A Grammar of Tariana

A Grammar of Tariana This is a comprehensive reference grammar ofTariana, an endangered Arawak language from a remote region in the northwest Amazonian jungle. Its speakers traditionally marry someone speaking a different language, and as a result most people are fluent in five or six languages. Because of this rampant multilingualism, Tariana combines a number of features inherited from the protolanguage with properties diffused from neighbouring but unrelated Tucanoan languages. Typologically unusal features of the language include: an array of classifiers independent of genders, complex serial verbs, case marking depending on the topicality of a noun, and double marking of case and number. Tariana has obligatory evidentiality - every sentence contains a special element indicating whether the infonnation was seen, heard, or inferred by the speaker, or whether the speaker acquired it from somebody else. This grammar will be a valuable source-book for linguists and others interested in natural languages. CAMBRIDGE GRAMMATICAL DESCRIPTIONS Editors: R. M. W. Dixon, Keren Rice This series is devoted to the publication of comprehensjve descriptive grammars of languages that have not previously been documented, and that have interesting and unusual characteristics which will expand our understanding of human language in all its ruversity and challenge the limits of current linguistic theory. Some of these languages are spoken by only a small number of people and can be considered endangered. Each grammar briefly introduces the society in which the language is spoken, and covers the key areas of phonology, morphology and syntax., together with typological and historical considerations. In each case, a sample text or texts in the language are provided, with full gloss and translatjon. A glossary of basic vocabulary is also included. The series aims to provide theoretical unguists in the various subdisciplines with reliable data and analysis which will provide a permanent and invaluable set of source materials. A GRAMMAR OF TARIANA, FROM NORTHWEST AMAZONIA Alexandra Y. Aikhenvald Research Centre for Linguistic Typology LaTrobe University, Melbourne rfll1 CAMBRIDGE ~ UNIVERSITY PRESS PUBLISHED BY THE PRESS SYNDICATE OF TilE UNIVERSITY OF CAMBRIDGE The Pttt Bu1ldmg, Trumpington Street, Cambndge CB2 IRP, United Kingdom CAMBRIDGE UNIVERSITY PRESS The Edinburgh Bu1ldmg, Cambndgc CB2 2RU, UK 40 West 20th Street, New York, NY 10011-4211, USA 477 Williamstown Road, Port Melbourne, VIC 3207, Australia Ruiz de Alarcon 13, 28014 Madrid. Spain Dock House, The Waterfront, Cape Town 8001, South Africa http://www.cambridgc.ors Cl Alexandra Y. Aikhenvald 2003 This book is in copyright. Subject to statutory exception and to the prov1s1ons of relevant collecuve licensing agreemenLs, no reproduction of any part may take place without the written permission of Cambridge University Press. ~- First Published 2003 LJU E It i A/EI'- !tiECIO A catalogue reC'ordfor this book 1s av~lbe System IMEX2c [TB] from the Brirish Librory Library of Congress Catalogumg in Publkalion data ISBN 0 521 82664 0 hardback ,. I. FACTI ·: -rr,_, (\.;117 .cA'~M Printed in the Umted Kmgdom at the University Press, Cambridge JYpefaceTimes NewRoman ll / 13.5pt. n-oc.-c J-ftOCEtENC IA ""4:.u/P·' 1 • •t 7 4 ftW to __JL_...:... /I li IV' " For my Tarianafamily Contents List oftables, schemes and diagrams xv Preface xvii Acknowledgements xiv Organisation and cross-referencing XXI List ofabbreviations xxii Map xxiv l.The language and its speakers 1.1 Linguistic profile ofTariana 1.2 Tariana and the multilingual setting of the Vaupes 1.3 Historical information about the Tariana 9 I .4 Social organisation l I 1.5 Ceremonies and beliefs 13 1.6 Naming practices 15 1.7 Nonverbal communication 16 1.8 What we know about the Tariana language 17 1.8.1 Denominations of lhe language 17 1.8.2 Previous studies of the language 18 I. 9 Materials and speakers 18 6 2. Phonology 25 2. 1 Segmental phonology 25 2.1. I Consonants 25 2.1.2 Vowels 32 2.2 Syllable structure, vowel sequences and the problem of diphthongs 34 2.2. 1 Syllable structure 34 2.2.2 Vowel sequences and diphthongs 35 2.3 Stress 3 7 2.3. 1 Primary stress 37 2.3.2 Stress in monomorphemic words 38 2.3.3 Stress in polymorphemic words 38 2.3.4 Secondary stress 39 2.3.5 Vowel reduction 39 2.4. Phonological word 40 2.4.1 Primary stress and prosodic classes of morphemes 40 2.4.2 Nasalisation 42 2.4.3 Aspiration 42 2.4.4 Vowel harmony 44 2.4.5 Word delimiting prosodic parameters in Tariana: a summary 45 2.5 Phonological processes 46 2.5.1 H-mctathesis 46 VII viii Content.r 2.6 2.7 2.8 2.9 2.5.2 Vowel fusions 47 2.5.3 Phonological processes within roots and within affixes Prosodic classes of morphemes and their properties 53 Pause marking 60 Phonological phrase 63 Intonation patterns 64 3. Word classes 66 3.1 Open word classes 66 3.l.l Verbs 66 3.1.2 Nouns 68 3. 1.3 Adjectives 72 3.1.4 Word class-changing morphological derivations 3.2 Manner adverbs and time words 77 3.3 Closed classes 80 3.4 Word classes and functional slots: an overview 81 4. Nominal morphology and noun structure 52 77 82 5. Noun classes and classifiers 87 5.1 Overview of the multiple classifier system 87 5.1.1 Organisation of the system 87 5.1.2 Noun classes 99 5.1.3 Classifiers as nominal derivational suffixes I 01 5.1.4 Numeral classifiers I 03 5.1.5 Verbal classifiers 104 5.1.6 Classi tiers with modifiers from closed classes I 05 5.2 Functions of classifiers 115 5.3 Repeaters and 'ad hoc' classifiers 117 5.3. I The use of repeaters I 17 5.3.2 The functions of repeaters 118 5.3.3 'Ad hoc' classifiers 119 5.3.4 Classifiers and the lexicon 121 5.4 Summary 121 6. Possession 122 6.1 Personal cross-referencing 122 6. I . 1 The system 122 6.1.2 The indefinite prefix 123 6.1.3 Impersonal cross-referencing 126 6.2 Obligatorily possessed nouns 127 6.3 Optionally possessed nouns 128 6.4 Archaisms, innovations and irregularities in obligatorily possessed nouns 129 6.4.1 Archaisms and innovations in marking inaJienable possession 129 6.4.2 Expansion of optionaJiy possessed nouns 129 6.4.3 Retention of possessive suffixes on optionally possessed nouns 131 6.4.4 A common Arawak irregularity in possession markjng 133 Contents 6.5 Possessive classifiers 133 6.5 .I Constructions with the possessive -ya- 134 6.5.2 Constructions with a generic possessed classifier -yarupe 6.6 Conclusions 137 139 7.1 General properties of case marking system 139 7.2 Semantics and function of core cases 140 7 .2.1 The marking of A, Sa, S0 I 40 7.2.2 The marking ofnon-A/SafSo 143 7.3 The semantics and function of oblique cases 148 7.3 .I Locative case 148 7.3 .2 lnstrumcntal-comitati ve case 150 7.4 Omission of oblique case markers I 54 7.5 Marking case more than once 155 7.6 Double case marking and double marking of syntactic function 7.6.1 Double case marking 158 7.6.2 Double marking of syntactic function I60 7.7 Summary 163 136 7. Case marking and grammatical relations 8. Number 164 8.1 Number marking 164 8.1.1 Inanimate nouns 164 8.1.2 Animate nouns 165 8.1.3 Kinship nouns 166 8.1.4 Uncountable nouns 168 8.1.5 Pluralia tantum 170 8.1.6 Singulatives 171 8.1.7 Associative plural 172 8.2 Plural marking on modifiers 173 8.3 Multiple marking of number 176 8.4 Number agreement 180 8.5 Summary 182 9. Further nominal categories 183 9.1 Tense 183 9.1.1 Nominal future 183 9.1.2 Nominal past 185 9. 1.3 Nominal and clausal tense marking: a comparison 9.2 Extralocality and restrietivity markers 188 9.3 Contrastive marker 189 9.4 Conjoining morphemes 191 9.5 Approximative 192 9.6 Diminutive and augmentative 193 9.6.1 Diminutive 193 9.6.2 Augmentative 194 9.7 Pejorative 195 187 I 58 IX x Contents I 0. Derivation and compounding I 96 I 0.1 Productive derivational suffixes 196 10.2 Semi-productive and non-productive derivational suffixes I 0.3 Deverbal nominalising affixes 200 I 0.4 Classifiers and nominal derivation 200 10.5 Compounding 201 197 I I . Closed word classes 203 11.1 Personal pronouns 203 11.2 Specifier articles 204 I I .3 Demonstratives 206 11.4 Interrogative-distributive kwa-lkwe- 208 11.5 Gestural deictic khi 213 11 .6 Distributive individuatiser napada 214 11.7 General indefinite pa:- 215 11.8 Numerals 217 11.9 Quantifiers 219 1 I .I 0 Connectives 221 11 . II Adpositions 222 12. Verb classes and predicate structure 234 12. 1 Vcrb classes 234 12.1 . 1 Ambitransitive, transitive, ditransitive and extended transitive verbs 235 12.1.2 Intransitive verbs 239 12.1.3 Prefixless verbs wilh two arguments, and extended intransitjve verbs 241 12.1.4 Secondary verbs 244 12.1.5 Copula verbs 250 12.2 The structure of a predicate 253 12.3 Verb root structure and verbal derivations 255 12.4 Defining a verbal word 256 13. Valency changing and argument rearranging mechanisms 258 13.1 Passive 258 13.2 Reflexive-reciprocal 263 13.2. 1 Reciprocal meanings 263 13.2.2 Reflexive meanings 266 13.3 Increasing transitivity 267 13.3.1 Morphological causatives 267 13.3.2 Causative serial verb constructions 274 13.3.3 Periphrastic causatives 275 13.3.4 Causative mechanisms in Tariana: a comparison 277 13.4 Argument-adding derivation 279 13.5 Argument-manipulating derivation 283 13.6 Summary 286 14. Tense and evidentiality 287 14.1 General remarks 287 Contents 14.2 Tense-evidentiality in affirmative clauses 289 14.2.1 Semanticsoftenses 289 14.2.2 Visual evidentials 293 14.2.3 Non-visual evidentials 296 14.2.4 Inferred evidentials 299 14.2.5 Reported evidentials 302 14.2.6 Lexical reinforcement of evidentials 304 14.2.7 Preferred evidentials and their cultural correlates 305 14.2.8 Omission of evidentials and speakers' competence 309 14.3 Tense-evidentiality in interrogative clauses 311 14.3.1 Visual interrogative evidentials 312 14.3.2 Non-visual interrogative evidentials 313 14.3.3 Inferred interrogative evidentials 315 14.3.4 Omission and variability of interrogative evidentials 317 14.3.5 Preferred interrogative cvidentials 318 14.4 Future 320 14.5 Final remarks 321 15 Aspect, Aktionsart and degree 324 15.1 Aspect markers 324 15.1.1 'Short duration. little by little' -ina 324 15.1.2 Habitual prescribed -hyuna 326 15.1.3 Customary -kape 328 15.1.4 Habitual repetitive -nipe 329 15.1.5 Anterior -nhi 330 15.1.6 Non-completed ongoing -daka 334 15.1.7 Non-completed ongoing proximate -sida 336 15.1.8 Already accomplished -sita 337 15.1.9 Repetitive -pita, -ta 339 15 .1.1 0 Completive -niki 340 15.1. I I Not quite completed -mana just about, almost' 342 15.2 Aktionsart enclitics 342 15.2.1 Aktionsart enclitics with corresponding verbs 343 15 .2.2 Aktionsart enclitics not used as independent verbs 348 15.2.3 Semantic and syntactic properties of Aktionsart enclitics 15.3 Degree markers 366 15.4 Emphatic markers 367 I 6. Mood and modality 371 16.1 Imperatives 371 16.1.1 Unmarked imperative 371 16. I .2 Proximate, distal and postponed imperatives 16.1.3 Detrimental imperative 374 16.1.4 Imperative by proxy 376 16.1.5 Conative precativc 377 16.1.6 Cohortativc 378 16. 1.7 Polite suggestion 378 16.1.8 Additional imperatives 379 16.2 Frustrative 380 372 363 xi xii Contents 16.3 Intentional 383 16.4 Apprehensive 384 16.5 Uncertainty 387 I 6.6 Conditional 390 16.7 Purposive 393 16.8 Counter-expectation 396 16.9 Declarative-assertive 398 17. Negation 400 17. 1 General characteristics of Tariana negation 400 17.2 Clausal negation in non-prohibitive clauses 400 17.3 Negative imperative 409 17.4 Negative prefix ma- 410 17.5 Inherently negative lexemes 411 17.6 Negative ne and negative response 4 19 17.7 Double negation 42 1 18. Serial verb constructions and verb compounding 423 18.1 Working definition of a serial verb construction 423 18.2 Defining properties of serial verb constructions in Tariana 424 18.3 Assymmetrical serial verb constructions 430 18.4 Symmetrical serial verb constructions 440 18.5 Ambient serial verb constructions 442 18.6 A comparison of three types of serial verb constructions 443 18.7 Verb compounding 446 19. Complex predicates 449 19.1 Epistemic complex predicate with repetition 449 19.2 Prolonged customary action 450 19.3 Complex predicate with a subordinator and repetition 451 19.4 Passive complex predicate with auxiliary 452 19.5 Apprehensive complex predicate with auxiliary 453 19.6 Admirative complex predicate 453 I 9.7 lrresultative complex predicate 454 19.8 Complex predicate of smaJI extent 456 19.9 Quasi-serial verb construction 456 19.10 Complex predicates with the optional subordinator kwe and purposive mood 457 19.11 Clause-like complex predicate 457 19.12 Complex predicates and serial verb constructions: a comparison 458 20. Participles and nominalisations 460 20.1 Participles 460 20.2 Nominalisations and converbs 461 20.2.1 The -nipe nominalisation 461 20.2.2 TI1e -mi nominalisation 465 20.2.3 The -ri nominalisation 467 20.2.4 Three nominalisations: a comparison 472 Contents 20.2.5 Nominalisations and subordinate clauses 20.3 Other derivational devices 473 xiii 472 21. Clause types and other syntactic issues 475 21.1 The structure of noun phrases 475 21.1 .1 Position of adjectival and closed class modifiers with respect to the head of a noun phrase 475 21.1.2 Discontinuous noun phrases 479 21.1.3 Appositional constructions 480 21.1.4 Headless noun phrases 482 21.1.5 Possessive and adpositional noun phrases 483 21.2 Coordination of noun phrases 484 21.3 Structure of predicates 487 21.4 Types of clauses 487 21.4. 1 Declarative clauses 488 2 1.4.2 Imperative clauses 502 2 I .4.3 Interrogative clauses 502 21.4.4 Exclamatory clauses 506 21.4.5 Dependent clauses: an overview 506 21 .5 Grammatical relations 507 21.5.1 Defining 'subjects' 507 21.5.2 Objects and obliques 511 22. Subordinate clauses and clause linking 515 22.1 Clauses marked with sequencing enclitics 515 22.1.1 Switch reference-sensitive sequencing enclitics 516 22.1.2 Non-switch reference-sensitive sequencing enclitics 525 22.2 Sequential -/w 528 22.3 Subordinate clauses marked with adpositions 532 22.4 Subordinate clauses and clause linking 534 23. Relative clauses 537 23.1 Grammatical properties of relative clauses 537 23.2 Content question words as markers of relative clauses 24. Complement clauses 547 24.1 Complement clauses marked with -ka 'subordinator' 24.2 Complement clauses marked with purposive 551 24.3 Complementation strategies 552 24.4 Interrogatives in complement clauses 556 543 54 7 25. Discourse organisation 561 25.1 Pragmatic basis for constituent order 561 25.1.1 Order of words in NPs and multi-word predicates 562 25.1.2 Pragmatic basis for clausal constituent order 562 25.1.3 Constructions with fixed constituent order 568 25.1.4 'Doubling' of personal pronouns 571 25.2 Ellipsis 573 25.3 Floating enclitics and discourse 574 xiv Contents 25.4 Repetition and sentence linking devices 576 25.4.1 Repetition 576 25.4.2 Sentence linking strategies 576 25.4.3 Discourse-organising phrase '1 am saying' 583 25.5 Grammatical properties of narratives and conversations 25.5.1 Narratives 585 25.5.2 Conversations 588 25.5.3 Code-switching 590 585 26. Issues in etymology and semantics 594 26.1 Tariana grammar and lexicon: an etymological perspective 594 26.1.1 Arawak grammar and lexicon in Tariana 594 26.1.2 Semantic restructuring under the innuence of East Tucano languages 595 26.1.3 Lexical borrowings 596 26.2 Semantic issues relevant for grammar 598 26.2.1 'Same' and 'different' 599 26.2.2 Generic and specific terms 603 26.2.3 Positional verbs 608 26.3 Remarks on Tariana lexical semantics 616 Appendix. The main features of the Tariana dialects Texts 630 Vocabulary 671 References 682 Index ofauthors, languages and subjects 690 620 XV Tables, schemes and diagrams Tables 2.1 Consonants 26 2.2 Phonotactic restrictions on the occurrence of consonants 31 2.3 Vowels 32 2.4 Phonotactic restrictions on the occurrence of vowels 33 2.5 Properties of affixes, proclitics and enclitics 42 2.6 Phonetic realisations of independent pronouns 45 3.1 Kinship nouns with fully suppletive vocative forms 69 3.2 Kinship nouns with partly suppletive vocative forms 70 3.3 Kinship nouns which form vocatives by subtraction of the gender sensitive suffix without stress shif1 70 3.4 Kinship nouns which form vocatives by stress shift 71 3.5 Kinship nouns which undergo no change in vocatives 71 3.6 Kinship terms which distinguish three forms 71 3.7 Word classes and functional slots in Tariana 81 5.1 Classifiers in Tariana 89 5.2 Classifiers with modifiers of closed classes 114 5.3 Reclassification o f a noun: uni 'river, water' 115 6.1 Cross-referencing prefixes and pronouns 122 7.1 Grammatical relations and core cases in Tariana 139 7.2 Oblique cases in Tariana 140 8.1 Personal pronouns in Tariana 177 9.1 Tense in Tariana nouns 183 9.2 Tense in Tariana participles 185 9.3 Properties of the diminutive marker with different word classes 194 I0.1 Derivational suffixes 197 10.2 Nominal ising affixes 200 I I. I Properties of closed classes 204 11 .2 Agreement forms of kwa-/kwe- 209 11 .3 Agreement forms of khi- 214 11 .4 Agreement forms of napada 214 11 .5 Agreement forms of kanapada 221 11.6 Properties of ad positions of the types A-1-1: a summary 224 12.1 Verbs used as primary and as secondary verbs 248 12.2 Verbs of feeling: a comparison 250 12.3 Prefixless and prefixed copula verbs 251 13.1 Valency reducing mechanisms and verb types 286 14.1 Evidentials and tense in affirmative clauses in Tariana 289 14.2 The semantics of evidentials: a summary 294 14.3 Evidentials and tense in interrogative clauses in Tariana 311 14.4 Evidentials and clause types in Tariana 322 xvi List oftables. schemes and diagrams 15.1 Aspectual markers in Tariana 325 15.2 Aktionsart enclitics with corresponding verbs 343 15.3 Aktionsart enclitics not used as independent verbs 349 15.4 Semantics of Aktionsart enclitics 364 17.1 Future and intentional in negative clauses 406 18.1 Verbs which impart aspectual meanings to asymmetrical serial verb constructions 432 18.2 Verbs which impart directional meanings to serial verb constructions 434 18.3 Order of components in assymmetrical serial verb constructions 440 18.4 Asymmetrical, symmetrical and ambient serial verb constructions 444 18.5 Properties of the two types of verb root compounding 448 18.6 Cliticised verb compounding and asymmetrical serial constructions 448 20.1 Nominal and verbal properties of normnalisations 472 22. 1 Switch reference-sensitive sequencing enclitics 516 22.2 Properties of subordinate clauses in Tariana 535 23.1 Comparison of main, relative and subordinate clauses 537 25.1 Possible functions of a common NP in two syntactically linked clauses, with some illustrative examples of deletion of its full occurrence 574 26.1 'Sameness' and 'differentness' in Tariana 603 26.2 The uses of -ni 'do' 607 26.3 Semantic features of positional verbs 609 Schemes 12.1 Predicate structure in Tariana: affixes and root 253 12.2 Predicate structure in Tariana: enclitics 254 13 .I Causativising mechanisms, verb types and their semantics 278 Diagrams 4. 1 Noun structure in Tariana 83 5.1 Semantics and form of articles with classifiers I 08 11.1 Semantics and form of demonstratives with classifiers 206 14.1 Preferred evidentials in Tariana 305 I 4.2 Preferences for evidentials in interrogative clauses 3 18 14.3 Evidentials in question-response: Tariana and East Tucano 3 19 xvti Preface This is a grammar of Tariana, a member of the Arawak family - the largest language family of Central and South America. Tariana was once a dialect continuum spoken by over 1500 people in various settlements along the Vaupes river and its tributaries. The subgroups of Tariana used to constitute a strict hierarchy, and each would speak a different dialect. The difference between individual dialects is comparable to that between Romance languages. As the Catholic missions and white influence expanded, the groups highest up in the hierarchy abandoned Tariana. As the result, the language is actively spoken by only about I00 people, mostly adults, representatives of the lowestranking group Wamiarikune, and is severely endangered. I started assembling data on the language in 1991 and have since worked with most speakers of this dialect of Tariana. As by-products of this work, we produced a bilingual Tariana-Portuguese dictionary, several collections of texts and a practical grammar, and launched a program for teaching Tariana in secondary school. The dedication of the consultants, their insights and their patience made it possible to carry out this daunting task. This grammar contains an analysis of the Tariana language, starting from a brief characterisation of language and its speakers, then going on to phonology, morphology, syntax, discourse organisation and semantics. The analysis is cast in terms of a cumulative typological framework of linguistic analysis - which employs 'the fundamental theoretical concepts that underlie all work in language description and change' (Dixon 1997: 128) and in terms of which significant typological generalisations arc postulated; this has come to be called basic linguistic theory. r have avoided employing any of the more specific formalisms (which come and go with such frequency). Every chapter of this grammar includes a presentation of the facts of the language interwoven with arguments for their analysis within a typological framework. No attempt has been made to separate pure 'description' from theoretical interpretation. In particular, numerous features of Tariana go against what was thought to be 'universal' for instance, having serial verbs and complex predicates of several distinct types as independent grammatical phenomena, or having gender and classifiers as overlapping means for noun categorisation. A typological perspective for each phenomenon is crucial for the analysis given here. Tariana is spoken in the multin~:,ra area of the Vaupes basin, characterised by strong cultural inhibition against language mixing - viewed predominantly in terms of loan morphemes. The long-term interaction between East Tucano languages and the Tariana dialect continuum has resulted in a rampant diffusion of patterns and calquing of categories- which include classifiers, tense-evidentiality, number marking and even pronominal cross-referencing. There is a strong areal convergence of patterns - rather than of forms- without, however, implying anything like the emergence of an 'identical' grammar. When appropriate, information on the arcally diffused or genetically inherited character of a pattern is included in this grammar. A full analysis of contact-induced developments in Tariana is dealt with in a separate study (Aikhcnvald forthcoming-a). xviii Preface Detailed exemplification is provided for every grammatical point. Most examples come from texts, and a few from spontaneous - or carefully directed - conversation. An additional objective is to convey as mucb of the Tariana culture as possible through using naturally occurring examples. 1 avoid using elicited sentences; elicitation was limited to lexicon and to paradigms. English glosses arc kept as close as possible to the glosses and explanations offered by my consultants. When no English equivalent is readily available, a Portuguese word is used, e.g. abiu-fruit or mucura-rat; some semantically complicated terms are explained in the texts. or in Chapter l. Readers should be warned against trying to draw conclusions concerning Tariana grammar and semantics from the study of translations. Examples, tables and diagrams are numbered separately within each chapter. Footnotes are numbered separately for each chapter. The orthography used in the examples from languages other than Tariana, and language names, follows that of the sources (w1less indicated otherwise). This grammar can be used as a sourcebook for further typological studies, and as a model for further grammars of languages of Arawak and of other families of the Americas. It is far from being the last word on Tariana - this grammar is intended to provide a sound systematic foundation for further studies, reanalyses and reinterpretations. lt is my hope that this book will encourage linguists to go out into the field and document languages threatened by extinction (before it is too late to do so), notwithstanding the difficulties and temporary frustrations which necessarily accompany a field worker. xix Acknowledgements My warmest thanks go to the Brito family who taught me their language and accepted me as a member of their family. When in July 1991 Graciliano Sanchez Brito ('Gara') appeared in a makeshift house in the township of Sao Gabriel da Cachoeira (state of Amazonas, Brazil) that 1 was sharing with a few students from the Federal University of Santa Catarina, little could I suspect that he would become my main teacher, and then my classificatory older brother. Later I met his brothers, Jovino and Jose Luiz, his sister Olivia, and his parents. His father, Candido Brito - one of the few traditional speakers and a real treasure-chest of Tariana lore and culture- was an indefatigable and patient teacher. And his mother, Maria Sanchez - a Piratapuya herself - adopted me as her daughter and taught me quite a few secrets about the life of Vaupes women. Gara's cousin, Rafael, the youngest speaker of the Santa Rosa dialect, has always been extraordinarily helpful, besides being great fun to have around. When I went to the villages of Tauarete and Santa Rosa, I met lsmael (Maye) Brito, Gara's older brother, and Leonardo Brito, Candido's younger brother. I learnt an enormous amount from both of them. It is hard to find words to express my gratitude to these people. I learnt a lot from all the other speakers of the Tariana dialect of Santa Rosa - who told me stories, or came to chat - Ricardo Brito (Dika) and his sons Emilio and Raimundo; Jose Manoel Brito and his children Sebastiao (Saba), Cristiano (Kiri), Joao {Nu), Clementina and Joaninha; Juvenal, Abelardo and their mother Amelia (a Guanano herself); Cristina, married to Abelardo; Jacinto Brito, and Batista Brito. Americo Brito, the oldest living speaker of Tariana, shared with me his eyewitness accounts of the offering feasts - he is the only living Tariana to have seen them. The women in Santa Rosa made an effort to communicate with me and help me in many ways. Some spoke Tariana, like Cecilia (Leonardo's wife). With others, our language of communication was rudimentary Tucano and Portuguese. 1 am deeply grateful to Edna Lucia, Maria do Carmo, Clemencia, Marilia, Silvana and Maurllia. Fcli Brito (Batista's son) knew no Tariana- I am grateful to him simply for being kind. And there are so many more people to thank -Eduardo Brito and his wife Antonia; Julia Brito and Nestor 13rito - for their hospitality and friendship My warmest gratitude also goes to the people of Periquitos, for teaching me their variety of Tariana, and for their perseverance in maintaining the language and speaking it to their children. I am grateful to the elders: Marino and Jorge Muni:t; to the younger generation: Domingo, lsmael, Batista, Dario, Vivaldo, Gustavo and Joao Bosco; and to the women of Periquitos - Alice, Maria Carlota, Ester Belinda and others - who are quietly determined to save Tariana. The future of the language is in their hands! I met Roni Lopez, a speaker of the Santa Terezinha dialect. just once. I am deeply grateful to him for his patience and interest in teaching me his dialect. , an expert on the I would also like to acknowledge my debt to Padre Casimiro Bck~ta area of the Upper Rio Negro and the Vaupes, who frequently helped with information and advice. He was the ftrst person to mention to me, back in 1990, that Tariana is still L xx Acknowledgements spoken by a very remote group, upstream from lauarete. And he has, ever since, been a real encyclopedia of the Upper Rio Negro region for me. I have also profited from discussions with Elias Coelho, Janet Barnes, Henri Ramirez and Stephen II ugh-Jones. I am grateful to all my ex-students and friends from Brazil, who accompanied me at the initial stages of fieldwork and provided me their friend ship and support - Simoni Valadares, Tatiana Santos, Marcos Baltar, Antonio Augusto de Souza Mclo, Cristiane Cunha de Oliveira and especially Rule Maria Coelho. This grammar would have been impossible without the assistance, support and friendship of Silvana and Valteir Martins. Lenita and Elias Coelho provided me with a second borne in Sao Gabriel da Cachoeira, looking after me, and giving me advice in every aspect of life, and concerning fieldwork in the remote areas of the Upper Rio Negro. My gratitude goes to all those people from the Upper Rio Negro area who taught me their languages: Humberto Baltazar and Pedro Angelo Tomas (Warckena); the late Candehirio da Silva (Bare); Afonso, Albino and Joao Fontes, Celestino da Silva, Cecilia and Laureano da Silva and the late Marcilia Rodrigues (Baniwa); the late Tiago Cardoso (Desano, Piratapuya), and also Alfredo Fontes (Tucano). Wannest thanks go to Sisters Aline Sienkiewicz, Daria Moratelli, Cristina, Rita and lreni - who looked after me in so many different ways, making their mission welcome for me- during my stays in lauarete. All those who participated in the Tariana literacy class and helped organise it deserve separate thanks. Marta Azevedo provided us with the infrastructure and food; she also organised a Pedagogical workshop which resulted in the creation of quite a number of stories in Tariana. And without GI6ria da Silva, Cecilia da Silva, Lourdes Brito, Galdino Pinheiro, Pedro Jesus, Padre Jesus, Padre Benjamim, Sister Daria Moratelli, Jaqueline Brito, Vanilde Brito, Nestor Brito and many, many others, this course would not have happened. I am most gratef'Ul to those who read through the whole draft of this book, or parts of it, and provided comments, corrections and ideas- Willem Adelaar, Avery Andrews, Timothy J. Curnow, Antoine Guillaume, Keren Rice, Tom Payne, Gunter Senft, Fritz Serzisko and Mary Ruth Wise. Invaluable comments on almost every page came from R. M. W. Dixon, without whose constant encouragement and support r would have never completed this grammar. Tonya Stebbins and Carol EI-Chaar carefully read through the draft of tlus book and corrected it. Adam Bowles formatted the volume with meticulous care. Thanks are equally due to them. Siew Peng Condon provided a wonderful working atmosphere at the Research Centre for Linguistic Typology. This book would have been scarcely possible without her. I gratefully acknowledge the Wenner Gren Foundation for a grant which made my fieldwork financially feasible. xxi Organisation and cross-referencing A brief overview of the linguistic profile ofTariana is found at the beginning of Chapter 1. Phonology is discussed in Chapter 2. Chapter 3 contains an outline of word classes. Chapters 4-20 deal with various aspects of morphology, white syntactic issues arc considered in Chapters 21-24. Chapter 25 deals with discourse organisation, and Chapter 26 provides a brief sketch of semantic issues, especially those relevant to the understanding of lhe grammar. Here and passim'-' stands for any morpheme boundary, that is, a boundary between a root and an affix, or between two roots. The symbol '=' indicates a boundary between a root and a elitic, or an affix and a clitic, or two clitics (see Chapter 2). The symbol · indicates a primary stress, and · indicates a secondary stress (obligatory on enclitics). Stress is marked on each example in Chapter 2 ('Phonology') and in other chapters only if it is relevant to the discussion. Since nasal vowels and long vowels are always stressed, stress on these is not marked. In Tariana there is a considerable amount of variation between certain allophones (discussed in Chapter 2). Examples of alternative pronunciations are yapt, napi 'bone', yitpu, iidpu 'stream', yama, nama 'two', pamziya, pamuna 'middle', -pidena, -pidana 'remote past reported', -naku, -nuku 'topical non-subject', -nuka, -naka 'present visual', di-kena, di-kenwa 'he begins'. There is also variation between long vowels and short vowels, e.g. ke:ri, keri 'moon', na:, na 'they went, they said'. Most of these variants depend on the age and proficiency of the speaker; every example in this grammar records the actual pronunciation by the consultant. Examples are numbered separately for each chapter. All the examples and texts are supplied with an interlinear morpheme gloss, and then translated into English. Homophonous morphemes (e.g. -nha 'pausal marker; interrogative present visual; imperative (containing doubt)') are differentiated by their glosses. The symbol '+' is used to indicate fused morphemes, e.g. direta 'he ordered' is glossed as 3sgnf+order+CAUS, its underlying form being di- (3sgnf) -ira (order) -ita (CAUS). Portmanteau morphemes are glossed with a ':', for instance, -peni (PL:ANIM) 'plural.anirnate'. All grammatical morphemes are glossed in small caps while lexical morphemes are glossed in lower case. Pronominal pretixes are shown as I sg, 3pl, in lower case. For polysemous morphemes, different translation in glosses corresponds to different meanings; for instance -kaka 'reflexive-reciprocal' is glossed as RIZFL when it has reflexive meaning and as RC when it has reciprocal meaning. Cross-references arc oft wo kinds: • those preceded by § refer to chapter and section number, e.g. § 11 .1 refers to section I of Chapter I I; • those beginning with a number refer to examples in the grammar, e.g. I I . I refers to example 1 in Chapter I I. xxii Abbreviations A AI3STR ACT.CONTR ADM ADV AFF ANTM ANT APPR APPROX ART ASS.PL AUG AUX CAUS CL COHORT COLL COM COMPL COND CONY CONTR CONTR.ACT COUNTEREX CURV CUST DECL OEM DIM DIST DISTRIB DS EMPH EX(IST) EXC EXTRAL f, fern, FEM FOC.A/S FR FUT transitive subject abstract action contrast admirative adverbial derivational affix animate anterior apprehensive approximative article associative plural augmentative auxiliary causative classifier cohortative collective comitative completive conditional con verb contrastive contrast of actions counter-expectation curved customary declarative demonstrative diminutive distal distibutive plural different subject emphatic existential excessive extralocaJ feminine focussed subject frustrative future FUT.CERT FUT.UNCERT GEN GENJNFR HAB HABIT HORT HUM IMP IMPV IMPV.DETR IMP.SEC INAN INDF INFR INS INT INTER TRRES LOC LOCN MASC NAT.PI IEN NCL NEG NEG.EX NF,nr NOM NOM.FUT NOM.PAST NONVlS NPOSS NON.INDIV 0 OBJ NUM.CL p P.REL PART certain future uncertain future generic generic inferred habitat habitual hortative human impersonal imperative detrimental imperative secondhand imperative inanimate indefinite inferred instrumental intentional interrogative irresultative locative locational masculine natural phenomena noun class negative negative existential non-feminine nominalising nominal future nominal past non-visual non-possessed non-individuated object objective case numeral classifier person past relative participle List of abbreviations xxiii PART.CONTR participant contrast passive PASS pausal PAUS pejorative PEJ perfective PERF plural PL, pi polite suggestion POL.SUG possessive POSS postponed imperative POSTP.IMPV potential POT precative PREC PRES present prohibitive PROH proximate PROX PURP.NONVIS purposive non-visual purposive visual PURP.VIS qualificative QUAL reciprocal REC REC.P recent past referential kinship REF term reflexive REFL relative REL REM.P remote past REP REPET s Sa S;o So SPEC.TNFR SUB SG, sg SINGL ss TH TOP.ADV TOP.NON.NS UNCERT VERT VIS VOC reported repetitive subject of an intransitive verb subject of an active intransitive verb subject of an intransitive verb with a non-canonically marked argument subject of a stative intransitive verb specific inferred subordinating singular singulative same subject thematic topic-advancing voice topical non uncertainty-subject vertical visual vocative MAP. ARAWAK AND TUCANOAN LANG UAGES IN NORTH-WESTERN AMAZONIA I The language and its speakers 1.1 Linguistic profile of Tariaoa Tariana is the only Arawak language spoken in the linguistic area of the Vaupes river basin in the territory of the Upper Rio Negro in northwestern Bra:t.il. All the other languages belong to the East Tucano subgroup of the Tucano family. The main feature of the Vaupes area is its obligatory multilingualism, dictated by the principles of linguistic exogamy (one has to marry someone who speaks a different language). There is a strong inhibition against 'language-mixing', viewed in terms of lexical loans. As the result of a long-term interaction, Tariana combines a few features inherited from ProtoArawak, with those diffused from East Tucano. Independent innovations in Tariana include granunatical phenomena divergent from those found in closely related languages and not explainable in terms of the influence of the East Tucano languages (see Aikhenvald 1999b, c). Tariana is polysynthetic and basically bead-marking (in the sense of Nichols 1986), with elements of dependent-marking -the latter due to the Tucano influence since most Arawak languages are entirely bead-marking (see Aikhenvald l999b). Tariana has twenty-five consonants and six vowels. As in many other Arawak languages of the area, there is a series of aspirated stops, preaspirated nasals and a glide. The vowels a, i, e and u have long and nasal counterparts, while o has only a nasal counterpart, and i has neither. The emergence of i and o, the two vowels with a limited distribution, is most probably due to the Tucano influence. Accent is contrastive, and of a pitch type, also a result of Tucano influence. An extensive set of phonological processes such as aspiration floating, h-metathesis and vowel fusion operate within a morpheme and across affix boundaries. Syllable structure is (C)V(h). Accent, translaryngcal vowel harmony, aspiration and nasalisation delimit the boundaries of a phonological word. See Chapter 2. Open word classes are verbs, now1s and derived adjectives. Underived adjectives form a closed class, with about twenty-nine members. They share some properties with nouns, some with stative verbs, and have some properties of their own. Adverbs and time words are semi-closed classes. Closed classes include demonstratives, intcrrogatives/distributives, quantifiers, over twenty postpositions, one preposition, and an unusual gestural deictic khi 'this size', which involves an actual demonstration of the shape or size of an object. A member of any open or semi-open word clac;s can occupy the predicate slot. Verbs have more morphological possibilities than members of other word classes used as predicates: for instance, only verbs can form imperatives. See Chapter 3. Nouns in Tariana are both derivationally and inflectionally complex. They can contain up to fifteen structural positions (one of which is a prefix position). Nominal categories include case, gender, number, classifiers, nominal tense (future and past), 2 I The language and its speakers cxtralocality, contrastivily, and various other affixes and clitics. 1.1 exemplifies an inflectionally complex noun (see Aikhenvald 1999d). 1.1 nu-we-du-ma-pe==yana-pe=tupe=pena=ne=se=misini=nuku Isg-younger.si bling-FEM-CL:FEM-PL=PEJ-PL=DIM:PL=NOM .fUT=COM =CONTR=ALSO:PART.CONTR=TOP.NON.A/S 'with my naughty future little younger sisters, too' Nouns have variable ordering of affixes, which results in meaning changes. Compare 1.2 and 1.3, where the same suffixes -da and -ma appear in a different order. 1.2 nu-kapi-ma-da I sg-hand-CL:SIDE.OF-CL:ROUND 'one palm of my hand' 1.3 nu-kapi-da-ma I sg-hand-CL:ROUND-CL:SIDE.OF 'one side of my finger' Plural and gender on nouns can be marked recursively. 1.4 shows that plural can be expressed up to five times within one noun. 1.4 nu-we-du-ma-pe==yana-pe=tupe=miki 1sg-younger.sibling-FEM-CL:FEM-PL=PEJ-PL=DlM:PL-NOM.PAST: PL 'my poor little dead younger sisters' The highly polysynthetic structure of Tariana nouns is an independent innovation of the language (neither Arawak, nor Tucano languages have this property)- see Chapter 4. Tariana has a very complicated system of over forty established classifiers which are used as noun class agreement markers, derivational affixes, nwneral classifiers and verbal classifiers (in terms of Aikhenvald 2000a). Example 1.5 illustrates a classifier -dapana 'house, habitat' (underlined) with an adjective, with a numeral, on a demonstrative, in a possessive construction, and on a verb. 1.5 pani-si ha-dapana pa-dapana DEM :INAN-CL:HAB one-CL:HAB house-NPOSS nu-ya-dapana hanu-dapana I sg-POSS-CL:HAB big-CL:HAB heku na-ni-ni-dapana-mahka wood 3pl-make-TOP.ADV-CL:HAD-REC.P.NONV1S 'This one big house of mine is made of wood.' An item is often introduced with just a noun, and then referred to with an appropriate classifier throughout a narrative. In this way classifiers are more frequent in discourse than actual nouns. Classifiers arc potentiaJly an open class, since any noun with an inanimate referent can be used as a 'repeater' (or 'self-classifier'). Repeaters also mark agreement with a topical noun, while established classifiers are used under other 1.1 Linguistic profile ofTariano 3 circumstances. Slightly different sets of classifiers are used with members of closed classes. The two-way gender opposition (feminjne vs. nonfeminine covering everything else) is used in personal pronouns Gust for third singular and all the plural forms) and verbal cross-referencing. Within a noun phrase, agreement can be marked twice. If a noun phrase contains another noun phrase as a modifier, noun class agreement with two distinct 'heads' - the head of the embedded noun phrase, and the head of the embedding noun phrase - is marked on the modifier. Thus, Tariana consistently follows the principle of multiple-layered marking of syntactic function, both within a clause (see below on double case), and within a noun phrase- see Chapter 5. Tariana distinguishes alienable and inalienable possession. Inalienably possessed nouns take person/gender/number prefixes. Besides first, second and thUd person, these have a form for 'fourth person', or impersonal, used for marking a generic referent, e.g. pa-kapi (lMP-hand) 'someone's hand'. lndefinite person marking replaces any other person marker on a noun used in a possessive construction, or as an argument of a postposition, e.g. nu-whida (lsg-head) 'my head', di-whida (3sgnf-head) 'his head', but ne: li i-whida (deer INDF-head) 'the head of a deer', nu-pumi ( l sg-after) 'after me', ne: ri ipumi (deer fNDF-after) 'after the deer'. The indefinite person marker is not used with verbs, while all the other person markers are. There is a possessive classifier construction which employs the generic noun yatupe 'thing, way' - see Chapter 6. Tariana is one of the very few Arawak languages with case marking for core syntactic functions. It developed the case markjng under Tucano influence (see Aikhenvald 1996a, 1999b, c). Personal pronouns with an animate referent distinguish subject and non-subject cases. Any non-subject topical constituent is marked with an enclitic -nuku (archaic form -naku). A subject in a contrastive focus is marked with a clitic -ne/-nhe. One case (-se) covers all locational meanings (locative, directional, elative). Another oblique case is instrumentallcomitative -ne (see Dixon and Ajkbenvald 2000, for a discussion of core versus oblique arguments). These case markers are optional and can combine with the marker -nuku 'non-subject topical constituent' if the constituent is topical (thus yielding a peculiar instance of marking the same syntactic function twice). Tariana also allows 'double case', that is, two distinct clausal functions can be marked on one noun phrase. Note that the double marking of syntactic function is distinct from marking the same function twice; see Chapter 7. Tariana nouns divide into several subclasses by their number marking. Number agreement in noun phrases and in clauses is obligatory only with human referents; see Chapter 8. Among typologically unusual nominal categories are nominal tense, extralocality and contrast; see Chapter 9. There are comparatively few productive derivational suffixes, with classifiers frequently used to form new nouns; see Chapter 10. Closed classes- articles. demonstratives, etc.- differ in their syntactic possibilities and in the subset of classifiers they require; see Chapter II. Tariana preserves the common Arawak division of verbs into transitive, intransitive active,which take prefixes cross-referencing A/Sa, and intransitive stative, which do not take cross-referencing markers. The person, number and gender of A/Sa are neutralised in negative forms . They are marked with the prefix ma- and with the suffix -kade (note 4 The /angt1age and its speakers that there is only one prefix pos1t1on per verb). A third class of predicates, those expressing feelings, physical states, etc., does not take cross-referencing markers. These predicates mark their subject with the non-subject case.' The Tariana verb is polysynthetic, with twenty-one structural slots (ten of which are enclitics). 1.6 illustrates a predicate with nine positions filled; see Chapter 12. 1.6 ma-siteta-kaka-kade-karu=pidana=pita=niki ma-sita-i-ta-kaka-kade-karu=pidana=pita=niki - underlying form NEG-smoke-CAUSl-CAUS2-REC-NEG-PURP.VIS=REM.P.REP=AGA1N= COMPLETELY '(We) did not make each other smoke (reportedly) again at all.' Tariana has a reciprocal and a passive derivation (strikingly similar to the passive in Tucano), and three types of causative. Morphological causatives are formed on intransitive verbs. The same morpheme with a transitive verb indicates the advancement of a peripheral argument to the core, and/or complete involvement and topicality of the 0 argument. Periphrastic causatives (indirect causation) and serial causative constructions (direcl causation) are applied to transitive verbs. An argumentmanipulating derivation marked on the verb is employed to promote a focussed constituent to surface subject. The underlying subject retains a number of subject properties, including verbal cross-reference- see Chapter 13. The Tariana verb has a vast array of categories to do with tense, evidentiality, aspect, mood and modality. A complex evidentiality (fused with tense) is, cross-linguistically speaking, one of the unusual features of Tariana (developed under the influence from Tucano languages). Four obligatory evidentiality speci'fications indicate the source of information obtained - visual, non-visual, inferred and reported. For instance, in describing an event such as 'The jaguar killed a man', use of the visual evidential would imply that the speaker saw this event happening. The non-visual evidential would be used iftbe speaker heard the noise of a man fighting the jaguar (or smelt the blood). The reported evidential would be used if someone told the speaker of the event, while the inferred evidential might be employed if the speaker had encountered a jaguar covered with human blood. Only visual, non-visual and inferred are distinguished in interrogative clauses. The four non-future tense specifications are present, recent past (used to talk about actions or states that started from two minutes to a few days ago), and remote past (used to refer to actions or states that started a long time ago and may be ' Tariana has lost the Proto-Arawak morphological split-ergativity (split S system in the sense of Dixon 1994: 71) marked through cross-referencing affixes (see further discussion in Aik.henvald 1995b, 1998b and 1999b) in the following way: • A=S. and possessor, expressed with cross-referencing prefixes • O=S. expressed with cross-referencing enclitics, where A is the subject of a transitive verb, 0 is the object of a transitive verb, S, is the subject of an intransitive active verb, and s. is the subject of an intransitive stative verb (see Dixon 1994: 70). Similarly to a number of other Arawak languages (for instance, Bare in the area of the Upper Rio Negro. or the Xinguan Arawak languages: see Aik.henvald 1999b), Tariana has retained the Proto-Arawak crossreferencing prefixes but lost the cross-referencing enclitics. /.1 Linguistic profile ofTariana 5 still continuing). Of the two future markers, one is used only with first person, and the other with any person. There is a reported future construction. No other evidentiality distinctions are made in the future; see Chapter 14. In addition to obligatory tense, Tariana marks a number of aspectual meanings to do with completion of an action or a state, e.g. anterior, ongoing, ongoing proximate, accomplished, completive; or to do with its duration, e.g. habitual, repetitive, short duration. Several dozen clitics describe varying facets of the action. some of which have highly specific meanings. These include 'do early', 'spill water', 'wag a tail in a friendly manner', 'step on something and feel sharp pain'. Most of these originate in verbal roots, and were calqued from Tucano. Clitics which mark the degree of action or state are the diminutive, augmentative, approximative ('more or Jess') and excessive; see Chapter 15. Tariana has an interrogative mood and a number of imperatives: simple (unmarked), proximate ('do here'), distal ('do there'), by proxy (order on someone else's behalf), precative ('please do'), cohortative ('Jet's do'), and detrimental ('do to your own detriment'). There are only two possibilities for the negative imperative in the negative: the general prohibitive, and the imperative 'by proxy'. Modalities include frustrativc ('do in vain'), apprehensive ('lest'), dubitative and conditional; see Chapter 16. Negation is marked differently for NSa (transitive and active intransitive) verbs and for S0 (stative intransitive) verbs; see Chapter 17. Serial verb constructions express aspectual, directionaL benefactive, causative and other meanings. The causative serial verb constructions are typologically quite unusual both verbs, which have different agents, are marked for the same subject. See 1.7 and Chapter 18. 1.7 du-a du-.iiha tfari-nuku 3sgf-make 3sgf-eat man-TOP.NON.A/S 'She fed (lit. make-eat) the man.' (lit. she-made she-ate the man). Complex predicates, used to express epistemic meaning, prolonged customary action, admirative, apprehensive, irresultative (that is, an action or state which does not quite amount to what it ought to), small extent etc., differ from serial verb constructions. Serial verb constructions require the same subject marking on all the components which cannot be separated by any intervening constituents or any markers of syntactic dependency; in contrast, complex predicates do not have these restrictions; see Chapter 19. The predicate of a relative clause is marked with a participle which distinguishes three tenses- past, present and future. Deverbal nominalisations are regularly formed on any verb, and employed as a complementation strategy; see Chapter 20. Clause linking is marked with sequencing enclitics, some of which are switch reference sensitive: the choice of an enclitic depends on whether or not the subject of lhe main clause is identical with that of the subordinate clause. The emergence of switch reference in Tariana is probably due to Tucano influence; see Chapter 22. Any argument except the possessor can be relativised. Interrogative pronouns are also employed in relative clauses, possibly under the influence ofTucano languages; see Chapter 23. Complement clauses can occur only in 0 function. Besides complement clauses, Tariana has a number of complementation strategies, one of which is a direct speech complement used after the verbs of speech; see Chapter 24. 6 I The language and its speakers Similarly to a number of other languages of the world (see Mithun 1987), Tariana does not have any basic constituent order. The rules for ordering constituents within an NP, a verbal complex, a clause or a sentence depend on types of constituents, the construction type and the pragmatics - for instance, the order within a noun-adjective phrase depends on the definiteness of the noun referent and how topically established it is. This is discussed in Chapter 25, together with different discourse genres and principles of code-switching. In spite of a strong cultural inhibition against lexical borrowing, Tariana does have a few borrowings from Lingua Geral (a lingua franca employed until recently throughout the Upper Rio Negro and the Vaupes area - see § 1.2), and from Tucano languages. These borrowings are fully nativised bound morphemes, many of them verbs. Tariana has also undergone a number of lexical shifts under East Tucano influence. The striking difference between the two dialects of Tariana, of Periquitos and of Santa Rosa, lies in the degree of 'acceptance' of loans from Tucano. While loans are not accepted - and are ridiculed - by the Santa Rosa speakers, a few Tucano loan morphemes have made their way into the Tariana of Periquitos. This is conspicuous in U1e speech of younger people; see Chapter 26. This grammar is basically focussed on the Santa Rosa dialect of Tariana. The main features of the other actively spoken dialect, that of Periquitos and of other Tariana dialects are summarised in the Appendix. The Tariana language of Santa Rosa is not being learnt by children, and shows a degree of obsolescence. The younger generation are still fluent in Tariana, but many prefer to use Tucano when speaking among themselves or to their non-Tariana spouses. They consistently use Tariana only when speaking to someone from their father's generation. The language spoken by younger speakers is characterised by calques and loan translations from Tucano, and, rarely, from Portuguese (Aikhenvald forthcoming-a). 1.2 Tarian a and the multilingual setting of the Vaupes The variety of Tariana described here is spoken as the fLrst (father's) language by fewer than one hundred people, considered members of two families in two settlements on the Vaupes river: the Brito fami ly of the Santa Rosa village (also known as Jukira-ponta, Tariana lwi-taku 'point of salt') and the Muniz family of the Periquitos village (also kno\\'11 as Kerekere-pani 'rapid of a parakeet', or Tupiyari-numana 'the mouth of a lizard').l A few Tariana speakers live in lauarete (Yawhipani (iaguar+CL:RAPID) 'the rapid of a jaguar'), a local mission centre, and in Ji-Ponta (Episitaku (stone.axeCL:POINT) 'the point of stone axe')- see Map. Over 1500 people in numerous settlements along the upper and the middle course of the Vaupes river identify themselves as ethnic 1 The old location for this settlement of the Muniz family of the Wamiacikune was Kerelcete-pani, 'the rapid of a parakeet', of which Periquitos is a translation into Portuguese (periquito is a word for parakeet}. It became swampy and uninhabitable a generation ago, and the Muniz moved to a nearby location called Tupiyarinumana 'mouth of a lizard'. The Portuguese name Periquitos, which had already figured in maps and official documents, has been transferred to this new location. 1.2 Tariana and the multilingual setting ofthe Vaupes 7 Tariana but no longer speak the language (cf. Aikhenvald 1996a and Moreira and Moreira 1994). The North Arawak languages are spoken on the Is;ana and its tributary, the Aiari (the Baniwa/Kurripako language, with HohOdeoe and Siuci as its main representatives), on the Vaupes (Tariana and Ilohodene) and on the Upper Rio Negro (Warekena and, formerly, Bare) (see AikhenvaJd l996a, 1999a-c, and forthcoming-c). Several languages of the East Tucano subgroup are spoken on the Vaupes (see Barnes 1999 and Aikhenvald 1999c). Traditionally, the most numerous East Tucano-speaking groups in the Brazilian Vaupes were the Tucano, the Guanano/Piratapuya and the Dcsano (there are also some Tuyuca and some Cubeo). Nowadays Tucano is gradually gaining ground as the lingua franca of the area. Linguae franche spoken in the region are Lingua Geral, or Nheengatu, and Tucano. Lingua Geral is spoken in the whole region of the Upper Rio Negro (see Rodrigues 1986: 102, Bessa Freire 1983, Moore et al. 1994). On the Vaupes, it is only understood by older people.J Lingua Geral is a creolised version of Tupinamba (Tupi-Guarani family) spread from the east coast of Brazil by white merchants and missionaries. It was the lingua franca of the whole Amazon region from the late seventeenth century up to the middle of the nineteenth century; its influence can still be seen in a few loan words in Tariana and other languages of the Vaupes (see Chapter 26). In t11e Vaupes area, it was gradually replaced by Tucano as a lingua franca from the early twentieth century, as a result of the language policy of Catholic missionaries and the civil authorities. IndoEuropean languages are represented by Portuguese and Spanish. They are also gaining ground as lingue franche of the region, especially among younger people. Multilingualism used to be- and to a great extent still is- a cultural norm over the whole Vaupes. Among the East Tucano peoples and the Tariana, marriage is exogamous. There are strict marriage rules, which are governed by language affiliation (see Sorensen 1972, Aikhenvald 1996a). Marrying someone who belongs to the same language group is considered akin to incesl Jackson (1974: 62), referring to the linguistic attitudes of the Bara, an East Tucano group of the Vaupes basin, reports being told: 'My brothers are those who share a language with me', and 'We don't marry our sisters'. Consequently language - which is acquired through patrilineal descent - is a badge of identity, together with the patrilineal descent. An Indian always identifies with their father's tribe and language. An individual generally knows between three and ten other languages of the region, including their own mother's language which would frequently coincide with his wife's language, and in addition Portuguese and/or Spanish. Since language identity is a symbol of ethnic identity, languages - even the most closely related ones - are kept strictly apart (cf. Sorensen I 972: 82). This creates a very strong impediment to lexical borrowings - in contrast to other multilingual situations. Each tribe, identifiable by a distinct language (e.g. Tariana, Piratapuya, Tucano etc.), is traditionally divided into several subtribes, hierarchically organised depending on whether they are descendants of 1 The Tariana tenn for Lingua Geral is Bare, which suggests that in the early days the Bare were identified with speakers of Lingua Geral. 8 I The language and its speakers the first, second, Lhird etc. son of an ancestor. The lower groups in the hicrnrchy nre said to be ex-Maku (and called 'underlings', or 'soldiers') due to the traditionally low status of the Maku groups in the Vaupes society. Their role is to serve the members of senior subtribes who are to be treated as elder brothers - for instance, lighting their cigars in the cigar-smoking ritual. Each group - and apparently each subtribe - has its own version of origin myths; there are dialectal differences between subtribes (see Appendix). All the Tariana are considered blood relatives (and called di-kesi-ni 'relative'), while members of other groups- which include Baniwa and all the East Tucano groups except for the Desano - are referred to as 'marriageable cousins' (see Aikhenvald I 999a). The basic rule of language choice throughout the Vaupes area is that one should speak the interlocutor's own language. According to the language 'etiquette' of the area, one is supposed to speak the language one identifies with - that is, one's father's language - to one's siblings, father and all his relatives, and one's mother's language to one's mother and all her relatives. However, during past decades the traditional pattern of language transmission in the Brazilian Vaupes has been affected by a number of factors. When Salesian missionaries established themselves in the area in the early 1920s, they imposed Western-style schooling on the indians, forcing children into boarding schools where they were made to speak just one language, Tucano. Salesians aimed at 'civilising' Indians. This implied not only making them into 'good Christians'. Salesians also considered the traditional multilingualism of the area a 'pagan' habit, and strived to make Indians monolingual'like other civilised people in the world'. The Tucano language was chosen because it was, numerically, the majority language. Salesian missionaries also practised forceful relocation of Indian settlements closer to mission centres - where the Indians could be more easily controlled - and amalgamation of different settlements, eliminating the traditional longhouse system and introducing European-style nuclear family houses. Another reason for the disintegration of traditional multilingualism was a breakdown of traditional father-child interaction: with the need for cash-flow, all the able-bodied men would go off to work for Brazilians -undertaking such tasks as collecting rubber and gold-mining - and as a result children would have a considerably reduced degree of exposure to their father's language. This resulted in the spread of Tucano, and, to a Jesser extent, of other East Tucano languages, to the detriment of Tariana (see Aikhenvald 2001c). The main consequence of the recent spread of the Tucano language in the Brazilian Vaupes is the gradual undermining of the one-to-one identification between language and ettmic group. Language has gradually ceased to be an emblem of ethnic identity, and the majority of languages other than Tucano have become endangered. The discrepancy between the number of those who belong to a tribe and those who actually speak the language is particularly marked in the case of Tariana. The spread of Tucano is also leading to the gradual disappearance of one of the most fascinating multilingual areas of the world, and the areal phenomena associated with it. /.3 Historical information about the Tariana 9 At present, Tariana is a highly endangered language. The language described here is actively spoken in just two locations - Santa Rosa and Periquitos - by representatives of one subgroup, who call themselves Wamia-riku-ne (I pl+float-AFF:PI. ACE-PL) 'the people of the place where we t1oated'.• See Appendix on the minor differences between these. This denomination reflects the origin myth. While the hierarchically higher groups of the Tariana appeared directly from the blood of Thunder, the latest arrivals- among them the Wamiarikune- appeared drifting on the water of the Apui falls after the Woman-Creator had smoked her sacred cigar. Some older people - Americo and the elders from Periquitos - pronounce the name of the group as Wamiarikine. The existence of a sacred name Wamiari, in the Periquitos group, could be used as a piece of evidence in favour of -ki in Wamiarikine being a separate morpheme; it could be cognate to the masculine suffix -ki, as in nu-dafipa-ki (I sg-near-MASC) 'the one who is near me; a wife's relative' (see B under §I 0.2, for further examples). Marino Muniz, one of the Periquitos elders, used Enu-dakini 'Thunder's grandson' and Enu-maki-ni-se1i (thunder Maku-MASC-SINGL), lit. 'the Maku ofThunder', as alternative names for the same group. Systematic efforts to start teaching Tariana at secondary school level started in 1999, when the author, together with the Brito family and the then Director of the Secondary School in lauarete, decided to introduce Tariana as a school subject. A course in Tariana language and literacy was held in lauarete in June 2000, with over 300 participants, most of whom were non-speakers or semi-speakers, jointly financed by La Trobe University with the lnstituto Socioambiental in Brasilia Regular teaching of Tariana as a second language is scheduled for 2003 at the secondary school in Jauarete; and further educational activities are being planned as part of the agenda for the recently established Indigenous Association for the Language and Culture of the Tariana of the lauarete Area (Associar;ao lndigena da Lingua e Cuftura Tariana do Disrrito lauarete, AILECTIDI), under the presidentship of Jovino Brito (see Aikhenvald forthcoming-f). 1.3 Historical information about the Tariana The first information about the population of the Rio Negro and Vaupes, including the first mention ofTariana, goes back to the second half of the eighteenth century (also see Aikhenvald 1999a and c). The analysis of Tariana place-names shows that 'historical' and 'mythological' place-names exist only in Tariana and not in any of the Tucano languages (see Aikhenvald 1996b). This conclusion - unexpected in an enviromnent of obligatory multilingualism - suggests that the Tariana might have arrived in the Vaupes from a predominantly monolingual context, and that they have adopted multilingualism fairly recently. • The Tucano-speaking Tariana still preserve some origin myths. lnfonnation from the Tucano-speaking Tariana groups is valuable for derennining differences in myths and stories between subtribes; however it cannot be completely relied upon. An impon.ant work based on Kaline tradition (third group in the Tariana hierarchy: see §1.4) is Moreira and Moreira (1994); however, one can only rely on what is conlinned by other sources (such as BrUzzi 1977; 1994; Biocca 1965; Amorim 1987; Stradelli I 890). I0 I The language and its speakers The first source we have which gives some indication about the spread of the tribes of the Vaupes is by Vigario Geral (Dr Jose Monteiro de Noronha; dated 1759: see Bruzzi 1977: 20-31). Another early source is Francisco Xavier Ribeiro da Sampaio (1774-5), who mentioned the Tariana, the Desano, the Guanano and the Uaupe on the Vaupes, and on the Papuri river. He was also the first historian to have observed the existence of social hierarchies in the Vaupes region. Alexandre Rodrigues Ferreira (1775-6) also noted the Cariana (Tariana?) on the Papuri, and the Uaupe' on the Vaupes river. Wallace (1853) mentions the Tariana in Siio Joaquim on the Vaupes river; and Martius (1867, vol. I: 567) pointed out the existence of the Tariana at Siio Jeronimo (now lpanore) on the Vaupes - characterising them as 'Nehmer, Rauber' ('takers, robbers'). See BrUzzi (1977: 20-31). A detailed account of the spread and early history of contacts of the Tariana is given by Koch-GrUnberg (1911: 49ft). He also points out the existence of two Tariana groups on the Vaupes: one in lpanore and the other in rauarete. 6 The report that the Tariana knew of gold and possessed it is repeated in several sources, contributing to the idea that the Tariana in particular were highly civilised and constituted a kind of elite among the indigenous population. According to Candido Brito (p.c.), they indeed used to get gold from the Andes. Incidentally, Tariana has a native word for gold, hiwaru. The Tariana (and other tribes of the Vaupes) were involved in the 'rubber boom' in the late nineteenth century. Hemming (1987: 315) describes the depopulation and depression of the whole region of the Upper Rio Negro in the mid-nineteenth century, due to epidemics. When in 1850 the province of Amazonas was separated from Para, the attention of the governors of the new stage turned to the tributaries of the Upper Rio Negro, and among them the Vaupes (with its tributaries, the Tiquie and the Papurl), where there was still a considerable Indian population, among them the Tariana. This period is marked by numerous attacks by the government on the 'nomadic' tribes of the region, including the Hohodene (Baniwa) and Maku. A new wave of baptism of the Indians started in 1852, with the appointment of Frci G rcg6rio Jose Maria de Bene, a mission director, who reportedly baptised a third of the estimated 2300 lndians in the Vaupes basin. Traders and governors used the rivalries between the tribes to capture lndian slaves; Indians were urged to leave their villages and to move to the main rivers. Wallace (1853) reports that the Tariana assisted in capnaring Indians of other tribes for traders; it appears that the Tariana were used for these purposes because of their high status among Indians of the region (also see Hemming 1987: 319). 1 The term Uaupt!, or Buopcs, could have been an umbrella term for indigenous peoples of the Vaup~s area. 6 He reports that the mission of Silo Jeronymo was built in lpanore in the second half of the eighteenth century, by Carmelites. This mission was visited by Johann Nanerer in July 1831, by Wallace in 1852 and by Spruce in 1853. The lpanore mission was rebuilt by Franciscans in 1880-3 (see Koch-Gr1lnberg 1911: 35). At lhat time, there were 330 inhabitants (62 houses), almost all of them Tariana. The mission of Silo Antonio in lauarete then had 402 people. The missionaries were expelled from the region in 1883, and the Indians returned to the lifestyle of their ancestors in longhouses. Permanent Salesian missions were established in 1925, which resulted in the disintegration of the Vaupes culture (see above). 1.4 Social orgamsation II According lo Koch-Gri.inbcrg ( 1911: 51), the Ta.riana language started to be ousted from everyday use in lauarete as early as the beginning of the twentieth century.7 The gradual spread of Tucano as a lingua franca started at the end of the nineteenth century (Giacone 1962: 7). 1.4 Social organisation Descriptions of the social organisation of the peoples of the Vaupes area can be found in Goldman (1979), C. Hugh-Jones (1979), S. Hugh-Jones (1979), Sorensen ( 1972; 1985), Jackson (1974; 1976; 1983), Galvao (1979), and BrUzzi (1977); additional information was obtained from my Tariana teachers and various other people of the region (including some non-Tariana-speaking Tariana). The Tariana and Tucano classificatory kinship systems are a variation of a basic Dravidian type (Sorensen 1972: 85; Goehner, West and Merrifield 1985: 59). The basic distinction is between cross and parallel cousins. Members of ego's generation are either parallel cousins considered classificatory siblings, or cross-cousins; only cross-cousins are regarded as marriageable. Consequently, any Tariana is referred to as a parallel cousin, and any member of a marriageable tribe is referred to as a cross-cousin. The 'unmarriageable' Maku are referred to as 'underlings'. The kinship terminology distinguishes five generations, with the sex of relatives being marked for each generation (cf. Jackson 1983: I 06). A member of an exogamous patrilineal phratry is eligible to marry a person from another exogamous phratry who is identified as speaking a different language. See Aik.benvald ( 1999a) for a full list of kinship terms. The traditional settlement pattern involved multifamily longhouses each including a patrilineage. As mentioned above, the process of replacing traditional longhouses with individual family houses began in the Brazilian part of the Vaupes region in the 1920s. There are no longhouses left in Brazilian territory; a local descent group is now necessarily split into Western-type nuclear families. The patterns of slash-and-bum agriculture require access to extensive territory, and the fight for resources may explain the raiding and feuding, memories of which live on (e.g., the wars between the Guanano and the ltjirimhenc subgroup of the Tariana). 7 'The male Tarianas of Yauarete speak only Tucano among themselves, which is the language of their mothers and daughters, since exogamy is strict in all the tribes of the Vaupes, i.e. a wife is taken from another tribe, very often from far away. However, when they go on a visit to other Tariana villages, at the reception and in conversations, they use Tariana which is perceived as a more appropriate ceremonial language, a kind of 'salon language', whereas Tucano is used in everyday affairs or during trips. The younger generation has already forgotten many Tariana words, and this is the best proof that this sonorant language [i.e. Tariana) is gradually on its way out, whereas Tucano is spreading more and more, and has already become a kind of''Lingua Geral" in the Lower and Upper Caiary-Vaupes and its tributaries Tiquie and Papurl, for tribes of different languages' {Koch-Grilnberg 191 I: 51, my translation). I The language and its speakers 12 Historically, the Tariana consisted of ten subgroups. Different subtribes used different sets of sacred names ('blessing names': see below and §3.1.2); there are also differences in at least some origin myths.' The following ten subgroups of Tariana are arranged in order of seniority, following BrUzzi (1977: 101-3). Some are named after a mythological being they descend from, and some after a totemic animaJ; the etymology of some names is unknown. Names ending in -ne are plural (-ne is the Tariana plural animate marker: for instance, the Britos are called Brito-ne; this marker is also used in the name lri-ne, the autodenomination of the Tariana: see § 1.8.1 ). 1. 2. 3. 4. 5. 6. 7. 8. Kwenaka (descendants of the first son; meaning unknown) Jtjitimhene (descendants of the second son; meaning unknown) Kali-ne (possibly, descendants ofKali, the creator: Kali-PLURAL) Paiphene (meaning unknown) Kumandene (Ducks; people of the Duck) Mali-ne (Guans) Kunuli-whi (feather ofKlinuli bird) Phitji-kawa, Phitji-kawa-pu (group of agoutis); or Phitji-ne 'Agoutis'; or Phitfikawa-pu-pe 'big feet of agouti' 9. Yawyari (people of jaguar) I0. Wamia-tiku-ne (1 pl+float-AFF:PLACE-PL) 'the people of the place where we noatcd' The exact hierarchy of the Tariana subtribes remains to be determined. Representatives of different groups give different reports concerning their relationships to other groups. According to Wamiarikune lore, the Tariana groups fall into two large classes: those who came floating out of the water at lpanore rapids, and those who appeared out of the smoke of the creator's cigar. The former - which include Kabana, KaJine, Nerikwa, Pukudane, Kunuliwhi and ltfirimhene, besides the ones listed aboveare higher on the hierarchy than the latter - which include Wamiarikune and Paiphene, in this order. The exact place of Kumadene and Adarune in the hierarchy needs to be further investigated. According to Galdino Pinheiro, the Paiphene -a group to which he belongs - are ranked below the Wamiarikune. (The names of subtribes Kari, Kwenaka. f'itjia (cf. Phitfi-ne) arc also quoted by Biocca 1965: 255- who does not put them in any order.) 1 For instance, in the story of the origin myth by descendants of a senior subtribe. Kaline, reflected in Moreira and Moreira ( 1994) and Biocca ( 1965), there were sometimes two, and sometimes three Tariana forefathers, while in the version told by the Wamiarikune they were always two. 1.5 Ceremonies and beliefs 1.5 13 Ceremonies and beliefs The Tariana share numerous beliefs, ceremonies and superstitions with other peoples of the Vaupes region, and also with the Baniwa of lyana. Here l will not attempt an exhaustive description ofTariana and Tucano ritual life; my intention is to highlight the most striking features necessary for understanding Tariana stories. Details are in Aikhenvald ( 1999a). Tariana religion is animistic. Due to the efforts of the Catholic church traditional religious concepts arc now intertwined with Christianity. The creator of the Tariana and of the whole world is believed to be Yapi-riku-ri (papirikuri) (lit. bone-LOC-REL) (cf. Baniwa Yapi-riku-ri 'the one who is on the bone'. Tucano O'il-ki (bone-son) 'son of the bone'). He is sometimes identified with the Christian God, e.g. Yapi-riku-ri di-namikada (3sgnf-die-CL:DA Y) 'Good Friday', lit. the day when Yapirikuri died. Kali is the creator and the master of manioc, and Kui (possibly related to Baniwa Kuwai: Bruzzi J994) is the master of various types of snuff and the protector of shamans. He is also the owner of the Yurupary flute (pili i-mina-ti flute INDF-master-MASC). The reason why it is forbidden for women to see the Yurupary flute is that both Yapirikuri and Kui dislike women (ka-duiha ka-na REL-dislike REL-want), who had 'sinned' by having allowed the Moon (Ke:Ji) to copulate with them. Ancient people of the 'olden days' are called Hipada nawiki ('stone people') since the traces of their adventures are sti ll there, in the form of stones on the rapids around the Vaupes. White people (yalana) appear in the origin myth as one of the Tariana subgroups; it says that, due to their 'knowledge', they managed to get hold of guns and clothing while other groups (e.g. Tariana, Cubeo, Desano, Piratapuya) were helpless, 'like children' (cf. Hugh-Jones 1988, on how white people were incorporated into the mythological cycle of the East Tucano group Barasano). The mythical 'fish-people' (kuphe-ne), are also conceived of as 'white'. They are 'in charge' of fishing and, sometimes, hunting. They belong to the other world (pa:-ehkwapi 'one/other-CL:WORLD') and typically appear disguised as big handsome white women or men. They lure unsuccessful hunters (husaite, Portuguese panema), or 'men not loved by women' (ina meninite) to live with them in their huge beautiful underwater houses; when the hunters come back to the earth they bring nothing but misfortune to their relatives. The 'fish people' also take away girls menstruating for the first time if they dare to run away from their seclusion in order to have a wash. The 'tish people' can be helpful to a human being in trouble: in one story a snake man helped a destitute widow get fish from a secret stream, on condition that she should not take too many; but once the allowed quantity was exceeded, the whole village was destroyed. Shamans - see Table 9 in Aikhenvald ( 1999a) - are divided into six groups, according to their strength and ability to inflict illnesses (kaipeli di-a 'illness he-gives') and to cure them (kaiperi di-susu 'illness he-sucks'). They sniff different kinds of snuff (generic when1, hipatu 'snuff). Shamans of a 'higher' category can do anything the lower category can do; however, each has his specialisation - for instance, the 'beginner', sakaka, specialises in 'attracting women'. The higher castes of shamans are known as yawi 1aguar' because of their ability to 'turn into jaguars and eat people' (see § 1.8.1, on Yawi as one of the erroneous names for the Tariana). These shamans also have the power of opening 'the pot of fever' so as to make their enemies fall iII and die. Other people with some magic powers (outside the hierarchy of shamans) are kanapa (REL-bless) 'blesser' (i.e. a person with the power to bless the tar of certain trees, 14 I The language and its speakers and to bless people so that some illnesses go away), yaku-si mina-ri (tNDF+talk-NPOSS master-MASC) 'master of speech; a magician who does not sniff snuff but can foresee the future', di-tape-lcani (3sgnf-cure-AGENTIVE.NOM) 'curer; someone who can cure diseases', and ma:m 'the master of ritual dances'. When a girl menstruates for the first time (hituka) she is put into an enclosure for three days, where she has to eat certain types of fruit, but not fish, game or pepper, and is forbidden to bathe. After three days she is let out, beaten, bathed and painted by an old woman, and an offering feast (see below) is organised. A boy, once he reaches puberty, is supposed to spend some time in the jungle proving he can hunt and fast; after that he returns to the village, gets beaten, bathed and painted by an old man, and participates in an offering feast. Female initiation is still practised, while male initiation was abandoned several decades ago. These initiation rites are similar across the Upper Rio Negro; cf. the description of the male initiation of Candelario da Silva, the last speaker of Bare, in Aikhenvald (1995a: 52-4), and cf. Jackson (1983) and S. HughJones ( 1979). The Tariana tem1 for an old man or woman who is performing the initiation ritual is di-tude-ta-kasi (3sgnf-hit+CAUS1-CAUS2-AGENTIVE} (this term is often translated as 'godmother', or 'godfather'); the child to whom the ritual is done is called nuri-da (lsg+son-CL:ROUND) (now used in the meaning of'godchild'). The offering ritual (known in Portuguese as Dabucuri festival, Tariana na-walitanipe (3pl-offcr-NOM)) is the most important part of Tariana culture; the modem Tariana regret the fact that it is not practised any more. Apparently, America Brito (see § 1.9) is the last Tariana to have participated, as a guest, in a real offering festival. It used to be organised by one village or longhouse if they had a surplus of produce, or as a part of the rite of initiation. The village or longhouse where the feast was going to take place was responsible for making manioc beer (Portuguese caxiri), while the others provided the food (see Reichel-Dolmatoff 1971, and especially the description in Btilzzi 1977: 303-4). After ritual dances organised by ma:ro 'the master of ritual dances', kahpi (a strongish variety of manioc beer) was distributed, and drinking carried on until the next day - this is why another term for the offering festiva.l is pa-ira-ni-kada (IMP-drinkTOP.ADV-CL:DA Y) 'day of drinking'. Drinking involved vomiting (inducing oneself to vomit), to be able to fully enjoy the feast. On the morning of the third day the cigarsmoking ritual took place, to mark the end of the feast. Male members of a low-ranking group used to light two big cigars (yema), each of which was held in a forked cigarholder (yemapu). The most important males (usually, the most powerful shamans) of the two participating groups (the 'host' group and the 'guest' group) offered the cigars to each other to smoke (na-siteta-kaka (3pl-smoke+CAUS-REC) 'make each other smoke'); after that the cigars were passed from more important to Jess important men. The shamans and the important men alternated in reciting episodes from origin stories (further details are in Bruzzi 1977: 312). Cigar-smoking was accompanied by putting snuff on each other's tongue and licking it from there (na-pefita-kaka (3pl-lick+cAUS-REC) 'make each other lick') (this is not mentioned by BrUzzi 1977). A subtype of 'offering' feast (offering of fruit, but not of meat or fish) involves the sacred flutes (pin) known as Yurupary flutes (Tucano miri, Baniwa Kuwai; seeS. HughJones 1979; Jackson 1983: 188-91; BrUzzi 1977: 313-15, on the geographical spread of the Yurupary cult and its characterisation as a cultural feature of the lyana-Vaupes region). Women are not allowed to see the flutes (and have to run away inside the 1.6 Namingpractices 15 longhouse as soon as they hear the sound of them); a woman who has seen a Yurupary nute must die. Numerous taboos concern a man's behaviour in the jungle. A hunter, if he wants to be successful, cannot have sex, or even think about women, before he goes on a hunting trip. Othenvise he may fall victim to the evil spirit iiamu who might appear to him in the image of a 'white woman' (i.e. a 'fish-woman': kuphe i-sa-do (fish rNDF-spouse-FEM)) and devour him, or 'take his heart away'. Similar sets of taboos apply to women who know the secret of making pots. For three days before they go to a faraway place (usually a clearing in the jungle), they are not supposed to have sex, eat, drink, defecate or urinate: otherwise the pots will break (Olivia Brito and Maria Sanchez, p.c.). Nowadays, very few women know these secrets, so female taboos are falling into disuse. The taboos concerning women's behaviour are linked to the traditional fear of women's menstrual blood (widespread in numerous other cultures across the world). Menstrual blood is believed to be the result of women's 'misbehaviour' with the Moon (ke:ri ka-sa-do (moon REL-spouse-FEM) 'moon marries [them]'). Similar to Tucano origin myths, Tariana origin myths mention an 'ancestor' woman believed to be the first to use the Yurupary flute, which she subsequently lost; however, this story is not well remembered. Since women continually 'misbehave' and cannot even see the Yurupary nute, they arc often referred to as ma:nihta-kadite (NEG+think/rcason-NEG+NCL:ANIM) 'the ones who do not reason' - see §26.3. Various other taboos are linked to the 'likes' and 'dislikes' of evil spirits (see Table I 0 in Aikbenvald 1999a). The most important of the evil spirits, namu, is known to have a liking for the blood of the first menstruation; this is the reason why girls who menstruate for the first time (hitu-kite 'menstruate.for.the.fust time-THEMATIC+CL:ANIM') have to be kept in seclusion and maintain dietary restrictions - lest iiamu comes and devours them. The evil spirits are believed to be especially dangerous at night-time - for instance, one is not supposed to make any appointments with another person after dark lest the evil spirit learn about it and come and eat a person up, disguised as the other person. The fear of getting close to 'beings who are different from us' is reflected in numerous stories about people living together with evil spirits, or snake-people, and striking deals with spirits of the jungle; all this involves moving into 'the other world' (pa:-ehkwapi) which is always ominous and not recommended. (See also Aikhenvald and Dixon 1998a: 255, for some putative correlations between the various taboos and restrictions, and linguistic phenomena, in Tariana.) 1.6 Naming p r actices All the Tariana have a Portuguese given name and a Portuguese family name; the given name is used to refer to people when they are absent; kinship terms are often used for address. Similarly to other peoples of the Vaupes region, the Tariana have a closed set of traditional personal sacred names called 'names of blessing' (Tariana pa-napa-nipe ipitana (IMP-bless-NOM INDF-name) 'blessing name', Tucano base'ke warne 'spirit name'). 'Names of blessing' are given to children (not to babies); they used to be employed in 'spells' and 'blessings', especially when children were sick. Personal names are a separate subclass of nouns (they have vocative forms, and no plurals). 16 I The language and its speakers The Iist of names of the Brito family, as far back as my consultants could remember, is given under 0 in §3.1.2. Children of Tariana women are not 'named' in Tariana, because they belong to a different language group and do not count as Tariana. Only some of the sacred names have a translation: Kumatharo means 'female duck', Tuiri means 'japu bird', Serewhari means 'lilac-tailed parrotlet', and Kumada means 'male duck'. It is unclear on what principle people are given names (the blessing names of my consultants are listed in Aikhenvald 1999a).9 Portuguese names also distinguish reference fonns from vocative fonns; and they are often abbreviated. Olivia is addressed as 0/i and referred to as 6/i, and Leonardo is referred to as Leo and addressed as Leona, while lsmael is both referred to and addressed as Maye. ln addition, each man has a nickname, usually the name of an animal, or an insect, or of a part of it, e.g. Namuritu 'wild peccary' for Ricardo, Newi 'otter' for Jose Manoel, Hema 'tapir' for Rafael, Hi:ri isipi 'rat's tail' for Jacinto, Paitji 'type of frog' for lsmael, Kumada 'duck' for Jose Luiz, Dume 'aracu fish' for Leonardo, or Wirikaru 'flycatcher', Portuguese 'bem-ti-vi', for Emilio (I have no such data for women). Some of these names are considered slightly comical and some slightly offensive, and people do not like their own nicknames. For instance, Candido Brito has never revealed his nickname to me, and Jovino was very reluctant in revealing, to the satisfaction of ot11ers, his nickname Kapatu 'bod6-fish' (a fish that sticks underneath the canoe). I was strongly discouraged from using Cristiano's nickname Naki 'evil spirit', Raimundo's Makini 'a Maku man', Juvenal's Ka:siri 'crocodile', and Marino Muniz's nickname Jnari 'mucura rat', which they were said to dislike. White people are also given nicknames for instance, Giacone (§ 1.8.2) was known as Maliapa 'white-haired monkey' (Portuguese macaco loiro), because of his white hair. The blessing names are hardly ever used in day-to-day life. When addressing one another, speakers mostly use appropriate kinship tenns or abbreviated versions of their Portuguese names. All the older speakers - except Leonardo, who occasionally uses Portuguese names - use kinship terms to address their relatives, keeping Tariana and Tucano fonns strictly apart. See §25.4.3, on the use of Tucano and Portuguese kinship tcnns in Tariana. 1.7 Nonverbal communication The patterns of nonverbal communication are similar across the Vaupcs area. Among the people, there is a lot of bodily contact as a sign of affection: indigenous women often walk around holding each other by the hand. Female relatives hug each other. Laughter often signals that the people are amused, but the source of this amusement is most often the cultural inappropriateness of something, frequently of language behaviour: when one unexpectedly switches languages or uses a Tucano or a Piratapuya ' I was given a name, Kumatharo 'female duck', only during the fourth field trip, and my son was named Tuin japu-bird; pied crested oropendola'. These names were said to be given 'to keep us safe during our trips because we fly a lot'. 1.8 What we know about the Tariana language 17 word while speaking Tariana, this is accompanied by roars of laughter. A discourse organising phrase 'I am saying this', frequently employed by narrators as a hedge in teUing a story, also provokes outbursts of laughter because it is considered somewhat inappropriate (cf. §25.5.3). People, especially women, used to laugh when I first started speaking Tariana (since it is unusual for a white woman to do so). Gestures are important in story-telling. A story-teller will accompany their talk with gestures indicating the direction of movement of people or of animals, and also the direction of the sun (as a way of showing what time it is). In day-to-day communication, lips are employed for pointing at things which are close and can be seen. Lip-pointing is accompanied with a head-tilt to point at things which are further away but still within the range of vision. Finger pointing is employed for more distant things, especially those which are too far away to be seen. A full palm is used to indicate directions (which is important during trips on boats) and the position of the sun (that is, time of the day). 1.8 What we know about the Tariana language 1.8. 1 Denominations of the language The Tariana refer to themselves as Tarici, Tariana or Tariana; -na is a suffix used in names of people (see §I 0.2). This term also combines with the singulative -seri: Tariaseri 'a Tariana man', or with -sa-do (spouse-FEM), as in Tatia-sa-do 'Tariana woman'. According to BrUzzi ( 1977: I 00), the suffix -na could be a loan from Lingua Gcral. Another autodenomination of the Tariana is lii-ne (blood-PL) 'the ones of the blood'. In Tariana, the Tariana language is referred to as waku 'our speech', or wa-yarupe 'our thing', or taria yarupe 'Tariana's thing' (cf. §26.2.2). The Tucano-speaking Tariana do not use the term lri-ne; they refer to the Tariana with the Tucano word Diroa ('blood') (see Moreira and Moreira 1994). Two other names are used to describe the Tariana in the literature: • The denomination lnene (ine-ne (evil spirit-PL) 'demons') is a pejorative name for the Wamiarikune subgroup of the Tariana, given to the Wamiarikune ancestors by their envious 'elder brothers' because of their 'impertinence'. KochGrilnberg (19 1L) reported that the Yurupary-tapuya ('people of the evil spirit', in Nheengatu) - also known as Jyaine, or iyemi 'evil spirits' - used to be a 'subtribc' of Tariana which was very low on the tribal hierarchy (see also Coudreau 18867, vol. 2: 160, 163). Loukotka (1968: 134) erroneously lists lyiiine, Yumparytapuya and Kumandene (sic) as one and the san1e tribe, distinct from the actual Tariana (whom he also calls Yavi, following Coudreau). • The name Yawi, or Yavi, 1aguar' was mentioned by Coudreau (1886-7, vol. 2: 474-6), and was correctly refuted by Koch-Griinberg ( 1911: 50); this term is only used to refer to a type of shaman. ~- 18 I The language and its speakers 1.8.2 Previous studies ofthe language The first short word-list in Tariana was collected by Johannes Natterer (20 July 1831 ), in Siio Geronimo (now Ipanon!). This list was believed to be lost (cf. Hemming 1987: 489-90; and Koch-Gri.lnberg 1911: 50, fu.3); but has been just recently rediscovered. Further lists in Tariana were collected by Wallace (1853) and Martius (1867, vol. I: 537). Koch-Grlinbcrg (1911) published a longish word-list, accompanied by a few phrases and a list of Tariana tribes - these data are interesting, but full of mistakes. Brtizzi ( 1961) published a reliable list of over one hundred words in Tariana. Brtizzi (1977) contains a list of Tariana subtribes, their location, and a detailed description of Tariana-Tucano cultural characteristics. A short grammar of Tariana with a list of phrases and a short dictionary was published by Giacone ( 1962). He started working on the Tariana language in 1946-7 with the speakers of the Tariana dialects of two high-class Tariana groups: Phit:fikawape, . Both spoken in the village of Dom Bosco, and Kabana (Kwenaka), spoken in ftai~u dialects are now extinct (see Appendix). In 1959, Giacone started revising his materials, with the help of Anibal Muniz, a fluent speaker of the dialect of Periquitos and a classificatory younger brother of the elders of the Periquitos community- Marino and Maxirniliano Muniz. As a result, Giacone's grammar is a peculiar mixture of different dialects ofTariana and is not a reliable source. A longish word-list in Tariana was collected by Alva Wheeler in the early 1960s (parts of it were published in Huber and Reed 1992), from a representative of the Adarune subtribe; this dialect is very different to the Santa Rosa dialect described here (see Appendix). Gonzalez-Nanez collected a l 00-word list from Roberto Brito (the elder brother of Candido Brito) who lives in Maroa, Venezuela; this list is deeply llawed. A typical example is the Spanish gloss porque (atirmativo), i.e. 'because', translated into Tariana as jiiida (in Gonzalez-Nanez's transcription). The word hiiida in Tariana means 'I don't know'. Apparently, when the researcher asked the Indian 'how do you say "because"?', the Indian honestly said 'I don't know', and the researcher mistook it for the required word. Some ethnographical materials on the Tariana were collected by Ettore Biocca ( 1965); he also made a few recordings of shamanic songs sung by an old man from Ipanore. Brilzzi also recorded a number of songs (see Bruzzi 1961 ). These songs are only partly intelligible to younger speakers. Older people, such as Candido Brito and Jose Manoel, seem to understand more. However, none of the speakers was able to translate any of these though they had no doubt as to their authenticity and have even made an attempt to learn them 'back'. Various aspects ofTariana grammar have been discussed in my publications, from a typological, areal and historical perspective (see References). Features ofTariana culture are discussed in Aikhenvald ( l999a), which contains a sample of texts. 1.9 Materials and speakers This grammar is based on the materials collected during five field trips between 1991 and 2000. The materials include over 1700 pages of texts, and also word lists and conversations. /.9 Materials and speakers 19 I have worked, or been in contact, with 90 per cent of the estimated thirty speakers of the Santa Rosa variety in Santa Rosa and in Jauarete. In addition, I worked with Marino, Jorge, Batista, Dario, Gustavo and Domingo Muniz, of the Periquitos variety (and contacted a few more representatives of the Muniz family in Iauarete), and with Roni Lopes, one of the remaining twenty or so fluent speakers of the Santa Terezinha dialect This grammar is followed by four stories in the Santa Rosa dialect, and a story from Periquitos. Speakers of the Santa Rosa variety vary in their proficiency, depending on generation and upbringing. There are only eight representatives of the older generation (60-80 years of age). Of these, only Jose Manoel Brito and Ricardo Brito live in Santa Rosa; Candido, Leonardo, Americo and Batista Brito live in Iauarete, Eduardo Brito lives in Sao Pedro (on the Middle Vaupes), while Roberto Brito lives in Maroa (in Venezuela). CANDIDO BRITO, my main consultant (about 70-80 years old) is perhaps the best and the most traditional speaker ofTariana. To my knowledge, he is the only person never to violate the rules for language choice. Though perfectly fluent in Tucano and in a number of other East Tucano languages (as well as in Lingua Geral), he never uses Tucano when speaking to his children, his younger brother Leonardo or his classificatory brothers, addressing them in Tariana and demanding the same from them. His wife, Maria, a Piratapuya herself, is proficient in Tariana (her mother's language). She rarely uses Tariana: most of her communication with her husband and children is in Piratapuya, while she speaks Tucano and regional Portuguese to her grandchildren. This may explain why her Tariana remains archaic. Candido Brito is an elaborate story-teller, and has a vast knowledge ofTariana place-names, culture and lore. His language is archaic: he never substitutes the indefinite person marker with other markers, he uses archaic gender forms of demonstratives, and does not apply vowel harmony to enclitics -naku 'topical non-subject' and -pidana 'remote past reported' (realised as -nuku and -pidena by younger people). He, his wife and his eldest daughter Olivia are equally proficient in Tariana kinship terms and relationships. LEONARDO ORITO, Candido's younger brother (born in 1949), is a fine story-teller, very proficient in Tariana. Leonardo's wife, Ceci1ia, a Cubeo, can speak and understand some Tariana. Leonardo speaks Tariana and Tucano to his two sons (Rafael and Gabriel), and just Tucano to all his daughters. Gabriel (now 15) understands Tariana, but answers in Tucano; so does Vanilde, his eldest daughter (now about 20). The younger daughters, Clarisse, Fatima and Maria Esther, speak and understand only Tucano. Rafael (born in 1973), a nuent speaker of Tariana, speaks to his father mostly in Tucauo, but now and again also in Tariana. Agewise, Leonardo is closer to Candido's children (he is just a year older than lsmael Brito), and so his speech is more similar to that of the younger generations in that it lacks some archaisms found with Candido: he does not use archaic demonstratives; but, like Candido, he uses the fonns -naku 'topical non-subject' and -pidana 'remote past reported'. His knowledge of culture and lore is good, but somewhat more restricted than that of Candido. JosE MANOEL BRITO (probably over 70) is a fluent speaker, and a reasonable storyteller. He is a widower, and now lives with one of his eight children (see below). l-Ie speaks Tariana to his son and Tucano to his daughter-in-law. His grandchildren appear to be the only children in the village (besides lreni: see below) to understand some commands in Tariana - he probably does speak Tariana to them now and again. He knows some things (but nowhere near as much as Candido) about traditions. His Tariana is looked upon as somewhat 'faulty'. His pronunciation is blurred, and he reduces all 20 The language and its speakers vowels. According to some, he grew up among the speakers of Hohodene, and consequently has never completely learnt how to use the evidential system; he consistently uses present reported -pida instead of remote past and recent past reported. Some of the forms be uses are archaic; be says -naku and not -nuku. RICARDO BRITO (Dika) (younger brother of Jose Manoel, also over 70) is a middling speaker, and an eager - but not a very articulate - story-teller. His pronunciation is somewhat blurred, and the reduction of unstressed syllables makes his speech difficult to understand, even for native speakers. He speaks Tariana and Tueano to his children, and Tucano to his daughter-in-law. Like Jose Manoel, he grew up among the speakers of Hohodene, and for this reason is said not to have mastered the system of cvidentials. The forms he uses arc archaic in other respects. BATISTA BRITO (Bati, probably 60) is fluent in Tariana. He is the only representative of the older generation of Wamiarikune in Brazil to have a good command of Portuguese, which he freely mixes with Spanish. He spent a number of years working in Colombia and Venezuela, and is known to show off his proficiency in 'the white man's language'. He is the only person to be mocked for inserting Spanish or Portuguese words into his Tucano or Tariana, because he is considered a 'show oft'. lie speaks only Tucano to his chi ldren, and Portuguese to his grandchildren. His wife Teresa (a Tucano) does not speak or understand Tariana, and neither do his chi ldren. His son Feli, a male nurse who lives in Santa Rosa and who is responsible for distributing government medicine, was the only male in the village never to participate in the language work, for the simple reason that he had no knowledge whatsoever of the language. Dati is a good and witty story-teller; however, he does not know as much as Candido about tradition and lore. lie told me that he only started learning Tariana when he was about eight. His Tariana is innovative and full of calques from Tucano, and he tends to employ the forms younger people use. This could be explained by his late acquisition of the language. AMERICO BRITO (Arne, well over 80) is the oldest speaker of Tariana in Brazil (he is approximately the same age group as Roberto Brito, Candido's elder brother, now in Maroa). He is extremely proficient in the language, and a marvellous story-teller; his knowledge of tradition equals that of Candido. Unlike Candido, he speaks only Tucano with his wife Maria (a Tucano) and with his children. Consequently, none of his children speaks Tariana. His lanb>uage is very archaic and his ways of expression elaborate. 1 have not worked with Roberto Brito, who now lives in Maroa in Venezuela and is married to a Guanano woman. None of his children speaks Tariana; his eldest daughter Celestina (Celeste) who lives in the Sao Paulo community on the Vaupes can understand the language but lacks the confidence to speak it. My main consultants among the younger generation (30-50 years of age) arc the children of Candido Brito. Their proficiency in Tariana varies. None of their children speaks Tariana; Candido's grandson Francismar, the son of his late son Francisco ('Chico'), understands some Tariana (he lives in the same house as Candido and gets some exposure to the language). Candido's eldest son, ISMAEL (Maye, born in 1950) is probably the best speaker of his generation. He is an elaborate story-teller, and has a vast knowledge of traditional stories and of the lexicon; his speech is morphologically complex. He alternates between archaic and innovative forms (e.g. -naku and -nuku), and is sometimes innovative himself (one of the peculiarities of his speech is overusing vowel harmony: for instance, he regularly pronounces negative marker -kade- as -kede-). He almost never I. 9 Materials ami speakers 21 displays any signs of the loss of the indefmite prefix. He speaks Tucano to his wife Margarida (a Tucano herself), and only recently started systematically speaking Tariana to h.is elder son Osmar (who seems to understand it). Ismael also speaks good Portuguese - he is a trained mechanic and a general handyman; however, his addiction to cane whiskey ('cachatya') hinders him from securing a good job. Candido's eldest daughter OLiVIA (born in 1953) is a fluent speaker, very caring and motherly to her siblings (including me, as a classificatory sibling). She has an amazing knowledge of the kinslup system; she speaks mainly Tucano with her daughter Laura and her nephew Francismar (in her care, after her younger brother Chico's death). Her Portuguese is reasonable (but not as good as that of her brothers). I fer Tariana is typical of a younger speaker and is full of calques from Tucano and - if she does not control herself - Tucano words creep into her narratives and especially spontaneous conversations. l had to recheck every piece of data she volunteered, and what she said was often corrected by Jsmael, Leonardo and Candido. Her younger brother GRACILIANO (Gara, born in 1954) was Lrained as a nurse in Manaus (the capital of the state of Amazonas) and is now in charge of the purchasing department of the hospital at Tauarete. He was my first teacher ofTariana. He is fluent in Tariana and speaks excellent, native-like Portuguese. He is a good and witty story-teller; however, his Tariana is marked by numerous calques from Tucano, and quite a few things he says are later corrected by Jsmael, Leonardo and Candido. For instance, he is the only person to insert an object NP between two components of a serial verb: once, when I repeated such a construction, I was told not to use it again because 'it was Gara who said it, it's Gara's stuff'. Since he is the only one in the family who proved successful in the white man's world, his brothers do not dare criticise him, even behind his back; but he is not considered an authority on language and culture. His younger brother JOVINO (born in 1960) is fluent in Tariana. He is the one who, in spite of having completed just six years of school, aspires to be the political leader of the Tariana people. He is a very good story-teller, and knows a lot about hunting and fishing, but not so much about the traditional culture. His Tariana is a representative of younger people's language, with Tucano calques and coordinating techniques. He often comes up with hypercorrections, and does not use the indefinite person marker. He is often not quite sure of his Tariana and seeks older people's advice. His wife, Gl6ria (a Piratapuya), does her best to communicate in regional Portuguese - rather than in Tucano - with her children; Jovino rarely does this. He speaks Tucano with his wife, children and younger siblings, Piratapuya with his mother and Tariana with his father and elder brothers. He has recently started using some Tariana when speaking to his eldest son, Helio. Candido's youngest son JosE Lutz (born in 1969) is a trained schoolteacher. He has a job as the librarian at the secondary school in lauarete (and has been chosen by the school principal to teach Tariana). He is Ouent in Tariana; however, he does not know much about the traditional life and practices, but is eager to learn: there are a few archaic fonns used by Candido that he is not familiar with. He is a very good storyteller. But the stories he tells are mostly 'white man stories' - such as Little Red Riding Hood and numerous stories about a man (usually not loved by women) who managed to marry a king's daughter after a series of trials. He speaks the younger people's variety, similar to Jovino (except for hypercorrections); be often appeals to Candido for advice about 'how to say it correctly'. He is probably the greatest zealot of the language 'norm' of aJJ the Tariana. 22 I The language and its speakers The youngest sibling, MARIA LoURDES spent a long time working as a maid in Curitiba, in southern Brazil. She spent the best part of her early childhood living with nuns. As a result, she can understand Tariana but is too shy to speak it (see § 1.2, on the high level of language proficiency required for people to be able to 'open their mouth' in a language). She has only just come back from southern Brazil to live with her parents and elder sister in Iauarete. Leonardo's eldest son RAFAEL (born in 1973), a teacher at the primary school at Santa Rosa, is rapidly developing his skills in Tariana, becoming more and more fluent (with the help of his father). He can tell a coherent albeit simple story in Tariana, and carry a good conversation. He was instrumental in co-teaching a course in Tariana for nonspeakers and semi-speakers, and translating numerous Brazilian songs and Catholic hymns into Tariana. His Tariana is full of Tucano calques. He speaks Tucano with his wife, Ednalucia, and Tucano, Portuguese and Tariana (in that order) with his young children. He and Jose Luiz were instrumental in teaching the Tariana language course in June 2000. Jose Manoel's children who Jive in Santa Rosa, SEBASTIAO (Saba), and CRISTIANO (Kiri) (both about 30), are reasonably fluent in Tariana. Saba - with whom Jose Manoel used to live - speaks better Tariana than Kiri; he has a good knowledge of hunting and fishing practices and is a reasonable story-teller. Due to his father's influence, he displays a significant degree of vowel reduction, and does not use 'the right evidentials'. In addition, his stories are full of Tucano calques; he tends not to employ the indefinite person marker and prefers analytic possession to synthetic (see Chapter 6). Kiri - with whom Jose Manoellives now - is an average story-teller. He seems to have difficulties in communicating in Tariana, and speaks only Tucano with his peers. He enjoys great Jose Manoel's daughters respect in the village, having been elected its chief ('capit~o). CLEMENTINA and JOANINHA who live in Iauarete and are married to Tucano men there, are good speakers: both complain of getting a bit 'rusty' for want of people with whom to speak Tariana. Jose Manoel's youngest son Nu (short for Joao) lives with his two other sisters (whom I haven't met) in Balaio, a Tucano settlement. Ile is a competent speaker and seems to have acquired a taste for writing in Tariana during the Tariana language course and pedagogical workshop in June 2000. Ricardo Brito's two sons (both about 30) live in Santa Rosa while his other four children are scattered over the Vaupes area. EMiLIO is one of the best story-tellers in the village; he is fluent and witty. He has a vast knowledge of the local flora and fauna. His Tariana is said to 'be influenced by HohOdcne' (a dialect of Baniwa of l9ana), of which he seems to have some knowledge: be consistently uses -pida ('present reported') instead of recent past and remote past reported, omits the indefinite person marker and uses a reduced system of classifiers. Similarly to the speakers of the Periquitos variety, he distinguishes a general animate numeral classifier (e.g. paile 'one (animate)') from a general inanimate one (e.g. paita 'one (inanimate) man-made object'), unlike the mainstream speakers of the Santa Rosa dialect who use the general inanimate form to cover animates and inanimate man-made objects (paita 'one (animate or man-made object)'). His wife, Silvana, understands basic Tariana (she is Tucano, but her mother was Tariana). His younger brother RAIMUNDO is a very good speaker, but he does not feel confident enough to tell a full story in Tariana. His Tariana is very similar to that of Emilio; they both speak Tariana and also Tucano with Dika, their father, and nothing but Tucano with their wives and children. I . 9 Materials and speakers 23 JuvENAL (Juvcnn), like his elder brother ABELARDO (Ave), had hcen thought by everyone not to speak any Tariana at all. Their father Jum (Americo Brito's younger brother) had died when they were little, and their mother Amelia brought them up aH on her own. Amelia (a Guanano herseU) - who now lives with Ave - speaks good Tariana. However, a couple of weeks after [ had arrived, Juvena volunteered to tell a story in Tariana, and did it very well, to the surprised appreciation of all the others. Juvcna's level of competence is similar to that of Rafael, and his lexical knowledge seems to be good; he is now very keen to speak the language. His eldest daughter, Anastasia, is also keen to learn the language, and while l was in the village, made an attempt to speak it. Juvena's elder brother Ave, the deputy-chief of the village, does not speak Tariana, but does understand the language; he was beginning to speak it when 1 had to leave. Abelardo's wife, CRISTINA (Kiri), is the daughter of Candido's late second cousin Julia married to Albino (a Hohodene; see below). They used to live close to Santa Terezinha on the Iauari river. Cristina is fluent in Tariana and in HoMdene, as well as in Tucano, but she is not a good story-teller. Her Tariana is a mixture of the Santa Rosa and Santa Terezinha varieties, with a strong influence from HoMdene: she uses Santa Terezinha 3sgnf prefix ri- instead of Tariana di-, and Santa Terezin.ha -peta instead of Santa Rosa -pita 'again'. Clearly she was having difficulty keeping her Baniwa HohOdene and her Tariana apart. She speaks HohOdene to her father, Albino (now about 100 years old). ALBINO is a fluent speaker of the Santa Terezinha dialect ofTariana, and of Baniwa Hohodene; he is still very active, but would not tell a story in Tariana because 'it is not his language'. Cristina's eldest daughter, lRENI (about twelve years old), is the only child in the village who can understand and say simple things in Tariana. She was born before Cristina married Ave, and her father is unknown, so, according to Olivia, Albino took the place of her father: he treated her as an ethnic Tariana and spoke Tariana to her (and continues to do so); this explains her knowledge of the language. JACINTO - who lives in the same house as Cristina, Ave, Amelia and Albino - is the son of Eduardo Brito (Candido's second cousin) and his wife Antonia (Hohodene). He speaks fluent HohOdene and Tucano, and also passable Tariana: he is considered a competent speaker. Jacinto is very shy (which is why whenever urged to tell a story he would simply go away), but, in spite of his young age (he is no more than 20), he is a very experienced hunter and fisherman. His Tariana is the Santa Terezinha, and not the Santa Rosa, variety. I briefly met Eduardo Brito, his wife Antonia (who live in the community of S~o Pedro on the Vaupes), and his children Marta and Laureano. Eduardo Brito is a fluent speaker. His wife and children arc fluent in Tariana; their Tariana is an interesting mixture with IIohOdcnc. The majority of the population of Periquitos (about fifty people, including smaJI children and women) are Ouent in Tariana. The three elders of the community, all in their mid-fifties - Marino, Jorge and Maximiliano Muniz - make it a point to speak nothing but Tariana with their children, and so do their children with their own children. In particular, all Marino's children - from 27-year old Domingo, the vice-chief of the village, to 10-year old Bosco - are very fluent. Lauro Muniz, a local school teacher, recently compiled an alphabet book in the Tariana of Periquitos (which 1 was asked to check for 'spelling mistakes'). The Periquitos elders are less eloquent story-tellers than the Britos and know less about their origins. For instance, they had difficulty remembering the name of their grandfather and even some of their own sacred names. However, the younger generation - in particular the village chief lsmael, the vice-chief 24 I The language and its .l'peakers Domingo, and Batista Muniz, all in their twenties - appear to be much more languageconscious than their peers among the Santa Rosa Tariana, who do not speak the language. They are most eager to maintain the language in their homes and to create their own program of language teaching. 2 Phonology Tariana has a large inventory of phonemes, compared with the closely related Baniwa of lyana and with neighbouring East Tucano languages. Tariana phonology bas undergone a strong areal impact from Tucano (see Aikbenvald 1996a and 1999c). 2.1 Segmental phonology The basic syllable pattern in Tariana is (C 1)V(C2) where C2 can be h, y and rarely n, and C, can be any consonant (see discussion in §2.2.1 ). A notable feature of Tariana is that phonotactic restrictions on the occurrence of consonants and of vowels depend on the type of morphemes (roots, affixes and clitic) summarised in Tables 2.2 and 2.4.' 2. I . I Consonants The phonological system of Tariana consonants is shown in Table 2. I. Consonants differ in frequency and in phonotactic restrictions (summarised in Table 2.2 at the end of this section). Tariana has labial, apico-dental and velar stops, with voicing distinctions only in the labial and dental series. Voiceless stops can be aspirated or unaspirated. In the remainder of this section l discuss the occurrence of consonants. A. STOPS. Voiceless bilabial stops- unaspirated and aspirated - occur in roots in initial and in medial position, e.g. pu:we 'capuchin monkey', pumeniperi 'sugar', dipe 'his meat'; kzJ:phe 'fish'; phirimftji 'cotton', phiripanakwari 'nightingale', yirphini 'thing'. The unaspirated p occurs in the iniLial position in suffixes, e.g. -pi 'classifier: long thin things', while ph occurs in both initial and medial positions, e.g. -phe 'classifier: leaflike', -mapha 'classifier: completely covered'. In enclitics, the phoneme p is found only in the enclitic-initial position, e.g. =pida 'present reported', =pada 'contrast',1 while its aspirated counterpart occurs only medially in one Aktionsart enclitic, =Jiphe 'hold firmly'. 1 Along similar lines, many Afroasiatic languages, for instance, Hebrew, distinguish between consonants which occur in roots, and those which occur in affixes (called 'servile', by medieval Hebrew grammarians (see e.g. Gesenius 1984: 83)). That is, phonological restrictions on occurrence of consonants depend on the type of morpheme. 1 In lhe enclitic =iupe 'diminutive plural', whose singular counterpart is =11iki, pe is a plural sumx, and not part of the enclitic. 2 Phonology 26 TABLE2.1 CONSONANTS BILARIAI. VOJCf:t.ESS STOP ASPIRATED VOICELESS STOP VOICED STOP p ph b ASPIRATED VOICED STOP APICQDENTAL AI'1CQ-LV~ PALATAL s m mh n n nh nh I LATh1tAL ASI'IRAlED SEMl·VOWEl. h f FLAP SEMI-VOWEL GLOTTAL tf VOICELESS AFFRICATE ASPIRATED NASAL OORSQVELAR k kh (g)_ I th d dh VOICELESS FRICATIVE NASAL !.AMINOPALATAL w y wh The voiced bilabial stop b is found in a few noun roots in the root-initial and rootmedial position, e.g. bamrina 'jungle mango (Calophy/lum brasiliensis, Jam. Guuiferae)', bibl:kha 'a type of curved vine', and in a few verbal roots in root-initial position. Most of these roots are Tucano loans (see §26.1.3), e.g. Tariana -besilti 'choose, revise', Tucano bese 'choose', Tariana -boleta 'defeatber', Tucano bure 'defealher, husk'; Tariana -bueta 'teach, learn', Tucano bu'e 'teach, learn'; Tariana -bueta 'blow fire', Tucano bue 'blow fire, smoke-dry'; Tariana -bata 'swing', Tariana -batuta 'swing with force'. This phoneme is also found in the initial position of the conditional enclitic =buhta, =bohta (which mjght also be a loan from Tucano: see §26.1 .3), in the epistemic ba, a loan from Tucano (see §25.5.3) and in the following Aktionsart enclitics (which probably go back to grammaticalised verbal roots: see § 15.2.2 and Table 15.3; and Aikhenvald 2000b): -bQsa 'break by smashing, split', =bili 'inside', =bala 'in every direction, everywhere' and =bisi 'into fire'. The phoneme b is rare in the majority of the North Arawak languages of the area (see b is restricted to a few lexical Aikhenvald 1999b). For instance, in Baniwa of ) ~an, items, at least some of which could be onomatopoeic, e.g. bukukuri 'owl'; it never appears in any grammatical morphemes. Voiceless apico-dental stops, unaspirated and aspirated, are found in roots, in affixes and in enclitics, both in initial and in medial position, e.g. tariwa 'traditional flute', turopa 'cone', di-tuda 'he breaks', -ite 'animate noun class marker', =ta 'again'; tlui:ro 'butterfly', thuime 'all', pethe 'manioc bread, beiju', -kuthe 'classifier for manioc bread', =tha 'frustrative', =tharil'precative'. The voiced unaspirated dental stop occurs in initial and in medial position in roots, in affixes and in enclitics, e.g. de:ri 'banana', dolo 'a fly', nu-dia 'l go back', nu-awada 'I think, I remember', -du 'feminine', =da 'apprehensive', =dakiJ 'yet (aspect enclitic)'. In informal speech, d alternates with r before or aficr a front vowel. Thus, =pidana 'remote past reported' is optionally realised as =pirana, negation -kade, -de (see B under §2.5.3) as -kare, or -re, and kida, kayda 'ready, finished' (see A under §2.5.3 on variation between ai/ay and i) as kira and kaira. This alternation is the result of a Tucano influence: in Tucano and most other East Tucano languages the intervocalic d surfaces as rafter a front vowel (Ramirez 1997, vol. I: 31 ). 2.1 Segmental phonology 27 The dental voiced aspirated stop dh has a limited distribution. Most often, it occurs as the product of the phonological process dV- + -h on a prefix-root boundary in wordinitial position, e.g. dhipa +- di-hipa (3sgnf-grab) 'he grabs', dhupa +- du-hipa (3sgfgrab) 'she grabs' (see §2.5.1). The Aktionsart enclitic =dha/Q 'unstick' with the corresponding verb -dha/a 'get unstuck', and the enclitic =dhuli 'be folded into two' (which does not have a corresponding verb) are the only instances of a morpheme-initial dh that does not come from a sequenced and h. Their etymology is unknown. The voiced velar g occurs only in loans from Portuguese, e.g. Graciliano, Gara (short for Graci/iano), Gabrie/'Siio Gabriel da Cachoeira'.1 Voiceless velar stops -aspirated and unaspirated - are found in roots, affixes and enclitics, in initial position and in root-medial position, e.g. kuheni 'crab', kali-si 'story', nu-kawa-na 'my leg', -ku 'classifier: folded cloth', -maka 'classifier: extended cloth', =kG 'sequential marker', =sil<il 'present inferred', khewaka 'deep', kulekha 'fishing line', -khi 'classifier: thin curved objects', =khuli 'wag one's tail' (Aktionsart enclitic). Kh is not found in aflix- or enclitic-medial position. B. FRICATIVES AND AFFRICATES. The voiceless apico-alveo-palatai fricatives occurs in initial and medial positions in roots, suffixes and enclitics, e.g. sipi 'blowgun', isa 'smoke', pusila 'clearing in jungle', -sata 'ask', -si 'non-possessed noun form', -se 'locative', =sa 'tightly', =bisi 'into fire'. The voiceless !amino-palatal affricate tf occurs in root-initial and -medial positions, e.g. iljida 'turtle', itfa 'hair', /fori 'man', atja 'men', matfa 'good', tfa 'Oh!', but not in affixes or enclitics. The contrast between s and tf tends to be neutralised before high vowels in polysyllabic words, e.g. nuketjipe, nukesipe 'my blood relatives', kuitji, kuisi 'mutum bird', rjipare, sipare 'urine'. In some frequently used words, rji is never pronounced as [si], e.g. tjinu 'dog' or irjiri 'game'. The phonemes tf and s contrast before i in a few examples, e.g. irji 'howler monkey', isi 'oil, fat', karutfi 'a smaller kind of aracu fish', Icarus! 'a bigger kind of aracu fish', kUtfiru 'a kind of evil spirit', kusim 'drum', and before a, e.g. isa 'smoke, wound', itfa 'hair'; they tend to be neutralised when there is no minimal pair.' The phoneme s has an allophone [ts] before u and o, e.g. sui-te, tsii-ite 'small (animate)', supe, fslipe 'type of fish', siideri, tsiideri 'edible caterpillar', some, /some 'a lot', and before a sequence of front vowels, e.g. sie, tsie 'firewood'. The phoneme tfcan be pronounced as [ts] before u and i, e.g. matji:te, matsi:te 'good (animate)', tjinu, tsinu 'dog', pa-tjimari , pa-tsimari 'one's son-in-law', pa:tji, pa:tsi 'other (of different kind)', tfUniri, fsliniri 'real one'. The only example of tfbefore o is the interjection tfo! 'Oh! l Just a few other Nonh Arawak languages have g as a phoneme. Its origin in Warekena of Xi~ is unknown (Aikhenvald 1998a), while in Reslgaro g corresponds to r or r in other Arawak languages (see Aikhenvald forthcoming-b). • This variation of sibilant phonemes in Tariana can be accoun1ed for by areal pressure from Tucano languages, most of wh1ch have just one sibilant phoneme (s), and also by language obsolescence, which typically results in excessive variability of segments (see Campbell and Muntzel 1989). Those who usually speak only Tucano at home tend to have more free variation between (and s; for instance, Batista Brito, an older speaker who does not have much opportunity for using the language, replaces (with s even in frjlti 'game'. Even Leonardo Brito, one of the most fluent speakers and a sophisticated story-teller who speaks Tariana at home about halfthe time, occasionally pronounced -smt 'also' as - rjim. 28 2 Phonology (surprise)'; this interjection is never pronounced as • W! There is no contrast between s and tjm the Periquitos dialect (see Appendix).s The glottal fricative h occurs in root-, affix- and enclitic-initial position, e.g. hek1ina 'tree', hanuite 'big (animate)', -hipa 'grab', -hupa 'wash', -hyu 'non-visual purposive', =hala 'be open' (Aktionsart enclitic), =hu 'away'. It appears in medial position in personal pronouns, e.g. diha 'he' (see Tables 2.6 and 8.1 ), in prefixed verbs in a slow speech register, e.g. di-hilitu (3sgnf-tish) 'he fishes' (normal speech register form is dhilitu), and in a few isolated lexemes, e.g. -duiha 'dislike', kehuri 'a type of ant' and kuheni 'crab'; it can also occur in coda position (sec §2.2.1 ). There are no roots with two h phonemes. The root-initial h is often not pronounced in rapid informal speech; thus, hema 'tapir' and halite 'early in the morning' often surface as [ema] and (alftej. The contrast between h and o word-initially can be illustrated with a few pairs like halite 'early in the morning' and alite (a variant of ali:te) 'the existing one'. C. NASALS. Unaspiratcd bilabial and dental nasals occur in all positions in roots and affixes, e.g. m: mawari 'anaconda snake', -mzikhua 'rinse (mouth)', amaku 'hammock', ~ rna 'classifier: feminine', -yami 'classifier: piece of cloth'; n: kni 'water', nawlki 'person', fnam 'woman', -na 'want', -ni 'topic-advancing derivation', -kwana 'classifier: plain objects', =yana 'pejorative', =na 'remote past visual'. The phoneme m occurs only in initial position in enclitics, e.g. =mana 'while', =mikiri 'nominal past singular nonfeminine'; n occurs in all positions in all types of morphemes. The lamino-palataJ nasal n occurs in root-initial and in root-medial positions, e.g. ntipi 'bone', namu 'minor evil spirit', newi 'oner,' lne 'major evil spirit', saiiape 'sweat', - ·nare 'disappear',6 -ken(w)a 'begin', and in just two enclitics, =nil 'step on something and feel pain' and =mana 'while'. Jt never occurs in affixes. This phoneme is in free variation withy if followed or preceded by a labial consonant, e.g. yapi, napi 'bone', yapu, napu 'stream', yama, nama 'two' and pamuya, pamui'ia 'middle'. There is no variation if a word contains two n, e.g. nona 'mad[ tish'. Each nasal in Tariana has an aspirated counterpart - mh, nh and nh.' Of these, mh occurs in root-initial and root-medial position in about a dozen morphemes, e.g. mhdida 'few, prohibitive', kO:mheru 'cucura fruit', wademhe 'later on', kemhani 'the one who walks'. This phoneme also occurs in initial position in just two enclitics: =mha 'present non-visual' and =mhe 'admirative'. The aspirated nasal nh occurs in the initial and medial positions of a few roots, e.g. nhtiri 'cacuri' (a type of fish net), nhesiri 'like', -tinhi 'understand, feel', and in the initial position of a few verbal enclitics, e.g. =nhi 'anterior', =nha 'visual present interrogative', =nhina 'non-visual apprehensive'. It does not occur in any affixes. I I s This pronunciation difference was the object of jokes by the speakers of the Santa Rosa variety, who maintained that the Periquitos variety underwent a stronger Tucano influence than their own. • The sign · indicates that the stress falls on the syllable preceding that of the root, e.g. mi-nare 'I disappear'. 1 A characteristic feature of North Arawak languages of the Ieana Vaupes is the presence of aspirated nasals. ( 9 v 2. 1 Segmental phonology 29 The aspirated lamino-palatal nasal iih is very rare: it appears in just three roots - two homophones -nba 'eat' and its derivatives,• -nha 'point at', and kuphiiiheda 'tumour'.9 The aspirated nasals often arise as a result of the phonological process of hmetathesis on the prefix- root boundary, e.g. ma-hema-kade (NEG-hear-NEG) - mhemakdde 'does not hear', nu-huma ( I sg-hear) - nhuma 'I hear', i'lama-hipa (two-CL:HUMAN) -+ nhamepa 'two (people)' (see §2.5.1 ). D. LIQUIDS. Tariana has two liquid phonemes: lateral I and flapped r. The apico-alveopalatallaterall appears in a limited number of lexical roots with unknown etymology, in word-initial position and word-medial position, e.g. leka 'break', lesa 'boil', /ape 'mud', dolo 'fly', kesole 'brown, muddy', -pala 'put', -sole 'carry', yakaso/o 'necklace', ta/ama 'flute', and in a few Aktionsart enclitics in medial position, e.g. =hala 'be open', =billa 'in every direction', =holo 'spill water', =khuli 'rub against someone in a friendly way, e.g. a cat; slip out of hand', =liphe 'hold firmly', but not in other grammatical morphemes. This phoneme also occurs in rare loan words from Indo-European languages, e.g. /Qma 'be aflame' (possibly, from Spanish llama 'flame') and lapi 'pencil' (from Portuguese lapis). The flapped r occurs in initial and medial positions of roots, suffixes and enclitics, e.g. -kurisa 'gnaw', thluo 'butterfly', ke:ru 'angry', -ri 'son'; -ri 'relativiser', -kari 'past participle', =weri 'crush by compressing (with a hit, or a bite)' (an Aktionsart enclitic: §15 .2.2 and Table 15.3). The flapped r does not occur in word-initial position, the only exceptions being riki 'pus' (which has a variant diki) and ritena 'more', also realised as ditena or as litena (see §18.5). It should be noted that the Proto-Arawak third person singular cross-referencing prefixes • m- 'third person singular feminine' and • ri- 'third person singular nonfeminine' (cf. Payne 1991; also cf. Baniwa of I<;ana m-, ri-) result in du- and direspectively in Tariana. Word-initial r is realised as d even in personal names of a Portuguese origin. One of the speakers whose full name is Ricardo is usually referred to as Dika. (See under D in §3.1.2. on the syncope of the last syllable and the coda of the second syllable characteristic for the formation of vocative forms in Tariana.) Similarly, in Tucano, the initial r or I of Portuguese loans is usually realised as rJd (prenasalisation is optional), e.g. Dosa, from Portuguese Rosa, or Dui, from Portuguese Luiz (see Ramirez 1997, vol. I : 32). (Portuguese pecado 'sin' was borrowed into Tariana as -pekam 'joke, play, mess around': see §26.1.3.) I mentioned in (A) above that d can be realised as [r] before or after a front vowel. Younger speakers occasionally pronounce r as d between two high front vowels, especially in Portuguese personal names, e.g. Emilio is pronounced as [Emirio] or as [Emidio]. The phoneme r optionally alternates with [d] in root-medial position if the preceding syllable contains a dental stop; for instance, nu-matara 'I leave, I stop' can be pronounced as [numatada) and =thara 'prccative' as [=thada]; or if the preceding syllable 1 This is one of the most stable roots in all the Arawak languages; it goes back to Proto-Arawak •mka (Payne 1991 ). See §26. 1.1. 9 The phoneme nh is also comparatively rare in Baniwa of !~an (c( Taylor 1991 ). Its comparative rarity in Tariana may be accounted for by language obsolescence phenomena and areal pressure of Tucanoan languages, which results in partial substitution of cognates with Baniwa, and subsequent 'loss' of items phonemes. which contain 'original' l~an-Vups 30 2 Phonology contains a flap; tarada 'alive' can be pronounced as [tarar<i] and nu-kurira 'I paint' as [nu-kurfda]. These phonetic processes apply more frequently with the speakers who have little, or no opportunity of speaking Tariana at home. The flapped r is often realised as a vibrant r before a high vowel, e.g. [inaru) 'woman', (karu] 'dangerous', [hfparu) 'toad', [ke:ri] 'moon', [itjiri] 'animal', [kt!:ru] 'angry', [kuruna) 'queen bee', [yapirikuri] 'traditional God' (lit. the one of the bone). Elsewhere r is the only realisation, e.g. [tarada] 'alive', [kamara] 'tcrmite',lmaredape] 'large (plural)', [yapuratu] 'long flute', [yurema] 'cobra', [yarak.i] 'caxiri or manioc beer' (the latter two words are probably borrowed from Nhecngatu). 10 If a word contains two flaps, the first tends to be pronounced as [1], e.g. warikiri 'young man' often surfaces as [walikiri]. E. GLIDES. Tariana has three glides: bilabial unaspirated w, aspirated bilabial wh and !amino-palatal y . The unaspirated bilabial w occurs in root-, afftx- and enclitic-initial and -medial positions, e.g. yawi jaguar', mawina 'pineapple', -dawa 'escape', wa- 'I pi prefix', -yawa 'classifier: hole', =wa 'postponed imperative', =wen 'smash'. This phoneme is optionally realised as a labio-dental fricative v before front vowels, e.g. [yawi], [yavi] 'jaguar'; [wika), [vika] 'on top of'; (pa:wc], (pa:ve] 'together'. The aspirated bilabial glide wh is found in initial and medial position in a few roots, e.g. whyume 'last', tawhi-pi 'macauba, Acrocomia aculeataJam. Palmae'; whetu 'type of snuft', ka:whi 'wake up, be early in the morning', kawhi 'manioc flour', and in affixinitial position in a few affixes, -whya 'classifier: canoes', -whi 'classifier: particles'; it is found in just one enclitic in medial position =kawhi 'do early in the morning'. ln word-medial and word-initial position wh is the result of the phonological process of 'b-melathesis' on a morphological boundary, e.g. Yawi-hipani (jaguar-waterfall) --> Yawhipani 'lauarete, lit. Jaguar's waterfall' (see under B in §2.5.1). The palatal glide y occurs in root-, affix- and enclitic-initial and -medial positions, e.g. ye: 'armadillo', yaka 'shrimp', piriya 'avocado', hawaya 'inga-fruit', eyawa 'firm ground', -yawa 'classifier for holes', -kuya 'classifier: river shore, straight stretch of river', =ya 'emphatic'. In word-initial position y can be realised as dy or t:ff before a nonlow vowel, e.g. yLitu, dyUtu, ctzJtu 'larva', nu-yena, nu-t:ffena 'I exceed'; y may be pronounced as !(before e in nu-tflna. 10 Emilio Brito, a fluent young speaker who was brought up in close contact with Baniwa of !~ an and with the Santa Terezinha dialect ofTariana, occasionally pronounces rasa voiced alveopalatal fricative 3 before i, e.g. di-mlna-n' 'his master' comes out as [ di-mfna-3i], a pronunciation typical for Baniwa of ! ~an and for the Santa Terezinha dialect. TABLE 2.2 PHONOTACTIC RESTRICTIONS ON THE OCCURRENCE OF CONSONANTS ROOTINITIAL AFFIXINITIAL ENCLITICINITIAL ROOTMEDIAL + + + + + + + + + + + + ( +) onJy noun roots + d + + + + + + dh (+)one root - k kh + + + + p ph b t th s tf h m rnh n nh ii iih + + + + + + + +(two roots) + + - - + + + + + + (+) two enclitics - - - + + + + + + ..L - - + + + - + + (+)two enclitics + + (+)enclitic + - ENCLITICMEDIAL + - + + (+) AFFIX- MEDIAL - + + restricted + + + + + +one root - - + + - - COMMENTS mostly in loans T + occurs word-initially as a result of h-metathcsis - often as a result of h-metathesis + - - + one enclitic + + + + + + - r + + 7 I w + + - - + + + + wh + + - + - (+) one enclitic y + + + + + + often as a result of h-metathesis only in three roots does not occur in word-initial position result of h-metathesis wordinitially and word-medially 32 2 Phonology The glottal stop, a phoneme in a number of East Tucano languages which is absent from North Arawak languages of the Upper Rio Negro, sporadically appears wordfinally after a, e.g. di-pa/a, di-pala? 'he put'. This is typical of some younger speakers who normally speak Tucano in their homes. Phonotactic restrictions on the occurrence of consonants in root-, affix- and encliticinitial and -medial positions are summarised in Table 2.2. I did not include any information on the occurrence of the phonemes in proclitics, since these are a very small class in Tariana (see §2.6). The symbol (+) in Table 2.2 indicates that the phoneme occurs in a limited set of items. Table 2.2 shows that all the aspirated consonants, especially the aspirated voiced stop, aspirated nasals and the aspirated bilabial glide, have a restricted distribution. Since most Tucano languages have no aspirated consonants, this may be due to the areal influence from Tucano. 2. / .2 Vowels The Tariana vowel phonemes are shown in Table 2.3. TABLE 2.3 VOWELS FRONT CLOSE-HIGH MID-CLOSE OPEN CENTRAL SIIORT NASALIS ED LONG SHORT i e i t: i e: 0 e NASALISED BACK LONG SIIORI NASALIS 1m LONG u ii u: a a a: 0 Every vowel, with the exception of i, has a nasal counterpart; neither i nor o has a long counterpart. Mid-central o and oand high back u, ii and u: are rounded. High front short vowel i occurs in root-, affix- and enclitic-initial, -medial and -final positions, e.g. iri 'blood', ira 'need', inam 'woman', hipam 'toad', ml-ira 'I drink', mariye 'knife', keri 'sun, moon', -ita 'inanimate classifier', =niki 'completed', =khani 'straight away'. Its long counterpart, i:, is found only in roots in a non-final position, e.g. i:-peri 'stinking', hi:ri 'rat'. Nasal 7is found only in root-medial and in root-final positions, e.g. sipi 'blow gun', hi'demonstrative animate', af'here'. Mid-front short e is found in root-initial position, and in root-, affix -and encliticmedial and -final positions, e.g. episi 'iron, stone axe', eyawa 'firm land', kepiria 'bird', kuphe 'fish', -ne 'plural'; -se 'locative', =thepi 'in(to) water', =ne 'instrumental'. In rootfinal position it has an allophone e, as in sic 'firewood' and tc'until'. Its long counterpart is found only in roots in non-final position, e.g. e:nu 'thunder', de:pi 'night', ne: ri 'deer'. Nasal e is found only root-mediaJiy and root-finally, e.g. yida 'mortar', -weya 'buy', meda 'in vain, counter-expectation', nuhepore 'my navel', ate 'big anteater', iheda 'blind snake'. High central i is extremely rare; it is found just in the augmentative enclitic =pi (possibly a Tucano borrowing and a variant of - pu: see §9.6) and in onomatopoeic ihmeni 'moan' (see also §2.7 on how this sound occurs in pausal forms). It appears as an occasional allophone of i in marawati, marawati 'a type of snuff', hilisi, hitisi 'tear', and -pit£, -piti 'chase away, kick'. Its emergence in Tariana is probably due to areal diffusion from Tucano (see Aikhenvald J996a). Mid-central short o is rare; it occurs in root-medial and root-final positions, e.g. d6lo 'fly', -holeta 'defeather', -khorena 'servant', and in the enclitic-medial and cnclitic-fmal 2.1 Segmental phonology 33 positions, =bosa 'break by smashing, split', =holo 'spill (water)', =kholo 'roll over'. The dialect of Santa Rosa shows free variation between o and u in the feminine derivational suffix -do, -du in nu-ketji-do, nu-ketji-du 'my feminine relative' (see §2.3.5). Its nasal counterpart o is quite rare; it is found in medial and final positions in the following roots: toke 'firefly', siwi.rikorena 'tapiriri, Tapirira guianensis, Jam. Anacardiaceae', nuito 'daughter! (vocative form)' (a minimal pair with nu-itu 'my daughter'), and -toreta 'roll into a thin roll, like a cigarette'. It appears in word-initial position in place-names of Tucano origin, e.g. Ororeana. TABLE 2.4 PHONOTACTIC RESTRICTIONS ON THE OCCURRENCE OF VOWELS i 1: j ROOTINITLAL AFFIXINrnAL ENCLITICINITIAl. ROOTMEDIAl. AFFIXMEDIAL ENCLITICMEDIAL ROOTFINAL AFFIX- OR ENCLITICFINAL + + + + + + + + + + + + + + - e e: ..e I + + + 0 - •o - u u: + ii a a: a + + + + + - - + + - - - - + + + + + + + + - + - - + - - - + - + + - + ~ - - + + -1 t-one enclitic - + - + + + - + + - - - - - + - + + + + - - The high back short u can occur in any position, except for the beginning of suffixes and enclitics, e.g. pupu 'a type of mosquito', kiJwhe 'woodpecker', uritu 'blue-winged parrot', -uka 'arrive', =tupe 'diminutive plural', =pu 'augmentative'. Its long counterpart u: occurs in root-initial and root-medial positions, e.g. u:ni 'water', yu:pia 'spider web', pu:we 'monkey (macaco prcgo)'; this phoneme never occurs in affixes or in enclitics. Long u: is optionally diphthongised, e.g. hu:site, huisite 'unlucky in hunting'. The nasal u is found in just a few roots, e.g. ui 'port', uhil-nawiki 'llohOdene (subgroup or Baniwa ofl9ana)', -yil'go up(strcam)', -yilta 'put upwards'. The low back short a is found in all positions, e.g. inam 'woman', -a 'go, say, give', awakcida 'jungle', -apa 'classifier: largish and long', =pidana 'remote past reported'. Its long counterpart is found only in roots, mostly in the non-final position, e.g. cl:pi 'anaconda snake', mawa:.ri 'snake', -da:na 'shadow', kci:whi 'be awake', -pita: 'paint', pa:pe 'maybe'. Nasal a is also restricted to roots, e.g. asi 'pepper', apa 'basket', tfli.ri 'man', ada 'grater', ate 'big anteater', mhaida 'few, prohibitive', -ya 'live, stay, wear', -wha 'bite'. The following minimal pairs confmn the phonemic status of long vowels: dinu 'he comes', di:nu 'he kills'; whepa 'we wash', whe:pa 'we answer'; eni 'spider', e:ni 'chier; nama 'they poison fish', na:ma 'they look for (something)'; duma 'she s leeps', du:ma 'she looks for (something)'. However, in informal speech long vowels arc often realised as II 34 2 Phonology short, and there is a significant degree of variation (e.g. u:ni, uni 'water', a:pi, api 'anaconda snake'). This could be the result of the areal influence of Tucano languages (which do not have phonological length) and of language obsolescence. Length distinctions tend to be maintained in the few cases of minimal pairs, e.g. numa 'I sleep' and nu:ma 'I seek'. An amusing illustration of how vowel length is being lost in Tariana comes from the treatment of an original minimal pair for long vowels: ka:whi 'be awake' (cf. Baniwa of T9ana ka:whi 'be awake') and lcawhi 'manioc flour' . According to my teachers, Tariana speakers used to playfully wake up their friends by knocking on a wall and saying in Tucano: po'ka, po'ka! Po'ka means 'manioc flour', and so the word play is based upon the loss of vowel length. All long vowels are stressed, and have high pitch, though not all stressed vowels are long (see §2.3). Phonotactic restrictions on the occurrence of vowels in root-, affix- and encliticinitial, -medial and -final positions are summarised in Table 2.4. Long vowels and nasal vowels do not occur in affixes and enclitics. Only short vowels a, i and u occur in prefixes. The large inventory of nasal vowels in Tariana is an innovation, since this is not found in other Arawak languages of the region, which usually have just four vowels, a, i, e and u. In a number of cases a nasal vowel in Tariana corresponds to a sequence of a vowel and a nasal consonant in Baniwa, e.g. Tariana -weya (variant: -wana), Baniwa -wen-ta, rapid register -weta, Warekena -wenita 'buy'; Tariana iii, Baniwa unai 'port'. 2.2 Syllable structure, vowel sequences and the problem of diphthongs 2.2.1 Syllable structure The basic syllable pattern in Tariana, as in most Arawak aanguages, is (C)V. Sy!Jablcs consisting of V are most frequently found in word-initial position. There are no restrictions as to the onset of syllables. Only y and h can appear in the coda position. The only instance of the (C)VN(asal) syllable structure is an idcophone tfin 'sound of a tinkling musical instrument'. The problem of glides in syllable onsets is considered under B in §2.2.2. The palatal y occurs syllable-initially, -media!Jy and -finally in roots, e.g. yeka 'rubber', ayni 'striped cuckoo', ayni 'wasp', kayni 'manioc', thuyme 'all', day 'be asleep', and in one enclitic =thily 'all'. The glottal fricative h appears in the coda position wordmedially in roots, where the onset of the following syllable is a voiceless stop t or k, e.g. -anih-16 'to think, to reason', ma-ra-tah-ka 'quick wave', ka-rah-ta 'lung', and in one enclitic =buh-ta 'conditional' (see 16.6). Syl tables of CV h-structure can arise as a result of the juxtaposition of two enclitics, one of which conta~s an aspirated stop or nasal (see §2.4.3), e.g. =-mah-kil 'recent past non-visual' (from =mha 'present non-visual' and =-kil 'recent past'), or =tah-kil 'frustrative recent past visual' (from =tha 'frustrative' and =ka 'recent past visual'). The fricative h in an unstressed word-initial or morpheme-initial syllable can appear in prevocalic or in postvocalic position, resulting in free variation between hV and Vh syllables, e.g. ehkwapi, hekwapi 'day, world', tarrida-ehkwapi, tarrida-hekwapi 'day of life, life' (cf. Baniwa of l9ana hekwapi 'day'), -hyume, -yuhme 'after:SS', hiporite, ihporite 'green (animate)', hiperiwi, ihperiwi 'type of hedge' (see under Bin §2.4.3). The appearance of CVh syllables in Tariana (not found in Baniwa) can be considered the result of the impact of Tucano languages, where CVh and/or CV? are the only 2.2 Syllable structure, vowel sequences and the problem ofdiphthongs 35 instances of a consonantal syllable coda (see discussion in Chapter 2 of Aikhenvald forthcoming-a). Vowel reduction and elision (see §2.3.5) in rapid speech can result in VC or eve syllables. The pre-tonic i in the verb -dusitti 'go back, swing on one's feet' is reduced to <J or to e in rapid speech. As a result of subsequent resyllabification, a CVC syllable arises: normal speech: rapid speech di-dusirti (3sgnf-go.back) 'he goes back' di-du-si-ta di-dus-ta When the pre-tonic i in the verbal clitic =sita 'accomplished' (§ 15.1.8) is reduced to e, a <J is inserted on the clitic boundary; this results in the creation of a phonetic VC syllable: normal speech: rapid speech di-pitci=kQ=sita (3sgnf-bathe=REC.P. VJS=PERF) 'he bathes' di-pi-ta-ka-si-ta di-ptak~s Asterisked vowels have restricted occurrence. 2.2.2 Vowel sequences and diphthongs In (A) we consider vowel sequences and their realisations. The problem of whether there are diphthongs in Tariana will be discussed in (B). A. VOWEL SEQUENCES AND THEJR REALISATIONS. Vowel sequences in Tariana cannot involve long or identical vowels. The following vowel sequences involve all short oral vowels, except for i: • ai, e.g. pa/'father!' (vocative)', kai 'ache'; • ia , e.g. dia 'he said, he went', -pia 'classifier: times'; • iu, e.g. diukti 'he arrived', -piu 'classifier: times' (variant of -pia); • ie, e.g. diema 'he stood'; • ua, e.g. dua 'she said, she went', kuripua '(there is) nothing'; • ui, e.g. hui 'like (food)', nuitu 'my daughter', hamlite 'big (animate)'; • ue, e.g. duema 'she stands', dhuepa 'she answers'. There is just one instance of the sequence ao- kaokUthe 'a type of manioc bread'. No other vowel sequences involve o. Sequences ei and ea are not attested; au is found only in hriu 'yes' and eu occurs only in two ideophones weu 'the sound of a rope winding', and koleu 'imitation of a bird's voice'. Sequences which involve nasal vowels are: ii-i, as in hdida 'l don't know', mhiiida 'prohibitive, few' and mhdisiki 'hunger, be hungry', and iii found only in iii 'port', and ei1 found in -mheu (a free variant of -mhe 'admirative'; see §19.6). These vowel sequences form two separate syllable nuclei. ln a very slow speech register each vowel is pronounced separately, and each of the vowels in the sequence can be stressed, depending on the stress pattern for each individual word, e.g. pai 'father!', but kai 'ache' (distinct from kay 'thus', a monosyllabic word of eve structure). The ftrst or the second vowel in a sequence can be reduced (u -+ u, i -+ i if unstressed, 36 2 Phonology e.g. ditr, dra 'he said, he went'; ia, ra 'you (pi) go', diukil, diului 'he arrived', hanuite, hamiTte 'big: animate', paikil, paikil'father (vocative of consideration)' (but cf. pai 'Dad (vocative)', not *pay), lulu, hau" or haw 'yes', kuripua, kuripiia '(there is) nothing'. There is an additional difference between a vowel sequence a-i and a diphthong ay. ln rapid informal speech a syllable ay becomes ey or e(), e.g. htiyku, he()ku 'tree', daykina, deykina, de()kina 'in the afternoon', hci-nay, himey, heine: 'this lake', hUnay, himey, hune(:) 'a type of potato', hipay, hipey, hfpe() 'land' (see also A under §2.5.3). This process does not apply to the vowel sequence a-i. The vowel sequence iii undergoes monophthongisation and becomes e with younger speakers, that is, mhiiida 'prohibitive' is pronounced as mheda and mhiiisiki 'hunger, be hungry' as mhisiki. There are no vowel sequences which contain long vowels. One would expect these sequences on a morpheme boundary when there is a prefix which ends in a short vowel, if the root starts with a long vowel (sec §2.5.1 ); note that in this case the rules for vowel fusion do not apply. In fact, the second vowel gets shortened, and a V 1 1- V2 : sequence becomes V 1 + V2, e.g. ai in na-itu (underlying form: na-f:lu) (3pl-steal) 'they steal' or na-inu (underlying form: na-i:nu) (3pl-kill) 'they kill', ui in dti-inu (from du-i:nu) (3sgfkill) 'she kills'. In rapid informal speech, the vowel sequence ai is pronounced as al or eT, as in naTtu, nellu 'they stole'. B. DIPHTHONGS AND THE PROBLEM OF GLIDES IN SYLLABLE ONSETS. We saw in §2.2.1 that y can occupy the syllable onset and the coda positions, while wand wh only occur in syllable onset, e.g. yawi jaguar', nu-kawana 'my leg', wha 'we'. Only in haw 'OK, yes' does w occur in the coda position. However, if a syllable contains a glide in the onset position, it can contain a preceding consonant, thus creating a syllable structure: C-ylw- V. There are a number of restrictions on the onset consonant, and on the vowel. lf the onset glide is y, the attested sequences are listed below (C-glide-V sequences are underlined). (i) p-y-a, or ph-y-a, e.g. na-12J!Q (3pl-hide) 'they hide', na-phl!ci (3pl-breathe) 'they breathe'; (ii) k-y-a, e.g. sifmimi 'swamp', di-dafsJE. (3pl-body+EMPII) 'his body indeed' (there is no sequence • khyV); (iii) s-y-a, s-y-, e.g. ~wa 'fire', pa-kaiJ!Q-nipe (IMP-scratch-NOM) 'itching', diweD!f!. 'he grinds', a~i 'pepperpot'; ~'striped cuckoo'; (iv) h-y-a, e.g. b:iQpa 'mountain'; (v) h-y-u, e.g. -!J.J!M 'purposive', -!rtY.me 'after:SS', -fnillna 'habitual', fu!y,kade 'not appear'; (vi) m-y-a, e.g. =mya 'only', Wamyarikune 'the Tariana subtribe of Santa Rosa and Periquitos', Myakadapana 'Ipanore'. In addition, if there are two glides in the onset position, the only possibility is for the bilabials w and wh to precede the !amino-palatal y; the syllables can contain either a or u, e.g. wvaka 'far', -wya 'extralocality marker on nouns', wvurina 'a type of tree', pawhya-nipe (IMP-breathe-NOM) 'breath, sorcery', whyume 'last'. 11 The symbol ii refers to an ultrashort phonetic realisation of the vowel. 2 J Stress 37 If the onset glide is w, the possible sequences are k-w-a, e.g. kwalw 'what', di-ekwa 'his face', di-kwa 'he is toasting (manioc flour)', di-kwti 'he hangs (in a hammock)', yak6rekwa 'door opening', -kwana 'classifier: plain'; k-w-e, e.g. kwe-peli 'whatever, all (things)', -kwema 'classifier: round things', and k-w-i, e.g. di-kwisti 'he scolds', kwitfi 'currassow' and -pukwi 'classifier: round and hollow'. These C-y-V and C-w-V sequences allow two alternative interpretations. We could consider them instances of CCV syllables thus necessitating a more complex syllable structure than the one presented in §2.2.1. Alternatively, we could consider ya, yu, ye, we, wa and wi as constituting complex syllable nuclei, i.e. diphthongs (and under this interpretation y should be written as i and w as u) thus necessitating a more complex paradigmatic array of vowels. There are no compelling reasons for choosing either of these. There is possibly just one argument in favour of the diphthong interpretation. As I show in §2.4.1, there are no independent phonological words of CV structure with a short vowel (with the exception of a few ideophones). The proclitic kwe 'what' (see §2.4.1 and §2.6) can be used as an independent phonological word. This suggests that the sequence we in kwe should be treated on a par with a long vowel, that is, as one vowel nucleus. There is a tendency towards simplifying kw to k. The noun yakOrekwa 'door opening' is pronounced as yakOrelw by all the Santa Rosa speakers, except for Leonardo Brito who has both pronunciations in free variation; the pronunciation yakorekwa is the one used in the Santa Terezinha dialect. The only instance of simplification of a sequence containing y as its onset is ye - e in -kyere 'classifier: island'. I have chosen to write the glides as y and w for two reasons: firstly, to allow for a uniform representation of the glides in the onset and in the coda positions, and secondly, to distinguish between glides in the onset position and a vowel sequence in cases such as the following minimal pair: -pukwi 'classifier: round and hollow', with a sequence k-w-i whlcb can be interpreted either as a syllable of CCV structure, or as a syllable of CV structure with a diphthong wi as its nucleus, and -lwpukU-i (turn-CAUS), which contains a vowel sequence IJ-i, and cannot be pronounced as [wi]. 2.3 Stress 2.3. I Primary stress Tariana distinguishes between primary and secondary stress. Primary stress is contrastive while secondary stress is not. A vowel with primary stress is characterised by higher pitch and greater intensity. Since the stressed syllable is realised with a high pitch, and the difference between many words lies in the positioning of this pitch, Tariana can be considered a pitch-accent language. Vowels which bear secondary stress do not differ in their pitch from non-stressed vowels (they differ from unstressed vowels -Vaupes area (Aikhenvald only in intensity). Pitch accent is a property of the I~an 1999b, c). Long vowels are always stressed. Primary stress can fall on the antepenultimate, penultimate or final syllable. The following examples illustrate minimal pairs djstinguished just by the placement of stress: pilw 'you laugh' - pikti 'you look'; piwa 'you play' - piwti 'your belly', pikwa 'you toast (manioc flour)' - pikwti 'you hang (in a hammock)'; Kayati 'old name for the Vaupes' - ktiyafi 'buoy (molog6)'; piweta 'you make enter' - pi-weta'you make a trap'; didieta 'he returns (something)' - didieta 'it (the wind) blows'; dipe 'his meat' - dipe 'he leaves'; mtiwi 'hook' - mawi: 'pari (type or 38 2 Phonology wood)'; ayni 'wasp' - aynf 'striped cuckoo', m;ya '1 stay, 1 wear' - nuya 'I hit'. The following items are differentiated by stress placement, and by vowel length: matji:te 'good' (its variant is matfoite; see under D II in §2.5.2) - ma:tjite 'bad'; ku:situ 'a dancing stick'- kusitu 'cutivaia (a mammal)'. Unstressed nasal vowels are rare; one example is iii 'port'. Stress shift indicates a change in grammatical meaning in the vocative forms of some kinship nouns (see Chapter 3), e.g. nuri 'my son', vocative nuri 'my son!' 2.3.2 Stress in monomorphemic words In disyllabic monomorphemic words the first syllable is most frequently stressed, e.g. dolo 'a fly', dzlpu 'a lizard', nana 'a type offish', e:nu 'thunder', yawi 'jaguar'. Polysyllabic monomorphcmic words are comparatively rare; their primary stress most frequently falls on the penultimate syllable, e.g. kusiwa 'a type of basket', turtipa 'cone', mawina 'pineapple', hanupe 'much, many' (see §8.1.5 on its structure), kanapfri 'grass', katanapiri 'worm', matamata 'freshwater fish (Pimelotus omatus)'. Tt can also fall on the antepenultimate syllable, e.g. kepiro 'bird', lniri 'traira-fish', pusatu 'sloth', i:natu 'woman' (this noun, however, contains a fossilised suffix -tu 'femjnine'), dapatatu 'cockroach', yakasolo 'traditional apron-like clothing'. There are a few examples of polysyllabic words with their stress on the last syllable, e.g. yapuratu 'long flute used at ritual offering', liamuritzJ 'wild caititu pig'. In irreguJar disyllabic plural forms (see Chapter 7) the penultimate syllable is stressed, e.g. l:na, Ina 'women' (plural of l:nam 'woman'), iltfo 'men', plural of tfori 'man'. However, these forms are bimorphemic. 2.3.3 Stress in polymorphemic words All roots have underlying stress, e.g. hanu 'big', ma: tji 'bad', matfo 'good'. Monosyllabic prefixed verbal roots can be stressed on the root, e.g. -ka 'to see', -a 'go', -ya 'hit', or on a pre-root vowel, e.g. - ·ka 'laugh', - 'ya 'stay, wear', - pa 'comb one's hair'. This is not predictable; that is, in a dictionary all morphemes are marked for their stress. All prefixes are lexically unstressed. A prefix can acquire stress only if it is attached to a root with the stress on the pre-root vowel, e.g. - 'ka 'laugh': di-k.a 'he laughs', -·nu 'to come': dl-nu 'he comes'; -'wa 'to try, to play': df-wa 'he tries, plays'. Examples with unstressed prefixes are nu-6 ( lsg-go) 'I go', di-ka (3sgnf-see) 'he sees', wa-ha (lplEMPH) 'we', wa-na (lpl-OBJ) 'us'. Suffixes can be either lexically stressed or unstressed. If a disyllabic word consists of two or more morphemes, the stress placement depends on accentual properties of morphemes. When an unstressed suffix combines with a root, the stress placement is unaffected, e.g. matfo 'good' + -ite 'noun class animate' -+ matji:te; ma:tji 'bad' + -ite 'noun class animate' -+ ma:tjite; episi 'iron, stone axe, metal' + -da 'classifier: round' -+ episi-da (metai-CL:ROUND) 'motor'. If a resulting word contains four syllables or more, stress shi fts to the antepenultimate syllable, e.g. pima 'sound' + -peri 'collective' -+ pimaperi 'voice'. The reduction of extraprosodic (ie. lexically unstressed) prefixes of V structure may go back to Pre-Tariana (This explains correspondence, such as Tariana pa:-, Baniwa a-pa: 'one', Tariana tfori, Baniwa a-ljiiiri 'man'.) When an underlyingly stressed monosyllabic suffix, or a disyllabic suffix stressed on the penultimate syllable, combines with a root, the stress moves onto the penultimate syllable, e.g. mawi 'hook', mawipi 'blowgun'; kUnuri 'a type of fruit', kunurina 'a 2 1 s1 Sj Ill le e yt m =n Tu 2.3 Stress 39 corresponding tree'; he()ku 'wood', heku-kena 'branch of a tree'; nu-kirwa + -na -> nukawa-na (Jsg-leg-CL:VERT) 'my leg'; i:ta 'canoe (as a class)', i:ta-whya (canoeCL:CANOE) 'an (individual) canoe'; hanu-maka (big-CL:EXTENDED.CLOTH) 'a big piece of extended cloth', hanupe 'much' + -ite 'noun class animate'-> hanupite 'the one who has a lot (of something)'. One monosyllabic suffix, -si 'non-possessed' (discussed in Chapter 6), is always stressed when it follows a root, e.g. pani-si 'house', herena-si 'illness', hini-si 'breast milk'. If a word contains a root and more than one stressed suffix (none of which is a djsyllabic suffix stressed on the last syllable), stress falls on the penultimate syllable, e.g. yani 'thing' + -maka 'classifier for cloth-like objects' + -si 'non-possessed' -+ yammak/Jsi 'article of clothing'; ka:ra (REL+Oy) 'flying' + -ka 'thematic' + -whya 'classifier for canoes' + -puna 'classifier for stretch of roads' -+ karakawhyapzina 'airport'. When a disyllabic suffix stressed on the last syllable combines with a root, the primary stress falls on the last syllable of the word, while the stressed syllable of the root now takes the secondary stress, e.g. hanu 'big' + -yawa 'classifier:hole' -+ hanuyawa 'big (hole)'. A disyllabic suffix stressed on the last syllable always retains its stress as the primary stress in the phonological word; thus, hanu (big) + -kina (CL:BRANCH) + -pada (CL:PJECE) results in hanu-kena-pada 'piece of a branch'. The stressed syllable of the preceding formative (as it was before the affix was added) takes on the secondary stress (as does -Jcena in this example). 2.3.4 Secondary stress Secondary stress is a characteristic of polymorphernic words, and of phonological words containing enclitics. Enclitics, but not proclitics, obligatorily bear a secondary stress (see §2.4). In addition, if a polymorphemic phonological word contains four syllables or more and is composed of stressed roots, the secondary stress falls on the syllable of the primary stress of the first component, e.g. matfo-maktira (good-dry) 'really dry, good and dry'. Monomorphemic words with four syllables or more (see examples in §2.3.2) do not have secondary stress. 2.3.5 Vowel reduction Tariana has vowel reduction in pre-tonic and post-tonic syllables which have no secondary stress whereby e, i, a go to ;;l. Vowel reduction occurs in normal and rapid speech registers, and not in slow, deliberate speech. Nasal vowels, and oral vowels u, o and j are never reduced. Pre-tonic vowel reduction takes place in the third syllable from the stress (counting left from the stressed syllable); or in the word-initial syllable. Consider the following examples of pre-tonic vowel reduction: yakOreka 'door' is realised as y:Jkoreka, yammakasi 'clothing' is realised as y;,mmalaisi (here vowel harmony may apply, and the word is realised as yurumakasi - see §2.4.4), and Kuphe-menane 'Piratapuya' as Kuphe- m;;t'lane. Pre-tonic vowel reduction also operates in syllables which immediately precede those with a secondary stress; trus can be illustrated with the following clausal enclitics (which take secondary stress: see §2.4): =makha 'recent past non-visual' realised as =m:icha; =tahlcQ 'frustrative+recent past visual' as =t:Jh/cQ. Post-tonic vowel reduction occurs in word-final syllables, e.g. yasene-+ ytisen;;J'the Tucano', mattimata-+ mat~ 'freshwater fish', hemdsiere -+ hemosier:> 'capybara'. 40 2 Phonology ln normal and rapid speech the post-tonic vowel sequence -ua is realised as either -u or -o, e.g. -ha-dua (-parent-FEM) 'mother' and -sa-dua (-spousc-F'EM) 'wife' are pronounced as -ha-du, -ha-do and -sa-du, -sa-do (see §I 0.2). ln polymorphemic words with the primary stress on the penultimate syllable, a word initial il, a or i may disappear, e.g. ildaro 'parrot', daru-phema (parrotCL:LEAF.LJKE+PAIR) 'a head cover made of parrots' feathers', from ildaro-phe-ima (parrot-CL:LEAF.LIKE-CL:PAIR), daro-iui 'River of a parrot (place-name)' (cf. §2.3.3). 2.4 Phonological word The main criterion for a phonological word in Tariana is primary stress. Different prosodic classes of morphemes have different rules with respect to stress - see §2.4.1. F'urther prosodic processes related to delimiting the boundaries of a phonological word are nasalisation (§2.4.2), aspiration (§2.4.3) and vowel harmony (§2.4.4). Prosodic parameters relevant for delimiting a phonological word are summarised in §2.4.5. Correlations between a phonological word and a grammatical word are outlined in §2.6 (see also Aikhenvald 2002b). 2. 4. 1 Primary stress and prosodic classes ofmorphemes A phonological word contains one primary stress; it may also contain one or more secondary stresses. Prosodic classes of morphemes are roots, affixes, proclitics and enclitics. Roots and proclitics can form a phonological word on their own, while affixes and enclitics cannot (see Table 2.5 and §2.6). Enclitics always take a secondary stress. Similarly to most Arawak languages, Tariana is predominantly suffixing. A small set of prefixes exists, consisting of A/Sa and possessor cross-referencing prefixes (Aikhenvald 1999b and §6.2), ma- 'privative' and lea- 'anributive, relative'. Prefixes can occur only with inalienably possessed nouns (see Chapter 6) and with transitive and active intransitive verbs (see Chapter 12). Prefixes and suffixes difTer in their accentual properties - sec §2.3.3. Monosyllabic phonological words ofCV structure must contain a long vowel, e.g. ta: 'a large wasp', ye: 'armadillo', ka: 'just like that', ma: 'let's go' (see §2.2.2 on kwe 'what'). There are just a few such monosyllabic phonological words. A few monosyllabic interjections contain short vowels, e.g. tfo, tfu 'oh, wow', be 'wow' (a loan from Tucano), ne 'Hey!' (possibly, a loan from Tucano). If the connective proclitic ne= (see §2.6) or the homophonous negative proclitic ne (used as a one-word negative response) form a phonological word of their own they are pronounced with a long vowel [ne:]. Affixes attach to a root according to the nature of the root: for instance, verbal affixes attach to simple verbs or serial verbs (see Chapter 18), and nominal aftixes, such as classifiers (see Chapter 5) and the locative -se (see §7.3.1 ), anach to nouns, or to noun phrases. A number of phonological processes occur on the aftixal boundaries (see §2.5). Some specific affixes, such as -dua, -du 'feminine' or -niri 'masculine' occur only with kinship nouns; other affixes occur only with some subgroups of verbs, some nouns and some adjectives, e.g. -ita 'causative', or classifiers. Tariana has a few proclitics and a large number of enclitics. Proclitics can form a phonological word on their own and never take a secondary stress. They can cliticise to the word which immediately follows them if they are not contrastive. Proclitics are ne= 'then (connective)', ne= 'negative', kay= 'thus' and kwa/kwe 'interrogative-distributive which, how, every' (see Chapter 11 ). In contrast, enclitics cannot form a phonological 1. 4 Phonological word 4I word on their own, and always have a secondary stress. No phonological processes, except for the fusion of homorganic vowels, occur on clitic boundaries (see §2.5). Enclitics fall into two types: (a) fixed position phrasal enclitics which attach to certain morphological classes of words or types of phrases (nouns, noun phrases, verbs), and (b) floating clausal enclitics. These are called floating because they do not have any fixed position and can attach to any constituent in a clause, provided lhis constituent is in contrastive focus. Phrasal enclitics include ::.nukU 'topical non-subject' (also pronounced as =naku, . depending on the speaker's generation: see under C in §2.4.4), =nuku or as = n~u - nel nhe 'focussed subject', =ne 'instrumental/comitativc', =iuki 'diminutive', =p1i 'augmentative, really', =ya 'really, emphatic', Aktionsart enclitics, e.g. =weti 'smash', =halo 'spill water' (see § 15.2), mood enclitics, e.g. =da 'apprehensive', and others. Clausal enclitics include tense-evidentiality, aspect and mood markers, e.g. - pidana 'remote past reported', =tho 'frustrative', =ti'Jpe 'third person in1perative', and a number of other miscellaneous enclitics, e.g. =wani 'emphatic'. lf a phonological word contains more than one enclitic, each of them has a secondary stress, e.g. di-tuk1i=pidana (3sgnf-fall=REM.P.REP) 'he fell', di-mku=pidana- pita (3sgnf-fa1J=REM.P.REP=AGAIN) 'he fell again'. The following minimal pair illustrates the difference between a suffix and an enclitic: iri-ne (blood-PL) 'those of blood, the Tariana' (see § 1.4, § 1.8. I) consists of a root and a suffix, and is a phonological word with one primary stress, while fri - ne (blood=comitative) 'with blood' consists of a root and an enclitic, and is a phonological word with one primary and one secondary stress. Another minimal pair is -wani 'classifier for abstract nouns and places' and =wani 'emphatic clitic'. Further properties of clitics are discussed in §2.6. A vowel which bears a secondary stress does not undergo vowel reduction (§2.3.5). Interjections (most of them idcophonic) have a primary stress but not lhe high pitch which generally accompanies it. This property is shared with Tucano (see Ramirez 1997, vol. I: 348). Most ideophones are monosyllabic, e.g. tjin 'sound of a tinkling musical instrument', weu 'sound of brusque movement>, hu 'away', {fit, tfo 'astonishment, unpleasant surprise', pua 'wow! (pleasant surprise)', kui, kwe 'excitement'. ldeophones are under! ined: 2.1. pima-pi dana .!fu! .!fu! diha inari sound-REM.P.REP (sounds) ART mucura.rat i-lhi-da leka-ka pima-pidana tho tho sound-REM.P.REP (sounds) INDF-eye-CL:ROUND burst-SUB 'There was a sound tftn! tftn! (a tinkling sound of a flute) ; when lhe eyes of the mucura-rat burst, there was a sound tho! tho!' 2.2 .!fu! phia ma:tft-pu-mha Oh you bad-AUG-PRES.NONVIS 'Oh, you are doing evil things to me.' nu-na lsg-OBJ pi-ni 2sg-do Classifiers of the 'repeater' type (i.e. lhose classifiers which 'coincide' with the head noun: see §5.3) have no primary stress and no high pitch. Their secondary stress falls on the same syllable as the primary stress in a corresponding independent noun. Thus, when a noun panisi 'house' is used as a repeater to mark agreement on an adjective matfa 'good', its primary stress is replaced by secondary stress, e.g. panisi matfo-panisi 'a 42 2 Phonology good house'. This property of classifiers-repeaters is also shared with Tucano (Ramirez J997,vol.l: 103). The properties of affixes, proclitics and enclitics are summarised in Table 2.5. See further discussion in §2.6. TABLE 2.5 PROPERTIES OF AFFIXES, PROCLITICS AND ENCLITICS PROPERriES INDEPENDENT PHONOLOGICAL WORD SECONDARY STReSS I'IIONOLOGICAL PROCESSES ON THE BOUNDARIES RESTRICTIONS ON COMB INA riON WITH ACONSTITUENT TYPE MOBILITY WITIIIN APHRASE OR A CLAUSE - SJ:m §2.6 AFFIXES no no PHRASAL CLAUSAL PROCLITICS ENCL.ITICS ENCLITICS yes no no yes yes no yes no no no yes no yes no no yes yes yes 2.4.2 Nasalisation The domain of nasalisation in Tariana is a phonological word. If a phonological word contains a nasal vowel O( a nasal consonant, all vowels (except 1), the voiced dental stop d, the lateral /, the flap rand the alveopalatal glide y within the word boundaries are replaced with a homorganic nasal in normal and in rapid speech registers: [r)-+ [n], [y] -+ [i'l], [d) - [n], e.g. -toreta 'roll dough' is realised as [toncta]; keniJwa-na 'a tree-like plant (unidentified)' as [kenowa-na); di-yii, di-na 'he lives' as (ni-na]; di-yii 'he goes up' as [ni-i'lil]; di-whii 'he bites' as [ni-wha). In very rapid infonnaJ speech, nasalisation may extend to phonological phrases, e.g. di-nu=niki yeda-se (3sgnf-come-COMPL downstrearn-LOC) 'he carne completely downstream' is realised as [ni-nu=niki nedea-se). This nasalisation spread in Tariana is the result of a 'Tucano accent' in Tariana, since nasalisation as a word prosody is a property of East Tucano languages (see Aikhenvald 1996a); it is not found in other Arawak languages of the area. 2.4.3 Aspiration Tariana has a set of aspirated stops and nasals and an aspirated bilabial glide. Table 2.2 shows that there are more restrictions on the occurrence of aspirated consonants than of unaspirated ones, especially within affixes and clitics. No roots, affixes or clitics contain two aspirated consonants. Here 1 first consider the loss of aspiration within a phonological word across clitic and prefix boundaries, and then discuss aspiration floating and the variation between pre-aspiration and post-aspiration. A. LOSS OF ASPIRATION WITHIN A PHONOLOGICAL WORD. If adjacent syllables across a clitic boundary contain aspirated stops, nasals or the bilabial glide wh, either of these, but not both at once, is realised as aspirated. For instance, a sequence of a prefix and a root di-pha (3sgnf-fall) and a clitic =lcha 'away' should have resulted in di-pha .,.kha; this sequence is realised as either [di-pha=ka] or [di-pa=kha]. Similarly, a phonological word consisting of a sequence of a root wyaka 'far', an enclitic - tho 'frustrative' and the enclitic =mhana 'remote past non-visual' should have resulted in wyaka=tha=mhana ('it is (perceived non-visually) as being far in vain') but is realised as either [wyaka=ta=mhana] or [wyaka=tha=mana]. 2.4 l'honological word 43 If a root contains the fricative h and a clitic contains an aspirated stop, this may result in the loss of aspiration within a clitic, with some speakers. For instance, Jose Manoel Brito, a very fluent elderly speaker of Tariana, consistently pronounces words like diayahya (3sgnf-swim)=kha (away) 'he swam away' as [di-ayahya-ka). This process is characteristic of clitic boundaries (see discussion in Aikhenvald 2002b). Aspirated voiced stops and aspirated nasals may lose !.heir aspiration in a non-wordinitial position, e.g. di-nu-mha (3sgnf-come-PRES.NONVIS) 'he is coming' can be pronounced as [di-nu-ma]. In contrast, mhema-kade (underlying form: ma-hima-kade (NEG-hear-NEG)) 'he does not hear' cannot be pronounced as [mema-kade] (note !.hat mema-kade (underlying form: ma-ima-kade (NEG-stand-NEG)) means 'he does not stand); mhamita-kade (NEG+bum-NEG) 'he does not bum' cannot be pronounced as [mamita-kadc], and dhima (3sgnf+hear) 'he hears' cannot be [dima] whlcb would mean 'he stands'. Free variation between aspirated dh and unaspirated din the two enclitics and the only root where dh occurs was observed with younger speakers, e.g. di-ya=dhuti (3sgnfhit=FOLD) 'he hit (the tree) so that it folded' is pronounced as [di-ya =durij, while =data, an Aktionsart clitic meaning 'get unstuck', is often heard as [=dala], and di-dhala (3sgnfunstick) 'he unstuck' can be pronounced as [di-dala]. The aspirated bilabial glide can be pronounced as unaspirated w in any position, e.g. [whyumej, [wyume] 'be last', [kcwhedonite], I kewcdonite] 'pregnant'. B. ASPIRATION FLOATING. Aspiration in Tariana is a prosodic phenomenon whose domain is less than a whole phonological word (see Aikhenvald 1996c for a similar account of aspiration in Bare, another Arawak language previously spoken in the Upper Rio Negro region). The domain of aspiration as a prosodic phenomenon is determined by the following rule which I call aspiration floating. In a sequence of two syllables (on a prefix- root boundary or within an enclitic), one of which contains an aspirated stop, nasal or bilabial glide and the other contains an unaspirated voiceless stop, the aspirated consonant may lose its aspiration; then the coda of the first syllable is h. The process does not apply if (a) the first syllable is not stressed, or (b) it does not contain a long vowel. Aspiration floating docs not apply within a root, cf. maratahka, not [maratakha] 'quick wave', karahta, not [karatha] 'lung', -anihta, not [-anitha] 'think, understand, reason'. An aspirated stop or nasal or wh on the prefix-root boundary is most often the result of h-metathesis (sec §2.5.1 ). Following this rule, kephunipe (underlying form: ka-iphu-nipe: REL-follow-NOM) 'four' is often pronounced as [kehpunipe), and mhepakilde (underlying form ma-hipakade NEG-wash-NEG) is pronounced as [mebpakltde] 'he/she did not wash', and dhipakaru (from di-hipa-karu 3sgnf-catch-PURP.YIS) is pronounced as (dihpahru] 'in order for him to catch'. In contrast, dhipakasu (underlying form di-hipa-kasu 3sgnf-catch-INT) 'he is going to catch' cannot be pronounced as [dihpakasu ], since the fust syllable of this word is stressed and thus condition (a) is not fulfilled. Similarly, mhe:pakade (underlying form ma-hepa-kade NEG-answer-NEG) 'he did not answer' is not pronounced as [me:hpakade] because the first syllable contains a long vowel and thus condition (b) does not hold. The form mhe:pakade can be pronounced with a short vowel (see above, §2.1 .2, on the Joss of vowel length) as mhepakade; speakers who do this can pronounce this word as [mehpakade], thus applying the aspiration floating. Along similar lines, the sequence of clitics =thaka (underlying form: =tha=ka) 'frustrative=rcccnt past visual' can be pronounced as [=tahka], and the sequence =mha: ka 'recent past non-visual' can be pronounced as [=mahka]. There is just one ,..... , 44 2 Phonology case in which aspiration floating takes place within what appears to be one morpheme: the enclitic =botha, =butha 'conditional' is often pronounced as =bohta or =buhta. However, it is likely that this clitic is actually a combination of two clitics =bo 'counterfactual' (a loan from Tucano boo 'counterfactual, potential') and =tha 'frustrative' (see Chapter 5 of Aikbenvald forthcoming-a). The application of aspiration floating within =botha may constitute an additional piece of evidence in favour of its bimorphemic origin: see § 16.6. In an unstressed syllable, word-initial h and the following high or central vowel can be metathesised, e.g. [biporite], [ihporite] 'green (animate)', [hiperiwi], [ihperlwij 'type of hedge', [hieriku], [iheriku] '(someone's) anus' (see further examples in §2.2.1). This does not happen if the word-initial syllable is stressed, e.g. [hima] 'you (plural) sleep' (not [ihma]), [hlpa] 'you (plural) grab' (not [ihpa]). The aspiration floating is thus indicative of a clitic boundary within a phonological word, and of a prefix-root boundary. This prosodic parameter can be used for determining the boundaries of a phonological word; sec §2.4.5. The h-metathesis (see §2.5.1) most frequently occurs on a prefix + root boundary, so that most aspirated consonants, especially the aspirated voiced stops, occur in the wordinitial position. Aspiration floating in this form is not attested in other North Arawak, or in East Tucano languages. Waltz and Waltz (1967: 30) note that Guanano (East Tucano) sometimes has 'fluctuation' of aspiration; this requires further analysis. North 1\rawak languages Baniwa of lcyana and Bare have aspiration as a prosodic feature, associated with h-metathesis. In Bare, but possibly not in Baniwa, aspiration is associated with the beginning of a phonological word. ln Warekena of Xie h-insertion is a word and phrase boundary marker. See Aikhenvald (l996c, 1995b and 1998b) for further discussion. 2.4.4 Vowel harmony Vowel harmony occurs within the limits of a phonological word, and is indicative of its boundaries. Tariana has three types of vowel harmony, all of which occur in normal and rapid speech. A. REGRESSIVE TRANSLARYNGEAL VOWEL HARMONY IN fNDEPENDENT PERSONAL PRONOUNS. Independent pronouns consist of a cross-referencing prefix and a deictic element -hct (see C hapter 11). Their underlying form is shown in the second column of Table 2.6. The way pronouns are pronounced in slow register is shown in the next column. The subsequent columns show how, in normal and in rapid speech register, a vowel identical to the vowel of the first syllable, i.e. to that of the cross-referencing prefix, is inserted after h. ln the rapid speech register the vowel of the first syllable elides, and aspirated consonants emerge as a result; this is shown in the last column. 11 The morpheme -ha triggers nasalisation of the following vowel; this process is widespread among the North Arawak languages of the area (see Aikhenvald 1996d). When pronouns get cliticised to the verb in the 'pronoun doubling' construction described in §25.1.4, the shorter, rapid speech register form occurs. 11 Biocca's (1965) recordings ofTariana shamanic songs contain instan.ces of a slow register realisation of third person singular non-feminine pronoun diha. 2.4 Phonologtcal word 45 B. WORlHNITlAL VOWEL HARMONY. lf the root begins with h and is preceded by a prefix, a vowel identical to that of the prefix is insened after h, e.g. slow register: [duhcni) ; normal to rapid [du-bueni] or (dhueni] (3sgf-ear) 'her ear'. This can be interpreted as an instance oftranslaryngeal vowel harmony. Two unstressed vowels at the beginning of a phonological word may be pronounced may be pronounced as identically to the second vowel, e.g. [yarumakasi], [y~rumaksi], [yurumakasi] 'clothing'; and [di-sepata] as (di-sapata] 'he suffers'. TABLE 2.6 PHONETIC REALISATIONS OF INDEPENDENT PRONOUNS UNDERL Yll\G FORM SLOW REGISTER l sg 2sg 3sgnf 3sgf Ipi 2pl 3pl nu-ha pi-ha di-ha du-ha wa-ha i-ha na-ha nuhii pihii dihii duhii wahii ihii nahii NORMAl. RF.GISTF.R nuhuii pihyii dihyii, dihiii duhuii, duhwii wahii ihyii nahii RAPID REGISTER nhuii,nhua phiii, phia dihyii, dhya duhwii, dhua whii,wha ihyii, hya nha C. WORD-FINAL VOWEL HARMONY WITHIN ENCLITICS. In normal to rapid speech, enclitics which have their secondary stress on the last syllable assimilate the vowel in the pre-tonic syllable to the one in the stressed syllable, e.g. -naku 'topical non-subject' becomes -nukU, -pidena 'remote past inferred' becomes -pidana. This only happens when these enclitics occur word-finally. Vowel harmony of this type is thus associated with Lhe end of a phonological word. Younger speakers tend to use the form with assimilated vowels as the only one, while older people have both. America Brito, the oldest living speaker of the Santa Rosa variety, pronounces -nakii as -n;;t/al. There is a certain amount of individual variation in bow the vowel harmony of this type applies among other proficient speakers. lsmael Brito and other younger speakers extended vowel harmony to the negative suffix, both word-finally and before a clitic, e .g. ma-nu-kilde (NEG-come-NEG) 'he/she did not come' was pronounced as [manukede], and to all the disyllabic words with a in the first syllable and e in the second, e.g. tyare 'indigenous cup' was pronounced as [tyere). 2.4.5 Word delimiting prosodic parameters in Tariana: a summary The following prosodic parameters in Tariana delimit the boundaries of a phonological word. Note that only the primary stress is word-defining; other parameters can be considered subsidiary. I . Primary stress: determines the number of phonological words (§2.4.1 ). 2. Secondary stress: indicative of clitic boundaries; indicative of the structure of a phonological word (§2.4. 1 and §2.3.4). 3. Nasalisation: operates within the boundaries of a phonological word (§2.4.2). 4. Aspiration floating: indicative of the beginning of a phonological word, and of clitic boundaries within it (i.e., of its internal structure) (§2.4.3). 5. Aspiration: indicative of the word-initial position (§2.4.3). I! 46 2 Phono/oJlJI 6. Vowel harmony: indicative of word-final, and of word-initial positions (§2.4.4: A,B,C). 7. Post-tonic vowel reduction: indicative of word-final position (§2.3.5). Phonological word boundaries have additional segmental properties. Phonological words hardly ever start with r. Restrictions on the occurrence of vowels and consonants in the root-, affix- and enclitic-initial and -medial positions are given in Tables 2.2 and 2.4. 2.5 Phonological processes Phonological processes of two kinds - h-metathesis (§2.5. 1), and vowel fusions (§2.5.2) -occur on the boundaries between a prefix and a root, a root and a suffix, and between two suffixes. The boundaries between parts of nominal compounds are treated in the same way as boundaries between roots and suffixes (see § 10.5 on the difficulties of distinguishing compounds from sufflxal derivations in Taria.na). Some processes (A2 and E in §2.5.2) occur on the boundaries between enclitics and roots and between enclitics. No phonological processes occur on proclitic boundaries. Root-internal phonological processes are discussed in §2.5.3. 2. 5.1 H-metathesis JI-metathesis is attested in a number of North Arawak languages of the Upper Rio Negro. See, for instance, Aikhenvald ( 1995a) on h-metathesis in Bare, and Valadares ( 1993) on h-metathesis in Kumandene, a dialect of Kurripako; note however that in Bare, but possibly not in Baniwa/Kurripako, this process is limited to word-initial position. ln Tariana, its application is independent of stress placement. Vowel fusion processes described in §2.5.2 always apply after h-metathesis. A root- or a suffix-initial h, when preceded by a prefix or a root, metathesises in the following ways: A. CV- + -hVX-+ ChVX-+ [aspirated C]VX, ifC is a stop, a nasal or a bilabial glide. Examples are: • prefix di- (3sgnf) + root -hima (hear) -+ dhi-ima -+ dhima 'he hears'; • prefix na- (Jpl) + root -hima -+ nha-ima -+ nhema 'they hear'; • pa: 'one' + -hipa 'classifier human' __, pha-ipa -+ phepa 'one (human)'; • rootyawi jaguar'+ root hfpani 'rapid'-+ yawhi-ipani __, yawhipani; • 'lauarete', lit. 'rapid of jaguar'. lf there are two syllables of CV structure, each containing a stop, a nasal or a bilabial glide, preceding the -hVX sequence, h-metathesis applies to the leftmost of these, e.g. nama 'two' + -hipa 'nunleral classifier: human' - nhama-ipa - nhamepa 'two (human ones)' (not •nhamhepa). If C is not a stop, a nasal or a bilabial glide, the h- metathesis does not apply. The suffix-initial h simply disappears and the processes of vowel fusion (see §2.5.2) apply, e.g. mada/i 'three'+ -hipa 'numeral classifier: human'-+ madalipa 'three (human ones)'. 2.5 Phonological processes 41 B. V 1- + hV 2X--+ hV 1V2X (--+ hV2X if V,=V2); that is, if a syllable preceding the boundary is of V structure, h ~umps' one syllable, and the preceding syllable acquires hV structure. Examples are: • prefix i- (2pl) + root -hipa 'grab' --+ hi-ipa --+ hipa 'you pl. grab'; • prefix i- (2pl) +root -hematha 'shout'--+ hiematha 'you shout'. The vowel coalescence which applies after h-metathesis does not result in the emergence of long vowels (see under A in §2.5.2). If the prefix has the voiced dental stop as an onset, h-metathesis takes place in normal and rapid speech, but not in slow deliberate speech, e.g. di-hipa --+ dhlpa 'he takes' (normal to rapid register), dihfpa 'he takes' (slow register); du-hupa--+ dhtjpa 'she takes' (normal to rapid register), duhupa (slow register). C. CV + VwhV --+ ChVwV --+ [aspirated C]Vw, when Cis a stop, a nasal or a bilabial glide, e.g.: • pa: 'one' + -iwhi 'classifier: thin particle-like things' --+ phewi 'one thin particlelike thing'. In this case vowel monophthongisation (a+i --+ e) applies after h-metathesis; vowel monophthongisation will not apply on the boundary between a root that ends in a: and a suffix that starts with -i (compare D II in §2.5.2). This rule does not apply for aspirated consonants other than wh, e.g. pa: 'one' + -iphina 'classifier: quarter'-+ pa-iphina 'one quarter',pa: 'one'+ -phe 'classifier: leaf-like'--+ pa:phe 'one Ieaflike thing'. The application of h-metathesis accounts for most occurrences of aspirated stops, especially of the voiced aspirated dh, and nasals, and of the aspirated bilabial glide (see Table 2.2). 2.5.2 Vowelji1sions Vowel fusions, i.e. the boundary processes which involve changes within vowel sequences, are (A) vowel coalescence, (B) monophthongisation, (C) vowel loss on prefix- root boundaries, (D) vowel loss and compensatory lengthening on suffix boundaries, and (E) the emergence of glides on suffix and enclitic boundaries. Vowel fusions apply after h-metathesis. A. VOWEL COALESCENCE. Vowel coalescence applies to identical vowels on prefixroot boundaries, on root-suffix boundaries, between suffixes, and between a root and an enclitic. AI. On a root- suffix boundary, independently of stressed placement: V,+V1 --+ V1• In the following examples, the initial vowel of the suffix is unstressed: • pani-si (house-NPOSS) + -ite (animate classifier) --+ panisite 'the one who belongs to the house'; • dfpumi- 'after him' + -ita 'causative' --+ dipumita 'he punishes'; • k/Jnari-i-thfda (mirror-INDF-eye) --+ konarithida 'mirror-eyed' (a description of an evil spirit). 48 2 Phonology In the following examples. the initial vowel of the suffix is stressed: • ma:tji 'bad' + -lpa 'classifier: big open space' -+ ma:tjipa 'bad (big open space, e.g. a field)'; • ma:tji- 'bad' + -ire 'noun class animate' -+ ma:tjite 'bad animate one'; • hawaya 'inga fruit' + -apa 'classifier: largish long objects' -+ hawayapa 'inga fruit which is largish and long'. A2. On a prefix- root boundary, if the initial vowel of the root is stressed, the rule is V1 + v1 -+ v1• Examples are: • nu-uma -+ mi:ma 'I look for' (a minimal pair with nu-ma 'l poison fish') ; wa-ara - wa: ra 'we fly' ; • di- (3sgnf) + -lsa (climb)-+ di:sa (3sgnf+climb) 'he climbs'. If vowel coalescence applies after h-metathesis, an A 1-type process applies, no matter whether the vowel of the root is stressed or not, e.g. di- (3sgn1) + root -hima (hear) -+ dhi-ima -+ dhlma 'be hears' (cf. other examples in §2.5.1 ). If there are two homorganic vowels on a root-enclitic boundary, A 1 applies, e.g. de.pi=iha (night=APPR) 'a little bit at night' (when it is just a little bit dark)-+ de :piha. B. MONOPHTHONGISATION. Monophthongisation applies on prefix- root boundaries and on root-suffix boundaries but not between suffixes. This process is applied independently of stress placement. a e, e.g.: na- (3pl) +-[sa (climb) -+ nesa 'they climb'; -dia 'return' + -ita 'causative'-+ -dieta 'make return'; ka- (relative) + -hima 'bear' -+ khema 'the one who hears' (here the monophthongisation applies after the h-metathesis); di-pitana 'his name' + -ita 'causative' -+ dipitaneta 'he names'; di:ra 'he drinks'+ -ira 'causative'-+ di: ret a 'he makes (someone) drink'; wapa 'we rot' + -ita 'causative' -+ wapeta 'we make rot'. 1 i-+ • • • • • • Monophthongisation does not take place if either i or a of the root is long (recaJI that affixes never contain long vowels; see §2.1.2 and Table 2.4), as in the following examples: • na- '3pl' + -l:ru 'steal'-+ na-i:tu 'they stole'; • na- '3pl' + -i:nu-+ na-i:nu 'they killed' (see also §2.2.2); • pa:- +-ita (one-CL:ANlM)-+ pa:ita, paila 'one (animate)'. Neither does the monophthongisation occur if the vowel a is stressed, as in: nama 'two' + -ita (two-CL:ANIM) -+ iiamaita 'two (animate)'. The exception to this is the specifier article dihci: • diha 'he, specifier article'+ -ita 'CL:rNAN'-+ dhera 'the inanimate one'; 2.5 Phonological processes • • 49 diha 'he; specifier article' + -ipu 'classifier: long hollow' -+ dhepu 'the long hollow one'; dihti 'he; specifier article'+ -itji 'bundle'-+ dhetji 'the bundle-like one'Y It always applies to a sequence a(.) + i which has arisen as a result of h-metathesis (see C in §2.5.1 ), e.g. pa:- 'one'+ -iwhi 'classifier: thin particle-like things' -+ phewi 'one thin particle-like thing (cf. phepa 'one (human)' in A under §2.5.1). Monophthongisation in Tariana appears to be fairly recent. A combination wapa 'we rot' + -ita 'causative' can only be realised as wapeta 'we make rot' in the present-day Tariana. However, in the Tariana shamanic chants recorded over fifty years ago (see § 1.8), the same verb appears interchangeably as wapeta 'we make rot' (with monophthongisation) and as wapaila (without monophthongisation), depending on the rhytlun. C. VOWEL LOSS ON PREFIX- ROOT BOUNDARIES. Prefixes contain only the short vowels a, u and i. On prefix- root bow1daries, the following rules apply, unless the A or B rules have already applied: Cl. (i) u- + -i-+ u, e.g.: • nu- 'I sg' + -!sa 'climb' -+ misa 'I climb'; • nu- 'lsg' + -ima 'sleep'-+ numa 'l sleep'. C 1 also applies after h- metathesis: • du- '3sgr + -hipa 'grab'-+ dhupa 'she grabs'. Cl. (ii) u- + -V -+ uY, ifV is a vowel other than stressed i, e.g.: -u + -i--+ ui, e.g. : • nu- 'lsg' + -iro 'drink'-+ mJira 'l drink'; -u + -e--+ ue, e.g.: • • du- '3sgr + -ema 'stand'-+ duema 'she stands'; du- '3sgr + -hepa 'answer' -+ dhuepa 'she answers'; -u + -a-+ ua, e.g: • nu- '1 sg' +-a 'go'-+ nua '1 go', nu- '1 sg' + -ara 'fly'-+ nuara 'l fly'. Sequences -u 1 -e can result in u for some speakers, both young and old, who do not have much chance of speaking Tariana regularly. Emilio Brito, Juvenal Brito and Batista Brito were using the form dhuka 'she dug' (underlying form: du- + -heka) instead of dhueka; other speakers were consistent in correcting them. 13 The classifier suffix 'bundle' is either -tji or -itji depending on the spealer. More traditional speakers, fruit' (from manalca ' a~i fruit' and e.g. Cfutdido Brito, prefer the form -tji, e.g. manalcalji 'bundle of a~i -tji 'CL:BUNDLE'). More innovative speakers use the form -1tji, e.g. manaketfi (from manalca · a~i fruit' and -itji 'CL:OUNDLE'), and dhetji 'the bundle-like one' cited above; this fonn is said to be 'from Baniwa'. 50 2 Phonology Since no long vowels are allowed withjn vowel sequences (see §2.2.2), a long vowel is realised as short, as in -u + -i: e.g. nu- 'I sg' + -i:tu 'daughter' -+ nu-itu 'my daughter'. This process does not operate on a root-suffix boundary, e.g. hanu- 'big' + -ite 'classifier animate' -+ hanuite (not *hanute) 'big animate one' (see D below). C2. (i) i- + -e -+ e if the h-metathesis applies, this process applies optionally: • di- '3sgnf' + -hepa 'answer' -+ dhepa or dhiepa 'he answers'; • di- '3sgnf' + -he 'enter'-+ dhe or dhie 'he enters'. C2. (ii) otherwise i- + -V-+ iV, e.g. : i- + -e-+ ie, e.g.: • di- '3sgnf + -ema 'stand' -+ diema 'he stands'; • di- '3sgnf + -eku 'run' -+ dieku 'he runs'; i- + -e:-+ ie, e.g.: • di- '3sgnf + -e:nene 'tongue'-+ di-enene 'his tongue'; i- + -u-+ iu, e.g.: • di- 'Jsgnf + -uma 'seek'-+ diuma 'he seeks'; • di- '3sgnf + -uka 'seek' -+ diukti 'he arrives'; •• i- + -a-+ ia, e.g.: • di- '3sgnf+ -a 'go'-+ dia 'he went'. C3. a- + -u-+ a:, e.g.: • na- '3pl' + -uma 'seek' -+ na:ma 'they seek'. C4. a- + - e-+ e:, e.g.: • wa- 'lpl' + -eku 'run'-+ we:ku 'we run'. T his process allows variable output, even with the same speaker: ai as in waiku 'we run', or ei as in weiku 'we run'. The variant we:ku is consistently used by older speakers. D. VOWEL LOSS AND COMPENSATORY LENGTHENING ON SUFFIX BOUNDARIES. These rules apply on root-sufflX boundaries and between suffixes, after rules A and B. (No suffixes begin with u ore: see Table 2.4.) D I. u • • • + -i -+ u, as in: di-rmisu 'he goes out' + -ita 'causative' -+ dimusuita 'he makes go out'; di-kapuku 'he turns' + -ita 'causative' -+ diiUJpukuita 'he makes turn'; hone 'that' + -tfku 'locative derivational affix' + -ima 'classifier: pair' rikUma 'the location of a side over there'. -+ hane- 14 Batista Brito, a fluent speaker who rarely uses Tariana, treats these sequences differently: for him -i + V _, V, e.g. di-elcu 'he runs' is pronounced as delcu, and di-uw 'he arrives' as duka. 2.5 Phonological processes 02. 51 u + -i--+ -ui, as in: • • hanu- 'big'+ -ire 'noun class animate' --+ hanuite 'big (animate)'; hanu- 'big'+ -ita 'noun class inanimate'--+ hanuita 'big (inanimate)'. u 03. ti + -i--+ (if there is a long vowel in the first syllable). as in: • he:leu 'wood'+ -ita 'CL:INAN'-+ hekUta 'paddle'; • ka .·yri 'thus'+ -ire 'NCL:ANlM'--+ kayrite 'an animate one who is like this'. 04 .•,; + -i-+ -i, as in: • hanu- 'big'+ -ipa 'classifier: large space'--+ hanfpa 'large (space)'; 'a lot'. 05. Li + -a- u, as in: • hanu 'big' + -aphi 'CL: HOLLOW'--+ hanuphi 'a big hollow object (e.g. a pan)'; • hanti 'big'+ -apa 'CL: LARGISII LONG'--+ hanupa 'big largish long object (e.g. a sandal)'. 06. u: +- f--+ u:, as in: • su:'small' + -ipa 'classifier: space'- su:pa 'small (space)'. 07. -ri: +-a--+ u:, e.g: • • su:'small' + -aphi 'classifier: hollow'--+ sfi:phi 'a small hollow object (a pan)'; sii: 'small'+ -apa 'classifier: largish long'-+ sil:pa 'small one (a sandal)'. 08. e + -a--+ • 09. e, as in: sumpe 'clay'+ -aphi 'classifier: hollow'-+ sumpephi 'a clay pan'. e + i--+ e, as in: • tidatu-phe 'parrot feather (parrot-classifier: leaf-like objects)' + -ima 'classifier: pair'-+ datuphema 'a head-ornament made of parrot feathers' (see §2.3.5 on the loss of if). 010. e + -i--+ I, as in: • • dilhi-sede-ite (3sgnf+eye-NEG.EX-NCL:ANIM)--+ dithisedite 'eyeless'; hanupe 'much'+ -ile 'noun class animate'--+ hanup{te 'the one who has a lot'. 011. a:, a+ -i-+ 1:, if i is unstressed (that is, monophthongisation does not apply): • matfo:- 'good' +-ire 'noun class animate'--+ matfi:te 'good (animate)'; • wya:- 'long'+ -ire 'noun class animate'--+ wi:te 'long animate one'; • alia 'exist'+ -ire 'noun class animate' -+ alite 'existing one'. This latter process is probably of recent origin, since some variation is allowed: for instance matji:te 'good (animate)' may surface as matfoite in slow deliberate speech, 52 2 Phonology especially when matji:le 'good (animate)' contrasts with ma:tjile 'bad (animate)' (underlying form ma:tji 'bad' + -ite 'NCL:ANTM'). We mentioned in §2.1.2 above that vowel length tends to be unstable; the optionaHty of the boundary process in this case is a means of 'reinforcing' the distinction in this minimal pair." E. WEAKENfNG AND LOSS OF iON SUFFIX AND ENCLITIC BOUNDARJES. If a root-final -i is followed by an enclitic-initial y, it gets deleted, e.g. nawiki 'person' + =ya 'emphatic' -+ nawikya 'really a person'. If a root-final -i is followed by affix-initial a, then i-+ y, e.g. da:pi 'cip6 vine', Dci:pyali 'the river of Cip6 vine'. When the same item is used as a suffix and as an enclitic (see §2.6), boundary vowel fusion applies only to the suffix. This is the case with -iha 'approximative': when it is used as a suffix, vowel fusion applies, as in kadihite (from kada-iha-ite) 'blackish' (see D ll above). When it is used as an enclitic, no vowel fusion takes place, e.g. kadfte-iha (with a variant kaditeyha: cf. §2.2.2) (black+NCL:ANIM-APPROX) 'blackish' (relative to other black thing - see §3.1.3), kadti-mhll-ihii, or kadamhayha (black-PRES.NONVISAPPROX) 'it is getting darkish'. 2.5.3 Phonological processes within roots and wUhin affixes Phonological processes within roots involve syllable contraction and monophthongisation (discussed under A below). The loss of the syllable ka within affixes is discussed under B. Tbese processes occur in normal and rapid register but not in slow deHberate speech. Unstressed pi can be deleted in rapid speech in the clitic ~ pidan, =pidakil and =pida which become =dana, =daka and =da respectively; this happens only with younger speakers. A. SYLLABLE CONTRACTION AND MONOPHTHONGISATION WITHIN ROOTS AND WJTHLN AFFIXES. In polysyllabic roots the unstressed sequence uwa gets contracted to a: in rapid speech, e.g. yuwapiku 'some time' becomes ya:piku. This process is more frequently attested with younger speakers. An unstressed wa becomes a if preceded by a consonant, e.g. di-kinwa, di-kina 'he starts', yakorekwa, yakoreka 'door opening' (see discussion in §2.2.2 above). This also happens within one Aktionsart enclitic: =boswa, =bosa 'break by smashing'. An unstressed sequence ua can be simplified to a with some young people, e.g. di-panua, di-ptinoa 'he sent' may be pronounced as di-panua and as di-pana. Vowel sequence ie may be monophthongiscd to e:. e.g. na-diena 'they paddle' becomes na-de:na. The manner adverb kayu 'thus, in this way, like' can be contracted to ke: or ki:, if used in an adpositional sense 'like' as in 2.3, or 2.4. Each of these examples can be rephrased with kayu if pronounced slowly. See further discussion under H in § 11 .11 . This pronunciation is characteristic of younger speakers and innovative older speakers (e.g. Leonardo Brito and Jose Manoel Brito). ts Jacinto Brito, raised by a Tariana-speaking father and a Baniwa-speaking mother, never applies this process, e.g. ewa·- 'yellow' + -ite 'NCL:ANIM' should yield ewi te 'a yellow animate object', while Jacinto pronounces thi s as ewaite. This has been pointed out to me as a Baniwa-typc pronunciation. 2.6 Prosodic classes of morphemes and their propertie.t 53 pi-ni 2sg-do 2.3 nu-a-mi ke:-nhi I sg-say-NOM this. way-ANT 'Have it done as I said.' 2.4 ka:paru ki: hiku-ite big.bcllicd.monkcy like appear-NCL:ANIM 'someone looking like a big-bellied monkey.' If kayu is used as an adverb contraction does not occur, e.g. 2.5: 2.5 yaphini=nha kuite thing=PRES.VIS.INTER !NTER+NCL:AN[M pima-ile diha sound-NCL:ANlM he 'What thing (is it) which sounds like this?' kayu thus In rapid informal speech a stressed syllable ay becomes ey ore(:), e.g. hdyku, he()ku 'tree', diha-nai, diha-nei, diha-ne: (ART-CL:LAKE) 'this very lake', Nanayo, Naneyo 'mythical female creator' (see further examples in §2.2.2). The syllable ay in an unstressed syllable tends to contract into i: , e.g. kayda, kida 'be finished, ready', kayu, ki: 'thus'. (Leonardo Brito has ke: as a variant of ki: and of ktiyu; see also §2.6 on the realisation of proclitic kay). Also see under C4 under §2.5.2.on waiku, weiku, we:ku. B. ELISION OF THE SYLLABLE ka WITHIN AFFlXES. The syllable ka is elided in disyllabic affixes if it is underlyingly unstressed and if the resulting word is not shorter than three syllables. For instance, di-sape-kasu (3sgnf-speak-INT) 'he is about to speak' is pronounced as di-sapesu, and ma-laiida-kade (NEG-endure-NEG) 'he/she did not endure' as ma-kiiida-de . The nounyaru-maka-si 'clothes', which consists of the rootya!U 'thing', a suffix -maka 'classifier for cloth-like objects' and another suffix -si 'non-possessed' is pronounced as yatumasi. The syllable -ka-, a part of the suffix -maka, is underlyingly unstressed; it acquires the primary stress in yam-maka-si following the stress rules formulated in §2.3 .3. The syllable ka is not elided if it is underlyingly stressed; for instance, kawhi-kasu (wakc.up-NEG.FUT) 'he/she won't wake up' cannot be pronounced as *kawhi-su (the unstressed suffix -kasu 'intentional' and the stressed -kasu 'negative future (used with prefixless verbs)' are a minimal pair (see further examples in § 14.4)). The elision of the ka syllable also does not apply if the resulting word would become shorter than three syllables; thus, mc1:-kasu (l+go-INT) cannot be pronounced as •ma:su. (The suffix -kasu is similar to -maka in yam-maka-si above in that it is underlyingly unstressed but acquires surface stress in agreement with §2.3.3.) Similarly, hyu-kade (appear-NEG) 'not appear; negative response' (Chapter 17) cannot surface as *hyude, and the negative kade often used as a negative response is never just *de (such elision was described for Tucano (Ramirez 1997, vol. I: 47)). 2.6 Prosodic classes of morphemes and their properties We saw in §2.4 that Tariana distinguishes roots, affixes, proclitics, and two types of enclitics (see Table 2.5). 1\ phonological word in this language may consist of one or 54 2 Phonology more than one grammatical word.'' If a phonological word contains a proclitic and/or an enclitic, then it may consist of more than one grammatical word. ROOTS can fom1 grammatical and phonological words on their own, and they have a primary stress (with the exception of interjections discussed in §2.4.1 ). Similarly to other North Arawak languages (see Aikhenvald 1999b), roots of different word classes have different syllable structure. Verbal roots can consist of one or two syUab1es. Examples of monosyllabic verbal roots are -a 'go, say, give', -ka 'see'. -nu 'come', -wa 'try', -na 'want', -rna 'poison fish', -yu 'sieve'. Examples of disyllabic verbal roots are -ulcu 'pick, sow', -inu 'kill', -dia 'become, return', -takha 'split'. Verbs which contain more than two syllables are usually derived, e.g. -rnatji-ka (bad-thematic) 'transform into something bad', -ifiu-piru (come up-move) 'move all over'. Most stative verb (S 0 ) roots are disyllabic, e.g. leka 'break', ka:whi 'wake up', halo 'be open', ira 'need'. Nominal roots are rarely monosyllabic (one example is ta: 'large wasp'); they are more often disyllabic or even polysyllabic, e.g. -numa 'mouth', -heni 'ear', -thaku 'nose', htiye 'antshrike (type of bird)', kuida 'chestnut', yakaso/o 'apron-like traditional clothing', ukara 'shrimp', kerawiki 'a type of snufi'. AFFIXES (prefixes and suffixes) cannot forn1 grammatical or phonological words on their own. Prefixes are all monosyllabic; they are cross-referencing prefixes (§6.1.1 ), relative-attributive ka- and negative rna-. Most suffixes are monosyllabic; some are disyllabic. Prefixes are atonic; some suffixes are stressed, and some are not (see §2.3.3). PROCLITlCS can form grammatical and phonological words on their own. They are monosyllabic. They are procliticised to the following word if they are not in focus. They never have secondary stress. Proclitics are: adverb kay= 'thus' (see §3.2), kwalkwe = 'interrogative-distributive' (see §11.4), connective ne = 'then; and' and ne = 'emphatic negation; negative response' (see §3.2, and also Chapters 17, 21 and 25). Example 2.6 illustrates kay= as a proclitic. It is typically used in the resumptive phrase 'so-they-did' (used in a sense close to 'after that'). 2.6 kay=na-ni na-yha nemhani=pidana na-yarni thus=3pl-do 3p1-swim 3pl+walk=REM. P .REP 3pl-die na-ruku na: uni-se 3pl-come.down 3pl+go water-LOC 'Having done this (after this), they got drowned and went down into the water.' Kay as an independent word is illustrated in 2.7. 2. 7 kay du-ni di-kesi-pe=nuku du-kalite=pidana thus 3sgf-do 3sgnf-friend-PL=TOP.NON.NS 3sgf-teii=REM.P.REP 'After she acted this way (not any other way), she told his friends .. .' 2 16 A definition of grammatical and phonological word is found in Dixon and Aikhcnvald (2002). A granunatical word is understood as a unit consisting of a number of grammatical elements wbicb (a) always occur together, rather than scattered through the clause (the criterion of cohesiveness); (b) occur in a fixed order; and (c) have a conventionalised coherence and meaning. Also see Aikhcnvald (2002a), for a funher discussion of words in Tariana. n 2.6 Prosodic classes of morphemes and their properties 55 In 2.8, ne 'then' is procliticised to the first constituent of the clause when the sequencing of actions is not in focus. 2.8 ne=dikwe-se=tiki mawari-dapana thcn=Jsgnf+near-LOC=DlM snake-CL:HAB na-na-ni-dapana 3pl-live-TOP.ADV -CL:HAB 'And near it there is a snake house where they live.' nha they The proclitic ne= in 2.9 is an independent phonological and grammatical word. Here it has a long vowel and is homophonous with the interjection ne 'Hey!' (see §2.4.1) used as an attention getting device, e.g. ne, 0/i! 'Hey, Oli (listen to me)!' (also see Chapter 3). As an independent phonological word, the connective ne= introduces an unexpected piece of information: what appeared to be just wasteland is in fact a dangerous spooky 'snake house'. 2.9 nawiki ma:-kade-wani=naka people NEG+go-NEG-CL:ABSTR=PRES.VIS karu-na=ma=naka ne: mawari-dapana=naka afraid-AFF=EXC=PRES. VlS and snake-CL:HAB=PRES. VIS 'It is a place where people do not go, it is dangerous and it is the snake house.' Unlike other proclitics, ne= 'then, and', can take the enclitic =nuku 'topical nonsubject' (see §7.2.2). Ne=nuku (then-TOP.NON.A!S) is used if sequencing is topical, i.e. is what the story is about. There is another connective, nese 'then', which can also take the enclitic =nuku. It is cognate with Baniwa of l9ana clause connective nheue 'then'. (See Appendix on how nese was reanalysed in the Periquitos dialect.) 2.10 ne=nuku naha na-k6na nheta=pidana then=TOP.NON.A/S they 3pl-lake.onto.lap Jpl+drag=REM.P.REP 'And then they took (a piece of gold they found) into their lap.' Interrogative-distributive kwa, kwe usually occupies the first position in the clause (see §25.1 ). It can be procliticised to an immediately following word with which it forms one constituent, as in 2.1 1; or it can occur with a clausal enclitic, as in 2. 12. 2.11 kwa=yaphini=nha wa-na khesarakana muck. around what=thing=PRES. VISJNTER 1pi-OBJ 'What is mucking around with respect to us? (asked the Moon's cousins)' 2. 12 kwe=mhade nu-ni nhua what=FUT.UNCERT lsg-do I 'What sha11 1 do (to feed my children)?' The negative proclitic ne 'emphatic negation' often occurs together with other negative markers, or inherently negative words in a clause (see § 17.6). It is homophonous with the connective ne. Unlike the connective, the negative ne never takes -se 'locative' or =nuku 'TOP.NON.AIS'. In 2.13 ne is cliticised onto the first word of the clause. 56 2 Phonology 2.13 ihya nu-enipe ne=mhaisiki ma-yami-kasu NEG-die-INT you lsg-chi1dren NEG=hunger 'You my children, (you) are not going to starve (lit. die from hunger).' The proclitic ne is most frequently cliticised onto a negative predicate, e.g. ne- hyukade=pidana (NEG=not.appear=REM.P.REP) '(nothing) appeared', or onto pa:'one', e.g.