominal argument, which looks like an agreement marker. Straightforward diagnostic techniques (Baker 2003b) can be applied to determine whether an element is a pronominal argument or not. More importantly for Tuyuca, I will argue that //the final vowels do not pattern as pronominal arguments-being agreement markers instead-then fused tense-evidentials must occur lower than TP. The premise behind this is simple: subject-verb agreement and tense morphology is marked for the entire clause, which means the clause must be constructed before agreement can be marked on it. In standard Minimalist accounts, agreement morphology "co-occurs" with tense (and case) morphology. In other words, the syntactic assignment of agreement (and case) is tied-up with the tense of the clause. I will present two models. One is based on a Cinque hierarchy (1999, Appendix E), the other is my model. I argue that both models can explain the data but mine does it in a less complex, more elegant way. > 62 non overt ~lthe arguments-instead-then 63 5 . 2 . 1 . A note on why T E is important Why is the order (T, E) or (E, T) important if the two categories are fused? The first answer is that assuming a lexically inserted fused tense-evidential morpheme brushes the issue of order aside. It does not provide an explanation for why the categories Tense and Evidential can be fused in the first place. In all cases of languages with evidential morphology only a small number of them fuse with Tense morphology (Table 12 for a nonexhaustive set of tense-evidential languages). It may in fact be the case that the Tuyuca lexicon inserts a fused tense-evidential morpheme into the syntax, but this does not explain how this morphology developed (tense-evidential is not a grammatical or lexical primitive, it must have some historical development for which the interaction between morphology and syntax are partly responsible). Additionally, investigating the ordered relation between T-E and seeking an explanation for how tense-evidential morphemes develop relates to concerns of meta-theoretical significance. Simply put, the investigation of morphological (and morphosyntactic feature) change in languages will shed light on the nature of the kinds of parameters by which languages may vary (Kemenade 2007; but see Baker 2008, Chapter 5, for arguments that purely syntactic parameters exist-and §5.4 here for a brief note on, and explication of, Baker's syntactic parameters). An explanation of how tense-evidential morphology is derived not only relies on a solid explanation for the interface between morphology and syntax, but also on a clearly articulated notion of language change and reanalysis in the formally oriented Minimalist Program. Until recently, language change and reanalysis was the proving ground for functionalist theories of language. But collaboration and competition between theories that constitute the two dominant 5.2.1. > oflanguages meta­theoretical exist-and 64 paradigms of linguistic research, broadly construed as Functionalist and Formalist (see Newmeyer 1998 for detailed definition of these terms), can only be a catalyst to progress in understanding how languages work. I do not pretend to have a "solid explanation" for the morphology-syntax interface or a "clearly articulated notion" of reanalysis in formalist terms, but I believe that investigating the ordered relation between T-E and the morphological instantiation of this relation will contribute to both an "explanation" and an "articulation" of morphosyntactic reanalysis. 5.3.1. Setting the stage In languages that have pronominal argument suffixes an apparent agreement marker is actually a full argument of the verb: it influences valency and is marked for case. In languages that do not have pronominal argument suffixes there is a co­occurrence restriction in the distribution of agreement markers, pronominals, and definite nominal arguments. This restriction on the co-occurrence of, for example independent pronouns and full NP arguments, can be seen from examples in German and English adapted from Mithun (2003: 236-237)-read with no intonational pauses: (51) German English a. Er beobachtet. b. He watches. 5.3. Tuyuca does not have pronominal argument suffixes Mein Vater beobachtet. Erj beobachtet die Kinder^ *Mein Vaterj er, beobachtet. *Erj beobachtet siej die Kinder,. My father watches. Hej watches the children^. *My father, hej watches. *Hej watches thenij the children,. r r morpho syntactic suffixes co-occurrence definite 237)--Eri Kinderk. Vateri eri Eri sie] Kinder]. English Hei childrenk. fatheri hei Hei themj childrenj. 65 (52) English a. My father, he sees. b. He sees them, the children. (53) Spanish a. Mi Papa vio los ninos. 'My father sees the children.' Brazilian Portuguese f. Meu Pai viu criancia. 'My father sees the children.' b. (El) vio. '(He) sees.' (Ele) viu. '(He) sees.' Vio los ninos. '(He/she) sees the children. h. criancia. '(He/she) sees the children.' *Mi papdj elj vio los ninosj. 'My dad he sees the children.' *Meu paij elej viu crianciaj. 'My dad he sees the children.' Mi Papai, elj vio los ninosj. 'My dad, he sees the children.' j . Meu Pai„ elej viu crianciaj. 'My dad, he sees.' German, English, Spanish, and Portuguese are very different from languages such as Walpiri, which treats all instances of fully specified referential noun phrases as nonessential. Jelinek (1984) calls this the Pronominal Argument Hypothesis. For example, the Ergative case-marked subject in (54a) 'the child' and in (54b) 'the man' are As seen in (51) there is a restriction on allowing subject or object NPs to co-occur with a pronoun for which the reference is the same. However these examples are perfectly acceptable with appropriate intonational pauses related to discourse and topic-focus phenomena, signaled by placement of the comma in (52). There are similar data in Spanish and Portuguese (53), which are pro-drop languages, i.e., they do not need to overtly express the subject. Instead, they rely on subject agreement morphemes to recover subjects that are not full NPs. drop English h. Spanish nifios. h. c. d. e. Via nifios. papai eli via nifiosj. Papa;, eli via nifiosj. Portuguese crian<,:ia. g. h. i. j. Viu crian<,:ia. paii e1ei crian<,:iaj. Pail' e1ei crian<,:iaJ• 66 adjoined, or dislocated, material not needed for core predicate-argument structure. They are similar to the fully referential nominals in (53e-j) and (52a-b). The difference between Walpiri and English, Spanish, and Portuguese is that the agreement on the verbs in (52), (53a-b), and (53f-g), i.e. watch-s, vi-o, vi-u, does not force the definite subject noun phrase, i.e., father, papa, pai, or the subject pronoun, i.e. he, el, ele, to be interpreted as adjoined or dislocated material - as they are in (53e-j) and (52a-b). Walpiri (adapted from Hale cited in Jelinek 1984: 60) ZERO-Child-ERG childj himself o ZERO-Child-ABS manj himself, Baker (2003b: 1) says of Jelink's (1984) analysis of Walpiri that [the] article introduced the Pronominal Argument Hypothesis into Principles and Parameters-style theories. In brief, her idea was that some languages have obligatory pronominal agreements/clitics that count as the arguments of verbs and other predicators. Full NP's in such languages are thus never themselves arguments; when present at all they have the status of optional adjuncts of some kind. Baker himself goes on to show that full NP's in languages with pronominal argument suffixes are dislocated; they are introduced into the syntactic derivation in adjunct positions and not as part of the verb phrase. From this, he applies the notion of dislocation as a diagnostic forjudging whether or not certain agreement suffixes are in fact subject arguments of the verb: if agreement forces subject NPs to be dislocated in a language then is nonconfigurational. He concludes that there are languages that fall s, u, e/, material- (54) a. Kurdu-ngku ka-zERo-nyanu ngarrka-0 nya-nyi child-ERG PRES-3SG.NOM-REFL man-ABS see-NONPAST 'Hei the childi sees himselfi (as) a man.' b. Kurdu-0 ka-zERo-nyanu ngarrka-ngku nya-nyi child-ABS PRES-3SG.NOM-REFL man-ERG see-NONPAST 'Hei the mani sees himselfi (as) a child.' I) ofWalpiri brief: c1itics dislocated: adjunct for judging L, L 67 in between those exhibiting the properties of German, English, Spanish, and Portuguese on one hand, and Walpiri, Mohawk, and Nahuatl on the other. The former set represents the configurational languages, the latter the nonconfigurational. In between the two extremes are partially configurational languages like Chichewa, Kinande, Slave, and perhaps Najavo. The goal here is to determine where Tuyuca fits in this typology: is it configurational, partly configurational, or nonconfigurational. 5.3.2. A look at Tuyuca subjects The diagnostic of subject dislocation can be applied to Tuyuca. With dislocated (or adjoined) subjects there is variability in the basic word order. For Tuyuca, this means that if one can find instances in which the canonical S(0)V order is violated without causing ungrammaticality, then this might be a case of a dislocated subject. However, even noncanonical word order is not a guarantee of dislocation, other diagnostics factor in; word order is just a good place to start. Unfortunately, relevant data for noncanonical word order do not exist in published sources-which is reflective of the data set, not the language. Despite the lack of noncanonical word order in the data set,2 2 this should not undermine the suggestiveness of the word order constraints in Tuyuca-they are still reliable. Thus, there are no instances in which a fully referential definite NP occurs postverbally, nor do subject pronouns ever deviate from the S(0)V pattern in Tuyuca data (55)-(62). This suggests that word order is an important organizing principle in Tuyuca syntax. (Glosses reflect the analysis in Chapter 4, the separation of agreement from 2 2 clauses is not productive - more data need to be collected. this. nonconfigurational. subjects O)sources-which non canonical 22 Tuyuca-O)22 There are actuall a few examples but they do not appear to be relevant to the purposes here. Also, because most of the nominative-accusative case marking in Tuyuca is covert, analyzing variable word order 68 I I interpret to be the genitive case /=ya/; see (58), (60), (61). Pak* sTa-yig4. kill-EVD.PST.SCD-3MSG Pak* yai-re2 3 sTa-yig-4. ACC.DEF EVD.sCD-3MSG (57) MS pak* wTma-n? nTt-rj+24 wii soe-yiri2 5? 2 S G be-OJ.MSG burn-rNT.EVD.PST Ko6=ya-g4 y+4-tuti-3FSG=GEN.SG-CLF.AN-MSG dog 1 SG-ACC.DEF SC0ld-EVD.PST.VIS-3MSG (59) Imi-a p*a-ra hea-w-a. man-PL.AN two-PL.AN.QUANTITY arrive-EVD.PST.VlS-3PL K**=ya-g4 t=ya-g4 rd 3MSG=GEN.SG-CLF.AN-MSG 2SG=GEN.SG-CLF.AN-MSG more-ADVR g4 nTT-T. be.big-GER-MSG be-EVD.PRES.VIS-3MSG 'His animal is bigger than your animal.' Ko6=ya-g4 k?f=ya-g4-kdrd 3FSG-GEN.SG-CLF.AN-MSG 3MSG=GEN.SG-CLF.AN-MSG-alike 2 3 reference, of a 2 4 nn-rji. rji wlmd-tji I 2 5 evidentials, and conformity to the Leipzig Glossing Rules. have also made more explicit the postpositional clitic representation, by using the equal sign as the boundary for clitic "=" (Leipzig Glossing Rules 2006), of what Barnes calls the possessive morpheme and (55) Paki yai sia-yfg-t. father jaguar kill-Evo.PST.sco-3MSG 'Father killed a jaguar.' (56) Paki yai_re23 sla-ylg-t. father jaguar-Acc.oEF kill-EVO.PST.sco-3MSG 'Father killed the jaguar (that had been killing the chickens).' Mil paki wim~-IJ+ nrt_IJ+24 wit yiri25? 2SG father child-M be-rn.MSG house bum-INT.EVo.PsT 'When your father was a child, did he burn down the house?' (58) ya-g-i- diyi yit-re tutf-w-i. 3FSG=GEN.SG-CLF.AN-MSG dog I SG-ACC.OEF scold-EVO.PST.VIS-3MSG 'Her dog barked at me.' Imi-~ pia-d man-PL.AN two-PL.AN.QUANTITY arrive-EVO.PST.VIS-3