A Grammar of Ngardi: As spoken by F. Tjama, M. Yinjuru Bumblebee, D. Mungkirna Rockman, P. Yalurrngali Rockman, Y. Nampijin, D. Yujuyu Nampijin, M. Mandigalli, K. Padoon, P. P. Napangardi, P. Lee, N. Japaljarri, M. Moora, M. Mudgedell and P. Smith 9783110752434, 9783110752328, 9783110752526, 2021939135

Citation preview

Thomas Ennever A Grammar of Ngardi

Mouton Grammar Library

Edited by Georg Bossong Bernard Comrie Patience L. Epps Irina Nikolaeva

Volume 92

Thomas Ennever

A Grammar of Ngardi As spoken by Freda Tjama, Margaret Yinjuru Bumblebee, Dora Mungkirna Rockman, Peggy Yalurrngali Rockman, Yunuja Nampijin, Dampa Yujuyu Nampijin, Maudie Mandigalli, Kathleen Padoon, Payi Payi Napangardi, Patricia Lee, Nyirrpingawurru Japaljarri, Mark Moora, Marie Mudgedell and Patrick Smith

The Ngardi knowledge in this book is the intellectual property of Ngardi people. This knowledge should only be used with the written consent of the intellectual property owners and with proper attribution.

ISBN 978-3-11-075232-8 e-ISBN (PDF) 978-3-11-075243-4 e-ISBN (EPUB) 978-3-11-075252-6 ISSN 0933-7636 Library of Congress Control Number: 2021939135 Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available on the Internet at http://dnb.dnb.de. © 2021 Walter de Gruyter GmbH, Berlin/Boston Typesetting: Integra Software Services Pvt. Ltd Printing and binding: CPI books GmbH, Leck www.degruyter.com

Acknowledgements The Ngardi language, the examples used and the knowledge expressed herein are all the intellectual property of the Ngardi people. While the analysis (and any of its inevitable shortcomings) are my own, many insights and observations about the language come from Ngardi speakers themselves. This input has had a substantive role in informing various analyses presented herein and this is made explicit in the prose and tracked in various footnotes. I express my gratitude and admiration to all of the Ngardi language experts with whom I worked. They invested so much of their time and energy into teaching me the Ngardi language, sharing their histories and way of life. All of this they did with the awareness that, without their efforts, much of their shared linguistic knowledge would go unrecorded. It is my hope that our time spent together has not just been a valuable two-way learning experience but also plays a part in the future longevity of the Ngardi language. Specifically, I would like to acknowledge Marie Mudgedell and Patrick ‘Jupiter’ Smith. This grammar could not and would not have been written without them. For their direct contributions to this work, I also thank Mark Moora, Payi Payi Napangardi, Barbara Sturt, Kathleen ‘Mayan’ Padoon, Irene Padoon, Bonnie Seela, Nelly Gordon, Biddy Timbinah, Patsy Mudgedell, Marie ‘Munyumunyu’ Gordon and Bessie Doonday. For their friendship and the various aspects of desert life which they taught me, I further thank Larry ‘Kamina’ Gundora, Joe ‘Helicopter’ Jungurrayi, Gary Njamme, Bruce Njamme, Shirley Yoomarie, Theresa Sunfly, Tossie Baadjo, Gracie Mosquito, Helen ‘Jirliwa’ Nagomara, Jimmy Tchooga, Angie Tchooga, Darren Mudgedell and Gregory Mudgedell. My own involvement with Ngardi documentation efforts came about through contact with Dr Felicity Meakins, Dr Tasaku Tsunoda and Lee Cataldi. I was initially directed by Felicity towards a set of recordings made by Dr Tsunoda some 50 years earlier, on south-eastern Kimberley cattle stations in the mid 1970s. Through making contact with Dr Tsunoda, Lee Cataldi and the Board of Warlayirti Artists, I embarked on a trip to the Balgo community to return copies of these early recordings and propose a descriptive documentation project which has been ongoing since 2015. A somewhat abridged first version of this grammar was written as a Masters thesis (Ennever, 2018). My revisions to the current grammar have benefited from comments and feedback from my supervisors Felicity Meakins and Mary Laughren, and from my examiners Jane Simpson and Jean-Christophe Verstraete. Throughout the process of writing this grammar my analysis benefitted from insights and comments from Mitch Browne, Amanda Hamilton-Hollaway, David Nash and David Osgarby. I am ultimately indebted to Erich Round for initially getting me interested and involved in linguistic enquiry. For assistance with resources, I thank Tasaku Tsunoda and Lee Cataldi for sharing their recordings and materials with me. I also thank Chikako Senge, Tom Honeyman, Tom Green, Joe Blythe, Elizabeth Alvey of the UQ Fryer Library, the staff https://doi.org/10.1515/9783110752434-202

VI 

 Acknowledgements

at AIATSIS Collections and Glenn Wightman of the NT Department of Environment and Natural Resources for sharing their knowledge on various logistical and resource matters. Financial support for this work at various points has come from the ARC Centre of Excellence for the Dynamics of Language, the University of Queensland, and the Australian Linguistics Society. I thank all of these bodies for their generous support. In Balgo, I thank Warlayirti Artists Aboriginal Corporation and Kapululangu Women’s Centre for their support and encouragement. On a personal level, I thank Sheryl, Aaron, Fiona, Nick, Poppy, Lauren and Ribnga of Warlayirti; Warlayirti chairs Jimmy Tchooga and Bonnie James; Tom Vidot of the Luurnpa School; Zohl dé Ishtar, SJ and the parntany of Kapululangu; Giancarlo Mazello of Boab Health; Father Ernest Kandie of the Balgo Parish; Kim Mahood and Willi Lampert. I thank all of you as well as many more friends too numerous to name. In Brisbane, I express my gratitude for the collegial support of my friends: Amanda Hamilton-Hollaway, Celeste Humphries, Claire Gourlay, Vivien Dunn, David Osgarby, Mitch Browne, Jayden Macklin-Cordes, Janet Watts and Jacqui Cook.

Contents Acknowledgementse 

 V

Abbreviations, symbols and conventionse  1 1.1 1.1.1 1.1.2 1.1.3 1.1.4 1.1.5 1.2 1.2.1 1.2.2 1.2.3 1.2.4 1.3 1.3.1 1.3.2 1.3.3 1.3.4 1.3.5 1.3.6 1.4 1.4.1 1.4.2 1.4.3 1.4.3.1 1.4.3.2 1.4.4 1.4.4.1 1.4.4.2 1.4.4.3 1.4.5 1.4.6 1.4.7

 XXI

 1 Introduction  Linguistic preliminaries   2 Language name   2 Linguistic region and dialectal variation   3 Linguistic classification   6 Ngardi in relation to Eastern Walmajarri and Nyininy   8 Typological profile of Ngardi   9 Physical environment: Bioregional characteristics   12 Environment and bioregional features   12 Traditional life (pre-European contact)   15 Climate   17 Prehistory and time depth of Aboriginal occupation   17 Ngardi in a socio-historical context   18 Written histories of Ngardi people   19 First contact with European settlers   20 Movement of Ngardi people onto missions and cattle stations  Establishment of the Balgo Mission and the mission period (1939–1983)   24 Post-Mission life in Balgo   26 The role of multilingualism: Past and present   28 The current grammar   30 General theoretical approach   30 Data collection and corpora   31 Recording types and methodologies   33 Lee Cataldi (1990s)   33 Thomas Ennever (2016–2020)   33 Previous research   33 Survey studies   33 Preliminary descriptive work   35 Descriptive work   36 Ngardi language experts and contributors   37 Example sentences   41 Overview of topics covered   42

 21

VIII 

2 2.1 2.1.1 2.1.2 2.1.2.1 2.1.2.2 2.2 2.2.1 2.2.1.1 2.2.1.2 2.2.1.3 2.2.1.4 2.2.1.5 2.2.2 2.2.3 2.2.3.1 2.2.3.2 2.2.4 2.2.4.1 2.2.4.2 2.2.5 2.2.5.1 2.2.5.2 2.2.5.3 2.2.5.4 2.3 2.3.1 2.3.2 2.3.3 2.4 2.4.1 2.4.2 2.5 2.5.1 2.5.2 2.5.3 2.5.3.1 2.5.3.2 2.5.4 2.5.4.1 2.5.4.2 2.6

 Contents

Phonology   44 Segmental inventory   44 Practical orthography   45 Minimal phonemic contrasts   47 Consonantal contrasts   47 Vowel contrasts   49 Allophonic variation   51 Plosives   51 Voicing   52 Duration   54 Release   58 Lenition   60 Summary of characteristics of plosives  Nasals   63 Liquids   63 Tap/trill /r/   63 Laterals /l/, /ɭ/, /ʎ/   64 Semivowels   65 Glides /j, w/   65 Retroflex approximant /ɻ/   68 Vowels   72 Allophony of /a/   72 Allophony of /i/   73 Allophony of /u/   73 Allophony of long vowels   74 Phonological processes   74 Rhotic metathesis   75 Syllable deletion   75 Phonemic alternations   76 Morphophonological processes   78 Vowel assimilation   78 Fortition   79 Word and prosodic structure   79 Syllable structure   81 Stem structure   80 Positional constraints   83 Stem-initial constraints   84 Stem-final constraints   86 Consonant clusters   88 Intra-morphemic consonantal clusters  Inter-morphemic consonantal clusters  Lexical stress   93

 62

 88  91

Contents 

2.6.1 2.6.1.1 2.6.1.2 2.7 2.7.1 2.7.2

 94 Acoustic correlates of lexical stress  Stress assignment within monomorphemic words   96 Stress assignment within polymorphemic words   97 Intonation   102 Intonational contour associated with questions   102 Intonational contour associated with vocative enclitic =wu 

3 3.1 3.1.1 3.1.2 3.1.3 3.1.4 3.1.5 3.1.6 3.1.7 3.1.8 3.1.9 3.1.10 3.2 3.3 3.4 3.5 3.5.1 3.5.2 3.5.3 3.5.4 3.5.5 3.6 3.6.1 3.7

 106 Word classes  Nominals   106 Nouns   107 Adjectives   107 Action nominals   109 Manner nominals   109 Quantifiers   109 Free pronouns   110 Demonstratives   111 Ignoratives   111 Directionals   112 Locational and temporal nominals  Preverbs   113 Verbs   115 Bound pronouns   116 Particles   117 Auxiliaries   118 Subordinators   118 Coordinators   118 Modal particles   119 Propositional particles   119 Interjections   119 Ideophones   120 Clitics   120

4 4.1 4.2 4.2.1 4.2.2 4.2.2.1 4.2.2.2 4.2.3 4.2.4 4.2.4.1

Nominals I: Case and nominal morphology  Introduction   122 Word and noun phrase structure   122 Structure of the nominal word   122 NP constituency   125 NP heads   128 NP dependents   129 Domain of case assignment   130 Complex NPs   131 Coordinated NPs   132

 113

 122

 104

 IX

X 

 Contents

4.2.4.2 4.2.4.3 4.2.4.4 4.2.5 4.2.5.1 4.2.5.2 4.3 4.3.1 4.3.2 4.3.3 4.3.4 4.3.5 4.3.6 4.3.7 4.3.8 4.3.9 4.3.10 4.4 4.4.1 4.4.2 4.4.3 4.4.4 4.4.5 4.4.6 4.4.7 4.5 4.5.1 4.5.2 4.5.3 4.5.4 4.5.4.1 4.5.4.2 4.5.4.3 4.5.4.4 4.5.4.5 4.5.4.6 4.5.4.7 4.6 4.6.1 4.6.2 4.6.3 4.6.4

Inclusory constructions   133 Embedded NPs   133 NPs in apposition   134 Nominal reduplication   135 Complete reduplication   135 Partial reduplication   139 Derivational suffixes   140 Associative -jarra    140 Agentive -pirri   144 Associative2 -warnu    145 Other -kariny    146 Like -payi    149 Group -malu    149 Denizen -nga(r)na, ngawurr    150 Inhabitant -wardingki, -mardaji   151 Kinship group -ngurra    152 Expert -mayin   152 Other nominal morphology   153 Number suffixes   153 Diminutive -pardu, -pawu, -nguniny    155 Attenuative -karrikarri    157 -puru ‘while’   158 Accompaniment -jarungu   159 Intensifier -nyayirni(n)    159 Non-productive suffix -pari   161 Nominal case   161 Approach to case-marking   161 Major functions of case   166 Case forms   167 Case functions   170 Absolutive -Ø   170 Ergative -rlu, -ngku, -ju, -tu, -u   171 Dative -ku   175 Locative -ngka, -rla, -ta, -ja, -a   180 Allative -kurra   185 Lative -ngkaku, -rlaku   188 Evitative -ngkamarra, -rlamarra, -tamarra, -jamarra, -amarra  Derivational case   190 Elative -jangka, -jangu   193 Ablative -ngulu, -ngurlu, -ngula   196 Perlative -wana   197 Proprietive -kulu, -kurlu   199

 189

Contents 

4.6.5 4.6.6 4.6.7 4.6.8 4.6.9 4.7 4.7.1 4.7.2 4.7.3 4.7.4

 203 Semblative -piya  Possessive -kuny   203 Personal possessive -punta   206 Privative -wangu   208 Lacking -jirraja   210 Multiple case-marking   212 Types of ‘multiple case-marking’   212 Adnominal double case-marking   214 Multiple case-marking involving locational adjuncts  Three levels of case stacking   218

5 5.1 5.1.1 5.2 5.2.1 5.2.1.1 5.2.1.2 5.2.2 5.2.2.1 5.2.2.2 5.2.2.3 5.2.2.4

 219 Nominals II: Nominal subclasses  Quantifiers   219 mapirri ‘both’   222 Demonstratives   223 Spatial demonstrative series   223 Proximal series minya, mula, mu- ‘this one, here’   226 Distal series yalu, yala, yali ‘that one, there’   228 Adverbial and recognitional demonstratives   230 kuja ‘thus’   231 kujanguny ‘in such a way, such a kind of’   232 Recognitional nyanung(u), nyantu ‘him, her, that one’   232 Recognitional jangu, yangka ‘aforementioned, that one (you know of)’   235 Ignoratives   236 nyampa, nyampu, nyarrpa ‘what’   237 wanji ‘which/where’   240 ngana ‘who’   241 nyangurla ‘when’   243 nyajangu ‘how many’   244 nganayi ‘whatsit’   244 Ignoratives and demonstratives in combination   246 Free pronouns   247 Ergative case-marking of free pronouns   249 Directionals   252 Cardinal directions   253 -purda ‘-wards’   258 -mirnikarra ‘side’   259 -mparra, -mpala ‘along, across’   260 -mpayi ‘long way’   261 Vertical directionals   261 kanka- ‘up, upwards’   262

5.3 5.3.1 5.3.2 5.3.3 5.3.4 5.3.5 5.3.6 5.3.7 5.4 5.4.1 5.5 5.5.1 5.5.1.1 5.5.1.2 5.5.1.3 5.5.1.4 5.5.2 5.5.2.1

 215

 XI

XII 

 Contents

5.5.2.2 5.6 5.6.1.1 5.6.1.2 5.6.1.3 5.6.1.4 5.6.1.5 5.6.1.6 5.6.1.7 5.6.1.8 5.6.1.9 5.7 5.7.1 5.7.2 5.7.3 5.7.4 5.7.5 5.7.5.1 5.7.5.2 5.7.5.3 5.7.5.4 5.7.5.5 5.7.5.6 5.7.5.7 5.7.5.8 5.7.5.9 5.7.6 5.8 5.9 5.10 5.10.1 5.10.1.1 5.10.1.2 5.10.1.3 5.10.1.4 5.10.1.5

kaniny ‘down, downwards’   265 Locational nominals   267 kujapurda ‘in such a direction, this way’   268 kalkurni ‘this side’   268 kamparri ‘ahead, (in) front’   269 purdangirli ‘behind’   269 kumpirnin, kumpirnan ‘rear, at the back’   270 kita ‘outside’   270 warririku ‘close’   271 mari, mariyawu ‘a long way’   272 wananganta ‘middle’   273 Temporal nominals   273 raakarra ‘sunrise’   274 rangarni ‘morning’   275 jalangu ‘today’   275 pirranginti ‘afternoon’   277 Other temporal nominals   278 jamuny ‘ready’   278 jayanta ‘a long time’   278 lurrij, nyamu ‘finish, no more’   279 mayarra ‘again’   281 naya ‘always’   281 ngaka, ngakarranyi ‘later’   282 wurra ‘still’   282 yalawar(d)ingki ‘a long time’   283 kamparri/kamparni ‘first’, purdangirli ‘later’   284 Derived temporals   284 Action nominals   284 Manner nominals   286 Kin terms and kin morphology   287 Kinship suffixes   293 Dyadic -rlangu   293 Anaphoric propositus -nyanu, -puru   294 Possessed kin -nguniny   295 Bereft -puka   296 Unknown kinship suffix -mala   296

6 6.1 6.1.1 6.1.2 6.1.3

 297 Bound pronouns  Forms of bound pronouns   297 Comparison of bound and free pronominal forms  Composition of subject forms   300 Composition of object/oblique forms   301

 299

Contents 

6.1.3.1 6.1.4 6.1.5 6.2 6.2.1 6.2.2 6.2.3 6.3 6.3.1 6.3.2 6.3.3 6.3.4 6.3.5 6.4 6.4.1 6.4.2 6.4.3 6.4.4 6.4.4.1 6.4.4.2 6.4.4.3 6.4.5 6.4.6

 304 Code-switched object bound pronouns  Composition of locational forms   305 Composition of the reflexive/reciprocal   309 Word order and positioning of bound pronouns   310 Initial position (no encliticisation)   311 Repeated encliticisation   312 Clitic copying   313 Functions of bound pronouns   313 Subject series   313 Object/Oblique series   314 Locational series   316 Reflexive/reciprocal pronouns =nyanu, =nyanungkurla   323 Cross-referencing with oblique and locational bound pronouns  Complex bound pronouns   326 Complex bound pronouns involving object/obliques   326 Complex bound pronouns involving locationals   329 Maximal pronominal complexes   330 Hierarchies of bound pronouns   333 Dual number replacement   333 Linear ordering   335 Non-subject cross-referencing   337 Metathesis   339 Epenthesis   340

7 7.1 7.2 7.2.1 7.2.2 7.2.3 7.3 7.3.1 7.3.1.1 7.3.2 7.3.3 7.4 7.4.1 7.4.1.1 7.4.2 7.4.2.1 7.4.3 7.4.3.1

 341 Verbal morphology  Overview of inflecting verbs   343 Inflecting verb inventory   343 A continuum of inflecting verb types   346 Ngardi verbal inventory in cross-linguistic perspective   347 Conjugation classes   348 Overview of inflectional categories   349 Analysis of verb roots, conjugation markers and suffixal boundaries  Composition of the narrative inflection   358 Parasitic inflections   359 Directional inflectional paradigm   360 Regular verbal inflections   362 Past   363 Imperfective past   366 Narrative   367 Imperfective narrative   370 Potential   371 Imperfective potential   373

 XIII

 325

 353

XIV 

 Contents

7.4.4 7.4.5 7.4.5.1 7.4.6 7.4.7 7.4.7.1 7.4.8 7.4.8.1 7.4.9 7.4.10 7.4.11 7.5 7.5.1 7.5.2 7.5.2.1 7.5.2.2 7.5.2.3 7.5.2.4 7.5.2.5 7.5.2.6 7.5.2.7 7.5.3 7.5.3.1 7.5.3.2 7.5.4 7.6 7.7

Obligative   374 Hortative   375 Imperfective hortative   377 Imperative   379 Irrealis   380 Imperfective irrealis   381 Infinitive   382 Imperfective infinitive   383 Present   384 Customary   387 Anticipatory   390 Directional verbal inflections   391 Interaction of TAM and directional inflections   391 Directional inflections: functions   393 Directional imperative   394 Directional imperfective imperative   396 Directional potential   398 Directional imperfective potential   399 Directional hortative   399 Directional past   400 Directional presentative   401 Associated motion   403 Prior motion move&x   410 Subsequent motion x&move   412 Additional verbal morphology: -kujarra ‘while’, -kariny ‘reciprocal’  Verbal inflections in a regional perspective   415 Verbal reduplication   418

8 8.1 8.2 8.3 8.3.1.1 8.3.1.2 8.3.1.3

 419 Complex predication  Types of complex predication   419 Serial verb constructions   422 Complex verb constructions   424 Preverbs and preverb types   425 Nominals in CVCs   435 Issues of homophony: Nominals and preverbs, verbs and verbalisers   437 Kriol verbs in CVCs   440 Productive inflecting verbs   444 Verbs of location, change of locative relation and possession  karri- be   447 ma- get   450 marda- have   456

8.3.1.4 8.4 8.4.1 8.4.1.1 8.4.1.2 8.4.1.3

 413

 446

Contents 

8.4.1.4 8.4.1.5 8.4.1.6 8.4.1.7 8.4.2 8.4.2.1 8.4.2.2 8.4.2.3 8.4.3 8.4.3.1 8.4.3.2 8.4.3.3 8.4.3.4 8.4.4 8.4.4.1 8.4.4.2 8.4.4.3 8.4.4.4

 458 wanti- fall  yirra- place   462 kiji- throw   465 yu- give   466 Verbs of locomotion   470 ya- go   470 yapartu- run   473 ka- carry   474 Verbs of contact/force   476 pu- hit   477 la- pierce   481 luwa- shoot   484 paja- bite   487 Other productive verbs   489 nga- eat   489 nya- see   491 wangka- speak   493 ngarri- tell   494

9 9.1 9.1.1 9.1.1.1 9.1.1.2 9.1.1.3 9.1.1.4 9.1.1.5 9.1.1.6 9.1.2

 498 Grammatical relations and argument structure  Grammatical relations   498 Defining grammatical relations: Arguments and adjuncts   499 Subjects   502 Direct objects   502 Indirect objects   503 Subjuncts   508 Adjuncts   509 Non-canonical arguments   511 Comparison of cross-referencing in Ngumpin-Yapa and northern Western Desert languages   514 Argument structure   516 Non-verbal predicates   516 Equative predications   517 Identification predications   517 Ascriptive predications   517 Copula constructions with karri- be   520 Verbal predicates   521 Impersonal clauses   523 Intransitive   524 Quasi-transitive   525 Extended intransitive   527 Transitive   530

9.2 9.2.1 9.2.1.1 9.2.1.2 9.2.1.3 9.2.2 9.2.3 9.2.3.1 9.2.3.2 9.2.3.3 9.2.3.4 9.2.3.5

 XV

XVI 

 Contents

9.3.2 9.4 9.4.1 9.4.2 9.4.3 9.4.4

Conative   531 Semi-ditransitive   532 Transitive complement   533 Ditransitive   534 Case and valency alternations   538 Valency alternations   538 Case alternations in bivalent clauses   540 Case alternations in trivalent clauses   543 Complex verb constructions and predicate argument structure  The unusual verbal predicate ngarri- tell   546 Secondary predicates   548 Depictives   548 Temporal clitics as a diagnostic for depictive secondary predicates   552 Resultatives   554 Possessive constructions   555 Internal possession   557 External possession   559 Prominent internal possession   561 Adnominal clitic possession   566

10 10.1 10.1.1 10.1.1.1 10.1.1.2 10.1.1.3 10.1.2 10.1.2.1 10.1.3 10.2 10.2.1 10.2.1.1 10.2.1.2 10.2.1.3 10.2.2 10.2.2.1 10.2.2.2 10.2.2.3 10.2.3 10.2.3.1 10.2.3.2

Auxiliaries, particles and modal constructions  Auxiliaries   567 Sequential auxiliary ngu ~ nga   570 Allomorphic variation   570 Simple clause usage of ngu ~ nga   571 Complex clause usage of ngu ~ nga   573 Topic auxiliary ma   574 Topic marking function   575 Emphatic auxiliary ka   577 Modality and modal particles   578 Uncertain mayi   582 Clausal scope   583 Phrasal scope   585 Interrogative functions   585 Unrealised wayi   586 Clausal scope   586 Phrasal Scope   587 Interrogative function   588 Dubitative parda   588 Clausal scope   588 Phrasal scope   590

9.2.3.6 9.2.3.7 9.2.3.8 9.2.3.9 9.2.4 9.2.4.1 9.2.4.2 9.2.4.3 9.2.4.4 9.2.4.5 9.3 9.3.1 9.3.1.1

 567

 544

Contents 

10.2.4 10.2.4.1 10.2.4.2 10.2.5 10.3 10.3.1 10.3.2 10.3.3 10.3.4 10.3.5 10.3.6 10.3.7 10.3.7.1 10.3.8 10.4 10.4.1 10.4.1.1 10.4.1.2 10.4.1.3 10.4.2 10.4.2.1 10.4.2.2 10.4.2.3 10.4.2.4 10.4.2.5 10.4.3 10.4.4 10.4.5 10.4.6 10.4.7 10.4.8 10.4.9 10.4.10 10.5 10.5.1 10.5.2 10.5.3 10.5.4 10.5.5 10.5.6 10.5.7

 591 Hypothetical ngarda  Clausal scope   591 Interrogative function   594 Modal particles and ignoratives   594 Propositional particles   596 Delimitive (=)jala   598 Assertive kala   600 maju ‘really’   601 ngayi ‘indeed’   602 Quotative nganta   604 Counterfactual kulanganta, kulangarra, kulanga, kulangu ‘mistaken belief’   605 Assumptive jangu   607 Conjunction function   607 Certainty junga   610 Temporal and propositional clitics   611 Subsequent/persistent =lku ‘then, still’   612 Persistent eventuality function   612 Subsequent eventuality function   614 Discourse-pragmatic function   615 Subsequent (=)wali ‘then’   616 Subsequent eventuality function   617 Topic shift function   617 Apodosis-marking function   619 Quantity/degree function   620 Discourse-pragmatic function   621 Iterative =rra ‘again’   621 Additive =rlangu ‘also’   622 Restrictive =mipa   624 Comparative =katu   625 Incremental =warru   626 Vocative =wu   627 Emphatic =yi   628 Presentational =nya   628 Interjections   629 Yuwayi, yuwawu, yuwa, yuu, yawu ‘yes, ok, alright’   630 Kujarna(wu) ‘that’s it, that’s the one, (you) got it’   631 Maju ‘certainly’   631 Kanti ‘oi, hey!’   632 Kari ‘perhaps’   632 Yawiyi ‘poor thing!’   633 Wali(lku), nyamu(lku), lurrij, pinij ‘that’s it, finished, enough’ 

 XVII

 634

XVIII 

 Contents

10.5.8 10.6 10.6.1 10.6.2 10.6.3 10.6.4 10.6.5 10.7 10.7.1 10.7.1.1 10.7.1.2 10.7.2 10.7.2.1 10.7.2.2 10.7.3 10.7.3.1 10.7.3.2 10.7.3.3 10.8 10.8.1 10.8.1.1 10.8.1.2 10.8.1.3 10.8.2 10.8.3 10.8.4 10.8.5 10.9 10.9.1 10.9.2 10.9.3

Waraa, warraa ‘wow, oh no, alas!’   635 Kriol-origin particles   636 Dubitative maitbi ‘maybe, perhaps’   636 Discourse marker yuno   637 Restrictive onli   637 Contrastive nawu ‘now, now unlike before’   638 Concessive wal   639 Commands   640 Direct commands   641 Imperatives   642 Prohibitives   644 Indirect commands   646 Directives   646 Admonitives   647 Illocutionary nominals   649 Wurna   650 Wurra   650 Yurru   651 Negation   651 Negative wakurra   652 Clausal negation   653 Phrasal negation   655 Interjection function   655 Negative kula   655 waku ‘no, nothing’   656 Negative suffixes: -wangu priv, -jirraja lack   657 Lexical items with a negative meaning   658 Interrogation   659 Polar Questions   659 Tag questions   660 Information questions   662

11 11.1 11.1.1 11.1.2 11.1.2.1 11.1.2.2 11.1.2.3 11.1.3 11.1.3.1 11.1.3.2

 664 Complex syntax  Finite subordinate clauses   665 Relation between subordinators and demonstratives  Kaji clauses   667 Temporal subordinate clause   669 Conditional subordinate clause   671 General subordinator (some speakers only)   673 Kuja clauses   674 Nominal and temporal relativisation   675 Complement ‘quotative’   677

 667

Contents 

 677 Insubordinate clauses  Nyampa clauses   678 Finite coordinated clauses   679 Asyndetic coordination   680 Sequential coordination ngu ~ nga   681 Apprehensional coordination ngarda   684 Contrastive coordination kala   686 Subsequent coordination =wali   688 an ‘and’   688 tumaji ‘because’   689 bat ‘but’   690 Non-finite subordinate clauses   691 Types of predicates in non-finite subordinate clause  Types of subordinating case   695 Prior subordinators   699 Locative -rla   699 Sequential locative -ngkarla   700 Elative -jangka   701 Associative2 -warnu   703 Simultaneous subordinators   704 Locative -rla   704 Simultaneous subordinator -karra   706 Allative -kurra   707 Privative -wangu   708 Subsequent subordinators   709 Dative -ku   709 Preparative -kungarnti   711 Evitative -ngkamarra   712 Concord subordinators   713 Ergative -rlu   713 Obviative -rlakarni   714 T- and C-subordinators in combination   715 Insubordinate usages of subordinating case   717

11.1.3.3 11.1.4 11.2 11.2.1 11.2.2 11.2.3 11.2.4 11.2.5 11.2.6 11.2.7 11.2.8 11.3 11.3.1 11.3.2 11.3.3 11.3.3.1 11.3.3.2 11.3.3.3 11.3.3.4 11.3.4 11.3.4.1 11.3.4.2 11.3.4.3 11.3.4.4 11.3.5 11.3.5.1 11.3.5.2 11.3.5.3 11.3.6 11.3.6.1 11.3.6.2 11.3.7 11.3.8

Appendix: Narrative Texts  References  Index 

 765

 745

 719

 693

 XIX

Abbreviations, symbols and conventions Conventions All phonological representations are rendered in IPA characters and may or may not be enclosed in forward slashes / . . . /. Phonetic forms are enclosed in square brackets [. . .]. The main line of example sentences is transcribed in the Ngardi orthography and is plain (non-italicised) text. Italics appearing in the main line of example sentences are reserved for loan words or borrowings, for example tayimap=ma-nani ‘tie_up=get-ipfv.pst’. Italics are also used for citing lexemes within text, for example mangarri /maŋari/ ‘vegetable food’. Suffixal (derivational and inflectional) boundaries in morphologically complex words are indicated by a dash ‘-’; preverbs which are compounded with a verb in a complex verb construction are represented with an equal sign ‘=’ (e.g. parli=pi-nya ‘find=hit-pst’), which is also used to indicate clitic boundaries (e.g. ya-nta=lu ‘goimp=pl.s’). Inflecting verb roots are glossed in small caps. Preverbs are glossed with no caps. Portmanteau forms are indicated by a period ‘.’ in the glossing line. Ungrammatical forms are indicated by an asterisk ‘*’, while translations which are uncertain are indicated by an initial question mark ‘?’

Abbreviations Abbreviations used in this work follow the Leipzig glossing conventions (Comrie, Haspelmath & Bickel, 2008), augmented by additional abbreviations as required. a abs abl accomp acs adj aff agent all AM anaph anticip asrt assoc assump att aux c, cat

transitive subject absolutive ablative accompaniment accessory adjectiviser affirmative agentive allative associated motion anaphoric propositus anticipatory assertive associative assumptive attenuative auxiliary catalyst

https://doi.org/10.1515/9783110752434-204

ipfv irr lct m move&x narr neg nmlz npst o O obl oblig obv paucal perl pfv pl

imperfective irrealis locational bound pronoun middle voice prior associated motion narrative negative nominaliser non-past object/oblique (pronominal) object of transitive clause oblique (pronominal) obligative obviative paucal perlative perfective plural

XXII 

cert comp compl compar condit conj contr cf cust CVC CM d.abl dat delim den dim dir du dual dub dyad emph emph.aux erg ela ep evit excl hith hort hyp imp incep incl incre inf intens inhab ipfv IO irr kgroup kposs lack lat loc

 Abbreviations, symbols and conventions

certainty (particle) complementiser completive comparative conditional conjunction contrastive counterfactual customary complex verb construction conjugation marker directional ablative dative delimitive denizen diminutive directional dual (pronominal) dual (nominal) dubitative dyadic emphatic emphatic auxiliary ergative elative epenthetic evitative exclusive hither hortative hypothetical imperative inceptive inclusive increment infinitive intensifier inhabitant imperfective indirect object irrealis kinship group suffix possessed kin lacking lative locative

pm po poss pot pp pposs pr priv prop prs prs.aux prst prtl pst quot recip recog refl rep restr result rdp s S sembl seq.aux seq_loc sg so ssub sub SVC T term top.aux tr twd uncert unreal voc x&move 1 2 3 ♀ ♂

possessum primary object possessive potential possessive phrase personal possessive possessor privative proprietive present present auxiliary presentative presentational past quotative reciprocal recognitional reflexive repetitive restrictive resultative reduplicant subject (pronominal) intransitive subject semblative sequential auxiliary sequential locative singular secondary object simultaneous subordinator subordinator serial verb construction temporal subordinator terminative topic auxiliary transitive toward uncertain unrealised vocative subsequent associated motion first person second person third person female male

1 Introduction Ngardi is a Ngumpin language within the Ngumpin-Yapa subgroup of the widespread Pama-Nyungan language family (Australia). It is the language of the Ngardi people whose traditional lands extend south of the modern-day community of Balgo around the Northern Territory/Western Australia border, spanning both the Tanami and Great Sandy deserts. Ngardi is a highly endangered language with less than ten remaining speakers and is no longer being acquired by children as a first language. While there has been a gradual shift towards English in recent years, transmission of the Ngardi language was arrested by complex social factors associated with colonial contact in the early to mid twentieth century. This resulted in a dramatic shift towards Kukatja, a Western Desert language (Valiquette, 1993). Today, different Ngardi people are speakers of a very broad range of languages including varieties of Warlpiri (Hale, 1982; Hale, Laughren & Simpson, 2015; Jagst, 1982; Nash, 1986; Simpson, 1983), Light Warlpiri (O’Shannessy, 2005), Jaru (Tsunoda, 1981b), Kukatja (Valiquette, 1993), Pintupi (Hansen & Hansen, 1978), Walmajarri (Hudson, 1978), Wangkajunga (Jones, 2011), (Fitzroy) Kriol (Hudson, 1985) and varieties of Australian Aboriginal English. Many Ngardi people continue to live close to their traditional lands in the communities of Balgo, Malarn (Mulan), Yaruman (Ringer’s Soak) and Kururrungku (Billiluna). Others yet have come to live further afield in such places as Kununurra, Fitzroy Crossing, Lajamanu and other communities in the NT. Generally, Ngardi people have shifted to using the lingua franca of each of these communities as their primary mode of communication. Such geographic dispersion and shifting language practices have resulted in Ngardi’s precarious status as an endangered language. Many of the legacy recordings on which this grammatical description is based were collected with a range of Ngardi men and women for whom Ngardi was their first language – but who have since passed away. The youngest fluent or semi-fluent speakers are now in their fifties. While there has been previous documentation of the language – the most significant being Cataldi’s (2011) dictionary – there has been no prior published reference grammar of the language. This introductory chapter begins by detailing a number of linguistic preliminaries including the typological profile (§1.1.5), language name and neighbouring languages (§1.1.2), and linguistic classification (§1.1.3). The wider physical environment (§1.1.5) and the socio-historical contexts in which Ngardi has been spoken (§1.3) are also detailed. The origins of this work, the data, previous research and contributors to the current grammar are all described in §1.4. The chapter concludes with an overview of the topics covered in the grammar and how they are divided between chapters (§1.4.7).

https://doi.org/10.1515/9783110752434-001

2 

 1 Introduction

1.1 Linguistic preliminaries 1.1.1 Language name ‘Ngardi’ is an endonym, used by Ngardi people to refer to themselves and the language they speak.1 Published references to the language appear with the following orthographic variants: Ngardi (Cataldi, 2011; Nash, 1996), Ngari (Berndt & Berndt, 1988; Nash, 1996; Wafer, 1980), Ngarti (Wrigley, 1989) Ngardily (Green, 1988), Ngartily (Nash, 1996) and Ngadi (Capell, 1940a). The spelling adopted here (Ngardi) follows the dictionary by Cataldi (2011). The variability in the representation of the medial consonant (r, rt or rd) has a basis in actual phonetic variation. Wafer (1980), for example, reports that the Ngardi speakers with whom he worked freely alternated between pronunciations with a retroflex flap [ŋɐ́ɽɪ] and a retroflex continuant [ŋɐ́ɻɪ], which motivates the spellings ‘Ngardi’ and ‘Ngari’ respectively. Many people in Balgo today pronounce the language name as [ŋɐ́ɖɪ] or [ŋɐ́ʈɪ] and hence it can adequately be captured by a phoneme /ʈ/ (stop voicing is non-contrastive in Ngardi) and orthographically represented as ‘rt’, i.e. ‘Ngarti’. Nevertheless, due to the presence of flapped articulations by some speakers, Cataldi’s orthography involving ‘rd’ is retained (see §2.1.1). With increasing community interest in language revitilisation, there may very well be revision to orthographic rd and spelling of the language name in the future. A yet additional variant, Ngardily, reflects the pronunciation of the language name by Peggy Rockman Napaljarri (a Ngardi-Warlpiri woman). Despite Peggy’s speech incorporating a number of Warlpiri borrowings, her idiolect of Ngardi is broadly consistent with the language described herein. Importantly, Ngardi/Ngardily (a Ngumpin language) is clearly distinct from the Yapa language Ngardilpa, a Western dialect of Warlpiri as described in Jagst (1975).2 The etymology of the word ngardi is uncertain. A potentially related lexeme in neighbouring Warlpiri is Ngardiji, a term used by Warlpiri to refer to the Kaytetye people (Laughren, Hale, & Warlpiri Lexicography Group, 2007, p. 767). The less clearly related term ngarlina ‘countrymen’ is found in Mudburra while further afield in some 1 The ISO 639-3 language code for Ngardi is ‘rxd’. 2 In addition to these orthographic and phonological variants, there are two additional terms that are sometimes used to refer to Ngardi people: Bunarra (Capell, 1940b; Oates & Oates, 1970; Terry, 1926) and Tchagilin (Terry, 1926). Within the linguistic literature, Capell (1940b) describes Bunarra as a variant of Eastern Walmajarri and places it in the territories described for the latter by Hudson (1978), while Oates and Oates (1970, p. 90) locate it in the area of Sturt Creek and south to Gregory Salt Sea. Terry (1926), in contrast, locates the ‘Boonarra tribe’ in areas overlapping with Tsunoda’s (2006) account of Nyininy and Ngardi territories. Potentially Bunarra may represent a more general term for multiple groups, indicating ‘those who live by winnowing grass seed in [ˈpanːa] or wooden dishes’ (Tindale, 1974). The term punarra in Ngardi means ‘stripe’ or ‘incision carved into a shield’ and, while Cataldi does not rule out its metaphoric extension to language, she has not heard its usage as a language name (Cataldi, pers. comm. to AUSTLANG, 2007).

1.1 Linguistic preliminaries 

 3

Mirndi (non-Pama-Nyungan) languages the forms ngarli (Ngarnka) and ngarlwi (Wambaya) are used for ‘speak’ or ‘talk’ (David Osgarby, pers. comm.). 1.1.2 Linguistic region and dialectal variation The Ngardi language is situated at the intersection of a number of language groups. To the west and north are the other members of the Western Ngumpin clade of Ngumpin-Yapa: Walmajarri and Jaru respectively. To the east are the Yapa languages Warlpiri and Warlmanpa (further east and north); and to the south and south-west are the Western Desert languages Kukatja, Wangkajunga and Pintupi.3 Map 1 illustrates the approximate locations of these languages. The extent of Ngardi country varies in published literature. Capell (1940b, p. 425) reports that ‘[Ngadi is] spoken about Lewis Creek, immediately south of Sturt’s Creek’, Oates (1975, p. 133): ‘[Ngardi is] west of Walmba [sic], about the W.A. border’, and McGregor (1988a, p. 204): ‘area from Balgo to Gordon Downs’.4 Cataldi (2011) provides the most comprehensive description and this account was affirmed by Ngardi speakers with whom I worked. Ngardi country is understood to be ‘bounded by Balgo and the southern shores of Lake Gregory in the North, about the NT/WA border in the east, the Mangkayi area in the south and just to the east of the Canning Stock Route in the west’ (Cataldi, 2011, p. i). A number of important Ngardi sites are situated south of Yagga Yagga: Tarlapunta, Yayiyarr (Twin Lakes), Ngantalarra, Nakarra Nakarra, Kunakurlu, Lamanparnta, Walkali and Mangkayi (Stansmore Range). While different speakers exhibit some variation in certain aspects of their speech, there is little in the way of ‘named dialects’ of Ngardi (although cf. the relationships between Ngardi and eastern Walmajarri and Nyininy discussed in §1.1.4). One exception is a discrete term ‘Ngaliya’ (sometimes written Ngalia) as used by some Ngardi people in Balgo. This term is used to refer to both a group of people and a language variety which is described as half Ngardi, half Warlpiri by Marie Mudgedell. Catherine Berndt (1965, p. 255) recorded the same term in use among Balgo residents over 50 years earlier to refer to peoples whose country lies to the east of Balgo in the area around Tanami Downs (formerly Mongrel Downs). This usage of the term appears somewhat distinct to Tindale’s (1974) report of ‘Ngaliya’, which he identified far to the south of this area (approximately west of Yuendumu).5 A less widely used term is 3 ‘Yulparija’, most likely an exonym meaning ‘southerners’ (also cited as ‘Julbre’), is placed immediately south of Ngardi in Hobson (1990) but this is corrected to the western side of Wangkajunga in Burridge (1996). 4 This same range of occupation is described in Wafer (1980) and attributed to Sandy Jupurrurla: ‘from Wirrimanu to Gordon Downs’. This region is, however, affirmed by multiple Ngardi and Jaru people as associated with Southern Jaru peoples. 5 Note that Tindale (1974) dismisses Catherine Berndt’s report of this usage and only recognises a group identified as Ngalia centred around an area west of Yuendumu, including Nyirrpi. This is

4 

 1 Introduction

Map 1: Ngardi and surrounding languages.

‘Nambulatji’, which is similarly identified as Ngaliya-Warlpiri and is possibly associated with the wider Tanami area (Patrick Smith (pers. comm. March, 2019), and see Piele as quoted in McGregor 2004, p. 164). A cross-cutting mode of identification for at least some residents of Balgo are terms which correspond to groups of people covering wider areas, above and beyond linguistic identification. From the perspective of some individuals living in Balgo the abstracted cardinal diagram in Figure 1 was established. Nyininy (Jaru), Ngardi and Ngalia were all identified as Waringarri. Ngalia is also sometimes grouped with Ngardilypa and certain westernmost Warlpiri people under

also the main usage of the term by Warlpiri people today. Note that both usages mentioned here are distinct from the Southern Western Desert language ‘Ngalia’ spoken in western South Australia.

1.1 Linguistic preliminaries 

 5

Waringari, kayirra N - Nyininy (Jaru) - Ngardi - Ngalia (Warlpiri)

Wanparta, kalarra W

Warnayaka, kakarra E

- Walmajarri?

- Ngalia (Warlpiri) - Ngardilypa (Warlpiri)

Warrmarla, kulirra S - Kukatja?

Figure 1: Cardinal-based macro-groupings of language groups (as recorded in Balgo).

the term Warnayaka. A similar denotation of Warnayaka is mentioned by Jagst (1982, p. 1), although compare the observation of Laughren et al. (1996) that it can refer to a central northern dialect of Warlpiri. With respect to the terms Wanparta and Warrmarla, there was little insistence on how these were associated with language groups, but it was suggested to me that Walmajarri go with Wanparta6 and Kukatja go with Warrmarla.7 Importantly, these terms were discussed and recorded in Balgo. How, and whether, such terms exhibit variable or deictically-sensitive reference based on someone’s locale is uncertain. For instance, one Ngardi man recorded in the 1970s when working on Nicholson Station referred to his birthplace at Mangkayi in the Stansmore Ranges, far to the south of Balgo, as belonging to Warrmarla. Still additional variation is recorded by the Berndts who noted at Birrindudu that Walmajarri people were being referred to as Warrmala – although the Berndts rightly identified it not as a reference to a specific language group but a ‘generalising’ label (Berndt & Berndt, 1987, p. 183).

6 A term ‘Wanparta’ is also listed in Westerlund (2015, p. 169) as an endoynm for those who speak Ngarla (a Ngayarda (Pama-Nyungan) language to the distant west of Ngardi country). 7 In Warlpiri, the term ‘Warrmarla’ can be used to refer to Western varieties of Warlpiri (Laughren et al., 1996, p. 2) or it can generically refer to ‘an armed band of men’.

6 

 1 Introduction

1.1.3 Linguistic classification Ngardi belongs to the Ngumpin clade of the Ngumpin-Yapa subgroup of the PamaNyungan language family (McConvell & Laughren, 2004).8 The subgroup takes its name from two widespread lexemes for the word ‘man’ in its member languages, ngumpin (as in Gurindji) and yapa (as in Warlpiri). While Ngardi is classified as a Ngumpin language, the term for ‘man’ in Ngardi is ngantany.9 A widely accepted internal phylogeny of the Ngumpin-Yapa subgroup is provided in Figure 2, after McConvell (2009).10 Warlpiri Yapa Warlmanpa Wanyjirra Jaru (Western/Wawarl) Ngumpin-Yapa

Western Ngumpin

Nyininy Walmajarri Ngardi Gurindji

Ngumpin

Malngin Victoria River Ngarinyman Eastern Ngumpin

Bilinarra Karranga Far Eastern Mudburra

Figure 2: Phylogeny of Ngumpin-Yapa (McConvell, 2009, p. 791).

8 Highly variable classifications of Ngardi are found in the literature. The earliest classification of Ngardi is that of Capell (1940a, p. 244), whose lexicostatistical analysis placed Ngardi within his proposed ‘Mudburra group’, consisting of Yulparija, Jaru/Nyininy, Wanyjirra, Mudburra, Malngin, Ngarinyman and Eastern Walmajarri. Capell (1962) revises this to a grouping of Ngardi, Warlmanpa, and Wanayaka (a Warlpiri dialect) within the larger ‘Western Desert group’. Later work by O’Grady, Voegelin and Voegelin (1966) positioned Ngardi within the ‘Ngarrka’ (Yapa) group – a classification mirrored in Oates and Oates (1970, p. 88). These later classifications were all proposed on the basis of lexicostatistical comparison, utilising Tindale’s (1954) wordlist of Ngardi. Tsunoda (1981b, p. 5) presents a lexicostatistical analysis of his own 104-item wordlist of Ngardi which grouped Ngardi most closely with Walmajarri (77%) and Jaru (66%). 9 This is contrary to Wafer (1980, p. 3), who claims the Ngardi term for ‘man’ is ngarrka. Cf. additional lexemes for ‘man’ in other Ngumpin-Yapa languages: mawun (Jaru), piyirn (Walmajarri) and ngarrka (Mudburra, Warlmanpa). 10 Mconvell’s (2009) phylogeny is augmented here with the positioning of Wanyjirra within Western Ngumpin languages (see Senge, 2015, p.8) and the inclusion of named varieties Nyininy and Malngin.

1.1 Linguistic preliminaries 

 7

While the Ngumpin languages Kartangpurru, Wanyjirra and Malngin are absent from this phylogeny, they are understood to form something of a dialect chain spreading from Jaru in the south-west to Gurindji in the north-east (Senge, 2015). Map 2 shows the relative locations of all languages within the Ngumpin-Yapa subgroup.

Map 2: The Ngumpin-Yapa language family.

A division of Ngumpin into Western and Eastern clades as in Figure 2 is not explicitly motivated in McConvell and Laughren (2004) but has been established on lexicostatistical comparison. Throughout this work, I will refer to Wanyjirra, Jaru, Walmajarri and Ngardi as a coherent group (Western Ngumpin) since they consistently share sets of morphosyntactic features to the exclusion of the eastern Ngumpin languages (Malngin, Gurindji, Bilinarra, Ngarinyman and Mudburra). A slightly different internal grouping is found in the phylogenetic analysis of Bowern and Atkinson (2012, p. 835), which groups Ngardi and Jaru at a level distinct from Walmajarri and places the eastern Ngumpin languages (Mudburra, Gurindji, Malngin, Ngarinyman and Bilinarra) at a closer relationship with the Yapa languages.11

11 Using similar phylogenetic methods but utilising data from phonotactics, Macklin-Cordes and Round (2015) produce a phylogeny that largely mirrors Bowern and Atkinson’s (2012) findings.

8 

 1 Introduction

In terms of speaker’s perceptions, Wafer (1980, p. 3) reports that the Ngardi speakers with whom he worked regarded Ngardi as being most similar to Jaru and Nyininy, then Walmajarri, and more distantly related to Warlpiri. When working with Ngardi speakers in Balgo in the 1990s, Cataldi (2011, p. 1) reported that speakers identified Warlpiri as most similar to Ngardi. In my own experience while working with speakers throughout 2016–2020, it was commonalities in grammar between Ngardi, Nyininy (southern Jaru) and eastern Walmajarri that were most frequently and explicitly pointed out to me. Speakers with whom I worked were also aware of commonalities (typically only lexical) between Ngardi and Kukatja as well as between Ngardi and Warlpiri.

1.1.4 Ngardi in relation to Eastern Walmajarri and Nyininy From a lexical and grammatical viewpoint, Ngardi is very closely related to Eastern Walmajarri and Nyininy and these three language varieties could be considered dialects of a single language (see Cataldi, 2011, p. i). However, they are considered different languages by the respective speech communities and so they are referred to as such in the discussion here. Additionally, despite sharing a high degree of lexical similarity, there are a number of grammatical features which distinguish these varieties. While I make no attempt to precisely quantify these differences, a selection of the main grammatical differences between these three languages includes i) tense, aspect and mood (TAM) categories and forms; ii) case markers; iii) bound pronouns; and iv) clausal negation. In the comparisons that follow, abbreviations are Eastern Walmajarri (eW), Jaru, (Ja), Ngardi (Ng) and Nyininy (Ny). Beginning with the verbal system, there is significant variation in the form, number and type of verbal inflections in Ngardi, Eastern Walmajarri and Nyininy. While all three languages share a ‘potential’ inflection of the form -Cu with a variety of modal functions, Ngardi and Nyininy have a discrete ‘irrealis’ inflection of the form -Nngi which is used to mark unreal or uncertain events, especially in hypothetical contexts (e.g. in past and non-past conditionals). In western varieties of Walmajarri and certain varieties of Nyininy (as formerly spoken at Gordon Downs) an inflection -Ca is used for imperative and other modal functions (e.g. past irrealis). In Ngardi, an inflection -Ca is strictly reserved for imperative functions. Additionally, eastern Walmajarri lacks any verbal directional morphemes (clitics or suffixes {-rni, -rra}). Nyininy possesses the ‘hither’ form -rni but it only occurs with imperative and potential tense values. Ngardi possesses both -rni (‘hither’) and -rra (‘hence’) and these combine idiosyncratically with TAM categories (see §7.5). Further comparison of TAM categories in Ngumpin-Yapa langugaes more generally is provided in §7.6. The forms and functions of case markers in these three languages also diverge. Some notable examples include the allative: -jawu (Ny, eW) vs -kurra (Ng); ablative -nginy (Ny, eW) vs -jangka (Ng); privative -mulu(ng) (eW, Ja) vs -wangu (Ng). In each

1.1 Linguistic preliminaries 

 9

case the ‘divergent’ Ngardi form is actually shared with the Yapa language Warlpiri. Ngardi, Eastern Walmajarri and Nyininy differ with respect to the case-marking of free pronouns. In Ngardi and eastern Walmajarri, all case marking attaches regularly to a single invariable root of the free pronoun, as per the Ngardi: ngaju-ngku (1sg-erg) and ngaju-ngka (1sg-loc). In Nyininy, however, non-core cases (i.e. dative, locative, allative, ablative etc) attach not to the expected root but to a modified stem. Thus the Nyininy equivalents: ngaju-nggu (1sg-erg) and nganinginy-ja (1sg-loc). In terms of their system of bound pronouns, eastern Walmajarri appears to be somewhat distinct from Ngardi and Nyininy. Eastern Walmajarri (like other varieties of Walmajarri) has a distinct series of bound pronouns used to cross-reference absolutive marked direct objects (e.g. =ji 1sg.o) as opposed to dative marked arguments (e.g. =ju 1sg.dat). Ngardi and Nyininy, however, use just a single series for crossreferencing absolutive and dative NPs except in the third person singular for which there is a distinction between-ϕ (3sg.o) and =rla (3sg.obl) respectively. Finally, clausal negation also varies across these three languages. Ngardi and Nyininy have a negative operator of the form wakurra, while Eastern Walmajarri has ngajirta. Eastern Walmajarri is similar to Ngardi and Nyininy however in forming negative commands via ngajirta + potential -Cu (cf. wakurra + potential -Cu in Ny and Ng). This can be contrasted with other varieties of Walmajarri as described in Hudson (1978) which form negative commands via ngajirta + imperative/irrealis inflection -Ca.

1.1.5 Typological profile of Ngardi The phonological system of Ngardi is unremarkable when compared to other languages within the Pama-Nyungan family. Ngardi exhibits many widespread phonemic properties of the language family: a three-way vowel system, numerous place contrasts across a paired series of nasal and oral stops, a lack of fricatives, and no voicing contrasts. With respect to place contrasts, Ngardi is of the ‘double-apical’ type (contrasting retroflex and non-retroflex apicals), while lacking a lamino-dental/ lamino-palatal distinction reported in other areas of the continent (Hamilton, 1996). Of particular interest is the presence of allophonic flapping of the retroflex approximant phoneme /ɻ/ – consistent with the southernmost varieties of Jaru (Tsunoda, 1981b) and constituting an interesting analogue to fully phonemicised flaps in western dialects of Warlpiri (Laughren & Ingram, 1999). Ngardi is further notable for lacking the sets of historic lenition changes observed for peripheral stop consonants in other Ngumpin languages (McConvell, 1997; McConvell & Laughren, 2004). Seven major classes of lexemes are recognised in Ngardi: nominals, preverbs, verbs, bound pronouns, particles, interjections, and clitics. Ngardi is morphologically agglutinative and exclusively suffixing. While Ngardi is notable for exhibiting patterns of case stacking (Austin, 1995; Plank, 1995), the language is fairly moderate on the spectrum of analytic↔polysynthetic word structure. Words rarely exhibit more

10 

 1 Introduction

than three or four morphemes. The complex verbal word in (1) has six morphemes, one of which is a portmanteau (of tense and aspect), and two of which are enclitics. (1)

Pirdiny=ma-nani-nyirra=lu=nganpa. arise=get-ipfv.pst-narr=3pl.s=1pl.excl.o ‘They made us stand up.’ (YMN: LC23b: 472309_475708)

Noun phrases (NPs) in Ngardi have ‘general number’ and can be cross-referenced with singular or non-singular bound pronouns when fulfilling argument roles (§6). Number of NPs can be optionally specified by dual or paucal suffixes (§4.4.1) or by use of a class of modifier NPs, termed ‘quantifiers’ (§5.1). Nouns are not classified by a gender or noun class system and NPs are not specified for definiteness. NPs in argument roles are marked for one of five grammatical cases (ergative, absolutive (unmarked), dative, locative, and allative). Non-argument NPs can occur with all of the grammatical cases (in non-argument contexts, e.g. secondary predicates §9.3) as well as the non-grammatical lative and evitative cases. Case marking in Ngardi typically exhibits complete concord across constituents of a noun phrase and only rarely shows evidence of phrase-final marking as found in the neighbouring Warlpiri. The morphological property of complete concord is demonstrated by the allative in (2). (2) Ya-nta-rra ngurrara-kurra nyuntu-punta-kurra wirrirl! go-imp-hence country-all 2sg-ppos-all back ‘Go back to your own country!’ (NMN: Manungka_01-035167) Ngardi, like Warlpiri, possesses a class of eight morphological suffixes termed ‘derivational case’: elative, ablative, perlative, proprietive, possessive, personal possessive, privative, and lacking (§4.6), which behave somewhat like case inflections (§4.5) on the one hand and somewhat like derivational suffixes (§4.3) on the other. An example is the elative -jangka which has a function typical of a regular spatial case suffix in (3) (it shows agreement across a noun phrase) but has a function more akin to derivational morphology in (4). (3) Ya-ni yalu-jangka murlukun-jangka. go-pst that-ela jar-ela ‘He got out of that glass jar.’ (MMN: TEN1-2017_016-02: 40743_44395) (4) Purdangirli-jangka-rlu purda=nya-nkura. behind-ela-erg listen=see-hort ‘People nowadays/young ones (lit. ‘those from behind’) must listen.’ (MAM: TEN1-2016_002-01: 470178_475997).

1.1 Linguistic preliminaries 

 11

Case suffixes in Ngardi are also used in a subordinating function to encode relationships between finite main clauses and non-finite subordinate clauses. So-called ‘subordinating case’ serves to mark control relationships between main and subordinate clause arguments as well as indicating various temporal or semantic relations between the two clauses. In (5), the elative case suffix -jangka marks the constituents of a nonfinite subordinate clauses, encoding prior relative tense and subject control. (5) Wirrirl=ya-ni paja-rnu-jangka purnu-jangka. back=go-pst bite-inf-ela wood-ela ‘He came back, having chopped the wood.’ (PSM: TEN1-2018_037-01: 1120922_1124405) Ngardi has properties of both a head-marking and a dependent-marking language (Nichols, 1986). Grammatical relations are signalled by a combination of case-marked NPs and cross-referencing bound pronouns. Frequent ellipsis of NPs is observed which results in arguments solely being registered by the bound pronouns. Ngardi is a morphologically ergative language. Nominals and free pronouns take case markers registering core relations of A vs S/O in an ergative case-marking alignment. Bound pronouns, in contrast, exhibit a nominative-accusative alignment of the core relations A/S vs O. All relevant syntactic processes such as reflexivisation, control of subordinate clauses and serial verb constructions provide evidence for nominativeaccusative syntax. Argument structure in Ngardi is further notable for the fact that a range of NPs marked with certain ‘non-core’ cases are also cross-referenced by the bound pronouns. Certain dative case-marked nominals are cross-referenced by one series of bound pronouns (labelled ‘oblique’ and identical to the series of forms used to cross-reference absolutive objects barring the 3sg), and certain locative or allative case-marked nominals are cross-referenced by another series of bound pronouns (labelled ‘locational’). Ngardi shares many of the same features of syntactic non-configurationality as described for Warlpiri (Austin & Bresnan, 1996; Hale, 1983). Evidence for rigid phrase structure above the noun phrase (NP) is limited – word order is syntactically free and likely only conditioned by discourse/information structure factors. This is an area of the grammar in need of much greater research than could be provided here. Like many languages of northern Australia, Ngardi has two discrete classes of verbs: an open, semantically rich class of uninflecting preverbs and a closed class of inflecting verbs. Different subclasses of preverbs with different properties can be identified but generally speaking they are restricted to the preverbal position as in (6). (6) Parli=pi-nya=rlipa=wu wirrpa karnti. find=hit=pst=1pl.incl.s=voc many bush_potato ‘We found lots of bush potato!’ (PSM: TEN1-2017_013-02: 841547_844182)

12 

 1 Introduction

The class of inflecting verbs comprises only 85 lexemes. These verbal roots inflect for tense, aspect, mood, directionality, and associated motion. The two directional features, ‘hither’ (towards the speaker) and ‘hence’ (away from the speaker), are notable for their non-orthogonal interaction with the TAM inflections. Ngardi also possesses two ‘associated motion’ (Koch, 1984) suffixes which appear to derive from historic complex predicate structures involving non-finite + finite verb combinations. Complex clauses in Ngardi take the form of both hypotactic and paratactic clausal relations. A major morphosyntactic division exists between finite and non-finite and subordination. Non-finite subordinate clauses obligatorily consist of a non-finite predicate (an infinitive verb, a preverb or even a nominal). Constituents of non-finite clauses are marked by case markers functioning as subordinators as described for the elative in (5). Case suffixes variously serve to link subordinate clauses to main clauses via temporal and other semantic relationships, as well as specifying the relationships between arguments of main and subordinate clauses (e.g. subject control, object control, non-subject/non-object control). Finite subordinate clauses comprise two finite clauses which are linked via a subordinator, some of which are homophonous with demonstratives. Finally, coordinate clauses in Ngardi may be formed overtly through the use of one of a small set of polyfunctional particles (with coordinative functions) or through simple apposition (asyndetic coordination).

1.2 Physical environment: Bioregional characteristics 1.2.1 Environment and bioregional features Ngardi country extends across both the western reaches of the Tanami Desert and the north-eastern region of the Great Sandy Desert, close to the border of Northern Teritory (NT) and Western Australia (WA). A striking feature of the Tanami region is its geological age – the weathered surface rock in these areas is some of the oldest in Australia and indeed the world, dating to the Proterozoic (Pre-Cambrian) (Pillans, 2018). The regolith (heterogenous surface layer) is highly weathered and comprises mineral profiles high in ferricretes and bauxites. These mineral profiles give the ground a distinct redness as found through arid central Australia and can be observed exposed in erosion-resistant caps on mesas (munturt) as in Image 1. The present-day community of Balgo itself is located atop an eroded ridgeline that drains from a wide area in a westerly direction towards Paruku (Lake Gregory), a large endorheic lake some 40 km west of the community. A short distance from the eastern edge of the lake is the small community of Malarn (Mulan, pop. approx. 100), while to its north is another community, Billiluna (or Kururrungku, pop. approx. 150). South and south-west of Balgo, lateritic gravel plains give way to extensive sand ridges consistent with those that typify the Great Sandy Desert (Mackay) bioregionalisation zone (see the Interim Biographic Regionalisation of Australia). This is the country in which

1.2 Physical environment: Bioregional characteristics 

 13

Image 1: Weathered slopes of erosion-resistant caps on mesas (munturt) south of Balgo.

the former community of Yagga Yagga and the outstations of Lamanparnta, Walkali and Piparr are located. The Great Sandy Desert is a vast area dominated by longitudinal tali ‘sand dunes’, which run some 1000 km across the Canning Basin. Aside from sandhills, the overall topography of Ngardi country is generally flat. The highest point within Ngardi country is found in the south around Mangkayi (the Stansmore Ranges) which rises to just 510 m above sea level (Blake & Yeates, 1976). South-eastern Ngardi country about the NT/WA border is dominated by extensive salt-lake chains – lakes White, Willis, and Hazlett. Map 3 shows Balgo in its local regional context. The major ecological zone throughout Ngardi country can be classified as ‘mixed shrub’ steppe (Beard & Webb, 1974) and is dominated by acacia species and spinifex (Cane, 1984). A wide diversity of acacias are found across Ngardi country including marntirla ‘sandhill wattle’, matu/wilpiya ‘silver witchety’, kirriliny ‘pindan wattle’ and manja ‘mulga’, to name but a few. Elsewhere on Ngardi country, large tracts of low-lying mungily ‘samphire shrub’ can be found, especially in the south-east of Ngardi country, due to their salt tolerance. Other widely distributed plant species in the region include kuntupungu ‘bloodwood’ from which the edible galls (ngarlukutu or kanta) are harvested, jipari ‘Sturt Creek mallee’ and malarn ‘river red gum’, after which the community to the west of Balgo is named. Hardwoods such as pangkurna or kunanturu ‘desert oak’ and yinirnti or kumpupanu ‘bean tree’ were traditionally

14 

 1 Introduction

Map 3: Balgo, Billiluna and Mulan and surrounding historic communities/outstations.

utilised by Ngardi people for manufacturing kana ‘digging sticks’, luwanja ‘coolamons’ and mirta ‘shields’. Some specific areas on Ngardi country have semi-permanent surface water or nyarna ‘deep water’ (for example Yayiyarr ‘Twin Lakes’) supporting water-dependent eucalpyts such as jarlipari ‘ghost gum’, while others areas have mulju or ngawiri ‘shallow soaks’. Larger trees can be found lining semi-seasonal creek beds. These primarily include species of eucalypts; yilangkiyi ‘blue mallee’, jalipari ‘river red gum’ and tinjil ‘coolibah’. Many other areas of Ngardi country are nevertheless devoid of significant quantities of readily accessible surface water or larger eucalypts for shade. Many researchers of Western Desert life have noted the formidable constraints on human habitation presented by the water and food scarcity of the region (see Cane, 1984, p. 393 for a summary).

1.2 Physical environment: Bioregional characteristics 

 15

1.2.2 Traditional life (pre-European contact) Prior to contact with non-Indigenous people, Ngardi people lived a semi-nomadic lifestyle consistent with that of many Aboriginal groups of the northern Western Desert region (see, e.g. Cane, 1984, 1987; Thomson, 1975). Availability of resources (namely the seasonal presence of food and water) played a major role in determining the distances people travelled, the sites visited, and the time spent in various locations. Resource-driven concerns were also set alongside a need to fulfil various social and ceremonial obligations across numerous sites throughout the year. Ngardi people – like their Kukatja and Pintupi neighbours – lived and moved in quite small groups, often consisting of just one or two family units (cf. Thomson, 1975). During wurrkarla ‘growing season after rain’, Ngardi people gathered in much greater numbers and would travel greater distances to attend ceremonies and to visit areas which had not been foraged over since the previous year (Cane, 1984, p. 394). When staying in temporary camps, Ngardi people would typically construct low-lying wirriny ‘windbreaks’, structurally supported by manja ‘mulga scrub’ and built up with yuka ‘spinifex’ and brushwood. Such structures were particularly necessary during jarrilyilyirla (cold (easterly) wind time), when the cold easterly winds sweep across the desert. Despite the fundamentally semi-nomadic lifestyle of Ngardi people, there is evidence that a range of food storage techniques and strategies had been developed – predominantly involving dry food stuffs (Cane, 1987; Thomson, 1962). Across the whole desert zone, lukarrara (Fimbristylis oxytachya), mungily (Tecticornia verrucosa) and other seed types were a staple of Aboriginal people’s diets. The collection, preparation (husking, winnowing, grinding) and cooking of seeds was a major preoccupation of Ngardi women (see relevant studies by Cane 1984, 1987). The seeds of various species of yuka ‘grass’ were utilised to make wanapirti ‘seed cakes’. Women would spend significant time winnowing seeds from chaff and sand/grit with a kulinjirri (a coolamon designed for winnowing) in the preparation of a paste which would either be eaten raw or cooked in a type of damper. Grindstones or jungari formed part of an essential toolkit for Ngardi people. Some were transported between camps, while other heavier base stones were left in seasonal camps so that grains could be readily processed in situ. Various species of acacias, including kilkirdi (Acacia holocericea), were also valued as sources of food by Ngardi people. However, many species of acacia seeds of this region are toxic and very little is known about how the Ngardi and Kukatja prepared such foodstuffs (Cane, 1987, p. 405). The woody stems and root systems of many acacias are also home to the prized laju ‘witchetty grubs’, which are extracted carefully with ngarnngu ‘hooked barbs’, made from whatever materials are available. Vast tracts of land were utilised and foraged over for their perennial crops of karnti or puurda ‘yam, bush potato’ (Ipomea costata), jalparr or kumpupaja ‘bush tomato’ (Solanum chippendalei), and jirrilpaja ‘bush carrot’ (Vigna lanceolata). Ngardi people continue to forage these bush foods when in season today. Ngardi people also monitor the activities of various animals that serve as seasonal indicators as to

16 

 1 Introduction

when conditions are most suitable for certain activities such as hunting and foraging. For example, when the Burton’s snake-lizard (Lialis burtonis) is heard producing its distinctive call ‘lirtlirt’ – from which it takes its Walmajarri/Ngardi name – it is considered a suitable time for puurda ‘yams’ to be harvested. Details of when various foodstuffs in this region are in season are provided in Cane (1987). While the women conducted much of the gathering of dietary staples, men would often be concerned with the hunting of larger game, dependent on seasonal and local availability. Particularly prized meats were various macropods jaji ‘kangaroo’, mala ‘rufous hare-wallaby’, parntarrngarna ‘rock wallaby’ and jiya ‘euro’, but also lizards such as jarrampayi ‘perentie’ and jarany ‘roughtail lizards’, and snakes such as walyarrangarnujarra ‘Ramsay’s python’. Ngardi men utilised a range of spear types – some specific to hunting, others for fighting. The kirnimiliny, for example, was a large hunting spear to which a jimpirla ‘stone spearhead’ would typically be attached using wajarn ‘spinifex resin’ and fastened with jiliwa ‘animal tendon’. Other spear types include the malmurru (a small, lightweight spear) and the wurrumpurru (a shovel-nosed spear). Ngardi country was once home to numerous smaller marsupials: kalatawurru ‘the desert rat kangaroo’, kanakarlumpayi ‘fat-tailed mouse’, nyarlku ‘bilby’, mingajurru ‘golden bandicoot’, milpatiri ‘spectacled hare-wallaby’, and the mala ‘rufous harewallaby’ to name but a few. Many of these species are now critically endangered, if not extinct. Some, such as the kakarraturl ‘northern marsupial mole’ are believed to have stable populations but are rarely spotted. The reduction in numbers of all marsupial species is primarily due to competition from and predation by introduced species. It is clear that traditional Aboriginal fire management practises also played an integral role in the ecological balance of landscapes across the Western Desert (Bird et al., 2008; Burrows et al., 2006) and Tanami (Nash, 1991; Vaarzon-Morel & Gabrys, 2009) regions, and their cessation has had dramatic effects on the ecology of these regions (Bolton & Latz, 1978). Some of the earliest satellite photographs of the wider region (c. 1953) reveal complex mosaics of small burnt patches of vegetation, producing a checkerboard of post-fire succession zones that both encouraged growth and limited the risk of large scale, devastating bushfires (see Burrows & Christensen, 1990). Later satellite images up to 1986 reveal that this mosaic has since been completely lost due to the last Pintupi people moving off their country west of Lake Mackay. In its place, satellite images have revealed large tracts of senescing vegetation alongside huge tracts of burnt country brought about by lightning-caused bushfires.12 In recent decades, various introduced species, including camels (kawujuwal), feral cats (pujikat) and wild horses (timana), have all contributed to the decline of native fauna (see Burrows, 2018). Wild horses predominate in northern and north-western tracts of Ngardi country (often centred around more permanent water sites), while 12 The study by Burrows and Christensen (1990) focused on an area west of Lake Mackay in Pintupi country (south of Ngardi country) – one of the last areas in Australia where Aboriginal Australians were living in an entirely subsistence fashion.

1.2 Physical environment: Bioregional characteristics 

 17

in the more arid zones to the south, large groups of camels predominate. Feral cats are found across all of these ecological zones and are found even within some of the most arid regions. Nevertheless, certain populations of many unique native species continue to thrive on isolated Ngardi country; including reptiles such as the striking katapurda ‘mountain devil’ (Moloch horridus), the deadly nguwa ‘desert adder’, and large numbers of jiya ‘euro or rock wallaby’ and marrany ‘dingo’.

1.2.3 Climate Balgo is located in a semi-arid zone that is within the influence of monsoonal seasons which dominate weather patterns across northern Australia. The mean maximum temperatures in Balgo (1950–2016) range from 26.1 °C in June to 38.7 in November/ December.13 Overnight mean lows range from 12.5 °C in July to 25.2 in January. Balgo exhibits some climatic features of a pre-monsoonal ‘build up’, with temperature peaks arriving in December prior to the heaviest rains. The wettest three months are January to March, with a mean maximum precipitation of 84.2 mm in February and a mean annual precipitation of 356 mm/year. The period following rains is known as wurrkarla (lit. ‘time of green’) when rapid growth of vegetation occurs and typically proceeds through March until May (Cane, 1987, p. 394). This is followed by pirriyarla or jarrilyiyirla, ‘cold weather time’, which is characterised by strong easterly and south-easterly winds. The driest three months are July to September. Rainfall is nevertheless highly unpredictable and annual rainfall has been recorded with totals as low as 90 mm (1958) and as high as 907 mm (1974) (Bureau of Meteorology). While the data provided here offer some indication of climate for Ngardi country in its northern extent, the influence of monsoonal weather patterns decreases as one moves south through Ngardi country. The southernmost tracts of Ngardi country are increasingly more arid with less annual precipitation. In the eastern regions of Ngardi country, annual average rainfall has been recorded around 200 mm (Kearney, 1985, p. 23). There are, however, no available climate data for central Ngardi country south of Balgo.

1.2.4 Prehistory and time depth of Aboriginal occupation When considering the time depth of Aboriginal occupation of Australia, it is generally agreed that the most extreme arid zones in Australia (including the Tanami and Great Sandy Deserts) were settled later than the rest of the continent (O’Connell & 13 Accurate climate data for Balgo is limited to the period 1950 to 2016. However, some precipitation data stretches back to 1940 and includes data gathered at both ‘Old Mission’ (30 km west) and present-day Balgo (Wirrimanu).

18 

 1 Introduction

Allen, 2012). While there have been exceedingly few archaeological studies on Ngardi country itself (although see Cane and Novak (1981)), neighbouring sites provide approximate indicators for the time depth of human occupation of the region. West of Ngardi country, at Parnkupirti ‘Site Three’ (eastern edge of Lake Gregory), stone artefacts associated with human occupation have been located in sediments dating to 37 thousand years ago and most probably in the vicinity of 45–50 thousand years ago (Veth et al., 2009). Taken in combination with evidence of habitation in country to the south-east of Ngardi country, south of Mt Liebig at Puritjarra, dating to 35 thousand years ago (Smith, 2006), as well as far south as the Nullabor Plain (e.g. Allen’s Cave, c. 40,000 years (Munt, Roberts & Gorman, 2018; Walshe, 1994)), there is sufficient archaeological evidence to support occupation of the entire north-west arid zone for a period approaching 50,000 years (Veth, McDonald & de Koning, 2018). The mounting evidence for long-term occupation of this region is set in contrast to the traceable time depths of Aboriginal languages spoken in the Western Desert region at the time of contact. Putative and highly approximate dates proposed for the expansions of both Pama-Nyungan languages, and (more locally) the Western Desert languages, are argued by some to be of a much more recent time depth at approximately 5000 years ago for Pama Nyungan (McConvell, 1996a; McConvell & Bowern, 2011), and approximately 1500 years ago for the Western Desert languages (McConvell & Laughren, 1996; Veth, 2000).

1.3 Ngardi in a socio-historical context The current, endangered status of the Ngardi language is a direct result of the recent history of Ngardi people and their contact with early colonists. While the histories of Ngardi people are complex and multifaceted, the primary factors that led to the endangerment of the Ngardi language were the encroaching influences of the pastoral and mining industries, along with the long-term direct effects of the Balgo Mission. On both the cattle stations and in the Balgo Mission, a great diversity of culturally and linguistically distinct groups were brought together into new, centralised communities. For complex socio-historical and demographic reasons, the Ngardi language was subordinated to the dominance of other traditional languages. In Balgo the dominant language was a variety of Pintupi referred to locally as Kukatja (see Cataldi, 2011, p. 1; Valiquette, 1993). Cataldi (2011, p. 1) points to the role of the dormitories, the segregation of children from their parents, and the linguistic repertoires of the first children gathered into the dormitories as being the deterministic factors in paving the way for the shift towards a Kukatjadominated speech community. Marie Mudgedell, a Kukatja-Ngardi speaker, reflected that Ngardi was even discouraged from being spoken by Kukatja people and she attributes this as one of the reasons for which Ngardi is no longer spoken by younger people. This section gives an overview of the history of Ngardi people and the historical developments that have given rise to the current sociolinguistic situation in Balgo.

1.3 Ngardi in a socio-historical context 

 19

First, a summary of the published literature which engages with Ngardi peoples is provided in §1.3.1. The earliest documentation of contact between Ngardi people and Australians of European descent is described in §1.3.2, the development of the mission in §1.3.4, post-mission life in §1.3.5 and a summary of multilingualism (past and present) in §1.3.6.

1.3.1 Written histories of Ngardi people There is exceedingly little written history which focusses on Ngardi people or the Ngardi language. Nevertheless, numerous incidental accounts of Ngardi people, their culture and their country can be found across a broad and diverse literature. Primary historical accounts from Ngardi people themselves can be found in Napanangka et al.’s (1997) Yarrtji: Six women’s stories from the Great Sandy Desert, a collection of stories and histories from six senior Ngardi and Kukatja women from Balgo and Yakka-Yakka. These oral histories include accounts of life prior to the arrival of the missionaries, settlement at the missions and accounts of first contact with Europeans. Many of these narratives form part of the Ngardi corpus and have informed the linguistic analyses in this grammar. Additionally, records of Ngardi people have been compiled as part of the due process of assembling cases for various land claims: Report No 22: ‘Warlpiri, Kukatja and Ngardi Land Claim’ (Kearney, 1985). Report No. 42 ‘Tanami Downs Land Claim’(Olney, 1992), and, more recently, Ngururrpa (Cane, 2006) and Lappi Lappi/Ngulupi (Tex on behalf of the Lappi Lappi and Ngulupi Claim Group v State of Western Australia, 2018). The history of the Balgo Mission itself has also been documented by a range of missionaries, (art) historians and anthropologists. Zucker (2005), Byrne (1989) and Ganter (2016) all provide accounts of the historical movements of the German Pallottine missionaries responsible for the development of the first Balgo Mission.14 Father Anthony Peile worked with Balgo people as a priest first in Halls Creek (1966–1973) and later in the Balgo community (1973 – late 1980s). Piele’s detailed and wideranging notebooks provide impressive documentation of the Kukatja language, with particular emphasis on Kukatja biocultural knowledge and concepts of health and healing. These notebooks formed the basis of later publications: A basic Kukatja to English dictionary (1993) edited by Hillaire Valiquette, and Body and soul: An Aboriginal view (1997) edited by Peter Bindon. There are a number of accounts examining the development of Balgo in the post-mission period, both with respect to the community (Dé Ishtar, 2005) as well as the development of art practices and the highly successful art industry (Watson, 1996, 1997, 2003; Carty, 2011). Anthropological accounts of Abo-

14 Primary sources for Ganter’s account predominantly come from the transcribed handwritten notes of Alphonse Bleischwitz and the biography of Brother Frank Nissl.

20 

 1 Introduction

riginal peoples who came to reside in Balgo can be found in Berndt (1970, 1972), Cane (1984, 1987), Poirier (2005) and McCoy (2008). Myers (n.d.) and Young, Marchevsky and Wick (1981) describe the burgeoning ceremonial life of Balgo in the 1980s, which came to be known as ‘Balgo Business’ (julurru) and was well known right across the Kimberley and into the Northern Territory. Musical traditions, including various song sets, belonging to Aboriginal people of the Kutjungka15 region (including Ngardi people) have been documented by a range of ethnomusicologists and linguists. Songs of various genres were recorded in Balgo by Richard Moyle (1997), Sonja Peter, Lee Cataldi and the women of the Manungka Manungka Association (e.g. Napanangka et al., 1997) and more recently by Turpin and Meakins (2019). In regards to social and governmental policy in the region, Cane (2016) provides an analysis and critique of governmental policy associated with the movement of Aboriginal people in the Balgo area to live on outstations in the late 1990s. Various geographical survey and documentation work of Ngardi homelands has also been conducted, timed with the beginning of oil and mineral exploration in the region in the late 1980s (see Cane, 2016 for a summary). Relevant archaeological reports on the prehistory of the wider region were mentioned in §1.2.4. Detailed archaeological studies which examine massacres of Aboriginal people in the Kutjungka region include Smith et al. (2017) and Smith (2016) for the massacre at Purrkuji (Sturt Creek Station) on Nyininy country (just to the north of Ngardi country).

1.3.2 First contact with European settlers There are few explicit written accounts of first contact between Europeans and Aboriginal peoples living on or near Ngardi country and none that reference Ngardi people directly. What can be surmised of the earliest settler contact with Ngardi people can be gleaned from examination of firsthand accounts (e.g. Napanangka et al. (1997)) and further appreciated from what has been documented under the history of the Pintupi (see Long, 1989), since Pintupi and Ngardi peoples’ contact history is closely connected. The Kimberley region in north-western Australia was one of the most recent regions within Australia to be subjected to non-Indigenous exploration and occupation (Smith, 2000). Prospecting cattlemen initially favoured the Kimberley for the region’s perceived suitability for cattle and an attitude that the large numbers of local Aboriginals would provide a ready source of labour (Bolton, 1958). The first forays of European explorers into Kutjungka lands was in 1855 with the expedition of Sir Augustus Charles Gregory who travelled down Sturt Creek as far as its terminus (Paruku), which he named Lake Gregory. European exploration of the Tanami and northern Great Sandy 15 The term ‘Kutjungka’ is used to refer to the region which encompasses the three communities of Balgo, Mulan and Billiluna – the residents of which all share familial, linguistic and cultural connections.

1.3 Ngardi in a socio-historical context 

 21

Desert regions began in the late 1870s when Peter Egerton-Warburton (1872–1874) traversed southern Warlpiri (Ngaliya) country as far west as ‘Red Cliff Pound’ and ‘Hidden Valley’ (near Lappi Lappi and Lake Hazlett) and later Nathaniel Buchanan (1880s) pioneered the first European overland routes through the Great Sandy Desert north of Lake Mackay.16 Carnegie’s expeditions across the Gibson and Great Sandy Deserts followed in 1896–1897, traversing Ngardi country south-to-north and finishing in Halls Creek. There were no documented expeditions directly into Ngardi country again until 1933 when Michael Terry ventured into a low-lying valley in the vicinity of Lake Hazlett (south-eastern Ngardi country) on his expedition across the NT/WA border. A similar area was traced again later in 1957 by Donald Thomson. The discovery of gold in the relative vicinity of Ngardi country, first to the north at Halls Creek in the 1890s, and later to the east at Yartuluyartulu (The Granites) and Jarnami (Tanami) in 1900, brought neighbouring Jaru and Warlpiri people in direct contact with European settlers (see Kearney, 1985; Olney, 1992). This was followed by a burgeoning interest of pastoralists in the viability of the south-eastern Kimberley as pastoral country. In 1906, Alfred Canning travelled through Walmajarri and Wangkajunga country (to the west of Ngardi country), in a bid to chart a route for droving cattle out of the Kimberley and on to southern markets. The eventual (but ultimately short-lived) Canning Stock Route profoundly affected Ngardi’s western neighbours (Carty, Davenport & Lafontaine, 2010). In addition to direct and hostile contact by Canning and his men, the spoiling of important water sites eventually led to the northward movements of people out of their desert homelands and onto various mission settlements and cattle stations to the north. By the 1920s cattle stations had been established right across the eastern Kimberley as far south as Sturt Creek, both at ‘Carranya’ and ‘(Old) Billiluna’ (see Map 4). Many Ngardi people had likely heard word of Europeans and began, for various reasons, to move on to cattle stations both in the NT and throughout the south-east Kimberley.17

1.3.3 Movement of Ngardi people onto missions and cattle stations A clear reconstruction of the historical events that led to Ngardi and other Aboriginal groups of the western Tanami and northern Great Sandy Deserts to occupy various mission sites, cattle stations and townships is a complex issue. Many social and environmental factors have been speculated upon as lying behind the historical movement of Western Desert groups out of the desert in the early to mid twentieth century and onto cattle stations and missions throughout the south-eastern Kimberley. These 16 Expeditions by Giles (1875–1876) and Warbuton (1875) predate Carnegie’s expedition but both passed to the south of Ngardi country. 17 Many Ngardi and Kukatja people also recall seeing stock animals for the first time before coming into contact with Europeans themselves.

22 

 1 Introduction

have variously included mentions of forced ‘round-ups’ and coercion as well as volitional, agentive movements by Western Desert peoples, potentially on the promise of more predictable and easier access to food and water sources (see Long, 1989; and Carty, 2011 for a summary of the relevant literature). Factors beyond direct contact with Europeans appear to have had an important role in the history of this region. Thomson (1975) and Long (1989) reference the effects of increasing water scarcity – in particular a decade-long regional drought which affected the country around Lappi Lappi (1895 to 1906). Such environmental factors encouraged the Kukatja to migrate east from an original occupation of lands westward from the area north of Lake Mackay to the Stansmore Ranges (Thomson, 1975, p. 6) – an area associated with Ngardi people. Carty goes as far as to conclude that ‘the social geography of the Kukatja was therefore being transformed by ecological and social pressures 50 years before the Pallottine missionaries arrived in the desert’ (Carty, 2011, p. 27). Whatever the complex social and historical motivations that instigated such dispersion, Ngardi men and women ended up working across a large number of cattle stations in the Kimberley and the NT (many owned by Lord Vestey), far to the north of their traditional country. Patrick Smith reports that Ngardi people lived in some numbers at the original sites of Billiluna station (Kilangkarra and Nyarna),18 Ruby Plains (Yuka), Flora Valley (Palngarna), Nicholson (Narrala), Sturt Creek (Purrkuji), Ringer’s Soak, Gordon Downs (Jukuj Krik), Spring Creek, and Birrindudu Stations (see also R. M. Berndt & Berndt, 1987). The location of these cattle stations with respect to the modern-day community of Balgo is presented in Map 4. The living conditions that greeted Aboriginal people on the stations were appalling. Firsthand accounts by the Berndts (1987) and a review of the literature by Smith (2000) provide a glimpse of the horrific conditions faced by Aboriginal men, women and children on cattle stations in the south-east Kimberley. Men and women worked for little more than meagre rations of sugar, flour, tobacco (ngunju) and tea (iki, ngurlukaji). Conditions on Birrindudu and Gordon Downs stations were singled out by the Berndts in particular as being particularly brutal: no access to nutritious food, labour exploitation, abuse of women, as well as murder and reprisal killings (for the hunting of livestock) characterised the new lives of Aboriginal people in the south-eastern Kimberly in the early to mid twentieth century. In addition to the oppressive conditions on the cattle stations, Walmajarri, Ngardi and Jaru people alive today recall the massacre of their grandparents which took place at Kaningarra and then at Purrkuji (near Sturt Creek Station) in the aftermath of the 18 There are multiple historical sites of the Billiluna homestead all within the Billiluna Pastoral Lease. The lease was first taken up by Joseph Condren on 4 May 1920 (Gard & Gard, 1990, p. 135). A windmill was erected at a site now referred to as Comet, not far from present-day Mulan. The original site of the first homestead was subsequently established at a place called Kilangkarra (on ‘Butcher Creek’, north of Mulan). The homestead was later moved to Nyarna (Lake Stretch, also known as Warnku) and then on to present-day Kururrungku (also known as Mindibungu).

1.3 Ngardi in a socio-historical context 

 23

Map 4: Map of cattle stations in the south-eastern Kimberley.

murders of Joseph Condren and Tim O’Sullivan in 1922.19 The oral history, passed down from the sole adult survivor, Riwarri20 and supported by recent archaeological investigations (Smith, 2016; Smith et al., 2017), recalls how a number of people were shot and other men, women and children were rounded up near Kaningarra, on the Canning Stock Route. The remaining survivors were walked in chains northeast as far as Old Denison Downs (near Sturt Creek Station) homestead. Here, ‘they were chained between trees in 19 The suspect, named Banjo, had been tracked west and north-west of Old Billiluna Station and was shot on 8 November 1922. Associated police reports make no reference to the massacres also enacted in the wake of the Condren/O’Sullivan murders, first at Warlikarrapungu (Godfrey Tank) and then later at Sturt Creek Station. See Smith (2016) for a full discussion of the evidence. 20 Note that Riwarri is the name as recorded in various sources (e.g. Smith, 2016), but at least Patrick Smith pronounced his name as Riwiyarri [ɻɪ́wɪjɐ̀ɾɪ].

24 

 1 Introduction

the goat yard and shot. The bodies were then burnt . . . Most of those killed were incinerated in the well, but not everyone. The bodies of women and children were incinerated at a separate area and others were burnt inside the goat yard’ (Smith, 2016, pp. 27–28). While maltreatment, exploitation and the murder of Indigenous people formed the dominant theme of settler–Indigenous relations throughout much of the southeast Kimberley, not all settler-Indigenous relations from that period are assessed negatively by Indigenous people of the Kutjungka region. Patrick Smith, for example, readily recalls the good treatment Jaru, Walmajarri and Ngardi received from Sam Hazlett and Jimmy Malabi at Palm Springs. Furthermore, survivors of the Sturt Creek massacre reported that they were given refuge by Dave Bickley at Mt Brophy, by Tom Bradshaw at Slatey Creek, and by Jack Skeen and Dave Piggley at Lewis Creek (Smith, 2000, p. 70, 2016, p. 31). Movement in and out of missions and cattle stations by men and women of older generations appears to have decreased rapidly after the first few decades of settler contact. The earliest accounts of contact made reference to a greater degree of freedom than came to be enjoyed on life within the mission or at the pastoral camps. Such accounts are found in the oral histories of Ngardi women recorded in the mid 1990s and these accord closely with the oral histories of Wangkajunga women (Andrews, 1996; Bolger, 1987). My mother take me all around. We bin go up and down, up and down. I bin getting big. I go back and grow at Balgo. Come there when I was a girl. We bin go to Mangkayi now, still girl. When I come back from Mangkayi had ngapurlu – breast like little tomatoes. Back to Balgo. Live at Balgo and Lake Stretch and Sturt Creek. I bin walking self now. (Napanangka et al., 1997, p. 103).

1.3.4 Establishment of the Balgo Mission and the mission period (1939–1983) The arrival of the Pallottine Missionaries and the establishment of the Balgo Mission in the late 1930s was by far the single most disruptive event for many Ngardi and Kukatja people. It brought the influence of Western ways of life further south than the southernmost cattle stations at the time – the southernmost pastoral station had been established on Sturt Creek at Dennison Downs Station by Stretch and Weekes (Carnegie, 1898, p. 357). The vision of the Church in venturing south from Beagle Bay and an initial site at Rockhole (south-east of Halls Creek) was to provide ‘a buffer’ between the traditional Aboriginal way of life and European life which was encroaching on their world – and, undoubtedly, to evangelise and bring Aboriginal people into the Church. However, the notion that the Church could somehow provide ‘a haven’ in which Ngardi, Jaru, Kukatja, Walmajarri, Wangkajunga and Pintupi people would be able to live their lives independent of the effects of European settlement would be a gross misrepresentation of the impact the mission had on people’s lives. Carty (2011, p. 47) goes as far as to conclude, in reference to the Balgo Mission, ‘while protecting

1.3 Ngardi in a socio-historical context 

 25

people from exploitation on the colonial frontier, the mission came to embody that very colonial frontier through practises which actively sought to change Aboriginal ways of life and foster dependence on an alien economy’. The formation of the first Balgo Mission is traced in detail by Ganter (2016). Following the initial recommendations of Father Ernest Worms, a German missionary team led by F. Aphonsus Bleischwitz left a site known as Rockhole (about 20 km southsouth-west of Halls Creek) in 1939 and ventured south with an intention to establish a mission site somewhere beyond the terminus of Sturt Creek and Lake Gregory on the fringes of the Great Sandy Desert. Initially, temporary mission camps were made at Comet (near modern-day Mulan) before another camp was made at Jalyuwarn, south-east of the southern end of Lake Gregory.21 This site, comprising a creek with a billabong and rock pools, is reported to have been occupied by a ‘60-strong Aboriginal camp’ upon the arrival of the missionary team – but no other details are noted (see Ganter, 2016). During the period 1939–1942, the mission’s livestock (mainly sheep and some goats) were moved between Comet and another site, Jumunturr (also spelt Doomendora), located south of Old Mission. The water here, however, was not of sufficient quality to support the team and so the operations of the mission were necessarily spread across three sites (Comet, Jalyiwarn and Jumunturr) until 1942 when the mission shifted to Palku (Balgo) east of Malarn (Mulan) where a permanent source of water had been established via the successful boring of a well.22 The former mission sites at Palku and Jalyiwarn are now referred to as ‘Old Mission’ or ‘Old Balgo’ by Balgo (Wirrimanu) residents today. By the time of the establishment of the Balgo Mission proper in 1942, significant disruption of Aboriginal peoples of the northern Western Desert region had already occurred, and across a vast area. According to Olney (1992, p. 11), by 1945 nearly all traditional owners associated with the Warlpiri/Kukatja land claim had left their traditional lands along the NT/WA border. While populations at Old Balgo Mission were somewhat transient in the mission’s early years, Ronald and Catherine Berndt’s anthropological work identified a wide array of linguistic groups coming to live there on a semi-permanent basis, including Kukatja, Manyjilyjarra, Walmajarri, Wangkajunga and Ngardi people (Berndt & Berndt, 1960, p. 2).23 Ngardi people continued to walk out of the desert and onto Balgo Mission well into the 1960s (see Long 1989, p. 29). By 1960, Berndt and Berndt (1960) reported that there were around 150 adults and children living on a semi-permanent basis in the vicinity of Old Balgo. Due to issues of water quality and water scarcity, as well as concerns regarding the operation of the mission and its independence from Billiluna Station (on whose lease21 Named ‘Bishopdale’ after Bishop Raible. 22 The origin of the place name ‘Balgo’ is not entirely clear. Ryan (1989) asserts it that is derived from the Kukatja term palku meaning ‘foul wind’. 23 Berndt and Berndt (1960) also make reference to another group, Ngabi, for whom I am unable to provide any further information.

26 

 1 Introduction

hold land it was located), the Balgo Mission was moved in 1965 under the supervision of Fr. John McGuire (Berndt & Berndt, 1960) to the site of modern-day Balgo.24 The new mission, completed in 1965, through the labour of local Aboriginal men, developed quickly. Balgo (or ‘New Mission’ as it was initially called by Aboriginal people) had an administration centre, clinic, school, monastery, teacher’s house, church, convent, laundry, boys’ and girls’ dormitories, a bakery, dining hall, slaughter house, store, workshop and a head stockman’s house. The mission, while increasingly more connected to the rest of the country (by both more reliable radio and later an airstrip), was largely self-supporting. It boasted a cattle industry, flocks of sheep and goats, pigs, chickens and gardens. By the late 1960s, the Aboriginal population had reached 300 and, by 1969, 400. Following the establishment of the DAA (Department of Aboriginal Affairs) in 1972 under the Whitlam government and increasing pressure under federal policy of ‘self-determination’, self-management became a condition of federal assistance to Aboriginal communities in the early 1980s. Less than 15 years after moving from Old Balgo, the Church was pressured into ceding its land administration to the Aboriginal people in 1983. The new governing body, Wirrimanu Aboriginal Corporation, was officially incorporated on 6 September 1984. While life in Balgo Mission was inescapably harsh and stories of brutality abound, it is also reflected upon with a certain fondness by many older Balgo residents. During the mission period, Aboriginal people remained fundamentally involved in the productive capacity of the community: they participated in the building of community infrastructure; the management of the mission’s livestock (cattle, goats, sheep and chooks), and the growing, harvesting and preparation of food in gardens and the kitchen. Since the closure of the mission and the shifting management of the Balgo community, Aboriginal people have faced a loss of involvement in day-to-day activities which govern their lives. The accompanying onset of welfare dependence also removed Aboriginal people’s active involvement in the day-to-day operations of the community.

1.3.5 Post-Mission life in Balgo Coinciding with the church’s loss of influence in the Balgo community, there was a nationwide movement to support Aboriginal people to move back and live on their traditional lands in the late 1970s and into the 1980s. This came to be known as the outstation movement (Peterson & Myers, 2016) and was typified by a widespread push (both on a local level by traditional owners and at a policy level by government) for the establishment of smaller settlements or outstations back on country.25 For Aborig24 Modern-day Balgo lies on Jaru country and is associated with the Luurn ‘kingfisher’ dreaming (Luurnpa in Kukatja). 25 ‘On country’ is a term that refers to Aboriginal people being on or living on their own traditional land.

1.3 Ngardi in a socio-historical context 

 27

inal people living in Balgo, this resulted in the establishment of Mulan (1979) to the west and Yagga Yagga (1985) to the south. Smaller outstations south of Yagga Yagga were also established at Piparr, Walkali and Lamanparnta around this time. Yagga Yagga in particular transformed into a particularly vibrant outstation – essentially becoming a community in its own right – with a permanent population of 150 residents by 1997 with numbers swelling to as many as 500 as during one period of initiation ceremonies in the early 1990s (Cane, 2016, p. 261). Many Ngardi people lived at Yagga Yagga during this time as it was situated within their homelands (unlike Balgo) and was close to many important cultural sites. Ultimately, however, the occupation of Yagga Yagga was relatively short-lived and its closure in 2005 is described by Cane (2016) as resulting from what he summarises as ‘an administration of suffering’ (cf. Peter Sutton’s The politics of suffering, 2011), involving not only systematic mismanagement, but a fundamental lack of interest on the part of governing and administrative bodies to genuinely listen to and respond to the concerns and challenges faced by Yagga Yagga (and Balgo) residents. In the years since the closure of Yagga Yagga, many Ngardi people have moved back into Balgo or have moved elsewhere throughout the south-east Kimberley or the Northern Territory. Balgo itself has transformed quite dramatically from its days as a Mission settlement where many Aboriginal people lived in makeshift lean-to’s made from discarded corrugated iron. Balgo residents are arranged into three main camps: ‘top camp’ on the eastern edge of town; ‘bottom camp’ on the western edge of town, near the Mulan road; and (the most recent) kayili ‘north’ camp on the northern edge of town. The modern community retains the Catholic school (Luurnpa Catholic School), the church and the parish house. Gone are the gardens, boys’ and girls’ dormitories, kitchens and nuns’ house. In their place are a conglomerate of buildings and infrastructure set up for external service providers: a clinic, a police station, a council office and post office. Over the years, Balgo has seen the success and the demise of a highly productive language program (in Kukatja) at the Walkali Centre within the Luurnpa Catholic School. Nevertheless, a number of key individuals in the community remain committed to language teaching in the school both within the school faculty and the community at large. On the eastern edge of the community, near the women’s jilimi (Law ground), Kapululangu Aboriginal Women’s Law and Culture Centre established a women’s centre which has served as a multifunctional organisation for nearly two decades. From the 1980s, the development of the Balgo art scene and the incorporation of Warlayirti Artists has played an integral role in the development of the community (Carty, 2011; Dé Ishtar, 2005; Poirier, 2005) and it continues to provide one of the very few income streams outside of Government welfare programs. The Balgo community received a newly renovated store in 2017 and a multi-million-dollar pool opened as recently as 2019. All residents buy all of their food, fuel, power cards (for domestic electricity), appliances and personal possessions from this one store. Small stores are also found in neighbouring Mulan (40 km) and Billiluna (110 km),

28 

 1 Introduction

but the nearest regional hub is that of Halls Creek (280 km). Balgo families regularly pool their resources to attend personal, social and business trips in the wider region. A number of highly successful artists have even travelled overseas as part of their internationally-recognised careers as artists.

1.3.6 The role of multilingualism: Past and present To my knowledge there are no monolingual Ngardi speakers today. All remaining Ngardi speakers are fluent in at least three languages. The multilingual status of Ngardi speakers appears to extend into pre-European contact times – oral histories give testament to high degrees of multilingualism being the normative mode of communicative practice. This would have been further fostered due to traditional intermarriage between discrete linguistic groups across the Western Desert (Holcombe, 2004). Many Ngardi speakers recorded over the last 60 years spoke and continue to speak Warlpiri and/or Jaru (Cataldi, 2011, p. 1).26 Numerous Ngardi people also have long-term connections to Walmajarri people who lived in the vicinity of Paruku (Lake Gregory) and are variously fluent in the variety of Walmajarri still spoken today in Mulan. By the same token, however, European contact and the fundamental social disruptions to Ngardi people also had a massive impact on how, when and where Ngardi people came to learn additional Aboriginal languages. Ngardi-speaker Tjama Napanangka thus reflects on coming to learn the Kukatja language only after she moved into Balgo Mission: (7) Nganimpa=rnalu kayirra-ngulu-jankga, wangka-nya-ngurra 1pl.excl=1pl.excl.s north-abl-ela speak-pst-narr Ngardi-mipa=lku. language_name-restr=then ‘Those of us from the north, we only spoke Ngardi.’ Wakurra=rnalu purda=nya-ngani nganayi-ku. Kula=rnalu=yanu neg=1pl.excl.s listen=see-ipfv.pst whatsit-dat neg=1pl.excl.s=3pl.o andasdan=ma-nani-nyirra=rnalu=yanu Kukatja understand=get-ipfv.pst-narr=1pl.excl.s=3pl.o language_name ‘We did not understand whatchacallem, we did not understand them, the Kukatja.’ (TJN: Manungka_02-022027: 2810322_2815198) Similarly, Bolger (1987) records an oral history of Warlpiri woman Violet Jimpirriya, who grew up on traditional lands away from direct contact with early missions or 26 Tsunoda (1981b, p. 17) reports that Ngardi people working as stockmen at Gordon Downs Station in the 1970s and 1980s also spoke Jaru.

1.3 Ngardi in a socio-historical context 

 29

cattle stations but eventually settled at Kurungal (Christmas Creek Station). In this process, she came to learn at least three new languages: Ngardi, Kukatja and Wangkajunga during her movements west across the NT/WA border and north on to the missions and stations of the south-east Kimberley. Prior to contact with kardiya she spent some time travelling in Warlpiri and Ngardi country around Mangkayi (Stansmore Range). Eventually, Jimpirriya arrived at Old (Balgo) Mission and, later, Billiluna Station. Here Jimpirriya makes the following comment: So we went to the camp and this Billiluna manager asked them to bring us up to the station next morning to give us clothes. So we was living there now in Billiluna. And I started to learn their language, Ngarti. Because when I came from Warlpiri area I was speaking Warlpiri and Ngaripi27 – two language. But when I came down here to Western Australia I spoke Ngarti, Wangkajunga, Kukatja, that’s all I spoke. (Bolger, 1987, p. 110).

Once Ngardi people had come to live at the mission and their patterns of movement across their country slowly decreased, it is clear that the intergenerational transfer of the Ngardi language was dramatically arrested. This was first and foremost due to the forced separation of families – and in particular the segregation of younger generations from older generations. Berndt and Berndt (1960, p. 2) reports on the situation in both the mission and the stations at the time: ‘for the youth on the station and in the dormitories, it seems possible that only about 5 to 10% of their time is spent with their parents and relatives in the native camp’. Marie Mudgedell reports that most of the children who grew up in the dormitories in Old Mission were not allowed to see their parents. Cataldi (2011, p. 1) summarises similar reports by the Balgo women with whom she worked. Of great consequence, then, was the fact that the dominant language of the children who were first placed into the dormitory system was a variety of a Western Desert language, which came to be referred to as Kukatja. It was in the dormitory context where many young children shifted from strictly speaking the language of their parents and instead started speaking the language of the first children taken into the dormitories – a communalect which is the basis of modern Kukatja. Marie Mudgedell and Mark Moora both affirmed that they learned Ngardi by spending time with old people (two generations above them), both on country and when ‘running away’ to live in camps near the stations rather than in the dormitories at Balgo. Patrick Smith likewise learnt Ngardi by spending much of his formative years outside of the mission dormitories with older Ngardi men with whom he worked as a stockman. Within the mission context, however, Ngardi quickly ceased to be spoken as a main language – instead being restricted to small groups for whom it was their first language. Kukatja instead quickly came to predominate as the lingua franca among the youngest generations who grew up in the mission dormitories from the late 1950s and early 1960s. The handful of Ngardi speakers who remain – and who

27 This may represent a spelling variant – or indeed a genuine phonetic variant – of Ngardilypa, a Western variety of Warlpiri detailed in Jagst (1975).

30 

 1 Introduction

contributed intimately to the publication of this grammar – are mostly elderly and spread across the south-east Kimberley and the Northern Territory. The language has not been a primary mode of communication for any significant group of individuals for over a decade and it is not being acquired by children. It is not the case that the community of Balgo became entirely monolingual within a generation, however. Valiquette (1993, p. 27) cites numerous languages spoken in the community during and preceding his work at the Luurnpa School in the early 1990s, including Ngardi, Walmajarri, Jaru, Warlpiri, Pitjantjatjara, Wangkajunga and Pintupi as well as varieties of English and Kriol. Nearly all of these languages continue to play a role in day-to-day communication and the expression of individuals’ identity and place in society today.

1.4 The current grammar 1.4.1 General theoretical approach The approach taken in the writing of this grammar can be said to broadly fall within the basic linguistic theory framework (Dixon, 1997). This is a descriptive-oriented framework that has emerged through grammatical descriptions (predominantly of Australian Aboriginal languages) with the loosely stated aim of ‘describing each language on its own terms’. Rather than being employed as a prescriptive theoretical framework, it has generally been taken up as a general approach to language description that builds on the cumulative analytical expertise of existing grammatical descriptions (Gaby, 2017, p. 22). To that end, the analysis of Ngardi grammar presented here has not been pursued purely in isolation, as if it were being approached as an entirely unique language in an undocumented language family. Rather, typological considerations from related and neighbouring languages frequently played a role in directing research questions, as well as informing and supporting the final analysis. This is an inherent entailment of analysing the grammar of a language since it requires a typological method (see Cristofaro, 2006). An additional aspect of this particular approach to language documentation has been the avoidance of positing formalisms or employing theoretical terminology that would reasonably date the grammar and render it opaque to future readers. One simple way in which this has been pursued has been the use of widely employed glosses and terminology whever appropriate. This approach better allows future linguists to revise existing analyses and make meaningful typological connections. Lastly, I have endeavoured to furnish the grammar with a generous selection of example sentences, as far as was reasonably possible. I believe this approach both better enables future linguists to assess Ngardi grammar in a manner somewhat independent from my own analytical conclusions, and moreover gives the reader a better sense of the ways in which the Ngardi language is used. However, the reader is cautioned that the corpus of examples in this grammar is by no means a representative

1.4 The current grammar 

 31

sample of naturalistic speech, nor are the examples a meaningful depiction of the culture, interests or priorities of the speakers that uttered them. The necessities of providing a grammatical description of the whole linguistic system have displaced any attempt to present a truly balanced socio-cultural representation of the language and how it is used across the broadest range of communicative contexts. 1.4.2 Data collection and corpora The language data which are drawn upon to exemplify the analysis in this grammar primarily originate from corpora compiled by four authors: Tasaku Tsunoda, Tom Green, Lee Cataldi, and me. Where necessary, I refer to these respective corpora as ‘Tsunoda’s corpus’, ‘Green’s corpus’, ‘Cataldi’s corpus’ and ‘my corpus’. Table 1 gives some indication of the provenance, content and size of the combined corpus of the Ngardi language. All recordings I made have been deposited with PARADISEC (http://. paradisec.org.au/) (these materials are being processed at time of print). The corpora of Tsunoda, Green and Cataldi are archived with the Australian Institute of Aboriginal and Torres Strait Islander Studies (AIATSIS) (https://collection.aiatsis.gov.au). My own audio-recordings were made on a Zoom H6 Portable Field Recorder. A variety of microphones were used including an XYH-6 XY microphone module, an MSH-6 Midside microphone capsule, and a Rødelink Filmaker kit for wireless recording via a Rødelink LAV lapel mic. In some cases, for example during transcription checking, speakers preferred not to be recorded and, as a result, notes were simply taken in notebooks.2829 Table 1: Overview of audio corpus of Ngardi language materials (as of 2019). Authors

Type(s)

Ennever, Thomas Cataldi, Lee; Peter, Sonja

Narratives, grammatical elicitation, 128hrs28 dictionary checking Narratives, procedural texts, dictionary 40hr checking, conversational data, ceremonial texts, song sequences Narratives 1hr Explanation of artwork, song sequence, 1hr biographical interviews, dictionary checking Word list elicitation, narrative text 5hrs

Green, Tom KLRC29 (various authors) Tasaku, Tsunoda

Duration (approx.) Dates collected 2016–2019 1990–1991, 2000–2001 1988 1984, 1989, 1998 1976

28 The total number of hours given here includes quite substantial periods of silence, conversations in English, and off-topic activity captured during a recording session. 29 Kimberley Language and Resource Centre.

32 

 1 Introduction

Table 2: Overview of written materials concerning the Ngardi language (as of 2019). Author

Type

Description

Ennever, Thomas

Masters thesis

Cataldi, Lee and Napanangka, Tjama Honeyman, Tom

Draft dictionary grammatical sketch

Dissertation submitted to the University of Queensland, entitled ‘Pronominal and nominal morphology of Ngardi: A Ngumpin-Yapa language of Western Austalia’ Substantial draft dictionary with examples extracted from written transcripts; grammatical sketch

Wafer, James ‘Jim’ Green, Tom and Rockman, Peggy Tsunoda, Tasaku

Wordlist

Honours thesis

Grammatical notes & wordlist Fieldnotes

Dissertation submitted to the Australian National University entitled ‘Topic and Focus in Ngardi’, based on Cataldi’s data A Summer Institute of Linguistics (SIL) wordlist Transcriptions of two Ngardi narratives; preliminary grammatical notes and a wordlist Grammatical notes, paradigms, wordlists

Dates collected 2018

1990–2011

2000

1980 1988

1975–1976

The variety of the Ngardi language described in this grammar is a result of providence and availability rather than design. The idiolects represented in Cataldi’s, Tsunoda’s and my own corpora all converge in likeness to a degree to which they can be confidently considered a single dialect of a single language. That this was appropriate from the perspective of speakers themselves was confirmed by all those with whom I worked throughout 2015 to 2019. The speech of Peggy Rockman (as in Green’s recordings) was noted by Ngardi speakers as involving notable mixing of Warlpiri words but these aspects were readily pointed out where appropriate by speakers. The same is true of other mixing practices of Ngardi with English, Kukatja, Walmajarri or Jaru vocabulary in the speech of various individuals. Since Ngardi is no longer used as a language of daily communication, this work is, in some respects, a salvage grammar in so far as I have attempted to capture the language as it was spoken by the last fluent speakers. My own fieldwork relied heavily on targeted elicitation in a bid to fill remaining gaps in grammatical analysis that were not elucidated from more naturalistic conversational data appearing in earlier recordings of the language. Nevertheless, the small number of speakers with whom I worked were still highly fluent speakers with a strong command of the language; some of them could freely supply extensive narrative texts and respond readily to semi-structured elicitation prompts.

1.4 The current grammar 

 33

1.4.3 Recording types and methodologies In this section I extrapolate briefly on the types of recordings created by Cataldi (§1.4.3.1) and myself (§1.4.3.2) – the two largest collections of recordings and those which from which the majority of examples for this grammar are drawn. 1.4.3.1 Lee Cataldi (1990s) Cataldi’s corpus includes a number of different recording genres which can broadly be categorised into three types. First are recordings of unprompted conversation between predominantly Ngardi and/or Kukatja women. For these recordings, Cataldi was absent. A subset of these are conversations recorded while women were playing cards. For reasons of privacy, some of these have been anonymised where appropriate and the associated audio has restricted access within the AIATSIS archive. Second are narrative recordings by a single individual recounting a story (sometimes with input from an additional listener). Cataldi was present during these recordings. Third are a number of elicitation-style recordings in which Cataldi prompts various Ngardi words from Warlpiri as a source language and receives responses from one or two women at a time. 1.4.3.2 Thomas Ennever (2016–2020) My own corpus involves a number of text types. A large proportion of the corpus involves targeted grammatical and lexical elicitation, generally using English as a source language or, on occasion, Jaru. In addition to direct translation-style elicitation, various non-verbal prompts were used to elicit various aspects of the language. These included various picture-prompted tasks, impromptu directional tasks, and picture-stimulus narration. Supplementing the elicitation-type texts are a number of autobiographical or biographical texts. In addition to the audio corpus, my own corpus includes many notes authored by myself as well as by Ngardi speakers (predominantly Marie Mudgedell), which provide new and revised translations of earlier recordings of the language.

1.4.4 Previous research The following sections provide a comprehensive summary of previous research into the Ngardi language. 1.4.4.1 Survey studies The earliest European records of the Ngardi language are in survey studies completed by Michael Terry, Arthur Capell and latter Norman Tindale. These are briefly described in turn.

34 

 1 Introduction

Michael Terry (1926) Michael Terry was an Australian explorer, surveyor, prospector and writer. Between 1923 and 1935 he led numerous expeditions throughout inland Australia and later published a number of books describing his experiences (Terry, 1925, 1927, 1931, 1987, among others). Terry (1926) provides a basic list of groupings of Aboriginal peoples from the areas around Gordon Downs Station and Billiluna Station at the terminus of Sturt Creek, which he documented as ‘Boonarra’ and ‘Tchagilin’, as highlighted in Table 3. Table 3: Terry’s (1926) lists of Aboriginal groups around cattle stations. Group Nanaroola Mootburra Wadderman Loonga Coorinji Bilinurra Mootburra Manoo Boonarra Boonarra Boonarra Tchagilin (language)30

Modern Orthography Mudburra Wardaman Gurindji Bilinarra Mudburra Malngin? Punurra Punurra Punurra

Location Newcastle Waters Montijinnie Dalamere Wyckham Wave Hill station near Mt Sanford East of Wave Hill Inverway (station) Flora Valley Halls Creek Gregory’s Sea Gordon Downs

Terry’s locations for the Bunarra overlap significantly with areas associated with Nyininy, Jaru and Ngardi peoples (Tsunoda, 1981b). Further evidence of this correspondence comes from the linguistic material Terry provides in association with the language groups highlighted above. On the basis of the lexical comparanda Terry collected, it appears that ‘Boonarra’ and ‘Tchagilin’ groups correspond quite closely with later data collected for the Nyininy dialects associated with the lands around Sturt Creek, Old Flora Valley and Gordon Downs (Tsunoda, 2006) and Ngardi (Cataldi, 2011; Tsunoda, 2006).30 Arthur Capell (1940) In his classificatory study, Capell (1940a, 1940b) provides a basic wordlist of Ngardi that shows close commensurability with the variety of Ngardi recorded about 35 years later by Tsunoda and some 60 years later by Lee Cataldi. Comparanda are presented in Table 4.

30 The term ‘Tchagilin’ may come from an Eastern Walmajarri or Ngardi word for moon, jakiliny.

1.4 The current grammar 

 35

Table 4: Comparanda of Ngardi wordlists collected by Capell, Tsunoda and Cataldi. Capell (1940b) man woman head eye nose mouth tongue stomach bone blood kangaroo opossum emu cry fly sun moon fire smoke water

ŋarga gandar waːlu milba mulju lira djälan̠ djaːla gidji djugän djädji djaŋana ___ djäŋilga ŋurin bṛaŋu jagan waḷu gundjuru ŋaba

Tsunoda (1975–1978) garnrda(rr) walu milba mulyu lirra jalany jala giji ___ jaji ___ ___ lungan ___ burangu jagiliny, yagarn warlu nguriny33 ngaba

Cataldi (2011) ngantany, ngarrka ngaringka wirrkirl, waluwarnu31 milpa mulyu lirra jalany jarla kiji jukan jaji jangany32

nguriny purangu yakarn warlu kunjuru ngapa

Norman Tindale (1952–1954) In his expedition through north-western Australia, Norman Tindale (1954) collected a 180-item Swadesh wordlist of Ngardi along with materials from 30 other tribes of the region. He also worked with Ngardi people and collected some anthropological notes while on his expedition to Haasts Bluff (Tindale, 1956).313233 1.4.4.2 Preliminary descriptive work Kenneth ‘Ken’ Hale (1967) Ken Hale made a short recording (of approximately 12 minutes) with Murray Japangardi in 1967 at Yuendumu. This audio file is archived at AIATSIS.

31 ‘Woman’s headdress’. 32 Desert brush-tail possum, Trichosurus vulpecula vulpecula. 33 The variant gunyjuru was collected at Sturt Creek and is also used by Patrick Smith.

36 

 1 Introduction

Tasaku Tsunoda (1975–1976) Tasaku Tsunoda made a number of recordings of Ngardi speakers in the course of his fieldwork with Jaru speakers. Much of his work was undertaken on various cattle stations in the southern Kimberley, including Gordon Downs, Nicholson and Sturt Creek stations. His combined corpus includes material on Wanyjirra (see Senge, 2015), Malngin (see Ise, 1999), Nyininy, Ngardi and Warlpiri. His fieldnotes from 1975–1976 include various grammatical notes for all these languages as well as comparative pronominal, nominal and verbal paradigms. Tsunoda’s audiofiles and fieldnotes have been archived with AIATSIS. Jim Wafer (1980) Jim Wafer interviewed a number of Ngardi speakers at Balgo in September of 1980. During this time, he elicited the ‘Wordlist for Australian Aboriginal Languages’ devised by SIL and collected some basic ethnographic information. Wafer worked with Charlie Jakamarra Gordon, Tomato Japangardi Gordon and Sandy Jupurrula (Wafer, 1980). Wafer’s materials are archived with AIATSIS. Matthew Wrigley (c. 1990) Matthew Wrigley recorded a wordlist with Popeye Jangala, Tchooga Napurrurla and Josephine Tchooga entitled ‘Ngarti/Bunarra’ while working with the Kimberley Language and Resource Centre (Halls Creek). This wordlist has been archived with AIATSIS. 1.4.4.3 Descriptive work Thomas Green (1988) Thomas Green collected some basic linguistic materials on Ngardi while working with Peggy Rockman Napaljarri at Lajamanu in August of 1988. Green compiled an impressive set of preliminary grammatical notes and a wordlist (Green, 1988), all of which are archived with AIATSIS. Lee Cataldi (1990–2011) Lee Cataldi continued to work with Peggy Rockman after Tom Green and also worked with a number of other Ngardi women at Wirrimanu (see list of contributors, §1.4.5). Cataldi conducted much of her elicitation work in the Warlpiri language, and was also involved in significant documentation of Warlpiri oral histories (Napaljarri & Cataldi, 1994). Much of the Ngardi language work was done in collaboration with Michelle McKenzie and Sonja Peter as part of the Manungka Manungka Association (the predecessor of the present-day Kapululangu Women’s Centre). Tjama Napanangka in particular is credited by Lee with the development of the Ngardi dictionary manuscript. Cataldi’s sketch grammar in the foreword of the dictionary

1.4 The current grammar 

 37

provides essential information on the nominal, pronominal and verbal systems of Ngardi. A preliminary analysis of the case system is present and there is discussion of similarities between Ngardi and Warlpiri and/or the Ngumpin languages. Significantly, Cataldi’s corpus includes a number of narrative texts, conversational data, and lexical elicitation for the dictionary. The current version of the dictionary was compiled in 2011 and contains a grammatical foreword (Cataldi, 2011). Cataldi’s corpus is archived at AIATSIS and includes both audio files and a selection of text files. An electronic copy of Cataldi (2011) A dictionary of Ngardi is available online via the University of Sydney e-Scholarship repository (https://ses.library.usyd.edu. au/handle/2123/21407) as of 2021. Tom Honeyman (2005) Tom Honeyman completed an Honours thesis investigating the placement of the pronominal clitic complex in the Ngardi clause according to information structure principles (Honeyman, 2005). Honeyman’s data comes exclusively from the Cataldi corpus.34

1.4.5 Ngardi language experts and contributors In this section I describe the contribution of Ngardi speakers35 either in groups or as individuals, organised by their method of involvement with the knowledge and analyses presented in this grammar. Tjama Freda Napanangka [TJN] †, Kuninyi (Rita) Nampijijin †, Nanyuma Napurrurla [NMN] †, Patricia Lee Napangardi [PLN] †, Yinjuru Margaret Anjule (Bumblee) Napurrurla [YMN] †, Mungkirna (Dora) Napaljarri [MDN]†, Damper Nampijin [YDN] †, Maatingali (Maati) Bridget Mudgedell Napanangka [MMJ] †, Martingali (Maudie) Nungurrayi [MMN] †, Payi Payi Napangardi [PPN] All of these women worked closely with Lee Cataldi in the recording and documentation of Ngardi language. Many of them were also central to the development of the book Yarrtji: Six Women’s Stories from the Great Sandy Desert in collaboration with Pamela Lofts and Sonja Peter (Napanangka et al., 1997). Tjama, in particular, is credited with much of the work that went into the Ngardi dictionary manuscript, providing translations, transcriptions and elucidating points of difference between Ngardi and surrounding languages. The considerable work of these women culminated in the first edition of the Ngardi dictionary (Cataldi, 2011). Of the ten women, I had the pleasure 34 Honeyman was also involved with the digitisation of the dictionary and the interlinearisation of some transcripts. 35 The symbol † is used to indicate those who, at the time of publication, have passed away.

38 

 1 Introduction

of meeting only Payi Payi Napangardi. She listened to and commented on a number of old Ngardi recordings which I played back to her for translation checking on my trips to Balgo between 2017 and 2019. Peggy ‘Yalurrngali’ Rockman Napaljarri [PRN] Peggy Rockman was first involved in the documentation of Ngardi with Tom Green in 1988. From 1992, she worked closely with Lee Cataldi in the recording, transcription and translation of Ngardi texts made by Ngardi speakers between (1989–1992) with the assistance of Michelle Mckenzie. Rockman was also heavily involved in related Warlpiri language work as part of the Yimikirli Project, resulting in the publication of Yimikirli: Warlpiri Dreamings and histories (Napaljarri & Cataldi, 1994). Many of the oral histories collected by Rockman and Cataldi were recorded in varieties of Warlpiri (Ngaliya, Warnayaka) but also in Ngardi, Kukatja, Walmajarri and Gurindji. A highly respected Law woman, Rockman – along with her sisters – was instrumental in securing the success of the Tanami Downs Land Claim (Olney, 1992). Spring Creek Mick [SCM] †, Jack Langgamarru [TJL] †, Jack Lightning [JLI] †, Yarduwuju ‘Saamy’ Jantujukurr [DJD] † All of these men were Ngardi speakers and recorded aspects of their language(s) during the 1970s with Dr Tasaku Tsunoda, both at Nicholson Station and at Halls Creek. I later also worked with the daughter of Jantujukurr – Marie Gordon. Tommy Skeen [TSK] †, Millie Skeen [MSK] †, Sambo Gordon [SGO] † Sambo Gordon recorded a little of the Ngardi language with linguist Joe Blythe in August of 1998. Later in August, Millie and Tommy Skeen also made some short recordings in Ngardi with Joe Blythe, detailing their artworks. Marie Mudgedell [MMN] and Patrick ‘Jupiter’ Smith [PSM] Marie Mudgedell has been involved with documenting and teaching language for nearly three decades. She assisted with the translation of a number of Ngardi texts collected by Sonja Peter and Lee Cataldi and is credited with Ngardi and/or Kukatja translations in a wide range of published formats including the book Yarrtji: Six Women’s Stories from the Great Sandy Desert (Napanangka et al., 1997), the film Tjawa Tjawa (Moora, 2015) and in numerous other capacities through the local school (Luurnpa Catholic School), the women’s centre (Kapululangu), the art centre (Warlayirti Artists) and the local church. Mudgedell is an accredited Kukatja translator and has previously worked as a Kukatja language teacher in the former language department at the local Luurnpa Catholic School (the Walkali Centre). In addition to her involvement in language documentation and education, Mudgedell is also a prominent Law woman and has played an increasingly central role within Kapululangu Woman’s Centre.

1.4 The current grammar 

 39

Marie Mudgedell speaks five languages: Ngardi, Kukatja, (Sturt Creek) Jaru, Eastern Walmajarri and English. One of the most senior men in Balgo, Patrick ‘Jupiter’ Smith is a stockman and a horse-handler. Despite being quick to profess that he never went to school, Jupiter holds an incredible wealth of linguistic and cultural knowledge. He has fluent command of at least eight languages: Ngardi, Jaru (Sturt Creek), Kukatja, Walmajarri (Billiluna/Mulan), Gurindji, Pintupi, Warlpiri and English. Jupiter spent much of his life working on cattle stations throughout the south-east Kimberley and the Northern Territory, and later helped the Mahood family set up their cattle station at Tanami Downs. Jupiter’s insights into the richness of expression within the Ngardi language and his ability to translate across multiple languages has greatly influenced the depth and richness of the analysis able to be presented in this work. The vast majority of my time in Balgo throughout 2016–2020 was spent working with both Marie Mudgedell and Patrick Smith. This work is a testament to their commitment to language documentation. Kathleen ‘Mayan’ Padoon † [KPA] I was lucky enough to work briefly with Kathleen ‘Mayan’ Padoon from 2017. Marie Mudgedell and I recorded an autobiographical piece for her in Ngardi as a complement to her art installations exhibited at DESART 2017. Mayan was an important Ngardi Law woman and a leading figure in the Manungka Manungka Women’s Association (later Kapululangu). She recorded Ngardi language and traditional songs and ceremonies for her country – Ngantalarra – both with Lee Cataldi in the 1990s and then again with me in 2017. Later in her life Mayan became a very successful artist – in 2019 she had a solo exhibition ‘Nakarra Nakarra’ at Alcaston Gallery in Melbourne. Irene Padoon [IPN] Irene Padoon is one of the daughter’s of Kathleen Padoon. She assisted me with Ngardi translations in 2018 and 2019. A speaker of at least four languages (Kukatja, Walmajarri, Jaru and Ngardi), Padoon helped generously with Ngardi dictionary and elicitation work throughout 2018 and 2019. Popeye Tchooga [PJA] Popeye Tchooga is a speaker of Ngardi, Jaru, Warlpiri and English. He has worked with many linguists including Lee Cataldi (documenting Warlpiri), Matthew Wrigley (documenting Ngardi) and, more recently, Josua Dahmen (documenting Jaru). His contribution to Ngardi language documentation were recordings made as part of a wordlist project conducted by the Kimberley Language and Resource Centre. Tchooga’s country is associated with the northern areas of the Tanami Downs Land Claim.

40 

 1 Introduction

Bonnie Seela [BSE] Bonnie Seela is a speaker of Jaru, English, Ngardi and some Kukatja. She worked with Marie Mudgedell and me in Balgo in 2018. Seela has also recently been involved in the documentation of the variety of Jaru spoken at Yaruman (Ringer’s Soak) with the linguist Josua Dahmen. Mark Moora † [MAM] Mark Moora was a senior Ngardi Law man for Ngururrpa country. Brother to Payi Payi Napangardi, Moora was born near Emily Springs, south of Balgo on Ngardi country. He was a speaker of Kukatja, Ngardi and English and has been an immensely important figure in not only native title work (Cane, 2006; Payi Payi & Ors on behalf of the Ngururrpa People v the State of Western Australia, 2007), but was also a driving figure in the establishment of the outstation at Yagga Yagga (Cane, 2016). Moora has been involved in the documentation of cultural knowledge over many years through the due process of native title work as well as heritage documentation work in land surveys. More recently he was also involved with the documentation of songs, stories and cultural knowledge (Lempert, 2018; Moora, 2015). Moora initially recorded some Ngardi language in 2016 during my first fieldtrip to Balgo and later helped check the accuracy of place names throughout 2017–2019. Barbara Sturt [BST] Barbara Sturt grew up around Flora Valley, Sturt Creek and Gordon Downs Stations. She is a speaker of Jaru, English and some Ngardi. Sturt has been involved in the documentation of the Jaru language for some years (e.g. the Jaru ethnobiological dictionary (Deegan et al., 2010)) and assisted with Ngardi translation and recording in 2017 in Halls Creek where she lives. Barbara was also a central member of the successful Jaru Native Title Claim (Sturt on behalf of the Jaru Native Title Claim v State of Western Australia, 2018). Marie Gordon Munyumunyu [MGO] Marie Gordon is a Ngardi, Jaru, Kukatja and English speaker, and daughter of Jantujukurr. Gordon and her late husband, Jack Gordon, helped record some Ngardi in Billiluna and assisted in dictionary elicitation, in 2018 in Balgo. Biddy Timbinah [BTI]; Nelly Gordon [NGO] Biddy Timbinah and Nelly Gordon are both multilingual Ngardi speakers. I visited them in Kununurra with Marie Mudgedell and Patrick Smith. We discussed the Ngardi dictionary project and both shared brief stories (recounted in Ngardi) of their early lives. Marie Mudgedell and I translated their stories.

1.4 The current grammar 

 41

1.4.6 Example sentences Example sentences in this grammar are numbered sequentially. Examples are glossed according to the conventions laid out by the Leipzig Glossing Rules (Comrie et al., 2008), where practicable. Examples are arranged with Ngardi orthography on the main line, glossing on a second line, followed by an English translation (typically my own) on a third line. On occasion, I have found it helpful to include either an alternative translation or a speaker’s own translation verbatim on a fourth line. Speaker’s translations are enclosed in double quotations “ . . . ”. The source of each example utterance is provided in brackets at the end of the entry, following the English translation. This citation includes information about the speaker (see the three letter codes in §1.4.5), the corpus and year, the session or filename and the timestamp of the utterance (in milliseconds). Example citation from my corpus: (MMN: TEN1- 2019_ 032-01: 3040_9090) Speaker Corpus Year Session + subsession Time References to examples cited from materials sourced from Cataldi’s, Tsunoda’s or Green’s corpus are slightly different. The main parameters of variation are that the year is not necessarily included in the filename and the session numbers differ somewhat. Example citation from the Cataldi corpus: (TJN: LC 21 3040_9090) Speaker Collector Filename Time Example citation from the Tsunoda corpus: (PJA: TT76_ 2501 3040_9090) Speaker Collector + year Filename Time Some examples do not come directly from audio-linked corpora but instead are sourced from Lee Cataldi’s (2011) A dictionary of Ngardi. Examples are cited with relevant page number and speaker codes (where known). The dictionary is available online via the University of Sydney e-Scholarship repository (https://ses.library.usyd. edu.au/handle/2123/21407) as of 2021.

42 

 1 Introduction

1.4.7 Overview of topics covered Chapter 2 describes the phonetics and segmental phonology of the language. This includes the description of the phoneme inventory, its orthographic representation, and the range of allophones available to each phoneme. Some limited phonological and morphophonological processes are described. The chapter also includes a description of the structure of the word – including syllable structure, stem structure, positional constraints on phonemes, and permissible consonant clusters. The chapter concludes with a preliminary analysis of lexical stress and a description of select intonational patterns. Chapter 3 introduces the structure of the noun phrase and surveys the derivational and inflectional morphology relevant to the nominal word class. The bulk of the chapter subsequently deals with the various forms and functions of the Ngardi case system. Chapter 4 details the various nominal subclasses in Ngardi: quantifiers (including numerals), demonstratives, ignoratives, free pronouns, directionals, locational and temporal nominals, action nominals, and kin terms. Chapter 5 introduces the system of pronominal enclitics or ‘bound pronouns’. This chapter describes the forms and functions of the bound pronouns, particularly with respect to their role in marking core grammatical relations which are further described in Chapter 8. Additionally, there is discussion of the positioning of bound pronouns within the clause and an analysis of the complex morphology involved with the realisation of multiple bound pronouns. Chapter 6 deals with morphosyntax of the verbal domain. This chapter establishes the two major classes of verbal words in Ngardi: the preverb and the verb. The subsequent focus is on the inflectional categories encoded by the inflecting verb: namely Tense Aspect Mood (TAM), as well as directional and associated motion inflections. Chapter 7 focuses on the formation of complex predicates. This includes a discussion of both serial verb constructions (SVCs) and complex verb constructions (CVCs). The bulk of the chapter deals with the latter since CVCs are one of the most common predicate types. CVCs are described in terms of their constituent members, which can variously include a range of different types of preverbs (uninflecting verbal elements) but also the use of nominals and Kriol origin verbs. This chapter also incorporates an event semantic analysis of the reduced set of inflecting verbs in Ngardi which are ‘productive’ in the formation of CVCs. Chapter 8 examines various simple clause structures. Language-specific definitions for grammatical relations are provided, and clause types are divided according to their valency (avalent, monovalent, bivalent and trivalent) and categorised with respect to their case-marking and pronominal argument properties. This analysis helps to chart canonical clause types (e.g. intransitive, transitive, ditransitive)

1.4 The current grammar 

 43

but also helps elucidate non-canonical variations (e.g. existential, copula, semitransitive, semi-distransitive, etc.). Chapter 9 presents a range of modal and propositional particles and enclitics which can be used to further modify a simple clause. Due to the function of these particles, this chapter deals with such constructions as negation, the formation of commands, and other modal constructions. Chapter 10 moves on to syntax beyond the simple clause and examines the way in which complex clauses are formed. This includes a discussion of finite and non-finite subordination, as well as types of coordination, involving both coordinators and asyndetic phenomena.

2 Phonology This chapter provides an overview of the phonetics and segmental phonology of Ngardi. The segmental inventory (§2.1), working orthography (§2.1.1) and minimal pairs/tuples illustrating phonemic contrasts (§2.1.2) are described. The range of allophones associated with each phoneme is presented in §2.2. Ngardi exhibits only a small number of phonological processes (§2.3) and some lexically restricted morphophonological processes (§2.4). The structure of the word is covered in §2.5 and includes discussion of syllable (§2.5.1) and stem (§2.5.2) structure; positional constraints on phonemes within words (§2.5.3) and permissible consonant clusters (§2.5.4). Some preliminary description within the prosodic domain is provided, in terms of lexical stress (§2.6) and intonation (§2.7). In describing the phonology of Ngardi, I have attempted to focus on presenting known variation in the phonetic data. While the overall phonological system is not notably distinct from many languages within the Ngumpin-Yapa subgroup the phonetics of these languages are largely undescribed. The description of Ngardi phonetics that follows involves primarily qualitative and impressionistic observations and so brings the usual limitations. Nevertheless, I hope it serves to provide a fruitful starting point for future research by focussing on observable phonetic variation underlying the phonological system (Gasser & Bowern, 2014).

2.1 Segmental inventory The inventory of consonant phonemes in Ngardi is set out in Table 5. The descriptive labels assigned to manner and place follow standard Australianist practice. The IPA symbol conveys a typical realisation of the phoneme. An exception is the series of plosives for which the voiceless IPA symbols alone are used. Orthography is enclosed in single quotation marks ‘ . . . ’. Table 5: Ngardi consonant phonemes. Apical Plosive Nasal Lateral Tap/trill Approximant

Laminal

Peripheral

Alveolar

Retroflex

Prepalatal

Bilabial

Velar

/t/ ‘t’ /n/ ‘n’ /l/ ‘l’ /r/ ‘rr’

/ʈ/ ‘rt’ ‘rd’36 /ɳ/ ‘rn’ /ɭ/ ‘rl’

/c/ ‘j’ /ɲ/ ‘ny’ /ʎ/ ‘ly’

/p/ ‘p’ /m/ ‘m’

/k/ ‘k’ /ŋ/ ‘ng’

/ɻ/ ‘r’ ‘rd’

/j/ ‘y’

/w/ ‘w’

36 For discussion of orthographic ‘rd’ see §2.1.1 https://doi.org/10.1515/9783110752434-002

2.1 Segmental inventory 

 45

The distinction between apico-alveolar /t, n, l/ and apico-retroflex /ʈ, ɳ, ɭ/ consonants is neutralised word-initially (see §2.5.3.1.1). Initial apicals are most frequently alveolar and the phonemic neutralisation is rendered with the nonretroflex series /t, n, l/. However, some lexemes are readily perceived with both alveolar and retroflex articulations; for example, /tuʈ/ ‘break’ is heard both as [ʈuʈ] and [tuʈ]. Yet other lexemes are produced consistently with quite deliberate word-initial retroflexes: /taka/ [ʈaka].37 This topic is therefore in need of further investigation. Orthographically, I follow the mainstream Australianist practice of using ‘t’ and not ‘rt’ for initial apicals; i.e. turt not rturt. Where it is necessary to phonetically represent initial apicals, I do so based on my own impressions; however, I make no further claims as to the phonetic nature of the neutralisation. The inventory of vowel phonemes is set out in Table 6. Ngardi distinguishes three vowel qualities (/i, a, u/), each of which makes a phonemic length distinction /iː, aː, uː/. Table 6: Vowel inventory.

High Low

Front

Back

i, ii

u, uu

a, aa

Vowel length has a very low functional load in Ngardi and minimal pairs with short vowels have only been identified within initial syllables of words. This positional constraint on a vowel length contrast is widespread in Australia and found in related languages Warlpiri (Ngumpin-Yapa) (Nash, 1986), Pintupi (Hansen & Hansen, 1969) and Wangkajunga (Jones, 2011, p. 23) (Western Desert); but also further afield in such languages as Arabana-Wangkanguru (Hercus, 1994), Yukulta (Keen, 1983) and Djambarrpuyŋu (Jepson, 2019).

2.1.1 Practical orthography The practical orthography used here is adopted from Cataldi (2011) and was developed in collaboration with the Ngardi speakers – specifically through consultation with Patricia Lee Napangardi. It closely follows the Warlpiri orthography and generally follows the voiceless series of IPA symbols p, k, t, rt, j for plosives /p, k, t, ʈ, c/.38 The orthography follows many well-established conventions for Australian languages

37 These observations for initial apicals parallel very closely the observations made by Senge (2015, p. 80) for Wanyjirra and Tsunoda (1981b, p. 37) for Jaru. 38 As is common in the orthographies of many Australian Aboriginal languages, the palatal plosive /c/ is represented with ‘j’ and the semivowel /j/ with ‘y’.

46 

 2 Phonology

(Dixon, 1980) – including the use of digraphs for three retroflex segments: r followed by the corresponding apico-alveolar series (t, n, l), i.e. rt, rn, rl (see the note on initial retroflexes in the preceding section). The retroflex approximant /ɻ/ is represented by a single r while the symbol rr is used for the apico-alveolar tap/trill /r/. Two other digraphs are used for palatal articulations: ly /ʎ/ and ny /ɲ/. When representing consonant clusters involving two retroflexes, only a single r-digraph is used on the left edge of the cluster. For example, /paɳʈa/ ‘the top’ is rendered as parnta and not parnrta. For palatal nasal + plosive clusters, no evidence for a clear contrast between homorganic /ɲc/ and heterorganic /nc/ clusters has been found. While it is possible to identify differences in articulation between types of clusters, the present work makes use of Cataldi’s (2011) orthography which does not disambiguate the clusters orthographically, i.e. nj is generally used for both /ɲc/ and / nc/. This approach is consistent with orthographic conventions for Warlpiri (see Nash, 1986, p. 10) but differs from those for most Ngumpin languages. An exception to this rule is the segmenting of multimorphemic words where a retroflex or palatal cluster is formed across a morpheme boundary. In these contexts, the two segments are both rendered as digraphs and are separated by a hyphen: e.g. lamparn-rtu /lampaɳ-ʈu/ ‘child-erg’, and parntany-ju /paɳʈaɲ-cu/ ‘old_woman-erg’. A single complicating factor in the Ngardi orthography is Cataldi’s (2011) orthographic decision to distinguish r, rt and rd in the Ngardi dictionary.39 Cataldi posits that rd is not a phonemic flap as in Warlpiri but is simply a ‘voiced’ retroflex plosive for which she heard ‘consistent differences in voicing properties from rt’.40 Orthographic rd is only found word-medially in Cataldi’s materials and no claim is made to a contrast outside intervocalic positions. Inspection of the data has revealed that there is significant intra- and inter-speaker variation in the pronunciation of lexemes containing Cataldi’s rd. Generally speaking, however, Cataldi’s rd can be assigned phonemically to either i) a retroflex glide /ɻ/ or ii) a retroflex stop /ʈ/. Both phonemes appear to have possible realisations as retroflex flaps [ɽ], for some speakers. i)

/ɻ/ as [ɻ] or [ɽ] purda /puɻa-/ [puɽa], [puɻa] ‘hear’

39 Green (1988) similarly transcribes Ngardi using ‘rd’ but worked from the assumption that the phoneme inventory of Ngardi was identical to Warlpiri and possessed phonemic flaps. 40 The phonetic realisation of intervocalic /ʈ/ is typically voiced [ɖ] and hence Cataldi’s proposal that there are consistent voicing differences between graphemes ‘rt’ and ‘rd’ does not hold up to scrutiny.

2.1 Segmental inventory 

ii)

 47

/ʈ/ as [ʈ], [ɖ] or (less commonly) as [ɽ] kardu /kaʈu/ [kaɖu], [kaɽu] ‘wife’

In the present work, I retain the use of rd for consistency with Cataldi (2011), barring one exception. In cases where rd only corresponded to a plosive in the data (without observed flapped articulations) these have been transposed to orthographic ‘rt’ in the present grammar. All other usages of ‘rd’ which occur in lexemes that show clear alternations with a flap have been retained. The nature of retroflex flapping is a highly interesting one in Ngardi and flaps arguably have a near-phonemic status for some speakers – a topic which warrants further instrumental investigation. Setting aside the use of ‘rd’, the Ngardi orthography is otherwise identical to that used for Walmajarri, Gurindji, Malngin, Warlmanpa and Warlpiri. Orthographies of other Ngumpin-Yapa languages are distinguished by the use of a ‘voiced’ series of graphemes for the plosives (d, rd, b, g as opposed to p, t, rt, k). This is the case for Jaru, Wanyjirra, Bilinarra and Mudburra.

2.1.2 Minimal phonemic contrasts This section presents minimal (or near minimal) pairs and tuples that illustrate phonemic contrasts among consonant (§2.1.2.1) and vowel phonemes. Contrasts are presented primarily in intervocalic positions only for reasons of space. 2.1.2.1 Consonantal contrasts Place contrasts are initially shown between the five places of articulation for plosives (Table 7) and nasals (Table 8). Table 7: Place contrasts among the plosive consonants.

/p/: /t/:

/ʈ/: /c/:

/t/

/ʈ/

/c/

/k/

panpa – pantaɳi ‘slow’ – ‘feel.pst’

wapira – waʈi ‘father’ – ‘collarbone’

piʎki – ciʎki ‘mud’ – ‘seed’

papar – pakar ‘mistake’ – ‘thorny’

ŋati – ŋaʈi ‘mother’ – ‘language name’

taka – caka ‘hand’ – ‘bum’

pata – paka ‘fall’ – ‘prickle’

wiʈi – wici ‘naked’ – ‘lizard sp.’

puʈaya – puka ‘bettong’ – ‘rotten’ ŋarca – ŋarka ‘enough’ – ‘comprehend’

48 

 2 Phonology

Table 8: Place contrasts among the nasal consonants.

/m/

/n/

/ɳ/

/ɲ/

/n/

/ɳ/

/ɲ/

/ŋ/

ŋama – ŋana ‘mother’ – ‘who’

ngama – ŋaɳa ‘mother’ – seq.aux.1sg.s

maɳa – ɲaɳa muɲcu – ŋuɲcu top. ‘toothache’ – ‘tobacco’ aux.1sg.s – ‘deep’

pina – piɳa ‘understand’ – ‘nectar’

pina – piɲa ‘understand’ – hit.pst

yani – yaŋi ‘go.pst’ – ‘one’

kaɳa – kaɲa emph. aux.1sg.s – carry.pst

puɳuŋgka – puŋuŋkaɭa ‘tree-loc’ – hit.inf.seq.loc

ɲaɳa – ŋaɳa ‘deep’ – ‘seq.aux.1sg.s’

No minimal pairs for alveolar and prepalatal nasals /n, ɲ/ have been observed in a prepalatal plosive /c/ context. As mentioned in §2.1.1, I generally follow Cataldi’s (2011) orthography and representing all nasal + palatal plosive clusters as ‘nj’. The following examples show contrasts between liquids and apical plosives (8), and between semivowels and plosives (9). (8)

Contrasts between liquids and apical plosives /r, l, t/ muru – mulu – mutu ‘reverse’ – ‘this one’ – ‘red ochre’ /r, ɻ, l, t, ʈ,/ ŋariɳi – ŋaɻi – ŋali – ŋati – ŋaʈi tell.pst – ‘belongings’ – 1du.incl – ‘mother’ – ‘language name’ /ɻ,, ɭ,, ʈ/ wiɻi – wiɭiɲ – wiʈi ‘floodout’ – ‘backbone’ – ‘naked’ /ɻ, l, t/ kiɻa – kila – kita ‘big’ – ‘dead bullock’ – ‘outside’ /r, ɻ/41 karu – kaɻu ‘creek’ – ‘wife’

41 The phonemes /r/ and /ɻ/ do not contrast word-initially, since the apical alveolar /r/ does not occur word-initially.

2.1 Segmental inventory 

(9)

 49

Contrasts between semivowels and plosives /j, w, c/ jaɻu – waɻujani – caɻu ‘slowly’ – ‘climb_go.pst’ – ‘talk’ /w, p, k/ ɻawa – ɻapa – ɻaka ‘underground’ – ‘headache’ – ‘hand’

There is only limited evidence for a contrast between the retroflex approximant /ɻ/ and the initial apico-alveolar plosive /t/. A putative minimal pair is /ɻuwaɻuwa/ ‘urine’ – /tuwa/ ‘store’, where the latter is a recent English borrowing. Nevertheless, many /ɻ/ initial words in Ngardi exhibit a phonemic alternation /ɻ/ ~ /t/, for example /ɻaka ~ taka/ ‘hand, five’ (and see §2.3.3). Many /ɻ/-initial words in Ngardi are cognate with flap-initial words in Western Warlpiri, or /ʈ/-initial words in Eastern Warlpiri. 2.1.2.2 Vowel contrasts Phonemic contrasts amongst the three vowels in word-initial and word-final syllables are presented in Table 9. Table 9: Contrasts between /i/, /u/ and /a/. Word-initial σ

Word-final σ

/i, u/ kita – kuta ‘outside’ – ‘short’

paɳki – paɳku ‘awake’ – ‘cross cousin’

/i, a/ piɳki – paɳki ‘cave’ – ‘awake’

paki – paka ‘wagon’ – ‘thorn’

/a, u/ kanci – kunci ‘leg’ – ‘bauhinia tree’

punta – puntu ‘promised spouse’ – ‘clean’

There is some limited evidence for a phonemic contrast between short and long vowels in Ngardi. Minimal pairs exist for /u, uː/ and /a, aː/ and are shown in (10). No clear examples have been identified for /i, iː/. I treat phonemic long vowels as bimoraic rather than as a sequence of two vowels.42 This helps capture the categorical absence of a sequence of two phonemic vowels of different qualities; *ia, *ua, etc.

42 Note, however, that phonemically long vowels are written as a sequence of two graphemes, e.g. /aː/ ‘aa’.

50 

 2 Phonology

(10) Contrasts between short and long vowels /u, uː/ puɻa – puːɻa ‘bush tomato’ – ‘bush potato’ /a, aː/ kaɻa – kaːɻa43 ‘saltwater’ – ‘east’ The analysis of phonemic vowel length affects the wording of the rule of allomorphy particular to the ergative and locative cases, namely, that it is sensitive to the moraic count of vowel-final stems rather than the syllabic count. Bimoraic stems take one allomorph while trimoraic or longer stems take another as illustrated with the locative allomorphy in Table 10 (see §4.5.3 for details). Table 10: Metrically-conditioned allomorphy of the locative case. Stem

Locative allomorph

/cuwa/ ‘carrot’ /puːɻa/ ‘yam’ /luwaɲca/ ‘coolamon’

-ŋka -ɭa -ɭa

Phonetic long vowels also occur as realisations of vowel-glide-vowel sequences both morpheme internally as well as across morpheme or word boundaries. These are discussed in (§2.2.4.1). There is of course the question of whether or not long vowels can be analysed as underlying vowel-glide-vowel (VCV) sequences where C is a glide that shares place of articulation with the neighbouring vowel, for example [iː] as /iji/. A long vowel analysis is preferred for the following reasons: i) There are some (admittedly limited) instances of (near-)minimal pairs (§2.2.4.1). ii) True phonemic long vowels are never realised as vowel-glide-vowel sequences even when uttered in citation contexts. iii) Certain phonemic vowel-glide-vowel sequences never show variation (i.e. /naja/ [nɐ́jɐ] but *[nɐ́ː]). iv) For those lexemes in which /aː/ is implicated, not only is there no acoustic trace of a glide; but it is further unclear whether it should be assigned to /j/, /w/ or even /ɻ/ at the phonemic level since none of the semivowels have quite the same co-articulatory relationship to /a/ as they do to /i/ or /u/. 43 While used by some Ngardi speakers, this is possibly a Jaru borrowing. The typical Ngardi form is /kakara/.

2.2 Allophonic variation 

 51

Nevertheless, the marginal status of phonemically long vowels is also reflected by their uneven distribution within the lexicon. Phonemically long vowels predominate in lexemes within the class of preverbs (11). Where phonemic long vowels appear within the class of nominals, they are largely restricted to animal or plant terms (12). (11)

caː- ‘agape’ jiːʎ ‘desperate’ raː- ‘clean’

(12) kaːɳka wiːɲ-wiɲ taɭkaːri jatalaːri paɭaːɳ

‘crow’ ‘grey falcon’ ‘mountain devil’ ‘mountain devil’ ‘turtle’

Long vowels are also found fairly frequently in English-derived loan words, likely as a means of retaining the phonetic features of the source language: (13) tiːpiri ‘billycan’ kuːl ‘school’ caːpala ‘prickle’

(literally ‘tea thing’) (from the English ‘sharp fellow’)

2.2 Allophonic variation In this section, the common phonetic realisations of phonemes are described. Wherever possible I provide specific examples from the corpus which instantiate the allophonic pattern being described. However, in some cases where an allophonic rule is of a sufficiently general nature, the data is abstracted away from.

2.2.1 Plosives Oral stop inventories of Australian languages can be categorised as to whether they make one or two place distinctions for plosives made with the tongue apex (apicals) and the tongue blade (laminals) respectively. Ngardi is a ‘double-apical’ /t/, /ʈ/ but single laminal /l/ language and lacks a lamino-dental plosive /t /̪ found in some Australian languages (Hamilton, 1996, p. 56), such as Arrernte (Pama-Nyungan) to the south-east and Kija (Jarragan) to the north. It is also conventional in descriptions of phonologies of Australian languages to further group the velar and bilabial plosives under a class of ‘peripherals’ (Dixon, 2002, p. 549). In Ngardi such a grouping is less

52 

 2 Phonology

well-motivated but is retained here out of descriptive convenience for a small number of phonotactic statements (§2.5).44 In the following subsections I describe the realisation of plosive consonants according to the following acoustic properties: voicing (§2.2.1.1), release (§2.2.1.3), duration (§2.2.1.2), and lenition (§2.2.1.4). In terms of their articulatory features, the bilabial plosive /p/ is articulated with both lips and limited, if any, lip rounding. The apico-alveolar /t/ is articulated with the apex of the tongue raised against the alveolar ridge. Preceding a retroflex in a following syllable, /t/ may show some anticipatory place assimilation. For the apico-postalveolar retroflex /ʈ/, the tongue apex is tipped further back and contact is made with the post-alveolar region. It has been claimed for some Australian languages that under faster and more casual speech, the retraction of the tongue apex is reduced (Butcher, 1995; Round, 2009, p. 42) and this appears to be the case for Ngardi too. The prepalatal plosive /c/ is articulated with the tongue blade rather than the apex. Contact is not as far back as the central area of the hard palate but occurs somewhere between the post-alveolar ridge and the prepalatal region. The term ‘lamino-alveopalatal’ is equally appropriate for this articulation. The velar plosive is the most posterior plosive and is articulated by raising the tongue body up against the velum. The exact nature of the acoustic cues to the various place of articulation contrasts are an area in need of further instrumental research.45 However, the following subsections provide some preliminary findings as to their varying acoustic properties. 2.2.1.1 Voicing Plosives in Ngardi are not contrasted in terms of their voicing properties and voiced and voiceless allophones occur for all plosives. Many factors appear relevant to the distribution of voiced and voiceless allophones, including speech rate and prosodic structure. Place of articulation also appears relevant. Peripherals /p, k/ tended towards voiceless realisations in a range of segmental environments more so than /c, t, ʈ/. This may relate to somewhat longer durational values for peripherals (at least for some speakers, see §2.2.1.2) and concomitant passive devoicing. Voicing of plosives also exhibits inter-speaker variation. Some speakers tend towards more voiceless realisations in all environments compared to other speakers. For example, Marie 44 This can be contrasted with Gurindji (McConvell, 1996b, p. 20) and Wanyjirra (Senge, 2015, pp. 92–96) where productive morphophonological lenition rules appear to target the peripherals as a coherent class. In Ngardi such lenition only applies allophonically in casual speech contexts. 45 There is some anecdotal evidence for partial place neutralisation rarely discussed in the literature on the phonology of Australian languages, namely, the neutralisation of the apical contrast in noninitial contexts and a loss of the prepalatal/retroflex apical distinction. Evidence for this comes from the types of transcription practises of native speakers: impermissible spellings, e.g. rny (i.e. hearing retroflexion with a palatal), and valid alternative spellings, e.g. Jupurrurla or Jupurrula (male subsection), kalarra or karlarra ‘west’ =rnalu or =nalu ‘we (not you)’, among others.

2.2 Allophonic variation 

 53

Mudgedell utilised voiceless (and longer) plosives more frequently and in a wider range of segmental environments than Patrick Smith, who exhibited not only more voiced realisations but also more lenited tokens (§2.2.1.4). The following generalisations can nevertheless be made with respect to voicing allophony, applicable to all plosives. Examples are presented below with various references to both a male (Patrick Smith) and a female (Marie Mudgedell) speaker. Plosives are generally: i) voiceless word-finally (/__#) (14) ii) voiceless in stop clusters (/__C, C__) (15) iii) voiced following nasals (/N__) (16) (14) /tirip/ ‘overnight stay’ /luric/ ‘finish’ /cupaʈ/ ‘warm’

[tɪɾ́ ɪp]

(MMN: TEN1-2018_015-03: 190030)

[lúɾɪc]

(PSM: TEN1-2018_009-01: 78533)

[cʉ́pɐʈ]

(MMN: TEN1-2018_019-01: 3772296)

(15) /kuckuc/ ‘bad situation’ /kak=pi-ɳa/ strike=hit-pst

[kʊ́ckʊc]

(PSM: TEN1-2016_005-03: 690194)

[kákpɪ̀ɲa]

(PSM: TEN1-2019_006-02: 1692837)

(16) /ɲuntu=n/ 2sg=2sg.s /paɳʈi=jira-ɳi/ smell-place-pst /lampaɳ/ ‘child’ /waŋka-ɲa=lu/ speak-pst=3pl.s /wanci-ɭa/ where-loc

[ɲʉ́ndun]

(PSM: TEN1-2016_005-03: 190107)

[pɐ́ɳɖi-jiɾaɳɪ]

(PSM: TEN1-2017_006-02: 79901)

[lɐ́mbɐɳ]

(PSM: TEN1-2016_005-03: 176020)

[wɑ́ŋgɐɲàlu]

(PSM: TEN1-2018_032: 1289232)

[wɑ́nɟɨɭa]

(MMN: TEN1-2019_005-06: 260258)

Moving beyond the environments listed above, much greater variability is found. Both voiced and voiceless realisations of plosives are found following the liquids /r/, /l/, /ɭ/, /ʎ/. Examples of variable realisations of peripheral plosives following liquids are presented in (17).

54 

(17)

 2 Phonology

Variably voiced allophones of peripheral plosives following liquids /wirpa/ [wɪ ŕ bɐ] (PSM: TEN1-2018038-02: 1192156) ‘many’ /wirkiɭ/ [wɪ ŕ gɨɭ] (PSM: TEN1-2018_038-02: 1384894) ‘hair’ /murku/ [múrku] (MMN: TEN1-2019_005-06: 340888) ‘boy’ /taɭkaːri/ [táɭgɐ̀ːɾi] (PSM: TEN1-2018_015-01: 84168) ‘mountain devil’ /palku/ [bálku] (PSM: TEN1-2016_005-03: 225723)46 ‘Balgo’

Under sufficiently emphatic or slow speech contexts (especially in single-word elicitation) plosives may be voiceless in more or less any environment, including post-nasal and intervocalic environments. In some cases, this appears to be a result of passive devoicing rather than active modulation of voicing. Voiceless intervocalic plosives were invariably of longer duration than voiced counterparts. Durations of voiceless peripheral tokens in particular were observed ranging between 100–200 milliseconds (ms); compare similar findings by Ennever, Meakins and Round (2017, p. 12) for Gurindji. For many tokens, a gradual decrease in intensity associated with voicing is observed consistent with the passive effects of a reduction in trans-glottal air pressure. Studies on the limits of passive voicing in plosives in different phonetic environments (Ohala, 1983; Stevens, 1998, pp. 465–466) reliably show that plosives are subject to passive devoicing only 25–100 ms after the onset of oral occlusion, depending on additional factors (subglottal pressure, muscular tension of the vocal tract and place of constriction). Word-medial plosives, perceived impressionistically as voiceless were investigated acoustically. The majority of voiceless tokens were on average longer than 50 ms and in fact were found with durations in excess of 100 ms. Indicative examples are provided in Figure 3. Spectrograms show 0–7000 Hz. 2.2.1.2 Duration The duration of plosives in intervocalic word-medial and intervocalic word-initial (i.e. plosives following a vowel-final word) contexts were investigated in order to determine any consistent differences between places of articulation. Duration values were automatically calculated within Praat (Boersma & Weenik, 2019) using an R-based script designed to extract durational values and estimate lenition properties based on relative changes in intensity (see Ennever et al., 2017 for details and justification). To begin, I focus on intervocalic, word-medial realisations of plosives in which the full range of segmental contrasts are observed and segmentation is the most straightforward. I report on 46 Initial /p/ in this token was voiced as it was preceded by a word-final vowel.

2.2 Allophonic variation 

 55

Figure 3: Passively devoiced intervocalic plosives /p/ (196 ms) in /ŋapa/ ‘water’ and /ʈ/ (124 ms) in / kaʈaɲ/ ‘coals’.

data from highly controlled single-word elicitation (one male and two female speakers (Figure 4)) as well as from casual, unprompted speech (one male speaker (Figure 5)).

Figure 4: Closure durations of intervocalic word-medial plosives (n=333).

The data in Figure 4 corroborate an impressionistic observation: there is substantial variation between speakers with respect to plosive durations. The speaker with

56 

 2 Phonology

the longest duration plosives (MMN) produced oral constrictions over 100 ms on average for /t/, /ʈ/ and /p/ compared to average durations between 70 and 90 ms for the two other speakers examined. MMN also produced voiceless plosives more often both in elicited speech and in casual speech contexts. Speakers also differed with respect to which places of articulation were associated with longer closure durations. For both MMN and BSE, velar /k/ had the shortest closure durations overall, while for PSM velar /k/ had the longest closure durations. A commonality was that all speakers consistently exhibited longer closure durations on average for the apicoretroflex plosives compared to the apico-alveolars. These data indicate that closure duration may play a role in distinguishing /t/ and /ʈ/. However, not only is there significant overlap in the range of values observed, but data collected from casual speech contexts (Figure 5) indicate that such tentative durational distinctions are not maintained across faster speech contexts. Figure 5 displays segment durations of intervocalic plosives as recorded during non-elicited, casual speech (one narrative) of a single male speaker, n=263. Since these data were recorded during casual speech, I include comparison between intervocalic word-medial and intervocalic word-initial tokens (i.e. both V_V and V# _V). Since the apicals are not distinguished word-initially, word-initial data were collapsed and presented under /t/ alone.

Figure 5: Duration of intervocalic plosives (initial and medial) in casual speech (n=263).

2.2 Allophonic variation 

 57

Figure 5 illustrates the unsurprising finding that under casual speech contexts all plosives were of shorter durations on average than in elicited contexts (compare plosive closure durations of PSM in Figure 4). In terms of positional effects, word-medial apical plosives were on average shorter than either the peripheral or palatal plosives in casual speech contexts. Somewhat unexpectedly, word-initial intervocalic plosives were not longer (and in some cases were even shorter on average, e.g. /c/ and /k/) than intervocalic medial plosives, with the exception of alveolar /t/. This runs counter to the expectation that consonants in the onset of a (primary) stressed syllable and at a word boundary would be less likely to be temporally reduced than in word-medial, unstressed positions (see e.g. Ségéral & Scheer, 2008). Outside of intervocalic contexts, plosives can also be reduced. Specifically, in nasal + plosive clusters, the acoustic properties of a discrete oral occlusion may be masked by delayed velic-raising. This results in the phonetic effect of post-nasal plosive deletion as shown in the spectrogram showed in Figure 6. Some evidence of the turbulence associated with the burst release of the palatal oral occlusion is visible leading into the onset of the vowel. Spectrogram shows 0–7000 Hz.

Figure 6: Deletion of a plosive in a nasal + plosive cluster /c/ in /waɲciɭa/ ‘where’.

Deletion of post-nasal plosives in favour of a prolonged nasal is reported in other Australian languages such as Pitjantjatjara and Burarra and appears to be a common manifestation of ‘preservative assimilation of nasality’ (Butcher, 1999, p. 480). These data represent a very preliminary investigation of plosive durations in Ngardi. Further work needs to be done to distinguish both the relationship between

58 

 2 Phonology

lexical stress and segment duration as well as higher-level prosodic factors (for example phrase-final lengthening). Nevertheless, based on this preliminary evidence, it does not appear that plosives are systematically distinguished by constriction duration and, moreover, there is notable inter-speaker variation with respect to segment durations. 2.2.1.3 Release Plosive releases are also considered relevant in cueing place of articulation contrasts since acoustic bursts carry acoustic properties associated with the resonating cavity formed behind the place of constriction (Ohala, 1990, p. 265). Plosive releases are restricted to prevocalic and prepause environments. Plosives occupying initial positions within clusters and word-final plosives are unreleased in Ngardi. The burst release properties of plosives in Ngardi can be broadly characterised as follows. Bilabial /p/ was found to exhibit only low intensity bursts (due to the lack of an anterior resonating cavity); the apicals exhibited short, low-intensity bursts. In contrast, velar /k/ and prepalatal /c/ are often realised with not only relatively strong bursts but were optionally realised as affricates.47 Affricated velars tend to occur in word-initial positions and are thus voiceless (i.e [k͡x]). Palatal /c/ tends to be realised as [c͡ç] (a voiceless, non-sibilant palatal-affricate). Affricated realisations are still produced within the prepalatal region, rather than being anterior in the vocal tract and produced as a canonical sibilant of the English variety, i.e [tʃ]. This impressionistic observation requires further instrumental study, however. A spectrogram illustrating affricated realisations of /k/ and /c/ within a single word is provided in Figure 7. Clear periods of frication can be observed upon the release of the oral occlusion. Spectrogram shows 0–7000 Hz. Duration of selection window is 670 ms.

47 Heavily fricated releases of /k/ were generally only identified for male speakers in the data.

2.2 Allophonic variation 

 59

Figure 7: Affricate realisations of /k/ and /c/ in /kaɲci/ ‘thigh’.

The resonant frequencies of the first three harmonics within bursts of different plosives were examined to determine any systematic differences by place of articulation. Table 11 presents the averaged frequencies of the first three harmonics of burst releases for each plosive preceding the vowel /a/. Normalised values were averaged from five tokens of each plosives, uttered by a single male speaker. Spectral slices showing average energy up to 10 kHz for each plosive are presented in Figure 8. Table 11: Resonant frequencies of first three harmonics in burst spectra; values (in Hz) averaged from five tokens of each plosive preceding /a/.

/k/ /p/ /c/ /t/ /ʈ/

H1

H2

H3

511 281 405 358 268

1982 1505 1596 1303 1317

2927 2787 2596 2177 2462

60 

 2 Phonology

Figure 8: Mean relative distribution of intensity in release burst spectra for plosives; mean taken from five normalised tokens of each plosive.

The burst spectra data in Table 11 and the spectral slices in Figure 8 show that bilabial /p/ is clearly distinguished from the other plosives by the relative lack of energy high in the spectrum and a relatively lower first harmonic frequency. Both /c/ and /t/ exhibit significant energy for frequencies up to 4 kHz before energy levels drop away. The prepalatal /c/ is differentiated from all other plosives by the retention of relatively higher energies in the higher frequencies greater than 5 kHz. 2.2.1.4 Lenition All plosives (with the exception of /c/) exhibit some degree of gradient lenition involving both temporal reduction and the loss of complete constrictions. Examples were found in both casual and highly targeted (single-word elicitation) contexts. The lenition patterns are provided in (18).

2.2 Allophonic variation 

(18) /p/ → [β̞], [w] /k/ →[ɰ] /t/ → [ɾ], /ʈ/ → [ɽ]

/ / / /

 61

V{#, -}__V V{#, -}__V V__V V__V

The predominant pattern of lenition in Ngardi is for the peripherals to be lenited to frictionless approximants at the same place of articulation and for this to apply intervocalically irrespective of word or morpheme boundaries. The peripherals /p, k/ lenite to their corresponding glide realisations [β̞]and [ɰ]. My general impression is that bilabial /p/ does not add a velar articulation under lenition (‘hence’ representation as [β̞]) but certain speakers do occasionally use a double articulated labio-velar articulation [w] consistent with morphophonological patterns of lenition found in other Ngumpin languages (e.g. Bilinarra: Meakins & Nordlinger, 2014, pp. 67–69). Allophonic lenition of the peripherals /p/ and /k/ is also found in Walmajarri (Hudson & Richards, 1969, p. 174) but involves fricative rather than approximant allophones. In Ngardi, generally only voiced approximants (i.e. [β̞] and [ɰ] in (18)) were observed, a finding consistent with studies of Gurindji (Ennever, 2014; Ennever et al., 2017). The articulatory movement involved in the target of an alveolar plosive /t/ is also reduced in casual speech contexts until it is indistinguishable from a tap [ɾ] (and hence indistinguishable from realisations of intervocalic /r/). The flapped allophone [ɽ] of the retroflex plosive /ʈ/ is limited to certain speakers. Unlike the tapped realisation of /t/, flapping involves a discrete articulatory mechanism. For flaps, contact with the passive articulator (generally the post-alveolar ridge region) is made with a forward, and continuous motion of a retracted tongue tip rather than a single ballistic motion towards (and away) from the alveolar region (see Derrick & Gick, 2011). Since the acoustic results are similar – shorter duration (than plosives), sustained voicing, minimal reduction in intensity throughout the segment, and no release burst – it is treated here as a type of lenition. While the lenition patterns set out in (18) imply a degree of categorical allophonic patterning, Ngardi appears to share very similar patterns of gradient lenition processes affecting plosives in intervocalic positions as found in Gurindji (see Ennever (2014) and Ennever et al. (2017)). Such plosive realisations can lack the full acoustic characteristics of a glide but nevertheless exhibit both persistent formant structure throughout the constriction period and a small relative decrease in intensity only. These were remarkably frequent throughout the speech of certain individuals in casual speech. Examples of such ‘incomplete closures’ are presented in Figure 9 for tokens of /k/ in morpheme and word-initial intervocalic environments respectively. Spectrograms show 0–7000 Hz and superimposed intensity curves show 0–90 dB. Approximated boundaries for partial constrictions have a duration of 50 ms and 35 ms respectively.

62 

 2 Phonology

Figure 9: Partially lenited examples of morpheme-initial /k/ in /jaŋi-kaɻiɲ/ one-other and wordinitial /k/ in /wali=lu kaɳa/ emph.aux=1sg.s.

Incomplete oral closures such as those illustrated in Figure 9 are quite frequent. Further research is required to determine whether there are certain word or morpheme boundaries that further condition the likelihood of phonetic reduction. It was impressionistically noted that the lenition of initial plosives within certain highfrequency verb stems (e.g. ka- carry, pi-/pu- hit) was somewhat more frequent than other morphophonological environments; for example, /paɭi=pi-ɲa/ [pɐɭɪβ̞iɲɑ]. It is possible that this particular morphological boundary permits a greater degree of phonetic reduction than other morpheme or word boundaries. 2.2.1.5 Summary of characteristics of plosives To summarise the features of plosives examined here: the four plosives are not consistently distinguished by voicing or duration properties – and in fact there is notable inter-speaker variation along these parameters. Voicing is only partly predictable by segmental environment and is likely to be affected by many compounded factors: speech rate, segmental duration and other prosodic factors not investigated here. Plosives were distinguished more robustly by their release properties (bursts and frication) and the prepalatal and velar plosives even exhibited affricate realisations. In terms of lenition, the peripherals /p/ and /k/ tended towards the highest rates of lenition, while the alveolars also showed significant reduction towards taps and flaps. The prepalatal plosive tended to not undergo reduction and was typically heard with

2.2 Allophonic variation 

 63

a fricated release in a range of segmental contexts. All plosives nevertheless showed reduction in terms of their duration during casual speech contexts.

2.2.2 Nasals The nasal series of stops in Ngardi are typically voiced and exhibit the same place contrasts as found for the plosive series. No significant allophony was observed in terms of manner of articulation, with the exception of word-final nasals which may be optionally devoiced or produced with creaky or breathy voice. Additionally, in nasal + plosive clusters, late velic raising may result in the (acoustic) deletion of the plosive and the ‘lengthening’ of the nasal component (an example of this process was provided in §2.2.1.1). Ngardi does not possess any pre-stopped nasals, which are widely distributed across Australia (Butcher, 1999; Fletcher & Butcher, 2014, pp. 109–110; Stoakes, Fletcher & Butcher, 2019) and are found as allophones in neighbouring Warlpiri, (Butcher & Loakes, 2008)) or as fully-fledged phonemes in such languages as Eastern and Central Arrernte (Breen & Dobson, 2005; Henderson, 2002).

2.2.3 Liquids 2.2.3.1 Tap/trill /r/ The liquid phoneme /r/ has the following allophones: − −

[ɾ], [ɹ] [r]

/ V_V / Elsewhere

(19) (20)

In intervocalic environments, an alveolar tap [ɾ] is the most frequent allophone but may be lenited to a continuant approximant [ɹ] which is more or less indistinguishable from [ɻ]. (19) /lira/ [líɾa] ‘mouth’ /tirip/ [tíɹip], [tíɾip] ‘overnight trip’ /paɻari/ [pɐ́ɻɐɹi] ‘seed sp.’

(SCM: TT76_1202: 146972) (BSE: TEN1-2018_014-01: 966097), (MMN: TEN1-2018_014-01: 1314934) (MMN: TEN1-2018_015-01: 1459789)

The trill allophone [r] is generally voiced but may exhibit devoicing or breathy voice in word-final positions.

64 

 2 Phonology

(20) /kirkir/ ‘brown’ /kaɻar/ ‘bush raisin’

[kírkir]

(PSM: TEN1-2018_009-01: 661019)

[kɐ́ɻɐr̥]

(MMN: TEN1-2018_015-01: 1450598)

Examples of tap [ɾ] and approximant [ɹ] realisations of /r/ in /tirip/ by two different speakers are provided in Figure 10. Duration of the [ɾ] token is 38 ms (selection window is 352 ms); approximate duration of the [ɹ] token is 65 ms (selection window is 348 ms). Spectrograms show 0–7000 Hz.

Figure 10: (a) Tap [ɾ] and (b) approximant [ɻ] realisations of intervocalic /r/ in /tirip/ ‘overnight stay’.

It is not unusual for taps and approximants to pattern as allophones in Australian languages despite the equally widespread distribution of a phonemic opposition between a retroflex approximant and an alveolar tap. Closely related languages Jaru (Tsunoda, 1981b, p. 25) and Bilinarra (Meakins & Nordlinger, 2014, p. 48) and the unrelated Tangkic language Kayardild (Round, 2009, p. 63) all show allophonic variation between taps and approximants. 2.2.3.2 Laterals /l/, /ɭ/, /ʎ/ The three lateral phonemes are generally voiced and do not exhibit any noticeable allophony in manner or place of articulation. As with the nasal and plosive series the alveolar/retroflex distinction is neutralised word-initially. Word-initial laterals do not appear to be articulated with any retraction of the tongue tip and appear to be ‘plain’ alveolars, i.e. [l]. This requires further instrumental verification, however. Word-final laterals are quite rare (see §2.5.3.2) but when appearing in this position they may be devoiced (as observed for nasals), e.g. /kuntaʎ/ [kuntaʎ̥] ‘daughter’, or exhibit oth-

2.2 Allophonic variation 

 65

erwise modified phonation properties, such as breathy or creaky voice. Pre-stopped laterals are reported allophonically for neighbouring Warlpiri (Loakes et al., 2008); however, no such articulations have as yet been observed in the Ngardi data.

2.2.4 Semivowels There are three semivowels in Ngardi: two glides /j, w/ (§2.2.4.1) and the retroflex approximant /ɻ/ (§2.2.4.2). The approximant /ɻ/ is classified alongside the glides as a semivowel as it exhibits similar phonotactic constraints (§2.5.4). 2.2.4.1 Glides /j, w/ The glides /j, w/ are both frictionless, voiced continuants. The glide /j/ is realised as [j] and, while it is aligned within the same ‘place’ series as the prepalatal (or ‘alveopalatal’) plosive /c/, strictly speaking it has a more posterior ‘palatal’ articulation. The labiovelar /w/ is typically produced as doubly articulated [w] with variable lip rounding. The glides /j, w/ are unique among the Ngardi consonants in that, for at least some words, they are optionally deleted word-initially, giving rise to vowel-initial words as illustrated in (21). The presence or absence of an initial glide is non-contrastive. (21) /wuɻucu/ ‘good’ /jiriɻiɲ/ ‘very hot’ /wura/ ‘wait’ /wuɳa/ ‘journey’

[ʊ́ɽoɟʉ], [wʊ́ɽoɟʉ] [jíɾɪɻiɲ, [íɾɪɻiɲ] [wúɾa], [úɾa] [wʊ́ɳa], [ʊ́ɳa]

If a glide is situated between two vowels of the same quality, glide deletion gives rise to phonetically long vowels within words (22) and across word boundaries in connected speech (23). The resulting long vowels are influenced by the wider segmental environment (§2.2.5). (22) a.

/awa/. /waɭawaɳu/ ‘ready’ /langa-waŋu/ ear-priv ‘deaf’

[wɑ́ɭɑ̀ːɳʊ] [lɐ́ŋɑ̀ːŋʊ]

66 

 2 Phonology

b.

c.

d.

/aja/ /majara/ [mɐ́jæɾa] [mǽːɾa] ‘again’ /pi-ɲa=ɳa=janu/ [píɲaɳɐ̀ːnʊ] hit-pst=1sg.s=3pl.o ‘I hit them.’ /iji/ /kaci=ji=n/ [káɟìːn] condit.1sg.o=2sg.s ‘If you (something) me.’ /uwu/ /wanti-ja=lu=wu/ [wɑ́ndijalʊː], [wɑ́ndijalɔː] fall-imp=pl.s=voc ‘You lot get down!’

(23) /Tuk-kamp-kura=ɳa ja-ni/ stock_camp-all=1sg.s go-pst ‘I went to the stock camp.’

[tókkɐ̀mgʊ̀rɐɳɐ́ːni]

Glide deletion between identical vowels gives rise to a neutralisation between phonemic vowel-glide-vowel sequences underlying long vowels. Some very restricted examples of minimal pairs can nevertheless be identified for /aː/ and /awa/ and /aya/ as shown in (24). However, no minimal pairs have been identified for /iyi/ and /iː/ nor for /uwu/ and /uː/. This is in no small part due to the limited number of words involving /uː/ and /iː/ and furthermore reflects an Australian-wide dispreference for contrasting long vowels with vowel-glide-vowel sequences (Hamilton, 1996). (24) a.

b.

/aː, aja/ waː ‘exclamation of surprise/fear’ waja ‘grill for cooking over coals’ /aː, awa/ raa=ja-ni clear=go-pst ‘it cleared up (e.g. weather)’ rawa=ja-ni separate=go-pst ‘he went separately’

In cases where a glide occurs between vowels of different qualities, glide deletion either gives rise to phonetic diphthongs (25) or vowel hiatus is resolved in favour

2.2 Allophonic variation 

 67

of the rightward vowel (with or without accompanying lengthening) (26).48 Hiatus resolution as shown in (26) has only been observed in morphologically complex words, operating across morpheme boundaries. The exact conditioning factors are unknown – the same phonological sequences in different morphemes never undergo the same vowel + glide deletion. (25)

a.

b.

c. d.

(26)

a.

/aji/

/kajili/ ‘north’ [ɐɪ] /maji/ ‘maybe’ /awu/ [ɐo], [oː] /ja-nku=ɭipa=wu/ go-pot=1pl.incl.s=voc [ɐwʊ], [ɐʊ] /kawuɳ/ ‘ashes’ /ija/ [ija], [ia] tijatija/ ‘magpie lark’ /uji/ [ʊɪ] /ŋu=ji=n/ seq.aux=1sg.o=2sg.s

/uwa/

[æɪ]

[ʷɑ]

[ɐː] b.

/uja/

[ɐː]

[kǽjilɪ], [kǽɪlɪ] [mɐjɪ], [mɐɪ] [jánguɭɪpɐ̀oː], [jánguɭɪpòː] [kɐ́wʊɳ], [kɐ́ʊɳ] [tíyadìya], [tíadìa] [ŋʊ́ɪn]

/jaŋi-kaɻiɲ-ku=wali/ [jɐ́ŋɪgɑ̀ɻɪɲgʷɑ́li] one-other-dat=then /kaʈija-kaɻiɲ-cu=wali/ [kɑ́ɽɪjɐgɑ̀ɻɪɲɟʷɑ́li] white_man-other-erg=then /waŋka-nyu-wangu/ [wɑ́ŋgɐɲɐ̀ːŋʊ] speak-inf-priv /majun-tu=janu/ [mɐ́jʉntɐ̀ːnʊ] wind-erg=3pl.o

A separate rule of morpheme-initial glide deletion has led to the partial reanalysis of certain complex verbs involving a glide-initial verb and a plosive-final preverb (27). The partial nature of the reanalysis of the complex verb construction as a single verb is further affirmed by the variable stress placement on the initial or pen-initial syllable (see §2.6.1.2.1). (27) a.

/p-wa/

b.

/k-wa/

/cup=wanti/ go_down=fall /luk=wanti/ lie_down=fall

[jʉpɐ́nti-] ~ [jʉ́pɐnti-] [lukɑ́nti-] ~ [lúkɑnti-]

48 Lexical data from Cataldi (2011) include words with the sequence /iyu/ and /uwi/; however, I do not have copies of the relevant audio files to determine their phonetic realisation.

68 

 2 Phonology

As with the other glide deletion processes described in this section, this process has not applied categorically. Certain plosive + verb stem-initial glides categorically fail to undergo deletion, for example, /tarp=wanti/ → *[tɐrpɐntɪ-]; or else exhibit variation, for example, /kapurk=wantinya/ → [kɑ́pʊrkɑ̀ntiɲɐ], or [kɐ́pʊrkwɑ̀ntiɲɐ]. 2.2.4.2 Retroflex approximant /ɻ/ The phoneme /ɻ/ has two allophones in apparent free variation: a continuant with a slightly retracted tongue tip [ɻ] and a flap [ɽ]. The flap [ɽ] is articulatorily distinct from the tapped allophone of /r/, [ɾ], since it involves continuous forward motion of a retracted tongue tip such that it strikes the post-alveolar region and subsequently continues its forward trajectory within the vocal tract (see Derrick & Gick, 2011). The allophones [ɻ] and [ɽ] do not appear conditioned by a segmental environment but instead appear to vary freely depending on the speaker. While the sounds are non-contrastive in Ngardi, some speakers such as Marie Mudgedell were aware of the differences. Figures 11–13 provide spectrograms illustrating variation between [ɽ] and [ɻ]. All spectrograms show 0–7000 Hz, with intensity contours overlayed (0–90 dB).

Figure 11: Flap [ɽ] (a) and continuant [ɻ] (b) realisations of yaluwardingki /jaluwaɻiŋki/ ‘a long time’ by a single female speaker; [ɽ] is 55 ms; no discrete boundaries for [ɻ].

A number of lexemes have been identified that exhibit particularly productive variability between [ɽ] and [ɻ]. An example set is provided in Table 12. The apparent free variation between [ɻ] and [ɽ] is complicated by two observations. First, there are lexemes involving /ɻ/ which seemingly never exhibit flapped

2.2 Allophonic variation 

 69

Figure 12: Flap [ɽ] (a) and approximant [ɻ] (b) realisations of intervocalic /ɻ/ in karapurta /kaɻapuʈa/ ‘westerly wind’ from a single male speaker; flap is 57 ms; no discrete boundaries for [ɻ].

Figure 13: Flap [ɽ] (a) and approximant [ɻ] (b) realisations of pirdiny /piɻiɲ/ ‘arise’ by two different speakers. Table 12: Examples of lexemes exhibiting [ɽ] ~ [ɻ] alternations. Lexemes

Realisations

Gloss

kirda /kiɻa/ pirda /piɻa/ puurda /puːɻa/ jingkirdi- /ciŋgiɻi-/ yumpardi /jumpaɻi/ pimirdi /pimiɻi/ -paɽu /-paɻu/

[kɨ́ɻa] [kɨ́ɽa] [piɻ á ] [píɽa] [púːɻa] [púːɽa] [ciŋgiɻi] [cíŋgiɽi] [jʉ́mbɑɻi] [jʉ́mbɑɽi] [pɪmɪɻɪ] [pɪmɪɽɪ] [pɐ̀ɻʊ] [pɐ̀ɽʊ]

‘big’ ‘full’ ‘bush potato’ ‘laugh’ ‘camp’ ‘aunt’ diminutive suffix

70 

 2 Phonology

allophones, e.g. /maɻumpu/ ‘hand’. Second, there are yet other lexemes which exhibit a phonemic alternation between /ʈ/ and /ɻ/ (see §2.3.3), and thus may be variably realised with a retroflex plosive [ɖ], a flap [ɽ] or a continuant [ɻ]. These facts may actually point towards a situation in which Ngardi was losing (or perhaps acquiring) a phonemic flap /ɽ/ but the change was not fully realised. In support of this idea are a number of correspondences between intervocalic /ɻ/ in Ngardi (realised as [ɽ] ~ [ɻ]) and flaps in Western Warlpiri on the one hand but retroflex approximants in Walmajarri and Jaru on the other. Some indicative examples are listed in Table 13 along with the observed allophonic realisations in Ngardi. Table 13: Comparanda sets for intervocalic /ɻ/ in Ngardi, Walmajarri, Jaru and Warlpiri (Western dialect). Ngardi

Walmajarri

Jaru

Warlpiri (Western)

kardany ‘hot coals’

/kaʈaɲ/ ~ /kaɻaɲ/ [ʈ], [ɽ]

/kaʈaɲ/

/kaɻaɲ/

/kaɽaɲcaɽu/ ‘bread’

kardu ‘wife’

/kaɻu/ [ɻ], [ɽ]

/kaʈu/

/kaʈu/

/kaɽu/

kirda ‘big’

/kiɻa/ [ɽ], [ɻ]

/kiʈa/

/kiʈa/

/kiɽa/

kardiya ‘white man’

/kaʈija/ ~ /kaɻija/ [ɖ], [ɻ], [ɽ]

/kaʈija/ ~ /kaɻija/49 [ɖ], [ɻ], [ɽ]

/kaɽija/

marda‘have, hold’

/maɻa/ [ɻ], [ɽ]

pirdiny‘arise’

/piɻiɲ/ [ɻ], [ɽ]

ngardahyp

/ŋaɻa/, /ŋara/ [ɽ], [ɻ],[ɾ]

/ŋaʈa/

raka /ɻaka/ ‘hand’

/taka/ ~ /ɻaka/ [ɻ],[ɽ],[ʈ]

/taka/50

-wardingki denizen

/-waɻiŋki/ [ɻ], [ɽ]

/maʈa/

/maɽa/

/ŋara/

/ŋara/ /ɽaka/ /-waɻiŋgki/

4950

Other Ngumpin-Yapa languages also have flaps in their phonetic inventories either as allophones of the retroflex approximant phoneme /ɻ/ or the retroflex plosive /ʈ/. Walmajarri exhibits the same pattern as Ngardi: flaps are allophones both of the plosive

49 A phoneme alternation /ʈ/ ~ /ɻ/ is also found for the cognate lexeme in Wanyjirra (Senge, 2015, 76). 50 Eastern Walmajarri only.

2.2 Allophonic variation 

 71

/ʈ/ and the approximant /ɻ/.51 In the variety of Nyininy spoken at Gordon Downs, /ɻ/ varies between an approximant [ɻ] and a flap [ɽ] (Tsunoda, 1981b, pp. 25–26) and the same variation is observed in the Western Desert language Pintupi (Hansen & Hansen, 1969, p. 156). Western Warlpiri is notable for possessing phonemic flaps which result in a three-way distinction between /ɽ/, /ɻ/ and /ʈ/ (Laughren & Ingram, 1999; Laughren, 2000).52 This resulted from some (but not all) retroflex plosives (/ʈ/) being re-phonologised as /ɽ/ except when followed by another retroflex in a subsequent syllable. In contrast to the Western varieties of Warlpiri, Lander and Hanson rivers dialects have flap allophones of retroflex plosives but these are not assigned independent phonemic status (Laughren & Ingram, 1999). The variable status of [ɽ] in neighbouring languages is summarised in Table 14. Table 14: Status of retroflex flaps in neighbouring languages. Language variety

Phonemic/allophonic status of retroflex flaps

Western Warlpiri

/ɽ/, /ɻ/ and /ʈ/ are contrastive word-initially /ɽ/ may lenite word-medially to [ɻ]

Eastern Warlpiri

[ɻ] is an optional allophone of [ʈ] in word-medial intervocalic positions

Wawarl and Nyininy Jaru

No reports of [ɽ]

Nyininy- Gordon Downs variety only

[ɽ] is an allophone of /ɻ/ in word-medial intervocalic positions

Walmajarri

[ɽ] (or [ɾ]?) is an allophone of /ɻ/ and /ʈ/

Ngardi

[ɽ] is an allophone of /ʈ/ and /ɻ/

Pintupi (Western Desert)

[ɽ] is an allophone of /ɻ/ (environment unspecified)

A more complete characterisation of the retroflex flap in Ngardi and its relationship to flaps in neighbouring languages warrants further investigation. The present analysis does not revoke the existing orthography which is sensitive to lexemes with an alternation ([ɻ] or [ɽ]) but nevertheless it is clear that Ngardi shows no evidence of a phonemic contrast in the available data.

51 Hudson (1978) uses the term ‘flap’ but does not make a distinction between taps and flaps. 52 In fact Warlpiri is only one of a handful of Australian languages that possess a place contrast between two non-lateral liquids (Hamilton, 1996, p. 66). Other languages include Amurdak, Kurrtjar, Wembawemba and Wergaia.

72 

 2 Phonology

2.2.5 Vowels All three short vowels show allophonic variation based on coarticulation with neighbouring consonants. As a general rule, the three vowels are typically more centralised within the vowel space than the three IPA symbols /i, a, u/ would suggest. Impressionistically assessed allophony of each of the short vowels /i, a, u/ is described in §2.2.5.1–§2.2.5.3. Phonemically long vowels have not been observed with a wide range of allophony and are described together in §2.2.5.4. All vowels are also subject to phonetic lengthening and modified phonation properties in phrase-final positions. This ranges from breathy or creaky voice (e.g. [a]) to fully devoiced (e.g. [ḁ]) vowels. Some referenced examples are listed in (28). (28) a. b. c. d.

/pirija/ [píɾijɐ̥ːːː] ‘cold’ /citakara=ya-nanta/ [cítɑkɑ̀rɐjɐ̀nɐntɐ̥] walking_stick=go-prs /puliki/ [pʉ́lɪkɪ̤] ‘cattle’ /mitwak-ku/ [mítwɔkkʉ̤] meatwork-dat

(PSM: TEN1-2018_014-01: 1515058) (PSM: TEN1-2018_015-01: 350318) (PSM: TEN1-2018_038-02: 116970) (PSM: TEN1-2018_038-02: 391506)

2.2.5.1 Allophony of /a/ The low vowel /a/ has a relatively wide range of allophonic realisations. A widely distributed allophone is [ɐ]. Preceding or following a velar consonant (/k/, /ŋ/ and /w/), /a/ is generally backed to [ɑ].53 (29) a. b. c. d.

/miɲa/ /paka/ /waɭu/ /ŋamaɳa/

[míɲɐ] [pɑ́gɑ] [wɑ́ɭʊ] [ŋɑ́mɐɳɐ]

‘this one’ ‘thorn, prickle’ ‘fire’ ‘yam’

Preceding or following palatal consonants have a raising (and centralising) effect on /a/ and it may be realised as [æ], [ɐ] or [ɛ]. (30) a. b. c.

/cacaji/ /kajira/ /janta/

[cɐ́ɟæɪ] ‘my mother’ [kǽɪrɐ] ‘north’ [jɐ́ndɐ], [jǽndɐ] ‘go!’

53 Some speakers additionally exhibit rounding of backed /a/. This is most perciptible in contexts where /a/ appears in the context of two adjacent peripheral consonants, e.g. /waŋ/ can be realised closer to [wɔŋ].

2.2 Allophonic variation 

 73

In unstressed positions (especially word-finally), /a/ may be reduced to [ə]. (31)

/parnta/

[pɐ́ɳʈə]

‘top’

2.2.5.2 Allophony of /i/ Typical realisations of /i/ range between [i] and [ɪ]. High front [i] is observed wordfinally or when preceded by a palatal (plosive or nasal) or alveolar lateral [l]. (32) a. b. c. d. e.

/jici/ /cipiɲ/ /lira/ /kinki/ /ɲilapunta/

[ɪ ́ci] [cípɪɲ] [lɪ ́ɾa] [kɪ ́nki] [ɲílapùnta]

‘gift’ ‘new, first time’ ‘mouth’ ‘sated’ ‘frog’

When preceding a retroflexed lateral /ɭ/, /i/ is heard as a somewhat more centralised allophone, approximately [ɨ] or [ᵻ], with accompanying ‘r-colouring’. (33) a. b.

/piʈiɭ/ /wiɭiɲi/

[pɪ ́ɖᵻɭ] [wí¯ɭiɲi]

‘strong, tough (of material)’ ‘hunting’

In the environment of a bilabial, unstressed /i/ can be realised as a backed and variably rounded vowel, approximatley [ʊ]. (34) /ciliwa/ [cɪ ́lɪwa], [cɪ ́lʊwa]

‘root’

2.2.5.3 Allophony of /u/ Phonemic /u/ is typically realised as [ʊ] in a wide range of segmental environments. Following the peripheral plosives /p, k/, it may have a realisation close to [u] but typically only in stressed syllables. (35) a. b. c. d. e.

/kuɲar/ /kurku/ /lucu/ /tuʈur/, /tuɻur/ /punta/

[kúɲar] [kúrkʊ] [lʊ́ɟʉ] [tʊ́ʈʊr] [tʊ́ɽʊr] [púnta]

‘dog’ ‘hole’ ‘ankle’ ‘descend’ ‘arranged spouse’

Following or preceding palatal consonants, /u/ may be fronted as far as [ʉ]. (36) a. b. c.

/cuʈul/ /cuwaʎ/ /cucu/

[cʉ́ʈʉl] ‘dust’ [cʉ́wɑʎ] ‘tall’ [cʉ́cʉ] ‘ceremony’

74 

 2 Phonology

The close-mid allophone [o] may occur after the semivowels (/w, ɻ/) and the tap/trill /r/ following /a/. Close-mid [o] may also occur preceding [w] if it follows a velar nasal [ŋ]. The semi-vowel [w] may or may not be perceptible in these contexts. (37) a. b. c. d.

/jawu/ /ŋuwa/ /jaɻu/ /laru/

[jɐ́o] [ŋóa], [ŋóʷa] [jɐ́ɻo] [láɾo]

‘yes’ ‘death adder’ ‘slowly’ ‘damp’

The close-mid allophone [o] has also been identified between velar [k] and a wordfinal trill [r]. (38) a. b.

/tiʎkur/ [tíʎkor] /kuŋkur/ [kúŋkor]

‘shoot’ ‘cold’

2.2.5.4 Allophony of long vowels Long vowels are mostly constrained to initial syllables of words (see §2.5.1). Only a limited number of allophones have been observed. For /aː/, both [aː] and [ɐː] are found. (39) a. b.

/caːʎ/ [cáːʎ] ‘desperate’ /palaːɳ/ [pɐ́lɐːɳ] ~ [pɐlɐ́ːɳ] ‘turtle’

For /iː/, only a centralised vowel [ɪː] has been observed. (40) /cɪːʎkara/

[cɪːʎkɐɾɐ]

‘branding’

For /u/, only a centralised vowel [ʊː] has been observed. (41) /puːɻa/

[pʊ́ːɻa]

‘yam’

2.3 Phonological processes Ngardi shows only a small number of phonological processes which modify the phonological structure within morphemes. These include rhotic metathesis (§2.3.1) and syllable deletion (§2.3.2). Some evidence is also presented for phonemic alternations (§2.3.3).

2.3 Phonological processes 

 75

2.3.1 Rhotic metathesis An optional rule of rhotic metathesis affects V1C2 in sequences /C1V1C2V2/ where C1 is a (retroflex) plosive, V1 is a high vowel [i] or [u] and C2 is [ɻ].54 The rule of metathesis swaps V1 and C2. Examples are (42) (a, b) and corresponding spectrograms are provided in Figure 14.

Figure 14: Spectrograms of (42) (a) and (b).

(42) a. b.

/paʈiɻi/ /kuʈuɻu/

→ →

[pɐɖɻiː] ‘teeth’ [kʊɖɻʊː] ‘fighting stick’

In articulatory terms, rhotic metathesis appears to be the result of early articulatory timing of a rhotic gesture. The retroflex approximant is phased sufficiently early that it overlaps (even precedes) the vowel to the point where it is perceived as a trilled or fricated release of the plosive. The remaining vowel is invariably longer than a monomoraic vowel which supports the analysis of metathesis rather than deletion. Similar processes have been observed in Wangkajunga (Pama-Nyungan) (Jones, 2011, p. 43).

2.3.2 Syllable deletion There are a small number of lexemes which exhibit variable deletion of syllables in sequences of different apical continuants /r/ and /ɻ/. Affected lexemes include the 54 This appears to be the sequence most conducive to metathesis. However, examples are also found where C1 was a bilabial plosive and V1 is /u/, e.g. /mumpari/ [mʊmbɻɐɪ].

76 

 2 Phonology

preverbs wirriri-55 ‘return, back’ and pirriri- ‘arrive’ as well as the verbal stem yurirri-. In both cases the latter syllable shows deletion as shown in (43). (43) Deletion of syllables in sequences involving rhotics in successive syllable onsets a. /wiriɻi=ja-/ ~ /wiri=ja-/ return=go b. /piriɻi=ja-/ ~ /piri=ja-/ arrive=go c. /juɻiri=ja-/ ~ /juri=ja-/56 move-around=go

2.3.3 Phonemic alternations There are a small number of alternations between two phonemes in certain words or morphemes. These phonemic alternations are not shared by all speakers and may be influenced by a speaker’s multilingual repertoire. The most robust alternation (in terms of the number of lexemes to which it applies) is an alternation between the retroflex approximant /ɻ/ and the retroflex plosive /ʈ/. This phonemic alternation occurs both word-initially and word-medially but not word-finally since /ɻ/ is an impermissible word-final segment (see §2.5.3.2). Similar phonemic alternations are reported in Jaru (Tsunoda, 1981b, pp. 31–32), Wanyjirra (Senge, 2015, pp. 75–76) and Bilinarra (Meakins & Nordlinger, 2014, pp. 54–55). Examples of lexemes exhibiting this alternation are provided in (44). (44) /ɻ/ ~ /ʈ/57 a. /kaɻija/ b. /tuɻur/ c. /kiɻa/

~ ~ ~

/kaʈija/ /tuʈur/ /kiʈa /

‘white man’ ‘descend’ ‘big, senior’

There are no other systematic phonemic alterations in the data. However, there are some additional lexically idiosyncratic alternations between phonemes worth men-

55 The preverb wirriri ‘back, return’ is further anomalous in that it exists in free variation with additional variants wirrirl [wɪɾɨɭ] and wirirr [wɪɾɪr ~ wɪɻɪr]. 56 In the case of the verb yurirri-, the reduced form has only been observed in combination with the present tense (i.e. yuri-rnanta) and showed retroflexion of the initial nasal of the expected present allomorph -nyanta. However, that this is an optional variant is confirmed by the presence of the expected form yurirri-nyanta in the corpus. 57 For those speakers who make use of flapped realisations of /ɻ/, the words in (44) may also be realised with intervocalic flaps, so, for example, [kɑɽija] for kardiya ‘white man’.

2.3 Phonological processes 

 77

tioning. The first is between a tap trill /r/ and the lateral /ʎ/ (45). Example (a) in (45) further shares another alternation between palatal and alveolar plosives /c/ ~ /t/ as shown in (46). (45) /ʎ/ ~ /r/ a. /ciʎk/ ~ /cirk/ ~ /tiʎk/ ‘drip’ b. /wurkal/ ~ /wuʎkal/ ‘green, month’ (46) /c/ ~ /t/ a. /ciʎkara/ ‘sting’ b. /cuŋkuc/ ‘set off’

~

/tiʎkara/

~

/cuŋkut/

In (47), alternations are presented between peripheral consonants /p/ ~ /k/. Three of the examples in (47) involve fossilised instance of the ‘simultaneous subordinating’ suffix -karra and could be analysed as metathesis. (47) /p/ ~ /k/ a. /carapici/ ‘tomato’ b. /kaʈiʈi/ ‘teeth’ c. /purkpara/ ‘split’ d. /turkpara/ ‘rub’ e. /ɻiripkara/ ‘rotate’

~

/carakici/

~

/partiɻi/

~

/purpkara/

~

/turpkara/

~

/ɻirikpara/

Lastly, (48) illustrates a lexically idiosyncratic alternation between the velar nasal and the velar plosive phonemes. (48) /k/ ~ /ŋ/ /kaɳʈar/ ~ /ŋaɳʈar/ ‘girl’

78 

 2 Phonology

2.4 Morphophonological processes There are only a very limited number of morphophonologically conditioned processes in Ngardi. These include lexically restricted instances of vowel assimilation (§2.4.1) and fortition (§2.4.2). Ngardi lacks the productive morphophonological processes of lenition, nasal-cluster dissimilation and vowel harmony found in various other Ngumpin-Yapa languages.

2.4.1 Vowel assimilation An optional rule of regressive vowel assimilation affects stem-final vowel /u/, fronting it to /i/ in the environment of a following palatal in just two morphological contexts. The first is preceding the object bound pronoun forms {=nyurra, =yanu, =nyanu}. Interestingly, of the possible stem-final /u/ vowels which may occur before {=nyurra, =yanu, =nyanu}, only the ergative allomorph -rlu and the final vowel of the 1pl.excl.s bound pronoun =rnalu have been observed consistently undergoing this change. (49) a. /=ɳalu=janu/

→ /=ɳali=janu/

(MMN: TEN1-2017_028-01: 3120791)

1pl.excl.s=3pl.o b. /kaɻija-rlu=janu/ → /kaɻija-ɭi=janu/ (PSM: TEN1-2016_005-03: 476853) white_man-erg=3pl.o c. /=ɳalu=ɲanu/ → /ɳali=ɲanu/ (YDN: LC24a: 370021) 1pl.excl=refl d. /=lu=janu/ → /lijanu/ (TJN: LC60a: 174464) 3pl.s=3pl.o The regressive changes illustrated in (49) are clearly not a widespread phonological rule in Ngardi since many other sequences involving /u/ in stem-final positions do not assimilate to following palatals. Ocassionally, the same vowel alternation will apply as a phonetic rule in connected speech across a different morphological boundary (e.g. the word boundary in (50)) but this appears less consistent than the assimilations in (49). (50) /muŋkiɳa-ɭu caci pi-ɲa . . . /→ [mʊŋkɪɳɑɭi cɐci pɪɲɐ . . . ] name-erg kangaroo hit-pst ‘Mungkirna (killed) a kangaroo.’ (YMN: LC23a: 582065) The second idiosyncratic regressive vowel assimilation rule operates on the /u/ vowel within the stems of three Class III verbs: yu- give, pu- hit and lu- cry. The root-final vowel exhibits regressive assimilation preceding a palatal-initial inflection; i.e.

2.5 Word and prosodic structure 

 79

pu-nganta hit-prs but pi-nya hit-pst. This rule is obligatory and is also paralleled by similar verbal paradigms in Warlpiri (Nash, 1986, p. 81). There are three verbal inflections which condition the root change (i.e three verbal inflections involve an initial palatal). These are the ‘past’ -nya /-ɲa/, ‘past narrative’ -nyangurra /-ɲaŋura/, and the ‘prior associated motion’ derivational suffix -nya(ni)nja- /-ɲa(ni)nca-/. The distribution of root vowels is not entirely regular, however, since a high vowel appears before the velar-initial irrealis suffix /-ŋi/, i.e. pi-ngi and yi-ngi rather than pu-ngi and yu-ngi.58

2.4.2 Fortition A restricted rule of fortition operates on initial palatal glides /j/ and replaces them with palatal plosives /c/ in the third plural bound pronouns /=janu/ and /=janu(ŋku)ɭa/ when preceded by /n/ (the only consonant that happens to precede these two bound pronouns). Compare the post-nasal plosive /c/ in (51) (b, c) with post-vocalic /j/ in (a). (51) a. b. c.

/ŋu=janu/ seq.aux=3pl.o /ŋu=n=canu/ seq.aux=2sg.s=3pl.o /ŋu=n=canuŋkurla/ seq.aux=2sg.s=3pl.lct

2.5 Word and prosodic structure Morphological constituents in Ngardi fall into one of two major classes: roots and suffixes. Ngardi does not have any prefixing morphology. The class of preverbs, which occur immediately preceding a verbal stem, is treated as a class of bound morphemes. Adverbial preverbs that may occur independent of the verb are treated as free, independent roots. The majority of roots fall into one of three major classes: verbal, preverbal (strong nexus, weak nexus and adverbial) and nominal. Of closed class lexical items, modal particles behave similarly to the nominal class in terms of their phonological wordhood, while the closed class of auxiliaries is anomalous in disobeying the disyllabic word mimimality constraint (§2.5.2). There is also a class of enclitics, which attach not to lexical roots but to complete (morphosyntactic) words. 58 This may be explained by these forms historically involving a cluster, i.e. piny-ngi; cf. other conjugations ngarn-ngi eat-inf and man-ngi get-irr. See §7.3 for a full discussion of verbal inflections.

80 

 2 Phonology

Morphological constituents in Ngardi do not have an entirely straightforward relationship to constituents in a prosodic hierarchy (Liberman & Prince, 1977; Selkirk, 1980). The relevant domain for stress assignment appears to be a fully inflected word (although stress assignment is sensitive to morphological structure). For simplicity, I will term this the ‘phonological word’.59 The phonological word in Ngardi is further characterised by a disyllabic minimum and the ability to occur independently (i.e. be separated by pauses). With the exception of interjections and auxiliaries, no phonologically well-formed word is monosyllabic. The various constraints on levels of prosodic structure are discussed in the following sections dealing with the syllable (2.5.1), the stem (§2.5.2) and positional constraints on types of segments (§2.5.3). Permissible consonant clusters are described in §2.5.4.

2.5.1 Syllable structure The basic syllable types in Ngardi are listed and exemplified in (52). (52) a. b. c.

CV /pa.ka/ ‘prickle’ CVC /kan.ka.ɳi/ ‘on top’ CVCC /ka.purk/ ‘dead’

Syllable types (i) and (ii) may occur in any word position. CVCC syllables have only been identified in word-final positions and generally only within the ‘preverb’ word class. There is a phonotactic constraint upon CVCC syllables such that the cluster is necessarily comprised of a liquid /ɭ, l, ʎ, r/ followed by a peripheral /k, ŋ, p/ (to the exclusion of /m/). Semivowels /w, j/ do not occur in any coda type. Onsetless syllables (53) are highly restricted in Ngardi (frequencies of word-initial vowels are presented in Table 19) and are only found word-initially. They are necessarily monomoraic and their vocalic nucleus is restricted to /a/ or /ai/. (53) a. b.

V /a.pi.nan/ VC /im.pa/

‘watermelon’ ‘alone’

59 Note that, in Warlpiri, there is clear evidence for distinguishing a ‘phonological word’ from a ‘prosodic word’ based on variable domains over which phonological processes occur (e.g. vowel harmony) as opposed to stress assignment (Pentland, 2004; Pentland & Laughren, 2004). Ngardi lacks any such phonological processes that exhibit clearly differential domains to prosodic processes and so I make no such distinction in the description here.

2.5 Word and prosodic structure 

 81

2.5.2 Stem structure While the majority of the possible syllable structures described in §2.5.1 are distributed amongst all types of morphological constituents in Ngardi, it is nevertheless necessary to specify morpheme sequence constraints specific to stems of different word classes. Three main groups can be identified based on minimal stem structure requirements as in Table 15. Since the subclasses of ‘preverbs’ are in fact conceived as idealised groupings of lexemes along a continuum of features (see §8.3.1.1), it is difficult to make precise statements about a class whose members are quite heterogenous. For example, most ‘weak nexus’ preverbs are consistent with group (ii) but a small number happen to have the stem structure condition of group (i). The groups illustrated in Table 15 also vary according to whether their members may appear as independent phonological words. This is discussed below. Table 15: Groups of word classes according to stem structure constraints. Stem word class

Examples

Minimal syllabic structure

i)

Verbal stems Auxiliaries Interjections

pu- hit ngu seq.aux yu ‘yes’

Minimally open monosyllables CV

ii)

Preverbs (strong nexus) Preverbs (weak nexus)

raa- ‘clean’ turt- ‘break’

Minimally heavy monosyllables CV, CVC, CVCC

iii)

Preverbs (weak nexus) Preverbs (adverbial) Nominal stems

jaku- ‘depart’ jakarr- ‘try’ minya ‘this’ kunyarr ‘dog’ kayirra ‘north’ parda dub

Minimally disyllabic CVCV

Modal particles

Stems in Group (i) share a monosyllabic stem minimum. Verb stems (a closed class of 81 lexemes) are minimally open monosyllables CV (e.g. pu- hit) – although the majority are open disyllabic stems CVCV (e.g. luwa- shoot). The few open trisyllabic stems CVCVCV (e.g. yapartu- run) have clearly resulted from the univerbation of a disyllabic preverb and monosyllabic verb (e.g. yapart=pu- ‘run=hit’).60 Verb stems cannot appear as independent words. The auxiliaries ngu, ma, ka and the small number of monosyllabic interjections (e.g. yu ‘yes’) are anomalous open monosyllabic stems as they may surface as independent words, violating the disyllabic word minimum in Ngardi.

60 See §8.4.4.4.2 for further discussion.

82 

 2 Phonology

Stems in Group (ii) share a monosyllabic word minimum involving a heavy syllable. Strong and (some) weak nexus preverb stems are other open syllables with long vowels, for example, raa- ‘clean’, or are closed syllables with simple or complex codas, e.g. jut- ‘set off’ or jarrk- ‘cross’. They cannot appear as uninflected independent words. Stems in Group (iii) share a disyllabic word minimum. Nominals, modal particles and adverbial and most weak nexus preverbs are all minimally disyllabic. Stems in Group (iii) are further characterised as being well-formed phonological words that may appear independently as uninflected words. In the following tables, I exemplify the permissible syllable structures arranged according to syllable count. Beginning with monosyllables, Table 16 presents the range of syllable shapes found in the corpus. Only verbs stems, interjections and preverbs allow monosyllabic stem types. Light, open syllables (CV) are found only in verb stems and interjections. All of the heavy monosyllabic stem types are necessarily strong or weak nexus preverbs. Adverbial preverbs are minimally disyllabic. Table 16: Permissible monosyllabic stem types. Type

Example

Gloss

Word class

CV

/pu-/

‘hit’

Verb

/ju/

‘yes’

Interjection

CVː

/caː-/

‘agape’

Preverb

CVC

CVCC CVːC CVCC

/luk-/

‘lie’

Preverb

/jup-/

‘dismount’

Preverb

/jur-/

‘motion_away’

Preverb

/ciʎk-/ /pirk-/

‘drip’ ‘throw’

Preverb Preverb

/caːʎ-/

‘jealous’

Preverb

/ciːʎ-/

‘burn’

Preverb

/jirk-/

‘name’

Preverb

Examples of the range of disyllabic structures found in word stems are presented in Table 17 (excluding disyllabic structures involving long vowels). Disyllabic stems are found in all stem classes (nominal, verbal, preverbal, interjections and particles). However, disyllabic stems with word-final complex codas (i.e. CVCVCC, CVCCVCC) are generally found only in the class of preverbs. For completeness, I include the additional syllable shapes generated by word-initial onsetless syllables.

2.5 Word and prosodic structure 

 83

Table 17: Permissible disyllabic stem types. Type

Example

Gloss

Word class

Type (onsetless σ’s)

Example

Gloss

Word class

CV.CV

/lu.wa/ /mi.ɲa/

‘shoot’ ‘this’

Verb Nominal

VCV

/i.ci/ /i.ta/

‘gift’ ‘small’

Nominal Nominal

/ka.la/

asrt

Particle

CV.CVC

/ca.kar/ /kaʈak/

‘try’ Preverb ‘billycan’ Nominal

VCVC

/i.kiʈ-/

‘rattle’

Preverb

CV.CVCC

/ka.rark/

‘bang Preverb together’

VCVCC

/i.ŋalk/

‘soup’

Nominal

/ŋa.cirk/

‘sneeze’

Preverb

CVC.CV

/cuɭ.pu/ /cam.pu-/

‘bird’ ‘kiss’

Nominal Verb

VC.CV

/im.pa/

‘alone’

Nominal

CVC.CVC

/kar.kur/ /kar.kaɲ/

‘utterly’ ‘chicken hawk’

Preverb Nominal

VC.CVC

/in.kil-/ /iŋ.kin/

‘reticent’ ‘tea leaves’

Preverb Nominal

Preverb

–

–

–

–

CVC.CVCC /kun.curk/ ‘cough’

Stems with long vowels (Table 18) are found in a range of word classes but are mostly found within the class of preverbs. Long vowels are almost entirely restricted to the initial syllable of a word with only very few exceptions. A similar distribution of long vowels is found in Warlpiri (Nash, 1986, p. 65). Lastly, long vowels do not occur in syllables with complex codas. Table 18: Permissible syllabic stem types involving long vowels. Type

Example

Gloss

Word class

CV:

/juː/ /raː-/

‘yes’ ‘clean’

Interjection Preverb

CVːC

/caː-/ /kiːr/

‘burn’ ‘brown-red’

Preverb Nominal

CV.CVː

/wa.raː/

‘wow’

Interjection

CVː.C.CV

/kaːɳ.ka

‘crow’

Nominal

CVːCCVC

/ŋaːk.man/

‘frog’

Nominal

CV.CVːC

/pa.ɭaːɳ/

‘turtle’

Nominal

CVːCVCV

/raː.kara/ /ŋa.ʎa.ʎa/

‘sunrise’ ‘pink cockatoo’

Nominal Nominal

CVː.C.CV.CV

/piːl.ka.ra/

‘bandicoot’

Nominal

84 

 2 Phonology

2.5.3 Positional constraints This section sets out the distributional constraints on consonants in Ngardi in terms of their occurrence in stem-initial (§2.5.3.1) and stem-final (§2.5.3.2) positions. Ngardi possesses a productive rule of epenthesis involving the syllable =pa which attaches to consonant-final stems in the environment of a following enclitic (§2.5.3.2.1). All consonants are permitted word-medially. 2.5.3.1 Stem-initial constraints Table 19 summarises the occurrence frequencies of all stem-initial phonemes (rounded to the nearest whole number) in a word list of 2613 entries. The distinction between apico-alveolar and apico-retroflex has been collapsed since the phonemic distinction is neutralised word-initially. The phonetic properties of this neutralisation are briefly described in §2.5.3.1.1. Table 19: Frequency of stem-initial phonemes.

Consonantal

Place

Phoneme

Frequency %

Peripherals

/k/ /p/ /ŋ/ /m/

16.4 12.6 8.6 10.7

Subtotal

48.3

/t, ʈ/ /n/ /l/ /r/

4.5 1.5 4.7 0

Subtotal

12.7

/c/ /ɲ /ʎ/

15.2 3.0 0

Subtotal

18.2

/j/ /w/ /ɻ/

7.3 12.1 2.0

Subtotal

19.4

/a/ /i/ /u/ /aː, iː, uː/ Subtotal

0.1 1.3 0 0 1.4

Apicals

Laminals

Non-Consonantal

Semi-vowels

Vowels

2.5 Word and prosodic structure 

 85

As illustrated in Table 19, not all phonemes occur word-initially in Ngardi. The lateral /ʎ/, the tap/trill /r/ and the high back vowel /u/ do not occur word-initially. Vowels /a, i/ only occur word-initially as part of onsetless syllables in a very small number of lexemes. The peripheral consonants are by far the most frequent place of articulation in word-initial positions. If the semivowel /w/ is added to the peripheral consonants, well over half (58%) of all words begin with a peripheral consonant. Apicals are comparatively rare word-initially, appearing in just over a tenth of the 2613 lexemes examined. 2.5.3.1.1 Apical neutralisation Based on my own impressions, neutralisation of the apical contrast word-initially is not ‘complete’, in the sense that both retroflex [ʈ] and apico-alveolar [t] phones occur. The majority of apical-initial words appear to be plain alveolars (i.e [t]), an impression consistent with descriptions for initial apicals in closely related languages Jaru and Wanyjirra. However, there are some lexemes which are clearly pronounced with an initial retroflex (irrespective of preceding segments) and this impression is further affirmed by occasional non-standard spellings by native speakers (initial orthographic ‘rt’). While the apical distinction is neutralised word-initially in all Ngumpin-Yapa languages, the description of the phonetic details of the neutralisation vary substantially, as illustrated in Table 20.61 Table 20: Description of the phonetics of word-initial apical neutralisation in Ngumpin-Yapa languages. Language62

Description of initial apicals

Jaru Walmajarri Warlpiri Bilinarra

Predominantly alveolar Retroflex Retroflex Retroflex but sometimes ‘halfway between [t] and [ʈ]’ Predominantly alveolar

Wanyjirra

While any convincing statements regarding the nature of initial apicals in Ngardi must await further instrumental work, it is worth pointing out that my perception of the majority of initial apicals as non-retroflex may have an articulatory explanation – since cues to consonantal retroflexion in Australian languages are often associated 61 Butcher (1995, p. 9) claims that across all Australian languages, the neutralisation is most commonly in favour of retroflex articulations. 62 References are: Jaru (Tsunoda, 1981b, pp. 36, 37), Walmajarri (Hudson & Richards, 1969, p. 175), Warlpiri (Nash, 1986, p. 70), Bilinarra (Meakins & Nordlinger, 2014, p. 60) and Wanyjirra (Senge, 2015, p. 80).

86 

 2 Phonology

with the formants in preceding vowels. Acoustically, while both apical articulations involve similar F2 attributes due to similar positioning of the tongue body, retroflexed articulations involve a falling F3 contour which corresponds to the movement of the tongue apex anteriorly towards the post-alveolar region.63 This effect is most pronounced if the preceding vowel is a low vowel. This is the property that is often referred to in impressionistic descriptions as the ‘r-colouring’ of preceding vowels (e.g. Tsunoda (1981b, p. 24) for Jaru). In contexts where there are no preceding vowels, the audible cues to the contrast are not able to be maintained. Suggestive evidence for cueing of retroflexion based on the preceding vowel can be adduced indirectly from phonotactics. The frequency of retroflex plosives following vowels of different qualities in different lexemes is presented in Table 21. Table 21: Frequency of retroflex plosives in different post-vocalic environments. Preceding V

a__

u__

i__

Lexical count % of tokens

41 51.9

21 26.6

17 21.5

Table 21 shows the retroflex consonants are twice as common following the low /a/ vowel than following either of the high vowels.64 The perceptual salience of retroflexion when in the environment of a preceding [a] has been noted in other Australian languages (see Hamilton, 1996, p. 49) and relates to the fact that low vowels have the least coarticulation with a retroflex apical tongue movement and that durations of [a] are longer on average than [u] or [i]. 2.5.3.2 Stem-final constraints Table 22 presents frequencies of permissible stem-final consonant clusters to the nearest tenth of a percent. All phonemes are permitted stem-finally except for the semi-vowels /w, j, ɻ/. The overwhelming majority of stems are nevertheless vowelfinal. The peripherals are very infrequent stem-finally, occurring in just 2.8 percent of the 2613 lexemes examined.

63 There are only very few cases in Australian languages where retroflexion has been shown to be cued by acoustic properties within the CV transition. See McGregor (1990, pp. 70–71) on Gooniyandi, Platt (1972, p. 7) on Gugada, and Blake (1979b, p. 190) on Pitta-Pitta. 64 Similarly in Warlpiri, retroflexion has been described as ‘most pronounced following [a], not as strong following [u] and ‘almost imperceptible’ following [i]; while in Walmajarri the alveolar-retroflex distinction is argued to be entirely neutralised preceding [i]’ (Hudson & Richards, 1969, p. 175).

2.5 Word and prosodic structure 

 87

Table 22: Frequency of root-final consonant phonemes. Manner

Place

Phoneme

Frequency (%)

Consonantal

Peripherals

k p ŋ m

2.0 0.3 0.3 0.2

Subtotal

2.8

t n l ʎ r

1.0 3.5 1.3 2.3 5.6

Subtotal

13.7

c ɲ

1.3 5.0

Subtotal

6.3

j, w, ɻ

0

Subtotal

0

a i

29.1 25.1

u aa ii, uu

19.3 0.1 Oblique > Object (see Tsunoda, 1981b) In Ngardi, as in Jaru (Tsunoda, 1981b, p. 131) and Wanyjirra (Senge, 2015, p. 338), these formal hierarchies are operative in at least three aspects of the morphology of bound pronouns: – rules of dual number replacement (§6.4.4.1) – rules of linear ordering (§6.4.4.2) – rules of cross-referencing for non-subject arguments (§6.4.4.3). 6.4.4.1 Dual number replacement A major constraint affecting pronominal sequencing is the impermissibility of two dual number clitics (1du.inc, 1du.excl, 2du, 3du) occurring within a single clitic complex. In these contexts, one member is replaced by a corresponding plural number – referred to here as ‘dual number replacement’.299 In Ngardi, the choice of which dual member is replaced is governed by a ranking of persons (1 > 2 > 3). In a given combination the higher ranking person remains in its dual form, while the lower ranking person is replaced by the corresponding plural. The relevant resulting forms were highlighted in Table 75. The following examples further illustrate dual number replacement, governed by the person hierarchy.

299 A constraint against the surfacing of two ‘dual’ (or two unit augmented) clitics is found in all Ngumpin-Yapa languages and is reported as a widespread feature of Australian languages (Blake, 1987, p. 105). Within the Ngumpin-Yapa group there is variation in the specifics of replacement. The facts described above for Ngardi are very similar to dual number replacement found in Western Warlpiri (Hale, 1973, p. 330) and the Western Ngumpin languages Jaru (Tsunoda, 1981b, p. 133–134) and Wanyjirra (Senge, 2015, pp. 341–346). This is in contrast to the Eastern Ngumpin languages Bilinarra, Gurindji and Mudburra (as well as Eastern Warlpiri) that possess a system in which any complex containing two non-singular/non-minimal subject and object forms is modified such that both forms are replaced with plural forms. In these languages the process has been referred to as ‘dual neutralisation’ (McConvell, 1980, p. 53).

334 

[i] (809)

 6 Bound pronouns

(1 > 2) Mapirri=ma-ni=rna=ngkupula, marda-rnanta=rlijarra=nyurra. both=get-pst=1sg.s=2du.o have-prs=1du.excl=2pl.o ‘I got you two together in the same place and we two are looking after you two.’ (NNA: ngurungka 30_4_91)

[ii] (1[inc] > 3) (810) Ngali-rlu ngali=rli=yanu kawarli=ma-ni 1du.incl-erg 1du.incl=1du.incl.s=3pl.o absent=get-pst ‘You and I, we lost two.’ (TJL: TT76_2601: 1445007_1449500)

kuriny. two

[iii] (1[exc] > 3) (811) Yi-nya=rli=yanu ngalijarra-rlu nyanung-kujarra-ku. give-pst=1du.incl.s=3pl.o 1du.incl-erg recog1-dual-dat ‘You and I gave (the dog) to those two.’ (PSM: TEN1-2017_028-01: 2695551_2698152) (812) Marda-rnanta=rlijarra=yanu. have-prs=1pl.excl.s=3pl.o ‘We two are looking after those two.’ (MMJ: LC31a: 662967_666519) [iv] (2 > 3) (813) Ngantu-kujarra-ku=npula=yanu warri-rnanta nyunpula-rlu? who-dual-dat=2du.s=3pl.o search-prs 2du-erg ‘Which two are you two looking for?’ (PRN: TG1_0034-01) The replacement of object duals in (809)–(813) can be compared with the replacement of subject duals in (814) and (815), again, in adherence to the person hierarchy. Note that (814) also involves an instance of the rare process of clitic copying (§6.2.3). [i] (1[incl] > 3) (814) Yi-nya=lu=ngalingku=lu kunyarr ngalijarra-ku. give-pst=3pl.s=1du.incl.o=3pl.s dog 1du.incl-dat ‘Those two gave the dog to us two.’ (PSM: TEN1-2017_028-01: 2826390_2828369) [ii] (1[exc] > 2) (815) Ngajarra=nta=ngajarra nyin=ma-ni. 1du.excl=2pl.s=1du.excl.o lose=get-pst ‘You two left/forgot him and me.’ (TJL: TT76_2601: 1587067_1589191)

6.4 Complex bound pronouns 

 335

While a number of examples provided above happen to involve clauses that contain additional nominal elements (namely free pronouns) which overtly specify the number of referents, this is not always the case (e.g. (815)). Utterances that involve dual number replacement and lack disambiguating free pronouns are generally pragmatically interpreted on the basis of preceding discourse or other contextual factors. 6.4.4.2 Linear ordering The linear ordering of individual bound pronouns within the pronominal complex is primarily sensitive to the person hierarchy. The complete set of linear orderings was presented in the tables of complex bound pronouns in §6.4.1 and §6.4.2. The following examples illustrate the widespread conformity to the person hierarchy by sampling a range of person/number combinations in complex bound pronouns: [i] 1 > 2: first person enclitics are sequenced before second person enclitics (816) Purda=nya-nganta=yi=n? listen=see-prs=1sg.o=2sg.s ‘Do you hear me?’ (Cataldi, 2011, p. 23) (817) Pu-ngku=rna=ngku kuja=n wangka-nyanta jayanta. hit-pot=1sg.s=2sg.o sub=2sg.s speak-prs always ‘I will hit you if you keep talking.’ (MMN: TEN1-2017_002-05: 1455406_1466756) (818) Ngaka=rna=ngkupula nya-ngku nyunpula. later=1sg.s=2du.o see-pot 2du ‘I’ll see you two later.’ (PRN: TG1_0034) (819) Nyuntu=ngajarra=n yi-nya kuyi. 2sg=1du.excl.o=2sg.s give-pst meat ‘You gave the meat to us two.’ (PSM: TEN1-2017_024-01: 3510435_3513367) (820) Wanji-rla=rna=nyurra jingka-ngampurr ka-ngku. where-loc=1sg.s=2pl.o child-paucal carry-pot ‘Where will I take you group of children?’ (MRN: LC35_b: 807139_809496) [ii] 1 > 3: first person enclitics are sequenced before third person enclitics (821) Yi-nya=rna=yanu kuyu. give-pst=1sg.s=3pl.o meat ‘I gave the meat to them.’ (MMN: TEN1-2017_001-01: 143842_146471)

336 

 6 Bound pronouns

(822) Jakarr nga=yi=lu pina=yu-ngani. attempt seq.aux=1sg.o=3pl.s teach=give-ipfv.pst ‘They were trying to teach me (Ngardi).’ (THS: TEN1-2017_003-05: 196730_202424) (823) Ma-ni=ngajarrangu=lu. get-pst=1du.excl.o=3pl.s ‘They came and got us both.’ (BSE: TEN1-2016_014-01: 420754_422751) [iii] 2 > 3: second person enclitics are sequenced before third person enclitics (824) Nyampa=n=janu yala=yi marda-rnanta karlarra? what=2sg.s=3pl.o that=emph have-prs west ‘What ones do you have there in the west?’ (LC56b: 330015_333423) (825) Ngu=ngku=lu nya-nganta. seq.aux=2sg.o=3pl.s see-prs ‘They are looking at you.’ (PJA: TT76_2601: 1212686_1214247) The ordering principles discussed here only affect the ordering of elements in the subject and object/oblique slots. The locational formative =rla which occupies the final position in the pronominal complex is never affected. Thus, compare (825) with (826): (826) Nyamu=lku=ngku=lu=rla karri-nyani. finish=then=2sg[lct]=3pl.s=lct be-ipfv.pst ‘That’s enough of them staying with you./They have been staying with you for long enough.’ (PSM: TEN1-2017_024-01: 2054548_2057828) Similarly, compare (816) with (827). (827) Nyuntu=yi=n=ku=rla yut=karri-nyanta ngaju-ngka. 2sg=1sg[lct]=2sg.s=ep=lct sit=be-prs 1sg-loc ‘You are sitting with me.’ (MMN: TEN1-2017_024-02: 890474_894050) A number of complex bound pronoun sequences nevertheless violate the person hierarchy, and instead follow the hierarchy of grammatical relations (828)–(831). Complex pronouns that are ordered according to this hierarchy predominantly involve a number of complexes involving the third plural subject forms =lu; the major exceptions being combinations of 3pl.s with 1sg.o (yi=lu) or 2sg.s (ngku=lu).300 300 Interestingly, in some of the Eastern Ngumpin languages – namely Gurindji (McConvell, 1980) and Bilinarra (Meakins & Nordlinger, 2014) – the hierarchy of grammatical relations appears to be the primary rule for clitic ordering, with the person hierarchy second.

6.4 Complex bound pronouns 

 337

(828) Wangka-nya=lu=ngajarra, kuyi ma=rnalu kupa-rnanta.301 speak-pst=3pl.s=1du.excl.o meat top.aux=1pl.excl.s cook-prs ‘They said to us, we are cooking the meat.’ (YMN: LC23a: 1280959_1285278) (829) Kulirra-ngulu yalu-rlu wirrpa-ngku, =lu=ngalypa nya-nganta. south-abl that-erg many-erg =3pl.s=1pl.incl.s see-prs ‘They (Sturt Creek mob) are looking at us from the south.’ (TJL: TT76_2601: 1500340_1505593) (830) Wurna=lku kunyku=lu=nganpa yu-ngani-nyirra. journey=then itchy=3pl.s=1pl.excl.o give-ipfv.pst-narr ‘For the journey, they gave us these itchy things to wear.’ (Cataldi, 2011, P. 23) (831) Wanji-rla=lu=nyurra ngatina and mama? where-loc=3pl.s=2pl.o M and F ‘Where are your mothers and fathers?’ (Cataldi, 2011, p. 24) In the case of two third person forms, the subject enclitic precedes a non-subject enclitic: (832) Nanikut ngu=lu=yanu ka-nya. goat seq.aux=3pl.s=3pl.o carry-pst ‘They brought the goats (as well).’ (MMN: TEN1-2017_005-02: 292112_294701) (833) Yupa-rnanta=lu=yanurla wirrpa-ngka. send-prs=3pl.s=3pl.lct many-loc ‘They are sending it with all of them.’ (PSM: TEN1-2017_021-03: 1995332_1998873) A single instance of a second singular subject enclitic preceding a first person object enclitic is found: n=ku=ngalingku. However, this was only recorded in elicitation and was described as ‘old Ngardi’, see (718). 6.4.4.3 Non-subject cross-referencing As described in §6.4.3, there are strict constraints on the possibilities of three-member pronominal clitic complexes. The third and final member in the sequence can only be the third singular oblique =rla. Therefore, in contexts where more than two nonsubject arguments are competing for registration in the clitic complex, at least one

301 Note that Bonnie Seela used an alternative ordering of 1du.excl.o>3pl.s (ngajarrangu=lu) (see Table 75 and (823)).

338 

 6 Bound pronouns

argument is excluded. The choice of which non-subject argument will be excluded is determined primarily by the hierarchy of grammatical relations, as follows. [i] Subjuncts outrank objects (834) Nyuntu-ngka=ngkurla yirra-ku (ngantany-ju) (ngaju). 2sg-loc=2sg.lct place-pot man-erg 1sg ‘That man will put me next to you.’ (PSM: TEN1-2017_024-02: 1921943_1923934) (835) Ka-nganta=rna=yanungkurla yalu-kurra ngantany-kurra, wirrpa-kurra. carry-prs=1sg.s=3pl.lct that-all man-all many-all ‘I will take (them) over to that big mob.’ (PSM: TEN1-2017_024-01: 2304517_2307948) [ii] Obliques outrank objects (836) Kuriny=pa=rlangu=rna=nyanu pu-ngku. two=ep=dyad=1sg.s=refl hit-pot I will kill two (goannas) for myself. (TJL: TT76_2201: 1219187_1220783) However, in the case of subjuncts in competition with object arguments, the person hierarchy again is relevant. Thus, a second person (direct object) outranks a third person (subjunct), as in (837). (837) Ngu=rna=ngku yirra-rnanta mula ngantany-ja wurrkaru-rla. seq.aux=1sg.s=2sg.o place-prs this man-loc three-loc ‘I am leaving you with these three men.’ (PSM: TEN1-2017_024-01: 1683411_1692809) By the same token, a second person locational outranks a third person object, as in (838). (838) Ngu=rna=ngkurla yirra-rnanta nyuntu-ngka jingka wurrkaru seq.aux=1sg.s=2sg.lct place-prs 2sg-loc child three ngu=n=janu ngampurr=marda-rnanku. seq.aux=2sg.s=3pl.o watch_out=have-ipfv.pot ‘I am leaving these three children with you, so that you can look after them.’ (MMN: TEN1-2017_024-01: 1772544_1808053) In conclusion, three morphological properties of complex bound pronouns can be accounted for with reference to person and grammatical relation hierarchies: i) dual number replacement ii) linear ordering of bound pronouns iii) cross-referencing of non-subject arguments.

6.4 Complex bound pronouns 

 339

6.4.5 Metathesis In Ngardi – as in other Ngumpin-Yapa languages – certain subject and object bound pronouns surface discontinuously in certain combinations. I follow descriptions of related languages which analyse this as a process of ‘metathesis’ of person and number (Hale, 1973, pp. 328–329; Tsunoda, 1981b, p. 135). Two types of metathesis can be observed: i) discontinuous number in subject pronominals: 1pl.excl.s =rna . . . =lu; 2du.s =n . . . =pula ii) discontinuous number in an object pronominal: 2du.o =ngku . . . =pula. In the case of (i), the subject enclitics =rnalu and =npula can be split if they co-occur with certain second or third person object/oblique enclitics. Examples are (839)–(841). (839) Nyuntu=rna=ngku=lu itaki=ma-nanta. 2sg=1excl.s=2sg.o=pl.s blame=get-prs ‘We are blaming you.’ (MDN: LC48b) (840) Ngu=rna=nyurrangku=lu warri-rnanta nyurrurla-ku, wirrpa-ku. seq.aux=1excl.s=2pl.o=pl.s search-prs 2pl-dat many-dat ‘We mob were looking for the lot of you.’ (PSM: TEN1-2017_028-01: 1880133_1885114) (841) Ngu=rna=pulany=ku=lu (warri-rnani). seq.aux=1excl.s=3du.o=ep=pl.s search-ipfv.pst ‘We mob were looking for those two.’ (PSM: TEN1-2017_028-01: 1805300_1808295) In the case of (ii), =ngkupula can be split by a third person subject bound pronoun, as in (842). (842) Wali, pu-ngku=ngku=lu=pula kuyi kila. well, hit-pot=2.o=3pl.s=du.o meat killer ‘They will kill an animal (a ‘killer’) for you two.’ (TJN: tjama 23-10-00) Example (843) demonstrates the splitting of person and number of both subject and object enclitics. (843) Nyunpula-ku ngu=rna=ngku=lu=pula warri-rnani wirrpa-ngku. 2du-dat seq.aux=1excl.s=2.o=pl.s=du.o search-ipfv mob-erg ‘A big group of us were looking around for you two.’ (PSM: TEN1-2017_028-01: 1797374_1801838)

340 

 6 Bound pronouns

The ability to split =rnalu and =ngkupula can be contrasted with =rlijarra (844), which is never split into putative person (=rli) and number (=jarra) morphemes, even in equivalent contexts. (844) Yi-nya=rlijarra=ngku kunyarr nyuntu-ku. give-pst=1du.excl.s=2sg.o dog 2sg-dat ‘He and I gave the dog to you.’ (PSM: TEN1-2017_028-01: 2548605_2550399) In Jaru, Tsunoda (1981b, 135ff.) proposes an additional rule of metathesis involving locational bound pronouns. This is necessary under his analysis, since he analyses locationals as rigid monomorphemic pronominal forms. However, I have proposed that =rla in Ngardi is always positioned in the final slot of the pronominal complex (Hudson, 1978, pp. 62–63) and indexes a locational bound pronoun compositionally with an object element – wherever it surfaces in the pronominal complex. Effectively, ‘locationals’ are only rendered discontinuous by: i) non-subject > object ordering and ii) the rules of metathesis already described above. Thus, the discontinuous expression of the locational bound pronoun in (845) is already accounted for by the rule of metathesis that splits subject/object person and number morphs. (845) Nganimpa=rna=ngku=lu=rla yut=karri-nyanta nyuntu-ngka warririku. 1pl.excl=1.s=2sg[lct]=pl.s=lct sit=be-prs 2sg-loc close ‘We are sitting close to you.’ (MMN: TEN1-2017_024-01: 1216181_1219565)

6.4.6 Epenthesis The epenthetic syllable =ku is found breaking up impermissible consonant sequences within certain complex bound pronouns. Specifically, it is inserted to avoid impermissible sequences of nasal + lateral phonemes. These sequences arise in: i) combinations of certain subject + 3sg.obl forms (e.g. (329), (822)), ii) combinations of subject and locationals (531), and iii) the compositional formation of locationals themselves (e.g. (773)). Epenthetic =ku insertion is applied following the rules of metathesis described in the preceding section and resolves any impermissible nasal + lateral phoneme sequences that arise from the postposing of subject number (e.g. (841)). Finally, there is one restricted usage of =ku separating a sequence of two nasals (718), but this form was only reported during elicitation and was described by Marie Mudgedell as ‘old Ngardi’.

7 Verbal morphology This chapter details the verbal system as it relates to verb roots, conjugations, as well as the forms and functions of the various verbal inflections. The class of ‘preverbs’ and the formation of complex verb constructions (CVCs) are also introduced here, although discussion of complex predication is greatly expanded upon in Chapter 8. The verbal inflectional system comprises a large number of inflectional suffixes that express various distinctions in tense, aspect, mood, directionality and associated motion. Verbal inflections show some limited interaction with auxiliary-like morphemes exhibiting clausal scope. These are restricted to the domain of modality and are discussed in detail in §10.2. ‘Verbs’ in Ngardi are a closed class of words that inflect for tense, aspect, mood, directionality and associated motion. The verbal inventory in Ngardi is restricted to 81 lexical roots. The semantic range of this inventory is augmented via the formation of a much larger number of multimorphemic predicates, known as ‘complex verb constructions’ (McGregor, 2002). Complex verb constructions are multimorphemic predicates, comprising either preverbs (846) or nominals (847) in combination with an inflected verb. (846) Ngantany-ju jingka purda=nya-nganta. man-erg child listen=see-prs ‘A man hears a child.’ (MMN: TEN1-2018_02-01: 758979_768592) (847) Nyuntu-ku=rna=ngku jina=marda-rnanta yut-warnu-rlu. 2sg-dat=1sg.s=2sg.o foot=have-prs sit-assoc2-erg ‘I am waiting for you while sitting down.’ (PSM: TEN1-2018_027-01: 2212757_2217355) CVCs may involve more than just a single preverb but there are restrictions on the kinds of preverbs which may combine in this fashion. Not all inflecting verbs exhibit the same productivity in the formation of CVCs. Some verbs may form many different CVCs in combination with different preverbs and are termed ‘productive’, while others do not readily co-occur with many preverbs and are termed ‘non-productive’. With respect to preverbs, there are also a range of preverb types which can be identified based on their degree of phonological, morphosyntactic and semantic integration with productive inflecting verbs. Additionally, nominals and borrowed Kriol verbs are observed combining with certain productive inflecting verbs to form CVCs. The verb itself may either host tense, aspect or mood suffixes – thus creating a finite inflected verb – or it may host a non-finite suffix deriving a non-finite verb stem. A special type of inflecting verb involves a lexical verb root inflected with one of two associated motion suffixes: one indicating a prior motion event, -ninja ‘move&x’, and another indicating a subsequent motion event, -nungka ‘x&move’. Both associated https://doi.org/10.1515/9783110752434-007

342 

 7 Verbal morphology

motion suffixes show evidence of being etymologically derived from full lexical verbs. The resulting verbs with complex event semantics host regular TAM suffixes as per other lexical verbs. Thus, compare the simple event semantics of the underived verb in (848) with the complex associated motion event semantics of the derived verb in (849). (848) Ma-nku=rna mangarri, jirri-ku. get-pot=1sg.s veg_food later-dat ‘I will get my food for later on.’ (MMN: TEN1-2018_004-02: 1621665_1623052) (849) Ma-ninja-nku=rna mangarri wirrpa. get-move&x-pot=1sg.s veg_food many ‘I will go and get lots of food.’ (MMN: TEN1-2017_002-04: 774899_779537) Non-finite verb forms are syntactically distinct from finite inflected verb forms due to their role in the formation of non-finite subordinate clauses through the attachment of case in subordinating function, as in (850). Non-finite verb forms are only briefly described in this chapter. They are further analysed in detail in the discussion of non-finite subordinate clauses (§11.3). (850) Minya kuyi jalangu-ku nga-rnu-ku. this meat now-dat eat-inf-dat ‘This meat is for eating today.’ (MMN: TEN1-2018_017-03: 103577_107438) The verbal inflectional system is divided into two partially distinct TAM paradigms. One paradigm, termed ‘regular’ (§7.4), involves a wide array of TAM inflectional suffixes and is generally found combining with all verbs. A second, termed ‘directional’ (§7.5), is characterised by the presence of a directional morpheme indexing a directionality feature, either ‘toward’ (-rni hith) or ‘away from’ (-rra hence) the speaker. Significantly, these directional forms interact idiosyncratically with TAM inflections and a subset of them are best treated as TAM-directional portmanteaux. These ‘directional inflections’ are further defective in that they have only been found combining with a subset of lexical verbs. This chapter is organised as follows. The verbal inventory along with membership of lexical roots into conjugation classes is described in §7.2. In §7.3 I overview the verbal inflections and motivate the decision to distinguish the ‘regular’ and ‘directional’ verbal paradigms. Sections §7.4 and §7.5 describe the basic functions of each verbal inflection within the regular and directional paradigms, respectively. The broader discussion of how various verbal inflections contribute to the co-construction of clause level modality is picked up on in §10.2. Within §7.6, some of the key similarities and differences between Ngardi and its neighbouring languages with respect to the verbal inflection system are noted. Some isolated examples of an otherwise non-productive pattern of verbal reduplication are described in §7.7.

7.2 Inflecting verb inventory 

 343

7.1 Overview of inflecting verbs Before proceeding with a discussion of the inflecting verb inventory in Ngardi, two preliminary comments are in order. First, with respect to verbal transitivity, Ngardi verbs are not fully described by a simple dichotomy of ‘transitive’ or ‘intransitive’ but rather require further sub-categorisation according to their valency and argument structures. This topic is not discussed until Chapter 8 but of significance here is that a single verb – or indeed a single complex verb – is not necessarily restricted to a single argument structure but may have multiple, alternative case frames. Second, with respect to the semantic analysis of verb roots, the single-word glosses provided are, by necessity, a gross simplification of the range of possible meanings expressed by the verb, and, in some cases, the gloss may fail to encompass certain meanings generated in a particular complex verb construction. To mitigate this, the reader is directed to §8.4 where the full range of semantics associated with each of the productive verbs is described in greater detail.

7.2 Inflecting verb inventory The inventory of inflecting verbs is presented in Table 80. Each of the lexical verb roots here was checked with Marie Mudgedell and Patrick Smith during 2016–2019 fieldwork. A small number of forms originally identified by Lee Cataldi in her corpus were not recognised by Marie Mudgedell and Patrick Smith and these forms are represented by an asterisk ‘*’ in Table 80. Additionally, there are a handful of forms listed by Lee Cataldi as inflecting verbs which I have not analysed as monomorphemic inflecting verbs for one of two reasons: i) they are analysable as preverb-verb combinations involving the productive verb =ma/=pa- get (e.g. walmakpa-, lurrijpa-, partpa-, kitpa-, nyitpa-, etc.) ii) they are analysable as various other preverb-verb combinations (e.g. parda-la ‘appear=pierce’). The assignation of verb roots to their respective conjugation classes is lexically determined. Unlike other Australian languages, there is no substantive correlation between conjugation class and valency (Dixon, 1980, 2002, p. 233, Koch, 1980, p. 157).302 One striking feature of the verbal inventory and conjugation class membership is the absence of highly irregular verbs. The only deviation from the class memberships as represented in Table 80 is the presence of an unexpected -lku potential allomorph on an otherwise regular class II verb jarnti- incise (instead of expected -ku). It is possible that this verb may belong to Class IV, however, there is insufficient data to make such 302 Within the two largest classes, class I has a majority of monovalent verbs while class II has a majority of polyvalent verbs.

ingkijama-

stay take laugh disbelieve be gaze laugh die smell embark glow speak fall whistle dance move

jarrantijarrijingkirdikapatikarrikulkurringarlarripaliparntiparditiripitiwangkawantiwinpirriyawarratiyurirrilick incise raise show tease traverse pour crush throw cut excrete cook rub shoot refine rub

light grind

ngurrjanyinju-/ nyunjunyipinyuri-* pajapalkapangiparntapayipilkapulapuntapurrjutanpa-* tirrkawakuwaljiwanjasqueeze knead bite poke DIG feel ask expose CALL gather blow accompany love slander pluck leave

choose kiss

III (ng)

nyaparapi-, puyapartuyi-, yu-

kali-, lusee follow hit run give

carry cry

IV (ny) ngaEAT

V (n)

ya-

lama-, pa- 303

(continued)

go

pierce get

303 The variant pa- only occurs in complex verb constructions, where it appears in free variation with =ma- (primarily in contexts with a causative function, see §8.4.1.2.2). Cataldi (2011, p. 246) glosses -pa- separately as an ‘inchoative’.

jampujarntijarralajirrijitijungkanyi-* junikanyikijikumakunakupalurrpaluwamajamapa-

II (Ø)

bring heat

I (y)

irrkijanka-

Table 80: Inventory of inflecting verbs.

344   7 Verbal morphology

I (y)

Table 80 (continued)

mardangajanganpingarringardingirrjan(g)irrjirringuka-

II (Ø) have void winnow tell sharpen huddle deafen swallow

warriwarrka wayiyarliyirapiyirrayunpayupasearch crawl bind wash hunt place sing send

III (ng)

IV (ny)

V (n)

7.2 Inflecting verb inventory   345

346 

 7 Verbal morphology

a determination. Its cognate form in Warlpiri is a ‘Ø’ class member (corresponding to Class II in Ngardi). Lastly, there are a small number of lexical verb roots contained in Table 80 which are homophonous with distinct preverbs. These lexical roots are discussed in §8.3.1.3.

7.2.1 A continuum of inflecting verb types Due to the combinatorial properties of inflecting verbs and their ability to enter into complex verb constructions where their semantic contribution to a predicate’s meaning may be very weak or non-existent, some studies of typologically similar verb systems choose to treat the inflecting verbal element as a type of ‘light verb’ (cf. Butt, 2010). Ngardi – like Warlpiri – possesses a range of verb types that can be said to exist along a continuum from semantically ‘light’ verbs to semantically ‘heavy’ verbs (Laughren, 2010). Verbs that can be characterised by their semantic contribution in this manner can also generally be characterised for their degree of syntactic productivity. Semantically heavy verbs are generally non-productive while semantically light verbs are generally productive (Laughren, 2010, p. 177). The ability of a single verb to appear both as an independent ‘simple verb’ (with its associated fully realised semantic content) and yet also appear in CVCs with somewhat ‘bleached’ semantic content, is variously analysed as cases of ‘homophony’ (Laughren, 2010) or treated as a polyfunctional feature of verbs (Meakins & Nordlinger, 2014). A subset of Ngardi verbs exhibits the same ‘homophony’ or ‘polyfunctionality’. The examples in (851) illustrate how a single lexical form can exhibit both ‘simple’ and CVC-usages across different valency types. (851) a. wantib. pu- ~ pic. yu- ~ yi-

fall (x) hit (x, y) give (x, y, z)

a′ pukany=wanti- sleep (x) b′ raly=pu- ~ pi- clean (x, y) c′ pina=yu- ~ yi- show/teach (x, y, z)

Analysts have chosen various different labels for instances of verbs in the pairs represented in {a, a′; b, b′; c, c′}. I will refer to verbs which appear in isolation as ‘simple verbs’ (or just ‘verbs’). Verbs appearing within complex verb constructions will be referred to transparently as ‘verbs in CVCs’. I treat formally identical verbs which may appear in both contexts, as exemplified in (851), as a single lexical verb. These verbs are grouped together as ‘productive inflecting verbs’ and are discussed in detail in §8.4. A significant number of Ngardi verbs do not exhibit the same degree of ‘polyfunctionality’ as schematised in (851). Instead, they appear to operate as simple verbs and only combine with adverbial preverbs when entering into CVCs. For example, yunpa- sing (x) and nyuri- knead (x, y) both appear to be best categorised as exclusively simple verbs: they are only found combining with adverbial preverbs and they

 347

7.2 Inflecting verb inventory 

do not exhibit any significant modification of their core semantics. I do not analyse any distinct ‘aspectual verbs’ as described for Warlpiri by Laughren (2010) (e.g. -ya as a path verb); however, this type of analysis of a highly grammaticalised ‘verb’ could be extended to the ‘associated motion’ stem-forming suffixes detailed in §7.5.3.

7.2.2 Ngardi verbal inventory in cross-linguistic perspective Ngumpin-Yapa languages generally have a small, closed class of inflecting verbs in the order of 30 to 40 lexical items. The Eastern members of the subgroup (e.g. Gurindji, Bilinarra, Mudburra) have the smallest inventories (Meakins et al., 2013 – around 30 to 35 – while the Western members (e.g. Wanyjirra, Jaru and Walmajarri) have around 40 (see Hudson, 1978, p. 43; Senge, 2015, p. 113; Tsunoda, 1981b, p. 76) as does the Yapa language Warlmanpa (Browne, 2021). Both Ngardi and Warlpiri (Nash, 2008) are notable exceptions with verb inventories of 81 and 130 respectively. Reported tallies of lexical verbs in Ngumpin-Yapa languages are presented in Table 81. Table 81: Sizes of inflecting verb inventory in select Ngumpin-Yapa languages. Yapa # of verbs

Western Ngumpin

Warlpiri

Warlmanpa

Ngardi

130

45

81

Eastern Ngumpin

Walmajarri Jaru 40

40

Gurindji

Bilinarra

Mudburra

33

23

34

The verb count for Ngardi could be somewhat altered (upwards) depending on how one chooses to analyse a number of three-syllable verb forms. Under the present analysis there are a number of identifiable preverb-verb combinations which are reduced by morphophonemic changes. Consider for example certain preverb-verb combinations involving wanti- fall in (852). (852) Morphophonemic realisation of CVCs involving wanti- fall lúk=wánti-ɲa → /lukánti-ɲa/ ~ /lúkanti-ɲa/ ‘he/she lay down’ láp=wánti-ɲa → /lapánti-ɲa/ ~ /lápanti-ɲa/ ‘he/she got a lift’ cúp=wánti-ɲa → /cupánti-ɲa/ ~ /cúpanti-ɲa/ ‘he/she got down’ These CVCs involve the deletion of the initial semivowel of the inflecting verb and stress shift away from the initial syllable of the preverb (although this appears to vary as in Walmajarri (Hudson, 1978, p. 8)). They are analysed as preverb-verb compounds and not monomorphemic forms as this is a regular morphophonological rule that applies consistently across preverb-verb boundaries, not only in Ngardi but also in Walmajarri and Jaru (Hudson, 1978, p. 50; Tsunoda, 1981b, p. 178). Furthermore, the inflecting verb in these compounds is still used synchronically as a ‘simple’ verb wanti- fall.

348 

 7 Verbal morphology

7.2.3 Conjugation classes The conjugation class of an inflecting verb determines the allomorphy of the TAM (regular and directional) and associated motion suffixes with which it combines. Certain conjugation classes share varying degrees of overlap in their inflectional allomorphy, however. The complete inflectional paradigms of each conjugation class are presented in §7.3. However, each of the five conjugation classes is maximally distinguished by the imperative and potential inflections, as repeated in Table 82. These are, in effect, ‘diagnostic inflections’ for conjugation class membership. Table 82: Conjugation classes and diagnostic inflections. I

II

III

IV

V

Number of verbs Example

18 wangkaspeak

52 kupacook

7 kacarry

1 ngaeat

3 yago

potential imperative

-ju -ya

-ku -ka

-ngku -ngka

-lku -nja

-nku -nta

The majority of Ngardi verbs belong to Class II, followed by Class I. Classes III, IV and V are notable for their monosyllabic members (with some longer stems being the product of recent historical derivations). Class IV has a single member, nga-. While Cataldi (2011, p. 2) chooses to treat nga- as an irregular member of Class II (and posit only 4 conjugation classes), I favour an analysis of five distinct conjugation classes (sec. Green (1988)) due to the diagnostic imperative (-ka vs -nja) and potential (-ku vs -lku) inflections. The presence of a single member in Class IV (nga- eat) is paralleled in the Yapa languages, Warlpiri and Warlmanpa. Other Ngumpin languages possess a second verb ma- ‘say’ (a root homophonous with ma- get) in this conjugation (McConvell & Laughren, 2004, pp. 169–170). Ngardi lacks this verb and instead has a Class I verb wangka- ‘speak’; a lexical root also found in Warlpiri and various Western Desert languages. The three (Class III) verbs with monosyllabic roots display the vowel alternation /u/ ~ /i/ (lu- cry, pu- hit and yu- give) conditioned by the place of articulation of the initial nasal in the following inflectional morpheme (e.g. pi-nya hit-pst but pu-ngani hit-ipfv.pst). A straightforward phonological generalisation very nearly covers the variation: V → /u/ __-/ŋ/, V → /i/ / __-/ɲ/. However, the irrealis forms represents something of an exception: pi-ngi, yi-ngi, li-ngi, rather than *pu-ngi, *yu-ngi, *lu-ngi. One other verb stem also exhibits the same alternation: yaparti- ~ yapartu-. This is the only trisyllabic verb stem in class III and the only polymorphemic stem exhibiting a root-final vowel alternation. A likely explanation for this is that this verb was historically a complex verb, comprising the preverb yapart- ‘run’ with the inflect-

7.3 Overview of inflectional categories 

 349

ing verb pu- hit but there has been subsequent loss of the initial consonant of the historic verb stem pu- and a loss of primary stress on the historic morpheme-initial syllable: i.e. *yápart=pú- → yápartu-. The case for this diachronic source of this trisyllabic suffix is strengthened by the occasional usage of yapart- as a preverb synchronically, combining with pu-: yápart-kàrra púnganta. Nevertheless, I do not analyse yápartu- as a synchronic combination of verb and preverb (i.e. yapart=pu), as I do for lúkanti- ( kangu ani > kangani (Koch, 2014, p. 177).

7.3 Overview of inflectional categories 

 357

manteau as I take this to be the most straightforward treatment available (although perhaps not the most analytically satisfying). The templatic structure of the regular inflecting verbal template is presented in (855). This template deals exclusively with inflecting verbal stems and associated derivational and inflectional TAM morphology. The various relationships between preverbs and verbs (i.e. CVCs) are described separately in §8.3 – but note that certain preverbs, labelled ‘strong nexus preverbs’ could be considered to be a type of ‘bound’ and left-compounding structure to the verb root. (855)

[verb root]-[cm.assoc mot]-([cm].[aspect1].[tense/mood])-[aspect2]-[dir] Portmanteau in gloss

Within the above template, I again include the ‘conjugation marker, CM’ recognised by Australianists. I do not formally identify it as a suffix in the morphological treatment of the Ngardi verbal paradigm but it is provided here for clarity’s sake. The tense/mood slot is defined positionally. It is not a functionally-unified slot but a formal position in a template which may host a wide range of verbal inflections with diverse functions within the domains of tense and mood. The aspect2 slot represents the peripheral attachment of the narrative suffix (-nyirra, -ngurra, nyurra) although it is not neatly classified as a purely ‘aspectual’ verbal feature (as discussed in §7.4.2). The expression of aspect2 is dependent on the expression of cm and a past tense value in the [tense1/mood] position – that is, aspect2 cannot concatenate directly with a root, nor with any other tense/mood value. The dir slot is the final position within the template and hosts one of two directional suffixes: -rni ‘hither’ or –rra ‘hence’. Importantly there are a number of tense/mood + directional combinations that cannot be analysed compositionally and are analysed instead as portmanteaux. This is addressed in detail in §7.5. The directional inflections have not been found to co-occur with the narrative (aspect2) inflection. The verbal template presented in (855) can be exemplified with the verb ma- get. The first set of examples in (856) demonstrate the various concatenations of the [cm], [aspect], [tense1/mood] and [tense2]: (856) Maximally analytic gloss ma-n-i get-(cm-pst) ma-n-an-i get-cm-ipfv-pst ma-n-i-nyirra get-cm-pst-narr ma-n-an-i-nyirra get-cm-ipfv-pst-narr

Translation ‘he went’ ‘he was going’ ‘he went (a long time ago)’ ‘he used to go (a long time ago)’

Employed gloss ma-ni get-pst ma-nani get-ipfv.pst ma-ni-nyirra get-pst-narr ma-nani-nyirra get-ipfv.pst-narr

358 

 7 Verbal morphology

The set of examples in (857) demonstrates various concatenations of [assoc mot] with the features: [cm], [aspect1], [tense/mood] and [aspect2]: (857) Maximally analytic gloss ma-n-inja-n-i get-cm-move&x-cm-pst ma-n-inja-n-an-i get-cm-move&x-cm-ipfv-pst ma-n-inja-n-i-nyirra get-cm-move&x-cm-pst-narr ma-nungka-ny-a get-x&move-cm-pst

Translation ‘he went and got it’ ‘he was going and getting it’ ‘he went and got it (a long time ago)’ ‘he got it and left’

Employed gloss ma-ninja-ni get- move&x-pst ma-ninja-nani get-move&x-ipfv.pst ma-ninja-ni-nyirra get-move&x-pst-narr ma-nungka-nya get-x&move-pst

It is possible – although attested only during elicitation – for the associated motion slot to comprise the two associated motion suffixes. This has only been observed with the ordering [root]-move&x–x&move-, and so far only with the verb ma- get. An example is provided in (858) and is discussed further in §7.5.3. (858) Yala=yanu ma-ninja-nungka-nya, wurna. that=3pl.o get-move&x-x&move-pst journey He went and got them and left, he set off. (PSM: TEN1-2019_005-01: 217705_222762) There are a number of inflections available to the [tense1/mood] slot which can be described as ‘parasitic’ (Matthews, 1991), in that they are best analysed as monomorphemic forms built off existing tense/mood inflections (§7.3.2). Having considered the compositionality of various inflections, the glossing conventions utilised elsewhere in the grammar can be summarised as follows: i) Conjugation markers are not glossed but integrated as the initial segment of a following inflection. ii) The aspectual formative (-(N)an) is not separated from TM inflections by a morpheme break but glossed as a portmanteau with the following TM inflection, i.e. -nani ipfv.pst. iii) The imperfective is the only marked aspect. I do not analyse a zero morpheme encoding a ‘perfective’ aspect – i.e. ma-ni get-pst vs ma-nani get-ipfv.pst. 7.3.1.1 Composition of the narrative inflection The narrative inflection is the only verbal suffix that displays a component of morphophonologically-conditioned allomorphy not entirely dictated by conjugation class. The full set of narrative inflectional forms are repeated in Table 88 alongside the past and imperfect past inflections off which they are built.

7.3 Overview of inflectional categories 

 359

Table 88: Past tense inflections. I (y)

II (-Ø)

III (ng)

IV (l)

V (n)

pst pst-narr

-nya -nya-ngurra, -nya-nyirra

-rni -rni-nyirra, -rni-nyurra

-nya -nya-ngurra, -nga-nyirra

-rni -rni-nyirra, -rni-nyurra

-ni -ni-nyirra, -ni-nyurra

ipfv.pst ipfv.pst-narr

-ngani -ngani-nyirra

-rnani -rnani-nyirra

-ngani -ngani-nyirra

-rnani -rnani-nyirra

-nani -nani-nyirra

The major allomorphic variation in the narrative morpheme, -nyirra/-ngurra, is phonologically determined. The allomorph -nyirra follows high-front vowels while -ngurra occurs elsewhere. Since the narrative suffix is morphotactically constrained to co-occurr with the past or imperfective past suffixes, the allomorphs -nyirra and -ngurra essentially co-occur with the past tense allomorphs -Ni and -nya, respectively; while the allomorph -nyirra is found following all allomorphs of the imperfective past (-ngani, -rnani, -nani). There is also an irregular variant, -nyurra (found in phonological contexts in which -nyirra is expected) co-occurring with verbal stems in all but one of the conjugation classes: ma- get, ya- go, karri- be, nga- eat, wanja- leave. It is possible that this irregular form may constitute an inflectional borrowing from Jaru or Walmajarri. The same ‘borrowing’ hypothesis is invoked for the unexpected appearance of the narrative form -nyurra in Mudburra (expected: -Nirra, -nyarra) attaching to Mudburra-origin verbs (Osgarby, 2018, pp. 42–43).312 The variant -nyurra therefore spans a wide geographic area right across the Ngumpin languages, from Mudburra in the (north)-east to Ngardi in the south(-west).

7.3.2 Parasitic inflections Within the regular (non-directional) verbal paradigm, Ngardi possesses a set of modal verbal inflections which are transparently built off another inflection within the verbal paradigm.313 Inflectional forms being ‘built off’ other paradigmatic forms have been referred to as ‘parasitic formations’ (Matthews, 1991) and are ‘actually relatively widespread in Australian verb[al] inflection[s]’ (Koch, 2014, p. 158).314 Parasitic formations in the Ngardi verbal system and their compositions are presented in Table 89.

312 In Mudburra the narrative inflection (-Nirra, -nyarra, -nyurra) is paradigmatically distinct from Ngardi in that it can attach directly to the verbal stem: e.g. nya-nyarra see-narr. Interestingly, one speaker of ‘Western Mudburra’ happened to use forms built off the past tense, directly comparable with forms in Jaru and Ngardi: e.g. kayi-nyangurra be-narr (Osgarby, 2018, p. 66). 313 Only the hortative is also present in the ‘directional inflectional paradigm’, see §7.3.3. 314 This phenomenon has also been referred to as ‘Priscianic formations’ (Matthews, 1972; Aronoff, 1992; Shimada, Nagano & Okubo, 2015)

360 

 7 Verbal morphology

Table 89: Parasitic inflections. Example (ma- get)

Formative + Increment

hortative

ma-nkura get-hort ‘let him get it’


temporal (Traugott, 1978) and it is reasonable to assume that this shift also entails a stage where both semantic interpretations (spatial and temporal) are available: (957) Extrapolated grammaticalisation path for directional morphology spatial > spatial & temporal > temporal One might speculate that the -rra suffix was to some degree ‘available’ for grammaticalisation since a directionally-unmarked form of the verb ya- is often pragmatically interpreted as movement away from the speaker, i.e. ya-nta ‘go (away)!’341 The pragmatic interpretation of directionally-unspecified lexical verbs with a conventionalised deictic meaning through opposition with another form (e.g. ya-nta go-imp vs ya-nta-rni go-imp-hith) is discussed further by Wilkins and Hill (2009). 7.5.2.1 Directional imperative The directional imperatives are formed compositionally via the affixation of -rni and -rra to the regular imperative inflection (-ya, -ka, -ngka, -ja, -nta) as exemplified with the locomotion verb ya- go in (958). (958) Jaja ya-nta-rni mu-kurra pawujeid-kurra. MM go-imp-hith this-all bough shade-all nga=rna=ngku juju pina=yi yu-ngku. seq.aux=1sg.s=2sg.o ceremony knowledge=emph give-pot ‘Grandson, come here to the bough shade so I can teach you songs/ ceremony.’ (MAM: TEN1-2016_002-01: 854015_857935) Examples (959)–(964) illustrate the directional inflections attaching to non-locomotion verbs. 341 This is equally true of Warlpiri (Laughren, 1978, p. 3).

7.5 Directional verbal inflections 

 395

(959) Yupa-ka-rni=pula. send-imp-hith=du.s ‘You two, send him over here.’ (PSM: TEN1-2017_020-01: 136728_138731) (960) Luwa-ka-rra=lu kuja-purda! shoot-imp-hence=pl.s thus-wards ‘You lot, throw it over that way!’ (PSM: TEN1-2018_023-02: 1334808_1339081) (961) Luwa-ka-rni minya warririku. shoot-imp-hith this close ‘Throw (the spear) this way, (at) this one close by.’ (PSM: TEN1-2018_023-02: 1424019_1426991) (962) Yut=karri-ya-rni. sit=be-imp-hith ‘Sit closer to me.’ (PSM: TEN1-2018_023-02: 2675101_2676773) (963) Yut=karri-ya-rra (kuja-purda). sit=be-imp-hence thus-wards ‘Sit further away.’ (PSM: TEN1-2018_023-02: 2736820_2739095) (964) La-nta-rra kuja-purda. pierce-imp-hence thus-wards ‘Dig over that way.’ (PSM: TEN1-2018_023-02: 965865_966980) As discussed in §7.5.2, there are certain contexts in which the translations provided were not inherently directional ‘(do) x away’ but rather aspectual or inceptive: ‘go ahead and x’ or ‘keep x–ing’. Thus consider the translations for (965)–(967): (965) Ya-nta-rra=pula. go-imp-hence=du.s ‘You two go away.’ (TJN: LC60a: 450773_453120) (966) Ya-nta-rra=lu. go-imp-hence=pl.s ‘You lot, keep on going.’ (PSM: TEN1-2018_028-01: 4642783_4644914) (967) Nga-nja-rra=lu, kuja=nta=rla ngarru nga-rnu-ku. eat-imp-hence=pl.s thus=2pl.s=3sg.obl happy eat-inf-dat ‘Go ahead and eat, you lot enjoy having a feed (lit. ‘be happy for eating’).’ (PRN: TG1_0034-01)

396 

 7 Verbal morphology

These observations are also highly relevant to the semantics of the directional imperfective imperative, discussed in §7.5.2.2. 7.5.2.2 Directional imperfective imperative The directional imperfective imperative suffixes (-Nantarni, -Nantarra,) are built off a set of forms identical to the present tense (-nanta, -rnanta, -nyanta, -nganta) in Ngardi. A suffix of the same form appears as a ‘continuous imperative/continuous irrealis’ in Jaru, Gurindji, Walmajarri and Mudburra. Therefore, despite lacking an ‘imperfective imperative’ in its regular paradigm, Ngardi has (or perhaps ‘retains’) an ‘imperfective imperative’ within its directional inflectional system. The major distinction in the usage of the imperfective directional imperative from the unmarked equivalent is that the focus of the illocution is to direct an action that is to continue (i.e. not just be enacted as a single punctual event). For instance, Marie provided the utterance in (968) as a hypothetical command given when someone is about to leave. The precise command was that the addressee follow behind in another vehicle (specified by the locational nominal purdangirli). The choice of the imperfective imperative serves to emphasise the continuous nature of a ‘following’ event rather than a punctiliar directive to ‘go’ or ‘come’. (968) Ya-nantarni purdangirli-ja! go-ipfv.imp.hith behind-loc ‘Come behind! (Lit. ‘Be coming behind!’)’ (TEN1-2017_019-01: 990718_1001444) An example involving a non-motion verb is provided in (969). (969) Yupa-rnantarni purdangirli! send-ipfv.imp.hith behind ‘Send it behind (after us)! (Lit. ‘Be sending it behind!’)’ (PSM: TEN1-2017_02001: 653012_656611) The semantics of the ‘hence’ imperfective imperative differ slightly from the ‘hither’ in an interesting way. Marie Mudgedell and Patrick Smith describe the use of -Nantarra as being used in contexts where the speaker is about to leave and is instructing the addressee to continue a given event/activity; Patrick Smith commented ‘(it’s for) when you get up and walk away’, as illustrated in (970)–(973). (970) Karlarra=lku, la-nantarra=wu. west=then pierce-ipfv.imp.hence=voc ‘(I am) going west, you keep on digging.’ (MDN: LC22a: 1233945_1237228)

7.5 Directional verbal inflections 

 397

(971) La-nantarra=wu wurna ngu=rna=wali. pierce-ipfv.imp.hence=voc journey seq.aux=1sg.s=then ‘You keep digging, I am heading off.’ (PSM: TEN1-2018_023-02: 1334808_1339081) (972) Nyuntu=wali luwa-rnantarra, ngu=rna ya-nku. 2sg=then shoot-ipfv.imp.hence seq.aux=1sg.s go-pot ‘You, you keep filing (it), I will head off.’ (PSM: TEN1-2018_023-02: 2109385_2111103) (973) Nya-ngantarra=lu ngalkin=ma-nantarra=lu=rla. see-ipfv.imp.hence=pl.s protect=get-ipfv.imp=pl.s=3sg.obl ‘You must keep on looking out, you must keep on looking after him.’ (DMN: LC42a: 1286416_1289156) There is the question as to whether or not the ‘hence’ morpheme necessarily encodes movement away in these contexts or whether it has shifted to a primarily continuous or repetitive aspectual function. During elicitation, I asked whether lanantarra may be used as a command by a speaker who was not about to depart, but intended for someone to continue their digging away from the speaker (i.e. I prompted with kuja-purda ‘this way’ and gestured away from our position). However, Marie Mudgedell affirmed that her own intuitions were that it was used solely for ‘when you go away’ – that is, the addressee understands that they are to continue digging in situ, while the speaker, by using this particular imperative inflection, was implicationally announcing their imminent departure. This being the case, one might speculate that a speaker’s departure (upon uttering a command) is a possible bridging context in which the event being described is enacted at some distance away from the speaker. This has possibly been subject to further grammaticalisation into the aspectual continuous ‘be x–ing’ (974) or even inceptive ‘go and x’ (975) functions. (974) Yunpa-rnantarra=lu! sing-ipfv.imp.hence=pl.s ‘You mob keep on singing!’ (IPN: TEN1-2018_028-01: 5087255_5088608) (975) Wakurra=n=ku=rla yinkil=yirra-ku, nga-rnantarra. neg=2sg=ep=3sg.obl unwilling=place-pot eat-ipfv.imp.hence ‘Don’t you be fussy about it, go ahead and eat them (sand frogs).’ (MMJ: LC31: 843413_845735) A functional shift in the spatial semantics of -rra is perhaps functionally supported by the fact that Ngardi has no other means to express a directionally unspecified imperative command with an imperfective aspectual specification since the bare form -Nanta

398 

 7 Verbal morphology

has developed into a regular present tense marker (unlike in the other Ngumpin languages in which it encodes an imperfective imperative). The directional imperfective imperative form of the motion verb ya- go is also used as something of an interjective farewell and, in this function, may host the vocative clitic =wu. Patrick Smith provided the typical (but hypothetical) conversational exchange shown in (976). (976) a.

b.

“Wurna=rnalu=wu.” journey=1pl.excl.s=voc ‘“We are going.”’ “Yuwayi, ya-nantarra=lu=wu!” yes go-ipfv.imp=pl.s=voc ‘“Yeah, see you mob!” (Lit. “You mob, go away!”)’ (PSM: TEN1-2018_02801: 4801493_4806909)

7.5.2.3 Directional potential The directional morphemes (-rni, -rra) are concatenated in a regular fashion with the allomorphs of the potential (-ju, -ku, -ngku, -lku, -nku). For motion verbs, the ‘hither’ suffix -rni occurs far more frequently on potential verbs, as in (937). This may relate to the fact that a motion verb unmarked for direction is pragmatically interpreted as ‘away’ from the speaker. (977) Ya-nku-rni=rna wirrirl. go-pot-hith=1sg.s back ‘I will come back.’ (TJL: TT76_2201: 1448215_1450432) The ‘hence potential’ was also difficult to identify due to its near homophony with the regular hortative ya-nkura ‘go-hort’ (§7.4.5). Many speakers freely lenite the alveolar tap /r/ ‘rr’ to a continuant [ɹ] such that the distinction between /ɻ/ ‘r’ and /r/ ‘rr’ is, at times, difficult to perceive. This is compounded by the fact that the semantic range of the hortative shows clear overlap with the meaning of the ‘away-directional-potential’. Examples are (978)–(981). (978) Makan ma=rna ya-nku-rra.342 morning top.aux=1sg.s go-pot-hence ‘I will head off in the morning.’ (PRN: TG1_0034-01)

342 The temporal makan ‘morning’ is a Jaru word. The Ngardi equivalent is rangarni.

7.5 Directional verbal inflections 

 399

(979) La-nku-rra=yanu kuja-purda. pierce-pot-hence=3pl.o thus-wards ‘He will shoot at them over that way.’ (PSM: TEN1-2018_023-02: 738935_740195) (980) Luwa-ku-rra kuja-purda. shoot-pot-hence thus-wards ‘He will throw it off that way.’ (PSM: TEN1-2019: 1937093_1944291) (981) Mangarri yupa-ku-rra=yanu veg_food send-pot-hence=3pl.o ‘He will send the food over to them.’ (PSM: TEN1-2018_023-02: 2284463_2287760) 7.5.2.4 Directional imperfective potential The directional imperfective potential combines regularly with the ‘hither’ directional suffix, as in (982) and (983). (982) Wurra ma ya-nanku-rni. later top.aux go-ipfv.pot-hith ‘She will be coming (later).’ (PSM: TEN1-2019_005-01: 931159_939054) (983) Kuyi ma ka-nganku-rni. meat top.aux carry-ipfv.pot-hith ‘He will be bringing the meat here.’ (PSM: TEN1-2019_005-01: 1141996_1150169) Only one clear case of the imperfective potential in combination with the away suffix (984) occurs in the corpus. However, as observed for the unmarked directional potential, it is generally difficult to distinguish this concatenation of morphemes from the imperfective hortative (§7.4.5.1). (984) Wakurra, ya-nani=wu. Wurra, ya-nanku-rra=wu. neg go-ipfv.pst=voc wait go-ipfv.pot-hence=voc ‘Oh no, he was going (leaving), he will be going away (the wrong way).’ (MMN: TEN1-2019_005-01: 911933_920199) [Context: Looking at tracks leading off in the wrong direction] 7.5.2.5 Directional hortative There is somewhat limited evidence for the hortative (-jura, -kura, -ngkura, -lkura, -nkura) combining with the directional suffixes. There are examples of the ‘hither’ suffix -rni combining regularly as in (985) and (986) but the current data does not extend to combinations involving all conjugation classes.

400 

 7 Verbal morphology

(985) Tardarr=ya-nkura-rni kaniny-kaniny. enter=go-hort-hith inside-rdp ‘Let him come inside.’ (PSM: TEN1-2017_026-01: 3040855_3043990) (986) Ngaringka yala ya-nkura-rni, mangarri ma-nku minya. woman that go-hort-hith veg_food get-pot this ‘Let that woman come over here, she can get some food.’ (PSM: TEN1-2018_03401: 213735_221165) There is only a single example of a hortative combining with the ‘hence’ directional suffix, as observed in (987). (987) Parda jakiliny-kariny-jaku, kaji=rnalu nganimpa=yi dub month-other-? condit=1pl.excl.s 1pl.excl=emph mariyawu=wali ya-nkura-rra. long_way=then go-hort-hence ‘It might be next month when we head off on that long trip.’ (PRN: FT1_a: 1527791_ 1533433) The imperfective hortative (-nyankura, -rnankura, -ngankura, -nankura) has not been attested in combination with either of the directional suffixes. 7.5.2.6 Directional past The directional past forms are built off a set of forms morphologically identical to the infinitive inflection in the regular paradigm (-rnu-, -ngu-, -nyu-, -nu-, §7.4.8).343 Examples of both ‘past hither’ and ‘past hence’ in combination with various verbs are provided in (988) – (991). (988) Ya-nurra=pula. go-pst.hence=3du.s ‘Those two went away.’ (MMN: TEN1-2017_021-01: 119679_122019) (989) La-nurni kalkurni-purda-rlu. pierce-pst.hith this_side-wards-erg ‘She was digging, coming this way.’ (PSM: TEN1-2018_023-02: 329524_336420)

343 As has been pointed out, forms cognate with these inflections are the basic past tense forms in Warlpiri.

7.5 Directional verbal inflections 

 401

(990) Mama-ngku=lu=nganpa ka-ngurni.344 F-erg=3pl.s=1pl.excl.o carry-pst.hith ‘The fathers brought us.’ (NNA: ngurungka 30_4_91) (991) La-nurra wanja-rnungka-nya yalu-ngka, Napanangka-rlu. pierce-pst.hence leave-x&go-pst that-loc ♀subsection-erg ‘Napanangka dug in that place, (then) she left and set off.’ (MMN: TEN1-2018_ 023: 256091_258000) 7.5.2.7 Directional presentative The semantic and temporal functions of the directional presentative appear to be more constrained than the usages of the regular present tense. Marie Mudgedell specified that -Nanyarni/-Nanyarra is only used at the time of the event – for instance while talking to someone on the phone who has asked about the whereabouts and movement of someone at that time. Marie’s observations match with other examples of the corpus wherein the events described by the presentative always appear synchronous with speech time. The label ‘presentative’ was chosen due to the functional similarity with the ‘presentative present’ in Warlpiri which describes events/states that hold at the time of the speech event (Laughren, 2017b, p. 3).345 The presentative is not used in some of the distinct contexts applied to the general present that do not refer to events synchronous with speech time (e.g. ‘historic present’ or for ‘planning/scheduling’, see §7.4.9). Examples of the ‘hence’ presentative are provided in (992) and (993). (992) Yala=lu ya-nanyarra. that=3pl.s go-prst.hence ‘They are still going.’ ‘(Dei) kip gorn.’ [Trs. Patrick Smith] (PSM: TEN1-2018_028-01: 4627464_4629171) (993) Yala-rla=pula ya-nanyarra. that-loc=3du.s go-prst.hence ‘(Look) there are those two are on their way over (now).’ (MMN: TEN1-2017_ 019-01: 811014_814282)

344 The kin term mama ‘father’ is a Kukatja word. The Ngardi equivalent is wapirra. 345 The form of the prs.dir base -Nanya has direct parallels to the ‘presentative present’ in Warlpiri for Warlpiri conjugations IIB (ka-nganya) and V (ya-nanya). Nevertheless, the presentative forms for the remaining Warlpiri conjugation classes lack the -(N)an element, either suffixing -nya directly to the stem (wangka-nya [I]) or else to a form reminiscent of the past tense (with u > i) paka-rni.nya [II]) and (nga-rni.nya [IV]).

402 

 7 Verbal morphology

Examples of the ‘hither’ presentative are provided in (994) and (995). (994) Ngapa ma yala kirda ya-nanyarni. water top.aux that big go-prst.hith ‘A big rain (storm) is coming.’ (PRN: FT1_b: 99708_101606) (995) Ya-nanyarni jalangu ya-nanyarni. go-prst.hith today go-prst.hith ‘He is coming today, he is coming.’ (PSM: TEN1-2017_019-01: 547801_550099) Verbal predicates inflected with the presentative are always conceptualised as ‘activities’ even if the lexical root itself may have inherent non-durative lexical meanings (i.e., accomplishments or achievements). In the following examples, the progressive forms of yupa- send (996), nga- eat (997) and luwa- shoot (998) have no indication that the endpoints of their activities were reached or their achievement were attained. (996) Yupa-rnanyarni=pula mu-kurra. send-prst.hith=3du.s this-all ‘Those two are sending it here.’ (MMN: TEN1-2017_020-01: 293330_297707) (997) Nga-rnanyarni.346 eat-prst.hith ‘He is eating, coming this way.’ (PSM: TEN1-2017_020-01: 1014065_1015975) (998) Luwa-rninyarni=nganpa nganimpa, nyampa-jangka? shoot-prst.hith=1pl.excl.o 1pl.excl what-ela ‘He is pelting (rocks) this way at us, what for?’ (PSM: TEN1-2019: 1631713_1636357)

346 This example provides some evidence that -rni and -rra could actually be analysed as associated motion markers (Koch, 1984) – since, in addition to marking path they are adding a motion component to an otherwise non-motion verbal predicate. However, these usages were only uncovered through elicitation and were not readily accepted for a very wide range of verbs. The canonical usage of the morphs -rni and -rra seems to be to add directional/orientation information to already inherently motion predicates. For this reason, I resist labelling them fully-fledged associated motion markers and retain that term for the suffixes -nungka and -ninja which – in their canonical usage – add a motion component to a wider range of non-motion verbs.

7.5 Directional verbal inflections 

 403

7.5.3 Associated motion The two derivational suffixes (-Nungka- and -Ninja-/-nyanja-) encode associated motion in Ngardi. ‘Associated motion’ is a grammatical category which refers to the encoding of motion in a grammatical rather than a lexical morpheme. Koch (1984) first observed and coined this term when observing certain verbal suffixes in the Australian language Kaytetye (Arandic), but it has since been recognised in many other Australian languages and beyond, with significant scholarly attention being paid to various languages of South America (Guillaume, 2016). In more recent typological work, associated motion has been defined as: A verbal grammatical category, separate from tense, aspect, mood and direction, whose function is to associate, in different ways, different kinds of translational motion (spatial displacement/ change of location) to a (generally non-motion) verb event. (Guillaume & Koch, In press, p. 1)

Associated motion suffixes in Ngardi serve to encode two kinds of motion co-events that differ based on their temporal relationship. I have labelled these ‘prior motion’ -Ninja-/-nyanja- move&x and ‘subsequent motion’ -Nungka- x&move, where x denotes the main verbal event, following Koch (1984).347 Ngardi does not possess an associated motion suffix that encodes ‘concurrent motion’, although this same notion can be expressed by a non-finite verb form of the ‘go’ verb marked with various case suffixes appropriate to its syntactic controller, e.g. ya-nu-rlu go-inf-erg (subjectcontrolled concurrent motion), yanu-kurra go-inf-all (object-controlled concurrent motion), see §11.3.348 The two associated motion suffixes do not have overt directionality/path specifications. Directionality is generally interpreted from context or is encoded elsewhere in the clause. Furthermore, the associated motion suffixes have generally not been found to combine with the ‘directional’ inflections – although 347 It is important to note that many examples of -Nungka suffix in the Cataldi corpus are translated by Cataldi (2011) as ‘go and x’ and are described as functioning identically to the dependent ‘go&V’ verb ‘nja-n-’ in Warlpiri. However, the presence of an additional associated motion suffix -ninja in Ngardi with a clear function of ‘prior motion’ (similar to Warlpiri) and more recent careful translations with Ngardi speakers throughout 2015–2019, suggests that the semantics of -Nungka involve a temporal ordering of subsequent motion ‘x&move’. Note also that Cataldi had listed two ‘verbs’ manunka‘get and take’ and maninja- ‘go and get’ and so there are some inconsistencies in her various translations. It is possible, however, that the suffix -Nungka- was in the process of developing into a more general associated motion suffix unspecified for temporal sequencing (i.e. with prior, simultaneous or subsequent meanings) given the less frequent use of -Ninja-/-Nanja- across different verb stems and TAM values. 348 Equally, I do not analyse other non-finite subordinate clauses (e.g. kupa-rnu-ngkarla nga-rni ‘having cooked he ate’) as cases of ‘associated motion’ since these do not form inflectional categories of a finite verb but rather complex clauses involving a finite main verb and a non-finite subordinate verb marked with case in subordinating function. Both -Ninja-/-nyanja- and -Nungka- are inflectional categories of a single, ‘finite’ inflecting verb and hence can be considered markers of associated motions.

404 

 7 Verbal morphology

some isolated forms were accepted during elicitation. In fact, the subsequent motion suffix -Nungka- has a default interpretation ‘x and go’ and has not been observed with the converse directional sense ‘x and come’. The semantic contrast between prior and subsequent associated motion suffixes is illustrated in (999) and (1000). (999)

Pi-nyanja-ni. hit-move&x-pst ‘He went and killed it.’ (YMN: LC23a: 1738121_1738841)

(1000) Pu-ngungka-nya, wurna-ngku. hit-x&move-pst journey-erg ‘He killed it and left, while on a trip.’ ‘Hit and run.’ [Marie Mudgedell’s trs.]: (PSM: TEN1-2018_033-03: 1246275_1247929) In all cases, the associated motion event is associated with the ‘subject’ (transitive or intransitive) and cannot be used to encode motion associated with non-subjects (i.e. objects, obliques or subjuncts). This restriction is widely reported for associated motion suffixes cross-linguistically (Guillaume, 2016, p. 91). However, examples of associated motion markers encoding motion associated with the object (O) of the verb expressing the lexical event can be found in Mparntwe Arrernte (Pama-Nyungan, Australia: Wilkins, 1991) and Cavineña (Tacanan, Bolivia: Guillaume, 2009). The forms of the associated motion suffixes in Ngardi are repeated in Table 94 for convenience. Table 94: Associated motion (AM) inflections by conjugation. AM type

I

II

III

IV

V

Subsequent

-nyungka-

-rnungka-

-ngungka-

-rnungka-

-nungka-

Prior

-nyanja-/ -nyanunja-

-rninja-

-nyanja-/ -nyaninja-

-rninja-

-ninja-

The formation of the associated motion suffixes is regular except for the prior motion suffix, where there is variation between a regular disyllabic form and an augmented form involving a medial -nu or -ni in Conjugations I and III respectively. In Conjugation I, there is variation between -nyanja- and -nyanunja-, while in Conjugation III, there is variation between -nyanja- and -nyaninja-. There is insufficient data to determine whether this variation applies to each verb root within each class or whether these variants are somewhat idiosyncratic to certain lexical roots. Formally speaking, the augmented forms involve what appear to be aspectually imperfective forms but no clear meaning contrasts have been identified.

7.5 Directional verbal inflections 

 405

Both of the associated motion suffixes likely have origins in complex verb constructions involving motion verbs but the details of their development appear to differ slightly. In both cases they appear to involve a (historic) non-finite verb form + finite motion verb – a common source for associated motion suffixes cross-linguistically (Guillaume, 2017) and one described for the Arandic languages (Koch, 1984; Wilkins, 1989, 270ff., 1991, 1997).349 The former (-Nu.ngka-)350 appears to derive form ka- carry and the latter (-Ni.nja-/-nya.nja-) from ya- go. Interestingly, these two finite verb stems are concatenated with different types of inflection: the synchronic non-finite suffix -Nu for x&move but the synchronic past tense -Ni/-nya and a foreign form -nja for move&x. The quality of the suffix-initial nasal in both cases is consistent with the conjugation class marker utilised in the synchronic past tense forms. The latter form likely involves an historic connection with Warlpiri since in Warlpiri non-finite verb forms are cognate with past tense forms in Ngumpin languages (Laughren, 2017b). Verb forms resulting from -Nungka- and -Ninja/-nyanja- suffixation take verbal TAM inflections from the conjugation classes of the verbs ka- carry (III) and ya- go (V), respectively. The finer details of the historical development of the two associated motion markers vary however. In the case of the prior motion suffix {-nya(nu)nja, -nya(ni)nja, -ninja-, -rninja}, an internal reconstruction is difficult. Synchronically a -ni formative has the status of a finite past tense suffix and Ngardi makes no other synchronic use of a form -nja, or even -n-ja. I set aside the issues of accounting for the variants -nyanunja- and -nyaninja- for Conjugations I and III, respectively, other than to note they may relate to the aspectual distinction -an vs -n (see §7.3.1). Comparison with neighbouring languages is fruitful, however. Warlpiri has verbal inflections which are formally and functionally very similar to the allomorphs -ninja-/-rninja. They are the Warlpiri non-finite suffix {-(ni)nja-} and the ‘inceptive’ {-(Ni)-nji} where (-Ni) refers to the ability of the inceptive to attach to either participial forms or a bare stem {-rninja-, -rnu-nju- and -nu-nju-} (Simpson, 2001, pp. 184–186; Laughren, 2010).351 The major difference in these inflections is the quality of the final vowel (which is further subject to regressive vowel harmony in the context of following high vowels). Functionally, the Warlpiri inceptive bears the closest relationship to the ‘prior associated motion’ suffix in Ngardi since it is a derivational suffix and semantically expresses

349 Cf. discussion of this construction in Wanyjirra and Jaru (Senge, 2015, pp. 472–476). 350 The full stop/period in -Nu.ngka- and -Ni.nja-/-nya.nja- marks an etymological morpheme boundary which is synchronically inactive. 351 Analyses of the Warlpiri inflection vary by author. Simpson (2001, pp. 184–186) labels an inflection {-Ninja, -Nunja-} as the inceptive; Laughren (2010, p. 174) treats the inceptive solely as -nji, which is optionally suffixed to participial inflections -Ni. Laughren also notes that the verb ya- go in Warlpiri also synchronically functions as a dependent (grammatical) verb =ya path which combines with the infinitive -ninja (as described by Simpson, 2001). The complex verb so formed adds a ‘path’ event synchronous with the event of the main (root) verb.

406 

 7 Verbal morphology

‘the idea that the event was initiated by some movement’ (Simpson, 2001, p. 184). Nash (1982, 1986) proposes that the inceptive in Warlpiri is the result of univerbation of the Warlpiri participial (non-finite) suffix with the verb ya- ‘go’. Simpson (2001, p. 184) formalises this analysis as in (1001). (1001)

*-(Ni)nja-ya -inf-go

→

-(Ni)nji-. incep

Note that the process of medial glide deletion and vowel hiatus reduction implied here is widely reported in this area of Australia (Koch, 2014, p. 177). The development of a form -ninja- in Ngardi appears to involve a similar process despite the obvious fact that Ngardi does not use a participial infinitive form of the shape -(Ni)nja-. It is possible that Ngardi may have borrowed a wholesale inflectional form (as an associated motion suffix) without Warlpiri-style morphophonology – that is the prior associated motion suffix in Ngardi is not subject to any regressive vowel assimilation as observed in Warlpiri. Given that Ngardi does not make any use of a verbal participial form -(Ni)nja outside of the prior motion associated motion suffix it would be difficult to argue for a discrete internal reconstruction. It is reasonable to conclude that the ‘prior motion’ associated motion suffix {-nya(nu)nja, -nya(ni)nja, -ninja-, -rninja} did not develop in isolation in Ngardi. A verbal morpheme -nja appears in many Western Desert languages and may have even been subject to areal spread given its occurrence with related functions in Warlpiri, Warlmanapa, and even Warumungu (Simpson, 2001, 179). In the Western Desert languages, -nja is synchronically a nominalizing morpheme (Clendon, 1988 for Manyjilyjarra; Goddard, 1985, p. 75 for Yankunytjatjara; and Jones, 2011, pp. 282–283 for Wangkajunga). In Manyjilyjarra the nominalizing -nja can modify a root that is tenseless (e.g. ya-ni-nja) or tensed (ya-nu-nja). Both forms can be further

inflected for aspect or in the latter case a past/future distinction (see Clendon, 1988, p. 194 for more details). Manyjilyjarra deploys these forms productively in a range of non-finite subordinate clauses. This includes the possibility to mark a nominalised ‘go’ verb with a locative case in a prior function – termed a ‘prior condition, pc’ suffix by Clendon, i.e. ya-ninja-ngka go-nom-pc ‘having gone’. While Manyjilyjarra does not appear to have directly associated a nominalised verb form ya-ninja- with another go verb ya- in a serial fashion, this is the grammaticalisation path that seems to have occurred in Warlpiri and Ngardi.352 Briefly, it has been speculated that the form -nja itself in Western Desert languages 352 It is further interesting that the inceptive is the only inflection in Warlpiri which exhibits a paradigmatic distinction between ostensible imperfective and perfective forms: e.g. ya-nin-ja-na-ni vs ya-ninja-, perhaps evidence that the -nja form was integrated into the language prior to the loss of the imperfective aspectual suffix found in all other Ngumpin-Yapa languages (McConvell & Laughren, 2004, p. 174).

7.5 Directional verbal inflections 

 407

originates from a continuous -n and a locative -ja in a ‘prior tense’ function (see Clendon 1988 for details). Thus, the modern prior motion suffix in Ngardi likely has a long history with multiple stages of grammaticalisation involving casemarkers in subordinating/tense marking functions (as argued for Western Desert) and verb serialisation (as argued for Warlpiri).

With respect to the subsequent motion suffix -Nungka ‘x&move’, a more transparent (and possibly more recent) grammaticalisation pathway is evident. Specifically, the subsequent motion suffix likely derives from the marking of a synchronic non-finite verb form with a locative subordinating case suffix encoding prior relative tense.353 Thus, consider the possible diachrony of ma-nungka-nya ‘he got it and left’ in (1002). (1002)

*[ma-nu-ngka get-inf-loc

ka-nya] carry-pst

→ haplology ka → ∅

ma-nungka-nya get-x&move-pst

This would involve an isolated instance of haplology kaka → ka. Evidence in support of this analysis and a more recent origin of x&move comes from the unusual asymmetries in the inflectional paradigms that may follow the two associated motion suffixes. While there is no a priori reason why the resulting verbal stem formed by either associated motion suffix could not take the full range of verbal inflections (conceivably including the directional suffixes), there are substantial gaps in the combinatorial possibilities of associated motion with TAM specifications.354 The subsequent motion suffix is only found combining with the past -nya and potential -ngku inflections. When attempting to elicit various other tensed forms, Patrick Smith switched to using the synchronically productive system of subordination via the locative -ngka/-rla (§11.3.3.1) or the sequential locative -ngkarla (§11.3.3.2). For example, when attempting to elicit ‘he is getting it and going’, the hypothetical form *ka-ngungka-nganta was rejected by Patrick Smith, who instead preferred the sequential locative case-marked non-finite verb forms as in (1003). (1003) Ka-ngu-ngkarla ka-nganta, or, ma-nu-ngkarla ka-nganta. carry-inf-seq_loc carry-prs or get-inf-seq_loc carry-prs ‘Having got it, he is taking it.’ (PSM: TEN1-2019_005-01: 565454_656577)

353 The same type of construction is observed as a fixed order construction with a ‘relatively tight boundary’ in Warlpiri (Simpson, 2001, p. 182). 354 This can be contrasted with the Warlpiri inceptive which is attested with all TM values and even includes an augmented imperfective form (ya-njinan-u, cf. form in (1013)) – a formal contrast that is not found in any other area of the Warlpiri verbal inflectional system (Laughren, 2010, p. 174).

408 

 7 Verbal morphology

Likewise, *manungkangku ‘he will get it and bring it’ was dismissed in favour of manungka kangku (get-inf-loc carry-pot), and *manungkangka! ‘get it and bring it’ was dismissed in favour of ma-nu-ngka ka-ngka (get-inf-loc carry-imp). Table 95 provides the attested co-occurrences of the AM suffixes with various TAM values, exemplified with the maximally attested verb ma- get. Table 95 clearly illustrates the discrete status of x&move as the less grammaticalised suffix as it is only found further concatenated with the potential mood.355 The grey shaded cells are populated with forms that speakers preferred as the appropriate forms for the elicited semantics but are distinguished by their productive status as a ‘complementised non-finite verb form + finite verb’. Table 95: Attested co-occurrences of TAM and associated motion suffixes.

Exemplar

Prior ‘move&x’

Subsequent ‘x&move’

ma- get

ma- get

ma-ninja-ni ma-ninja-ni-nyirra ma-ninja-nku ma-ninja-nta ma-ninja-ngi

ma-nungka-nya

Unmarked (Perfective) pst pst-narr pot imp irr inf

ma-nungka-ngku ma-nu-ngka ka-ngka

imperfective cust prs ipfv.pst ipfv.pst-narr ipfv.pot hort ipfv.irr anticip ipfv.inf

? ma-ninja-nanta ma-ninja-nani ? ? ? ? ? ?

ma-nu-ngka(rla) ka-nganta

In nearly all cases, attempts to elicit associated motion inflections in combination with the directional inflections (§7.5) were unsuccessful. I nevertheless include examples (1004) and (1005) which were accepted and repeated during elicitation. The exact semantics of their events are not well understood.

355 In defence of the position that x&move is a less grammaticalised AM inflection and possibly has a more recent diachronic origin, it is worth reiterating that, in the concatenation of both AM suffixes, it is the ‘less grammaticalised’ -Nungka- which occurs in the most rightward suffixal position: i.e. ma-ninja-nungka- < ma-ninja-nungka ka-.

7.5 Directional verbal inflections 

 409

(1004) Kartak, lurrij=ma-ni=rna ngurlukaji, ma-nungka-ngka-rra yala. cup finish=get-pst=1sg.s tea get-x&move-imp-hence that ‘The cup, I have finished (my) tea, get this (cup) and take it away.’ (MMN: TEN1-2019_005-01: 446579_456017) (1005) Ma-ninja-nu-rni. get-move&x-pst-hith ‘He came and got it (?)’ [Translation uncertain] (PSM: TEN1-2019_005-01: 268042_270177) Asymmetries in the general productivity of the associated motion suffixes also extend to the degree to which they are attested with the full inventory of verb roots. Table 96 provides an exhaustive list of all the verb forms with which the associated motion suffixes have successfully been identified to date. Table 96: Verbs attested with associated motion suffixes. AM suffix

Verbs attested with AM suffix

-ninja move&x

karri- be, kiji- throw, kupa- cook, ma- get, pangi- dig, pu- hit, payi- ask, wangka- speak, wanja- leave, yirra- place, ya- go356

-nungka x&move

karri- be, kupa- cook, ma- get, nga- eat, pu- hit, wangka- speak, yirra- place

As is clear from Table 96, AM suffixes do not appear to be readily compatible with predicates expressing stative event types. The predicate karri- which can form stative complex predicates (cf. §8.4.1.1) is only found expressing activities (i.e. ‘stopping’, ‘resting’) when suffixed with an associated motion inflection as in (1006) and (1007). (1006) Yaru=rnalu karri-nyungka-nya wurna ma=rnalu quiet=1pl.excls be-x&move-pst journey top.aux=1pl.excl.s ya-ni kankarra marru-kurra=wali, pinij. go-pst up.a house-all=then complete ‘We rested then left, we went up to the homestead, that’s it.’ (MNN: LC24b: 189356_194400)

356 The corpus contains only a single instance of ya- inflected with the move&x suffix (yaninjani) and its meaning/translation is unclear.

410 

 7 Verbal morphology

(1007)

Karri-nyanja-ni=pula paparti-rlangu, ya-ni=pula. be-move&x-pst=3du.s b-dyad go-pst=3du.s ‘They (the two brothers) stopped on the way.’ (Patrick’s translation: PSM: 119919_125370)357

There are no instances of actual stative event types (e.g. pina=karri- ‘know’) cooccurring with associated motion markers. 7.5.3.1 Prior motion move&x The prior motion suffix -ninja/-Nanja, encodes an event as having being preceded or initiated by motion as in (1008). (1008) Nya-nyanja-ni=rna. see-move&x-pst=1sg.s ‘I went and saw him.’ (MMN: TEN1-2018_033-03: 1591059_1593259) A closely parallel function is described for the Warlpiri ‘inceptive’ -nji: ‘it expresses the idea that the event denoted by the verb is initiated, probably by some movement’ (Simpson, 2001, p. 184). Warlpiri actually possesses two formally similar constructions, one involving the participial form of a verb compounded with either ya- go (1009) or parnka- move_rapidly, both of which may have ‘path’ interpretations, and the other the more grammaticalised ‘inceptive’ -njV (1010),358 which Laughren (2010, p. 175) analyses as an ‘aspectual verb’.359 (1009) Kuyu ka=rlipa paka-rninja ya-ni. meat pres=1pl.incl.s kill-inf go-npst ‘We are going along killing game.’ [Warlpiri] (Nash, 1986, p. 44)

357 In (1007), move&x is translated in a way that is suggestive of a ‘concurrent’ event. However, this is perhaps a result of a conceptualisation of the motion event being a larger macro event (a long trip) that, while being initiated prior to the event expressed by the verb root (karri-), is not terminated at this point. This conceptualisation appears to be affirmed by the re-specification of a motion event in the following clause. 358 The Warlpiri inceptive is argued to have derived historically from a compound *-ntya-ya-ni, i.e. the combination of a participial form with the non-past form of the verb go: -ya-ni go-npst (Simpson, 2001, p. 184). 359 Warlpiri actually possesses an additional associated motion construction involving the addition of the suffix -yalpi to the participial form marking prior return motion. The suffix -yalpi does not occur as an independent verb but is likely due to influence of Arandic languages (cf. -yalpe in Kaytetye (Simpson, 2001, p. 182)).

7.5 Directional verbal inflections 

(1010)

 411

Kuyu=lpa luwa-rnunju-nu ka-ngu=rnu=lpa. meat=imperf shoot-incep-npst carry-pst=hith=imperf ‘He went and killed some animal and brought it back.’ [Warlpiri] (Simpson, 2001, p. 184)

Of the two Warlpiri constructions above, the inceptive inflection (1010) represents the more grammaticalised of the pair, since the participial form in (1009) alone may also combine with parnka ‘run, move fast’. The Ngardi associated motion suffixes, as illustrated in (1011) and (1012), clearly resemble the Warlpiri inceptive, yet are clearly Ngardi forms based on some of the unique TAM inflections and the fact that they never undergo regressive vowel assimilation (cf. Warlpiri -rninja → -rnunju / __-Cu; → -rninji / __-Ci (Simpson, 2001, pp. 184–185)). (1011)

Kala ngati ngajupala-rlu pi-nyaninja-ni jaji kunyarr-kurlu-rlu. asrt M parent-erg hit-move&x-pst kangaroo dog-prop-erg ‘And my mother would go off hunting kangaroo with a dog.’ (YMN: LC23a: 244207_248752)

(1012)

Yala=rna kuyi nga-rninja-ni, wurna-ngku. that=1sg.s meat eat-move&x-pst journey-erg ‘I bin go eat, then go.’ (Patrick Smith’s translation: PSM: TEN1-2018_033-03: 1777239_1781354)

Further examples of the prior AM suffix combining with different TAM values are presented in (1013)—(1015). (1013)

Ngurrpa ma-ninja-nani, ka-ngu-ngkarla=yanu yu-ngani, ignorant get-move&x-ipfv.pst carry-inf-seq.loc=3pl.o give-ipfv.pst yirra-rni. place-pst ‘She was going and getting the ignorant ones/novices, having got them she sent them and placed them (there).’ (YDN: Maungka_02-022027)

(1014)

Ngapa ma-ninja-nta=yi. water get-move&x-imp=1sg.o ‘Go and get water for me.’ (PRN: TG1_0034-01)

(1015) Ya-nku=rna ma-ninja-nku warlu yala, wirriri=ka-ngku-rni=rna. go-pot=1sg.s get-move&x-pot fire that back=carry-pot-hith=1sg.s ‘I will go and get that firewood, I will bring it back.’ (PSM: TEN1-2018_033-03: 2655523_2659470)

412 

 7 Verbal morphology

7.5.3.2 Subsequent motion x&move The subsequent motion suffix -Nungka- encodes the fact that the event of the lexical verb is followed by a motion event as in (1016). (1016)

Ma-nungka-ngku=pulany kuriny. get-x&move-pot=3du.o two ‘(The car) will get and take those two.’ (MMN: TEN1-2018_033-03: 2697429_2700651)

The use of a ‘carry’ verb in a context that resembles a variously-grammaticalised associated motion function can also be found in Western Desert languages. In Yankunytjatjara, verbs compound productively with -kati (whose free verb sense is ‘carry, bring, take’) to form an associated motion construction with a meaning ‘do while going along’ (Goddard, 1985, pp. 118–120).360 In Warlpiri, a large number of preverbs may combine with the ka- carry to form complex verbs which have an intransitive ‘go’ or ‘move’ meaning (Laughren, pers. comm.). In Ngardi, examples of -Nungka sometimes reflect a semantic analysis of the associated motion type as actually x&carry or x&cause_to_move in contrast to a purely motion event x&move. That is, in (1017), involving ma- get, ‘get x and carry x’ is semantically equivalent to ‘get x and go’. (1017)

Ya-nani-nyirra kayini-mpala, puliki=yanu nguru=pu-ngani, go-ipfv.pst-narr north-along cattle=3pl.o steal=hit-ipfv.pst ma-nungka-nya-ngurra. get-x&move-pst-narr ‘He went along the north side, he was stealing bullocks for them, he grabbed and took them.’ (Yinjuru: LC: 168)

In fact, manungkanya in (1018) was translated by Marie Mudgedell as ‘stole’. (1018)

Yala ma-nungka-nya warlu, nyampa-jangka-rlu? that get-x&move-pst fire what-ela-erg ‘He stole that wood – what for?’ (MMN: TEN1-2018_033-03: 2601213_2604238)

However, the fact that the suffix has undergone a semantic shift to ‘x&move’ rather than purely a synchronic interpretation of ‘x&carry’ is clear from the examples in (1019)–(1020), and in particular (1019) where there is no possible interpretation of an associated induced change of location ‘carry’ event.

360 The same situation is found in Wirangu (Hercus, 1999, p. 130).

7.5 Directional verbal inflections 

(1019)

Nya-ngungka-nya=rna=yanu. see-x&move-pst=1sg.s=3pl.o ‘I saw them and left.’ (MMN: TEN1-2018_033-03: 1541311_1542960)

(1020)

Karri-nyungka-nya. be-x&move-pst ‘He rested and went.’ (BDO: TEN1-2017_023-03: 596671_598034)

(1021)

Pirdiny=karri-nyungka-nya wurna. arise=be-x&move-pst journey ‘He got up and went.’ (PSM: TEN1-2017_023-03: 662593_666596)

 413

The move&x suffix is generally incompatible with the carry verb ka- as the root verb, but it was accepted in a past tense form during elicitation in (1022). (1022)

Waku ma ka-ngungka-nya. neg top.aux carry-x&move-pst ‘Oh no, he took it and left.’ (MMN: TEN1-2019_005-01: 539949_545679)

7.5.4 Additional verbal morphology: -kujarra ‘while’, -kariny ‘reciprocal’ There are at least two additional verbal inflections within the Ngardi verbal system: -kujarra ‘while’ and -kariny ‘again’. These two inflections are somewhat anomalous and poorly attested. Both forms are homophonous with nominal inflections: the ‘dual’ suffix (§4.4.1) and the ‘other’ suffix (§4.3.4), respectively. Both forms attach outside the regular TAM inflections described in the preceding sections. Beginning with -kujarra, Cataldi (2011, p. 128) glosses it as an enclitic form =kujarra with three related senses: ‘around, all around, everywhere’, ‘while’, and ‘consequently’. A revised assessment of the data indicates that it likely has only temporal senses. Additionally, I analyse this form as a positionally restricted verbal suffix, in contrast to Cataldi’s enclitic analysis. All examples of this suffix involve attachment to a fully inflected verb but it never appears outside any other enclitics, nor does it co-occur with stems from any other word class. Examples interpreted with a ‘simultaneous’ temporal function are presented in (1023) and (1024), while the ‘consequently’ sense is presented in (1026). In its latter function, the translation provided by Cataldi (2011) is entirely parallel with the function of the prior tense subordinators -rla ‘locative’ (§11.3.3.1) and -ngkarla ‘sequential locative’ (§11.3.3.2) when attached to non-finite verb forms – the distinction here being that -kujarra attaches to a finite-inflected verb form.

414 

 7 Verbal morphology

(1023)

Maitbi=rna worried=ya-nanta-kujarra, jarla=rna ngawu. maybe=1sg.s worried=go-prs-while stomach=1sg.s sick ‘As I go about stressing, I feel sick to my stomach.’ (PSM: TEN1-2018_001-01: 798371_802450)

(1024)

Yalu-jangka-rla=yi nga=rlipa nga-lku jukurru nawu that-ela-loc=emph seq.aux=1pl.incl.s eat-pot orange now ma-nku-kujarra. eat-pot-while ‘After that, we will eat bush oranges as we pick them.’ (PRN: FT1_b: 245832_249448)

(1025)

Nyangurla-karra=wali ngapa minya waku karri-ju? when-?=then water this neg be-pot Ngu=rna=yanu ya-nku-kujarra jingka-ku maju. seq.aux=1sg.s=3pl.o go-pot-while child-dat really ‘When will this rain stop? I want to go for my children’. (PRN: FT1_b: 419616_427332)

(1026)

Yalu-rla=rnalu karri-nya, pi-nya-kujarra karri-nya=rnalu tirrip. that-loc=1pl.excl.s be-pst hit-pst-conseq be-pst=1pl.excl.s camp ‘We stayed there, having got (the goanna), we slept.’ (NMN: LC20a).

The second ‘outer’ verbal inflection, -kariny ‘reciprocal’, is less well understood. Identical in form to the nominal suffix ‘other’ (§4.3.4), it has been identified in complex conditional clauses and marks the event of the apodosis as being reciprocal in nature, as indicated by the translations in (1027) and (1028).361 (1027)

Kaji=yi=n pungku ngu=rna=ngku pu-nganta-kariny. condit=1sg.o=2sg.s hit-pot seq.aux=1sg.s=2sg.o hit-prs-recip ‘If you hit me, I’m hitting you back.’ (PSM: TEN1-2018_029-01: 2293256_2298295)

(1028)

Kaji=yi =n yu-ngku ngunyju, ngaju=rna=ngku condit=1sg.o=2sg.s give-pot tobacco 1sg=1sg.s=2sg.o yu-ngku-kariny=parda wirrirl. give-pot-recip=dub back ‘If you give me tobacco, I might give you some back again in return.’ (PSM: TEN1-2019_005-01: 1334071_1351347)

361 The function of -kariny ‘recip’ appears to be very similar to instances of -ra ‘again’ on verb forms in Jaru (see Tsunoda, 1981b, p. 209).

7.6 Verbal inflections in a regional perspective 

 415

For (1027), Marie Mudgedell provided an alternative protasis involving a potential verb form (pu-ngku-kariny) rather than a present tense verb form (pu-nganta-kariny). Both are considered grammatical. The latter constitutes an ‘intentional’ or ‘scheduling’ usage of the present tense, as described in §7.4.9.

7.6 Verbal inflections in a regional perspective The verbal inflections described for Ngardi in §§7.3–7.5 are particularly interesting when one considers the inflectional systems of neighbouring languages, both within Ngumpin-Yapa and Western Desert subgroups. Formally and semantically, the verbal inflectional system of Ngardi is structurally the most similar to its closest genealogical neighbours: both Eastern Walmajarri and Nyininy (Western Ngumpin). On the other hand, there are a number of features that are notably distinct from Western Ngumpin languages and resemble (at least historically) aspects of the Warlpiri (Yapa) verbal inflectional system. Finally, there are some distinct inflectional features that align closely with the northern Western Desert languages and are not found within any other Ngumpin-Yapa language. In this section, I will briefly survey each of these relationships in turn. Some of the key shared characteristics of the Western Ngumpin verbal systems (Ngardi, Nyininy, Wawarl (Jaru), Walmajarri) can be summarised as follows: – an aspectual distinction between ‘perfective’ (unmarked -N-) and ‘imperfective’ (-N.an-) tense/mood inflections – the use of -Nu as a non-finite inflection (and NOT past tense, compare and contrast Yapa languages) – a distinction between a general ‘present’ tense and a ‘customary/habitual’ (excludes Nyininy) – the use of {-Na, -Ni} as a past tense inflection – the presence of a narrative ‘tense’ {-ngurra, -nyirra, -nyurra} encoding temporally distant events (analysed as ‘completive’ in Walmajarri). Ngardi is nevertheless distinct from Western Ngumpin through the following features: – discrete inflections marking a formal distinction between irrealis (-Nngi) and imperative (-Nta) functions362 – a number of idiosyncratic TAM inflections appropriate to forms inflected for the directionals -rni ‘towards’ and -rra away – two productive associated motion suffixes -Nungka and -Ninja/-nyanja. 362 Eastern Walmajarri (as spoken at Billiluna/Mulan) lacks a discrete irrealis form -Nngi but uses the form -Cu rather than (-Ca) for irrealis/potential functions. Additionally, in the Gordon Downs dialect of Nyininy, a single verb form covers imperative and ‘irrealis/potential’ functions. However, in Wawarl Jaru (as spoken around Halls Creek today) there is a discrete irrealis form, as in Ngardi.

416 

 7 Verbal morphology

Ngardi shares the following features with Warlpiri to the exclusion of Western Ngumpin: – the presence of a ‘presentative’ form (-Nanya) in some part of the verbal paradigm – an inflection -Nu with a past tense value (only with the directional forms in Ngardi) – stem forming suffixes encoding associated motion. Table 97: Inflectional categories in Ngumpin-Yapa languages. Western Ngumpin

Distinct Narrative Distinct Habitual Irrealis/Imperative homophony Directional suffixes (any kind) Associated Motion suffixes

Eastern Ngumpin

Yapa

Ngardi

Walmajarri

Nyininy

Wawarl

Wanyjirra

Gurindji

Warlpiri

✓ ✓

✓ ✓

✓ –

✓ ✓

– –

– –

– –

✓

✓

–

–

✓

✓

✓

– –

✓

✓

–

–

✓

–

✓

–

✓

–

✓

✓

Shifting away from comparison of Ngardi within the Ngumpin-Yapa subgroup, there are at least two key aspects in which the Ngardi verbal inflectional system shows clear influence from Western Desert languages: the presence of the obligative -Cumarra and anticipatory -Nngirda inflections. Formally related inflections are not found in any other Ngumpin-Yapa languages363 but feature in neighbouring northern Western Desert languages Kukatja (Valiquette, 1993, p. 456), Wangkajunga (Jones, 2011, pp. 201–202) and Yulparija (Burridge, 1996, pp. 37–39). Indeed, Ngardi’s Western Desert neighbours, Yulparija, Wangkajunga, Kukatja, Pintupi and Manyjilyjarra all appear to have a number of ‘parasitic’ verbal inflections that comprise various increments to the irrealis form of a verb and serve to express more marked functional categories as: characteristic, admonitive, hypothetical, obligative and unrealised (Jones, 2011, pp. 198–202).364 The limited set of parasitic forms in Ngardi are presented alongside related forms in Western Desert languages in Table 98. Notice that the hortative is

363 There is a singular exception. The form pu-ngka-marra hit-irr-marra is identified in Walmajarri (Hudson, 1978, p. 31). Hudson describes it as a highly unusual instance of an evitative case suffix (‘projected reason’ in her terms) attached to an inflected irrealis form. Like the Ngardi form it is found in the context of a prior imminent event but lacks deontic force. 364 In Pintupi, the entire verbal inflectional system has been analysed as being built off of two ‘classifiers’ – those inflections built off what is formally analogous to just a bare conjugation maker – and those inflections built off a form that is partially analogous to a ‘potential inflection’ in languages to the north (Hansen & Hansen, 1978, p. 157).

7.6 Verbal inflections in a regional perspective 

 417

included here as well as a parasitic inflection; however, this inflection is also found throughout the Ngumpin languages. Table 98: Comparison of parasitic inflections built off an irrealis and/or potential inflection in Ngardi and neighbouring northern Western Desert languages (illustrated with ya- ‘go’). Ngumpin

Northern Western Desert

Increment

Ngardi

Yurlparija

-marra -ngaRa

ya-nkumarra ‘obligative’

ya-nkumarra ya-nkumarra ‘subjunctive’ ‘hypothetical’ ya-nkungarra ‘obligative’

-rda ~ -ra

ya-nanngirda ya-nkurta ‘anticipatory’366 ‘negative optative’

ya-nku-rta ‘warning’

–

–

[]-[fut]-rta ‘prohibitive’

-ra

ya-nkura ‘hortative’

ya-nkura ‘unrealised’

ya-nkura ‘want’367

–

ya-nkura ‘hortative’

ya-nkura ‘optative’

Wangkajunga Kukatja

Pintupi

ya-nkumara ya-nkumara ‘unreal’365 ‘subjunctive’

Manyjilyjarra

ya-nkungara ‘irrealis’

In Ngardi, the obligative only marks hypothetical events situated prior to reference time (x should have y’d). In Wangkajunga the same temporal restriction appears to be observed. The brief description of -Cumara provided by Valiquette (1993, p. 455) for Kukatja suggests it encodes a past attempt (failed action) – however Kukatja speakers Marie Mudgedell and Patrick Smith affirmed that a primary sense is (at least) obligation ‘should have x’ed’ and this would bring it in line with the description of the function of the cognate form in Wangkajunga (Jones, 2011).368 Shifting to the anticipatory, the Ngardi inflection -Nanngirda is clearly paralleled in function by the suffix -rta labelled ‘warning’ in Wangkajunga (Jones, 2011, p. 180). Formally it is claimed to ‘follow verbal inflections’; however, the available examples involve it appearing only as an increment to a ‘future/irrealis’ -lku. Finally, with respect to the encoding of directionality within the verbal system, while Ngardi encodes two directional distinctions via verbal inflection (‘hither’ and

365 Some speakers use -tjakumarra (glossed contrafactual) as an alternative to -nkumarra for the ‘unreal’ inflection in Kukatja (Valiquette, 1993, p. 456). 366 Here the ‘base’ inflection is the Ngardi ‘irrealis’ (i.e. -Nanngi) rather than the potential (-Cu). 367 This form is unlabelled by Valiquette but can be found in examples within the Kukatja dictionary with (at least) an optative function: yankura ‘(I) want to go’ (Valiquette, 1993, p. 353). 368 Further afield in Ngarla (Ngayarta, Pama-Nyungan: Westerlund, 2015), a possibly related form -nyamarta, -rnamarta is analysed as an anticipatory (i.e. primarily epistemic functions (deductive/ assumptive)). However, on the balance of available evidence, such functions in Ngardi appear to be covered by a separate suffix -Nanngirda (glossed antic) while -Cumarda (oblig) appears to have primarily deontic functions.

418 

 7 Verbal morphology

‘hence’) across a number of TAM categories, it nevertheless lacks the highly articulated system of directional proclitics found in neighbouring northern Western Desert languages which may distinguish up to four directional values. For example, in Kukatja, preverbal proclitics distinguish the following directional values: ‘away’, ‘toward’, ‘around’ and ‘across the field of vision’ (Valiquette, 1993, p. 460). In this respect Ngardi patterns more closely with the Ngumpin-Yapa languages – although there is considerable subgroup-internal variation on how the directional suffixes interact with other feature values of tense, aspect and mood.

7.7 Verbal reduplication Inflecting verbs in Ngardi do not exhibit productive reduplication of any kind (cf. Warlpiri (Nash, 1986, pp. 136–147)). Here, however, I report on one lexically restricted instance of reduplication found in the corpus involving an historical coverb-verb sequence (yapart=pu- ‘run=hit’) that has univerbated into a single inflecting verb (yapartu- run). Reduplication is of the first syllable and the onset of the second, that is, the first CVC sequence as illustrated by (1029). (1029)

Yap-yaparti-nya, mariyawu, wakurra=rlipa nya-lku. rdp-run-ps long_way neg=1pl.incl.s see-pot ‘He ran and ran a long way, we will be unable to see him.’ (PSM: TEN12018_009-04: 784040_787842)

The primary function of verbal reduplication appears to be one of intensification but with limited data it is difficult to rule out iterative, durative or pluractional functions. When I asked Patrick Smith about whether it was possible to say ‘yap-yap-yapartinya’, he responded as in (1030). (1030) Yala wirrpa. Big mob bin yap-yap-yaparti-nya. that many many pst rdp-rdp-run-pst ‘That’s lots. (Like) a large group ran and ran.’ (PSM: TEN1-2018_ 828999_832900).

8 Complex predication Complex predication is a type of verbal predication comprising two or more predicational elements which project a unitary syntactic structure (Butt, 2010, p.  2).369 The term ‘complex predicate’ subsumes a range of typologically and structurally diverse phenomena that have variously been defined in the typological and descriptive literature (see Amberber, Baker & Harvey, 2010; Baker & Harvey, 2010; Bowern, 2006; Butt, 1995, 2010). Two distinct constructions in Ngardi meet the criteria of complex predication: serial verb constructions (SVCs, §8.2) and complex verb constructions (CVCs, §8.3).370 SVCs are the rarer of the two constructions and comprise predicational heads from the same lexical class – that of the (inflecting) verb. CVCs are by far the more frequent in Ngardi and share many of the properties of what are often termed Light Verb or Coverb Constructions (Bowern, 2006; Butt & Geuder, 2001). In addition to the description of the syntax of complex predication, this chapter also includes an event-semantic analysis of the 20 or so inflecting verbs that combine productively with preverbs (and in some cases nominals and Kriol verbs) to form CVCs. Throughout §8.4, I group and analyse each of the productive verbs in semantically coherent ‘classes’ based on semi-formal criteria of co-patterning with sets of preverbs and features of argument and thematic structure. Evidence is presented for the claim that these verbs have a primary ‘event categorisation’ function, following SchultzeBerndt (2003).

8.1 Types of complex predication Both SVCs and CVCs can be classified as types of ‘complex predication’ due to the fact that they combine multiple predicational elements in the formation of a unitary syntactic structure. SVCs and CVCs differ in a number of significant ways. SVCs (§8.2) minimally comprise two finite verbs (i.e. the two predicational heads are sourced from the same lexical class). SVCs are also further defined in semantic, inflectional and prosodic terms (Aikhenvald & Dixon, 2006). Semantically, the sequence of ‘serial verbs’ is considered to conceptualise a single macro-event. Inflectionally, SVCs typically share a single TAM value (i.e. they share the same inflection) and they cannot be negated independently (i.e. they share polarity). Prosodically, the

369 Although cf. Baker and Harvey (2010) for a dissenting view. 370 A yet third type is that of serial coverb constructions which can be found in a number of Daly River languages (Hoffman, 2015a). While Ngardi permits more than a single preverb to be combined in a complex predicate, this is generally restricted to an adverbial preverb combining with a strong nexus coverb. This is distinct from the more productive coverb serialisation as in MalakMalak (Hoffman, 2015b). https://doi.org/10.1515/9783110752434-008

420 

 8 Complex predication

intonational properties of SVCs are consistent with monoclausal, simple verb counterparts. SVCs are further categorised based on their argument sharing properties: for example, ‘same-subject serialisations’ or ‘switch-subject serialisations’. All SVCs in Ngardi minimally share a subject but may differ as to whether additional arguments are shared by both predicational heads. CVCs (§8.3), unlike SVCs, are a type of multipartite predication involving lexemes from two distinct word classes. They combine a member of a large set of semantically rich lexical items which do not inflect for TAM (termed ‘preverbs’) and a member of a small set of lexical items which inflect for TAM and typically only have generic or event-schematic semantics (termed ‘verbs’). The resulting predicational structure is monoclausal and has only a single set of pronominal arguments. In Ngardi, all verbs which appear in CVCs also appear as simple, independent verbs. Two special subtypes of CVCs in Ngardi are also identified. These are CVCs in which a preverb is not used as the uninflecting lexical element, but instead either a nominal (§8.3.1.2) or a Kriol verbal loan (§8.3.1.4) is used in its place. Within Australia, complex predicates are a well-recognised areal feature of languages of northern Australia (Schultze-Berndt, 2000) and are found in a broad swathe of phylogenetically diverse languages stretching from the Pilbara across to the Gulf of Carpentaria, traversing both Pama-Nyungan and non-Pama-Nyungan languages alike (Bowern, 2006, 2014; Schultze-Berndt, 2003). The CVC subtype of complex predicate phenomena has been described in some detail for languages of the Jaminjungan group (Schultze-Berndt, 2000, 2003) and for Wardaman (Yangmanic: Wilson, 1999). Further south, CVCs are found in all members of the Ngumpin-Yapa subgroup and in many of the Western Desert languages (see Amberber et al. (2010) for a summary of the relevant literature).371 With respect to the southern boundary of this areal phenomena – within the Western Desert languages – the functional load of CVCs appears to gradually reduce while the use of SVCs appears to increase (McConvell & Laughren, 2004; McGregor, 2004; and Nash, 2008). SVCs in contrast have generally only attracted interest in languages outside of Australia – particularly in isolating languages of Asia, West Africa, as well as in the Americas and in Oceanic and Austronesian languages (Aikhenvald & Dixon, 2006; Crowley, 2002). SVCs, like CVCs, are also found in a number of Australian languages across northern Australia, with a particular concentration in languages of the southern Gulf of Carpentaria area, for example Waanyi (Garrwan: Laughren, 2016c, p. 90), Wambaya (Mirndi: Nordlinger, 1998), and Yukulta (Tangkic: Keen, 1983). Isolated instances also appear elsewhere across the country (e.g. Dyirbal in north-eastern Australia, Dixon, 2011). Additionally, the northern Western Desert languages (e.g. Wangkajunga: Jones, 2011; Kukatja: Valiquette, 1993; and Pintupi: Hansen & Hansen,

371 Complex predicates have also been described for languages well outside of the Northern area – for example Diyari (Austin, 1981a).

8.1 Types of complex predication 

 421

1978) are notable for possessing both types of complex predicates (both SVCs and CVCs). The use of SVCs in particular is described as being ‘extremely productive’ in these languages (e.g. Wangkajunga: Jones, 2011, p. 209) and ‘near ubiquitous’ in the case of Yankunytjatjara (Goddard, 1988, p. 177).372 Lastly, serialisation of either a ‘coverb-like’ or ‘verb-like’ element is also found throughout the Daly River languages (Western Arnhem Land). However, the details vary significantly between languages and the terms ‘coverb’ and verb’ are perhaps less appropriate for these languages (Nordlinger, 2017). Within Ngardi, complex verb constructions (CVCs) are the most pervasive and frequent form of predication, as per all other languages of the Ngumpin-Yapa subgroup. However, there are two aspects in which complex predication in Ngardi is somewhat atypical of the Ngumpin-Yapa subgroup. First, Ngardi is notable for being able to integrate nominals directly into complex verb constructions as if they were preverbs – but with a somewhat reduced set of inflecting verbs. The main examples of this type involve inflecting verbs which could be argued to have ‘grammatical’ or ‘derivational’ functions: karri- be (as a stative/inchoative) and ma- get (as a causative). Nominals may also occur in CVCs with a select number of additional inflecting verbs (e.g. pu- hit and marda- have) with less obvious ‘derivational’ functions. The ability for nominals to appear within CVCs with a select number of inflecting verbs is also shared with Walmajarri and, to some extent, Warlpiri. The ability of nominals to form complex predicates with inflecting verbs problematises the distinction between nominals and preverbs (in both Ngardi and Walmajarri) and, in some ways, bears some similarities to noun incorporation. Second, the (limited) presence of SVCs in Ngardi is highly anomalous for a Ngumpin-Yapa language. No other Ngumpin-Yapa language has reported the use of serial verbs as a form of complex predication and it is in fact one of the typological features by which they are generally distinguished from the Western Desert languages to the south. Indeed, the usage of SVCs by Ngardi speakers may potentially represent a multilingual or contact influence since many speakers of Ngardi have a northern Western Desert language within their multilingual repertoire. Having defined the range of complex predication types in Ngardi, the following sections detail and exemplify serial verb constructions (§8.3) and complex verb constructions (§8.3) in turn. The ability for nominals and Kriol verbs to be integrated into complex verb constructions is discussed separately in §8.3.1.2 and §8.3.1.4.

372 The details of serialisation in Yankunytjatjara differ slightly from what is observed in Ngardi, however. Serialisation is not accomplished by simply juxtaposing verbs inflected for any number of identical inflections. Instead, an invariable serial form of a verb is used.

422 

 8 Complex predication

8.2 Serial verb constructions Serial verb constructions (SVCs) in Ngardi are a type of complex predication in which two or more finite-inflected verbal predicates share a single set of pronominal clitics and form a single clause. Finite verbs in SVCs share the same TAM features although this feature-sharing does not extend to associated motion or directionality.373 Further positive evidence for their status as monoclausal SVCs, rather than a type of multiclausal construction, is the absence of any intonational break which might align with a clause boundary. With respect to their argument structure, all of the SVCs in Ngardi meet the definition of same-subject serialisation (Crowley, 2002, p. 40), that is, verbal predicates within SVCs all share identity of their subjects. All SVCs in Ngardi encode a close sequential temporal relationship between the two events depicted by the two verbal predicates. The clausal ordering of verbal predicates is always iconic of the sequence of events (cf. verb serialisation in Wambaya where anti-iconic ordering is possible (Nordlinger, 2014)). Furthermore, Ngardi serial verbs are ‘symmetrical’ (Aikhenvald, 2006) insofar as there are no stringent restrictions on the semantic content of each verb. Examples (1031)–(1034) represent a range of semantically diverse verbs whose combination is only constrained by logical, pragmatic constraints. (1031) Ma=rna=rla ka-ngku yu-ngku. top.aux=1sg.s=3sg.obl carry-pot give-pot I will carry and give it to him. (PSM: TEN1-2018_034-02: 1465289_1467646) (1032) Ya-nku=rna ngapa=wali pirda-kurlu-rlu nga-lku. go-pot=1sg.s water=then full-prop-erg eat-pot Full up (with food), I will go and drink some water. (TJL: TT76_2201: 1314805_1317736) (1033) Nga-rni=lu luk=wanti-nya. eat-pst=3pl.s lie_down=fall-pst They ate and lay down. (YDN: Manungka_022027) (1034) Pangi-rni=rna pi-nya. dig-pst=1sg.s hit-pst I dug and found it. (TJN: LC50a: 395720_399523) Ngardi exhibits both ‘tight’ and ‘loose’ serialisation (Goddard, 1988, p. 180). In tight serialisation (1031), no intervening material occurs between the two serialised pred-

373 Two minor exceptions to the co-identity of verbal TAM are discussed in this section.

8.2 Serial verb constructions 

 423

icates and neither predicate takes arguments or modifiers peculiar to itself. In loose serialisation, intervening material is permitted between the two serialised predicates, including bound pronouns (1032)–(1034). In loose serialisations, predicates may also select non-subject arguments peculiar to themselves. In (1035), for example, the reflexive object enclitic is only selected for by the transitive verb nya- see and not the initial verb ya- go. Similarly, in (1036), only the initial verb luwa- shoot takes the non-subject enclitic =rla in its predicational structure as a benefactive third person ‘for her’. (1035) Ya-nku=rna=nyanu wapirra nya-nku. go-pot=1sg.s=refl F see-pot I will go and see my father. (PSM: TEN1-2018_033- 1638160_1640436) (1036) Kuyi=pula=rla luwa-rni-nyirra… nga-rni-nyirra… paja-rni-nyirra. meat=3du.s=3sg.obl shoot-pst-narr eat-pst-narr bite-pst-narr The two of them speared meat for her, (they) ate some, and cut it up. (MMN: Manungka_01-035155: 1157783_1164377) The following examples also illustrate the possibility for a verb exhibiting associated motion morphology to be serialised with another verb (1037) as well as for two imperative verb forms to be serialised together (1038). (1037) Turt=ma-nungka-nya=rla yi-nya. break=get-x&move-pst=3sg.obl give-pst ‘He broke it and went and gave it to her.’ (BSE: TEN1-2017_017-01: 422641_424934) (1038) Turt=pa-nta=rla yu-ngka jirrilpaja. break=get-imp=3sg.obl give-imp carrot ‘Break up the bush carrot and give it to him.’ (MMN: TEN1-2017_002-02: 730024_733916) In all of the illustrations of SVCs observed so far, the verbs share identical TAM inflection. There are two exceptions in the data to this generalisation. The first is the serialisation of the potential and hortative verbs (1039), and the second is the serialisation of past and imperfective narrative verb forms, an example of which can be seen in (1065). I have no explanation for these beyond to suggest that this may provide evidence that the potential and hortative are perhaps more closely linked in their featural specification than I have analysed here; and that verbal aspect may not be subject to the same strict constraint of featural co-identity as tense or mood.

424 

 8 Complex predication

(1039) Yalu-rla=lku=rnalu kupa-ku, nga-lku, ya-nankura. that-loc=then=1pl.excl.s cook-pot eat-pot go-ipfv.hort ‘There then we will cook, (we) will eat and (we) will come back.’ (TJN: Manungka_01-035155_3a: 877472_881559

8.3 Complex verb constructions The canonical complex verb construction (CVC) in Ngardi involves the combination of a verb with at least one preverb in the formation of a single predicational structure.374 Preverbs are a heterogenous class of verbal words which do not inflect for tense, aspect, mood, directionality or associated motion. By this inflectional criterion, preverbs are clearly distinguished from verbs. Preverbs, like nominals, may take case marking in subordinating or secondary predicating functions, and so the distinction between preverbs and nominals can be somewhat less clear than the distinction between verbs and preverbs.375 On the one hand, there are some clear formal differences that apply to certain exemplars of preverbs and nominals. Most preverbs, for instance, are unable to take case marking in a relational function (i.e. they cannot function as arguments of main clauses). Like nominals (and nominalised verbs), most weak and adverbial preverbs can receive case marking either in a subordinating (§11.3.2) or secondary predicational (§9.3) function, relating a preverbal predicate to a range of arguments including transitive subjects (1040), intransitive subjects (1041) and objects (1042). (1040) Pila=ma-ni=wali yapart-u. follow=get-pst=then run-erg ‘He chased them, (he) in full flight.’ (PSM: TEN1-2018_010-02: 1260437_01276092) (1041) Pukany-wangu=rnalu ya-nku kaputa-rla Three-ways-wana, sleep-priv=1pl.excl.s go-pot dark-loc place_name-perl Camooweal-wana, marriyawu-rla. place_name-perl long-way-loc ‘We will go, sleepless, through the night, past Threeways and Camooweal, a long way.’ (PRN: TH1_0034-01)

374 Argument structure of clauses is discussed in much greater detail in Chapter 7. Of import here, however, is the observation that while preverbs generally do not project their own argument structure, there are some significant exceptions to this rule. These will be summarised in the relevant preverb subsections as well as in subsections detailing the various inflecting verbs where relevant. 375 This arises in no small part due to the fact that the class of nominals itself is functionally and semantically highly heterogenous – a feature of quite a number of Australian languages, as discussed by Dench (1995).

8.3 Complex verb constructions 

 425

(1042) Ngaringka-rlu nya-nganta pukany-kurra. woman-erg see-prs sleep-all ‘A woman watches (a child) sleeping.’ (PSM: TEN1-2018_010-02: 800376_804735) However, there are a small number of preverbs that nevertheless receive case marking in a clear relational function, such as ergative marking indexing a transitive subject, e.g. pukany-ju ‘sleep-erg’ in (1043): (1043) Pukany-ju ma=yi pi-nya. sleep-erg top.aux=1sg.o hit-pst ‘I am sleepy (lit. ‘sleep is biting me’).’ (PSM: TEN1-2018_029-01: 446878_449340) Due to their typical semantic content (expressing events or states), preverbs do not inflect for number (e.g. -kujarra dual) and cannot take any of the derivational morphology available to nominals (e.g. -jarra associative (§4.3.1) or -pirri instrument (§4.3.2)). However, this criterion does not distinguish preverbs from certain nominals which likewise denote events or states and rarely inflect for number. Therefore, while it is possible to point to exemplars within the class of nominals or exemplars within the class of preverbs, the boundary between the two is by no means clear cut. 8.3.1.1 Preverbs and preverb types Moving away from the distinction between preverbs and nominals, we can now also consider the differences within the class of preverbs and demarcate them into three subtypes: strong nexus preverbs (§8.3.1.1.1), weak nexus preverbs (§8.3.1.1.2) and adverbial preverbs (8.3.1.1.3).376 The central concept that is used to differentiate preverb types is ‘tightness of nexus’ (McConvell, 1996b; Schultze-Berndt, 2000, 536ff.), a term denoting the degree of morphosyntactic, semantic and phonological integration of a preverb with an inflecting verb. Importantly, these three subtypes are conceived as idealised positions along a continuum – they represent clusters of properties which serve as useful yardsticks for describing the variation in a formally and semantically diverse class of lexemes. A cline of preverb types has been recognised in all Ngumpin-Yapa languages on the basis of their variable phonological, semantic and morphosyntactic behaviour (see Osgarby (In press) for a summary of the relevant terminology).377

376 The term ‘adverbial preverb’ is adopted from Senge (2015, p. 145). 377 In other Ngumpin languages such as Bilinarra, certain morphology has been identified as uniquely selecting preverbal/coverbal hosts and thus serves as a further definitional criterion for preverbs (especially in contrast to adverbs). Ngardi lacks any morphology which uniquely selects preverbal hosts.

Grammatical case

Derivational suffix378

Number

Predicational Modifier

Predicational Head

Referential Modifier

Referential Head

√

√

√

ngaju ‘I’ nyuntu ‘you’

√

√

√

√

√

mu‘this’ yalu‘that’

Demonstra tive

√

√

√

√ √

√

√

√

√

√

√

√

kirda ‘big’ wuruju ‘good’

Adjective

√

√

pamarr ‘rock’ jingka ‘child’

Noun

√379

√

kaniny ‘inside’ jalangu ‘today’

Locational Temporal

√

√

√

wirlinyi ‘hunting’ tirrip ‘camping’

Action Nominal

√

√

ya-nu‘go’ pu-ngu‘hit’

Infinitive verb

√

√

√

jakarr ‘try’ wirrirl ‘back’

Adverbial preverb

√380

√

√

turt ‘in pieces’ purluk ‘submerge

Weak nexus

378 The various derivational suffixes actually show much more variable selection of lexical hosts than is afforded under a single label here. 379 This is limited to the derivational case -kariny deriving a spatial/temporal nominal. 380 This is limited to certain preverbs like pukany ‘sleep’ as illustrated in examples such as (1043)

Morphology

Syntax

Examples

Free pronoun

Table 99: Prototypical functions and features of nominals, infinitive verbs and preverbs.

(continued)

√

jut‘set_off’ luk‘lie’

Strong nexus

426   8 Complex predication

Reduplication

Resultative -karra

Secondary pred. case

Subordinating case

Derivational case

Nongrammatical case

Table 99 (continued)

√

√

Free pronoun

√

√

Demonstra tive

√

√

√

√

Noun

√

√

√

√

√

√

Adjective

√

√

√

√

Locational Temporal

?

√

√

√

√

Action Nominal

√

√

√

√

Infinitive verb

√

Adverbial preverb

√

√

√

√

√

Weak nexus

Strong nexus

8.3 Complex verb constructions   427

428 

 8 Complex predication

Table 99 summarises the various syntactic and morphological features that distinguish preverbs types and subtypes of preverbs from subtypes of nominals. The morphosyntactic criteria are adapted variously from Dench (1995), Meakins and Nordlinger (2014), and Osgarby (2018). The features of each preverb subclass are examined in subsequent subsections (§§8.3.1.1.1–8.3.1.1.3). It is important to note that while all three subtypes of ‘preverbs’ are grouped here as a single lexical class for descriptive purposes, there is the persistent problem that at least some strong nexus preverbs arguably lack the status of a fully independent lexical word since they violate the general disyllabic minimum word constraint. Examples include luk-, lap- and jup-, as previously listed in Table 16. These forms, and other strong nexus preverbs, can never occur without their accompanying verb with which they form a single phonological word (e.g. for the purposes of stress assignment). With these theoretical issues in mind, the following subsections detail the various features of the different preverb types. 8.3.1.1.1 Strong nexus preverbs Strong nexus preverbs are characterised by the following cluster of properties. They: – combine with a small number of ‘productive inflecting verbs’ (§8.4) which primarily have an ‘event-classification function’ – immediately precede the verb and do not permit any intervening clitics, bound pronouns or other morphology – are minimally heavy monosyllables (i.e. with codas or long vowels) – may not receive primary lexical stress. – generally add the bulk of the semantic information to the CVC (sometimes almost entirely so) – may modify the ‘expected’ thematic and syntactic roles associated with a verb in its simple form One of the main features of strong nexus preverbs is that they only combine with ‘productive’ inflecting verbs (§8.4). For example, the strong nexus preverb kily- ‘cool’ combines with only two verbs: ma- get and wanti- fall. In both cases, the major semantic structure comes from the strong nexus preverb and only a basic event schema and general argument structure is provided by the inflecting verb. In the case of kily=ma-, ma- get only contributes a causative function ‘x makes y (cold)’ as in (1044), while in kily=wanti-, wanti- projects an impersonal clausal structure ‘x becomes (cold)’ referring to the weather, as in (1045). (1044) Kily=ma-ni ma=yi pirriya-rlu. cold=get-pst top.aux=1sg.o cold-erg ‘The cold (weather/wind) is making me cold.’ (MMN: TEN1-2017_023-03: 1899675_1910236)

8.3 Complex verb constructions 

 429

(1045) Kily=wanti-nyanta. cool=fall-prs ‘It’s getting cold.’ (MMN: TEN1-2017_023-03: 2004519_2014229) Some strong nexus preverbs only co-occur with a single inflecting verb. The following preverbs, for example, are only found with the inflecting verb pu- hit (see §8.4.3.1). [i]    kuny ‘wait’ (1046) Mangarri-ku=rla ita-yita-pardu-rlu kuny=pu-nganta. veg_food-dat=3sg.obl small-rdp-dim-erg wait=hit-prs ‘A child is waiting for food.’ (TJL: TT76_2501: 2052310_2055473) [ii]    kart ‘tell, narrate’ (1047) Kart=pu-nganta=rna=ngku kuyi-kurlu. narrate=hit-prs=1sg.s=2sg.o meat-prop ‘I am telling you about the meat.’ (MMN: TEN1-2017_023-03: 1273845_1282082) [iii]    jup ‘dismount’ (1048) Minya nawu=rnalu jup=wanti-nya. this now=1pl.excl.s dismount=fall-pst ‘Now (in this picture) we got out of the car.’ (MMN: TEN1-2017_020-02: 136810_146173) A final feature of some strong nexus preverbs is their ability to project an argument structure which is generally not observed for the verb with which it combines when it occurs as a simple verb. For example, jut ‘set off’ is a strong nexus preverb than combines with pu- (which typically projects a transitive predicate argument structure in its simple form) to form a CVC that is intransitive, as in (1049). (1049) Yalu-jangka=rnalu jut=pi-nya nganayi-kurra. that-ela=1pl.excl.s set_off=hit-pst whatsit-all ‘After that we headed off (north) to whatsit place.’ (TJN: Manungka_01-035166: 551797_561991) 8.3.1.1.2 Weak nexus preverbs Weak nexus preverbs are characterised by the following key properties. They: – combine with (sometimes numerous) productive inflecting verbs (e.g. turt ‘in pieces’ in Table 100 and purluk ‘submerge’ in Table 101)

430 

– – – – –

 8 Complex predication

may be separated from the inflecting verb by bound pronouns (1050) but generally precede the verb receive a primary lexical stress (e.g. jaku /cáku/ ‘try’ in (1050)) are typically disyllabic or longer (e.g. jaku ‘try’) but may be heavy monosyllables (e.g. turt ‘in pieces’) can undergo reduplication with the function of pluractionality (1052) (some) can also function predicatively (e.g. pina ‘know’ in (1064)) and/or form a resulative secondary predicate when suffixed with -karra (e.g. pina-karra ‘know-result’ in (1065))

The ‘looser’ phonological and syntactic integration of weak nexus preverbs (in contrast to strong nexus preverbs) is primarily illustrated by the possibility for clausal material to intervene between preverb and verb, including the bound pronouns as in (1050). In these contexts, the preverb always receives a primary lexical stress. (1050) Jaku=lu yaparti-nya. depart=3pl.s run-pst ‘They took off.’ (PSM: TEN1-2017_009-04: 1004659_1006917) In terms of event semantics, CVCs formed with weak nexus preverbs exhibit more transparent semantic compositionality. This can be exemplified with two different preverbs. The first example, turt ‘in pieces’, specifies various events in which a figure undergoes an action resulting in it being distributed into multiple pieces. It combines with a range of verbs of location, verbs of force and some select additional verbs as presented in Table 100 and exemplified in (1051)–(1056). A secondary sense (or perhaps homophone) of turt, meaning ‘remove’, is also provided for completeness. Table 100: Attested combinations of weak nexus preverb turt ‘in pieces’ with various verbs. Verbs of Location#

turt ‘in pieces’ turt ‘remove’ #

Verbs of Force

Other

karribe

ma-/paget

pajachop

puhit

ladig

luwashoot

kumacut

ngaeat

‘be in two’

‘break in half’

‘bite/cut in half’

‘crush’

‘pierce and crush’

‘smash by throwing’

‘cut into ‘lose a pieces’ relative’

‘pull out’381

‘cut out’

This classification includes verbs of location and verbs of change of locative relation (see §8.4.1)

381 Eastern Walmajarri has a preverb turt- which only co-occurs with a get verb ma-/pa- and has a sense ‘pull out via roots’.

8.3 Complex verb constructions 

 431

[i] turt=karri(1051) Nyanungu=lku=pula=nyanu turt=karri-nya. recog1=then=3du.s=refl pieces=be-pst ‘Then those two themselves were twins.’ (TJN: LC21a: 1260752_1263364) [ii]    turt=luwa(1052) Luwa-rni karli-kurlu-rlu turt=luwa-rni yala turru. shoot-pst boomerang-prop-erg pieces=shoot-pst that bird ‘He hit it with a boomerang, he shattered its (leg).’ (MMN: TEN12017_017-1: 311556_318139) [iii] turt=paja(1053) Turt=paja-rni. pieces=bite-pst ‘He bit it in half.’ (MMN: TEN1-2017_017-01: 282578_285446) [iv] turt=pu(1054) Yalu-ngku ma turt=pi-nya kuyu. that-erg top.aux pieces=hit-pst meat ‘That one broke the meat in half (with his hands).’ (MMN: TEN1-2017_01701: 266731_269679) [v]   turt=nga(1055) Turt=nga-rni. pieces=eat-pst ‘She lost him (her husband).’ (Cataldi, 2011, p. 302) [vi] turt=kuma(1056) Kuyi kuma-rni ngantany-ju ngirnti turt=kuma-rni kuyi, kangaroo. meat cut-pst man-erg tail pieces=cut-pst meat kangaroo ‘The man cut the tail, he cut that meat in half, kangaroo.’ (PSM: TEN1-2017_01701: 207707_219809) Weak nexus preverbs are also generally able to undergo reduplication which serves a pluractional function – either indicating repetition/iteration of the event or plurality of a participant as in (1057). (1057) Yuwayi, turt-turt=paja-rnu-ku, lurrij. yes pieces-rdp=cut-inf-dat complete ‘Yes, in order to cut it all up into pieces.’ (MDN: LC22a: 1367133_1370685)

432 

 8 Complex predication

Another example of a weak nexus preverb is purluk which encodes a figure as (partially) submerged below a surface. It is integrated into a range of CVCs in a highly compositional fashion with verbs of location, verbs of change of locative relation as well as verbs of force, as summarised in Table 101 and as exemplified in (1058)–(1063). Table 101: Attested verbal combinations of weak nexus preverb purluk ‘submerge’. Verbs of location/change of locative relation purluk

Verbs of force

karri- be

wanti- fall

yirra- place

kiji- throw

la- pierce

‘swim, float around’

‘fall, dive into’

‘submerge, place within’

‘throw and plop’ ‘dunk, submerge’

‘pierce (sand, soft ground)’

[i] purluk=karri- ‘swim, float around’ (1058) Yala=lu purluk=karri-nyanta ngapa-ngka. that=3pl.s submerge=be-prs water-loc ‘They are lying/floating around in the water.’ (MMN: TEN1-2017_017-01: 947934_951713) [ii] purluk=wanti- ‘fall or dive into (water or fire)’ (1059) Purluk=wanti-nya=pula ngapa-ngka. submerge=fall-pst=3du.s water-loc ‘The two fell into the water.’ (MMN: TEN1-2017_016-02: 295136_297447) [iii] purluk=yirra- ‘submerge, place within (water or fire)’ (1060) Warlu-ngka=rna purluk=yirra-rni tipa. fire-loc=1sg.s submerge=place-pst billycan ‘I put the billycan in the fire.’ (MMN: TEN1-2017_017-01: 965008_969025) [iv]

purluk=kijia) ‘throw and plop’ (1061) Purluk=kiji-rni=rna fishingline ngapa-kurra. submerge=throw-pot=1sg.s fishing_line water-all ‘I threw the fishing line into the water.’ (MMN: TEN1-2017_017-01: 905117_909599) b) ‘dunk/submerge’ (1062) Yalu-jangka=rlipa purluk=kiji-ku ngapa-ngka. that-ela=1pl.incl.s submerge=throw-pot water-loc ‘Then we soak them (seeds) in water.’ (TJN: Cataldi, 2011, p. 273)

8.3 Complex verb constructions 

[v] (1063)

 433

purluk=la- ‘pierce’ Purluk=la-ni=lu, kana-kurlu-rlu kurlarda-kurlu-rlu. submerge=pierce-pst=3pl.s digging_stick-prop-erg spear-prop-erg ‘They pierced it (the soakage in a sandy creek bed) with a digging stick, with a spear.’ (MDN: LC48b)

Finally, certain weak nexus coverbs (e.g. pina ‘know’) can also function both as main clause predicates in verbless ascriptive clause types (1064), (§9.2.1.3) and as secondary predicates (§9.3), including resultatives when affixed with -karra result’ (1065).382 (1064) Pina=wali=rna mangarri-ku kupa-rnu-ku. know=then=1sg.s veg_food-dat cook-inf-dat ‘I know how to cook food.’ (MMN: TEN1-2018_012-02: 1202687_1208223) (1065)

Nga=rnalu ya-ni nya-ngani-nyirra pina-karra.383 seq.aux=1pl.excl.s go-pst see-ipfv.pst know-result ‘We wold go along watching, until we had learned.’ (MMJ: LC06)

Weak nexus preverbs in verbless ascriptive clauses can be contrasted with copula ascriptives (§9.2.2) as in (1066) where karri- be forms a CVC with pina which allows tense information to be added to the clausal predicate. (1066) Wakurra=rnalu=rla pina=karri-nya-ngurra yalu-ku neg=1pl.excl.s=3sg.obl knowledge=be-pst-narr that-dat jaru-ku. language-dat ‘We didn’t used to understand that language.’ (TJN: Manungka_02-02202)

382 In Walmajarri pina is a nominal with the meaning ‘ear’. This body part is metaphorically associated with cognition and the mind and can form CVCs meaning ‘to hear’, ‘to learn’, ‘to know’ and so forth. The semantic link between ear and knowledge is also found in other Australian Aboriginal languages. However, in Ngardi, the nominal lexeme for ‘ear’ is the distinct body part term langa. 383 This is a rare example of a serial verb construction (§8.2) in which the two verbal predicates differ in their tense inflection.

434 

 8 Complex predication

8.3.1.1.3 Adverbial preverbs Adverbial preverbs are characterised by the following properties. They: – are syntactically and phonologically independent from the inflecting verbs (1067)–(1069) – obey the disyllabic minimum word constraint in Ngardi (unlike strong nexus and some weak nexus preverbs) – receive a primary lexical stress as an independent prosodic word, e.g. jakarr /cákar/ ‘try’ in (1067)–(1069) – may occur preceding (1069) or following (1067) the verb, and be separated by bound pronouns (1068) – may combine with an existing preverb + verb combination (1067) – do not modify the predicate argument structure of the verb they combine with.384 (1067) Ngu=yi=lu pina=yu-ngani jakarr. seq.aux=1sg.o=3pl.s knowledge=give-ipfv.pst try ‘They were trying to teach me (Ngardi).’ (THS: TEN1-2017_003-05: 67499_73135) (1068) Jakarr ngu=rna pila=ma-nani-nyirra. try seq.aux=1sg.s follow=get-ipfv.pst-narr ‘I tried to follow him around (but failed).’ (PSM: TEN1-2017_009-04: 793303_795166) (1069) Jakarr ma=rna purda=nya-nganta Ngardi. try top.aux=1sg.s listen=see-prs language_name ‘I don’t understand/am unable to understand Ngardi.’ (PSM: TEN1-2017_0101: 2028997_2031458) Adverbial preverbs in Ngardi are functionally reminiscent of adverbs in other languages. Adverbial preverbs often serve to modify the spatial (1070), (1071), temporal (1072) or modal (1069) interpretations of an event. [i] wirrirl ‘return’ (1070) Yalu-jangka=rnalu kakarra wirrirl jungkuj=pi-nya. that-ela=1pl.excl.s east back depart=hit-pst ‘After that (the ceremony) we headed back east.’ (YMN: LC23a: 544718_547435)

384 The reader is referred to §9.2.4.4.2 for a discussion of the ability of the adverbial preverb jakarr- ‘try’ to modify the argument structure of a verb. This is an isolated feature of one preverb; adverbial preverbs generally do not modify the predicate argument structure of the verbs they combine with.

8.3 Complex verb constructions 

 435

[ii] jarrap ‘one leg crossed over another’ (1071) Minya jarrap yut=karri-nyanta. this leg_crossed sit=be-prs ‘This one, he is lying with one leg crossed over the other.’ (PSM: TEN1-2018_ 007-03: 3784255_3789735) [iii] jamun ‘soon’ (1072) Jamun=pa=rnalu yunpa-rni. soon=ep=1pl.excl.s sing-pst ‘We just started singing.’ (BSE: TEN1-2018_017-03: 1046804_1048284) While lacking any grounds for a discrete class of adverbs, adverbial preverbs share a number of properties with locational and temporal nominals which have adverbial functions. Temporal and location nominals are generally only distinguished from adverbial preverbs by the ability of the former to take spatial case-marking in a main clause (adjunct) function. 8.3.1.2 Nominals in CVCs In addition to preverbs, nominals in Ngardi may also combine with certain inflecting verbs to form complex predicates.385 The ability for nominals to directly combine with inflecting verbs without any morphological adjustment is a feature shared with Walmajarri (Ngumpin), Warlpiri (Yapa) and certain northern Western Desert languages such as Wangkajunga. This property of Ngardi complex predicates can be exemplified with the nominal walmak ‘lightning’ which appears as an argument in (1073) and as part of a CVC in (1074). (1073) Walmak-ta rayin=karri-nyani. lightning-loc afraid=be-ipfv.pst ‘He is frightened of lightning.’ (MMN: TEN1-2018_026-01: 395586_400370) (1074) Walmak=ma-nanta yala motika. lightning=get-prs that car ‘That car’s lights are flickering.’ (MMN: TEN1-2018_030-01: 1116573_1120614)

385 In some Ngumpin languages, there are specific derivational suffixes which attach to nominals to derive preverbs. This derivational morphology provides an avenue through which a lexical nominal root may enter into a complex predicate construction, e.g. -g in Wanyjirra and Jaru (Senge, 2015).

436 

 8 Complex predication

The majority of instances of nominals forming CVCs involve one of two inflecting verbs: transitive ma-/pa- get or intransitive karri- be. However, examples of nominal CVCs with other verbal predicates have been found. For example, the nominal jina ‘foot’, may combine with ya- go (1075) and marda- have (1076). (1075) Jina=rnalu ya-nani-nyirra.386 foot=1pl.excl.s go-ipfv.pst-narr ‘We travelled by foot.’ (BSE: TEN1-2017_014-01: 32897_34563) (1076) Nyuntu-ku=rna=ngku jina=marda-rnanta yut-warnu-rlu. 2sg-dat=1sg.s=2sg.o foot=have-prs sit-assoc2-erg ‘I am waiting for you while (I am) sitting down.’ (PSM: TEN1-2018_027-01: 2212757_2217355) While it may be possible to argue that jina is a direct object of marda- in (1076), with the compositional semantics of ‘to hold one’s feet’ being interpreted as ‘to wait’, the same cannot be said of (1075).387 Similarly, in Walmajarri a cognate lexeme jina ‘foot’ combines with verbs pu- ‘hit’ and -ma- ‘vblz’ to mean ‘track’ (Hudson, 1978, p. 46). This kind of complex verb formation is thus reminiscent of some of the properties of noun incorporation (see Mithun (1984, p. 855)). In addition to the various nominal types described above, action nominals (§5.8) can also form CVCs as per wurna ‘journey’ in (1077). (1077) Parli=pi-nya=rna kuriny jarany kuja=rna wurna=ya-nani. find=hit-pst=1sg.s two roughtail sub=1sg.s journey=go-ipfv.pst ‘I found two roughtail lizards while I was travelling.’ (MMN: TEN1-2018_05-01: 3144281_3150597) Action nominals are distinct from other nominals, however, in that they can freely be used as the main predicate in non-finite subordinate clauses (§11.3.1) as per jiilykarra ‘brand’ in (1078) and (1079), with or without an associated non-finite main verb. (1078) Jiilykarra=la-nu-jangka, ya-ni, yupa-rni, puliki. brand=pierce-inf-ela go-pst send-pst cattle ‘After branding, (they) went, and sent the cattle (away).’ (PSM: TEN1-2018_ 038-01: 544655_555812)

386 The Kriol term ‘putruk’ < English ‘foot walk’ is interchangeable with jina here. 387 That is, ya- go is an intransitive simple verb that does not subcategorise an absolutive object. Alternatively, one might argue that jina ‘foot’ is in a part-whole (appositional) construction with the subject: ‘we, the feet walked’.

8.3 Complex verb constructions 

 437

(1079) Yumpaly=pa=lu luk=karri-nya ngantany, jiilykarra-jangka tired=ep=3pl.s lie_down=be-pst man brand-ela puliki-jangka. cattle-ela ‘Tired from branding the cattle, the men slept.’ (PSM: TEN1-2018_038-01: 583016_585193) Action nominals may also function as the sole predicating element in a main clause as per jiilykarra ‘brand (an animal)’ in (1080) – however, note that a number of nominals also have this function in simple ascriptive clauses. (1080) Mayarra=lku=lu pirranginti, pirranginti jiilykarra nawu. again=then=3pl.s afternoon, afternoon brand now ‘Then in the afternoon, in the afternoon they (turned to) branding.’ (PSM: TEN1-2018_038-01: 967509_971097) 8.3.1.3 Issues of homophony: Nominals and preverbs, verbs and verbalisers Two inter-related issues are thrown up by the ambiguous nature of nominals in CVCs in Ngardi and other Ngumpin-Yapa languages. The first is whether or not ostensible nominals appearing in CVCs are true nominals or whether they should be analysed as homophonous preverbs. That is, one could propose two separate lexical items: for example, a nominal lexical root jina ‘foot’ and a lexical preverb jina ‘by foot’. This avoids positing that CVCs can comprise both preverbs and nominals alike. This approach is taken by Nash (1986) for Warlpiri, for example. In Ngardi such an analysis would generate a significant amount of homophony. Second, given that nominals can enter into CVCs but sometimes only with one or two inflecting verbs, there has also been the claim that it is the inflecting verb which exhibits formal homophony between verb and derivational verbaliser. In Warlpiri, for instance, bona fide nominals (those not argued to have preverb counterparts) can combine with one of two verb forms, both analysed as derivational ‘verbalisers’: -ma- ‘causative’ and -jarri- ‘inchoative’ (Nash, 1986, p. 42), the former being homophonous with a verb of the same conjugation ma- get. Jones (2011) similarly treats relevant inflecting verbs in Wangkajunga as homophonous.388 An alternative approach is found in Jaru, where Tsunoda (1981b, p. 122) identifies certain verbs which can

388 To be precise, Nash (1986, p. 42) analyses nominals used as preverbs in Warlpiri as a case of ‘double-bracketing’ of a lexical item’s word class. The nominal bracketing is always ‘primary’ or the ‘inner bracketing’ to which an outer preverb specification is added. Nash likens his analysis to ‘zero-derivation’. For Wangkajunga, Jones (2011, p. 158) analyses two homophonous ma- verbs: a maget verb in a complex verb (compound in her terms) and a ‘causative’ -ma- but acknowledges that distinguishing the two is difficult.

438 

 8 Complex predication

‘behave like grammatical verbs’ and labels them ‘semi-grammatical’ (p. 123). This approach views a single verbal lexeme as polyfunctional in nature and is also best fitted to the Ngardi data. Ngardi lacks evidence for any purely verbalising lexical verb forms. Unlike Warlpiri, Ngardi does not have any exclusively dependent verb forms but instead appears to be able to associate a subset of productive inflecting verbs with nominals in the formation of complex predicates.389 In Ngardi, the most frequently used verbs in CVCs involving nominals are ma-/ pa- get and karri- be. Importantly, there are no formal grounds for distinguishing instances of a ‘verbalising’ usage of these verbs from a ‘non-verbalising’ usage. Moreover, karri- be is unlike Warlpiri jarri- ‘inchoative’, in that it is an independent verb in its own right, functioning as a stative predicate. In these respects, CVCs in Ngardi appear to be typologically more similar to Walmajarri and Western Desert languages and less similar to Warlpiri or other Ngumpin languages. Wanyjirra and Jaru, for instance, require the suffixation of a derivational morpheme (-g ‘translative’ or ‘predicative’ with variants /-gaɻa(g), -waɻa(g)/) to nominals in order to form CVCs (Senge, 2015, pp. 594–596; Tsunoda, 1981b, p. 119). See, for example, the necessary derivation of the nominal yura ‘good’ (1081) in order to form CVCs in Wanyjirra and Jaru. In Ngardi the semantic equivalent can be expressed by directly associating the nominal wuruju ‘good’ with a verb ma- get (cognate with the Wanyjirra and Jaru verbs). (1081) a. Gangirriny-ju ngu=yi yura-g man-ana. sun-erg real=1sg.o good-trnsl get-prs ‘The sun makes me better.’ [Wanyjirra] (Senge, 2015, p. 106) b. Yura-g ma-ni. good-pred make-pst ‘He made him better.’ [Jaru] (Tsunoda, TT76_2601: fieldnotes) c. Nyuntu=yi=n wuruju=ma-ni, maparn-piya=wali. 2sg=1sg.o=2sg.s good=get-pst healer-sembl=then ‘You made me better, like a healer.’ [Ngardi] (TT76_2601: 544337_553138) In fact, Ngardi possesses no other productive verb formation processes of the type nominal > verb, zero derivation or otherwise. Instead, there is only productive complex verb formation through the association of lexemes from various word classes including nominals with an inflecting verb. In cases where nominals are integrated into complex predicates, the nominal provides the core substantive semantics while the inflecting verb provides only schematic information: specifically, Aktionsart and

389 Instances of pa- (treated as an allomorph of ma- get) could be considered an exception to this but, like Walmajarri pa-, the data point to a phonological basis to the variation between ma- and paand as such the allomorphic analysis is preferred.

8.3 Complex verb constructions 

 439

the predicate-argument structure of the complex predicate. The majority of these nominals depict states, as in (1081). The Ngardi verbs ma-/pa- and karri- have essentially causative and inchoative functions in these contexts but these meanings are not a direct property of the verbs themselves; rather they are a result of their integration into a CVC. (1082) Kirda=karri-nya=rna Gordon Down-ta. big=be-pst=1sg.s place_name-loc ‘I grew up on Gordon Downs Station.’ (MMN: TEN1-2017_029-02: fieldnotes) (1083) Yuwayi, ngaju-ku=yi kirda=ma-ni jamirdi-rlu, ngaju-punta-rlu. yes 1sg-dat=1sg.o big=get-pst MF-erg 1sg-pposs-erg ‘Yeah, my own grandfather brought me up.’ (BSE: TEN1-2017_014-01: 10437_16380) While the nominal kirda clearly forms a complex predicate with both karri- and ma- in (1082) and (1083), it has not been observed productively forming a complex predicate with other inflecting verbs. Attempts to elicit such combinations typically resulted in kirda being interpreted as an argument of a simple verbal predicate, as in (1084). (1084) Kirda=rna marda-rnanta. big=1sg.s have-prs ‘I’ve got a big (bit) of (tobacco).’ (MMN: TEN1-2017_023-03: 2194434_2202440) The following tables provide some examples of nominals combining with the semi-grammatical verbs ma- get and karri- be (Table 102), and nominals combining with other less productive verbs (Table 103). Table 102: Nominals found in complex predicates involving ‘semi-grammatical verbs’ ma- get and karri- be. Nominal

Meaning

CVC combinations

Meaning

parnkaj

‘open’

parnkaj=karriparnkaj=pa/ma-

kirda

‘big, senior, important’

ngayirra

‘sorrowful, worried ’

tapu

‘stubborn’

kirda=karrikirda=mangayirra=karringayirra=matapu=ma-

‘be open, available’ ‘open something; let something out’ ‘grow up’ ‘grow someone up’ ‘be in a state of worry, dejection’ ‘worrying about someone/thing’ ‘keep, retain’

wuruju

‘good, healthy’

wuruju=karriwuruju=ma-

‘feel good, be good’ ‘heal, fix, improve’

440 

 8 Complex predication

Table 103: Examples of nominals found in complex predicates with additional verb types. Nominal

Meaning

CVC combinations

Verb gloss

Meaning

yarr

‘atop’

yarr=pu-

hit

‘make tall, stretch’

ruju

‘play, corroboree’

ruju=wangkaruju=karri-

speak be

‘make fun of (verbally)’ ‘play’

rayin

‘fear’

rayin=karrirayin=nyarayin=yapartu-

be see run

‘be frightened’ ‘watch in fear’ ‘flee in fear’

raly

‘hair, fur’

raly=puraly=nya-

be see

‘pluck’ ‘envy, be jealous of’

kawurr

‘hot, warm’

kawurr=karrikawurr=pu-

be hit

‘get/become hot’ ‘make (something) hot’

tipiny

‘wooden stick’

tipiny=pu-

hit

‘skewer/pierce’

jaru

‘word, language’

jaru=yu-

give

‘instruct, tell, farewell’

kakarra

‘east’

kakarra=wantikakarra=karri-

fall be

‘turn east’ ‘shifts/extends east’

8.3.1.4 Kriol verbs in CVCs Kriol verbs, like nominals, can also enter into CVCs in Ngardi. This type of verb borrowing into a complex predicate structure, kown as a ‘light verb strategy’ (Wohlgemuth, 2009, p. 184) is widely reported in the literature on verb borrowings. Examples of Kriol loan borrowings are italicised in (1085) and (1086). (1085) Marrka=ma-ni boltimap=ma-ni lurrij. strong=get-pst bolt.tr.compl=get-pst complete ‘He tightened it, he tightened it up completely.’ (YMN: LC24b: 273624_275992) (1086) Kaontim=ma-nta=yi=lu, kuraj=pu-ngka=yi=lu. count.tr=get-imp=1sg.o=pl.s count=hit-imp=1sg.o=pl.s ‘Count them for me, count them for me.’ (Cataldi, 2011, p. 142) There appears to be a clear grammatical and functional equivalence between Kriol verbal loans and Ngardi preverbs or nominals. Speakers appear to be aware of these correspondences, especially given the frequency of sequential clauses involving monolingual and code-switched complex verb constructions. For example, in (1085), marrka is a Ngardi nominal and boltimap is a Kriol verb while in (1086) kaontim is a Kriol verb and kuraj is a Ngardi preverb.

8.3 Complex verb constructions 

 441

During dictionary checking sessions, Patrick Smith often provided a definition of a Ngardi preverb + verb combination by providing an equivalent Kriol verb within a CVC. Examples are: jarrara ‘approach’ (Ngardi) and rijim (Kriol) in (1087), and lintij ‘sharpen’ (Ngardi) and japinim (Kriol) in (1088). (1087) a. Jarrara=ma-nku=rna. approach=get-pot=1sg.s ‘I will catch up to him.’ (PSM: TEN1-2018_035-02: 520583_522301) b. Rijim=ma-nku=rna. catch=get-pot=1sg.s ‘I will catch up to him.’ (PSM: TEN1-2018_035-02: 522115_524693) (1088) a. Lintij=pu-nganta yala. sharpen=hit-prs that ‘He is sharpening that one.’ (PSM: TEN1-2018_035-02: 229423_231164) b. Japinim=ma-nanta jangu. sharpen.tr=get-prs indeed ‘He is sharpening it, that’s right.’ (PSM: TEN1-2018_035-02: 236389_237550) As demonstrated in (1086)–(1088), the choice of (Ngardi) verb in the mixed KriolNgardi verb complex does not always correspond to an equivalent monolingual Ngardi CVC. This is especially true given the tendency for borrowed Kriol verbs to primarily co-occur with the basic transitive get verb ma- when forming transitive bilingual CVCs. While all Ngumpin-Yapa languages have been found to borrow Kriol verbs using a light-verb strategy, the parameters of this borrowing vary across the subgroup (Bavin & Shopen, 1985; Hamilton & Ennever, 2017; O’Shannessy & Meakins, 2012). Gurindji (Ngumpin), for instance, integrates Kriol loans as if they were native coverbs. They combine with the full range of inflecting verbs (typically selecting semantically compatible inflecting verbs) and can occur in the same range of clause positions as weak nexus coverbs (before and after the verb). This can be contrasted with Warlpiri and Warlmanpa, where Kriol verbs are treated as if they were functionally nominals: they only enter into light verb constructions via the dependent (or ‘syntactic’) verbs =jarri (‘inchoative’; =ja in Warlmanpa) and =ma (‘causative’) and must occur in the immediately preverbal position. Mudburra represents a more extreme case yet. Kriol verbs are integrated with just a single Mudburra verb: ka- which functions essentially as an auxiliary: providing nothing more than a host for verbal inflection. Semantics, transitivity and lexical aspect are all provided by the borrowed Kriol verb. Ngardi appears to occupy the middle ground on this spectrum. Kriol verbs are integrated with a reduced subset of Ngardi inflecting verbs. Both karri- and ma-, which are used ‘monolingually’ to derive CVCs from nominals, are the most frequently used Ngardi verbs used when integrating Kriol loan verbs. However, as many as eight other inflecting verbs have also been observed combining with Kriol loan verbs. The

442 

 8 Complex predication

number of Kriol loans and their co-occurrences with Ngardi verbs in the corpus is provided in Table 104 (Hamilton & Ennever, 2017). Table 104: Number of different Kriol loan verbs combined with Ngardi verbs. Transitive verbs Verb # of Kriol verbs

mado >55

pi- ~ puhit 9

yirraplace 3

Intransitive verbs ka carry 2

yago 9

karribe 7

wantifall 2

pangidig 1

Like strong nexus preverbs, Kriol verbs are found only in the immediately preverbal position. The inflecting verb nevertheless contributes transitivity, basic event types and sometimes some associated semantics. This situation of Kriol loan verb integration with a reduced subset of inflecting verbs is mirrored in Wanyjirra (Senge, 2015, p. 469). Examples (1089) and (1090) constitute a rare example of a transitivity contrast in Kriol verb integration. (1089) Pinij=karri-nya. complete=be-pst ‘It was finished.’ (PLN: LC47a: 297190_298502) (1090) Kaji=rnalu pinijim=ma-ni juju... condit=1pl.excl.s complete.tr=get-pst ceremony ‘When we finished the ceremony...’ (LC23b: 733036_737583) Examples of Kriol verbs in combination with the other inflecting verbs are provided in (1091)–(1095). These examples illustrate that the meaning of the complex verb construction is sometimes determined not only by the loan verb but is also compatible with the core meaning of the Ngardi inflecting verb. (1091) Minya ngantany, motika pujim=yirra-rnanta.390 this man car push.tr=place-prs ‘This man, he is pushing the car.’ (PSM: TEN1-2017_016-03: 673763_688806) (1092) Yalu-ngulu juwag rolimap=pi-nya. that-abl swag roll.tr.compl=hit-pst ‘Then he rolled up his swag.’ (LC23a: 964589_966412)

390 The absence of ergative marking on the NP [minya ngantany], coupled with the presence of a pause, indicates that this NP is extra-clausal.

8.3 Complex verb constructions 

 443

(1093) Jarri-nya=rlanyanta yala ngari, nyanungu-rlu wali take-pst=3sg.lct that belongings recog1-erg indeed wirim=ka-nganta. wear.tr=carry-prs ‘He took the clothes from him, (and) he is now wearing them.’ (MMN: TEN1-2017_006-01: 1190901_1194546) In some cases, the compositional semantics of the Kriol verb and the Ngardi inflecting verb are less transparent. Thus, contrast the motion event in (1094) (which logically combines ranawei ‘run away’ with the ya- go verb) with the achievement event in (1095) (which combines win ‘win’ with the same verb, ya- go). This latter example is also somewhat unusual insofar as the second clause exhibits another loan word, sport, which is assumedly functioning as a direct object of the mixed verbal predicate win=ya-ni. In monolingual contexts, ya- go cannot independently select a direct object. (1094) Nyanyi, ranawei=ya-ni. before run_away=go-pst ‘A long time ago, she ran away.’ (PSM: TEN1-2016_005-03: 36343_40778) (1095) Wakurra=rnalu win=ya-ni Balgo na, no Halls Creek neg=1pl.excl.s win=go-pst place_name now neg place_name ngana=lu win=ya-ni sport. who=3pl.s win-go-pst sport ‘We, Balgo, we didn’t win, no, it was Hall’s Creek who won the sporting contest.’ (MMN: TEN01-2017_014-02: 319174_323975) The corpus shows variation as to whether the Kriol verb is integrated with its transitive marker -im when combing with a transitive Ngardi verb. Compare the presence of transitive -im as part of the Kriol verbal borrowing in (1096) and its absence in (1097). (1096) Halpim=ma-nani=rna dayiningal-u-rla, werak-kurra. help.tr=get-ipfv.pst=1sg.s dining.hall-ep-loc work-all ‘I helped them in the dining hall, working.’ (KAP: LC43_20001125_b: 155826_160470) (1097) Ngu=lu=yanu halp=ma-ni Balgo-mob. seq.aux=3pl.s=3pl.o help=get-pst place_name-group ‘The Balgo mob helped them.’ (BST: TEN1-2016_005-01: 491987_494187) While variation in the presence of transitive -im within Kriol verbal borrowings occurs in Ngardi transitive clauses, no examples of Kriol verbs with the transitive -im suffix

444 

 8 Complex predication

are found combined with intransitive Ngardi verbs. This is also reported to be the case for similar verb borrowings in Walmjarri (Hudson, 1978, p. 175). Certain other examples of Kriol morphology can be identified within mixed complex verbs in Ngardi but these too appear to be examples of frozen morphology. The Kriol completive suffix -ap (Hudson, 1983; Meakins & O’Shannessy, 2016; Schultze-Berndt, Meakins & Angelo, 2013) occurs infrequently (see (1085)–(1092)), while another aspectual suffix -aot is only found once (1098). Other Kriol verb forms that resemble phrasal verbs in English have been observed being borrowed wholesale: for example layidimku < ‘let him go’, lukapta < ‘look after’, dernaran < ‘turn around’, jingkabat < ‘think about’. I have observed no instances of the Kriol ‘continuative’ suffix -bat being borrowed in with a Kriol verb to form part of a code-switched complex predicate construction. (1098) Takimaot=ma-nanya motika, lurrij. take.tr.compl=get-cust car complete ‘He would service the car, and that’s it done.’ (YMN: LC23b: 267992_272965) There are no examples of borrowed Kriol verbs taking any case markers in subordinating contexts. One partial exception is the presence of a Kriol loan found inflected with the simultaneous subordinator suffix -karra in (1099). In this context however, the function of the suffix appears more akin to -karra in other Ngumpin languages where it functions as an aspectual preverb suffix (Meakins, 2016). (1099) Ngu=rna jart-karra=kiji-rni, warlu. seq.aux=1sg.s start-ssub=throw-pst fire ‘Then I will start a fire.’ (TJL: TT76_2201: 1326474_1330369) Finally, an interesting type of ‘double marking’ has been observed in a code-switched clause (1100). Here, features of both tense and person/argument status are rendered in both Kriol and Ngardi in discrete morphological exponents: past: {bin, -nya} and first singular subject: {ai, =rna} in Kriol and Ngardi respectively. This kind of ‘double, bilingual marking’ is rare. (1100) Ai bin born=wanti-nya=rna. 1sg.s pst born=fall-pst=1sg.s ‘I was born.’ (DJD: TT76_2201: 2134293_2136836)

8.4 Productive inflecting verbs A difficult task in the description of complex verb constructions in Ngardi is accounting for the highly productive distribution and functionality of some inflecting verbs. On the one hand, Ngardi possesses a large set of verbs which are relatively non-

8.4 Productive inflecting verbs 

 445

productive – such as ngarlarri- laugh. These non-productive verbs can be assigned a straightforward gloss and can be translated in a fairly consistent manner. However, the productive inflecting verbs occur in a much wider array of semantic contexts and it is more difficult to isolate their core semantic contribution to the event depicted by a complex predicate. This section will briefly attempt to illustrate some of the semantic variation observed in the ‘productive inflecting verbs’ in Ngardi. The approach I take here follows previous work that has emphasised the ‘event categorisation’ function of inflecting verbs in similar complex verb constructions in typologically diverse Australian languages (e.g. Jaminjung: Schultze-Berndt, 2003; Wagiman: Wilson, 1999; Bilinarra: Meakins & Nordlinger, 2014; and Wangkajunga: Jones, 2011; see also McGregor’s (2002, pp. 101–147) survey across language families). To begin, some additional terminological comments are warranted. Throughout this section I will make reference to ‘Participants’. I take Participants to be inherent to individual predicative lexemes. As a matter of practicality, the labels which I apply are generally selected so as to adequately apply to a class of verbs (e.g. Recipient, Undergoer) but I do not appeal to any universal set of ‘thematic roles’ in the description of different Participants. The labels deployed are only intended as shorthand for referring to predicate-specific roles (e.g. ‘giver’, ‘narrator’, ‘receiver’) (see Marantz, 1981) – which may on occasion be used for clarity. Such predicate-specific labels have also been termed ‘individual thematic roles’ (Dowty, 1989). The following list of 19 verbs are those that are classified as ‘productive inflecting verbs’. These are verbs that systematically form complex predicates with strong/weak nexus preverbs, and/or (action) nominals. They are divided into subgroups which broadly categorise the semantics of the event type they depict but are based predominantly on formal evidence. The primary evidence is combinatorial in nature in that it involves examining the (semantic) types of preverbs with which an inflecting verb combines. A secondary criterion is that of argument structure. Each subgroup generally exhibits significant overlap in possible argument structures even though there is a degree of within-group variation. The resulting groups of productive verbs are presented in Table 105. Table 105: Productive inflecting verbs. Verbs of location and change of locative relation karrikijimamarda-

be throw get have

wantiyirrayu-

fall place give

446 

 8 Complex predication

Table 105 (continued) Verbs of Locomotion yayapartu-

go run

ka-

carry

pierce shoot

pajapu-

bite hit

ngarriwangka-

narrate speak

Verbs of Force laluwa-

Other Productive verbs nganya-

eat see

Non-productive verbs (those excluded from Table 105) include certain verbs which are identified with just one or two preverbs (e.g. kuma- cut in turt=kuma- ‘break into pieces’, and kupa- cook in turn=kupa- ‘bake in ground oven’ and karrkurr=kupa- ‘cook right through’) and those verbs which only combine with preverbs of the ‘adverbial’ type (e.g. wirrirl ‘around’ or jakarr ‘try’). Non-productive verbs typically do not display any substantial deviation from their core semantics and can be provided with consistent translations barring occasional idiomatic usages. In the subsections that follow, I set out the range of event types classified by each class of productive verb. Each verb is examined in a separate subsection and occasional further subsections are provided for additional senses or event types associated with a single verb. At the close of each subsection I provide tabulated summaries of the types of complex verb constructions in which each productive inflecting verb appears.

8.4.1 Verbs of location, change of locative relation and possession A selection of seven verbs have been grouped in this section: karri- be (§8.4.1.1); ma- get (§8.4.1.2), marda- have (§8.4.1.3), wanti- fall (§8.4.1.4), yirra- place (§8.4.1.5), kiji- throw (§8.4.1.6) and yu- give (§8.4.1.7). A similar grouping of semantically related event types is proposed for Jaminjung (Schultze-Berndt, 2000, p. 220). This group of verbs shares close semantic and formal similarities relating to the expression of location/existence (e.g. karri-) and those expressing possession (e.g. marda- have and ma- get). In addition, in regard to the three verbs wanti- fall, yirra- place and yu- give, it will all be argued that they share a common semantic component of ‘change of locative relation’. The formal relationship of these seven inflecting verbs can be observed in their integration into CVCs involving preverbs that encode the configuration of a figure

8.4 Productive inflecting verbs 

 447

with respect to a location.391 Many postural preverbs occur with most members in this class. For example: pikala ‘support on head, support above’ combines with at least ma- (1101), marda- (1102) and yirra- (1103). (1101) Kankarra pikala=ma-nungka-nya. up.a on_head=get-x&move-pst ‘She got (the water) up on her head and went.’ (PPN: Manungka_01-035160: 1580047_1581580) (1102) Yatalayari, puju-kulu, wirtpala-kulu, nyanka-ngka mountain_devil spine-prop horn-prop neck-loc pikala=marda-rnanta. on_head =have-prs ‘The mountain devil has spines and horns which it carries across the head on the neck.’ (Cataldi, 2011, p. 188) (1103) Mary, Nakarra-ku=pa=rlangu, and Jajayi-ku=pa=rlangu name ♀subsection-dat=ep=also conj name-dat=ep=also pikala=yirra-rni. on_head=place-pst ‘He put Mary and Nakarra’s and Jajayi’s (father) on top of the camel.’ (YDN: Cataldi, 2011, p. 48) Verbs of this class are also typified by the fact that they may co-occur with a (spatial) case-marked NP denoting the (end) location of the figure depicted in the event. Examples for each include: karri- (1108), ma- (1115), marda- (1129), wanti- (1135), yirra- (1145), kiji- (1157) and yu- (1161).392 Each of these verbs has a number of additional senses which diverge in various ways from their locational sense. These will be described in turn (§§8.4.1.1–8.4.1.7). 8.4.1.1 karri- be The intransitive verb karri-, when appearing as a simple verb, categorises an event in which the subject is at rest at a location. It is a ‘neutral’ positional verb and is neither specified for ‘posture’ (e.g. ‘sit’, ‘stand’, ‘crouch’) nor is it specified for a particular location type (e.g. under, on). It is frequently used to predicate existence (i.e. underspecified location) in an existential construction ‘there is x’ as in the

391 I follow Schultze-Berndt (2000) in using Talmy’s (1985, p. 61) concept of ‘Figure’, which refers to a ‘participant which is located’. 392 Although it is more common for yu- give to take dative-marked NPs denoting the (generally animate) endpoint of motion.

448 

 8 Complex predication

subordinate clause in (1104) or prolonged presence in a location (cf. English ‘stay’) as in (1105). (1104) La-nku=rna kaji wirrpa karri-ju. pierce-pot=1sg.s condit many be-pot ‘I will dig until there are a large number (of bush potatoes).’ (JLI: TT76_2201: 1418550_1425505 (1105) Karri-nyani-nyirra=rnalu naya-naya. be-ipfv.pst-narr=1pl.excl.s always-rdp ‘We stayed there for a long time.’ (UMN: LC23a: 1768063_1773718) In the formation of CVCs, karri- be may combine with preverbs of spatial configuration as per the preverb yut ‘sit’ in (1106). (1106) Yut=karri-nyanta=lu wirrpa. sit=be-prs=3pl.s many ‘A big group is sitting down.’ (PSM: TEN1-2017_015-02: 681290_683194) 8.4.1.1.1 Stative and inchoative functions In addition to combining with preverbs to categorise events of spatial configuration, karri- also freely combines with most nominals. In some Ngumpin-Yapa and Western Desert languages, a formally similar ‘verbaliser’ has been termed an ‘inchoative’ (e.g. -jarri- in Walmajarri (Hudson, 1978, p. 45) and -arri-, -rri- in Wangkajunga (Jones, 2011, p. 172)). The range of nominals with which karri- combines is broad but frequently involves those describing states such as nyurnu ‘sick, dead’ (1107) and ngurrku ‘happy’ (1108); attributive nominals, such as kirda ‘big, senior’ (1082); temporal/spatial nominals, such as warruru ‘night’ (1109); ignoratives such as nyarrpa ‘what’ (1110); and cardinals, such as karla ‘west’ (1111). (1107) Ngaju ma=yi kunyarr nyurnu=karri-nya. 1sg top.aux=1sg.o dog dead=be-pst ‘My dog died.’ (TJL: TT76_2201: 782815_784951) (1108) Ngurrara-rla ngurrku=karri-nya=pula. country-loc happy=be-pst=3du.s ‘Those two were happy on country.’ (MMN: TEN1-2019_005-06: 372085_374268) (1109) Karri-nya=rnalu warurru=karri-nya. be-pst=1pl.excl.s night=be-pst ‘We stayed and it became dark.’ (TJN: Manungka_02-02202)

8.4 Productive inflecting verbs 

 449

(1110) Nyarrpa=karri-nya=pula? what=be-pst=3du.s ‘What did those two do?’ (NMN: LC21b: 523737_530123) (1111) Jungarni=wali, mayarra=lku karla=karri-nya. straight=then again=then west=be-pst ‘Straight then this time (it needs to reverse), but yet again (the vehicle) turned west.’ (PSM: TEN1-2018_029-01: 999467_1023059) There are a small number of eventualities categorised by karri- which represent events rather than states or incipient states (inchoatives). These are primarily events involving bodily functions such as nimij=karri- ‘blink’ or ‘wink’ (1112) and appear to be alternatives to CVCs involving expected ma- get or pu- hit (i.e. nimij=ma- and nimij=pu- are also both attested without any obvious meaning differences). (1112) Nimij=karri-nyanta. blink=be-prs ‘He is blinking.’ (MMN: TEN1-2018_035-02: 634291_636334) 8.4.1.1.2 Idiomatic usages Examples of idiomatic extension of event types categorised by karri- include the postural predicate lapart=karri- in (1113). (1113) Jarla-kariny ma=rna lapart=karri-nyanta. stomach-other top.aux=1sg.s lie_flat=be-prs ‘I am having a change of heart (lit. ‘my own stomach is lying down’).’ (LC42b: 577414_580566) 8.4.1.1.3 Summary of CVCs States (general) jaa=karrijalukuly=karrijany-jany=karrijarr=karrilurrij=karri-

abs is open abs is broken abs is stuck abs is stopped abs is finished

ngampurr=karriparnki=karriturr=karriwuruju=karriwuruly=karri-

abs is careful abs is awake abs is all together abs is good, healthy abs is quiet

450 

 8 Complex predication

States (postural) lapart=karriluk=karri-

abs lies flat abs lies down

pirdiny=karriyut=karri-

abs gets up abs sits

States (emotional) minyirri=karrinurrku=karri-

abs is ashamed abs is happy

ngunykarra=karrirayin=karri-

abs is sulking abs is afraid

Inchoative (sample only) kirda=karritalji=karri-

abs grows up/becomes big abs becomes stiff

Bodily activities ngurrkarra=karrinimij=karri-

abs snores abs blinks/winks

Ignorative nyarrpa=karri-

what happen(ed)?

Idiomatic (jarla) lapart=karri-

abs has a change of heart

8.4.1.2 ma- get The verb ma- get in its simple form has a basic meaning that relates to affectedness and contact. This verb is highly polyfunctional and covers the following basic event types: affectedness and contact (§8.4.1.2.1), caused state (§8.4.1.2.2), non-physical interaction (§8.4.1.2.3), sound emission (§8.4.1.2.4), a verbaliser for loan words (§8.4.1.2.5) and idiomatic usages (§8.4.1.2.5). Examples and a summary of CVC types are presented in §8.4.1.2.7. 8.4.1.2.1 Affectedness and contact The verb ma- can be used to describe events in which the subject ‘acquires’ an entity either conceptualised as possession or by immediate (physical) control. Thus, in (1114), ma- is used to describe an event of a boy picking up a dog.

8.4 Productive inflecting verbs 

 451

(1114) Murrku-ngku ma-ni yala kunyarr ngamurlu=marda-rnani. dog-erg get-pst that dog embrace=have-ipfv.pst ‘The boy picked up that dog and held him in an embrace.’ TEN12017_016-02: 95980_99445) Events of physical contact can be further specified by preverbs expressing manner of action, for example turrku- ‘grab’ in (1115). (1115) Turrku=ma-ni yala kunyarr ka-nya ngapa-kurra. grab=get-pst that dog carry-pst water-all ‘He grabbed that dog and pulled him into the water.’ (MMN: TEN1-2017_01201: 199198_202617) 8.4.1.2.2 Caused state (causative function) The verb ma- get is frequently used as a ‘causative’ grammatical verb. In this function it readily combines both with nominals, for example kirda ‘big’ (1116) and preverbs such as nyin- ‘lose’ (1117). (1116) Yawiyi, ngati, kuja=yi kirda=ma-ni. poor_thing M sub=1sg.o big=get-pst ‘Poor thing, it was my mother who brought me up.’ (DMN: LC42b: 906712_910318) (1117) Nyin=ma-ni=rna. lose=get-pst=1sg.s ‘I lost him.’ (BSE: TEN1-2018_012-02: 1247741_1249189)

Cataldi (2011, p. 246) also records an additional ‘verbaliser’ -pa, which she terms ‘inchoative’. Unlike ma-, -pa- does not also occur as an independent lexical verb, i.e. *pananta ‘he is getting it’. However, in all other respects, the use of ma- and -pa- in CVCs is identical and they can be treated as variants of the one verb.393 In fact, in neighbouring Walmajarri, -pa is simply treated as the allomorph of a do verb ma- in the environment following a consonant (Hudson, 1978, p. 14). I likewise gloss instances of -pa- as an allomorph of ma-, i.e. get. The variable appearance of -pa- forms recorded in the most recent Ngardi data may actually reflect that this is an aspect of Ngardi grammar undergoing change at the time of documentation. When checking verb forms with Marie Mudgedell and Patrick 393 Moreover, Cataldi’s use of the term ‘inchoative’ is somewhat misleading in that ‘pa-’ does not always denote an inceptive event but rather verbalises a range of nominals with different event structures in a fashion identical to ma-.

452 

 8 Complex predication

Smith in 2017–2018, forms listed by Cataldi as containing -pa- (e.g. lurrijpa-rnanta, partpa-rnanta) were often rejected in favour of forms with ma- (lurrij=ma-nanta, part=ma-nanta). However, despite this, Marie Mudgedell and Patrick Smith produced a small number of unprompted tokens involving pa-, for example nyit=pa-rnanta- ‘tie_up=get-prs’. The verb ma- can be used in a range of contexts and with a range of preverbs that are shared by the verb of change of locative relation yirra- place. Often there is no substantial change in meaning. For instance, turr=yirra- together=place and turr=matogether=get were considered to be homophonous, meaning ‘to gather together’. When encoding events of caused state, the verb ma- may combine with a number of spatial preverbs that typically combine with verbs of locomotion (§8.4.2). The effect of combining these preverbs with ma- is to generate an event of change of state involving induced motion. Thus, yirrit=ya- emerge=go, ‘come out’ can be contrasted with yirrit=ma emerge=get, ‘x pulls y out/causes y to be out’. Postural preverbs which combine with ka- carry can also combine with the verb ma- to encode an event of an ‘induced change of locative relation’. For example, ngamurlu=ka- embrace=carry means ‘carry in (one’s) arms’ but ngamurlu=ma- embrace=get means ‘cause x to be in one’s arms’. The events categorised by ka- and ma- here can be further distinguished from CVCs involving marda- have (§8.4.1.3) in that the latter depicts a state of possession or physical contact/spatial contiguity. Thus ngamurlu=marda- embrace=have describes an event ‘hold/have in one’s arms’. 8.4.1.2.3 Motion and non-physical interaction In addition to expressing direct physical affectedness and contact, ma- also categorises a range of event types that encompass non-physical, but highly affected, interactions between entities. For instance, ma- combines with a range of preverbs of pursuit as in (1118) and (1119). (1118) Ngantany-ju pila=ma-ni jarrampayi tatakurr-kurra. man-erg follow=get-pst goanna exhausted-all ‘A man chased a goanna until it was exhausted.’ (PSM: TEN1-2018_027-01: 3707454_3715070) (1119) Puliki=lu majurrum=ma-nani ngantany-ju. cattle=3pl.s muster=get-ipfv.pst man-erg ‘The men were mustering the cattle.’ (PSM: TEN1-2018_038-01: 201521_204337) Despite ma- get having a canonical transitive argument structure, there are a number of examples where ma- forms an intransitive predicate, for example in (1120) where it encodes a motion event. Such examples are unusual in that otherwise simple motion events usually involve a verb of locomotion, e.g. ya- go.

8.4 Productive inflecting verbs 

 453

(1120) Japurrk=ma-ni (puruk) walya-kurra purnu-jangka. jump=get-pst frog ground-all tree-ela ‘The frog jumped down to the ground from the tree.’ (MMN: TEN1-2019_00506: 229747_232673) 8.4.1.2.4 Speech and sound emission The verb ma- shares a property of the verb of force pu- hit and the verb wangka- speak in being able to categorise a small number of events of sound emission. These event types include both human speech (1121) as well as human (1122) and non-human (1123) vocalisations. (1121) Nyanyikarnu=lu yirdi=ma-nani-nyirra nyurnu-nyurnu-rlu janjuru. long_time=3pl.s name=get-ipfv.pst-narr dead-rdp-erg gidgee ‘A long time ago it was called janjura by our ancestors.’ (TJN: LC28) (1122) Minya ita jingkirdi=ma-nanta.394 this child laugh=get-prs ‘This child is laughing.’ (MMN: TEN1-2017_006-01: 1354135_1357450) (1123) Rung=ma-nani-nyirra. bark=get-ipfv.pst-narr ‘(That dog) had been barking.’ (PSM: TEN1-2018_035-02: 896699_898302) This function is also extended to ‘signalling with the hands’, as in (1124). (1124) Yala ma warrpaj-warrpaj=ma-nanta. that top.aux wave-rdp=get-prs ‘That one is waving/signalling with his hands.’ (PSM: TEN1-2018_035-02: 1349542_1352143) The ability of ma- to subcategorise absolutive subjects in CVCs as in (1121)–(1124) is anomalous when considering its function as a canonical transitive verb of manipulation/acquisition (x gets y) in its simple, independent form. When comparing this form to neighbouring languages, it is tempting to suggest that Ngardi has two discrete homophonous forms (i.e. ma- get, a transitive verb with an associated causative function, and masay, an intransitive verb of sound emission). This is observed in both related Ngumpin languages, (e.g. Jaru: Tsunoda, 1981b, pp. 81–82) and northern western Desert languages (e.g. Wangkajunga: Jones, 2011, p. 177). However, beyond these specifics CVCs in Ngardi, ma- is never used as a verb of speech in its simple form. Attempts to elicit it as such 394 Note that this preverb is doubly classified as a lexical verb root (Class I) (see Table 80).

454 

 8 Complex predication

were rejected by Ngardi speakers as ‘Jaru talk’ and would be corrected to the Ngardi verb wangka- speak. Furthermore, for those Ngumpin languages with two homophonous verbs of the form ma-, the verb of speech emission belongs to a separate conjugation class, characterised by an imperative inflection akin to that of conjugation IV (i.e manja). This is not the case in Ngardi; all uses of the ma- verb belong to the one conjugation class. Interestingly, the ability for a get verb ma- to classify an event of sound emission (in CVCs only) is also shared by Warlpiri (Nash, 1986, p. 246), which similarly shares wangka- as a generic verb of speech. McConvell and Laughren (2004, p. 170) hypothesise that the Yapa languages inherited a verb of speech *ma-nyja (say-imp) but it subsequently shifted to the conjugation of ma-nta (get-imp) and was lost outside of complex verbs involving preverbs of speech production. Such an account appears equally wellsuited to Ngardi. 8.4.1.2.5 ‘Verbaliser’ for loan words The get verb ma- is the most common inflecting verb used to integrate Kriol loans via a ‘light verb strategy’ (Wohlgemuth, 2009, p. 184). Numerous examples of ma- in this function were provided in §8.3.1.4, along with the discussion of Kriol verb borrowings more broadly. 8.4.1.2.6  Idiomatic usages One notable idiomatic usage of -ma is its role in the formation of the avalent clause type walmak=ma ‘it is lighting’, as in (1125). (1125) Walmak=ma-nanta. lightning=get-prs ‘It is lightning.’ (MMN: TEN1-2017_002-02: 1215983_1218404) 8.4.1.2.7 Summary of CVCs Due to the very numerous CVCs involving ma-, the survey of event types provided in the preceding subsections is preliminary and indicative at best. There may be many more event types than indicated here. In fact, the ‘event classification function’ of productive verbs is the least clear for ma- when compared to other verbs simply because ma- is so productive as a general transitive predicate. This is also reflected in the overwhelming usage of ma- when borrowing Kriol verbs into CVCs in Ngardi (§8.3.1.4). The following charts summarise some of the key event types described in the preceding subsections and provide further examples of relevant, attested CVCs. Affectedness and contact (acquisition) turrku=manguru=ma-

erg grabs abs erg steals abs

ngamurlu=mawarru=ma-

erg holds abs in arms erg collects abs from all around

8.4 Productive inflecting verbs 

 455

Affectedness and contact (physical impact) nyirtily=maturt=ma-

erg pinches abs erg breaks abs

Affectedness and contact (manipulation) turrku=mangardak=manyit=manyuri=ma-

erg grabs abs erg manufactures abs erg ties up abs erg kneads abs

palu=mawarru=mayirrit=ma-

erg rubs abs erg gathers up abs erg pulls abs out

Motion and non-physical interaction japurrk=mapila=mamajurrum=ma-

jurra=mawaya=ma-

abs jumps erg follows abs erg musters abs

erg hunts out abs erg bothers/pesters abs

Causative/caused states395 jaa=majalkun=maluu=maparnki=mapirrka=ma-

erg opens abs erg softens abs erg makes abs strong erg wakes up abs erg constructs abs

turn=maturr=mawuruju=mawuruly=ma-

erg covers abs (in ashes for cooking) erg brings abs together erg heals/mends abs erg makes abs quiet

Speech yirdi=ma-

erg insults/name calls abs

Sound emission jingkirdi=manyinypurrk=marung=ma-

abs laughs abs snorts abs barks

warrpaj=mangaparayi=ma-

abs signals with hand abs yawns/sighs

395 Note that the number of CVCs relevant to this category is far larger than indicated here. In addition to some of the preverbs presented here, ma- can more or less combine with any nominal in a causative function.

456 

 8 Complex predication

Idiomatic walmak=ma-

lightning striking (avalent clause)

8.4.1.3 marda- have The verb marda- have has two primary linked senses: possession (i.e. ‘x has y’) and physical contact/contiguity (i.e. ‘x holds y’). The possessive sense could be argued to be secondary in that the encoded possessive relationship necessarily involves immediate contact/control between the possessor and possessed. That is, mardahave is not a generic means of expressing ‘permanent possession/belonging’ so much as ‘temporary control/use’ (cf. Schultze-Berndt, 2003, p. 227). When expressing events of possession, the possessor is invariably the subject of marda- clauses and the possessed is invariably the object. The possessive sense freely encompasses types of possession that would be semantically categorised as ‘alienable’ (1126) or ‘inalienable’ (1127). (1126) Ngarrja=rna marda-rnanta. enough=1sg.s have-prs ‘I have enough (money).’ (PSM: TEN1-2018_021-01: 3769583_3773671) (1127) Ngaju-ngku=rna marda-rnanta kirda=nyayirniny milpa. 1sg-erg=1sg.s have-prs big=intens eye ‘I have really big eyes.’ (MMN: TEN1-2018_026-01: 1297356_1309570) When encoding events of physical contact/contiguity, marda- can co-occur with preverbs that further specify the locus of the ‘holding’ event, for example julij- ‘on the hip’ as in (1128). Alternatively, the locus of the ‘holding’ event can be specified by a spatial case-marked NP as in (1129). (1128) Julij=marda-rnanta luwanja, parda jingka. on_hip=have-prs coolamon dub child ‘The woman would hold a coolamon on her hip or perhaps a child.’ (YDN: LC24b) (1129) Lirra-ngka marda-rnanta. mouth-loc have-prs ‘He is holding it in his mouth.’ (PSM: TEN1-2018_001-01: 320994_324268) The event schematics of marda- also extend to metaphoric or non-literal events of holding, namely emotional states associated with caring, nurturing or support . This

8.4 Productive inflecting verbs 

 457

is observed both when marda- occurs as a simple verb as in (1130) and in certain CVCs, as for example warda=marda- ‘care for’ in (1131). (1130) Nga=pula marda-rni-nyirra ngawu. seq.aux=3du.s have-pst-narr sick ‘The two of them looked after the sick one.’ (MMN: Manungka_01-035155) (1131) Jarla=rna=ngku wangka-nyanta warda=marda-rnanta=rna=ngku. stomach=1sg.s=2sg.o speak-prs care_for=have-prs=1sg.s=2sg.o ‘I am speaking to you from my stomach (lit. ‘my stomach speaks to you’), I care deeply about you.’ (DMN: LC42b: 1322310_1325379) There are some idiomatic usages of marda- that do not neatly fall under the primary senses of possession, physical contact or emotional support. For example, jirri=marda- ‘x shows/offers y (to z)’ involves a transfer event that allows for a recipient in the dative case (1132). Note that the theme (or entity being offered) in this construction is typically under the physical possession of the subject. (1132) Jirri=marda-rnanta(=rla). show=have-prs=(3sg.obl) ‘He is showing it (to him).’ (MMN: TEN1-2017_001-02: 1332677_1338384) A second idiomatic usage of marda- is the CVC jina=marda ‘wait for’ (1133). Compositionally, it can be transliterated ‘holding one’s feet’ but it is a clear synonym of the CVC kuny=pu- ‘wait=hit’. Both of these CVCs subcategorise a dative indirect object. (1133) Jina=marda-ka=rla, ya-ni ma ruwaruwa. foot=have-imp=3sg.obl go-pst top.aux urinate ‘Wait for him, he went to the toilet.’ (PSM: TEN1-2018_027-01: 249473_251871) Lastly, marda- have has an idiomatic usage in its simple form: in combination with an ergative-marked body part instrument – langa-ngku ear-erg – marda- encodes an event ‘to listen to someone’, as in (1134). (1134) Langa-ngku=rna=ngku marda-rnanta. ear-erg=1sg.s=2sg.o have-prs ‘I am listening to you (lit. ‘I, the ears, are holding you’).’ (DMN: LC42b: 348435_350658)

458 

 8 Complex predication

8.4.1.3.1 Summary of CVCs Possession palka=mardajaku=marda-

? has in one’s possession erg steals abs into one’s possession

Locus of physical contact/contiguity kulykuly=mardajulij=marda-

erg holds abs in mouth erg holds abs on hip

Metaphoric ‘holding’ warda=mardangampurr=ma-

erg cares for abs erg looks out for abs

Manipulation of held objects jakurla=mardajirri=mardaturrku=marda-

erg takes back/holds back abs erg shows/displays abs erg holds out/proffers abs

8.4.1.4 wanti- fall When operating as a simple verb, wanti- fall has a primary sense of encoding a downwards-oriented motion event. Like the verbs of locomotion, events encoded by wanti- may lexically specify a source of motion – typically with an elative (1135) or ablative (1136) case-marked nominal. (1135) Ngapi=yi wanti-nya pamarr-ngurlu. F=1sg.o fall-pst rock-abl ‘My father fell off the rock.’ (PRN: TG1_0034-01) (1136) Jup=wanti-ya=lu purnu-jangka, kankarra. dismount=fall-imp=pl.s tree-ela up.a ‘You lot get down from that tree up there.’ (PSM: TEN1-2018_010-01: 357374_376281) However, unlike verbs of locomotion, events categorised by wanti- rarely specify a goal. This is possibly due to the pragmatic expectation that the majority of falling events share an inherent goal: the ground.

8.4 Productive inflecting verbs 

 459

Two additional comments are warranted for wanti-. First, despite the gloss fall, the motion events encoded by wanti- can be either involuntary or voluntary (e.g. wumpurr=wanti- ‘falls headfirst’, jup=wanti- ‘dismounts’). Second, despite frequently being associated with events associated with downward motion, wanti- is not strictly delimited to categorising only these events. There are some instances where a vertical component of motion is either minimal or absent entirely. For instance, jup=wanti can be used to mean ‘get out of a car’, which only involves a very minor vertical displacement (from the floor base of a vehicle to the ground) and could alternatively be conceptualised as a ‘standing up’ or a ‘stepping out’ event. Indeed, the primary sense appears to be that of ‘dismount’ or ‘depart from a mode of locomotion’. (1137) Minya nawu=rnalu jup=wantinya, this now=1pl.excl.s dismount=fall-pst nya-ngani=rnalu=rla yarla-ku. see-ipfv.pst=1pl.excl.s=3sg.obl yam-dat ‘Here (in this photo) we got out (of the car), and we were looking for bush potatoes.’ (MMN: TEN1-2017_020-02: 136810_146173) Furthermore, there are some clear instances of non-motion events in which wantiis implicated. In addition to its primary motion sense, wanti- also encodes static/ postural event types that are situated close to the ground (1138), changes of various states (1139), and, somewhat idiomatically, events of birth (1100), falling ill (1140) and passing away (1141).396 (1138) Tajput=wanti-nyanta. bog=fall-prs ‘It is stuck/bogged.’ (Cataldi, 2011, p. 292) (1139) Kily=wanti-nyanta. cold=fall-prs ‘It’s getting cold.’ (PSM: TEN1-2017_023-03: 2050802_2054635) (1140) Wakurra kala=rnalu wanti-nya-ngurra nyurnu, waku. neg asrt=1pl.excl.s fall-pst-narr sick no ‘We didn’t get sick, nothing.’ (TJN: Manungka_01-035151: 798847_802667)

396 This metaphoric usage is also encoded lexically in the preverb-verb combination yululu=wanti ‘abs passes away’.

460 

 8 Complex predication

(1141) Yuwayi, ngaju=rna wapirra nyin=ma-ni yalu-ngka=lu wanti-nya [...] yes 1sg=1sg.s F lose=get-pst that-loc=3pl.s fall-pst ‘Yeah, I lost my father, they (all) passed away in that place.’ (PSM: TEN12019_001-02: 295970_302634) The fact that downwards motion is not intrinsic to the meaning of wanti- in certain CVCs is made further apparent by its combination with the preverb jarr- (1142) meaning ‘to come to a halt’ or the Kriol ternim (1143) meaning ‘to turn around’. In both cases, the ‘motion event’ does not involve a vertical directional component but simply a change of orientation or movement. (1142) En nyanungu yaparti-nya-ngurra, jarr-jarr=wanti-nya-ngurra. and recog1 run-pst-narr stop-rdp=fall-pst-narr ‘And that one had been running, and then came to a stop.’ (TJN: Manungka_01-035154) (1143) Nyanungu-rla ternim=wanti-nya. recog1-loc turn.tr=fall-pst ‘That one turned around for him.’ (MMJ: MAATI1) Similarly, the event of riding in a saddle is encoded somewhat idiomatically by the CVC laja=wanti- ride=fall (1144). (1144) Jadel-a laja=wanti-nya-ngurra kankarni. saddle-loc ride=fall-pst-narr up.b ‘He rode along up there in the saddle.’ (TJN: LC20a) These patterns of CVC formation suggest that wanti- is not strictly specified for its downward motion (although it prototypically involves events of this type) or voluntariness but simply entails that a figure ‘comes to be in a locative relation with respect to a location’ as described by Schultze-Berndt (2000, p. 232) for a similar functioning verb in Jaminjung (-irdba- ‘fall’). Generally speaking, CVCs involving wanti- almost invariably exhibit an intransitive predicate argument structure; however, there are a small number of avalent predicate argument structures (kuru=wanti- ‘(it) becomes dark’ and kily=wanti/ mum=wanti- ‘it becomes dark’, see §9.2.3.1) and at least one extended intransitive (lap=wanti-, abs follows after loc, see §9.2.3.4).

8.4 Productive inflecting verbs 

 461

8.4.1.4.1  Summary of CVCs Downwards-oriented movement yut(-jut)=wantijirrk=wantijit/jut=wantijup=wantijuruk=wantijurr=wantikalkurr=wanti-

abs sits down abs drips down abs lowers/ducks down abs dismounts abs swims abs descends abs falls away (sand)

luk=wantinyinkarra=wantipurluk=wantitalypurt=wantiwumpurr=wantiyinjilpi=wantiyuruk=wanti-

abs lies down abs swims (& showers) abs jumps into water abs sinks down abs falls headfirst abs drips abs pours/rains down

Postural state proximal to the ground lapart=wantiparntaparnta=wantipukany=wanti-

abs lies flat abs lies down ill abs sleeps

tajput=wantitirrip=wantiyurpij=wanti-

abs is stuck abs camps abs sits back on heels

Change of state397 kily=wanti398 born=wantijarr=wantijupu=wantikapurrk=wantikarrak=wantikuru=wanti-

(it) becomes cold abs is born abs stops abs stops abs sleeps/is unconscious abs becomes grey (it) becomes overcast/dark

mum=wantiramuk=wantinguriny=wantiwak=wantikarlana=wantikarlaniyin=wantitaly=wanti-

(it) becomes dark (it) becomes dark abs turn around abs develops breasts abs turns west abs sets (sun) abs fell and broke

Pursuit lap=wanti-

? follows after loc

Idiomatic laja=wantitarrjak=wanti-

abs rides erg? secures abs

397 Cataldi (2011, p. 6) describes wanti- as having a ‘change of direction’ meaning in CVCs. 398 This is an example of an English loan being incorporated into a CVC.

462 

 8 Complex predication

8.4.1.5 yirra- place The transitive verb yirra- as a simple verb can be consistently translated as ‘x places y’ or ‘x puts y’. The core meaning of this verb can be rendered more schematically as ‘x causes y to be at z’, that is, an induced change of location (§8.4.1.5.1). The optional argument z is a location that can be specified by a locative-marked nominal, as for example yalu-rla that-loc in (1145). (1145) Yalu-rla yirra-ka ingkarna, ngu=rlipa kupa-ku. that-loc place-imp kindling seq.aux=1pl.incl.s cook-pot ‘Put that kindling down there, so that we can cook.’ (PSM: TEN1-2018_00401: 473976_479665) In its simple verb form, yirra- place is not specified for directionality (e.g. ‘put up’ or ‘put down’) nor the configuration or alignment of the object/entity being placed/ positioned. Typically, such additional specifications are either understood from context or encoded through combinations with preverbs. Furthermore, yirra- may combine with certain preverbs or nominals expressing mental states to encode transfer of information (§8.4.1.5.2). There are some additional CVCs involving yirrawhich are best categorised as idiomatic (§8.4.1.5.3). 8.4.1.5.1 Induced change of locative relation When encoding an induced change of locative relation, yirra- may enter into CVCs that involve preverbs that emphasise the directionality of the positioning event as in (1146) and (1147), or the manner of the positioning event, as in (1148) and (1149). Preverbs of the latter type may emphasise the function or resulting configuration of the figure(s), as in (1150). (1146) Wirrkirl yalu jit=yirra-ka=lu! head that lower=place-imp=pl.s ‘You lot, put (your) heads down!’ (PSM: TEN1-2018_010-01: 85426_89429) (1147) Nga=pula purnu-ngka tarn=yirra-rni. seq.aux=3du.s tree-loc hang_up=place-pst ‘Those two put him up in the tree.’ (TJN: LC49a: 257286_260374) (1148) Wajarn jut=yirra-ka=lu=wu kirnimiliny-ku! resin_wax fasten=place-imp=pl.s=voc spear-dat ‘Apply/fasten on that wax for the spear!’ (PSM: TEN1-2018_010-01: 701944_704499)

8.4 Productive inflecting verbs 

 463

(1149) Pija=yirra-ka, ngu=rlipa yut=karri-ju. flatten=place-imp seq.aux=1pl.incl.s sit=be-pot ‘Make it flat, so that we can sit down.’ (PSM: TEN12018_004-01: 536223_540559) (1150) Turr=yirra-ka=yanu puliki! together=place-imp=3pl.o cattle ‘Muster the cattle!’ (PSM: TEN12018_010-02: 1127623_1142954) 8.4.1.5.2 Transfer of cognitive state An additional, related event type to induced change of locative relation is that of transfer of cognitive state. In these event types, the figure is not a physical entity but instead a cognitive entity (e.g. information), as in (1151)–(1153). (1151) Nyampa=n marda-rnanta jirri=yirra-ka! what=2sg.s have-prs show=place-imp ‘Show me what you are holding!’ (PSM: TEN12019_007-02: 110336_116259) (1152) Karntuwangu, jangu=rnalu pina=yirra-rni kakarra tree_sp recog2=1pl.excl.s knowledge=place-pst east pantarra purnu. tall tree ‘Karntuwangu is one which we have noticed in the east, it is a tall tree.’ (LC28b) (1153) Wakurra=yi=nta yilayi=yirra-rnanku ngaju. Wuruju=ma-nta! neg=1sg.o=2pl.s slander=place-ipfv.pot 1sg good=get-imp ngaju-kariny ma=rna wuruju–– yut=karri-nyanta. 1sg-other top.aux=1sg.s good sit=be-prs ‘You lot shouldn’t be slandering me! Be nice! I am simply sitting here by myself, peacefully.’ (PSM: TEN1-2018_012-03: 2576872_2587348) 8.4.1.5.3 Idiomatic usages Finally, there are a small number of usages of yirra- which are best characterised as idiomatic. While this includes certain CVCs (see summary in §8.4.1.5.4), it also includes yirra- in its simple form in the expression ‘to make food’, as in (1154).

464 

 8 Complex predication

(1154) Minya=rna yirra-rni mangarri. this=1sg.s place-pst veg_food ‘Here I was making breakfast.’ (MMN: TEN1-2017_020-02: 31223_39327) 8.4.1.5.4 Summary of CVCs Change of locative relation iliri=yirrajarnaka=yirrajirriku=yirrajit=yirrajurr=yirrajut=yirralapart=yirramitimiti=yirramalany=yirramunturrmunturr =yirraparnta=yirraparntik=yirrapart=yirrapija=yirrapikala=yirra-

erg buries/caches abs erg places abs on one’s back erg keeps abs for later erg lowers abs erg puts abs down erg fastens abs erg lays down abs erg immobilises abs erg loans abs to dat

pirrip=yirrapuru=yirraputari=yirraramuk=yirratakurl=yirratardarr=yirratarn=yirraturn=yirrawalypi=yirra-

erg covers up abs erg places abs out of sight erg collects abs erg covers up abs erg gathers in abs erg inserts abs erg hangs abs up (high) erg buries abs erg adorns abs with bark

erg heaps up abs

waraly=yirra-

erg hangs abs (on loc)

erg places abs down erg spreads out abs erg lathers abs on loc erg flattens out abs erg places abs on one’s back/shoulders

wipiriny=yirrawirrim=yirrayimpiyimpi=yirrayuruk=yirrayut=yirra-

erg burrows ? erg wears abs erg shakes abs erg pours abs erg positions abs

Transfer of cognitive state jirri=yirramilki=yirrapina=yirrayilayi=yirrakunka=yirra-

erg shows abs (to dat) erg shows abs (to dat) erg notices abs erg slanders abs erg pays back abs

Idiomatic and other uses jarra=yirrakulpa=yirraparnti=yirrayinkil=yirrajujupa=yirratirrngij=yirramarrka=yirra-

erg warms abs by fire ? vomits erg puts smell on abs abs makes a fuss about dat erg makes love to abs erg blocks abs ? stops

8.4 Productive inflecting verbs 

 465

8.4.1.6 kiji- throw The verb kiji- throw is only marginally classified as a productive inflecting verb and has been attested with a relatively small number of preverbs in CVCs (see §8.4.1.6.1). It is grouped here with verbs of change of locative relation as it can combine with a small number of preverbs indicating the manner of an induced change of locative relation. Moreover, kiji- throw and yirra- place can form CVCs with some of the same preverbs. The primary distinction in the resulting event types appears to be that of manner (i.e. projectile motion (kiji-) vs accompanied motion (yirra-)) but there also appears to be a component of intentionality involved – namely, whether the figure comes to be located at a specific location (yirra- place) or not (kiji- throw). Thus yuruk=yirra pour=place is typically used to refer to controlled acts of pouring; the locational endpoint of the figure (the water) can often be specified as a locative (e.g. kartak-ta cup-loc). This is in contrast to yuruk=kiji pour=throw which describes an event of tipping or pouring (out) as a means of disposing a liquid. A number of induced change of location events categorised by kiji- tend to be typified by a locative goal that is undefined, as in (1155) and (1156). (1155) Yuruk=kiji-ka! pour=throw-imp ‘Pour it out!’ (Ennever, 2018, fieldnotes) (1156) Nyampa-ku=n=ku=rla minya jaku=kiji-rni? what-dat=2sg.s=ep=3sg.obl this depart=throw-pst ‘Why did you throw that out?’ (PSM: TEN1-2017_010-01: 604814_608640) In such contexts, kiji- can be schematised simply as ‘x causes y to move away via a projectile motion’. Nevertheless, kiji- still patterns like the other verbs of induced change of location in that it may optionally specify a locational goal via a spatial casemarked NP as in (1157). (1157) Kiji-rni=lu fishing line ngapa-kurra. throw-pst=3pl.s fishing_line water-all ‘They chucked the fishing line into the water.’ (MMN: TEN1-2017_012-01: 13072_20635) The verb kiji- throw also categorises at least two bodily events associated with ingestion: breathing heavily (ngayirr=kiji- breathe=throw) (1158) and quickly drinking or ‘sculling’ (nyurnun=kiji- swallow=throw) (1159). Both events appear to involve rapid ingestion and can be contrasted with verbal predicates of lesser intensity: ngayirr=ma- ‘breathe’ and nguka- swallow. These events are clearly distinct from the projectile events described above in that they do not involve an entity being projected away from an agent.

466 

 8 Complex predication

(1158) Minya ma=rna ngayirr=kiji-rnanta, wurna-jangka, yumpaly. this top.aux=1sg.s breathe=throw-prs travel-ela tired. ‘I am breathing heavily from travelling, tired.’ (PSM: TEN1-2018_021-02: 546691_555468) (1159) Nyurnun=kiji-rni, yi-nya=rna=ngkupula kuriny-ku. swallow=throw-pst give-pst=1sg.s=2du.o two-dat ‘She sculled it all; (even though) I gave it to the two of you.’ (MMN: TEN12017_002-05: 385093_392906) Interestingly, the CVC nyurnun=kiji- appears to be able to be metaphorically extended to contexts of non-literal ingestion, but rather events involving the surrounding of some medium around the affected figure (1160), not dissimilar to the English expression ‘swallowed up’ (by the ground, water, etc.). (1160) Kayirra-ngulu ya-nanta, jarrara=ma-nani-nyirra, north-abl go-prs approach=get-ipfv.pst-narr nyurnun=kiji-rni ngantany ngapa-ngku. swallow=throw-pst man water-erg ‘It (the water) is flowing from the north; it came closer and closer and engulfed the man.’ (TJN: Cataldi, 2011, p. 244) 8.4.1.6.1 Summary of CVCs Manner of motion jaku=kijijurr=kijipata=kijiyuruk=kiji-

erg throws abs out erg deposits abs down erg spills abs erg pours out abs

Bodily actions (ingestion) ngayirr=kijinyurnun=kiji-

abs breathes erg swallows abs erg engulfs abs

8.4.1.7 yu- give The verb yu- give categorises a range of events that broadly fall into one of three main event schemas: transfer of possession (§8.4.1.7.1), transfer of information (§8.4.1.7.2) (x  causes y to be at z) or induced motion (§8.4.1.7.3) (x causes y to move). A small

8.4 Productive inflecting verbs 

 467

number of constructions do not fit neatly into any of these event types and are considered idiomatic (§8.4.1.7.3). 8.4.1.7.1 Transfer of possession In its simple form, the verb yu- is a trivalent predicate that denotes the transfer of possession of an object/entity to a recipient. The various argument structures available to this predicate are discussed in more detail in §9.2.3; however, recipients can be marked either as absolutives (erg gives abs, abs) or as datives (erg gives abs to dat). Examples of yu- categorising events of ‘transfer of possession’ can be found with the verb appearing in its simple form (1161) and in combination with preverbs that specify the manner in which an entity is transferred, such as purdayan ‘reciprocal’ as in (1162). (1161) Nyuntu-ngku=n=ku=rla yi-nya kuyi yalu-ku. 2sg-erg=2sg.s=ep=3sg.obl give-pst meat that-dat ‘You gave the meat to him.’ (MMN: TEN1-2017_024-01: 3728830_3733933) (1162) Kaja=n=ku=rla yu-ngku-rni purdayan. ♀S =2sg.s=ep=3sg.obl give-pot-hith reciprocal ‘You will give it back in kind to (your) son.’ (MDN: LC22b: 462909_469716) Examples of yu- in (1161) and (1162) can be contrasted with the non-productive verb of transfer yupa- send which involves an animate, human theme and does not require a recipient – but may exhibit an allative-marked NP as a goal, as in (1163).399 (1163) Minya-rlu jingka yupa-rni (ngati-kurra warlalja-kurra). this-erg child send-pst M-all family-all ‘This one sent the child back to his own mother.’ (PSM: TEN1-2017_008-01: 1542437_1547865) 8.4.1.7.2 Transfer of information/cognitive state In addition to encoding events of transfer of possession, yu- may also encode events of transfer of information or a cognitive state. In many ways, this represents a subtype of transfer of possession that involves a non-physical entity; and one that involves a change in the knowledge state of the patient. This event classification includes complex predicates covering the notions of English ‘teach’ (pina=yu- knowledge=give) (1164) and ‘tell’ (jaru=yu- word=give, (1165)), but also ‘compel/oblige’ (jinjin=yu- oblige=give), (1166). Emotional states do not appear to be categorised

399 An exception is the metaphoric karrarda=yupa ‘x gives a fright to y’.

468 

 8 Complex predication

in this fashion (i.e. *rayin=yu- fright=give) and are instead categorised by pu- hit (rayin=pu- fright=hit, ‘give someone a fright’). (1164) Nga=yi=lu jakarr pina=yu-ngani seq.aux=1sg.o=3pl.s try knowledge=give-ipfv.pst Ngardi-ku. language_name-dat ‘They tried to teach me Ngardi.’ (THS: TEN1-2017_003-05: 116355 _133068) (1165)

Jala=rnalu ngawu=karri-nya kuja=yi jaru=yi-nya indeed=1pl.excl.s bad=be-pst sub=1sg.o word=give-pst wurna-ku. journey-dat Indeed, we were upset when he told me (we) had to go. (TJN: Cataldi, 2011, p. 68)

(1166) Nga=pulany jinjin=yi-nya-ngurra nguku-ku. seq.aux=3du.o oblige=give-pst-narr water-dat ‘He made the two of them get water.’ (Cataldi, 2011, p.75) 8.4.1.7.3 Induced motion The verb yu- also categorises events of induced motion. This has only been observed in CVCs formed with a small set of preverbs which specify the manner of the induced motion: for example ikirt- ‘sway’ (1167) and purrku- ‘push’ (1168). (1167)

Ikirt=yu-nganta purnu mayawun-tu, kirda-ngku. sway=give-prs tree wind-erg big-erg ‘The big wind is swaying the trees back and forth.’ (PSM: TEN1-2017_009-03: 44011_53005)

(1168) Wakurra=nta yala purrku=yu-nku. neg=2pl.s that push=give-pot ‘You lot don’t push him away!’ (MMN: TEN1-2019_007-01: 379801_382411) 8.4.1.7.4 Idiomatic usages Finally, there are a small number of CVCs involving yu- give which are best characterised as idiomatic. For example, in combination with the preverb yapart ‘run’, this verb describes an event ‘x races y / x gives y a race’ (1169).400 400 The same idiomatic expression is observed in Sturt Creek Jaru involving the Jaru preverb puruja ‘run’ (Patrick Smith, pers. comm., 2017).

8.4 Productive inflecting verbs 

 469

(1169) Ngu=rlijarra=nyanu yu-ngani-nyirra yapart. seq.aux=1du.excl.s=refl give-ipfv.pst-narr run ‘We raced each other./We gave each other a race.’ (MMN: TEN1-2018_029-01: 986882_1028999). Two events of changed state are also categorised by yu- which would otherwise be expected to be categorised by a transitive verb such as ma- get or even yirra- put: wumpu=yu- ‘singe’ (1170) and kanakana=yu- ‘winnow’ (1171). (1170) Nga=rnalu wumpu=yu-ngku. seq.aux=1pl.excl.s singe=give-pot ‘Then we singe it in a fire.’ (MDN: Cataldi, 2011, p. 336) (1171) Ngu=rnalu mangarri junta ngu=rnalu seq.aux=1pl.excl.s veg_food onion seq.aux=1pl.excl.s kanakana=yu-ngku, kuja, kuja-piya. winnow=give-pot thus thus-sembl ‘And we clean/winnow wild onions in this way, like this.’ (Nyurrungka: Cataldi, 2011, p. 103) 8.4.1.7.5 Summary of CVCs Transfer of an entity injanu=yukuntil=yu-

erg shares out abs (to dat?) erg distributes abs (to dat?

Induced motion ikirt=yuimpi=yu-

erg shakes abs back and forth erg shakes abs

purrku=yuwiti=yu-

erg pushes abs erg rotates abs (of a skewer)

Transfer of cognitive iInformation jaru=yujinjin=yu-

erg instructs abs erg obliges abs (to dat)

Idiomatic kanakana=yuwumpu=yu-

erg winnows abs erg singes abs

pina=yu-

erg teaches dat

470 

 8 Complex predication

8.4.2 Verbs of locomotion There are three productive verbs of locomotion: ya- go (§8.4.2.1), yapartu- run (§8.4.2.2) and ka- carry (§8.4.2.3). I use the term ‘locomotion’ following Schultze-Berndt (2003, p. 252) which in turn corresponds closely to Talmy’s (1985) ‘translational motion’. All three verbs share the feature that they involve self-propelled motion events which proceed along a path of motion (Schultze-Berndt, 2003, p. 253). A path of motion event is one where a figure occupies more than two positions across more than two points in time. This definition is required to distinguish a path of motion from merely a change of locative relation (see Langacker, 1990, p. 155). All verbs of locomotion are unspecified for directionality and are further characterised by their ability to combine productively with the directional inflections (§7.3.3) which indicate the direction of movement with respect to the speaker. All verbs of locomotion can readily be supplied with a source and a goal, the former marked ablative or elative and the latter marked allative or dative (in the case of animates). The events denoted by ya- go and yapartu- run are essentially only distinguished by velocity of motion, although the former appears to enter more productively into combinations not always observed for the latter. Finally, ka- carry is distinguished more clearly from the two other productive locomotion verbs in its expression of ‘accompanied locomotion’. 8.4.2.1 ya- go As a verb of locomotion, ya- is unspecified for directionality but may combine productively with the directional suffixes to generate the deictic motion events ‘come’ and ‘go (away)’. The verb ya- is also unspecified for manner of motion. This can be contrasted with the productive locomotion verb yapartu- run and the non-productive verb of locomotion warrka- crawl, both of which are specified for manner. Nevertheless, a motion event encoded by ya- can be further specified for manner by direct integration of a nominal, as demonstrated with jina ‘foot’ in (1172), or by a preverb, as demonstrated with ikirtikirt ‘rattle’ in (1173). (1172) Jina=rnalu ya-nani. foot=1pl.excl.s go-ipfv.pst ‘We went by foot.’ (YDN: LC24a: 849040_857290) (1173) Ikirt-ikirt=ya-nanta. rattle-rdp=go-prs ‘(The tree) is swaying.’ (DMN: LC48b: 775224_778360)401

401 Here ‘tree’ is metonymic for ‘branches/leaves’.

8.4 Productive inflecting verbs 

 471

Additionally, as a verb of locomotion, ya- combines with preverbs expressing path of motion as in (1174)–(1176). This feature of verbs of locomotion is shared with a number of verbs of change of locative relation. (1174) Turdurr=ya-ni=lu pamarr-jangka. descend=go-pst=3pl.s hill-ela ‘They came down from the hill.’ (BSE: TEN1-2017_015-02: 774152_775715) (1175) Murrku waru=ya-ni purnu-ngka kankarra. boy climb=go-pst tree-loc up.a ‘The boy climbed up the tree.’ (MMN: TEN1-2017_016-02: 180117_181766) (1176) Ngaju-ku=yi ngumparna pirriri=ya-ni kayirra-ngulu. 1sg-dat=1sg.o husband arrive=go-pst north-abl ‘My husband arrived from the north.’ (MNN: LC24b: 24675_28033) Verbs of locomotion can also be used to categorise the spatial extension of linear objects (cf. Talmy’s ‘Coverage Path’ (Talmy, 1996)) in a manner similar to such English expressions as ‘the road goes around to the left’. This metaphoric extensional use  of a locomotion verb to a non-motion event can be seen in the event encoded by the complex verb lankarr=ya-, which describes an event of ‘splitting’ (1177). The linear extent of a ‘split’ or a ‘crack’ is conceptualised as proceeding over a path of motion. (1177) Ngu=rna yangi kala ma=yi lankarr=ya-ni yangi. seq.aux=1sg.s one asrt top.aux=1sg.o split=go-pst one ‘I do actually have one (boomerang), but the other one split/broke.’ (TJL: TT76_2201: 1940014_1943160) A further, somewhat idiomatic, usage of ya- is found in cases where ‘spatial extension’ is metaphorically applied to time. Thus, consider the Kriol-Ngardi CVC wayit=yawait=go in (1178) (and see also a similar construction involving the Kriol stative noun ‘worried’ in (1023)): (1178) Marlurlu-kurlu-rla wayit=ya-nani-nyirra kulirra-jangu-ku, jingka-ku. initiate-prop-loc wait=go-ipfv.pst-narr south-ela-dat child-dat ‘In the company of the initiates, (they) were waiting for those children from the south.’ (TJN: LC21a: 1158310_1163101) This same temporal extension is perhaps also relevant in a small number of remaining idiosyncratic usages, for example linti=ya- begging=go ‘x begs y’ in (1179) or

472 

 8 Complex predication

ngunykarra=ya- ‘x sulks’ in (1180). These events did not clearly involve a motion event but were translated with progressive aspectual semantics. (1179) Ngu=rnalu=yanu linti=ya-nani-nyirra puluku-kurra. seq.aux=1pl.excl.s=3pl.o beg=go-ipfv.pst-narr cattle-all ‘We were begging them for(?) cattle.’ (PPN: Manungka_01-035167) (1180) Ngunykarra=ya-nanta, nyampa-jangka=parda! sulk=go-prs what-ela=dub wakurra=yirla wankga-nyanta. neg=1sg.lct speak-prs ‘She is being all grumpy over something or other! She won’t speak to me.’ (PSM: TEN1-2018_012-02: 13883_19759) 8.4.2.1.1 Summary of CVCs Manner of motion ikirt=yajina=yapanpa=ya-

abs sways abs goes by foot abs goes softly

Path of motion or change of relative location jaku=yajarrk=yajirrmin=yajurr=ya-402 jurru=yamurru=yapirriri=yapurrpkarra=ya-

abs goes away/escapes abs crosses/traverses loc abs goes straight abs descends abs goes through abs goes backward abs arrives abs splits ways/scatters

riri=ya-

abs goes around

takurr=yatardarr=yaturdurr=yaturr=yaturu=yawaru=yawilayi=yawirriri=ya-/ wirrirl=yayirrit=ya-

abs goes inside abs enters abs descends abs congregates abs goes through abs ascends abs goes around abs returns/goes back abs emerges

402 Some speakers exhibit variation between [cur] and [tur] for jurr-. Note that there is also a distinct preverb homophonous with the latter pronunciation, i.e. /tur-/ as [tur] in turr=ya- ‘(come) together’.

8.4 Productive inflecting verbs 

 473

Extension of linear object/time lankarr=yapinpin=yawayit=ya-

abs splits abs swells/bruises abs waits

Other linti=yangunykarra=yawayit=ya-

erg begs abs abs sulks abs waits (for dat)

werak=yawurna=ya-

abs works abs journeys

8.4.2.2 yapartu- run The verb yapartu- run categorises a set of event types not dissimilar from ya- go but appears to be somewhat less productive in its event categorising function. The major distinction is that yapartu- in its simple form is specified for the manner of its locomotion – namely it describes motion events of greater velocity. For example, jaku=ya- ‘x sets off’ is distinguished from jaku=yapartu- ‘x runs off’. The verb yapartu- typically combines with preverbs that further specify the manner of locomotion, for example in wuruly=yapartu- ‘run quietly’ (1181), or else a path of motion, for example jarrk=yapartu- ‘rush/run across’, as in (1182). (1181) Domitri-jangka=rna wuruly=yaparti-nya. dormitory-ela=1sg.s quiet=run-pst ‘I stealthily ran away from the dormitory.’ (BSE: TEN1-2017_014-01: 410623_413055) (1182) Jarrk=yaparti-nya=rna (ngapa-ngka). cross=run-pst=1sg.s water-loc ‘He rushed across the water.’ (PSM: TEN1-2017_009-02: 397191_404413) While yapartu- generally encodes locomotion events that are defined as self-propelled motion events, there are nevertheless a range of CVCs in which yapartu- is used where the subject is not truly self-propelled but are in some sense autonomous, as is the case with certain man-made objects: (1183) Windmil kurruri=yapartu-nganta. windmill round=run-prs ‘The windmill goes around.’ (PSM: TEN1-2018_020-01: 563570_568255) In addition to literal locomotion, yapartu-, can also be used to describe the static extension of linear objects (1184), as described for ya- go (§8.4.2.1).

474 

 8 Complex predication

(1184) Boa-ja ngayi wangka-nyani ka rod, yapartu-nganta bore-loc indeed speak-ipfv.pst emph.aux road run-prs rod kakani-mpala. karlani-mpala. road east-along west-along ‘At the bore, she spoke of a road that runs east to west.’ (LC22a: 1732738_1740555) 8.4.2.2.1 Summary of CVCs Manner of motion wuruly=yapartujurrpa-jurrpa=yapartu-

abs runs away quietly abs runs with little strides

Path of motion jaku=yapartujarrk=yapartujirrmin=yapartujurr=yapartujurru=yapartu-

abs runs off/flees abs rushes across abs runs straight by abs runs away/down abs runs directly

parda=yapartupirriri=yapartutardarr=yapartuturr=yapartuyirrit=yapartu-

abs runs up (a gradient) abs rapidly arrives abs runs inside abs rushes together abs rushes out

8.4.2.3 ka- carry The verb ka- carry exhibits a number of features that pattern with the other verbs of locomotion. First, ka- carry is unspecified for directionality. It freely combines with both the hither (-rni) and hence (-rra) directional suffixes to generate the deictically specified meanings equivalent to English ‘bring’ and ‘take’ respectively. The basic semantic event type encoded by ka- is an ‘induced change of location’ (cf. SchultzeBerndt’s (2003, 270ff.) ‘accompanied locomotion’). The non-productive verb jarri‘take away/steal’ is also an induced change of location verb and is further specified for directionality but is not productive in the formation of CVCs.403 The ‘accompaniment’ component of the locomotion event can be variously categorised as one of physical contact (e.g. ‘carrying’ or ‘holding’) or it may simply be one of ‘continued co-presence’ (cf. Schultze-Berndt, 2003, p. 270). Like the other verbs of locomotion, clauses involving the ka- carry verb may lexically encode a source (1185) and/or a goal (e.g. (1115)).

403 There are (near-)synonyms for jarri- in the form of the CVCs: jarrily=ma- and nguru=ma-.

8.4 Productive inflecting verbs 

 475

(1185) Timana=li=yanu ka-nya mula-jangka ngurra-jangka. horse=3pl.s=3pl.o carry-pst this-ela country-ela ‘They took the horses from this country.’ (PSM: TEN1-2017_00801: 515436_520163) In addition to being able to specify the locus of the carrying event via a case-marked NP, clauses containing the ka- carry verb may also involve one of a range of different preverbs that specify the locus of the carrying, typically a body part, e.g. julij ‘carry on hip’ (1186). (1186) Ngaringka-rlu julij=ka-nganta ita. woman-erg carry_on_hip=carry-prs small ‘This woman is carrying the child on her hip.’ (MMN: TEN1-2017_016-03: 176360_185891) In the case of the CVC pilaka=ka- ‘carry on back/shoulders’, it can be observed that it has undergone semantic extension to the novel event types: ‘transport by car’ or ‘ride a horse’ as shown by (1187). (1187) Nyanyi-ku=rnalu pilaka=ka-nganta. before-dat=1pl.excl.s shoulder=carry-prs ‘We used to ride (horses).’ (PSM: TEN1-2018_017-03: 382654_384244) There are some distinct (and possibly idiomatic) usage of ka- in the categorisation of some events that relate to cognitive states. For example, pina=ka- ‘x shows y’ denotes a transfer of information event (1188) and is an apparent synonym for pina=yuknowledge=give (as shown previously in (1164)). Other examples of idiomatic CVCs involving ka- are provided in §8.4.2.3.1. (1188) Nya-ngani-nyirra=rnalu nganimpa-rlu, see-ipfv.pst-narr=1pl.excl.s 1pl.excl-erg pina=ka-nya=lu=nganpa. knowledge=carry-pst=3pl.s=1pl.excl.o ‘We looked at it, they showed it to us.’ (TJN: LC49a: 411122_415421)

476 

 8 Complex predication

8.4.2.3.1 Summary of CVCs Manner of induced change of location jarnaka=kajitijitikarra=kapilaka=kapikala=ka-

erg piggybacks abs erg drags abs erg carries abs on back erg carries abs on back/head

ngamurlu=ka- erg carries abs in arms warnkurr=ka- erg carries abs over the shoulders julij=kaerg carries abs on the hips

Idiomatic nguny=kangampurr=ka-404 pina=kayarrk=ka-

erg is sulky with abs erg looks out for abs erg teaches abs (about dat), erg shows abs, abs erg aims at dat

8.4.3 Verbs of contact/force There are four productive ‘verbs of contact/force’: pu- hit (8.4.3.1), la- pierce (§8.4.3.2), luwa- shoot §8.4.3.3), and paja- bite (§6.9.3.4). Verbs of force are all transitive in their simple form and all share a common semantic component in that they categorise events involving an event of impact and a highly affected ‘impacted’ participant. The verbs of force are primarily distinguished from one another by their specification of the shape of the instrument used in the impact event: la- pierce, luwa-shoot and paja- bite are differentiated by instruments involving a point, a blunt object, and an edge, respectively. The verb pu- is semantically the most polyfunctional and not only categorises the widest array of impact events but also extends to event types lacking a literal impact event. Similarly, while luwa- shoot can categorise events involving impacts with non-pointed objects (e.g. ‘grind’) it may also describe projectile events independent of their impact (cf. English ‘throw’). Finally, many CVCs formed by maget (classified according to its primary function as an ‘induced change of locative relation verb’) share many event and semantic characteristics with this class and indeed some CVCs involving ma- are entirely interchangeable with pu- with no change in meaning. This is discussed in §8.4.3.1.

404 This CVC is synonymous with ngampurr=marda- watch_out=have and ngampurr=ma- watch_ out=get.

8.4 Productive inflecting verbs 

 477

8.4.3.1 pu- hit The inflecting verb pu- combines with a large number of preverbs with widely varying semantics. Generally speaking, many of the event types categorised by pu- in CVCs reflect the same transitive argument structure (erg-abs) exhibited by pu- when functioning as a simple verb. However, there are a number of complex predicates involving pu- hit – particular those expressing movement, performance or bodily actions – that are monovalent and select a single intransitive subject (abs). A similar division between transitive and intransitive complex predicates with a single verb is observed in Walmajarri (Ngumpin) with the verb pu- hit (Hudson, 1978, p. 54) and in Wangkajunga (Western Desert) with the verb puwa- hit (Jones, 2011, p. 160). Some isolated examples of complexes involving pu- with extended intransitive case frames (erg-dat) and ditransitives (erg-abs-dat) are also provided below. When occurring as a simple verb, pu- can have a range of meanings, the majority of which involve a subject impinging upon another, as in (1189). (1189) Purangu-rlu ngu=yi pu-nganta murna. sun-erg seq.aux=1sg.s hit-prs arm The sun is burning my arm. (PSM: Ennever, fieldnotes, 2018) In addition to its core meaning as a verb of physical impact (1190), pu- can categorise a wide array of event types including manipulation (1191); performance (1192); movement (1193); bodily actions (1194); cognitive states (1195); seeking and pursuit (1196); and speech and sound emission (1197). It is also used as a means for integrating Kriol loan verbs, for example, (1092), but is somewhat less frequent than ma- get in this function (see §8.4.1.2.5). [i]      Physical impact (1190) Kak=pi-nya. knock_out=hit-pst ‘He knocked him out.’ (PSM: TEN1-2019_006-02: 1692837_1712434) [ii]   Manipulation (1191) Purlpurl=pu-ngka=yi, marlirri-rlamarra. cover=hit-imp=1sg.o cold-evit ‘Cover me up (with a blanket) from the cold.’ (PRN: TG1_0034-01) [iii]    Performance (1192) Nga=lu pitpirriya=pi-nya-ngurra. seq.aux=3pl.s dance=hit-pst-narr ‘They were dancing.’ (MMJ: LC31)

478 

 8 Complex predication

[iv]   Departure (1193) Mayarra=lku=rnalu jungkut=pi-nya, kakarra Tartpu-kurra. again=then=1pl.excl.s set_off=hit-pst east place_name-all ‘Again we set off east to Tartpu.’ (YMN: LC23a: 100821_107647) [v]   Bodily action (1194) Nyampa-jangka=rna ngajirrk=pu-nganta? what-ela=1sg.s sneeze=hit-prs ‘Why am I sneezing?’ (PSM: TEN1-2018_021-01: 2870272_2876063) [vi]    Cognitive state (1195) Nyampa-ku=n kanginy=pu-nganta? what-dat=2sg.s not_recognise=hit-prs ‘Why don’t you recognise him?’ (PSM: TEN1-2017_008-01: 3690719_3692710) [vii]    Seeking and pursuit (1196) Parli=pi-nya=rna kuriny jarany kuja=rna wurna ya-nani. find=hit-pst=1sg.s two roughtail sub=1sg.s journey go-ipfv.pst ‘I found two roughtail lizards while I was out hunting.’ (MMN: TEN12018_025-01: 3144281_3150597) [viii] Speech and sound emission (1197) Kart=pu-ngka=rlanyanta. narrate=hit-imp=3sg.lct ‘Tell that story to him.’ (MMN: TEN1-2017_023-03: 1356716_1360719) With respect to many of the above event types, pu- appears to be interchangeable with ma- with no obvious change in meaning, as in (1198).405 (1198) a. Jangku=pi-nya. promise=hit-pst b. Jangku=ma-ni. promise=get-pst ‘He promised him.’ (PSM: TEN1-2017_010-01: 379828_402532)

405 Interestingly, there are quite a number of preverbs that combine with ma-do/get in Jaru but whose cognate preverbs in Ngardi seemingly only combine with pu- hit. A single example is lindijmanan ‘flake-do/get’ (Jaru) but lintij-punganta flake-hit (Ngardi) for ‘flake (a rock)’.

8.4 Productive inflecting verbs 

 479

This apparent interchangeability of ma- and pu- is observed across nearly all event types: physical impact (turt=ma-, turt=pu- ‘break’), bodily actions (e.g. nyinpurr=ma-, nyinpurr=pu- ‘blow one’s nose’), speech and interaction (e.g. warrpa=ma- and warrpaj=pu- ‘to beckon/wave’), among many others. However, on occasion, there are clear cases where there is a contrast in meaning between events categorised by maand pu- involving the same preverb. Consider (1199) and (1200). (1199) Ngaju ma=rna nyin=ma-ni. 1sg top.aux=1sg.s lose=get-pst ‘I lost them (my sandals).’ (PSM: TEN1-2018_001-01: 1478582_1480573) (1200) Langa=lu=nganpa nyin=pu-nganta. ear=3pl.s=1pl.excl.o lose=hit-prs ‘They are deafening/hurting our ears.’ (PSM: TEN1-2017_009-02: 107727_114426) An interesting metaphoric extension of the pu- verb is found in the complex predicate juwaj=pu- (erg rejects abs) as in (1201). (1201) Yala ma juwaj=pi-nya, nyampa-jangka-rlu? that top.aux reject=hit-pst what-ela ‘That one rejected him, what for?’ (PSM: TEN1-2019_006-02: 1834850_1837265) The usage in (1201) does not neatly fall into the event types described above, but a cognate form in Walmajarri (juwaj=pu-) has a literal meaning ‘spit’. In Eastern Walmajarri the ‘reject’ meaning is also observed but with a distinct inflecting verb (juwaj=pa- or juwaka-) (Richards & Hudson, 2012). Potentially, this meaning is historically and metaphorically linked to an expectorating event which would fall under the bodily action event types commonly categorised by pu- hit. The verb pu- hit also provides multiple examples of how the argument structure of a CVC is not always solely derived from the inflecting verb but may be modified or coconstructed in conjunction with a particular preverb. The verb pu- is attested with monovalent, bivalent and trivalent predicate argument structures depending on the preverb types with which it occurs. Intransitive argument structures are found across the event types of ‘bodily actions’, ‘speech/sound emission’, as well as ‘movement’ and ‘performance’. Transitive argument structures are found in event types involving ‘physical impact’ or ‘manipulation’ as well as some ‘perception, seeking and pursuit’ event types. Finally, a trivalent argument structure is found for the specific combination kart=pu- in which the event of narration has a trivalent argument structure mapped on to the semantic roles of: narrator, theme and recipient. These participants may be encoded with either an erg-abs-dat case frame or an erg-abs-all case frame (with the former involving oblique cross-referencing of the recipient and the latter involving locational cross-referencing).

480 

 8 Complex predication

8.4.3.1.1 Summary of CVCs Physical impact (force/violence) jirlmung=punak=pujura=pukalpurr=pulankarr=pulurrij=pumunturrmunturr=punyin=puparntu=pu-

erg breaks abs erg causes abs to convulse erg strikes at abs erg pushes abs down erg splits/cuts up abs erg kills abs erg massacres abs erg deafens abs erg brushes off abs

pilkurr=pupurdayan=pu-

erg strikes abs on head erg retaliates against abs

tilykurr=putipiny=puturt=puwaly=puwawu=puwilyk=pu-

erg shoots abs erg skewers abs erg crushes/breaks abs erg thrashes abs erg singes abs erg kills abs

Manipulation iliri=pujarnan=punguru=pukalya=punum=pukarrkurr=pukawurr=pu kuraj=pulintij=pulukurr=puparntik=pu-

erg hides abs erg cleans abs erg steals abs erg paralyses abs erg shuts up abs ? crushes ? warms ? counts erg sharpens abs erg washes abs erg spreads out/flattens abs

palu=puparr=pupurlpurl=puraa=puraly=pururaj=puwaparl=puwarda=puwarrit=puwarrparl=puyarrara=puyuruly=pu-

erg extinguishes abs erg applies abs erg covers up abs erg cleans/opens abs erg plucks abs ? sprays erg spreads out abs erg blocks abs erg scatters abs erg scatters abs erg pours out abs erg removes abs

Performance jurrka=puyawluyu=pu-

erg performs abs erg performs ceremony

pitpirriya=pu-

abs dances

Movement jaku=pujilky=pujirrk=pujungkuj=pujut=pujuwantarra=pu-

abs departs abs drips out abs drips out abs sets off abs jump abs stretches out (legs)

tititi=puyapart=pu-406 turriny=pujirriny=pujurrka=puyuruk=pu-

abs shivers/shakes abs runs abs boils abs move together abs dances abs falls over

406 This CVC is rarely used. Instead, the simple verb yapartu- run is more common.

8.4 Productive inflecting verbs 

 481

Bodily action ngajirrk=pupuny=pu-

abs sneezes abs farts

mampat=puwarrpaj=pu-

abs waves goodbye ? beckons/waves

Cognitive state rayin=pu-

erg frightens abs

waja=pu-

kanginy=pu-

erg doesn’t recognise abs

pinarrikarra=pu-407

erg conceals/ disguises abs erg teaches abs

Perception, seeking and pursuit parli=pumardu=pumilya=pu-

erg finds abs erg pursues abs erg looks at abs (?)

kuny=pungunyi=pu-

erg waits for dat erg chases abs away

Speech and sound emission jangku=pujarak=pukart=pu-

erg promises abs (abs) erg berates abs erg narrates abs to dat erg narrates abs to all

liti=pumurr=purung=pu-

erg lies to abs abs rumbles ? barks

8.4.3.2 la- pierce The verb la- pierce is a verb of contact/force that is specified for the shape of instrument used in the impact event. Specifically, la- denotes an event of contact that involves a relatively pointed object in event types of spearing, piercing/poking (1202), digging (1203), stabbing (1204), or skewering (1205).408 (1202) Paka-ngku=yi la-ni. thorn-erg=1sg.o pierce-pst ‘The prickle pierced me.’ (MMN: TEN1-2017_025-02: 915488_917532)

407 A more common expression for ‘teach’ in Ngardi is pina=yu-; however, when eliciting from Jaru with binarri-g bungan ‘he is teaching him’, this was provided as the Ngardi equivalent. 408 I am unaware of any semantic differences in digging events encoded by la- spear vs pangi- dig.

482 

 8 Complex predication

(1203) Kakarra-mirnikarra la-nani-nyirra, kayirra-mirnikarra. east-side pierce-ipfv.pst-narr north-side ‘They used to dig there along the east side and along the north side.’ (TJN: Manungka_01-35166: 494508_498618) (1204) Pila=ma-nungka-nya la-ni. follow=get-x&move-pst pierce-pst ‘He chased and got him, he stabbed him.’ (TJN: LC05) (1205) Tirriny=la-nani-nyirra=rna. hole=pierce-ipfv.pst-narr=1sg.s ‘I was piercing/skewering (the seeds).’ (MMJ: LC31: 595418_600327) There is one clear case where the impact event categorised by la- does not involve a clear point but rather a blunt object: jiilykarra=la- brand=pierce that is used to describe the ‘branding’ (of cattle), as in (1206). Potentially, it is the near-resemblance of a branding stick with a spear that may have led to this event categorisation of an event type introduced via contact with Europeans.409 Note that this example also illustrates that la- pierce does not always require the instrument to categorically penetrate the impacted participant. (1206) Yard-ta rangarni-rangarni=lu jiilykarra=la-ni. yard-loc morning-rdp=3pl.s brand=pierce-pst ‘Early in the morning, they branded (them) in the yard.’ (PSM: TEN12018_038-01: 412715_417011) In some isolated cases, the impact event is of secondary importance to the associated projectile event. An emphasis on a projectile event can be coerced by inflectional directional suffixes which modify the path of motion rather than modifying the impact event itself (1207). The following example also illustrates that, as a verb of force, lamay take instruments specifying the implement used in the projectile/piercing event. (1207) La-nku-rni=ngala yalu ngantany-ju kularda-kurlu-rlu. pierce-pot-hith=1pl.incl.o that man-erg spear-prop-erg ‘That man might shoot towards us with a spear.’ (PSM: TEN1-2018_02302: 661015_678128)

409 In Jaru a cognate preverb jilygarra ‘brand’ is able to form CVCs with both a ‘pierce’ verb la- and a ‘cook’ verb gamba-.

8.4 Productive inflecting verbs 

 483

8.4.3.2.1 Idiomatic usages There are some additional, idiomatic usages of la- pierce. The first is its categorisation of a ‘spitting’ or ‘squirting’ event, as in (1208). (1208) Nya-ngankura jukany yala, kuyurr=la-nku kaniny-jarra=lku. see-ipfv.hort blood that squirt=pierce-pot inside-assoc=then ‘He will see that blood, it will squirt out from inside.’ (TJN: Cataldi, 2011, p. 150) Conceivably the categorisation in (1208) results from the physical specifications of ‘squirting’ or ‘spitting’ events generally involving a narrowed constriction of the mouth. The resulting liquid projection is also typically narrow and focussed (i.e. similar to the event schema of a ‘pointed’ object or projectile). The flow of liquid in this event can be contrasted with a low velocity flowing event which would be encoded by junpi- pour or a dispersed projected liquid: yarli- wash. A second, somewhat unusual, usage of la- pierce is its categorisation of an event ‘x tastes/tests y’ in the complex verb kut=la- ‘x tastes y’, as in (1209). It is possible that this usage is conceived from an event type of poking or contact (poke uncooked food, make a small bite, etc.) but, synchronically, it appears to be able to be used as a generic means of expressing the concept of ‘taste’ or ‘try’. This usage is most suitably described as idiomatic. (1209) Ka=rna kut=la-ni wuruju. emph.aux=1sg.s taste=pierce-pst good ‘I tasted it, and it was good.’ (TJN: Manungka_01-035159: 836601_839480) Finally, la- is also implicated in an idiomatic expression for the rising of the sun in combination with parda- ‘rise’, as in (1210). Patrick Smith considered parda=la‘rise=pierce’ and pirdiny=karri- arise=be ‘gets up’ to be synonyms in response to a prompt of ‘sunrise’.410 (1210) (Purangu) parda=la-nanta. sun rise=pierce-prs ‘The sun is getting up.’ (PSM: TEN1-2017_009-02: 434575_436990) Some limited additional idiomatic usages are summarised in §8.4.3.2.2.

410 Patrick Smith further equated these events with the Walmajarri CVC ngirra=manu ‘to add colour’ which is likely to refer to the ‘first light’ brightening the sky.

484 

 8 Complex predication

8.4.3.2.2 Summary of CVCs Impact (pointed instrument) jiilykarra=lajurru=lapart(part)=lapurluk=la-

erg brands abs erg spears abs right through ? dots/paints erg pierces/stabs abs

tajtaj=latirriny=lawiti=lawunpurr=la-

erg bangs abs erg threads abs erg skewers abs erg stabs abs

Idiomatic kuyurrk=laparda=lakut=laparda=la-

? squirts appears erg tastes abs abs appears

mingki=larangkayin=lawaparl=la-

? points erg examines abs erg spreads out/combs abs

8.4.3.3 luwa- shoot The verb luwa- shoot, despite its gloss, has at least two primary senses. One sense categorises a projectile event (e.g. ‘throw’, ‘shoot’) and a distinct second sense categorises certain impact events (e.g. ‘grind’, ‘rub’). Both event types are characterised (and distinguished from la- pierce) by the fact that they involve non-pointed instruments, including, for example boomerangs (1211) and rocks (1217). The verb luwa- shoot is also notable for a significant number of idiomatic usages described in §8.4.3.3.1.411 (1211) Luwa-rni karli-kurlu-rlu. shoot-pst boomerang-prop-erg ‘He struck (that bird) with a boomerang.’ (MMN: TEN1-2017_017-01: 311556_318139) In (1211), the projectile is registered as a nominal case marked with the proprietive. This adjunct-encoding of the projectile distinguishes events encoded by luwa- shoot from events encoded by kiji- throw since, in the latter case, the projectile is encoded as the direct object when it appears as overt NP. The ability of luwa- to categorise a projectile event (without explicitly requiring an impact event) is confirmed by examples such as (1212). In most cases the event describes a situation in which an actor initated a projectile event with the purpose of ‘killing’ or ‘maiming’ the impacted participant (which also distinguishes it from the change of locative relation verb kiji- throw).

411 Note that, in Jaru, a cognate verb luwa- spear is used as the sole lexical verb in ‘Avoidance Language’ (replacing all other lexical verbs) when a male speaks with his mother-in-law (see Tsunoda, 1981b, 216ff.). Unfortunately, I have no data on avoidance language usage or characteristics in Ngardi.

8.4 Productive inflecting verbs 

 485

(1212) Ngantany-ju ngu=yi luwa-rni wurna-kurra. man-erg seq.aux=1sg.o shoot-pst journey-all ‘The man shot at me while I was going along.’ (IPN: TEN1-2018_025-01: 747292_756487) In addition to the projectile events described above, luwa- also categorises impact events that are characterised by the application of friction and/or blunt force, for example: ‘rub’ in (1213), ‘grind’ in (1214), ‘rasp/file’ in (1215) and ‘knead/mash’ in (1216). In its ‘grind’ and ‘file’ senses, luwa- displays partial synonymy with the non-productive verbs jama- grind and lurrpa- rub – both of which can be used to describe similar events. (1213) Yuwa, luwa-rni=pula kungkala. yes shoot-pst=3du.s firestick ‘Yes, the two rubbed the firesticks.’ (NNM: Cataldi, 2011, p. 228) (1214) Pamarr yala=lku puuj-ja nganayi, mangarri jangu=lu stone that=still bush-loc whatsit veg_food recog2=3pl.s luwa-rnani. shoot-ipfv.pst ‘Those stones are there in the bush, they would use those ones for grinding food (seeds).’ (YDN: Cataldi, 2011, p. 164) (1215) Yala ma luwa-rnanta karla. that top.aux shoot-prs boomerang ‘He is sharpening that boomerang (with a rasp).’ (PSM: TEN1-2018_02302: 2100933_2103626) (1216) Ngarlany jangu nga-rnanta=rnalu ngarlany, kuyi-kurlu nyanyi rice recog2 eat-prs=1pl.excls rice meat-prop before luwa-rnani ngapurlu-kurlu. shoot-ipfv.pst milk-prop ‘That ngarlany (rice), we would eat it, with meat, and in the old days (we) would mix/mash it with milk.’ (MMN: TEN1-2018_004-02: 5982308_5998849) The primary unifying aspect of the different events exemplified above appears to be the application of friction between an instrument and an object. Interestingly, a

486 

 8 Complex predication

related verb form luwa- appears in Warlpiri, Jaru and Walmajarri but only in Walmajarri can it be used in at least one of its senses for events of ‘grinding’.412 In addition to the events described above which involve discrete impact or projectile events, there are some limited instances where both are necessarily invoked within the one CVC. In (1217), luwa- arguably categorises a projectile event while the preverb jalykuly-‘smash’ encodes the impact event. (1217) Pamarr-kurlu-rlu jalykuly=luwa-rnanta. rock-prop-erg smash=shoot-prs ‘He smashes (the car’s windscreen) with a rock.’ (PSM: TEN12017_010-01: 1621358_1624493) 8.4.3.3.1 Idiomatic uses The verb luwa-has a number of more idiomatic usages. First, the verb luwa- can be used metaphorically to describe the effects of wind on certain entities, including fire, as in (1218). In this case the affected object warlu ‘fire’ is not literally ‘speared’ but merely forced in a new direction. (1218) Mayuwun-tu luwa-rni karlarra warlu. wind-erg shoot-pst west fire ‘The wind pushed the fire west.’ (MMN: TEN1-2017_025-02: 81249_893173) This metaphoric usage is also observed in the CVC pirnpirn=luwa- spread=spear, as in (1219) and ikirt=luwa- sway=spear as in (1220). This latter CVC appears to be equivalent to ikirt=yu- sway=give, as exemplified in (1167). (1219) Mayawun-tu luwa-rnanta ngurlu, pirnpirn=luwa-rnanta. wind-erg shoot-prs seed spread=shoot-prs ‘The wind is blowing the seeds, it is spreading them.’ (Cataldi, 2011, p. 267)

(1220) Ngapa-ngku mayawun-tu ikirt=luwa-rnanta purnu. water-erg wind-erg sway=shoot-prs tree ‘The rain and wind are making the tree sway backwards and forwards.’ (TJN: Cataldi, 2011, p. 42).

412 In Warlpiri luwa- can be used (amongst other senses) to refer to ‘igniting by friction’ and ‘winnowing’ seeds but seemingly not to the act of ‘grinding’ seeds with a rock.

8.4 Productive inflecting verbs 

 487

An idiomatic usage of luwa- in its simple form includes the encoding of an event of ‘calling’, specifically the sound emissions of various animals, as in (1221). (1221) Murungkurr “wirtirtirt” luwa-rnanta. bird_sp wirtirtirt shoot-prs ‘The murungkurr bird makes a “tirt-tirt-tirt” noise.’ (TJN: Cataldi, 2011, p. 165) 8.4.3.3.2 Summary of CVCs Impact/projectile (non-pointed instrument) jutu=luwaturt=luwa-

shoot=spear (< Kriol jutu) erg smashes abs

yungkut=luwa-

erg fatally strikes abs

erg spreads abs about erg causes abs to sway

turrkparra=luwataly-taly=luwa-

? rubs (with hands) ? knocks up against

Idiomatic pirnpirn=luwaikirt=luwa-

8.4.3.4 paja- bite The verb paja- bite encodes a type of impact event in which the agentive component (agent or instrument) of the impact is specified for an edge. As a simple verb, pajacan be used to describe events where the ‘edge’ is a body part (e.g. teeth (1222)) or an instrument (e.g. an axe (1223)). (1222) Kunyarr-u ma=yi paja-rni. dog-erg top.aux=1sg.o bite-pst ‘The dog bit me.’ (JLI: TT76_2201: 755947_757898) (1223) Ngalipa-rlu=rlipa paja-rnanta warlu-ku. 1pl.incl-erg=1pl.incl.s bite-prs fire-dat ‘We were chopping (wood) for the fire.’ (MMN: TEN1-2017_023-01: 1094089_1106523) The type of impact event described by paja- also varies. In its simple form, and in the CVC wunpurr=paja-, paja- can encode a stabbing event not dissimilar to one encoded by la- pierce (1224). (1224) Kaniny-jarra la-ni, wunpurr=paja-rni=warru=nyanu. inside-assoc pierce-pst stab=bite-pst =incr=refl ‘He speared (deep) inside, he really stabbed himself.’

488 

 8 Complex predication

Alternatively, the CVC turt=paja- describes an event whereby an entity is separated into two components as a result of the impact event as in (1225). (1225) Turt=paja-rni. break=bite-pst ‘He bit it in half.’ (MMN: TEN1-2017-017-01: 282578_285446) In (1225), the preverb turt- contributes the semantics of ‘break’ but paja- specifies an impact event and an edge – in this particular case it has only been observed with the body part instrument ‘teeth’. Breaking events involving a snapping event with the hands as an instrument are alternatively encoded by the CVCs turt=pu- break-hit or turt=ma- break=get, while the breaking of an entity via a projectile is encoded by turt=luwa- break=shoot. Finally, paja- bite is not restricted to forceful ‘chopping’ or ‘cutting’ events (i.e. high velocity impacts) but extends to a number of other events construed more broadly as ‘impact’, as long as they involve an edge. For example, events of ‘sharpening’ or ‘whittling’ (generally lower velocity, high-friction events) can be encoded by the simple verb paja- as per this extract from a procedural text (1226): (1226) a. En kurupa=rlipa warlu-ngka kupa-rnanta. and crowbar=1pl.incl.s fire-loc cook-prs ‘And we heat the crowbar in the fire.’ b. Yalu-jangka=rlipa marda-rnanta, paja-rnanta=rlipa. that-ela=1pl.incl.s have-prs bite-prs=1pl.incl.s ‘After that we take a hold of it (the crowbar) and we sharpen it.’ (MDN: LC22a: 1034398_1040355) The verb paja- also categorises a small number of events which are not canonical impact events but are rather bodily actions associated with the mouth. In the case of nuku=paja- kiss=bite (1227), it is possible to view the lips somewhat metaphorically as an ‘edge’ involved in an ‘impact’ event (i.e. kissing). The extension of this compositional analysis to another bodily event, jalk=paja- ‘taste=bite’ in (1228), is somewhat less convincing, however. (1227) Nuku=paja-rni=rna jingka lamparn. kiss=bite-pst=1sg.s child little_one ‘I kissed that small child.’ (MMN: TEN1-2018_001-02: 891807_900375) (1228) Jalk=paja-ka lirra-ngku! taste=bite-imp mouth-erg ‘Taste this! (Lit. ‘Try this with your mouth!’)’ (PSM: TEN1-2017_010-01: 1179751_1184461)

8.4 Productive inflecting verbs 

 489

A final idiomatic event type is paja- in its simple verb in the reflexive clause in (1229) ‘to have a cramp’. This is one example of an event categorised by paja- in which there is no overt ‘edge’ nor any literal impact event. (1229) Paja-rni=rna=nyanu bite-pst=1sg.s=refl ‘I have a cramp.’ (Cataldi, 2011, p. 246) 8.4.3.4.1 Summary of CVCs Impact (edged instrument) wunpurr=pajaturt=pajajirlmung=paja-

erg wounds abs erg breaks in half abs erg cuts up abs

Bodily action nuku=pajajalk=paja-

erg kisses abs erg tastes abs

Idiomatic jalkuly=paja-

erg softens abs

8.4.4 Other productive verbs 8.4.4.1 nga- eat The transitive verb nga- is glossed eat but covers two types of ingestion: eating and drinking. Inhaling is not encoded by nga-. This event is instead categorised by maget or kiji- throw in CVCs involving the preverb ngayirr- ‘breathe’. The two senses available to nga- can be used without the addition of preverbs. Disambiguation of the specific event is primarily achieved through context or the specification of the object being ingested. Significantly, there are additional verbal predicates available for expressing ‘drink’ – the inflecting verb nguku- or the near-synonym CVC nguku=kiji- ‘drink quickly’, an event which appears to be primarily differentiated by speed of ingestion. When appearing as a simple verb, nga- appears to have three metaphoric or idiomatic secondary senses of ‘murder/slaughter/defeat’ (1230), ‘clean out’ (Cataldi,

490 

 8 Complex predication

2011, p. 204), and ‘love someone’ (1231). These additional senses have primarily been reported by Cataldi (2011) and are not common in the corpus. (1230) Nga-rni-nyirra=yanu, lurrij, jingka, parntany-parntany, ngantany, eat-pst-narr=3pl.o finish child old_woman-rdp man nga-rni-nyirra=yanu lurrij, kaji ngurra=mipa karri-nya. eat-pst-narr=3pl.o finish condit camp=restr be-pst ‘It murdered them, all of them, children, women, men, it murdered them until only the camp remained.’ (Cataldi, 2011, p. 204) (1231) Nga-rni-nyirra=yanu. eat-pst-narr=3pl.o ‘He was loving them.’ (PSM: TEN1-2017_010-01: 1894432_1896046) In CVCs, the range of preverbs occurring with nga- are notably restricted in type and number and appear only to specify the manner of ingestion, as in (1232). (1232) Kuny-kuny=nga-rni, nga-rnanta=rnalu ngarlu. suck-rdp=eat-pst eat-prs=1pl.excl.s sugarbag ‘We used to suck it, (thus) we would eat the honey.’ (TJNL LC28a: 287712_29223) A small number of idiomatic usages are listed in §8.4.4.1.1 and include an idiomatic expression for ‘sunrise’ encoded by the CVC rangkarr=nga- first_light=eat, as in (1233). (1233) Kakarrara kaji rangkarr=nga-rni, rarrpa=karri-nya east condit first_light=eat-pst dawn=be-pst ‘In the east when the sun rises and the sun dawns.’ (TJN: Manungka_01-035155) 8.4.4.1.1 Summary of CVCs Manner of ingestion kuny-kuny=ngawaapapa=ngajaku=nga-

erg sucks abs ? consumes everything erg steals away and eats abs

purangu. sun

8.4 Productive inflecting verbs 

 491

Idiomatic lurrij=nga-413 turt=ngarangkarr=nga-

erg murders abs erg loses abs ? rises (the sun)

8.4.4.2 nya- see The verb nya- primarily categorises events of perception. As a simple verb, nyacanonically refers to vision (1234) and often appears in CVCs that further specify the manner of visual perception, for example lirrki- ‘stare sidelong’ (1235) and jirrpi‘watch out for’ (1236). (1234) Nyunpula nga=rna=ngkupula nya-nganta. 2du seq.aux=1sg.s=2du.o see-prs ‘I am looking at you two.’ (PJA: TT76_2601: 448816_453024) (1235) Nyampa-ku=yi=n lirlki=nya-nganta? what-dat=1sg.o=2sg.s stare_sidelong=see-prs ‘Why are you staring sidelong at me?’ (PSM: TEN1-2018_012-01: 1092690_1095848) (1236) Kula=lu=yanu jingka jirrpi=nya-nku.414 neg=3pl.s=3pl.o child watch_out=see-pot ‘They will not look out for the children (as they should).’ The meanings of nya-, however, go beyond visual perception. For example, nya- is involved in the CVC purda=nya ‘listen=see’ meaning ‘hear/listen’ or ‘understand’. This CVC itself has multiple senses with various case frames. All of the event types encoded by purda=nya relate to cognitive processes and include (at least) ‘listen/ hear’, ‘listen out for’ (1237), (1238) ‘understand’ (1239), and ‘believe’ (1240). (1237) Ngaju ma=rna=ngku purda=nya-nganta nyuntu. 1sg top.aux=1sg.s=2sg.o listen=see-prs 2sg ‘I am listening to you./I hear you.’ (JLI: TT76_2201: 666607_670003)

413 While this meaning appears to be somewhat conventionalised as ‘murder’, lurrij- is nevertheless an adverbial preverb and can also combine with nga- to simple mean ‘eat x up completely’, i.e. ‘to finish a meal’. 414 The clausal negator kula in this example is a Warlpiri borrowing. The typical Ngardi clausal negator is wakurra.

492 

 8 Complex predication

(1238) Ngantany-ju=rla purda=nya-nganta jingka-ku. man-erg=3sg.obl listen-see-prs child-dat ‘A man is listening out for a child.’ (MMN: TEN1-2018_026-01: 784938_795991) (1239) Kardiya-rlu nya-nganta kulanganta, kulanganta “oh, nice painting”. white_man-erg see-prs cf cf oh_nice_painting ‘Whitefellas look at it and mistakenly think, “oh, nice painting”.’ Wakurra=lu purda=nya-nganta that’s walya. neg=3pl.s listen=see-prs that_is country ‘They don’t understand that that’s country.’ (MAM: TEN1-2016_002-03: 41886_349430) (1240) Wakurra=yirla ngaju-ngka purda=nya-nganta. neg=1sg.lct 1sg-loc listen=see-prs ‘She doesn’t believe me.’ (PSM: TEN1-2018_030-01: 2579703_2581700) The following CVCs involving nya- are unusual in that they do not denote primarily perceptual events but rather (emotional) states (1241), or events of solicitation (1242). (1241)

Nyampa-ku=yi=nta maka-maka=nya-nganta? what-dat=1sg.o=2pl.s anger-rdp=see-prs ‘Why are you all angry with me?’ (MDN: Cataldi, 2011, p. 255)

(1242) Maju, ngunyju yu-ngka=yi, jaaly-jaaly=nya-nganta=rna=ngku=rla. indeed tobacco give-imp=1sg.o envious=see-prs=1sg.s=2sg.o=3sg.obl ‘Indeed, give me some tobacco, I beg you desperately for it.’ (MDN: Cataldi, 2011, p. 48) The preverb tirrk- can be added to the CVC purda-nya which appears to generate an event type with a reflexive argument structure depicting an emotional state ‘to worry oneself’ (1243) rather than a simple act of perception. (1243) Nyampa-ku=n=nyanu tirrk=purda=nya-nganta? what-dat=2sg.s=refl worry=listen=see-prs ‘What are you worrying yourself over?’ (Cataldi, 2011, p. 298) It is possible that the CVCs in (1241)–(1243) primarily denote events of perception and this is masked by the choice of English translation. That is to say, it is uncertain whether the following alternative translations are possible: ‘Why are you watching me in anger?’, ‘I watch you jealously (for) it.’, and ‘Why are you anxiously perceiving yourself?’

8.4 Productive inflecting verbs 

 493

8.4.4.2.1 Summary of CVCs Visual perception jirrpi=nyakirnik=nyalirlki=nya-

erg looks out for abs erg stares at abs ? stares sidelong

paparr=nyapina=nyayangkarra=nya-

? mistakes someone (for something) erg recognizes abs ? watches abs pass by

Non-visual perception maly=nyaparnti=nya-

? ponders ? erg sniffs/smells abs

purda=nyaraly=nya-

abs understands abs* erg observes abs in jealousy

* Various alternative case frames

Idiomatic jaaly-jaaly=nyamaka-maka=nya-

erg begs abs erg is angry with abs

8.4.4.3 wangka- speak The verb wangka- speak categorises a range of event types that relate to sound emission, including human speech, as previously illustrated in (1184), but also non-human vocalisations such as barking, as in (1244).415 (1244) Rung=wangka-nya. bark=speak-pst ‘He (the dog) barked.’ (BSE: TEN1-2018_012-03: 1652163_1655910) The verb wangka- may also categorise events of non-verbal communication, as in (1245). (1245) Yala ma warrpaj-warrpaj=wangka-nyanta, nyampa-ku? that top.aux wave_hand-rdp=speak-prs what-dat ‘That one is beckoning with his hands, what for?’ (PSM: TEN1-2018_03502: 1354511_1356995)

415 Alternative CVCs for ‘(to) bark’ in Ngardi are rung=pu- (bark=hit) or rung=ma- (bark=get).

494 

 8 Complex predication

The act of (hand-)signing appears to be variously encoded via a CVC which integrates the nominal marumpu ‘hand’ or else utilises the simple verb wangka- in combination with a proprietive-marked nominal as illustrated in (1246).416 (1246) Marumpu=wangka-nyanta nawu, marumpu-kurlu wangka-nyanta. hand=speak-prs now hand-prop speak-prs ‘He is signing now, he is speaking with his hands.’ (MMN: TEN1-2019_00501: 1635374_1642061) 8.4.4.3.1 Summary of CVCs Speech and sound emission iyindij=wangkajaaly=wangkarawu=wangkaruju=wangkalirra=wangkajaaly=wangkajarla=wangkajarrak=wangka-

abs laments, complains erg gossips about abs abs cries erg teases abs abs speaks literally abs gossips abs speaks passionately abs speaks angrily/shouts

jilyirr=wangkamarumpu=wangkarung=wangkayawuru=wangkayupuru=wangkawarlka=wangkawarrpaj=wangka-

? speaks offensively abs handsigns abs barks abs speaks softly abs speaks quickly abs tells lies abs handsigns

8.4.4.4 ngarri- tell The verb ngarri-, despite being a verb of speech, does not appear to pattern with the same set of preverbs observed with wangka- speak. For example, it is not found co- occurring with preverbs expressing manner of sound emission. Instead, ngarritell categorises a more restricted set of events associated with conveying information to a recipient. Unlike wangka- it combines with pina ‘knowledge’ as in (1247) to encode a transfer of information – and in this function it appears similar to pina=yuknowledge=give.417

416 Note that the generic term for hand in Ngardi (and Warlpiri) is taka. However, taka is not integrated into any CVC meaning ‘to handsign’. 417 Importantly, ngarri- tell encodes an act of speech unlike yu= give. Thus pina=yu- may imply teaching by any method (including non-verbal modes), while pina=ngarri- implies that a recipient is informed via speech.

8.4 Productive inflecting verbs 

 495

(1247) Pina=ngarri-ku=rna=ngku waljirri. knowledge=tell-pot=1sg.s=2sg.o Dreaming ‘I will tell you a Dreamtime story.’ (PRN: TG1_0034-02) The verb ngarri- can be used in its simple form to refer to the act of ‘calling’ or ‘referring’ to something by its name as in (1248); or it can combine with the nominal yirdi ‘name’ to encode an event of ‘name-calling’, as in (1249). (1248) Kardiya-rlu ngarri-rnanta bloodwood. white_man-erg tell-prs bloodwood ‘Whitefellas call it bloodwood (referring to kuntupungu).’ (PSM: TEN12018_033-01: 15586_25296) (1249) Kuja=n jayanta yirdi=ngarri-nanta. sub=2sg.s always name=tell-prs ‘That person, whom you are always calling names.’ (DMN: LC42b: 170075_172091) 8.4.4.4.1 Summary of CVCs Transfer of information iyindij=ngarriyirdi=ngarringurr-ngarriparta=ngarripina=ngarri-

? gossips erg calls abs names ? reports that erg coerces abs to silence erg informs/tells abs abs

8.4.4.4.2 Preverb/verb doublets and grammaticalisation of preverbs A number of the productive inflecting verbs described in the previous subsections are candidates for being situated along a cline of ‘grammaticalisation’ in the sense that they appear to have some functions typically associated with verbalising or valency changing derivational morphology. Schultze-Berndt (2000, p. 541) makes a similar point for Jaminjung: One could imagine that some of the high-frequency verbs in Jaminjung, which have several polysemous senses and also form complex verbs which have to be regarded as idiomatic, would be good candidates for further grammaticalisation along these lines. These include -arra put, -mili get, -ma hit and -yu(nggu) say/do.

With respect to the Ngardi data, such candidates include karri- be (§8.4.1.1), ma- get (§8.4.1.2), pu- hit (§8.4.3.1) and wanti- fall (§8.4.1.4).

496 

 8 Complex predication

Additionally, there are a number of non-productive inflecting verbs which are likely historically derived via the univerbation of erstwhile CVCs. Laughren (2016a), for example, has drawn attention to the fact that there are formal and semantic pairings between certain preverbs and verbs synchronically in Warlpiri. Preverb-verb ‘doublets’ of a similar nature are found in Ngardi and are listed exhaustively in Table 106, cited in their present tense forms. As argued for Warlpiri, it is quite plausible that the complex verbs are the older forms and the simple verb forms are the more recent innovations. Indeed, such a process of univerbation of complex verb forms is also observed at a more systematic level in other northern Australian languages such as Gooniyandi (McGregor, 1990) and Nunggubuyu (Heath, 1984). Table 106: Preverb/Verb Doublets in Ngardi.418 Preverb + Verb

Gloss

Verb

Gloss

irrkal=ka-nganta jingkirdi=ma-nanta parnti=ma-nanta yapart=pu-nganta nyuri=ma-nanta

accompany=carry-prs laugh=get-prs smell=get-prs run=hit-prs knead-get-prs

irrki-nyanta (I) jingkirdi-nyanta- (I) parnti-nyanta (I) yapartu-nganta (III)418 nyuri-rnanta (II)

accompany-prs laugh-prs smell-prs run-prs knead-prs

The verbal form yapartu- ~ yaparti- is a simple lexical verb and its origin as a preverb-verb combination yapart=pu- has been described in §8.3.1.3. An example of the preverb yapart is provided in (1250). (1250) Ngu=rlijarra=nyanu yu-ngani-nyirra yapart. seq.aux=1du.excl.s=refl give-ipfv.pst-narr run ‘We were racing each other.’ (MMN: TEN1-2018_029-01: 986882_1028999) A number of additional verbs are likely to have also derived historically from historical preverb-verb combinations but for which the synchronic evidence is less clear. Specifically, a number of synchronic simple verbs in Ngardi can be identified which all terminate in an element -rri: ngarlarri- laugh, yurirri- move, karri- be, ngirrjirrideafen, nirrjirri- tremble and winpirri- whistle. A-rri verbal morpheme is used as a productive verbaliser in Western Desert languages, for example Wangkajunga (Jones, 2011) and a similar verbaliser may have played a role in Ngumpin-Yapa prehistory (see Nash, 2008).419 Interestingly, there exists (historically complex) mono-

418 CVCs involving yapartu- are described in 8.4.2.1. 419 See also Laughren (2016c) and Schultze-Berndt (2003, p. 540) for a discussion of cases of two verbal components becoming one in languages of Northern Australia.

8.4 Productive inflecting verbs 

 497

morphemic verbs in neighbouring Warlpiri which were never fused in Ngardi. For example, Warlpiri has a simple verb parntarri- ‘crouch’ while Ngardi only has a CVC: parnta=karri- ‘crouch=be’. While the origins of Warlpiri verb roots have been examined in detail in Nash (2008), the interesting similarities and differences in the Ngardi data call for further investigation of the origin of verb roots in Ngumpin Yapa languages more broadly.

9 Grammatical relations and argument structure The Ngardi clause minimally consists of a predicate and (unless elided) the core arguments it subcategorises for. The clausal predicate itself can be a fully inflected verb or a non-verb (such as a preverb or nominal). Core arguments optionally take the form of case-marked NPs but all NPs representing core arguments can nevertheless be elided from the clause in a process of null anaphora. Clausal arguments remain interpretable through the presence of cross-referencing bound pronouns which typically attach to some primary constituent – typically a verb or a dedicated auxiliary or propositional particle. Realisation of bound pronouns in the pronominal complex is dependent on factors of animacy, person/number and specific combinations of core arguments as described in Chapter 5. In this chapter, the relationship between case-marked NPs, bound pronouns and grammatical relations will be defined in language-specific terms. As mentioned above, a central feature of the Ngardi clause is the possibility for the elision of any and even all non-predicate constituents of a clause. This basic feature means the venerable criterion of ‘optionality’ cannot be applied as central to any distinction between would-be ‘arguments’ and ‘adjuncts’ in Ngardi. Instead, the bound pronouns and their relationship to case marking on NPs are used as the diagnostic means for defining a range of argument types: subject, object, indirect object and subjunct. The chapter is organised as follows. Grammatical relations in Ngardi are first defined and exemplified (§9.1.1) and are subsequently compared with the crossreferencing possibility of case-marked NPs in Ngumpin-Yapa languages (§9.1.2). Basic clause types in Ngardi are then described in terms of argument structure and the sets of possible case frames available to different predicate types (§9.2). This includes discussion of non-verbal predicates (§9.2.1), copula constructions (§9.2.2) and verbal predicates (§9.2.3). Secondary predicates are described in §9.3. This chapter concludes with a description of the syntactic expression of possession in Ngardi (§9.4). Possession is included in this chapter since its description relies heavily on the explanation of grammatical relations and argument structure presented here. Other topics relevant to simple clauses such as negation and questions/interrogation are described in Chapter 8 due to the fact that these clausal constructions involve morphology pertinent to the expression of clausal modality.

9.1 Grammatical relations Grammatical relations (such as subject, object, indirect object) are often deployed as primitives in both theoretically explicit frameworks (e.g. Relational Grammar (Blake, 1990; Perlmutter, 1980, 1983; Perlmutter & Rosen, 1984)) as well as descriptive grammars. In the case of Australian languages, a priori concepts of subject, object, https://doi.org/10.1515/9783110752434-009

9.1 Grammatical relations 

 499

indirect object and ‘arguments’ and ‘adjuncts’ are consistently used in grammars (see  Meakins  & Nordlinger, 2014 and Gaby, 2017 for recent examples). The main parameters of typological difference within Australian languages are reducible to whether or not ‘grammatical relations’ are predominantly understood to be encoded via verbal morphology (i.e. ‘head- marking’ as observed in many non-Pama-Nyungan languages) or via case morphology (i.e. ‘dependent-marking’ as observed in many Pama-Nyungan languages) (see Nichols, 1986; Nordlinger, 2014). Importantly, however, the Australian continent is also home to some languages  – among which are all of the Ngumpin-Yapa languages – which exhibit properties of both head-marking and dependent-marking (see Nordlinger, 2014). Despite possessing extensive case morphology, an analysis of grammatical relations in Ngumpin-Yapa languages via case morphology alone is hindered by: i) rampant ellipsis of NPs (commonly referred to as ‘null anaphora’), ii) the polyfunctionality of case forms, and iii) the fact that the enclitic bound pronouns do not cross-reference the same categories as observed in the nominal case-marking paradigm. In fact, one must conclude that the bound pronouns do not mark the same ‘case categories’ that are superficially observed as case marking on NPs in the clause. In this vein, a substantial literature has been amassed on the topic of morphological ergativity and the status of ‘deep’ or ‘syntactic ergativity’ (see Laughren, 2017e for a summary) as well as influential revisionist work on the issue of how to resolve ‘surface’ morphological marking with respect to ‘deeper’ case relationships that correspond to underlying syntactic relations (Goddard, 1982; Legate, 2008b). This has also led various analysts to propose that ‘true arguments’ in these languages are not the case-marked NPs (cf. Simpson, 1991; Legate, 2002) but rather the ‘cross-referencing pronominals’ themselves (Jelinek, 1984; Laughren, 1989; Meakins, 2014). A third alternative has been to treat neither case-marked NPs nor the pronominal clitics as the true arguments, but instead propose phonologically null arguments that simultaneously relate to both overt representations (Baker, 1996; Pensalfini, 2004). This approach posits that functional features of an ‘argument’ (person, number, syntactic case, etc.) are assigned to the bound pronoun on the one hand while encyclopaedic (i.e. lexical) features are assigned to adjoined NPs. In this grammar, I take the pronominal argument view that grammatical relations are primarily indexed by the bound pronouns, due to the frequency of NP ellipsis in Ngardi. Case-marked NPs, when they do occur, are viewed as compositionally encoding grammatical relations and as a means of lending prominence to a referent within the discourse.

9.1.1 Defining grammatical relations: Arguments and adjuncts A long-standing distinction in the discussion of ‘grammatical relations’ is that of ‘arguments’ and ‘adjuncts’ (see Haspelmath, 2014, for a recent summary). This bifur-

500 

 9 Grammatical relations and argument structure

cation in grammatical relations hinges on the observation that certain linguistic expressions in a clause are more central or obligatory to the predicate (Dixon, 1980, p. 294) while others are more peripheral to the predicate and modify the event ‘as a whole’ (see  Dowty, 2000). The criterion of obligatoriness is often framed as both a semantic and syntactic one.420 That is, arguments are ‘obligatory’ in that they are required in order to make logical sense of the predicate – for example English kill implies a killer (agent) and one killed (patient). In configurational languages, these constituents are also ‘required’ by the syntax of the language since they fill structural syntactic positions – that is, complements are sisters to heads while adjuncts are sisters to intermediate projections. In Ngardi, many of the classic criteria (e.g. obligatoriness, valency-changing processes, clausal positions, proximity to the verb) are either inappropriate or difficult to apply. As emphasised already, there are barely any NPs which are strictly ‘obligatory’ in a syntactic sense and most can be freely elided under agreement with a pronominal clitic, as in (1251).421

(1251)

Nga=rna=pulany marda-rnanta. seq.aux=1sg.s=3du.o have-prs ‘I have two (girls).’ (RJD: LC32a: 136478_137854)

In fact, a clause in Ngardi may entirely consist of a predicate and encliticised pronominal clitics, as in (1252) and (1253).

(1252)

Marda-rni=yi=lu. have-pst=1sg.o=3pl.s ‘They looked after me.’ (PLN: LC47a: 38091_40091)

(1253)

Ngurrpa=rna. ignorant=1sg.s ‘I don’t know.’ (TJN: LC28_20000810_a: 1302689_1304024)

Furthermore, Ngardi, like all Ngumpin-Yapa languages, lacks evidence for VP constituents which can be used to examine criteria like ‘closeness to the verb’, or ‘positional restrictions’. Unlike in configurational languages, subjects and object relations in Ngardi cannot be defined based on their positions within structural hierarchies (e.g. head or complement of a syntactic phrase). Ngardi also lacks a voice alternation

420 Although grammatical relations (GRs) are nearly always distinguished from semantic roles (SRs) or thematic roles (θ-roles). 421 Exceptions include: cognate objects and secondary objects, all of which lack cross-referencing and must be realised as overt NPs. See similar observations by Simpson (1991, pp. 339–345) for Warlpiri.

9.1 Grammatical relations 

 501

involving passive or applicative type constructions.422 With this in mind, the available criteria for determing argument/adjunct status are listed in (1254).

(1254) i) ii)

Criteria for argument/adjunct status cross-referencing and case marking behaviour under reflexivisation a. subject arguments ‘control’ reflexives b. non-subjects arguments are ‘replaced’ by reflexive clitics under reflexivisation c. adjuncts are not involved in syntactic reflexivisation.

The first criterion refers to the co-occurrence properties of case on NPs (where present) and the bound pronouns. The second criterion, ‘behaviour under reflexivisation’, refers to properties observable in the pronominal clitic complex in reflexive/ reciprocal constructions. Specifically, it refers to the ability for arguments and only arguments to either: bind reflexives (subjects), or be themselves replaced by reflexive clitics (objects, indirect objects and subjuncts). Adjuncts are not affected by syntactic reflexivisation. The set of grammatical relations identified by these criteria are presented in Table 107. Table 107: Canonical grammatical relations in Ngardi. Grammatical Relations

Subsection

Case marking

Pro. clitic

Subject (A, S)423 Object (O)424 Indirect object Subjunct Adjuncts

§9.1.1.1 §9.1.1.2 §9.1.1.3 §9.1.1.4 §9.1.1.5

erg, abs abs dat loc, all Various

sbj obj obl lct —

Bind refl args ✓ — — — —

Substituted with refl clitics — ✓ (o refl) ✓ (o refl) ✓ (lct refl) —

Table 107 lists the features of the main grammatical relations in Ngardi. Excluded from Table 107 are a number of grammatical relations which do not fit neatly into these categories. These include ‘primary’ and ‘secondary’ objects in double object constructions (Dryer, 1986) which are discussed separately as ‘non-canonical arguments’ in §9.1.1.6.

422 The only valency alternating strategy is essentially the choice of inflecting verb in combination with a preverb (e.g. wuruju=karri- good-be ‘to be good/well’ vs wuruju=ma- good=get ‘to make better’). 423 Subjects of quasi-transitives (§9.2.3.3) are a non-canonical subject type (see §7.1.1.7). 424 Secondary objects in double object constructions are a non-canonical object type and are discussed in §9.1.1.6. Double object constructions themselves are described in detail in §9.2.3.9.

502 

 9 Grammatical relations and argument structure

Canonical grammatical relations in Ngardi include two types which are somewhat anomalous in the broader literature. Firstly, my use of the term ‘indirect object’ encompasses not only the marking of recipients in classic three-place predicates, but also extends to a range of additional dative case-marked NPs cross-referenced by the oblique bound pronouns. Secondly, certain locative and allative marked NPs are cross-referenced by a unique series of locational bound pronouns. These are referred to as subjuncts, a term adopted from Tsunoda’s (1981b, pp. 112–115) description of Jaru. The primary feature which defines these two classes of relations as arguments is their cross-referencing by a unique bound pronoun. In most cases (barring so-called ‘ethical datives’), they are combinatorially restricted to a subset of predicate types and their thematic role is not entirely predicted from their case marking alone – all properties generally considered to be associated with arguments. 9.1.1.1 Subjects Subjects have the following properties: (i) They are cross-referenced by a subject pronominal clitic (∅ in the 3sg), unless inanimate. (ii) When realised as overt NPs, they are ergative case-marked or appear unmarked (absolutive case). (iii) They can bind arguments encoded by the anaphoric enclitics =nyanu/=nyanungkurla in reflexive/reciprocal constructions. Examples include (1255)–(1257).

(1255)

Ngantany-ju=lu kuma-rnanta puliki. man-erg=3pl.s cut-prs cattle ‘The men are cutting up a bullock.’ (MMN: TEN1-2018_027-01: 743967_763983)

(1256)

Ngantany=pa=lu ya-nanta wirlinyi. man=ep=3pl.s go-prs hunting ‘The men are going hunting.’ (PSM: TEN1-2017_023-01: 894716_899203)

(1257)

Yalu-kujarra-rlu=pula=nyanu kuli jup=ma-ni. that-dual-erg=2du.s=refl anger begin=get-pst ‘Those two started fighting each other.’ (YMN: LC23a: 1445989_1449385)

9.1.1.2 Direct objects Direct objects have the following properties: (i) They are cross-referenced by object pronominal clitics (Ø for 3sg), unless inanimate or in a ditransitive clause in which case they may lack cross-referencing. (ii) When realised as overt NPs, they are absolutive case-marked (i.e. unmarked).

9.1 Grammatical relations 

 503

(iii) The cross-referencing object series bound pronoun is replaced by the anaphoric enclitic =nyanu in reflexive/reciprocal constructions. Examples include (1258)–(1260).

(1258)

Ngu=pulany yala-kujarra pi-ngi wati-kujarra. seq.aux=3du.o that-dual hit-irr man-dual ‘He was about to kill those two men.’ (TJN: LC20a)

(1259)

Turn=ma-nanta=rna yarla jangilany-ja, warlu-ngka. cover=get-prs=1sg.s yam fire-loc fire-loc ‘I am covering up the bush potatoes in the ashes of the fire.’ (MMN: TEN12017_020-02: 660436_670805)

(1260) Kunyarr-u=lu=nyanu

wirrpa-ngku pangi-rnanta. dog-erg=3pl.s=refl many-erg dig-prs ‘That big group of dogs are scratching themselves.’ (PSM: TEN1-2017_022-01: 47642_50345)

Note that in (1260) an overt absolutive NP kunyarr ‘dog(s)’ registering the direct object cannot appear in the clause. 9.1.1.3 Indirect objects Indirect objects have the following properties: (i) They are cross-referenced by object/oblique bound pronouns (=rla for 3sg). (ii) When realised as overt NPs they are dative case-marked. (iii) The cross-referencing object/oblique bound pronoun is replaced by the anaphoric clitic =nyanu in reflexive/reciprocal constructions. One function of the indirect object relation is to encode the thematic role of recipient in canonical three-place predicates (e.g. yu- give) as in (1261).

(1261)

Kuyi=rla yi-nya murtilya-rlu kunyarr-ku. meat=3sg.obl give-pst boy-erg dog-dat ‘The boy gave the meat to the dog.’ (NMN: LC21b)

Indirect objects also appear in extended intransitive clauses (e.g. wangka- abs ‘speaks to/for’ dat) as in (1262) or ‘conatives’ – either of the obligatory type (e.g. kuny=pu- erg

504 

 9 Grammatical relations and argument structure

‘waits for’ dat) (1263), or the variable type425 (e.g. nya- erg ‘looks for’ dat) (1264) – all of which are further described and exemplified in §9.2.3.4.

(1262)

Japanangka=yi wangka-nya, “ya-nku=rli!” ♂subsection=1sg.o speak-pst go-pot=1du.incl.s ‘Japanangka said to me, “Let’s go!”’ (BSE: TEN1-2017_015-01: 95179_98349)

(1263)

Nganayi-ku ngatina-ku=rla kuny=pu-nganta ita-ngku. whatsit-dat M-dat=3sg.obl wait=hit-prs small-erg ‘The child is waiting for (his) mother.’ (SCM: TT76_1301: 39476_41914)

(1264) Nya-ngku=rna=rla

ngapa-ku. see-pot=1sg.s=3sg.obl water-dat ‘I am looking for water.’ (MMN: TEN1-2017_002-05: 1129818_1132767)

Nevertheless, the ‘indirect object’ relation in Ngardi extends well beyond the typologically common thematic roles described above. Specifically, my use of the term ‘indirect object’ extends to cover what have also been termed ‘external objects’ (Simpson, 1991, p. 380) or ‘adjunct/ethical datives’ (Hale, 1982; Simpson, 1983, 1991) in Warlpiri. According to Simpson (1991, p. 379), the function of an ‘ethical dative’ is to ‘indicate that the action or state denoted by the verb has some relation to another participant external to it’. It includes various relations sometimes described as ‘(un)intentional benefactive or adversative’ as well as ‘possessor ascension’ (Simpson, 1991, p. 379). These datives provide a type of information that is not strictly ‘semantically obligatory’ to the predicate (Koenig, Mauner & Bienvenue, 2003) but nevertheless, they are registered in the clause as if they were equivalent to other canonical arguments (see  similar comments by Simpson (1991, p. 380) for Warlpiri). The fact that NPs bearing such semantic roles behave like indirect objects in Ngardi is not unusual. English, for instance, permits certain ‘benefactives’ with certain verbs to be ‘promoted’ to the structural positions of ‘objects’, as in the alternation: ‘Tom baked bread for Jayden’ and ‘Tom baked Jayden bread’. Compare then the identical pronominal indexing of the ‘indirect objects’ in ditransitive (1261), extended intransitive (1262) and conative (1263) clauses above with the indirect objects below with various functions of: topic of knowledge/speech (1265), affected entity (1266), (1267) benefactive (1268) and external possession (1269).

425 The obligatory type refers to predicates which can only select an erg-dat case frame, while the variable type refers to predicates which can take erg-dat or erg-abs.

9.1 Grammatical relations 

 505

(1265)

Mu-ku=rna=yanu wangka-ju ngantany-ku pawungurru-ku. this-dat=1sg.s=3sg.o speak-pot man-dat many-dat ‘I will speak about/for these man.’ (SGO: T1091_1: 373633_379483)

(1266)

Parntany-parntany-ku=yanu pirriya=karri-ju. old_woman-rdp-dat=3pl.o cold=be-pot ‘It will be cold for those old ladies.’ (PSM: TEN1-2018_030-01: 993879_998059)

(1267)

Ngu=rna yangi, kala ma=yi lankarr=ya-ni yangi. seq.aux=1sg.s one asrt top.aux=1sg.s split=go-pst one ‘I (have) one, but (the other) one split on me.’ (TJL: TT76_2201: 1940014_ 1943160)

(1268) Kupa-rni=nta=rla

nyanungu-ku pirranginti. cook-pst=2pl.s=3sg.obl recog1-dat afternoon ‘You lot cooked it for him yesterday (afternoon).’ (MMN: TEN1-2017_030-01: 1249268_1252760)

(1269)

Ngaju-ku=yi panji karri-nyani nganayi-rla. 1sg-dat=1sg.o WB be-ipfv.pst whatist-loc ‘My brother-in-law was staying there.’ (PSM: TEN1-2018_038-01: 52435_62029)

All of the dative NPs in (1261)–(1269) are treated identically by Ngardi syntax. All of them are able to be cross-referenced by an object/oblique bound pronoun and are expressed by the reflexive clitic -nyanu in reflexive-reciprocal constructions. Consider (1270)–(1272).

(1270)

(Ngaju-ku) wapirra=rna=nyanu nya-nku. 1sg-dat F=1sg.s=refl see-pot ‘I will see my father.’ (PSM: TEN1-2018_033-03: 1651816_1656065)

(1271)

Marda-rni=nyanu nyanungu-ku warlalja. have-pst=refl recog1-dat own ‘He kept it for himself.’ (PSM: TEN1-2017_028-01: 2320342_2325311)

(1272)

Jaja-kujarra=pula=nyanu wangka-nya Warlpiri. MM-dual=3du.s=refl speak-pst language_name ‘The two grandmothers spoke Warlpiri to each other.’ (YMN: LC23_20000801a: 889170_891092)

It is not entirely clear to what extent an overt dative NP can co-occur within a clause in which an indirect object has undergone syntactic reflexivisation. In Warlpiri it is

506 

 9 Grammatical relations and argument structure

argued that only ‘ethical (adjunct) datives’ datives can appear in clauses in which an indirect object has undergone syntactic reflexivisation (Simpson, 1991, p. 172). In Ngardi there are some constraints. No indirect object of a ditransitive predicate has been observed co-occurring with a reflexive bound pronoun with which it is co-referent. However, indirect objects representing benefactives have been observed co-occurring with cross-referencing reflexive clitics within a clause. Thus, in (1273), the predicate yirra- place does not take any other type of dative NP other than those bearing the role of ‘benefactives’. While one could argue that the overt NP is in fact a secondary predication on the absolutive object ‘house’ meaning ‘my own’, this would be an unexpected use of the dative. Adnominal possession relating two NPs in an internal possession construction is typically encoded by the possessive -kuny or by the personal possessive -punta (see §9.4.1).

(1273)

Yirra-ku=rna=nyanui ngaju-kui. place-pot=1sg.s=refl 1sg-dat I built (a house) for myself. ‘?I built my own (house), for myself.’ (PSM: TEN1-2017_022-01: 967904_970075)

At this juncture, it is also worth considering whether conatives (§9.2.3.6) (e.g. nya[erg sees abs] vs nya- [erg looks for dat]) are fundamentally any different to certain types of indirect objects added to canonically intransitive predicates, that is, where the action denoted by a canonically intransitive verbs is directed at an ‘intensional object’426 (e.g. ya- ‘go (for)’, ngarrurda=karri ‘be happy (for)’, etc.). In Ngardi, as in English, these types of intransitive verbs can be contrasted with intransitive verbs which generally cannot (wanti- fall, jup=wanti ‘dismount’, ruraj=karri- ‘spray up (dirt)’).427 The former are analogous to certain transitive verbs which can variously register an argument as either a direct object or a ‘conative-type’ dative NP. Some indicative examples are provided in Table 108. All of the arguments in the final column of Table 108 are treated identically by Ngardi syntax and are not distinguished from any other type of indirect object. All of them are able to be cross-referenced in the bound pronoun by the ‘oblique’ clitics and are subject to replacement by the reflexive -nyanu under reflexivisation, as in (1274) and (1275).

426 I use the term ‘intensional object’ as an extension of Simpson’s (1991, p. 326) use of the term to refer to dative objects of verbal predicates which encode a ‘purpose’ (e.g. warri- search meaning ‘to search for’). 427 Although consider such metaphoric exceptions in English as ‘to fall for someone’ meaning ‘to be attracted to someone’ and ‘to jump for joy’ meaning to be ‘really excited/pleased’

9.1 Grammatical relations 

 507

Table 108: Dative NPs in intransitive and transitive clauses as ‘conatives’.

ya- go lu- cry ngarrurda=karri‘happy=be’ nya- see la- pierce pangi- dig purda=nya- ‘listen=see’

Intransitive

Transitive

Conative

Case of NPs

Bound Pronouns

Case of NPs

Bound Pronouns

Case of NPs

Bound Pronouns

abs

s

erg-abs

s-o

erg-dat

s-obl

‘x goes for y’428 ‘x cries for y’ ‘x is happy for y’

‘x goes’ ‘x cries’ ‘x is happy’ ‘x sees y’ ‘x digs up y’ ‘x digs up y’ ‘x understands/hears y’

‘x looks for y’ ‘x digs for y’ ‘x digs for y’ ‘x listens out for y’

(1274)

Li-nya=pula=nyanu. cry-pst=3du.s=refl ‘Those two cried for themselves.’ (MMN: LC03)

(1275)

Purda=nya-nya=nyanu. listen=see-pst=refl ‘She has listened to herself.’ (DMN: LC42b: 511989_516504)

Similarly, canonical transitive predicates can add an additional indirect object in a type of ‘dative applicative’ construction. Examples of reflexivisation of indirect objects in trivalent clauses include (1276) and (1277).428

(1276)

Marda-rni=nyanu nyanungu-ku warlalja. have-pst=refl recog1-dat own ‘He kept (it) for himself.’ (PSM: TEN1-2017_028-01: 2320342_2325311)

(1277)

Minya mangarri=rli=nyanu kupa-rnanta ngali-ku, this veg_food=1du.incl.s=refl cook-prs 1du.incl-dat kuriny-ku ngali-ku. two-dat 1du.incl-dat ‘You and I are cooking this food for the two of us.’ (PJA: TT76_2601: 673172_680124)

428 Dative marking of animate goals in conjunction with the predicate ya- go is rare. Animate goals are more commonly registered as a subjunct with a locative case-marked NP and a locational bound pronoun. There is a slight difference in semantics however, as locative subjuncts do not encode purpose.

508 

 9 Grammatical relations and argument structure

The variability in clause structures involving indirect objects in Table 108 and the ‘additional arguments’ in (1276) and (1277) can be contrasted with some predicates which are only found with indirect objects – that is to say they are never found in simple intransitive (abs) or transitive (erg-abs) clauses. For instance, the negative predicator waku always takes an abs-dat case frame, as in (1278).

(1278)

Waku=rna=rla ngaju ngunyju-ku. neg=1sg.s=3sg.obl 1sg tobacco-dat ‘I don’t have any tobacco (lit. ‘I am not of tobacco’).’ (PSM: TEN1-2018_03001: 482667_486150)

Additional examples of intransitive and transitive predicates which only take indirect objects are provided in §9.2.3.4 and §9.2.3.6, respectively. To summarise then, I classify all the types of cross-referenced dative NPs described above as indirect objects (a type of argument) following Simpson (1991) and not ‘adjunct datives’ (Hale, 1982; Simpson, 1983) since I reserve the latter for usages of the dative in which there is no possibility for cross-referencing (i.e. dative marked nominalised verbs functioning as purpose clauses, dative NPs marking temporal assignment etc.; see §9.1.1.5). 9.1.1.4 Subjuncts Subjuncts have the following properties: i) They are cross-referenced by locational bound pronouns. ii) When realised as overt NPs, they are case-marked with the locative or the allative. iii) The cross-referencing locational pronominal clitic is replaced by the reflexive/ reciprocal anaphoric enclitic =nyanungkurla (see §6.3.4). Examples include (1279)–(1281). (1279)

Ngu=rna=ngkupularla yut=karri-nyanta nyunpula-rla. seq.aux=1sg.s=2du.lct sit=be-prs 2du-loc ‘I am sitting with you two.’ (PSM: TEN1-2017_028-01: 1477238_1479595)

(1280)

Yangi ya-ni=yanurla wirrpa-kurra. one go-pst=3pl.lct many-all ‘One fella went off towards that big mob.’ (PSM: TEN1-2016_006-01: 314841_320844)

(1281)

Minya=rna=nyanungkurla makarta yirra-rnanta langka-ngka. this=1sg.s=refl.lct hat place-prs head-loc ‘I put the hat on my head.’ (PJA: TT76_2601: 640186_645787)

9.1 Grammatical relations 

 509

The thematic roles assigned to subjunct arguments vary according to the predicates that select them. The examples above all involve straighforward extensions of the core meaning of the locative as a semantic case that marks locations (1281), and of the allative as a semantic case-that marks goals (1280). Nevertheless, subjunct uses of the locative case extend well beyond expected spatial case meanings. Subjuncts can encode static and dynamic accompaniment (1279), (1376) (cf. the comitative case), movement or transfer away from an entity (1373)–(1375) (cf. the elative case), recipients of speech predicates (1341) (cf. the dative case), and sources/origin of fear (1343), shame (1344) and (dis)belief (1389) predicates (cf. the ablative or elative cases). None of these thematic roles are inherently predictable from the meaning of the locative case alone and are dependent on their co-occurrence with specific predicates. Interestingly, the same types of thematic roles are found encoded with distinct ‘subjuncts’ in Jaru (Tsunoda, 1981b, pp. 113–115) and the ‘accessory’ in Walmajarri (Hudson, 1978, pp. 62–68). 9.1.1.5 Adjuncts Adjuncts have the following properties: (i) They are not cross-referenced by pronominal clitics. (ii) They are semantically independent from the clausal predicate and may be freely added to modify either the clausal event or a single argument of the predicate. (iii) They may appear with a wide variety of case suffixes in different functions, including the ergative and absolutive (as secondary predicates). Other typical case suffixes used are dative, locative, elative, ablative, allative, and proprietive, among others. Examples include (1282) and (1283). (1282)

Pamarr-u=nganpa luwa-rnani-nyirra yapart-karra. rock-erg=1pl.excl.o shoot-ipfv.pst-narr run-ssub ‘He pelted us with rocks while we ran.’ (TJN: tjama 23-10-00)

(1283)

Mangarri-kurra=rnali ya-ni kayirra-purda. veg_food-all=1pl.excls go-pst north-wards ‘We went north to the food.’ (TSM: TT75_1301: 1177496_1180540)

Multiple adjuncts can co-occur within the one clause. In (1284), for example, an allative marked NP marking a locational endpoint and a dative marked non-finite subordinate clause marking purpose co-occur within a single clause. This example illustrates the principle of iterativity – adjunct NPs are unrestricted and can be added freely to a clause where semantically plausible.

510 

 9 Grammatical relations and argument structure

(1284)

Ka-ngku-rni=lu ngantawu-kurra kupa-rnu-ku. carry-pot-hith=3pl.s shade-all cook-inf-dat ‘They will bring it here to the shade for cooking.’ (MDN: LC22b: 328236_331261)

The possibility for multiple adjuncts to co-occur within a single clause extends to multiple adjuncts marked with the same case. For example, in (1285), the two NPs marked with the allative are not a complex NP agreeing in case. Instead, they represent two distinct adjuncts – the first (mu-kurra) encodes a locational endpoint of the motion event expressed by ya- go and the second (kuli-kurra) encodes an intransitive subject-controlled use of the allative subordinator marking purpose.429 (1285)

Ya-nta-rni mu-kurra kuli-kurra. go-imp-hith this-all anger-all ‘Come here to fight!’ (PSM: TEN1-2018_021-01: 618438_622687)

Adjunct dative NPs include datives of temporal assignment, as in (1286) and (1287), and datives marking non-finite verbs or preverbs in purpose clauses, as in (1288) and (1289), respectively. These are distinguished from indirect objects by the fact that they are never cross-referenced in the pronominal complex. (1286)

Yirra-ka-rra=lu=ngala ngaka-ku. place-imp-hence=pl.s=1pl.incl.o later-dat ‘Put it aside for us for later.’ (MDN: LC22b: 362117_364758)

(1287)

Ngaka-ku mangarri iliri=yirra-rni. later-dat veg_food hide=place-pst ‘He buried the food for later.’ (PSM: TEN1-2017_009-03: 128794_133496)

(1288)

Ngapa=rna ma-nku yarli-rnu-ku. water=1sg.s get-pot wash-inf-dat ‘I will get water for washing.’ (PSM: TEN1-2017_026-02: 830725_834719)

(1289)

Ngantany ya-nanta yut-ku. man go-prs sit-dat ‘A man goes to sit down.’ (PSM: TEN1-2018_029-01: 1368248_1372614)

429 Note that since the action nominal kuli can be used as a referential nominal ‘a fight’ or as a predicate ‘to fight’, this example could also mean ‘Come here, to the fight!’. However, in order to express the notion ‘Come here to this fight!’ the discrete proximal demonstrative minya would be used (see §5.2.1.1). For discussion of the possibility for action nominals to be the sole constituent of a non-finite subordinate clause, see §11.3.1.

9.1 Grammatical relations 

 511

Note that, in some cases, a dative NP may be ambiguous as to whether it is interpreted as being an overt realisation of a main clause indirect object or a subject of a non-finite subordinate clause, as in (1290). However, in this clause, the object/oblique bound pronoun in the main clause can only be interpreted as cross-referencing an indirect object, and not a constituent of a non-finite subordinate clause. (1290)

Minya=rna=ngku kuyi ka-ngani, nyuntu-ku nga-rnu-ku. this=1sg.s=2sg.o meat carry-ipfv.pst 2sg-dat eat-inf-dat ‘I brought this meat for you, (for you) to eat.’ (PSM: TEN1-2018_030-01: 1027548_1031821)

9.1.1.6 Non-canonical arguments In addition to the canonical arguments and adjuncts described in the previous sections, there are a number of non-canonical arguments in Ngardi. Generally speaking, these are NPs which exhibit case marking typically associated with arguments; however, they are either cross-referenced with an unexpected bound pronoun or else not cross-referenced at all. Their labels and properties are illustrated in Table 109. Table 109: Non-canonical grammatical relations in Ngardi. Type

Case marking

Crossreferencing

Bind refl args

Substituted with refl clitics

Primary objects Secondary objects Ergative-controlled adjuncts Absolutive, and dativecontrolled adjuncts

abs abs erg abs, dat

obl — — —

— — — —

?430 — — —

9.1.1.6.1 Non-canonical objects in double-object constructions Two main deviations from canonical relations occur in double-object constructions (§9.2.3.9) involving what I term ‘primary objects’ and ‘secondary objects’, following Dryer (1986). Double-object constructions in Ngardi comprise an ergative subject and two absolutive objects (primary and secondary) with the thematic roles of recipient and theme, respectively. This argument structure is available to just a single verbal predicate: yu- give. The two ‘non-canonical objects’ can be compared to direct objects and indirect objects in simple transitives in terms of both their case-marking and pronominal cross-referencing, as in Table 110.

430 Unfortunately, I have no data to indicate whether primary objects are able to replaced with a reflexive clitic.

512 

 9 Grammatical relations and argument structure

Table 110: Encoding and marking alignments of ‘objects’ in Ngardi. Thematic Role

Case Marking

Bound Pronoun

Direct object Primary object Secondary object

do po so

patient theme recipient

abs abs abs

o — obl

Indirect object

io

recipient, possessor, etc . . .

dat

obl

On initial inspection of case marking of direct, secondary and primary objects, Ngardi is unremarkable. The language exhibits the cross-linguistically common ‘neutral’ alignment in terms of case marking between direct, secondary and primary objects – all of them being registered as absolutive arguments. Unusually though, all three object types are registered differentially by the enclitic bound pronouns (by an object bound pronoun, by no bound pronoun, and by an oblique bound pronoun, respectively). Such ‘tripartite’ marking of do, po and so is reportedly ‘rare’ and argued to be ‘uneconomical’ (Malchukov, Haspelmath & Comrie, 2010, p. 7) in the world’s languages, although it is found in at least one other Australian language: Kayardild (Evans, 1995, p. 336). Clear, diagnostic examples of the case and cross-referencing of primary (po) and secondary objects (so) are exceedingly rare in the corpus due to ellipsis of NPs and the fact that the distinction between object and oblique cross-referencing is only apparent in the third singular. The examples (1291) and (1292) illustrate the encoding of primary and secondary objects since the relevant NPs are overt and third singular. These can be compared with the ‘non-diagnostic’ examples (1293) and (1294) since the NPs are not third singular. Subscripts po and so are used to identify primary and secondary objects. [i] Diagnostic: Secondary object is 3sg and cross-referenced with 3sg.obl =rla (1291)

Itaso=rlaso mangarripo yi-nya ngati-ngku. small=3sg.obl veg_food give-pst M-erg ‘Then the mother gave the kid the food.’ (JLI: TT76_2201: 471018_473113)

(1292)

Nga=rna=rlaso minyapo [ngaju-ku ngajayi]so. 1sg=1sg.s=3sg.obl this 1sg-dat yB ‘I will give my brother this.’ (TJN: Manungka_01-035160: 470937_475876)

[ii] Non-diagnostic: Secondary object is non-singular and cross-referenced with third dual =pulany and third plural =yanu (1293)

Yi-nya=n=pulanyso yalu-kujarraso kuyipo. give-pst=2sg.s=3du.o that-dual meat ‘You gave those two the meat.’ (MMN: TEN1-2017_024-02: 89097_91176)

9.1 Grammatical relations 

(1294)

 513

pulikiso Ya-nta-rra=lu, minya ma=rna=yanuso go-imp-hence=pl.s this top.aux=1sg.s=3pl.o cattle ngapapo yu-ngku. water give-pot ‘You lot clear out, I am going to give the cattle some water.’ (PSM: TEN12018_027-01: 641873_655945)

9.1.1.6.2 Ergative-controlled adjuncts In addition to marking transitive subjects, ergative case can mark a range of other constituents, namely certain types of secondary predicates (§9.3), as well as ergative marked subordinate clauses (§11.3.6.1). Despite being marked with a traditionally ‘grammatical’ or ‘structural’ case, these NPs are like adjuncts in all other respects – they are not cross-referenced in the pronominal complex, they are not supplied a definite interpretation if elided from a clause, they are not highly restricted (they may co-occur with any semantically compatible verbal event), and they are conceivably iterable. In fact, the assignment of ergative marking in these cases is often understood as a type of ‘agreement’ (e.g. Simpson, 1991, p. 145), or else a ‘referential use’ of case (Dench & Evans, 1988). In this grammar, these are treated as ‘secondary predicate’ usages of the ergative (see §4.5.4.2.2). While infinitive verbs, action nominals and preverbs can all take ergative case-marking, they cannot appear as sentential subjects (see similar comments by Simpson (1991, p. 110) on Warlpiri). Thus, consider the ergative-marked action nominal wirlinyi in (1295). (1295)

Karli=rna paja-rni, wirlinyi-rlu. boomerang=1sg.s bite-pst hunting-erg ‘I cut a boomerang, while out hunting.’ * ‘I, the hunter, cut a boomerang.’ (TJL: TT76_2201: 1786234_1790780)

9.1.1.6.3 Absolutive and dative adjuncts Unmarked adjuncts in the absolutive include a range of secondary predications which show ‘agreement’ with canonical objects (§9.3). Despite being marked with a traditionally ‘grammatical’ or ‘structural’ case, these forms are in all other respects akin to adjuncts – they are not cross-referenced in the pronominal complex, they are not generally supplied a definite interpretation if elided from a clause, they are not highly restricted (they may co-occur with any semantically compatible verbal event), and they are conceivably iterable. Absolutive-controlled adjuncts are generally constrained to types of depictive secondary objects, while dative-controlled adjuncts include both nominalised verbs, and temporal nominals, as were exemplified previously in (1286)–(1289).

514 

 9 Grammatical relations and argument structure

9.1.2 Comparison of cross-referencing in Ngumpin-Yapa and northern Western Desert languages The Ngumpin-Yapa and northern Western Desert languages vary with respect to what range of case-marked NPs can be cross-referenced by bound pronouns and function as clausal arguments.431 In this section, the ability of non-core cases (dative, locative, ablative, etc.) to be cross-referenced under certain conditions (animacy/person/ combinatorial hierarchies in the bound pronoun complex) is compared across the Ngumpin-Yapa and northern Western Desert languages.432 Table 111 presents the observed patterns of permissibility for non-core cases to be cross-referenced given the  equisite conditions (i.e. animacy, constraints on combinations of bound pronouns, etc.). The possibility for registration is further listed with a descriptive label used for the relevant series of bound pronouns: indirect object IO, dative dat, oblique obl, accessory acs, locational lct, ablative abl, and avoidance avoid. Table 111: Cross-referencing of case morphology in Ngumpin-Yapa and northern Western Desert languages.

Yapa Eastern Ngumpin

erg

abs

abs

dative

locative

allative

ablative

Warlpiri

s

s

o

dat

—

—

—

Warlmanpa

s

s

o

dat

—

—

—

Mudburra

s

s

o

obl

obl

obl

—

Bilinarra

s

s

o

io

—

—

—

Gurindji

s

s

o

obl

lct433

lct

lct

Wanyjirra434

s

s

o

obl

obl

obl

obl

431 Note that the southern languages of the Western Desert (e.g. Yankunytjatjara, Pitjantjatjara and Gugada) use overt nominals and free pronouns more than they utilise cross-referencing (see summary in Jones, 2011, p 149). 432 The use of canonically ‘local cases’ in the marking of certain verbal arguments (e.g. addressee, recipient, experiencer, etc.) is also a feature of case-rich systems in other language groups of the world – for instance the Tsezic languages of the Nakh-Daghestanian language family (Forker, 2010). 433 McConvell (1996b, p. 84) takes a compositional analysis of what I have labelled a locational ‘series’ (i.e. o(bject) + i(ndirect)o(bject)) and describes the combined bound pronoun as crossreferencing ‘affected locations’ (animate, generally human, NPs). McConvell comments: “what affected locations have in common is that they are cross-referenced by clitics, unlike other locations, and in a way distinct from either S, O, or IO clitics. Those persons which normally take O clitics (i.e. all except third singular) are marked with the O clitic + IO clitic -rla” (1996b, p. 84). Affected locations may be case-marked by allative, ablative, comitative or dative cases. 434 Senge (2015, 535ff.) also reports that animate NPs case-marked with the comitative can be crossreferenced by the oblique (obl) series of bound pronouns. Note that the comitative in Wanyjirra (and Gurindji) -gunyja is historically related to a dative/genitive form with a locative increment: i.e. *-gu-ny-ja.

9.1 Grammatical relations 

 515

Table 111 (continued)

Western Ngumpin

Northern Western Desert

erg

abs

abs

dative

locative

allative

ablative

Jaru435

s

s

o

obl

lct

lct

lct

Ngardi

s

s

o

obl

lct

lct

—

Walmajarri

s

s

o

obl

acs436

—

—

Kukatja

s

s

o

dat

acs

acs

avoid

Wangkajunga

s

s

o

dat

lct

lct

abl

Pintupi

s

s

o

dat

acs

acs

avoid

The grammatical status of NPs that are cross-referenced with bound pronouns other than S and O forms has been described variously for Ngumpin-Yapa and Western Desert languages. They are referred to variously as ‘indirect object and subjuncts’ (Tsunoda, 1981b, pp. 112–115) for Jaru, ‘oblique and additional arguments’ (Senge, 2015, 535ff.) for Wanyjirra, ‘oblique arguments’ (Meakins & Nordlinger, 2014, p. 363; Osgarby, 2018, p. 145) for Bilinarra and Mudburra, ‘indirect objects’ and ‘affected locations’ for Gurindji (McConvell, 1996b, pp. 56–57), and external objects (datives only) for Warlpiri (Simpson, 1991). All authors acknowledge that they appear to have a special status outside of canonical arguments and canonical adjuncts. For Bilinarra (Ngumpin) and Warlpiri (Yapa), even though no non-core cases except the dative are cross-referenced, Hale (1982, p. 272) notes that certain verbal predicates ‘select’ an argument in the locative in Warlpiri, while Meakins and Nordlinger (2015, p. 363) make the same observation for Bilinarra. A major distinction exists in the Ngumpin languages between those languages which have two paradigmatically distinct series for non-core cases (dative/oblique and locational/accessory) and those which have only one (generally labelled dative/ oblique). The former generally corresponds to the Western Ngumpin languages while the latter is found for all other Ngumpin-Yapa languages. Despite having a single paradigmatic series for non-core cases, Mudburra still allows marking of a range of noncore case-marked NPs including locative, allative and even ablative, but with the same series of bound pronouns used to cross-reference dative NPs. Warlpiri is an interesting case in which pronominal sequences formally analogous to the ‘locational’ series in Ngardi and Jaru are observed (with =rlajinta in the 3sg rather than rlanyanta) except

435 Jaru is analysed as having two allative forms and two ablative forms. All four can be cross-referenced by a locational bound pronoun (see Tsunoda, 1981b, p. 103). 436 Hudson (1978) refers to instances of the locative case that can be cross-referenced in Walmajarri as ‘accessory case’ and those that cannot as ‘locative case’. In addition, there is something of a choice in Walmajarri in the formation of the accessory bound pronoun. It is either built off a form that is used for the O relation bound pronoun or else a form that serves as the base for the obl relation. Note that the accessory bound pronoun involves the addition of a final -nyanta, whereas the obl form involves a final -rla.

516 

 9 Grammatical relations and argument structure

(with the exclusion of the 3sg =rlajinta) they are not construed of a single argument. In the case of =rlajinta this form alone can cross-reference either two dative NPs or a single dative NP (generally of the ‘conative’ type) but it is never observed cross-referencing any NP marked with a locative, allative or ablative case.

9.2 Argument structure Discussion of predicate argument structures is divided according to the lexical class of the predicating element. In Ngardi, the predicational head can either be a non-verb (§9.2.1) including nominals or preverbs; a verb (§9.2.3); or a ‘copula-like’ construction (§9.2.2).

9.2.1 Non-verbal predicates Verbless clauses may involve either a nominal, or a preverb functioning as a predicate. There is a close relationship between verbless clauses and clauses in which a copula verb karri- appears with a basic meaning ‘x is (y)’ or ‘x exists/there is x’. Copula constructions such as these will be described in §9.2.2. The major distinction between the two is that verbless clauses have no ability to inflect for TAM and hence there is no overt specification of these features in verbless clauses. Temporal and aspectual meanings are generally interpreted from context, although they tend to be interpreted as predications that hold at utterance time – or else are temporally unbounded (e.g. existential ‘there are x here’). Non-verbal predications are generally not interpreted with modal meanings, nor with illocutionary force. Two significant exceptions to this latter generalisation are negative commands comprising non-finite verb forms (§10.7.1.2) and illocutionary nominals (§10.7.3)). Three subtypes of nominal predications can be identified based on their semantic functions, following Simpson (1983, p. 203) for Warlpiri: (i) equative predications (one nominal is equated with another) (ii) identification predications (one nominal is identified as another, or a member of a class) (iii) ascriptive predications (ascribe a stative property to something). The lexical semantics of a particular nominal largely determine the types of predicative structures it may appear in. Thus, attributive adjectives appear in ascriptive predications while referential nominals (namely nouns), by contrast, are more frequent in equative or identification predications. Nearly any kind of nominal can be used predicatively in Ngardi (as in Warlpiri, Hale, 1983, p. 35). Free pronouns and demonstratives generally only have equative functions when functioning predicatively (§9.2.1.1). Other nominals (for example locationals, temporals and those depicting states or qualities) often serve as ascriptives. Many nominals can nevertheless be used in each of these predication types.

9.2 Argument structure 

 517

9.2.1.1 Equative predications Equative predications express a relationship of equivalence between two referents, i.e. ‘x = y’. Two examples of this clause type are the proximal demonstratives minya and mu(lu)-, used to identify or introduce an entity as in (1296) and (1297). (1296)

Minya ma ngawirri. this top.aux soak ‘Here/this is the soakwater.’ (MMJ: Manungka_01-035151: 1251276_1253821)

(1297)

Yuwayi, minya=yi ngurra Ngardi. yes this=1sg.o country language_name ‘Yeah, this is my Ngardi country.’ (MAM: TEN1-2016_002-01: 166563_171506

In Warlpiri, a characteristic intonation has been identified as distinguishing equative predications from identifications and ascriptives. In Ngardi this does not appear to be a unique feature of equative clauses. 9.2.1.2 Identification predications Identification predications identify an entity as being a member of a larger set. This type of predication is very similar to an equative predication ‘x = y’ but is distinguished by the fact that ‘x’ forms a subset of ‘y’. For example, in (1298), the subject (ngaju) is identified as a member of both the larger set depicted by the kin term Japaljarri and the additional (secondary predicate) Ngarrkajarra-wardingki (‘someone from the Two Man Dreaming’). Both the argument and predicate NPs in identification predicates are typically referential, that is, headed by referring nominals – typically from the noun subclass. (1298)

Ngaju Japaljarri Ngarrkajarra-wardingki. 1sg ♂subsection man.du-inhabit Ngurrara=yi Sturt Creek. country=1sg.o place_name ‘I am Japaljarri, I am of the Two Man Dreaming. My country is Sturt Creek.’ (PSM: 3111_16410)

9.2.1.3 Ascriptive predications Ascriptive predications ascribe a state or property to a referent. Such predications denote various states of knowledge (1299), emotion (1300), or even an impersonal state ‘it (will be) cold’ as in (1301). (1299)

Pina=wali=rna mangarri-ku kupa-rnu-ku. knowledge=then=1sg.s veg_food-dat cook-inf-dat ‘I know how to cook.’ (MMN: TEN1-2018_012-02: 1202687_1208223)

518 

 9 Grammatical relations and argument structure

(1300)

Nga=rna jaalyjaaly kuli-ku. seq.aux=1sg.s jealous.rdp fight-dat ‘I am hankering for a fight.’ (PSM: TEN1-2018_019-01: 1177118_1181460)

(1301)

Kaji pirriya rangarni, warlu=wali=rna ma-nku. condit cold tomorrow, fire=then=1sg.s get-pot ‘If it is cold tomorrow, I will get firewood.’ (PSM: TEN1-207_026-01: 2026395_2032966)

Predicative NPs which may form ascriptive predications are frequently of the adjective subclass of nominals, as in (1299)–(1301). However, case-marked NPs can also function as the main clausal predicate. Such ascriptive clause types can involve the dative expressing possession (1302) (or even an elided dative NP (1303)), the proprietive expressing ‘having/accompaniment’ (1304) and privative expressing ‘lacking’ (1305) or the locative expressing ‘existence at a location’ (1306). (1302)

Yala ngu=yi ngaju-ku motika. that seq.aux=1sg.o 1sg-dat car ‘That’s my car.’ (PSM: TEN1-2018_030-01: 117603_120157)

(1303)

Yirdi=rla Joe Mahood. name=3sg.obl name ‘His name, (was) Joe Mahood.’ (TEN1-2018_009-01: 11827_14988)

(1304)

Yala ma ngaringka jingka-kurlu. that top.aux woman children-prop ‘This woman has kids.’ (PSM: TEN1-2018_030-01: 176535_178578)

(1305)

Nyuntu ma=n milpa-wangu. 2sg top.aux=2sg.s eye-priv ‘You are blind.’ (MMN: TEN1-2018_030-01: 248702_252185)

(1306)

Yala-piya yala-rla kurrku-kariny. that-sembl that-loc hole-other ‘There is another water basin (soak) like that there.’ (MDN: LC22b: 395183_397947

Indirect objects (§9.1.1.3) are the most common non-subject argument types in ascriptive clauses. Such clause types are referred to as ‘extended’ nominals by Austin (1989). When they are combined with a copula verb they essentially constitute an ‘extended intransitive clause type’ (§9.2.3.4). They can be divided according to the following

9.2 Argument structure 

 519

predicational subsets: a negative predicate [i]; emotion predicates [ii]; knowledge predicates [iii]; and having/lacking predicates [iv]. Lastly, nominal predicates – like verbal predicates – allow for the expression of semantically compatible ‘affected entities/benefactives’ [v] and even dative-case marked non-finite subordinate clauses [vi]. [i] Negative predicate wakuThe negative nominal waku can function predicatively and select a dative casemarked indirect object as its argument. This predicate type is discussed in a wider discussion of negation in §10.8.3. [ii] Emotion predicates Dative case-marked indirect objects can mark the theme of nominal predicates expressing emotion. Examples involving jaaly ‘jealous of/about’ and ngarru ‘happy for’ were provided in (1300) and (1311), respectively. [iii] Knowledge predicates Similar to those described in [2], dative case-marked indirect objects can mark the theme of nominal predicates, expressing mental states such as pinarri ‘knowledgeable in’ (1307) and (1308) and ngurrpa ‘ignorant of’ (1309). (1307)

Yuwayi, pinarri=rna=yanu ngaju holot-ku pamili-ku, yes knowledge=1sg.s=3sg.o 1sg whole_lot-dat family-dat manija-ku wayif-pa-ku. manager-dat wife-ep-dat ‘Yeah, I knew them, the whole family, the manager, the wife.’ (PSM: TEN12018_009-01: 604505_609666)

(1308)

Minya ngantany ngu=rla pinarri Ngardi-ku. this man seq.aux=3sg.obl knowledge language_name-dat ‘This man knows Ngardi.’ (MMN: TEN1-2018_026-01: 116053_125527)

(1309)

Waku, ngurrpa mani-ku. neg ignorant money-dat ‘No, I didn’t know about money.’ (PSM: TEN1-2018_038-01: 477750_482046)

[iv] ‘Having’ and ‘lacking’ predicates ‘Having’ and ‘lacking’ clauses involve nominal predicates marked with the proprietive and privative case suffixes, respectively. They indicate that the property or entity referred to by the predicate is possessed or not possessed by the subject. Examples were provided in (1304) and (1305) for ‘having’ and ‘lacking’ predications, respectively. [v] Nominal predicates allowing for ‘benefactives’ or ‘datives of respect’ Dative case-marked indirect objects functioning as ‘benefactives’ can be predicated of nominal predicates, as in (1310) and (1311).

520 

 9 Grammatical relations and argument structure

(1310)

Kulanganta wuruju wirrpa-ku. cf good many-dat ‘I thought it was sufficient (food) for everyone/the group (but it wasn’t).’ (MMN: TEN1-2017_026-01: 174835_182567)

(1311)

Ngaju ma=rna=rla ngarru papartu-ku. 1sg top.aux=1sg.s=3sg.obl happy eB-dat ‘I am happy for my brother.’ (MMN: TEN1-2018_030-01: 859907_870310)

[vi] Nominal predicates allowing for ‘purpose clauses’ Finally, non-verbal clauses headed by nominals can take both a dative indirect object as well as control a non-finite subordinate clause headed by a nominalised verb. Nominalised verbs are never cross-referenced (although dative-marked referential nominals are) and are treated as adjuncts. (1312)

Jungari lukarrara-ku luwa-rnu-ku. grinding_stone seed-dat shoot-inf-dat ‘A jungari is for grinding seeds./A jungari is for seeds, for grinding.’ (MMN: TEN1-2018_025-01: 1590761_1599539)

Dative-marked non-finite subordinate clauses can also indicate ‘in what respect’ an ascriptive predication holds. This can be observed in (1313), jirniny ‘abs is displeased with respect to dat’ in (1313) and (1314) with the stative preverb inkilji ‘abs is reticent with respect to dat’. (1313)

Jirniny ma=rna yapart-ku. dislike top.aux=1sg.s run-dat ‘I don’t like running.’ (TEN1-2018_019-01: 2676165_2677790)

(1314)

Nyanungu-kujarra=pula inkilji yu-ngu-ku. recog1-dual=3du.s reticent give-inf-dat ‘Those two don’t like sharing.’ (TEN1-2018_019-01: 368164_377499)

9.2.2 Copula constructions with karri- be Copula constructions can be formed with the stative inflecting verb karri- be (see §8.4.1.1). This is in addition to the functions of karri- within intransitive clauses (§9.2.3.2) encoding ‘staying’ or ‘living’ events, as in (1315).

9.2 Argument structure 

(1315)

 521

Karri-nya-ngurra=lu Ngarlikujarra-rla jayanta. be-pst-narr=3pl.s place_name-loc long_time ‘They were staying there at Ngarlikujarra for a long time.’ (TJN: Manungka_01-035151: 80068_98495)

In its copula function, karri- appears in both ascriptive ((1316) and (1317)) and existential ((1318) and (1319)) clauses. (1316)

Ngaju=rna karri-nyanta ngarru Ngardily wangka-nyu-ku. 1sg=1sg.s be-prs happy language-name speak-inf-dat ‘I am happy to be speaking Ngardi.’ (PRN: TG1_0034-01)

(1317)

Nyurnu-wangu=rnalu karri-nya-ngurra. sick-priv=1pl.excl.s be-pst-narr ‘We didn’t used to (get) sick.’ (TJN: Manungka_01-035151)

(1318)

Pamarr=pa=lu yala karri-nyanta. rock=ep=3pl.s that be-prs ‘Those (grinding) stones were (there).’ (MMN: TEN1-2017_004-03: 159404_162812)

(1319)

Wakurra ngapa karri-nyanta kaniny wiyil-ta. neg water be-prs inside well-loc ‘There is no water inside the well.’ (MMN: TEN1-2017_005-02: 59970_68485)

The distinction between the copula clauses exemplified above and ‘verbless clauses’ described in §9.2.1 essentially hinges on the ability to index TAM information within copula clauses. For this reason, copula clauses tend to emphasise the temporal setting of a state of a subject (ascriptive clauses) or the temporal setting of the existence of a subject at a certain location (existential clauses). Verbless clauses, in contrast, express these conditions/states without temporal boundedness.

9.2.3 Verbal predicates Verbal predicates in Ngardi can be categorised according to their valency and argument structure. One major division in verbal predicates in Ngardi is whether they may take an ergative-marked subject or an unmarked (absolutive) subject. All simple verbs generally fall into one of these two classes (a singular exception is described in §9.2.4). This observation corresponds to the widespread claim that verbs in Australian languages are often strictly ‘transitive’ or strictly ‘intransitive’ (Blake, 1987, p. 12; Dixon, 2002, p. 178; Evans, 1989; Hale, 1982; Laughren, 2017e). However, beyond this strict

522 

 9 Grammatical relations and argument structure

division, there is substantial variation in the types of bivalent predicates in Australian languages (Austin, 1982, 1997). This is described in §9.2.4. First we turn to the range of basic clause types in Ngardi. Tables 112–115 provide an overview of the subclasses of Ngardi verbal predicates, along with their subcategorisation frames, cross-referencing properties and up to three indicative examples (simple or complex verbs). Subscripts (e.g. ergx, sbjx) are employed to make explicit the relationship between case marking, cross-referencing and predicate-specific semantic role. The tables are arranged based on valency and the number of arguments a predicate may select (between 0 and 3). Note that the case frames described below bear a striking similarity to those identified in Jaru (see Tsunoda, 1981b, p. 101). Table 112: Avalent predicates. Clause type

Case frames

Cross-referencing

Examples

Impersonal

nil

nil

ngurn-ngurt=ya- ‘(it) thunders’

Table 113: Monovalent predicates. Clause type

Case frames

Cross-referencing

Examples

Intransitive

[absx]

sbjx

karri- ‘x sits’ tardarr=ya- ‘x enters’ yapartu- ‘x runs’

Table 114: Bivalent predicates. Clause type

Case frames

Cross referencing

Example

Quasi-Transitive

[absx-absy]

sbjx437

wangka- ‘x speaks y (language)’ ruju=karri- ‘x performs/plays y’438

ExtendedIntransitive

[absx-daty]

sbjx-obly

tirrka- ‘x is happy/has love for y’ jina=marda- ‘x waits for y’ purda=nya- ‘x listens out for y’

[absx-locy]

sbjx-lcty

wangka- ‘x speaks with y’ yut=karri- ‘x sits with y’ rayin=karri- ‘x is frightened by y’

[absx-ally]

sbjx-lcty

ya- ‘x goes up to y’ yapartu- ‘x runs up to y’

437 Since the contextual semantic role of ‘thing spoken’ is invariably inanimate (and generally singular), there is no overt registration of this argument in the bound pronoun. 438 Where y has the contextual semantic role of ‘instrument played’ or ‘ceremony performed’

9.2 Argument structure 

 523

Table 114 (contiuned) Clause type

Case frames

Cross referencing

Example

Transitive

[ergx-absy]

sbjx-objy

nga- ‘x eats y’ nya- ‘x sees y’

Conative

[ergx-daty]

sbjx-obly

warri- ‘x searches for y’ nya- ‘x looks for y’

[ergx-locy]

sbjx-lcty

jurlurr=ma- ‘x gives a fright to y’

Table 115: Trivalent predicates. Predicate type

Case frames

Cross-referencing

Examples

Semi-Ditransitive

[absx-absy-datz]

sbjx-oblz

wangka- ‘x speaks y to z’

[absx-absy-locz]

sbjx-lctz

wangka- ‘x speaks y to z’

[absx-absy-allz]

sbjx-lctz

ngarri- ‘x tells y to z’

Transitive Complement

[ergx-absy-absz]

sbjx-obj (compl)z

ngarri- ‘x calls y, z’

Ditransitive

[ergx-absy-absz]

sbjx-oblz

yu- ‘x gives z, y’

[ergx-absy-datz]

sbjx-(objy-oblz)

yupa- ‘x sends y to z’ jinjin=yu- ‘x obliges y to z’

[ergx-absy-locz]

sbjx-(objy-lctz)

yupa‘x sends y with z’ purda=nya‘x learns y from z’ jarri- ‘x takes y from z’

[ergx-absy-allz]

sbjx-lctz

ngarri- ‘x tells y to z’

In the tables above, a number of predicates are listed in more than a single clause type. The possibility for verbs to exhibit case and valency alternations will be discussed in §9.2.4. 9.2.3.1 Impersonal clauses Impersonal (avalent) clauses lack overt referential arguments. Only a small number of predicates have been identified in this clause type and all express meteorological or temporal conditions (1320)–(1323). These clauses are distinguished from intransitive clauses with unrealised third person singular arguments by the fact that impersonal clauses cannot take overt nominals as subjects. These clause types overwhelmingly involve the be verb karri- or the fall verb wanti- functioning as basic change of state predicates.

524 

 9 Grammatical relations and argument structure

(1320)

Ngurn-ngurt-karra ya-nanta. rumble-rdp-ssub go-prs ‘It is thundering.’ (PSM: TEN1-2018_030-01: 1332055_1334051)

(1321)

Rangkarr=karri-nyanta=wali. sunrise=be-prs=then ‘It is getting light.’ (PSM: TEN1-2018_030-01: 1493619_1494919)

(1322)

Kuru=wanti-nya. dark=fall-pst ‘It became overcast/fell dark.’ (TJN: Cataldi, 2011, p. 148)

(1323)

Kily=wanti-nya. cold=fall-pst ‘It is getting cold.’ (MMN: TEN1-2017_023-03: 2004519_2014229)

It is worth considering that the above examples may not actually represent bona fide impersonal clauses.439 Other expressions involving temperature (e.g. pirriya=karri‘to be cold’) have been identified with overt referential subjects (e.g. ngaju ‘I’) and when used to refer to climatic conditions they perhaps could be understood to involve an unexpressed third singular NP representing a larger environ (e.g. ngurra ‘the country’). Hale (1982, p. 231) has made a similar suggestion for analogous clauses in Warlpiri that refer to climatic conditions – arguing that they can be understood as implicationally involving an unexpressed argument, for example, ngurra ‘camp, home country’ or nguru ‘sky’. Predicates which lack overt subjects can still take non-subject arguments, for example in the role of affected datives, as in (1324). (1324)

Pirranginti=karri-nya=yi. afternoon=be-pst=1sg.o ‘The day was getting away from me (lit. ‘it became afternoon on me’).’ (TJL: TT76_2201: 1980359_1981938)

9.2.3.2 Intransitive Intransitive clauses comprise an absolutive subject NP and a subject bound pronoun (ø in the 3sg), as in (1325).

439 Another putative impersonal predicate walmak=ma- ‘(it is) lightning’ can also take ngapa ‘water’ as a subject or else be used in a semantically distinct monovalent predicate type: walmak=ma-nanta yala motika ‘that car is emitting a flickering light’.

9.2 Argument structure 

(1325)

 525

Jangilany pali-nya. fire die-pst ‘The fire went out.’ (PRN: TG1_034-02)

All verbs of locomotion (§8.4.2) can form intransitive clauses, as well as certain verbs of location/change of locative relation – for example, the stative verbs karri- be and wanti- fall. Intransitive clauses involving motion and activity events are frequently elaborated with adverbial preverbs and/or locational adjuncts, as in (1326)–(1328). (1326)

Hospital-ta=wali ngati ngaju-ku=yi karri-nyani-nyirra. hospital-loc=then M 1sg-dat=1sg.o be-ipfv.pst-narr ‘My mother was staying at the hospital.’ (NGO: TEN1-2017_011-01: 282909_285661)

(1327)

Jingka purnu-jangka wanti-nya. child tree-ela fall-pst ‘The child fell out of the tree.’ (BSE: TEN1-2017_017-01: 454510_457662)

(1328)

Wirrirl=ya-ni=rli Kururrungku-kurra. back=go-pst=1du.incl.s place_name-all ‘We two went back to Billiluna Station.’ (PSM: TEN1-2016_005-03: 714580_716896)

Some intransitive complex predicates are anomalous in being formed from inflecting verbs that are transitive in their simple (non-complex verb) usage. These include certain CVCs formed from the (otherwise transitive) pu- hit verb: wirrirl=pu- ‘travel around’, jungkuj=pu- ‘jump about’, pitpirriya=pu- ‘dance’, and lawu-lawu=pu- ‘float’ (see §9.2.4.4 for further discussion). An example involving jut=pu- ‘set off’ is provided in (1329). (1329)

Wukayi-wukayi=rna jut=pi-nya mulu Tutpungu-kurra. tired-rdp=1sg.s set_off=hit-pst this place_name-all ‘Very tired, I set out towards Tutpungu.’ (PPN: Manungka_01-035167)

9.2.3.3 Quasi-transitive Quasi-transitive clauses are similar to transitive clauses in selecting an absolutive direct object – however, they are distinguished by selecting absolutive subjects. Objects of quasi-transitives are semantically restricted and are low in individuation (cf. Hopper and Thompson, 1980), meaning the object selected by the predicate is not semantically distinct from the situation described by the predicate (e.g. ‘to sing a song’). Objects which have low individuation are often also considered to be

526 

 9 Grammatical relations and argument structure

lower on scales of transitivity than canonical transitives (Austin, 1982, p. 46). Many clauses of this type have been labelled ‘cognate object constructions’ (Austin, 1982; Simpson, 1991, p. 343) and are found in all of the Ngumpin-Yapa languages as well as in numerous other Australian languages (e.g. Diyari, Bayungu, Guugu Yimidhirr, and Yidiny).440 I use the term ‘quasi-transitive’ since the object need not be ‘cognate’ with the verb itself (although see (1330)).441 Only a restricted set of predicates is found in quasi-transitive clauses. Many happen to be predicates involving performance or speech: ruju=karri- ‘perform/play’ (1330), yunpa- ‘sing’, and wangka- ‘speak’ (1331). Predicates involving performance or speech typically occur with the object juju ‘ceremony’ or specific names of songs and ceremonies, while the predicate wangka- typically takes the object NP jaru ‘story, language’ or else the names of languages. (1330)

Ruju=rnalu ruju=karri-nyanta ngantany.442 play=1pl.excl.s play=be-prs man ‘All of us men were performing corroboree.’ (PJA: TT76_2601: 501521_505690)

(1331)

Ngantany wirrpa=lu wangka-nyanta Ngardily. man many=3pl.s speak-prs language_name ‘Lots of Aboriginal people speak Ngardi.’ (PRN: TG1_0034_01)

An additional example involves lu- ‘cry’, which may take the cognate object jirlmirl ‘tears’. Note that it is possible to express jirlmirl as either a cognate object (1332) or a proprietive case-marked adjunct (1333). (1332)

Jirlmirl lu-nganta=n. tears cry-prs=2sg.s ‘You are crying tears.’ (PSM: TEN1-2018_001-01: 1662713_1664356)

(1333)

Jirlmirl-kurlu lu-nganta. tears-prop cry-prs ‘He is crying (with) tears.’ (PSM: TEN1-2018_001-01: 1665250_1668054)

Some predicates can take objects with seemingly low individuation but nevertheless never take an absolutive subject due to the higher ‘inherent transitivity’ of the verbal 440 See Senge (2015, pp. 575–576) and Meakins and Nordlinger (2014, 368ff.) for descriptions of cognate objects in the related languages Wanyjirra and Bilinarra, respectively. 441 See a similar usage of the term ‘quasi transitive’ by Gaby (2017, p. 376) for Kuuk Thaayorre. 442 The action nominal ruju itself is polysemous in having both dynamic ‘perform/play’ and referential ‘(a) play, performance’ senses.

9.2 Argument structure 

 527

predicate. Thus in (1334), pangi- dig takes a semantically related object kurrku ‘a hole’, but the subcategorised subject is strictly ergative-marked. (1334)

Ngalipa-rlu=rlipa pangi-rnanta kurrku ngu=rlipa 1pl.incl-erg=1pl.incl.s dig-prs hole seq.aux=1pl.incl.s ma-nku karnti. get-pot bush_potato ‘We were all digging holes, so that we could get bush potato.’ (PSM: TEN12017-023-02: 1142166_1155600)

9.2.3.4 Extended intransitive Extended intransitive predicates select a subject in the absolutive and one additional argument in the dative, locative or allative cases. The term ‘extended intransitives’ was first coined by Austin (1988) to refer to intransitive predicates which also select a dative argument (see also Plank, 1995, p. 366). I follow Wilkins’ (1989) expanded use of this term to refer to any intransitive clause which is associated with additional arguments. Three different subtypes of extended intransitive clauses can be identified based on their case frames and cross-referencing properties as shown in Table 116. Table 116: Subtypes of extended intransitives. Subtype

Case Frame

Cross-referencing

I II III

abs-dat abs-loc abs-all

s-obl s-lct s-lct

[i] abs-dat Extended intransitives that select a dative argument include i) predicates of speech/ sound emission (e.g. wangka- speak, ngarri- tell, ngarlarri- laugh; some of these alternatively take subtype II), ii) certain predicates of knowledge (these may appear without their stative verbs as ‘nominal predicates’, §9.2.1), iii) various predicates of emotion (e.g. tirrka- love), and iv) numerous complex verbs formed by the stative be verb (e.g. maki=karri- ‘joyful’, ngurrku=karri- ‘very happy’, luyurr=karri- ‘sad’).443 The dative arguments in these clauses are all instances of ‘indirect objects’ (§9.1.1.3). Examples are (1135)–(1137).

443 Note that similar emotion predicates expressing fear (e.g. rayin=karri-) and shame (e.g. minyirri=karri-) are extended intransitives of subtype II, taking abs-loc case frames rather than abs-dat.

528 

 9 Grammatical relations and argument structure

(1335) Wangka-nya=yi kapulu ngaju-ku kuja ya-nani-nyirra karla-ngurlu.444 speak-pst=1sg.o eZ 1sg-dat sub go-ipfv.pst-narr west-abl ‘My sister said to me that she had come from the west.’ (PRN: FT1_b: 535088_538345) (1336)

Maki=karri-nya=pulany yalu-ku ngantany-kujarra-ku, parntany. happy=be-pst=3du.o that-dat man-dual-dat old_woman ‘The old woman was happy for those two men.’ (TJN: LC49a: 162506_167579)

(1337)

Tirrka-rnanta=nyurra Nangala-warnu-ku Napaljarri. love-prs=2pl.o ♀subsection-assoc2-dat ♀subsection ‘Napaljarri loves/feels good for all you Nangalas.’ (MMN: TEN1-2017_002-02: 169070_175975)

(1338)

Nyanungu=pula=rla ngarlarri-nyani-nyirra. recog1=3du.s=3sg.obl laugh-ipfv.pst-narr ‘Those two were laughing at her.’ (TJN: LC49a: 198404_200905)

For many of these predicate types, the extended intransitive (abs-dat) case frame is an alternative to a simple intransitive (abs) case frame. This is true of verbs of speech as well as the complex verbs formed via the stative verb karri- be. However, there is at least one example of a verb for which the abs-dat frame is the only one available: tirrka- love, as in (1339).445 (1339)

Tirrka-rnanta=rla ita-ku jingka-ku. love-prs=3sg.obl small-dat child-dat ‘She is happy for/loving that little child.’ (TEN1-2017_002-02: 254802_260067)

[ii] abs-loc Extended intransitives that select a locative argument are found with three main types of verbal predicates.446 Stative predicates (karri- be, yut=karri ‘sit’, etc.) can select 444 Note that the status of the dative NP ngaju-ku here is ambiguous – it could either be the indirect object of the predicate wangka- speak or else the possessor of the subject NP kapulu ‘elder sister’. 445 Interestingly in Warlpiri, the same is observed: yulka-mi ‘love’, rdanpa-rni ‘accompany’, and yura-ka-nyi ‘sneak up on’ all obligatorily take an abs-dat case frame (Simpson, 1991, p. 317). With respect to obligatory datives in extended intransitives in Warlpiri, Simpson (1991, pp. 318–319) shows that these arguments can control subordinated -kurra clauses (an object complementiser) and that these are further marked dative: i.e. [nominalised verb]-kurra-ku. Unfortunately, there is insufficient data to verify this in Ngardi. 446 Hale (1982, p. 273) notes that certain verbs in Warlpiri also ‘select’ the locative case as part of their basic argument structure. However, unlike in Ngardi, locative NPs in Warlpiri are never crossreferenced with a bound pronoun of any kind.

9.2 Argument structure 

 529

a locative argument marking an inherent (animate) location (1340). Predicates of speech (wangka-, ngarri-, etc.) can select a locative argument marking the (animate) addressee/recipient of the speech event (1341). Motion predicates can register animate sources/destinations of motion as locative arguments, as in (1342). Finally, predicates of fear (rayin=karri-, (1343)) and shame (minyirri=karri, (1344)) can select a locative argument marking the (animate) source of the emotion (although the evitative case can also be used in this context without cross-referencing). The locative NPs in these clauses all have the status of ‘subjuncts’ (§9.1.1.4). (1340) Yala=yi=lu=rla yut=karri-nyanta ngaju-ngka. that=1sg[lct]=3pl.s=lct sit=be-prs 1sg-loc ‘They are sitting with me.’ (MMN: TEN1-2017_024-01: 964155_971197) (1341) Nyampa-jangka=yi=n=ku=rla ngirrka-ngirrka=wangka-nyanta? what-ela=1sg[lct]=2sg.s=ep=lct shout-rdp=speak-prs ‘Why are you shouting at me?’ (PSM: TEN1-2018_030-01: 2156114_2162337) (1342) Ya-nku=rna=rlanyanta. go-pot=1sg.s=3sg.lct ‘I will go away from her.’ (MMN: TEN1-2017_026-01: 1132397_1135021) (1343) Rayin=karri-nya=rlanyanta yalu-ngka, kurrkurr-ta. afraid=be-pst=3sg.lct that-loc owl-loc ‘He is afraid of that one, the owl.’ (MMN: TEN1-2017_016-02: 212312_217107) (1344) Ngantany=pa=rlanyanta minyirri=karri-nyanta ngaringka-rla. man=ep=3sg.lct shame=be-prs woman-loc ‘The man feels shame on account of the woman.’ (PSM: TEN1-2019_005-01: 1553082_1560141) [iii] abs-all The final type of extended intransitive involves a subjunct argument in the allative case.447 This subtype is restricted to verbs of locomotion (e.g. ya- go (1345)) and of induced change of locative relation (e.g. ka- carry (1346)). The allative NPs in these clauses mark an animate terminus/goal of motion and have the status of ‘subjuncts’ (§9.1.1.4).

447 Wilkins refers to similar arguments in analogous clause structures in Arrernte as ‘inherently directed’ (Wilkins, 1989, p. 221), while Osgarby (2018, pp. 155–157) employs the terms ‘inherent location’ and ‘inherent destination’ for similar clausal arguments in Mudburra.

530 

 9 Grammatical relations and argument structure

(1345) Parda=yi paja-rnngi kunyarr-tu kaji=rna=rlanyanta warririku ya-nngi. dub=1sg.o bite-irr dog-erg condit=1sg.s=3sg.lct close go-irr ‘The dog might have bitten me, had I gone closer to it.’ (PSM: TEN1-2018_03401: 3119037_3123961) (1346)

Ka-nganta=rna=yanungkurla yalu-kurra ngantany-kurra, wirrpa-kurra. carry-prs=1sg.s=3pl.lct that-all man-all many-all ‘I will take (that group) to them.’ (PSM: TEN1-2017_024-01: 2304517_2307948)

Note that, in (1346), the direct object of the predicate ka- carry is not registered overtly anywhere in the clause – neither as an overt NP, nor in the pronominal complex – where it has been displaced by a subjunct argument (the allative marked ‘goal’). 9.2.3.5 Transitive The basic transitive clause involves an ergative subject and an absolutive object. These arguments are cross-referenced by the subject and object bound pronouns respectively. Transitive predicates are numerous. Some examples include paja- bite (1347), ka- carry (1348), and ma- get (1349). (1347) Minya ngantany paja-rni kunyarr-u. this man bite-pst dog-erg ‘The dog bit this man.’ (BSE: TEN1-2017_016-03: 603727_611663) (1348) Timana=lu ka-nya kakana. horse=3pl.s carry-pst east ‘They took the horses east.’ (KAP: LC43b: 106124_108150) (1349) Warlu=rna ma-nani pirranginti. fire=1sg.s get-ipfv.pst yesterday ‘I was collecting wood yesterday.’ (PSM: TEN1-2017_026-01: 1613877_1620721) Simple transitives can be further modified by adjuncts such as the locational or temporal modifiers in (1348) and (1349), respectively. Indirect objects arguments in the dative case can also be added to otherwise simple transitives. These may be overt NPs with cross-referencing bound pronouns (1350) or bound pronouns alone with no overt NPs (1351). If the dative NP is inanimate, it is not cross-referenced at all (1352). (1350) Nyurrurla-rlu, kupa-ka=lu=rla yalu-ku. 2pl-erg cook-imp=pl.s=3sg.obl that-dat ‘You lot, cook it for him.’ (TEN1-2017_030-01: 658290_661527)

9.2 Argument structure 

 531

(1351)

Karli ngantany-ju paja-rnanta=nyanu yalu-rlu. boomerang man-erg bite-prs=refl that-erg ‘That man is cutting out a boomerang for himself.’ (TJL: TT76_2501: 278825_282796)

(1352)

Ngalipa-rlu=rlipa paja-rnanta warlu-ku. 1pl.incl-erg=1pl.incl.s cut-prs fire-dat ‘We were all cutting wood for the fire.’ (MMN: TEN1-2017_023-01: 1094089_ 1106523)

9.2.3.6 Conative Conatives are transitive clauses whose ‘objects’ are dative rather than absolutive case-marked (Hale, 1982; Laughren, 1988; Simpson, 1991).448 Conatives are also very similar to ‘extended intransitives’ but are distinguished by the fact that their subjects are ergative rather than absolutive. Examples are (1353) and (1354). Note that in the latter, the ergative subject is only revealed by the secondary predication yut-warnu-rlu ‘while sitting down’ agreeing with the elided subject (ngaju-ngku 1sg-erg). The dative NPs in these clauses have the status of indirect objects (§9.1.1.3). (1353) Ngantany-ju=rla kirrit=pu-ngani ngapa-ku. man-erg=3sg.obl search=hit-ipfv.pst water-dat ‘A man looks around for water.’ (MMN: TEN1-2018_026-01: 446299_452057) (1354) Jina=marda-rnanta=rna=ngku yut-warnu-rlu, (nyuntu-ku). feet=have-prs=1sg.s=2sg.o sit-assoc2-erg 2sg-dat ‘I am waiting for you while (I am) sitting down.’ (MMN: TEN1-2018_027-01: 2206488_2211410) Predicates which select an erg-dat case frame can be divided into two categories: those for which this is the only available predicate argument structure,449 for example warri- search, jina=marda- ‘wait for’, and kirrit=pu- ‘search, look around for’; and second, those for which (erg-dat) is an alternative case frame to the standard transitive (erg-abs) – an alternation pattern referred to here as a ‘conative alternation’. An example is the verb nya- see with the case frames: erg sees abs (1355) and erg looks for dat (1356).

448 For a discussion of the term ‘conative’ see Vincent (2013). The use of the term ‘conative’ as used here to describe a dative case alternation was originally taken from Athapaskan linguistics by Hale (1982). For summaries of related clausal alternations in English see Levin (1993). 449 These are referred to as ‘verbs with obligatory dative arguments’ in Simpson’s (1991, p. 326) description of Warlpiri.

532 

 9 Grammatical relations and argument structure

(1355) Ngantany-ju nya-nganta jaji. man-erg see-prs kangaroo ‘A man sees/watches a kangaroo.’ (MMN: TEN1-2018_026-01: 728003_732926) (1356) Ngantany-ju=rla nya-nganta jaji-ku. man-erg=3sg.obl see-prs kangaroo-dat ‘A man looks for a kangaroo.’ (MMN: TEN1-2018_026-01: 705898_710681) The contrast between the status of the participant with the role of theme in (1355) and (1356) is realised both in the case system (absolutive vs dative) and in the cross-referencing system (object – unmarked in third singular – vs oblique). The semantic effect of the erg-dat case frame typically involves positioning the theme as being in some sense less affected by the event expressed by the main predicate. The range of predicate types which exhibit conative alternations in Ngardi is similar to many other languages in Australia. A conative alternation for nya- see in particular is common to all Ngumpin-Yapa languages (for examples see Warlpiri: Hale, 1982, pp. 248–250; Simpson, 1991, p. 328; and Jaru: Tsunoda, 1981b, p. 149). However, in Warlpiri, conative alternations extend even to verbs of contact/force (§8.4.3). This is not the case for all verbs of contact/force in Ngardi. For example, while luwa- shoot as a simple verb can appear with either erg-abs or erg-dat case frame, the verb la- pierce cannot. Instead, in order to construct a conative clause type with la-, the adverbial preverb jakarr must co-occur with la- in a complex verb construction (see discussion in §9.2.4.4.2). 9.2.3.7 Semi-ditransitive The semi-ditransitive clause involves an absolutive subject, an absolutive (cognate) object, and either a dative indirect object (1357), a locative subjunct (1358), (1359) or an allative subjunct (1360). This clause type has only been found with the verbs wangka- speak and ngarri- tell. (1357) Kuja=rna=ngku wangka-nyanta jaru. thus=1sg.s=2sg.o speak-prs story ‘Like this, I am telling a story to you.’ (PSM: TEN1-2017_009-01: 719957_722755) (1358) Nyuntu=yi=n=ku=rla wangka-ju ngaju-ngka Ngardi? 2sg=1sg[lct]=2sg.s=ep=lct speak-pot 1sg-loc language_name ‘Will you speak Ngardi with me?’ (PSM: TEN1-2017_026-02: 1021616_1024518)

9.2 Argument structure 

 533

(1359) Ngaju ma=rna=yanurla ngantany-ja wirrpa-ngka jaru 1sg top.aux=1sg.s=3pl.lct man-loc many-loc story ngarri-rnani. tell-ipfv.pst ‘I was telling a story to the many men.’ (PSM: TEN1-2018_030-01: 1760805_ 1765309) (1360) Ngaju ma=rna=yanurla ngantany-kurra ngarri-rnani 1sg top.aux =1sg.s=3pl.lct man-all tell-ipfv.pst ‘I was telling a story to the men.’ (PSM: TEN1-2018_030-01: 1804597_1808452)

jaru. story

In some special cases, the absolutive object is not a cognate object but rather quoted or reported speech functioning as the theme of a speech event, as in (1361). Note that as an exclamative, this ‘reported speech’ lacks any formal subordination typically associated with reported speech (e.g. a kuja complementiser). (1361)

‘“Waraa”=nyanu ngarri-rni-nyirra.’ wow=refl tell-pst-narr ‘“Oh gosh!” he said to himself.’ (TJN: Manungka_01-035159: 797324_800139)

9.2.3.8 Transitive complement Transitive complement clauses involve an ergative subject, an absolutive object and an additional complement that agrees in case with the absolutive object. The most common predicate which exhibits this clause structure is ngarri- ‘erg calls abs abs(complement)’ but also includes marda- ‘erg keeps abs abs(complement)’. In these constructions, the direct object may variously be: i) realised in the bound pronoun (1362), ii) realised as an NP but not cross-referenced in the bound pronoun (1363), or iii) lack any overt expression in the clause at all (1364). Objects and their complements are bolded in the following examples. (1362) Jakuru=yi=lu ngarri-rnanta. jakuru=1sg.o=3pl.s tell-prs ‘They call me jakuru.’ (PSM: TEN1-2018_004-01: 1134877_1136514) (1363) Ngunyju=rlangu=lu ngarri-rnani iki. tobacco=also=3pl.s tell-ipfv.pst shavings ‘They would also call loose tobacco iki (little bits/chips of wood).’ (MMN: TEN1-2018_004-01: 254265_256633)

534 

 9 Grammatical relations and argument structure

(1364) Kardiya-rlu ngarri-rnanta ankle, nanimpa=rnalu ngarri-rnanta white_man-erg tell-prs ankle 1pl.excl=1pl.excl.s tell-prs luju.450 ankle ‘Whitefellas call (this one) ankle, we call (it), luju.’ (PSM: TEN1-2019_003-010: 780525_791299) Transitive complement clauses bear some similarity to ditransitive clauses that also involve an erg-abs-abs case frame – so called ‘double object’ constructions. The two are differentiated in the types of arguments they represent, however. Transitive complement clauses involve a single absolutive object that is cross-referenced by an object clitic (i.e. a canonical object) and a complement that agrees in case but is not cross-referenced. Double-object constructions, however, involve a ‘primary object’ that is never cross-referenced by an object clitic and a ‘secondary object’ argument that is cross-referenced by an oblique clitic but is anomalously realised as an absolutive case-marked NP (i.e. a non-canonical indirect object) (see §9.1.1.6 for discussion). 9.2.3.9 Ditransitive Ditransitives include a range of clause types as shown in Table 117. Many simple and complex verbs expressing transfer of entities, things or knowledge are ditransitive predicates and occur with more than one of the following subtypes. Only a single verbal predicate (yu- give) has been identified exhibiting the argument structure of Subtype I. Subtype II is the most common and most predicates that can occur as Subtypes I, III, IV can also occur as Subtype II. Some of the common alternations in case frames for trivalent predicates discussed below are summarised in §9.2.4.3. Table 117: Subtypes of Ditransitive Clauses. Subtype

Case Frame

Cross-referencing

Example

I II III IV

[ergx-absy-abs] [ergx-absy-datz] [ergx-absy-locz] [ergx-absy-allz]

sbjx-obly sbjx-(objy-oblz) sbjx-(objy-lctz) sbjx-(objy-lctz)

(1365), (1366) (1368)–(1370) (1372)–(1376) (1377)

[i] erg-abs-abs Subtype I corresponds to clauses which are typically referred to as ‘double object’ constructions. They consist of an ergative subject and two absolutive arguments: a ‘secondary object’ and a ‘primary object’, expressing a theme and recipient respectively (Dryer 1986). The secondary object is distinct from all other object types in that it is never 450 In these two clauses the objects are unexpressed third singular arguments.

9.2 Argument structure 

 535

cross-referenced by a bound pronoun. The primary object is also an anomalous object type in that it is cross-referenced with an oblique rather than an object bound pronoun in the third singular (i.e. =rla rather than =∅). This is only apparent in the third singular as in (1365) and (1366). Recall that for all other person/number categories, the object bound pronoun is not distinguished from an ‘oblique’ form, see (1367). In these examples the subscripts po and so mark ‘primary object’and ‘secondary object’ respectively. (1365)

mangarriso yi-nya ngati-ngku. Itapo =rlapo small=3sg.obl veg_food give-pst M-erg ‘Then the mother gave the kid the food.’ (JLI: TT76_2201: 471018_473113)

(1366) Kala yu-ngku nga=rna=rlapo minyaso [ngaju-ku ngajayi]po. contr give-pot seq.aux=1sg.s=3sg.obl this 1sg-dat yB ‘But I will give my brother this.’ (TJN: Manungka_01-035160: 470937_475876) (1367)

Yi-nya=rli=yanupo kunyarrso yalu-kujarrapo.451 give-pst=1du.incl.s=3pl.o dog that-dual ‘You and I gave the dog to those two.’ (MMN: TEN1-2017_024-02: 749209_760395)

The fact that what I term a ‘primary object’ displays different cross-referencing to a monotransitive ‘direct object’ means my use of the term does not fully align with Dryer’s (1989) use of the term ‘primary object’ which he deployed to describe a type of object relation which was marked identically in monotransitive and ditransitive contexts. However, Dryer did not address languages which deploy the precise mixture of head-marking (cross-referencing bound pronouns) and dependent-marking (case suffixes) as found in Ngardi. Since the case-marking properties of what I term ‘primary objects’ are entirely consistent with Dryer’s ‘primary objects’, and the fact that it is the ‘primary object’ which is cross-referenced (albeit with an unexpected form) and not the ‘secondary object’, I retain the use of the term but I draw attention to their unique cross-referencing properties.452

451 The object bound pronoun in this example has undergone dual number replacement (see §6.4.4.1). 452 Note that certain verbs of transfer have been analysed as complex verbs, e.g. pina=yuknowledge=give. Were such predicates to be analysed as simple verbs (i.e. yu- give), such clauses would necessarily be analysed as secondary object constructions. For example, the following clause would be translated ‘A man gives the boy knowledge for singeing (a kangaroo)’ under an analysis where pina is a nominal that does not form a complex verb with yu- give. Ngantany-ju=rla pina=yu-nganta jingka wawu=pu-ngu-ku. man-erg=3sg.obl knowledge=give-prs child-dat singe=hit-inf-dat A man teaches a kid how to singe (a kangaroo). (IPN: TEN1-2018_025-01: 1766927_1781648)

536 

 9 Grammatical relations and argument structure

[ii] erg-abs-dat Ditransitive Subtype II is what is typically considered a canonical ditransitive – this clause type comprises an ergative subject, an absolutive object and a dative indirect object. This clause type is found with the common ditransitive verbs yu- give and yupa- send. These verbal predicates depict events in which an agent causes some entity to move to/from a goal/source. In these clause types, either the absolutive object (1368) or the dative indirect object (1369) can be registered in the bound pronoun depending on person/number and animacy (see §6.4.4). Formally, there is no distinction between the argument registration of canonical three-part predicates (e.g. yugive) and canonical transitive predicates that add an additional indirect object, for example, kiji-, erg throws abs to dat; ka-, erg brings abs to/for dat (1369); ma-, erg gets abs for dat (1370). (1368)

Yu-nganta=rna=pulany jingka-kujarra mu-ku ngaringka-ku. give-prs=1sg.s=3du.o child-dual this-dat woman-dat ‘I am giving two kids to this woman.’ (MMN: TEN1-2017_002-05: 2307586_2316897)

(1369)

Minya=rna=ngku kuyi ka-ngani nyuntu-ku nga-rnu-ku. this=1sg.s=2sg.o meat carry-ipfv.pst 2sg-dat eat-inf-dat ‘I was bringing this meat for to you to eat.’ (MMN: TEN1-2018_030-01: 1054437_1074034)

(1370)

Ngantany-ju ma-nta=lu=nyanu mangarri. man-erg get-imp=pl.s=refl veg_food ‘You men, get some food for yourselves.’ (MMN: TEN1-2017_022-01: 77857_280800)

A variant of the ditransitive subtype II is the possibility for an object complement to appear alongside an object with which it agrees. Thus, in (1371), marda- have has the predicate argument structure: erg ‘keeps’ abs abs (complement) for dat. Alternatively, jardi could be analysed as a type of secondary predicate controlled by the object (see §9.3). (1371)

Jardi=rna=nyanu marda-rnanta ngaju-ku. safe=1sg.s=refl have-prs 1sg-dat ‘I am keeping it safe for myself.’ (MMN: TEN1-2018_019-01: 2248221_2253283)

[iii] erg-abs-loc Ditransitive type III consists of an ergative subject, an absolutive object and a locative subjunct. The locative NP in these constructions generally marks the animate recipient

9.2 Argument structure 

 537

in the transfer of an entity, an object or a knowledge state. As with the other ditransitive clause types, this subtype can be used with verbs of literal or cognitive transfer: e.g. jirri=yirra- ‘show’: erg shows abs to loc (1372), jarri- take: erg takes abs from loc (1373), ma- get: erg gets abs from/out of loc (1374), (1375). The locative subjunct NP is typically registered in the bound pronoun at the expense of the direct object. (1372)

Ngu=rna=rlanyanta jirri=yirra-rni ngaringka-rla. seq.aux=1sg.s=3sg.lct show=place-pst woman-loc ‘I showed it to the woman.’ (MMN: TEN1-2017_027-01: 517265_526634)

(1373) Yalu-ngka=rna=rlanyanta ngantany-ja jarri-nya kipara-kujarra kuriny. that-loc=1sg.s=3sg.lct man-loc take-pst turkey-dual two ‘I took the two bush turkeys from that man.’ (MMN: TEN1-2017_027-01: 430392_437290 (1374) Ma-ni=n=ku=rlanyanta? get-pst=2sg.s=ep=3sg.lct ‘Will you get it from him?’ (LC56b) (1375)

Ma-nta=yirla minya nyampa. get-imp=1sg.lct this something ‘Get this thing (a thorn) out of me.’ (PSM: TEN1-2019_003-10: 275723_284175)

A special case of Type III ditransitive clauses involves the use of a locative subjunct to mark an entity which accompanies the theme on a path of motion rather than registering the source or goal of the theme. This is found as an alternate case frame for the otherwise prototypical Subtype II ditransitive, e.g. yupa- send: erg sends abs with loc, as shown in (1376). (1376)

Yupa-rni=yirla ngaju-ngka. send-pst=1sg.lct 1sg-loc ‘He sent it with me.’ (MMN: TEN1-2017_020-01: 812702_816707)

All predicates which appear in this kind of ditransitive clause are also attested in simple transitive clauses lacking the source, goal or accompanying entity of the transfer event. [iv] erg-abs-all Subtype IV is nearly identical to Subtype III in that it involves an ergative subject, an absolutive object and a subjunct. In this type, however, the subjunct argument encoding the recipient is marked with the allative, rather than the locative case. This ditransitive clause subtype is rare in the corpus. The following example (1377) is notable

538 

 9 Grammatical relations and argument structure

for the fact that the expected ergative marking on the subject free pronoun (ngaju) is absent (see §5.4.1).453 (1377)

Ngaju ma=rna=yanurla ngantany-kurra ngarri-rnani jaru. 1sg top.aux=1sg.s=3pl.lct man-all tell-ipfv.pst story ‘I was telling a story to the men.’ (PSM: TEN1-2018_030-01: 1804597_1808452)

A final variant of ditransitive clauses involves certain predicates of speech, wherein the theme of the speech event is realised with a proprietive marked NP as in (1378) (rather than an absolutive). Proprietive NPs are never registered in the auxiliary, however, and have the status of adjuncts. (1378)

Kart=pu-nganta=rna=ngku kuyi-kurlu. narrate=hit-prs=1sg.s=2sg.o meat-prop ‘I am telling you about the meat.’ (MMN: TEN1-2017_023-02: 1356716_1360719)

9.2.4 Case and valency alternations A striking feature of Ngardi clausal syntax is a number of valency and case alternations available to a number of different verbal predicates – both simple and complex. The following sections first deal with valency (§9.2.4.1) and case (§§9.2.4.2–9.2.4.3) alternations specific to a single predicate (be it a simple or complex verb), while §9.2.4.4 examines how the formation of complex verbs can involve modification of expected argument structures in comparison with simple verb counterparts. This section concludes with a discussion of some of the interesting properties of the special verbal predicate ngarri- tell (§9.2.4.5). 9.2.4.1 Valency alternations Valency alternations are found for a number of simple and complex verbs. Only a single verb (wangka- speak) exhibits all three main valency types (monovalent, bivalent and trivalent). All other predicates that exhibit valency alternations do so between monovalent and bivalent clauses, or between bivalent and trivalent clauses. Table 118 provides some major examples of predicates exhibiting valency alternations. Descriptions of the clause types involved were provided in the preceding section (§9.2.3). Examples (1379)–(1384) illustrate how a single predicate (wangka- speak) can be found with as many as six different case frames. 453 The verbal predicate ngarri- tell typically selects ergative subjects. This is a case of optional ergative marking on free pronouns.

9.2 Argument structure 

 539

Table 118: Select verbs with valency alternations. Monovalent

Bivalent

Trivalent

wangkaspeak

abs speaks (1379)

abs speaks abs (1381) abs speaks with loc (1382) abs says to dat (1380)

abs speaks abs to dat (1383) abs speaks abs with loc (1384)

jankaburn

abs burns (1391)

erg burns abs (1392)

ruju=karriplay=be

abs plays (1899)

abs performs abs (1330)

yago

abs goes (1328)

abs goes for dat (1666) abs goes towards loc (1345) abs goes towards all (1280)

yupasend

erg gives abs (1394)

erg sends abs to dat (762) erg sends abs with loc (1396)

maget

erg gets abs (1301)

erg gets abs for dat (1370) erg gets abs from loc (1374)

pina=kaknow=carry

erg shows abs (–)

erg shows abs to dat (1188) erg teaches abs to loc (790)

[i] wangka- (abs speaks) (1379) Yupuru=rna wangka-nyanta. fast=1sg.s speak-prs ‘I speak fast.’ (PRN: TH1_0034-01) [ii] wangka- (abs speaks to dat) (1380) Japanangka=yi wangka-nya. . . ♂subsection=1sg.o speak-pst ‘Japanangka said to me. . .’ (BSE: TEN1-2017_015-01: 95179_98349) [iii] wangka- (abs speaks abs) (1381) Nganimpa=rnalu kayirra-ngulu-jangka, wangka-nya-ngurra 1pl.excl.s=1pl.excl.s north-abl-ela speak-pst-narr Ngardi-mipa=lku. language_name-restr=then ‘Those of us from the north, we would only speak Ngardi.’ (MMN: TEN1-2017_003-01: 4545_7163) [iv] wangka- (abs speaks with loc) (1382) Wakurra=yi=n=ku=rla ngaju-ngka wangka-ju. neg=1sg[lct]=2sg.s=ep=lct 1sg-loc speak-pot ‘Would you please not talk to me?’ (MMN: TEN1-2017_026-01: 150198_154749)

540 

 9 Grammatical relations and argument structure

[v] wangka- (abs speaks abs to dat) (1383) Wangka-nyanta=pula=nyanu Ngardi. speak-prs=3du.s=refl language_name ‘Those two speak Ngardi to each other.’ (MMN: TEN1-2017_022-01: 615275_621266) [vi] wangka- (abs speaks abs to/with loc) (1384) Nyuntu, wangka-ya=yirla Ngardi! 2sg speak-imp=1sg.lct language_name ‘You, speak Ngardi with/to me!’ (PSM: TEN1-2017_026-02: 969069_975199) 9.2.4.2 Case alternations in bivalent clauses The most common case alternation in bivalent clauses in many Australian languages is an alternation between absolutive and dative objects in ‘transitive’ clauses. Alternations of this kind are labelled ‘conative alternations’ and have been the subject of much discussion in the literature (Hale, 1982; Laughren, 1988; Levin, 1993; Simpson, 1991). First brought to wider typological attention in Hale’s description of Warlpiri, conative alternations have been identified in many Australian languages (see citations in Blake, 1987, pp. 28–29; Evans, 1995, pp. 344–345; and for recent examples see Senge, 2015, p. 567 for Wanyjirra; Gaby, 2017 for Kuuk Thaayorre). In Ngardi, this type of case alternation is available to a range of verbal predicates expressing perception events. An example involving a simple verb is nya- ‘see’ (Table 119) and an example involving a CVC is purda-nya ‘hear, understand’ (Table 120). The CVC purda=nya‘hear, understand’ also permits an argument structure involving a locative subjunct (erg-loc). Table 119: Argument structures of nya- see. nya- see

Transitive (1385)

Conative (1388)

Case frame Bound pronoun Semantic role

[ergx-absy] sx-oy x sees y

ergx-daty sx-o/obly x looks for y

Table 120: Arguments structures of purda=nya- ‘listen, hear, understand, believe’. purda=nya- ‘listen’

Transitive (1387)

Conative (1388)

Subjunct (1389)

Case frame Bound pronoun Semantic role

[erg-abs] sx-oy x hears/understands y

[erg-dat] sx-o/obly x listens out for y

[erg-loc] sx-lcty x believes y

9.2 Argument structure 

 541

Examples (1385)–(1386) illustrate the clause types available to the simple verb nya‘see’. (1385)

Nga=rna=ngkupula nya-nganta, nyunpula. seq.aux=1sg.s=2du.o see-prs 2du ‘I see you two.’ (PJA: TT76_2601: 448816_453024)

(1386) Nya-ngka=lu=rla wirrpa-ngku kirda-ku. see-imp=pl.s=3sg.obl many-erg big-dat ‘All of you mob look for that old man.’ (PSM: TEN1-2017_023-01: 438419_ 440770)454 Examples (1387)–(1389) illustrate the clause types available to the CVC purda=nya- ‘to hear, understand, believe’. Note that alternate argument structures have a significant effect on the semantic interpretations of the predicate in each clause. (1387)

Ngantany-ju jingka purda=nya-nganta. man-erg child listen=see-prs ‘A man hears a child.’ (MMN: TEN1-2018_026-01: 758979_768592)

(1388) Ngantany-ju=rla purda=nya-nganta jingka-ku. man-erg=3sg.obl listen=see-prs child-dat ‘The man is listening out for the child.’ (MMN: TEN1-2018_026-01: 784938_795991) (1389) Wakurra=yirla ngaju-ngka purda=nya-nganta. neg=1sg.lct 1sg-loc listen=see-prs ‘She doesn’t believe me.’ (PSM: TEN1-2018_030-01: 2579703_2581700) In Warlpiri, erg-abs and erg-dat case alternations are available to a wide range of verbs, including those that sit high on ‘transitivity hierarchies’ (Hopper & Thompson, 1980; Tsunoda, 1981a, 1985). In fact, in Warlpiri, conative alternations are observed not only for verbs of perception but even high transitivity verbs of impact and concussion (Hale, 1982; Simpson, 1991), including la- pierce.455 In Ngardi, only a single verb of force has been observed exhibiting a conative alternation: luwa- (erg spears abs; erg shoots at dat). Other verb of force do not 454 Note that Ngardi also has a distinct inflecting verb warri- search which is most frequently found with an [erg-dat] case frame. 455 In Warlpiri, conative dative arguments are further unique in being able to take the ‘extended’ dative cross-referencing bound pronoun =rlajinta. See Nash (1991) for different analyses and functions of the pronominal form -rlajinta in Warlpiri.

542 

 9 Grammatical relations and argument structure

exhibit conative alternations. For example, la- pierce cannot take an erg-dat case frame when occurring as a a simple verb. However, the erg-dat case frame becomes available when the preverb jakarr ‘failed action’ is combined with la- in a complex verb construction. Thus, in Ngardi, ‘conative’ case frames are only available to the ‘highest transitivity’ verbs (e.g. la- pierce) if their semantics are modified by a preverb in the clause – as is the case with jakarr ‘try to’ in (1390). (1390) Jakarr=la-ni ngu=rla ngantany-ju jaji-ku. try=pierce-pst seq.aux=3sg.obl man-erg kangaroo-dat ‘A man tried to spear a kangaroo.’ (PSM: TEN1-2018_026-01: 1198811_1203873) All of the case alternations presented above involve variable case marking of nonsubject arguments. This fact reveals an interesting property of Ngardi in that nearly every verbal predicate strictly selects a single subject type: an ergative subject (i.e. transitive) or an absolutive subject (i.e. intransitive). Only a single lexical verb – janka ‘burn, heat up’ – can appear in clauses with different types of subjects: ergative (A) and absolutive (S) as exemplified in (1391) and (1392) respectively. All other verbal predicates in Ngardi strictly select exclusively ergative subjects or exclusively absolutive subjects.456 (1391)

Janka-nya lamparn jarrilyili-rla, jangu kuja=lu jingka burn-pst little_one cold-loc recog2 sub=3pl.s child naya warlu-ngka karri-nyani always fire-loc be-ipfv.pst ‘The child burns in cold weather time, you know, like how children always sit too close to the fire.’ (TJN: Manungka_02-02202)

(1392)

Purangu-rlu janka-nya ngawu-karra. sun-erg burn-pst bad-result ‘The sun burnt him until he was ill.’ (DMN: LC42a: 701665_704305)

This type of ‘double classification’ of a verbal predicate is an example of a patientive ambitransitive (Dench, 1991, p. 167; Mithun, 2000, pp. 88–114; Dixon, 2002, pp. 176–177). Patientive ambitransitives are predicate types in which the syntactic ‘object’

456 There is some very limited evidence that wangka- speak can occasionally select an ergative subject (despite generally being subcategorised for an absolutive subject). Consider (332) and the combination with raly ‘tease’ in §8.4.4.3. However, this is a very common verb in the corpus and it overwhelmingly appears with absolutive subjects (when overt subjects appear). Note also that the alternation between ergative and absolutive subjects for janka- generates a meaningful different in the sense of the predicate. This is not the case with the isolated alternations for wangka- speak.

9.2 Argument structure 

 543

in the transitive clause (1392), corresponds to the syntactic ‘subject’ in the intransitive clause (1391). Types of NPs which may occupy the S role in the intransitive use of janka- are semantically constrained to referents which are themselves nominals depicting fire (e.g. warlu ‘fire’) or nominals that be subject to burning (e.g. yuka ‘grass’) – although this extends to a wider notion of ‘to be affected by fire/heat’ as in the examples above (that is, ‘to be sunburnt’, ‘to have sunstroke’, ‘to have burns’). Interestingly, the presence of a patientive ambitransitive verb with the semantics ‘burn/cook’ in the lexicon is something of a widespread areal feature – appearing as the only doubly classified verb in Wangkajunga (Jones, 2011, p. 153), Warlmanpa (Nash, 1979), and Warlpiri (Hale, 1982, pp. 240–241) and one among a limited set in Ngumpin languages Walmajarri (Hudson, 1978, p. 53) and Jaru (Tsunoda, 1981b, p. 76) as well as languages further afield, such as Arrernte (Wilkins, 1989, p. 225) and Kalkatungu (Blake, 1979a, p. 51). 9.2.4.3 Case alternations in trivalent clauses Another common type of case alternation is between trivalent predicates which have alternative case frames involving either indirect object or subjunct arguments. Semantically this generally involves the substitution of a participant with the semantic role of recipient for one with the semantic role of accompaniment. To exemplify, the predicate yupa-send has three possible case frames: erg-abs (1393), (1394); erg-abs-dat (1395) and erg-abs-loc (1396). The dative NP in (1395) is an indirect object while the locative NP in (1396) is a subjunct. (1393)

Yupa-ka-rni kaniny-kaniny. send-imp-hith inside-rdp ‘Send him this way, inside!’ (PSM: TEN1-2017_026-01: 3128673_3130601)

(1394)

Yupa-rni=pulany kayirra-ngulu, wayitpali-rlu. send-pst=3du.o north-abl white_man-erg ‘The whitefella sent the two of them from the north.’ (NMN: LC20a)

(1395)

Minya=rna=rla yupa-rni mu-ku ngaringka-ku. this=1sg.s=3sg.obl send-pst this-dat woman-dat ‘I sent it to this woman.’ (PSM: TEN1-2017_008-01: 1655032_1660036)

(1396)

Yupa-rnanta=yi=lu=rla ngaju-ngka. send-prs=1sg[lct]=3pl.s=lct 1sg-loc ‘They sent it with me.’ (MMN: TEN1-2017_021-03: 1701964_1704640)

544 

 9 Grammatical relations and argument structure

9.2.4.4 Complex verb constructions and predicate argument structure While Ngardi lacks any applicative or other valency-changing morphology, the formation of complex verb constructions (§8.3) allows for considerable flexibility in transitivity values of a given clause. Generally speaking, preverbs do not carry transitivity information and are not typically ‘valency changing’. Instead, their association with different inflecting verbs allows for their appearance in clauses with varying transitivity values (§9.2.4.4.1). Some select examples of preverbs which do exhibit valency-modifying functions are presented in (§9.2.4.4.2). 9.2.4.4.1 Inflecting verbs as valency changers The ‘valency changing’ function of inflecting verbs in the formation of CVCs was first mentioned in §8.3.1.3 where the inflecting verbs ma- get and karri- be were both described as having functions as a ‘causative/transitive verbaliser’ and ‘an inchoative/intransitive verbaliser’, respectively. Of all the inflecting verbs, these two verbs most frequently combine with nominals to form complex verb constructions. They are briefly re-exemplified in Table 121 with a focus on their function as a valency-altering strategy. Table 121: Inflecting verbs as valency changers for nominals/preverbs expressing states.

wuruju ‘good’ kirda ‘big, senior’ marrka ‘strong’

pirdiny ‘get up, arise, awake’

Intransitive karri- be

Transitive ma- get

wuruju=karriabs is good kirda=karriabs becomes big/grows up marrka=karriabs is strong

wuruju=maerg heals/improves abs kirda=maerg brings up abs marrka=maerg makes abs strong abs tightens/fastens abs pirdiny=maerg causes abs to be upright/awake

pirdiny=karriabs comes to be upright abs gets up

Postural preverbs (e.g. luk ‘lie down’, pukany ‘sleep’) also commonly appear with intransitive verbs (e.g. karri- be, wanti- fall) or with the transitive verbs yirra- place or pu- hit expressing a ‘caused position’ event as exemplified in Table 122.

9.2 Argument structure 

 545

Table 122: Inflecting verbs as valency changers for postural preverbs. Intransitive karri- be luk ‘lie down’ parnta ‘on top’

luk=karriabs is lying down parnta=karriabs is on top

jurr ‘position down’ pukany ‘sleep’

jurr=karriabs is low

purluk ‘submerge’

purluk=karriabs swims

pukany=karriabs sleeps

wanti- fall

Transitive yirra- place

luk=wantiabs lies down

luk=yirraerg lies abs down parnta=yirraerg places abs on top jurr=yirraerg places abs down

pukany=wantiabs lies down to sleep purluk=wantiabs dives in

pukany=yirraerg puts abs to sleep purluk=yirraerg puts abs to sleep

pu- hit

parnta=puerg places abs on top jurr=puerg puts abs down pukany=puerg puts abs to sleep

9.2.4.4.2 Preverbs as valency changers In many Australian languages with complex verb constructions, it has been found that the association of preverbs or ‘coverbs’ with inflecting verbs can result in a complex predicate whose valency is not that which would be typically expected of the inflecting verb in isolation (see Reid (2000) for Ngan’gityemerri and Schultze-Berndt (2000) for Jaminjung). This is also observed in Ngardi. For example, the canonically transitive verb pu- hit exhibits monovalent argument structures in select CVCs, as shown in Table 123. Table 123: Anomalous argument structures of CVCs involving pu- hit. Bivalent

Monovalent

pu- hit

erg hits abs

—

jut=pujungkuj=puwirilyi=punyinypurru=ma-

— — — —

abs sets off abs starts/sets off abs walks around erg blows nose

kuny=pu-

erg waits for dat

—

The same phenomenon is observed with some combinations of ma- get with various preverbs expressing types of sound emission and communication as well as some other distinct preverbs as shown in Table 124.457 457 It is worth repeating here that other Ngumpin languages have a contrast between two lexical verbs ma- say/do and ma- get which belong to separate conjugational classes. This is not the case in Ngardi (see the associated discussion in §8.4.1.2.4).

546 

 9 Grammatical relations and argument structure

Table 124: Anomalous argument structures of CVCs involving ma- get. Bivalent

Monovalent

ma- get

erg gets abs

—

jingkirdi=manyinypurru=marung=mawarrpaj=ma-

— — — —

abs laughs abs sniffs abs barks abs signals with hand

ngaparayi=majapurrk=ma-

—

abs yawns/sighs abs jumps

Generally speaking, all of the monovalent argument structures generated from ma- in CVCs predictably involve absolutive subjects, as in (1397). (1397)

Wakurra, japurrk=ma-ni kankarni kirda-nguniny ngaakman. neg jump=get-pst above big-dim frog ‘Oh no, the big frog jumped on top.’ (MMN: TEN1-2019_005-06: 229747_232673)

However, there are some select additional examples where a monovalent complex verb takes an ergative-marked subject (e.g. nyinypurru=pu- ‘to blow one’s nose’). In (1398), the subject is non-overt but its ergative status is indicated by the secondary predicate kungkurru-kurlu-rlu which agrees in case with the elided ergative subject NP. (1398)

Nyinypurru=pu-ngani=rna, kungkurru-kurlu-rlu. nose_blow=hit-ipfv.pst=1sg.s sick-prop-erg ‘I was blowing my nose, while I was sick.’ (PSM: TEN1-2018_001-02: 1088552_1091524)

Additionally, as shown by kuny=pu- (erg searches for dat) in Table 123, there are also cases in which a preverb does not alter the valency of a verb, but it does alter the expected case frame and argument structure of the predicate. The clearest example of this involves the adverbial preverb jakarr ‘failed action’ and the verb of force lapierce. The verb la- in its simple form is a simple transitive verb with semantics of ‘dig’ and ‘shoot’. It cannot select a dative-marked indirect object. However, in combination with preverb jakarr, the resulting complex verb subcategorises a ‘conative’ case frame in which the theme of the event is encoded as a dative NP and registered as an oblique bound pronoun, as shown in Table 125. 9.2.4.5 The unusual verbal predicate ngarri- tell The verbal predicate ngarri- tell stands out as an unusual predicate for at least two reasons. First, it can be used in transitive clauses with an additional object comple-

9.2 Argument structure 

 547

Table 125: Argument structures of simple and complex verbs involving la- pierce. Predicate Simple

la- pierce

Complex

jakarr=la- ‘shoot at’

Case frame

Bound pronoun

erg shoots abs/ erg digs abs erg shoots at dat

s-o s-obl

ment (§9.2.3.8) of the type: erg calls/names abs, abs(complement). Second, it can register a topic of speech as a reflexive argument and this will be registered in the pronominal complex in addition to the subject and indirect object as in (1399). This constitutes a rare example of a three-member pronominal clitic complex (§6.4.3) and the only one not involving the third person singular oblique clitic =rla. (1399) Ngarri-rni=yi=lu=nyanu ngantany-ngampurr-u. tell-pst=1sg.o=3pl.s=refl man-paucal-erg ‘The men spoke to me about themselves.’ (PSM: TEN1-2018_028-01: 409321_422417) There is another instance of a reflexive clitic where it refers to the recipient and not the theme (1400), although this single example is admittedly somewhat idiosyncratic. (1400) Mula-ngka=nyanu ngarri-rnanta. here-loc=refl tell-prs ‘It is being recorded here (lit. ‘it is telling (it) to itself here’).’ (NNM: LC33b: 962871_965507) If the recipient of ngarri- tell is registered not as a dative indirect object, but as a locative subjunct, then it is not possible for the recipient as well as the theme to be registered in the bound pronoun: (1401) Ngaringka-rlu=nyanu ngarri-rni ngantany-ja. woman-erg=refl tell-pst man-loc /(*ngaringka-rlu=yanu=nyanu=rla) woman-erg=3pl[lct]=refl=lct ‘The woman spoke with the men about herself.’ (PSM: TEN1-2018_028-01: 745732_752884) It may be possible to consider the reflexive clitic in (1399) as a type of external possession marking, that is, ‘the men told me their (story)’. However, it is notable that it is only this predicate which has been observed permitting this particular bound pronominal complex and if it the relationship is simply one of possession, there

548 

 9 Grammatical relations and argument structure

should not be any reasons why it could not be added freely to other clause types. Consider also the same restrictions observed in the pronominal complex in Walmajarri (Hudson, 1978, p. 68) and similar use of reflexives in expressing a topic of verbs of speech in Wanyjirra (Senge, 2015, pp. 586–587).

9.3 Secondary predicates Secondary predicates are adjunct, nominal (or nominalised) expressions which express a temporary property of a participant associated with the time frame of the main predicate event (Schultze-Berndt & Himmelmann, 2004). Secondary predicates form a separate predicative element to the main predicate but occur within the same prosodic unit. They are distinct from serial verb constructions (§8.2), which also involve multiple predicative elements in a single clause, since they are non-finite and do not share the arguments of the main predicate. There are two distinct types of secondary predication in Ngardi: depictives (§9.3.1) and resultatives (§9.3.2). As defined by Schultze-Berndt and Himmelmann (2004, pp. 65–66) depictives ‘designate a state of affairs which hold at the same time as the eventuality encoded by the main predicate while resultatives designate an eventuality which is a consequence, or result, of the eventuality encoded by the main clause predicate’. In Ngardi, resultatives are morphologically distinct from depictives in that they are introduced by a specific suffix -karra. Depictives, on the other hand, can be very difficult to distinguish from discontinuous NPs and/or NPs in apposition since they agree in case with their controller. Some select temporal clitics nevertheless can distinguish depictives from other NPs which agree in case. This is discussed in §9.3.1.1

9.3.1 Depictives Depictives are a type of secondary predication which serves to modify an argument in a manner similar to NP modifiers. They are distinguished from NP modifiers by the fact that they are a type of predication that specifically holds at the time of the eventuality expressed by the main clausal predicate, rather than modifying a head of a noun phrase (see Schultze-Berndt & Himmelmann, 2004, p. 60). Depictives agree in case with the argument of which they are predicated, termed ‘controller’. Depictives are most frequently controlled by subjects, whether they be within ditransitive (1402) transitive (see (1415)) or intransitive (1403) clauses. (1402)

Wakurra=yi=n kapanku-rlu yi-nya. neg=1sg.o=2sg.s quickly-erg give-pst ‘You didn’t give it to me straight away.’ (DMN: LC42b: 362480_364331)

9.3 Secondary predicates 

(1403)

 549

Tarn=yirra-rnanta=rla iliri ngarda wirriri=ya-nankura hang=place-prs=3sg.obl hide hyp back=go-ipfv.hort karrmantawangu. hungry ‘He is putting it safely away for her (the mother) in case she comes back this way hungry.’ (PSM: TEN1-2018_007-03: 65573_84335)

Depictives can also be controlled by objects as per the NPs warlawarnu ‘cooked’ and kunyka ‘raw’ in the consecutive clauses in (1404).458 (1404)

Kupa-ku=rnalu warla warnu ka-ngankura. cook-pot=1pl.excl.s cooked carry-ipfv.hort ‘We will cook and carry some (bush potatoes) back cooked.’ Yirda=rnalu ka-ngankura kunyka. another=1pl.excl.s carry-ipfv.hort raw ‘We will (also) bring back another lot raw.’ (TJN: Manungka_01-035155: 792421_795689)

The types of nominals which may function as depictives are diverse but frequently include nouns (1405), adjectives (1406), and action/manner nominals (see (1402)). Additionally, independent use of preverbs (1407) and (case-inflected) non-finite verbs (1408) can both be considered types of depictives but this type is discussed in greater detail within the treatment of ‘non-finite subordinate clauses’ (§11.3). (1405)

Ngaju=rna=yanu jingka-ngku nya-ngani.459 1sg=1sg.s=3pl.o child-erg see-ipfv.pst ‘As a child, I used to watch them.’ (PSM: TEN1-2017_007-02: 85847_88784)

(1406)

Parrajardu-rlu=rna ngapa nga-rnanta. thirsty-erg=1sg.s water eat-prs ‘Thirsty, I am drinking water.’ (PJA: TT76_2601: 466420_470551)

458 Both secondary predications are of objects: in the first clause an elided nominal karnti ‘bush potato’ and in the second clause yirda ‘another one/the other one’. Importantly, the nominal yirda has the special property of referring to an additional member of a logical set. Here it is the same referent referred to in the previous clause. One is forced to interpret warlawarnu ‘cooked’ in the initial clause as being a temporary property predicated of karnti (i.e. some bush potatoes were cooked) since yirda makes anaphoric reference to additional members of a wider set, i.e. ‘some others being raw’. 459 This example includes an instance of a free pronoun that lacks ergative case-marking. See §5.4.1 for discussion.

550 

 9 Grammatical relations and argument structure

(1407)

Yut=karri-nyani-nyirra pukany yalu-wana=lku. sit=be-ipfv.pst-narr asleep that-perl=still ‘He was lying down, asleep, still near them.’ (PPN: Manungka_01-035160)

(1408)

Kupa-rnu-wangu-rlu yi-nya=rla. cook-inf-priv-erg give-pst=3sg.obl ‘Not having cooked (it), he gave it to her.’ (TJN: Manungka_01-035160: 639747_649030)

In addition to simple nominal stems showing agreement with their controllers, depictives can also take the form of case-inflected NPs. Examples involve the privative (1409), proprietive (1410), and locative (1411) case suffixes. Case suffixes in these contexts essentially derive stative predicates from eventive ones. When controlled by transitive subjects they receive further (ergative) case marking in agreement, as in (1409) and (1410). Case-marked NPs functioning as depictives can also be object controlled. Thus, in (1411), the nominal waku- is locative case-marked but is controlled by an absolutive object of the main predicate kiji- throw. (1409)

Iji-wangu-rlu nga-lku, ngaju-kariny-ju=lku. gift-priv-erg eat-pot 1sg-other-erg=then ‘I will eat it without giving it away, (I will eat) all by myself.’ (TT76_2201: 1961893_1964790)

(1410)

Ya-nku=rna ngapa=wali pirda-kurlu-rlu nga-lku. go-pot=1sg.s water=then full-prop-erg eat-pot ‘Full up (with food), I will go and drink some water.’ (TJL: TT76_2201: 1314805_1317736)

(1411)

Ya, kiji-ka=yi waku-ngka. yes throw-imp=1sg.o nothing-loc ‘Yes, throw it to me, (me) with nothing.’ (LC56b: 823716_827692)

Further examples of secondary predicates which show multiple case-marking are included below. They are generally restricted to examples involving temporal (1412) or locational (1413) predications. Example (1413) is a rare instance of a secondary predication controlled by a dative main clause argument. Further examples were provided in the discussion of multiple case-marking involving locative adjuncts (§4.7.3).

9.3 Secondary predicates 

 551

(1412)

Junguny-ju jaku=ma-nanta=lu=ngala munga-ngka-rlu mouse-erg depart=get-prs=3pl.s=1pl.incl.o night-loc-erg mangarri, nga-rnanta. veg_food eat-prs ‘The mouse steals our food in the dark of night and eats it.’ (MDN: Cataldi, 2011, p. 50)

(1413)

Kupa-rni=rna, ka=rna=yanu kupa-rni-nyirra stockman-ku cook-pst=1sg.s emph.aux=1sg.s=3pl.o cook-pst-narr stockman-dat ngurra-ngka-ku. camp-loc-dat ‘I would cook, I used to cook for the stockmen at the stock camp.’ (YMN: LC23b: 149262_155066)

In Warlpiri, it is argued that the use of the ergative on locationals enforces the interpretation that the subject (in addition to the event and the object affected) was situated at the location, whereas in the absence of the ergative it is possible that the subject was not (Hale, 1982, p. 268; Simpson, 1991, pp. 207–208). How this fine-grained semantic distinction would readily apply to temporal secondary predicates (e.g. mungangka-rlu dark-loc-erg) is not entirely clear.460 A subset of depictives, termed ‘circumstantials’, are sometimes distinguished on the basis that they provide a temporal or semantic link with the event of the main clause but have a somewhat looser relationship to the clausal event than other depictives (Simpson, 1983, pp. 380–382). In Ngardi, such types of ‘secondary predicates’ include non-finite subordinate clauses marked with the ergative, locative or elative (sometimes referred to as T-Complementisers). An example is provided in (1414) but see §11.3.6.1 for further details. (1414)

Minya ngaringka-rlu kupa-rnu-karra-rlu jingka marda-rnani-nyirra. this woman-erg cook-inf-ssub-erg child have-ipfv.pst-narr ‘Cooking, the woman was holding the baby.’ ‘The woman – in the act of cooking – was holding the baby.’ ‘The woman was holding the baby while cooking.’ (PSM: TEN1-2018_027-01: 2600670_2611955)

460 There are some differences in analyses as to how absence of ergative on locationals or temporals is to be interpreted in Warlpiri. Hale (1982, p. 268) argues that the absence of the ergative in these contexts is used precisely to indicate that the spatial case expression is predicated of the absolutive (i.e the unmarked argument). Simpson (1991, pp. 241–246), however, argues that the absence of the ergative reflects the fact that the spatial case expression is predicated of an event rather than a participant.

552 

 9 Grammatical relations and argument structure

9.3.1.1 Temporal clitics as a diagnostic for depictive secondary predicates While a working definition for depictives was provided in §9.3.1, distinguishing depictives from types of complex NPs can be difficult (Laughren, 1992; Schultze-Berndt, 2006; Simpson, 2005). This is due to both frequent deletion of NPs and the fact that constituents of complex NPs can be discontinuous. Consider the nominal parrajardu in (1415) and the two possible readings which can be associated with it. (1415)

Parrajardu-rlu=rna ngapa nga-rnanta (ngaju-ngku). thirsty-erg=1sg.s water eat-prs 1sg-erg ‘Thirsty, I am drinking water.’ [Secondary predication reading] ‘I, the thirsty one, am drinking water.’ [Attributive modifier reading] (PJA: TT76_2601: 466420_470551)

The nominal parrajardu in (1426) can be interepreted either as: (i) a secondary predication on the subject, expressing a state of the subject holding at the time of the main event (ii) a modifying NP agreeing in ergative case with the transitive subject (ngaju-ngku). These two interpretations are variously referred to as depictive or attributive readings or, in Hale’s (1982) terms, merged and unmerged. In Ngardi, as in Warlpiri, there is no overt syntactic difference between merged and unmerged readings of complex NPs (Hale, 1982).461 Despite the relative lack of syntactic tests for disentangling attributive and depictive functions of nominals, there are some specific contexts in Ngardi in which only depictive readings are possible. First, action nominals and weak nexus preverbs can only be used with secondary predicative meanings. This derives from their inherent lexical semantics as dynamic predicates. For other nominal subtypes (namely adjectives and nouns), NPs agreeing with main clause arguments have clear secondary predicate functions when encliticised with the diagnostic temporal clitics =lku and =wali, both of which may have either phrasal (NP-level) or clausal (predicate-level) scope. I will briefly exemplify with the clitic =lku.462

461 Ideally, with a large enough corpus of naturalistic data involving similarly case-marked NPs, it might be possible to conclude whether the majority of ‘discontinuous NPs’ are actually better analysable as secondary predications. 462 Laughren (1992) shows that reflexivisation places constraints on how and where secondary predicates can be deployed in Warlpiri. Laughren compares this with part-whole constructions that exhibit greater freedom with respect to reflexivisation. Part-whole relations are considered a type of secondary predicate by Laughren but are treated as ‘external possession’ (9.4.2) in this grammar. Unfortunately, I have no data to clearly indicate how secondary predication interacts with syntactic reflexivisation in Ngardi.

9.3 Secondary predicates 

 553

One function of the temporal clitic =lku (§10.4.1) is to modify a proposition as being true at two successive referent times: T1 and T2. The clitic =lku exhibits scope over the constituent to which it attaches and can roughly be translated as ‘still x, still as of yet x’. When marking an NP expressing a state it essentially describes a ‘persistent’ state. Thus, consider wanyji=lku in (1416). (1416) Ya-ni=rna wirrirl junga=lku=rna=yanu wanji=lku timana go-pst=1sg.s back recog2=then=1sg.s=3pl.o alive=then horse parli=pi-nya ngaju-ku. find=hit-pst 1sg-dat ‘I went back and I found my horses, still alive.’ *’I went back and then I found my alive horses.’ (PSM: TEN1-2016_005-03: 396454_399799) In (1416), the speaker (Patrick Smith) was not seeking to contrast one set of object referents (i.e. his alive horses) with another set (his ‘dead horses’), but rather a persistent state of a single referent: ‘his horses still living’. There is an implication here that his horses were not going to be alive at some later time (T2) (many other horses had been put down). Despite the expectation, the proposition is true at T2 (they were still alive) and this is marked on the predicate wanji by the encliticisation of =lku. Importantly, the scope of the clitic =lku over the proposition ALIVE(my horses) is not possible if wanji itself cannot function as a secondary predicate on the argument timana.463 A second illustration is provided by the constituent tayimap=pa=lku (a Kriol borrowing) in (1417). This constituent is a secondary predicate which takes jina minya ‘these feet’ as its argument. Again, the temporal clitic =lku has scope over the state expressed by the nominal tayimap. It expresses a persistent state which continued to hold at the time of the main clause predicate. Note that the argument jina minya is itself in a part-whole relationship with the clausal argument ‘them’ registered in the bound pronoun by the third person plural =yanu. (1417)

[Jina minya tayimap=pa=lku jayin-ta]=li=yanu ka-ngani. foot this tied_up=ep=then chain-loc=3pl.s=3pl.o carry-ipfv.pst ‘They were taking them with their feet still tied up in chains.’ *’They were still taking [‘the ones chained by their feet’].’ (PSM: TEN1-2017_006-02: 125928_129829

463 Note that =lku can exhibit scope over a VP but in this context it invariably appears encliticised to the verbal word.

554 

 9 Grammatical relations and argument structure

Additionally, the clitic =lku also has a surprisingly antonymic function describing a state as being previously false at an earlier time (T1) and yet subsequently true at a later time (T2). This function is actually the precise function of a cognate enclitic =lku in Warlpiri (Nash, 1986, p. 56). Again, =lku in this function is able to take a secondary predicate NP within its scope, as in (1418). (1418)

Ngaju=rna ngaju-kariny=pa=lku pukany=karri-ju. 1sg=1sg.s 1sg-other=ep=then sleep=be-pot ‘I will sleep by myself then. *I myself will sleep then.’ (DMN: LC32b)

In conclusion, evidence from temporal enclitics suggest that there are grounds for proposing that Ngardi has both depictive secondary predicates and discontinuous complex NPs but these are necessarily hard to distinguish in the vast majority of cases.

9.3.2 Resultatives Resultatives are a type of secondary predication that describe a state of an object that has a causal relationship to the event depicted by the main clausal predicate. That is to say, the main event is understood to bring about the state expressed by the resultative NP. Resultatives have a closer relationship to the main verbal predicate than depictives and can affect the interpretation of the clausal event. Specifically, resultatives mark the telic endpoint of an event or state. Resultatives are compatible with verbs exhibiting a range of aktionsart properties, including unbounded events such as activities. Resultatives in Ngardi are encoded by a single suffix -karra, homophonous with the subordinating case suffix -karra (§11.3.4.2), and can be a range of arguments including intransitive subjects (1419), transitive subjects (1420), and transitive objects (1421). Unlike depicitives, they do not agree in case with their controller. (1419)

Ya-ni=rna yumpaly-karra. go-pst=1sg.s tired-result ‘I went until I was tired.’ (MMN: TEN1-2018-023-02: 6527960_6529243)

(1420)

Ngantany-jui nga-rni kuyi pirda-karrai. man-erg eat-pst meat full-result ‘The man ate meat until he was full.’ (MMN: TEN1-2018_025-01: 2071712_ 2077517)

9.4 Possessive constructions 

(1421)

 555

Ngantany-ju pila=ma-ni goannai tatakurr-karrai. man-erg follow=get-pst goanna exhaustion-result ‘The man chased a goanna to the point of exhaustion.’ (PSM: TEN1-2018_02701: 3707454_3715070)

The ability for resultatives to be predicated of any clausal argument can be illustrated by comparing (1422) with (1420), in which the main clausal arguments are rearranged by using the verb ma- get in a causative function. (1422)

Kuyi-ngku pirda=ma-ni ngantanyi pirda-karrai. meat-erg full=get-pst man full-result ‘The meat filled up the man until he was full/satisfied.’ (TEN1-2018_025-01: 2047517_2052486)

In cases where a resultative occurs within a clause whose main predicate depicts an achievement, the verbal event is interpreted as being iterated until the time point expressed by the resultative state, (i.e. ‘dead’ in (1423)). The ‘subsequent’ nature of the resultative is further reinforced by the use of the ‘then’ enclitic (§10.4.1). (1423)

Ngantany-ju la-ni jajii nyurnu-karra(=lku)i. man-erg pierce-pst kangaroo dead-result=then A man speared the kangaroo until it was dead. (MMN: TEN1-2018_025-01: 1874992_1886370)

9.4 Possessive constructions Possessive constructions in Ngardi cover a range of relations between two NPs in Ngardi and are not strictly restricted to the expression of possession or ownership but extend to kinship, body parts, and use/manipulation of tools. There are a number of morphosyntactically distinct possessive constructions in Ngardi. Possession can be expressed at the clause level by the verb marda- have, where the possessor (pr) is encoded as an ergative subject and the possessum (pm) as an absolutive object, as in (1424). (1424)

[Ngaju-ngku]pr=rnapr marda-rnanta [kirda-nyayirniny milpa]pm. 1sg-erg=1sg. have-prs big-intens eye ‘I have really big eyes.’ (MMN: TEN1-2018_026-01: 1297356_1309570)

If the possessor is a head of a noun phrase, there are several morphological strategies for encoding a possessive relation, including the proprietive suffix (§4.6.4), the dative suffix -ku (§4.5.4.3), the possessive suffix -kuny (§4.6.6), the personal possessive suffix,

556 

 9 Grammatical relations and argument structure

-punta (§4.6.7), and two kinship suffixes: the anaphoric propositus -nyanu (§5.10.1.2) and the possessed kin -nguniny (§5.10.1.3). In this section, four syntactically distinct strategies are described for encoding possession termed i) internal possession, ii) external possession, iii) prominent internal possession (Bárány, Oliver & Nikolaeva, 2019; Corbett, 2017; Meakins & Nordlinger, 2017), and iv) adnominal clitic possession. These terms are adopted as they reflect the morphosyntactic properties of the constructions rather than their semantic or functional usages. Elsewhere, the first two construction types approximately correspond to what have been described as ‘alienable’ and ‘inalienable’ possession, respectively (Dixon & Blake, 1983, 122ff.). Inalienable possessive constructions are also referred to as ‘part-whole constructions’, reflecting the common semantic relationship between possessum and possessor (Hale, 1981; Laughren, 1992; Tsunoda, 1995). All of these terms are semantically-oriented labels reflecting the semantic basis of some of the common relationships they encode. They are problematic in Ngardi at least insofar as it is entirely possible for the same nominal/referent to be either internally or externally possessed and so I label possessive constructions based on their morphosyntactic properties rather than inferred semantic relationships. In the subsections that follow I employ the terms possessor (pr), possessum (pm) following Payne and Barshi (1999). I indicate the entire NP which contains the possessive relation with pp (possessive phrase). The morphosyntax of each possessive construction is summarised in Table 126 with reference to the case marking and pronominal cross-referencing of the possessor (pr), the possessum (pm) and the possessor phrase (pp). Table 126: Morphosyntax of possessive constructions. Adnominal case marking Internal (§9.4.1)

pr pm

-kuny poss; -punta pposs; -ku dat ∅

-nyanu, -nguniny

— —

pr

—

✓

pm

—

—

pp External (§9.4.2)

∅

Cross-referencing in argument positions

pp

✓ nil

✓(As obl)

Prominent internal (§9.4.3)

pr

-ku dat

pm

—

Adnominal clitic (§9.4.4)

pr

—

✓

pm

—

—

pp

pp

—

✓ (As obl) —

9.4 Possessive constructions 

 557

9.4.1 Internal possession Internal possession constructions are defined by all marking of possession occurring internal to the noun phrase (NP). Should the possessive construction happen to occupy an argument position within the clause, the entire possessor phrase is cross-referenced. Bound pronouns are not used to mark the possessor. Two subtypes of internal possession can be identified based on whether the possessor or the possessum is case marked for the possessive relationship. Beginning with the latter, the possessum may be marked as being possessed by the ‘anaphoric propositus’ suffix -nyanu or the ‘possessed kin’ suffix -nguniny. The types of semantic relations that may be encoded in this way are restricted to kin and close relations. Examples were provided in §5.10.1.2 and §5.10.1.3 respectively and one is repeated in (1425). (1425)

Kuyi=pula=rlapp [ngati-nyanu-ku]pp ka-ngu-ngkarla meat=3du.s=3sg.obl M-anaph=3sg.obl carry-inf-seq.loc yi-nya-ngurra. give-pst-narr ‘The two gave the meat to their mother, having carried it.’ (MMN: LC3)

Alternatively, the possessor may be marked and the possessum left unmarked. Possessor-marking suffixes include the possessive -kuny, the personal possessive -punta or the dative -ku. Since adnominal case-marked NPs in Ngardi are nearly always further marked with case indicating their role in the clause, the possessive -kuny suffix is most commonly used in marking internal possession (no case suffixes can follow the dative case). The following examples illustrate possessive NPs with various case roles of subject (1426), object (1427), indirect object (1427) an allative subjunct (1429), and a locative adjunct (1430). [i] Subject (1426) Paja-rni=lupp [ngaju-kunypr-ju kunyarrpm-u]pp. bite-pst=3pl.s 1sg-poss-erg dog-erg ‘My dogs killed it.’ (PSM: TEN1-2017_018-01: 804629_807537) [ii] Object (1427) [Nangala-kunypr paniyaps]pp rulany=la-ni.464 ♀subsection-poss eye jerk=pierce-pst ‘Nangala’s eye got scratched.’ (TJN: Cataldi, 2011, p. 152)

464 This is also one of the rare examples of -kuny marking possession of a body part in the data.

558 

 9 Grammatical relations and argument structure

[iii] Indirect object (1428) Minya mangarri yu-ngka=rla pp [nyuntu-kunypr-ku jamirdipm-ku]pp. this veg_food give-imp=3sg.obl 2sg-poss-dat MF-dat ‘You, give this food to your father!’ (MMN: TEN1-2017_018-01: 516465_525736) [iv] Subjunct (allative-marked) (1429) Yaparti-nya=rlanyanta pp [Nakarra-kunypr-kurra, ngatipm-kurra]pp. run-pst=3sg.lct ♀subsection-poss-all M-all ‘He ran towards Nakarra’s mother.’ (PSM: TEN1-2017_018-01: 1168921_1173931) [v] Adjunct (locative-marked) (1430) Ya-nta-rni=lu [ngaju-kunypr-ja motikapm-rla]pp. go-imp-hith=pl.s 1sg-poss-loc car-loc ‘Come in my car!’ (MMN: TEN1-2017_018-01: 940682_946487) An example of the more restricted personal possessive -punta (which only marks close possessive relationships between kin, Dreamings and country) is provided in (1431). (1431)

[Ngaju-puntapr-rlu ngatipm-ngku]pp=yi ka-nya mariyawu. 1sg-pposs-erg M-erg=1sg.o carry-pst long_way ‘My mother took me all the way.’ (MMN: fieldnotes, TEN1-2017_029-02)

The dative is only rarely found in this type of possessive construction and necessarily can only occur when the possessor phrase is not assigned an overt relational case (e.g. ergative) since the dative cannot appear as an ‘inner case’ in multiple casemarking.465 A rare example of the dative marking a possessor in an internal possession construction is provided in (1432).

465 Tsunoda similarly notes that possessive usages of the dative in Jaru without concomitant cross-referencing are rare (Tsunoda, 1981b, p. 196). This is also the case in Pintupi, where all possession marked with -ku obligatorily requires possessor cross-referencing, while the ‘possessive relator’ -kunu (cf. Ngardi -kuny) is only used in internal possession constructions (Hansen & Hansen, 1978, p. 54).

9.4 Possessive constructions 

(1432)

 559

nganayi-rla [Ngaju-kupr ngatipm]pp wangu=karri-nya yalu-rla=lku 1sg-dat M nothing=be-pst that-loc=then whatsit-loc kulirra Gordon Down-ta. south place_name-loc ‘My mother passed away there, at whatstitplace, there in the south, Gordon Downs.’ (NGO: TEN1-2017_011-01: 31300_36842)

Despite examples of the dative marking adnominal possession as in (1432), the possessive also can be used in contexts where no further case marking appears, as in (1427). The dative is only marginally used as an adnominal case in possessive constructions.

9.4.2 External possession In external possession constructions, there is no NP-internal marking of the possessive relationship.466 Instead, the possessor and possessum are placed in an appositional relationship (see §4.2.4.4) and the possessor alone is cross-referenced in the bound pronoun, should it occupy an argument position. The nature of the semantic relationship between types of NPs which may be ‘externally possessed’ is generally that of ‘part-whole’ relations (e.g. body parts) as in (1433) but also extends to such additional categories as kin relations and one’s name (1434). (1433)

Timanapr=rna=yanupr lirrapm kirnik=nya-ngani-nyirra. horse=1sg.s=3pl.o mouth stare=see-ipfv.pst-narr ‘I stared at the horses’ mouths.’ (MDN: Cataldi, 2011, p. 125).

(1434)

Puyuku [minya ngurra]pr yirdipm. place_name this camp name ‘The name of this place is Puyuku (lit. ‘Puyuku, this place, name.’).’ (TJN: LC21a: 792787_796729)

The fact that it is the possessor (or ‘whole’) that is cross-referenced when this type of possessive construction occupies argument roles is demonstrated by the number and person features of the object bound pronoun in (1435). 466 Payne and Barshi (1999, p. 3) provide the following definition of external possession crosslinguistically: ‘constructions in which a possessor-possessum relation is expressed by coding the possessor as a core grammatical argument of the verb and in a constituent separate from that which contains the possessum’. Functionally, external possession is typically used to encode possessive relationships that involve an ‘indissoluble relationship between two entities’ (Chappell & McGregor, 1995, p. 4).

560 

 9 Grammatical relations and argument structure

(1435)

Ngarda=yipr wiriny=ma-nngi jina-kujarrapm. hyp=1sg.o bind=get-irr foot-dual ‘He might bind up my two feet (lit. ‘he might bind me up, the two feet’).’ (PSM: TEN1-2018_037-01: 196807_213432)

Widespread patterns of null anaphora in Ngardi frequently result in the lack of an overt possessor nominal Instead, a bound pronoun serves as the only exponent of the possessor. Examples are (1436)–(1438). (1436) Jarlapm=rnapr=rla ngawu=karri-nya. stomach=1sg.s=3sg.obl bad=be-pst ‘I felt very bad for her (lit. ‘I, the stomach, felt bad for her’).’ (DMN: LC42b: 1048306_1052034) (1437)

Paka-ngku=yipr la-ni jinapm. prickle-erg=1sg.o pierce-pst foot ‘A prickle poked my foot (lit. ‘a prickle poked me, the foot’).’ (PSM: TEN12018_001-01: 1421033_1423634)

(1438)

Purangu-rlu=ngku pu-ngku wirrkirl ngawu-kariny. sun-erg=2sg.o hit-pot head bad-other ‘The sun will strike your head (until) you are really ill (lit. the sun will strike you, the head, (until) you are really ill.’).’ (DMN: LC42a: 742711_747419)

If the possessive phrase occupies an object position in the clause and the possessor is co-referent with the subject, then the object bound pronoun indexing the possessor can optionally be replaced by =nyanu. Thus, compare (1439) and (1440). It is unclear if there is a slight meaning different between reflexivised vs non-reflexivised strategies. (1439)

Ita=n warlu yirra-ku kuja ma-nku=nyanupr takapm. small=2sg.s fire place-pot sub get-pot=refl hand ‘You should make a small fire so that he can get his hands (warm).’ (PSM: TEN1-2018_031-03: 410625_416863)

(1440)

Parraj=ma-ni=rna=yipr jinapm. burnt=get-pst=1sg.s=1sg.o feet ‘I burnt my feet (e.g. on hot sand).’ (MMN: TEN1-2017_002-01: 1092009_1094010)

9.4 Possessive constructions 

 561

Finally, when the possessive phrase occupies a subjunct argument role, the possessor will be cross-referenced with a locational bound pronoun, as in (1441).467 Identical possessive constructions involving subjunct arguments are reported in Jaru by Tsunoda (1981b, 198ff.).

(1441)

Purnu pantarra, yuwayi, nyampa parda=yirlapr wilayi ya-nanta tree foliage yeah what dub=1sg.lct around go-prs ear-hole-rlapm, tajitaji. . . brain-loc plant_sp ‘A tree, with good foliage, yes, this one perhaps is going around in my brain, it is the tajitaji. . . (i.e. the name is going around in my head).’ (Cataldi, 2011, p. 252)

As with object bound pronouns marking possessors, locational bound pronouns marking possessors can also undergo reflexivisation when co-referent with the subject of the clause, as in (1442). In this example, an unexpressed possessor (ngaju-ngka) is cross-referenced by a first person singular locational bound pronoun (=yirla), which in turn has triggered reflexivisation and been replaced by a locational reflexive enclitic (=nyanungkurla). (1442)

Minya=rna=nyanungkurlapr makarta yirra-rnanta langka-ngkapm. this=1sg.s=refl.lct hat place-prs head-loc ‘I put on the hat on top of my head.’ (PJA: TT76_2601: 640186_645787)

Again, this is consistent with reflexivisation processes reported in Jaru (Tsunoda, 1981b, 152ff.). Compare (1442) with the Jaru example in (1443). (1443)

Ngaju-nggupr nga=rna=nyununggula magarda yaan-an langka-ngkapm. 1sg-erg cat=1sg.nom-m-loc hat put-prs head-loc ‘I put a hat on my head.’ [Jaru] (Tsunoda, 1981b, p. 153)

9.4.3 Prominent internal possession Meakins and Nordlinger (2017) draw attention to a particular type of possession construction that is particularly prevalent in Ngumpin-Yapa languages and that demon467 In (1441), since there is registration of a first singular possessor in the bound pronoun, the unexpressed possessor must be ngaju-rla 1sg-loc. The unexpressed free pronoun cannot be of the form ngaju-kuny-ja 1sg-`poss-loc since this would involve internal possession and possessors in internal possession constructions are not cross-referenced in the bound pronoun.

562 

 9 Grammatical relations and argument structure

strates properties of both internal and external possession – a construction they term ‘possessor dissension’ but which has been described in recent typological literature as ‘prominent internal possession’ (Bárány et al., 2019).468 As with internal possession, case marking is used to indicate the possessor. However, these constructions are distinct from internal possession in two ways. First, the grammatical dative case is used to mark the possession relation on the possessor, rather than either of of the non-grammatical suffixes: -kuny poss or -punta pposs . Second, when these possessive phrases appear as arguments in a clause, not only is the entire possessive phrase cross-referenced by a bound pronoun appropriate to its clausal role, but the possessor itself is also registered syntactically in the bound pronoun (as an object/ oblique).469 Examples which include an NP referring to the possessor are provided in (1444) and (1445). However, it is most common for the possessor NP to be elided, leaving the oblique bound pronoun as the only indicator of the possessor, as in (1446) and (1447). (1444)

Minya=yipr=lupp [ngaju-kupr warlaljapm]pp. this=1sg.o=3pl.s 1sg-dat family ‘These are my family/friends of mine’. (MNN: TEN1-2017_018-02: 389577_393467)

(1445)

[Ngaju-kupr=yipr=pulapp wapirra-kujarrapm]pp ya-ni puliki-kurra. 1sg-dat=1sg.o=3du.s F-dual go-pst cattle-all ‘My two fathers went towards the bullocks.’ (TJN: LC06a)

(1446)

Wanji-rla=lupp=nyurrapr [ngatinapm en mamapm]pp.470 where-loc=3pl.s=2pl.o M and F ‘Where are all your mothers and fathers?’ (TJN: LC06)

468 This has also been labelled ‘oblique possession’ in Bilinarra and Gurindji (Meakins & Nordlinger, 2017) and ‘inherent possessor raising’ in Mudburra (Osgarby, 2018, pp. 168–170). 469 Outside of the Ngumpin-Yapa subgroup of Australian languages, this type of external marking of modifiers is somewhat unusual. In the typological literature Corbett flags this type of construction as ‘verb agreement with possessives’ (Corbett, 2006, p. 21), while the syntactic construction itself – beyond just possessive constructions – has been described as a ‘trigger happy agreement’ (Comrie, 2003). 470 The Kukatja kin term mama ‘father’ is used here in place of the expected Ngardi wapirra.

9.4 Possessive constructions 

(1447)

 563

[Mamapm-nguniny]pp=pa=lupp=nganpapr ya-ni-nyurra wirlinyi, F-kposs=ep=3pl.s=1pl.excl.o go-pst-narr hunting kuyi-kurra.471 meat-all ‘Our fathers went hunting for meat.’ (TJN: Manungka_01-035151L: 543793_551299)

In the Ngardi corpus, prominent internal possession appears to be the preferred method for encoding alienable possession involving human possessors. Most of the examples in the corpus involve first person singular possessors but examples of plural possessors (1448) are also found. (1448)

Yuu, ngu=nganpapr wirrpapr-ku. yes seq.aux=1pl.excl.o many-dat ‘Yeah, it’s (the Dreaming) for the whole lot of us.’ (MAM: TEN1-2016_002-01: 276070_278607)

Note that when the entire possession phrase is third singular (as above) and is in an A, S or O role, it will not be realised overtly in the bound pronoun and thus it appears that only the possessor is cross-referenced, as in (1449). (1449)

Ngantu-ngku=yipr pi-nya [ngaju-kupr kunyarrpm]pp? who-erg=1sg.o hit-pst 1sg-dat dog ‘Who killed my dog?’ (PSM: TEN1-2017_009-01: 391674_393787)

This is also true of the verbless clauses in (1450)–(1452). (1450)

Ngantupr-ku=rlapr [minya card]pm? 3sg-dat=3sg.obl this card ‘Whose card is this?’ (Cataldi, 2011, p. 214)

(1451)

[Minyapm=nyurrapr pulikipm nyurrurlapr-ku]pp nawu. this=2pl.o cattle 2pl-dat now ‘These bullocks are all yours now.’ (PSM: TEN-2016_005-03: 658881_663777)

471 This example is the only instance of a prominent internal possession construction involving possessive marking on the possessum (-nguniny). It is unclear whether (the unexpressed) dative marked possessor nganimpa-ku ‘ours’ could also co-occur in a single clause where the possessum is also marked.

564 

 9 Grammatical relations and argument structure

(1452)

Jangu nga=rna=nyurrapr [kurrijipm nyurrurlapr-ku]pp. recog2 seq.aux=1sg.s=2pl.o in_law 2pl-dat ‘Indeed, I am your parent-in-law.’ (PPN: Cataldi 2011, p. 244)

Cross-referencing of the possessor phrase is also morphologically constrained by the fact that the bound pronominal complex prevents the registration of more than two bound pronouns. Thus, in contexts where the possessor phrase is a non-subject argument in a transitive clause, there is competition between the possessor phrase and the (prominent internal) possessor for cross-referencing in the bound pronoun. The corpus, however, only contains examples involving object possessive phrases in which the possessum is third singular and in these contexts the possessor is cross-referenced: (1453)

Nyin=ma-ni=pula=nyanupr [jajapm nyanungu-kupr]pp. lose=get-pst=3du.s=refl MM recog1-dat ‘Those two lost their own grandmother.’ (MMJ: LC03)

If (1453) had involved a plural object (i.e. their grandmothers), there would be competition between registering the possessor phrase (=yanu) or the prominent internal possessor (=nyanu). In Jaru both of these possibilities are realised and the selection has to do with whether or not the ownership is being emphasised (see Tsunoda, 1981b, p. 154). One remaining issue in identifying instances of prominent internal possession is the fact that, in some cases, they are not clearly distinguished from instances of dative case-marked NPs functioning as relational arguments of the clause (i.e. benefactives or indeed malefactives) rather than clear possessive constructions. Consider, for example, (1454) and (1455). (1454)

Wakurra minya, minyapm ma=yanupr mula-ngarna-kupr. neg this this top.aux=3pl.o here-den-dat ‘Not this (country), this country belongs to people from here.’ (MMJ: LC31a: 457225_460673)

(1455)

Kunyarrpm=pa=yipr ngaju-kupr lus=karri-nya. dog=ep=1sg.o 1sg-dat lose=be-pst ‘My dog died (he passed away).’ (TJN, Cataldi, 2011, p. 150)

While the translations reflect possessive constructions, it is difficult to definitively state that the dative is marking adnominal possession rather than marking a relation between the NP and the clausal event. Thus, compare the translations above with the following alternatives: ‘this land is for them’ (1454) and ‘my dog died on me’ (1455). These observations about the marginal usages of the dative as an adnominal case can

9.4 Possessive constructions 

 565

be contrasted with the possessive -kuny, which functions exclusively in internal possession contexts and is never syntactically registered in the bound pronoun. This same ambiguity arises in contexts of embedded possessive phrases in Ngardi (see also §4.2.4.3). Ngardi possesses a morphological constraint against a sequence of a dative case suffix followed by any other case suffix, including a second dative case suffix (*-ku-ku). Thus, the only strategy for embedding more than a single possessive phrase is by a sequence of the adnominal possessive (-kuny) or personal possessive (-punta) followed by the dative (-ku). Again, possessive constructions involving the dative are indistinguishable from alternative readings of the dative as a relational case, marking a benefactive or malefactive relationship to the main clausal predicate. While problematic for the current analysis, this type of functional polysemy (between possessives and benefactives) is not uncommon cross-linguistically (Lichtenberk, 2002; Rapol, 2010).472 (1456)

Yalapm=rlapr [nyuntu-kuny-ku kunyarr-ku]pr kuyipm. that=3sg.obl 2sg-poss-dat dog-dat meat ‘That is your dog’s meat./That meat is for your dog.’ (MNN: TEN1-2017_02401: 94755_99649)

(1457)

Yalapm=rlapr [nyantu-punta-ku]pr. that=3sg.obl recog1-pposs-dat ‘That is his dog’s meat./That (meat) is for his own (dog).’ (MMN: TEN121017_024-01: 183424_186379)

(1458)

Minyapm=rla ngaju-punta-kupr [nyantu story]pm kuja=yi this=3sg.obl 1sg-pposs-dat recog1 story sub=1sg.o Jampijin-tu ngaju-kuny-ju ngarri-rnani warlalja nyupa-ngku. ♂subsection-erg 1sg-poss-erg tell-ipfv.pst family partner-erg ‘This is my husband’s story./This story belongs to my husband, it is the one that my own husband, Jampijin, told me.’ (PPN: Cataldi, 2011, p. 131)

Beyond the types of embedded possessive constructions presented above, it is not possible to have any further levels of possessive embedding in Ngardi (i.e. *[[w’s, x]’s y]’s z).

472 The key observation for (1456)–(1458) is that the possessive relationship in the embedded lower phrase must be encoded adnominally (i.e. with -punta or -kuny) while the possessive relationship in the higher phrase must be encoded by a grammatical case (i.e. with dative -ku) and registered syntactically in the bound pronoun as an argument.

566 

 9 Grammatical relations and argument structure

9.4.4 Adnominal clitic possession There is a marginal construction in Ngardi in which a pronominal clitic is used to index a possessive (adnominal) relationship, separate from – and in addition to – the clause-level bound pronouns. This can be most clearly exemplified by the example in (1459) where two separate bound pronoun complexes occur within a single clause. (1459)

Nga=pula jingkil-pala karri-nyani ngaju-ku=yi. seq.aux=3du.s single-adj be-ipfv.pst 1sg-dat=1sg.o ‘My two parents were single (back then).’ (NGO: TEN1-2017_011-01: 45602_47721)

Additional, unusual examples exist in which an NP co-referent with the implied possessor is registered as an unmarked and topicalised argument of the clause, as in (1460) and (1461). Note that these two examples are somewhat similar to external possession in that the nouns marking the possessor and possessum are placed in simple apposition. However, the possessor is cross-referenced in the pronominal complex as an object/oblique, despite filling the subject role of the clause. In regular external possession, the possessor would be cross referenced with the subject bound pronoun =rna. (1460)

Ngajupr=yipr ngamapm 1sg=1sg.o m ‘My mother ran.’ (TJN: TJAMA6)

(1461)

[Ngaju pr ma=yipr kunyarrpm]pp 1sg top.aux =1sg.o dog ‘My dog died.’ (JLI: TT76_2201: 782815_784951)

yaparti-nya. run-pst

nyurnu=karri-nya. die=be-pst

The presence of pronominal clitics being used with adnominal functions is also reported in Mudburra (Osgarby, 2018, p. 93) and at least some Western Desert languages (e.g. Warnman: Burgman, 2003, p. 35) but is otherwise not widely documented for Australian languages.

10 Auxiliaries, particles and modal constructions The focus of this chapter is the description of the class of uninflecting lexemes in Ngardi: the auxiliaries, the modal and propositional particles, and enclitics. Since these play a central role in the expression of clausal modality, various modal constructions are also detailed in this chapter. The chapter begins by examining word order and second position phenomena (§6.2). The classes of uninflecting lexemes are then described in detail, including auxiliaries (§10.1), modal particles (§10.2), propositional particles (§10.3), propositional and temporal enclitics (§10.4), interjections (§10.5), and Kriol-origin particles (§10.6). This chapter also deals with select modal constructions including commands (imperatives, directives, prohibitives, and admonitives (§10.7)), negation (§10.8), and interrogation (§10.9).

10.1 Auxiliaries Auxiliaries473 in Ngardi are monosyllabic, uninflecting particles which serve as phonological hosts for the bound pronouns. Unlike other particle types, auxiliaries in Ngardi (when present) obligatorily host pronominal clitics. Ngardi possesses three auxiliaries, as listed in Table 127: sequential ngu/nga §10.1.1, topic ma §10.1.2, and emphatic ka §10.1.3. Auxiliaries lack syntactic functions, and are only distinguished from one another by some distributional tendencies and some secondary discourse-semantic functions from which their labels are derived. Table 127: The auxiliaries. Form

Label

Syntactic function

Discourse functions

ngu, nga ma ka

seq.aux top.aux emph.aux

— — —

Sequential marker Topic marker Emphasis marker

473 I use the term ‘auxiliary’ to refer to the independent, uninflecting root to which bound pronouns attach. This usage can be found in grammatical descriptions of Ngumpin languages (see Senge, 2015 and Meakins & Nordlinger, 2014). In Warlpirist literature however, such a ‘root’ is typically referred to as an ‘auxiliary base’ and the entire complex of ‘base + bound pronouns’ is termed ‘the auxiliary’ (e.g. Nash, 1986). ‘Catalyst’ is another term that has been used by some authors for equivalent parts of speech in other Ngumpin-Yapa languages (e.g. Jaru: Tsunoda, 1981b, p. 124; Mudburra: Capell, 1956, pp. 68–69; and Gurindji: McConvell, 1996b, p. 34). https://doi.org/10.1515/9783110752434-010

568 

 10 Auxiliaries, particles and modal constructions

Auxiliaries in Ngardi are grammatically optional. Their presence or absence has no bearing on the interpretation of tense, aspect or mood of a clause.474 Tense, aspect and mood distinctions are encoded by verbal inflections (§7.4) and modal particles (§10.2). In a study of 1000 lines of Ngardi utterances of naturalistic conversation, Honeyman (2005) found that an overt auxiliary was utilised only 10 percent of the time. While Honeyman’s study was not necessarily representative of all speakers (e.g.  another sample of 200 utterances of a single male speaker recorded in 2015 showed auxiliaries used in 20 percent of clauses), it is indicative of the optionality of the auxiliary in Ngardi. In the absence of an auxiliary, the bound pronouns encliticise to the preceding word. The optionality of auxiliaries in Ngardi and the positioning of the bound pronouns in each context is exemplified by the minimal pair (1462) and (1463) for ma and the minimal set (1464), (1465) and (1466) for ngu. The emphatic auxiliary ka is rare in the corpus and wholly absent from the speech of some speakers. (1462) Ngaju=yi=n purda=nya-nganta. 1sg=1sg.o=2sg.s listen=see-prs ‘You are listening to me.’ (JLI: TT76_2201: 637505_639571) (1463) Ngaju ma=yi=n purda=nya-nganta. 1sg top.aux=1sg.o=2sg.s listen=see-prs ‘You are listening to me.’ (JLI: TT76_2201: 630324_632733) (1464) Mapa-ku=rlipa=nyanu. rub-pot=1pl.incl.s=refl ‘We will rub ourselves (with ochre).’ (SCM: TT76_1301: 467834_470150) (1465) Mapa-ku ngu=rlipa=nyanu. rub-pot seq.aux=1pl.incl.s=refl ‘We will rub ourselves (with ochre).’ (SCM: TT76_1301: 473738_475926) (1466) Ngu=rlipa=nyanu rabim=ma-nku. seq.aux=1pl.incl.s=refl rub.tr=get-pot ‘We will rub ourselves (with ochre).’ (SCM: TT76_1301: 486927_488680) The optionality of auxiliaries distinguishes them from both modal and propositional particles whose (non)occurrence has meaningful semantic content. Distinctions between auxiliaries and modal and propositional particles are further reinforced by their distri474 Unlike Ngardi, the presence or absence of auxiliaries in Walmajarri (Ngumpin) and Warlpiri (Yapa) is grammatically meaningful for the interpretation of clausal TAM. For example, pa and ngu in Walmajarri are used (amongst other functions) to distinguish indicative and interrogative moods (Hudson, 1978, p. 78).

10.1 Auxiliaries 

 569

butional properties. Auxiliaries, when they do appear, obligatorily host the bound pronouns. Propositional and modal particles can appear in clauses where they do not host the bound pronouns and, in addition, can have phrasal scope over an adjacent constituent. Auxiliaries cannot co-occurr with propositional or modal particles with clausal scope but can co-occur with the clausal negator wakurra.475 See, for example, (1467) and compare the optional presence of ma in the negative clauses in (1468) and (1469). (1467)

Wakurra ngu=lu paja-ku waku. neg seq.aux=3pl.s bite-pot neg ‘They won’t bite, nothing.’ (PSM: TEN1-2018_038-01: 869750_871578)

(1468) Wakurra ma=rna Walmajarri. neg top.aux=1sg.s language_name ‘I am not Walmajarri.’ (MMN: TEN1-2018_030-01: 782539_786950) (1469) Ngaju wakurra=rna Walmajarri. 1sg neg=1sg.s language_name ‘I am not Walmajarri.’ (MMN: TEN1-2018_030-01: 771857_776734) A final feature of the auxiliaries is that they may appear even when the bound pronoun is non-overt (i.e. third singular subject), as in (1470)–(1472).476 (1470)

Jarranti-nyani minya-rla ma payilpayi=luwa-rni, kankani-mpala. stay-ipfv.pst this-loc top.aux getting_close=shoot-pst up-along ‘He stopped where he thought he had it but it was up on the top.’ (Cataldi, 2011, p. 261)

(1471)

Nga karrmantawangu karri-nya-ngurra. seq.aux hungry be-pst-narr ‘So he stayed there, hungry.’ (TJN: LC05)

(1472)

Wakurra, wurna ka ya-ni Kumunjayi Pring-kurra. neg journey emph.aux go-pst place_name-all ‘No, she did not go to Alice Springs.’ (TJN: LC32b)

The auxiliares can be partially distinguished from one another by some positional tendencies and additional semantic and discourse-related roles they may serve. 475 Auxiliaries can co-occur with propositional or modal particles when the latter are functioning with phrasal scope. 476 The ability for auxiliaries to appear independently means they disobey the otherwise strict disyllabic minimum word constraint in Ngardi.

570 

 10 Auxiliaries, particles and modal constructions

Namely, ngu ~ nga appears frequently in a secondary clause in closely coordinated, paratactic clauses with sequential interpreations, while ma appears to have topic-marking functions in some contexts (perhaps at some stage of diachronic change). These will be exemplified in the following subsections. 10.1.1 Sequential auxiliary ngu ~ nga The sequential auxiliary ngu ~ nga appears in both simple and complex clause contexts. The appearance of ngu ~ nga in these contexts correlates – but is co-extensive – with its clausal positioning as presented in Table 128. Table 128: Simple and complex clause functions of the sequential auxiliary ngu ~ nga. Simple clause (§10.1.1.2)

Complex clause (§10.1.1.3)

Function

Null

Sequential (temporal or logical)

Default Position

Second position

First position

In simple clauses, ngu ~ nga serves as a semantically neutral phonological host for the pronominal clitics and may be freely elided from a clause. It can appear in a range of clausal positions; however, second position appears to be ‘default’. In multi-clausal contexts, the sequential auxiliary functions somewhat similarly to a coordinative or a complementiser (see similar comments by Honeyman (2005, pp. 55–57)). Semantically it encodes a sequential relationship (either logical or temporal) between two coordinated clauses. The default clausal position for the auxiliary in this function is clause-initial. The simple (§10.1.1.2) and complex (§10.1.1.3) clause functions of ngu ~ nga will be detailed and exemplified in turn but first, the variation in the form of the auxiliary is described in §10.1.1.1. 10.1.1.1 Allomorphic variation In the Ngardi corpus, there is intra-speaker variation between ngu and nga as the preferred form of this auxiliary. In (1473), for instance, Popeye Jangala equivocates on his word choice and switches (within the one clause) between using ngu and nga with no implied difference in meaning: (1473) Nga=n=janu, nga=n=janu–– nya-nganta. seq.aux=2sg.s=3pl.o seq.aux=2sg.s=3pl.o see-prs ngu=n=janu nya-nganta ngu=n=janu. seq.aux=2sg.s=3pl.o see-prs seq.aux=2sg.s=3pl.o ‘You are watching them.’ (PJA: TT76_2601: 1191518_1196980).

10.1 Auxiliaries 

 571

In neighbouring Ngumpin languages, the variation between cognate auxiliaries ngu and nga is either argued to be phonological (e.g. Wanyjirra: Senge, 2015, p. 485) or semantic/grammatical (e.g. Walmajarri: Hudson, 1978, p. 58).477 Perhaps due to the fact that Jaru possesses only the auxiliary nga, both Marie Mudgedell and Patrick Smith assert that nga is a Jaru word and ngu is a Ngardi word. Nevertheless, both Marie and Jupiter use both ngu and nga in their Ngardi speech. For instance, in a sample of seven narrative texts Jupiter produced 21 tokens of nga to 17 tokens of ngu, while in a sample of six narratives Marie produced 3 tokens of nga to 23 tokens of ngu. In the absence of any further disambiguating evidence they are simply treated as forms in free variation and both glossed seq.aux. 10.1.1.2 Simple clause usage of ngu ~ nga In simple clauses, ngu ~ nga serves as a semantically neutral host for bound pronouns, as in (1474) and (1475).478 There is a tendency for the auxiliary to occupy second position in simple clauses. (1474)

Nyunpula-rlu ngu=nta=ngajarra nya-nganta kuriny-kariny-ju, 2du-erg seq.aux=2pl.s=1du.excl.o see-prs two-other-erg nyunpula-rlu. 2du-erg ‘You two are looking at us two, you two are.’ (TJL: TT76_2601: 1657081_1662160)

(1475)

Ngaju nga=rna wangka-nyanta Ngardi. 1sg seq.aux=1sg.s speak-prs language_name ‘I am speaking Ngardi.’ (SCM: TT76_1301: 74689_76970)

Other analyses of Ngumpin languages have treated a similar auxiliary (ngu ~ nga) as a marker of ‘realis’ mood (Wanyjirra: Senge, 2015, p. 485) or a marker of ‘declarative’ constructions (and see Capell’s (1956) analysis of a declarative auxiliary ba in Mudburra). For each of these languages, such an auxilliary can nevertheless occur in non-realis clauses and so it is necessary to propose that the auxiliary can be ‘neu-

477 In Wanyjirra, the form nga ‘occurs before pronominal clitics beginning with /l/ whereas the form ngu occurs before any pronominal clitics’ (Senge, 2015, p. 485). In Walmajarri, ngu is an (irregular) allomorph of pa (along with ma) and contrasts both with the grammatically distinct auxiliary nga (with allomorphs ngarta and ∅) which marks interrogative (or admonitive) mood and with the absence of an auxiliary which marks imperative mood, among other modalities (see Hudson, 1978, 58ff.). 478 The cognate ‘catalyst’ forms in Jaru and Gurindji are both described as semantically ‘meaningless’ host elements.

572 

 10 Auxiliaries, particles and modal constructions

tralised’ for modality in certain contexts.479 Other authors treat the ‘declarative’ as something of a ‘default value’ assigned to the auxiliary (e.g. Osgarby (2018, p. 103) for Mudburra; McConvell (1996b, p. 94) for Gurindji).480 For Ngardi, I treat all simple clause usages of ngu ~ nga as semantically null. Modality is determined separately via the presence or absence of certain modal particles and inflections on the verb and not the presence or absence of any of the auxiliaries per se. Examples illustrating the range of clausal modalities in which ngu ~ nga is found are provided below. Note that in clauses exhibiting irrealis verbal inflections (1476), the ngu ~ nga auxiliary has no effect on either epistemic or deontic interpretations available to the verbal inflection. Examples of ngu ~ nga appearing in clauses with ‘future’/intentional and hortative meanings are provided in (1477) and (1478) respectively. (1476)

Ngu=rlipa ya-nngi. seq.aux=1pl.incl.s go-irr ‘We should go.’ (Epistemic) ‘We should have gone.’ (Epistemic)481 ‘We might go.’ (Deontic) (PSM: TEN1-2018-01: 2348472_2352558)

(1477)

a.

b.

(1478)

“Karri-ju=rlipa yangi-ngka ngurra-ngka.” be-pot=1pl.incl.s one-loc camp-loc ‘(The boss said) “We will all stay in the one camp.”’ “Wakurra, ngu=rna karri-ju ngaju-kariny.” neg seq.aux=1sg.s be-pot 1sg-other ‘(I responded) “No! I will stay by myself.”’ (TJN: LC5)

Ngu=rlipa ya-nku, wurna=wali. seq.aux=1pl.incl.s go-pot journey=then ‘Let’s go, (let’s) set off now.’ (PSM: TEN1-2017_008-01: 2147058_2149414)

As an auxiliary, ngu ~ nga is generally incompatible with modal particles operating with clausal scope (i.e. parda, ngarda, mayi, wayi), with the exception of the Kriol loan maitbi as in (1479). It can – albeit rarely – occur after negative (1480), and propositional (1481) particles.

479 See e.g. Wanyjirra, where ngu ~ nga can optionally surface in irrealis mood instead of the irrealis auxiliary with no obvious change in meaning (Senge, 2015, p. 418). 480 In Mudburra, the presence or absence of ba is meaningful in disambiguating imperative and potential usages of an otherwise syncretic inflection, however. 481 The irrealis is not constrained to ‘prior’ temporal reference. However, clear examples of ngu + IRR with past time reference are not observed in the data.

10.1 Auxiliaries 

 573

(1479)

Maitbi ngu=n wirrirl parda ya-nku yalu-jangka. maybe seq.aux=2sg.s back dub go-pot that-ela ‘Perhaps you will go back to Halls Creek from there.’ (TT75_1301: 1478700_1480488)

(1480)

Wakurra ngu=lu paja-ku waku. neg seq.axu=3pl.s bite-pot neg ‘They won’t bite, not at all.’ (PSM: TEN1-2018_038-01: 869750_871578)

(1481)

Kulanganta nga=lu la-nanta nganayi-piya carrot-piya. cf seq.aux=pl.s pierce-prs whatsit-sembl carrot-sembl ‘They were supposedly digging up, like a, you know, like a carrot.’ (PRN: FT1_a: 1394751_1398443)

Both ngu and nga can surface without any overt pronominal arguments, as is the case when subjects (and objects) are 3sg, as in (1482) and (1483). (1482)

Ngaringka-rlu ngu kupa-rnanta mangarri. woman-erg seq.aux cook-prs veg_food ‘A woman is cooking.’ (PA: TT76_2601: 988910_993746)

(1483)

Kit=pa-ni nga luka-ngku motika. stuck=get-pst seq.aux mud-erg car ‘The mud trapped the car.’ (MMN: TEN1-2017_002-04: 74536_80946)

10.1.1.3 Complex clause usage of ngu ~ nga One discernible function of the otherwise semantically null auxiliary ngu ~ nga is to link multiple clauses in a sequential relationship either of a temporal (1484) or purposive (1485) nature, hence the gloss ‘sequential auxiliary’. In these contexts, ngu ~ nga is always clause-initial and the ordering of clauses is strictly iconic of the temporal/ logical ordering of events – a feature consistent with coordinate structures in many of the world’s languages (Diessel, 2001). This function is described further in §11.2.2. (1484)

Kupa-rni=pula, nga=pula ka-nya-ngurra. cook-pst=3du.s seq.aux=3du.s carry-pst-narr ‘They cooked it, then they took it.’ (TJN: LC21a: 881514_884880)

(1485)

Yunpa-ka=rla jingka-ku ngu pukany=karri-ju. sing-imp=3sg.obl child-dat seq.aux sleep=be-pot ‘Sing to the child so he can go to sleep.’ (MMN: TEN1-2017_002-03: 1746418_1757907)

574 

 10 Auxiliaries, particles and modal constructions

10.1.2 Topic auxiliary ma The topic auxiliary ma, like ngu ~ nga, also serves as a semantically null morphological host for bound pronouns. Its optional grammatical status is exemplified by (1486) and (1487). (1486)

Wakurra ma=rna Walmajarri-kurlu. neg top.aux=1sg.s language_name-prop ‘I am not Walmajarri./I am not a Walmajarri speaker.’ (MMN: TEN1-2018_ 030-01: 802461_806130)

(1487)

Ngaju wakurra=rna Walmajarri. 1sg neg=1sg.s language_name ‘Me, I’m not a Walmajarri (person).’ (MMN: TEN1-2018_030-01: 771857_776734)

The topic auxiliary may either appear after an initial constituent as in (1486) or it may be clause-initial as in (1488). (1488)

Ma=rna walyka=karri-ju. top.aux=1sg.s cool=be-pot ‘I will cool down.’ (PSM: TEN1-2018_037-01: 1197364_1198578)

The ma auxiliary has previously been described as a ‘negative auxiliary’ (Cataldi, 2011) or a general ‘irrealis’ (Honeyman, 2005 and Ennever, 2018). The full range of clauses in which it appears are, however, broader than these labels suggest. The label ‘topic auxiliary’ has been adopted since there is evidence it may enhance the topical (or sometimes focal) status of an initial constituent (see Honeyman, 2005) in some contexts. The topic auxiliary can optionally follow the clausal negator wakurra (1486) and (1487), a propositional particle such as assertive kala (1489), and can also appear in clauses involving tense (e.g. present (1490)) and modal (e.g. potential (1493) and (1494)) inflections. (1489)

Kala ma=pula nyanungu-kariny-ja, kuja jaja=yi asrt top.aux=3du.s recog1-other-loc sub MM=emph nyin=ma-ni, nyanungu-ku=pulanya. lose=get-pst recog1-dat=3du.o ‘Yeah, those two will stay with that other one, that one who has (also) lost (and is bereaved of) their grandmother.’ (MMN: Manungka_01-035155: 450313_454434)

10.1 Auxiliaries 

 575

(1490)

Ya-nku=rlipa ngampurr ma=rlipa wirilyi=pu-nganta. go-pot=1pl.incl.s several top.aux=1pl.incl.s hunting=hit-prs ‘We (want) to go, so that we will all be hunting (together).’ (MDN: LC42a: 215553_218931)

(1491)

Wirriri=ya-nku ma=rli kulirra ngurrara-kurra. back=go-pot top.aux=1du.incl.s south country-all ‘You and I will go back to our country.’ (MMJ: LC31: 613440_616905)

There are some select contexts in which the topic auxiliary ma- appears with particular frequency – and to the exclusion of the sequential auxiliary. First, in verbless clauses ma appears to be the auxiliary of choice, as in (1492) (and see (1486) and (1487) above). The sequential auxiliary is rarely observed in this context. (1492)

Wirrpa ma minya puliki. many top.aux this cattle ‘There were lots of cattle.’ (PSM: TEN1-2018_038-01: 972217_975114)

Second, ma is also the preferred auxiliary of choice following a negative interjection waku as in (1493) or clausal negator wakurra in (1494). The sequential auxiliary is rarely observed in this context (one example in elicitation only). (1493)

a.

b.

(1494)

“Ya-nku=n timana-rla, kamul-rla?” go-pot=2sg.s horse-loc camel-loc ‘“Will you go by horse, by camel?”’ “Wakurra, waku ma=rna ngaju ya-nku=rna walarnu.” neg neg top.aux=1sg.s 1sg go-pot=1sg.s on_foot ‘“No, not I, I will go by foot.”’ (MDN: LC22_b: 877440_879996)

Waku=rnalu karri-nyanta. neg=1sg.s be-prs ‘We are not staying.’ Waku, wurna ma=rnalu wirriri=ya-nku. neg journey top.aux=1pl.excl.s back=go-pot ‘No, we will journey back instead.’ (MMJ: LC31_a: 633222_640064)

10.1.2.1 Topic marking function In addition to its role as a generic pronominal host, there is some evidence that ma also has a discourse-related function as a topic marker. The clearest examples of this are instances of ma in declarative clauses sequenced after demonstratives, either in cases where the demonstrative is used to refer to a real world referent (e.g. minya ma

576 

 10 Auxiliaries, particles and modal constructions

in (1495)), or in a metalinguistic usage as a response to elicitation, (e.g. yala ma in (1496)).482 Many of the examples of ma in verbless clauses described in the preceding section could also be viewed as ‘topic-marking’ in function. (1495)

Minya ma jukurr-jangka kuyi=lku. this top.aux Dreaming-ela meat=then ‘This meat is from the Dreamtime.’ (MMJ: LC31: 829949_832736)

(1496)

Yala ma pilypin=pu-nganta. that top.aux scrape=hit-prs ‘That one is pilypinpunganta.’ (TJL: TT76_2501: 164327_169331) [Prompt: Any word like scraping? Like a bullock skin?]

More broadly, this contextual usage of a form ma bears many similarities to a clitic which has been labelled a ‘topic’ marker in many other Ngumpin-Yapa languages (e.g. Bilinarra: Meakins & Nordlinger, 2014, pp. 391–397; Gurindji: McConvell, 1996b, p. 98; Mudburra: Osgarby, 2018, pp. 184–186; Wanyjirra: Senge, 2015, p. 512; and Warlmanpa: Browne, 2021, pp. 340–343). For Gurindji and Mudburra, McConvell (1996c, p. 301) suggests that ‘pronominal encliticisation to an initial constituent with a “topic suffix” -ma marks double focus or contrast’, and furthermore, that ‘such a discourse function combines features of “topic switch” and contrastive focus’. What makes identifying ma as a unique topic marker difficult is the fact that many have argued that it is ‘the initial position’ itself (or the the position leftwards of the bound pronouns) in Ngumpin-Yapa languages which is argued to convey focus (Austin & Bresnan, 1996; Laughren, 2002) or which Simpson (2007) argues to be analysable in terms of ‘relative prominence’ (focus) as opposed to relative newness (topic). It is therefore difficult to assert that the encoding of discourse functions like ‘topic’ are strictly derived from a constituent’s position or by the auxiliary itself. For Ngardi, I have been unable to convincingly and consistently distinguish all uses of ma as a null auxiliary from a ‘topic-marking’ clitic and so I have simply glossed all instances of ma as a putative ‘topic auxiliary’ top.aux. Ngardi ma is clearly distinguished from ma in other languages in that it is clearly not a suffix and can freely appear utterance-initial. It is possible that further prosodic research might distinguish different usages of ma in Ngardi clauses. It is worth noting that there is significant variation as to whether ma is phonologically and prosodically integrated with the preceding word. Some tokens of non-initial ma form prosodic words with

482 The same kind of construction has been observed in dictionary-elicitation contexts in other Ngumpin-Yapa languages and has been referred to as a ‘topic-setting’ construction by Meakins and Nordlinger (2014, pp. 181–182, 260) or ‘contrastive initial attachment’ (McConvell, 1996b, p. 98).

10.1 Auxiliaries 

 577

preceding elements while others are prosodically separated, even sometimes with significant pauses, as in Figure 24, the waveform for (1497). (1497)

Jarany ma=rna kupa-rni. roughtail top.aux=1sg.s cook-pst ‘I cooked the roughtail lizard.’ (TJL: TT76_2201: 1951914_1954294)

Figure 24: Waveform of (1497) demonstrating independent ma.

10.1.3 Emphatic auxiliary ka The ‘emphatic auxiliary’ ka is poorly understood. Ngardi speakers with whom I worked identified it as a Warlpiri form, yet it was used liberally within Ngardi speech spoken by Ngardi/Warlpiri speakers in the Cataldi corpus.483 Its auxiliary status is affirmed by its monosyllabicity, its clausal positioning and the fact that it obligatorily hosts the bound pronouns. Cataldi (2011, p. 96) describes ka as having a propositional value: ‘right, especially, this one’. Its function in some contexts appears similar to the ‘topic-setting’ usages of ma. (1498)

Kurrwa ka=rnalu ngarri-rnanta stone_axe emph.aux=1pl.excl.s tell-prs ‘We call it kurrwa.’ (YMN: LC40a: 692563_694171)

483 The Warlpiri auxiliary ka= interacts with verbal TAM to index imperfective aspect in conjunction with nonpast verb forms (in past forms lpa= is used) (Hale et al., 2015 p. 1678; Nash, 1986, p. 229).

578 

 10 Auxiliaries, particles and modal constructions

Honeyman (2005) contends that ka in Ngardi marks indicative mood; however, it is not an obligatory clausal element and there are examples of it also occurring in interrogative clauses.484 Like the other auxiliaries, it can appear either clauseinitially (1499) or in second position (1500), although the latter is by far the more frequent (as  for the topic auxiliary ma). It is found in clauses containing verbs inflected for a range of tense/aspect categories and is compatible with negative clauses (1501). (1499)

Ka=rna=yanu kupa-rni-nyirra stockman-ku ngurra-ngka-ku. emph.aux=1sg.s=3pl.o cook-pst-narr stockman-dat camp-loc-dat ‘I used to cook for the stockman in the camp.’ (YMN: LC23b: 149262_155066)

(1500)

Luk=karri-nya ka=rnalu. lie_down=be-pst emph.aux=1pl.excl.s ‘We lay down.’ (YDN: LC24a: 238989_240713)

(1501)

Wakurra ka=rla yi-nya. neg emph.aux=3sg.obl give-pst ‘He did not give her to him.’ (MDN: LC42b: 1068063_1069935)

10.2 Modality and modal particles Modality in Ngardi is encoded both through modal inflections appearing as suffixes on the verb and via independent and uninflecting modal ‘particles’. Within the verbal system, Ngardi possesses six inflections which have modal functions: the imperative, potential, irrealis, anticipatory, hortative, and obligative. These were introduced and detailed in §7.4. Table 129 sumarises the full range of modal meanings available to each of the six inflections. As can be observed in Table 129, the potential and irrealis cover the widest array of modal meanings in Ngardi. A signficant feature shared by both these inflections is that they both allow for epistemic and deontic readings as illustrated in (1502) and (1503). That is to say, in Ngardi, notions of ‘might x’ are generally not distinguished from notions of ‘should x’.

484 An example is (543).

 579

10.2 Modality and modal particles 

Table 129: Modal inflections and their meaning categories. Potential Irrealis

Imperative

Anticipatory

Hortative

Obligative

(pfv)

-Cu

-Nngi

-Ca

(NIL)

-Cura

-Cumarda

ipfv

-Nanku

-Nanngi

—

-Nanngirda

-Nankura

—

Temporal Reference

fut

—

—

—

fut

past

Meaning Uncertainty Categories Desire Intention Hypothetical

✓

✓

—

✓

—

✓

✓

Form

Capacity Expectation Command

✓ ✓ ✓ ✓ ✓

Advisability/ ✓ Obligation Admonition ✓ Hortation

✓

✓

✓ ✓

✓

✓

✓

✓

?

✓

(1502)

Kuyi ma=rna ka-ngku. meat top.aux=1sg.s carry-pot ‘I might take the meat.’ (Epistemic reading) ‘I should take the meat.’ (Deontic reading) (PSM: TEN1-2018_028-01: 4520937_4525912)

(1503)

Kuyi ma=rna ka-ngi. meat top.aux=1sg.s carry-irr ‘I might take the meat./I might have taken the meat.’   (Epistemic reading) ‘I should take the meat./I should have taken the meat.’   (Deontic reading) (PSM: TEN1-2018_028-01: 4510575_4512049)

The pair of examples above also illustrate one major difference between the potential and irrealis inflections. While both can be used for expressing epistemic uncertainty involving events sequenced after reference time, it is only the irrealis that can be used for expressing epistemic uncertainty involving events prior to reference time. Thus, in (1504), the relation between event time and reference time is underspecified by the irrealis inflection, and so two temporal readings are available depending on context.

580 

 10 Auxiliaries, particles and modal constructions

(1504)

Nga=rna=ngku pi-ngi. seq.aux=1sg.s=2sg.o hit-irr ‘I might beat you.’ (After reference time) ‘I might have beaten you.’ (Prior reference time) (TJN: LC56b)

In addition to the verbal inflections, Ngardi possesses a small closed class of four uninflecting particles (mayi, wayi, parda, ngarda) which, amony other functions, encode modal meanings. All four modal particles in Ngardi are polyfunctional and display a signifcant degree of functional overlap with one another. Three modal markers (excluding ngarda) share the ability to modify the epistemic possibility/certainty associated with a clause or a phrase. Furthermore, all modal particles have additional roles; variously functioning as question words, apprehensive markers (in a coordinated biclausal construction), interjections, or intensifiers. The full range of morphosyntactic roles of the modal markers are summarised in Table 130.485 Note that the negative particles wakurra and kula are also classified as modal particles by word class but are discussed separately in §10.8. Table 130: Morphosyntactic roles of modal particles. Clausal

Biclausal

Extra-clausal

Phrasal

Role

Modal marker

Question word

Coordinator

Tag/ Interjection

Modal marker

Ignorative intensifier

Status

Word

Word

Word

Word

Word/ Clitic

Clitic

mayi wayi parda ngarda

✓ ✓ ✓ ✓

✓ ✓

✓

✓ ✓ ✓

✓ (✓) ✓

(✓)

✓

Many of the discrete morphosyntactic roles listed in Table 130 can be exemplified with mayi as in (1505)–(1509). Each function has slightly distinct distributional properties which will be discussed further in the relevant section for each modal particle: mayi (§10.2.1), wayi (§10.2.2), parda (§10.2.3) and ngarda (§10.2.4).

485 As will be discussed in the subsections that follow, not all modal particles are used consistently by all speakers in the corpus. The significant functional overlap between different forms is possibly an artefact of their dialectal or idiolectal basis.

10.2 Modality and modal particles 

 581

[i] Clausal: Modal marker (1505) Paja-ku mayi ngaju. bite-pot uncert 1sg ‘He might bite me.’ (MMN: TEN1-2018_021-02: 1011805_1014731) [ii] Clausal: Question word (1506) Mayi=n marda-rnanta? uncert=2sg.s have-prs ‘Do you have any (food)?’ (MMN: TEN1-2017_026-02: 929061_931035) [iii] Extra-clausal: Tag word (1507) Minya, mayi? this uncert ‘This one, yeah?’ (YMN: LC56b) [iv] Phrasal: Mood marker (local scope) (1508) a. Yalu-kurra wanji-kurra kuja=rnalu ya-nani-nyirra? that-all where-all sub=1pl.excl.s go-ipfv.pst-narr ‘What place was it that we used to go?’ b. Mayi kayirra, nganayi-kurra Jirrmin=pa=rnalu uncert north whatsit-all place_name=ep=1pl.excl.s ya-nani-nyirra. go-ipfv.pst-narr ‘It was north, to you know, Jirrmin (that) we would go.’ (TJN: Manungka_01-035166: 514692_520009) [v] Phrasal: Ignorative intensifier (a special instance of [iv]) (1509) Yirra-ka=rla nyampa=mayi, teabag. place-imp=3sg.obl what=uncert teabag ‘Put some something in it for him, that teabag.’ (PSM: TEN1-2017_027-01: 1018491_1031948) The extension of markers of modality to question/ignorative functions in particular is found in other Ngumpin-Yapa languages (e.g. Warlpiri: Laughren, 1982a) and is reported to be a cross-linguistically common polyfunctionality (Palmer, 2001, p. 172). In Ngardi, mayi and wayi can function as question words, although only mayi occurs extra-clausally as a tag question. The modal particles parda and ngarda never have this function. Two of the modal particles (mayi, parda) are further interesting for frequently encliticising directly to ignoratives either in their nominal or interrogative senses and function as intensifiers of epistemic uncertainty. Lastly, ngarda (§10.2.4) is

582 

 10 Auxiliaries, particles and modal constructions

distinct from the other modal particles in having a unique apprehensive function in certain biclausal constructions. In their clausal functions, the modal markers typically mark an event as being situated somewhere on a scale of possibility (an epistemic function) and encode modalities sometimes termed ‘speculative’ (Palmer, 2001, p. 25) or ‘dubitative’ (Bybee, 1985, p. 179). However, a subset is also compatible with deontic modal meanings typically centred around intentionality and obligation or in a restricted case, apprehension. These are compared in Table 131. Table 131: Clausal functions of modal markers. Modality

Meaning Category

ngarda

parda

mayi

wayi

Epistemic Deontic

Possibility Obligation Intention Apprehension

✓ — — ✓

✓ — ✓ —

✓ ✓ ✓ —

✓ ✓ ✓ —

All modals co-occur with at least some tense inflections and at least some modal inflections. The observed patterns of co-occurrence are presented in Table 132. Some of the gaps in Table 132 are likely an artefact of the available data – especially since some inflections (such as the anticipatory) are infrequent in the corpus. Of all the modals, wayi was by far the most infrequent as a clause-level marker of mood while ngarda was found in combination of a wide number of TAM inflections and was a frequent clause-level marker of modality. Having summarised the distributional and functional features of the modal particles, I now exemplify them in turn.

10.2.1 Uncertain mayi As a marker of epistemic uncertainty, the particle mayi can have scope over a clause (§10.2.1.1) or a phrase (§10.2.1.2). The modal mayi also has interrogative functions (§10.2.1.3) as a question word or tag.486

486 Interestingly, a particle mayi is found in a number of other Ngumpin-Yapa languages, but with highly diversified functions. In Wanyjirra mayi is argued to be the sole member of a class of ‘prepositional particles’ and introduces an NP in a ‘like’ or ‘semblative’ relation (Senge, 2015, pp. 496–497). In Warlpiri, mayi only occurs in second position of the clause, hosts pronominal clitics and has an evidential function expressing a presumption ‘presumably x’ or ‘must be x’. Both of these usages of mayi in related languages are distinct from the modal and interrogative functions of mayi in Ngardi.

10.2 Modality and modal particles 

 583

Table 132: Modal particles and attested TAM combinations. Inflectional category

Meaning category

mayi (§10.2.1)

wayi (§10.2.2)

parda (§10.2.3)

ngarda (§10.2.4)

Tense

prs pst ipfv.pst pst-narr ipfv.pst-narr hab

✓ ✓

✓

✓ ✓

✓ ✓

pot irr ipfv.irr anticip hort ipfv.hort

✓

Modal

✓

✓ ✓

✓ ✓

✓ ✓ ✓

✓ ✓ ✓ ✓ ✓ ✓

10.2.1.1 Clausal scope As a modal marker with clausal scope, mayi encodes dubitative modality – that is, it encodes a speaker’s uncertainty towards the clausal event. When exhibiting clausal scope, it must host any overt bound pronouns (1511) but is otherwise not constrained to a strict clausal position. Second position placement is the most frequent but not obligatory. Thus, compare the positioning of mayi in (1510) and (1511). (1510)

Mayi yunpa-rni, jirrkanka=karri-nya=wali. uncert sing-pst heal=be-pst=then ‘Perhaps he sang him, he is healed now.’487 ‘It could be he sang him, he is healed now.’ (PSM: TEN1-2018_019-1: 2853426_2856677)

(1511)

Pila=ma-nani-nyirra mayi=yanu. follow=get-ipfv.pst-narr uncert=3pl.o ‘He might have been following them.’ ‘He could have been following them.’ (DMN: LC42b: 471891_474419)

When exhibiting clausal scope, mayi combines with a range of TAM inflections, including the (imperfective) past (1510), present (1512), and imperfective past narrative (1513) inflections.

487 Here, ‘sang’ refers to a widespread custom of directing a chant or song at someone, with the intention of healing or causing harm or death.

584 

 10 Auxiliaries, particles and modal constructions

(1512)

TEN:

(1513)

Pila=ma-nani-nyirra mayi=yanu. follow=get-ipfv.pst-narr uncert=3pl.o ‘He might have been following them.’ (DMN: LC42b: 471891_474419)

Yangi-kariny, ngayirr, ngayirr? one-other breathe breathe ‘And another word ngayirr, ngayirr?’ [Prompt] MMN: Kunyarr mayi yala ngayirr-ngayirr=ma-nanta, yapart-jangka. dog uncert that breathe-rdp=get-prs run-ela ‘A dog might be breathing heavily, from running.’ (MMN: TEN1-2018_021: 571118_577666)

The combination of mayi and the present tense can be used for one of two main scenarios. In one context, it is used to describe a possible event that the speaker genuinely believes to be potentially occurring (simultaneous to speech time) but for which there is insufficient evidence to indicate certainty. An example of this type appears in (1571). Alternatively, the use of an epistemic modal particle and the present tense is used to generate deliberately hypothetical, yet plausible, events which the speaker does not necessarily expect to be currently occurring (but could be). This type of usage was found in elicitation contexts where an event type was being prompted as an example utterance. The Ngardi response was set in the present tense but within the scope of the epistemic modal mayi to indicate its hypothetical nature. Unlike parda, mayi is not found with a wide range of verbal modal inflections, instead only appearing in combination with the potential as in (1514) and (1515). The modal mayi constrains the possible interpretations of the potential suffix to epistemic possibility/uncertainty – it cannot have an intentional/realis reading.488 (1514)

Kuyi mayi itaki=ma-nku. meat uncert track=get-pot ‘He might track that game.’ *’He will track that game.’ (PSM: TEN1-2018_029-01: 1692872_1697237)

488 This can be contrasted to a clausal construction involving the dubitative parda + irr verbal inflection which is used with past intentional meanings ‘x was going to y’.

10.2 Modality and modal particles 

(1515)

 585

Paju-ku mayi ngaju.489 bite-pot uncert 1sg ‘He might bite me.’ * ‘He will bite me.’ (MMN: TEN1-2018-021-02: 1011805_1014731)

10.2.1.2 Phrasal scope When exhibiting phrasal scope, mayi expresses (un)certainty/possibility that its phrasal host has the thematic role expected from its clausal context. In this function it encliticises to the final constituent of the phrase. Crucially, in such contexts the speaker is not casting doubt over the occurrence of the event itself but whether certain arguments (or adjuncts) have the role expected of them from their clausal context. An example is (1516). (1516)

Dina=mayi janka-nya warlu-ngka, taka=mayi janka-nya. lunch=uncert heat-pst fire-loc tucker=uncert heat-pst ‘It might be some lunch he cooked on the fire, it might be some tucker he cooked.’ (PSM: TEN1-2017_006-01: 411791_415343)

A special phrasal usage of mayi is its encliticisation to ignoratives in their declarative or interrogative usages. In this specific context, mayi appears to emphasise a speaker’s lack of certainty regarding the identity/content of the referent encoded by the ignorative. An example is provided in (1517). Further examples and discussion of the interaction of modals and ignoratives are provided in §10.2.5. (1517)

Nyajangu=mayi, pobala? how_many=uncert four How many exactly (was it), four? (BSE: TEN1-2017_015-01: 362005_364304)

10.2.1.3 Interrogative functions The modal mayi also serves as a question word in polar question clauses (1518) and as an extra-clausal tag, as in (1519) and (1520). See §10.9.1 for further discussion. (1518)

Mayi=n marda-rnanta? uncert=2sg.s have-prs ‘Do you have any (food)?’ (MMN: TEN1-2017_026-02: 929061_931035)

489 Note that the clausal positioning of mayi in (1514) is ambiguous as to whether it has scope over the NP kuyi or whether it has clausal scope.

586 

 10 Auxiliaries, particles and modal constructions

(1519)

Ya-nku dokta jalangu, mayi? go-pot doctor today tag ‘That doctor is coming today, yeah?’ (YMN: LC33b: 98315_104772)

(1520)

Yuwayi, nyamu mayi? yes finish tag ‘Yes, finished now, yeah?’ (TJN: LC21a: 1353496_1354988)

10.2.2 Unrealised wayi The modal particle wayi ‘unrealised’ exhibits substantive functional overlap with mayi. It similarly encodes the dubitative modality when operating with clausal scope (§10.2.2.1) or casts doubt over the thematic role of a phrase (§10.2.2.2). Also, as for mayi, wayi can be used with an interrogative function (§10.2.2.3). ‘Unrealised’ wayi is not used as a tag morpheme. 10.2.2.1 Clausal scope The modal wayi exhibits the same distributional and functional features of mayi when functioning with clausal scope. It hosts pronominal clitics and encodes epistemic possibility/uncertainty. The only distinction between mayi and wayi in their clausal functions is their clausal TAM co-occurrences. The particle wayi occurs most frequently with the potential (1521), (1522) and past (1523) inflections. A single example of wayi co-occurring with the customary is provided in (1524). (1521)

Wayi=pula parli=pu-ngku. unreal=3du.s find=hit-pot ‘Perhaps those two will find it.’ (TJN: LC32b)

(1522)

Ngaju wayi=rna yalkiri-rla karri-ju. 1sg unreal=1sg.s sky-loc be-pot ‘Perhaps I was in the sky then (i.e. hadn’t been born yet).’ (NGO: TEN1-2017_011-01: 49272_51147)

(1523)

Wayi=rna ya-ni Halls Creek-kurra nganayi-ku unreal=1sg.s go-pst place_name-all whatsit-dat wal waku. well neg ‘I was going to go to Halls Creek, to the hostel but no.’ (PLN: LC47a: 262683_265936)

Hostel-kurra, hostel-all

10.2 Modality and modal particles 

(1524)

 587

Ngu=lu wayi ma-nan. seq.aux=3pl.s unreal get-cust ‘Perhaps they used to do it.’ (PRN: FT1_a)

There are some limited examples of wayi co-occurring with the irrealis. In this context it can encode past obligation (1525) or past intention (1526), as observed for parda. (1525)

Wayi=rlipa ma-nngi warlu pirranginti, pirriya unreal=1pl.incl.s get-irr fire yesterday cold ma=rlipa karri-nyani. top.aux=1pl.incl.s be-ipfv.pst ‘We should have got some wood yesterday, we were cold.’ (PSM: TEN1-2018_033-03: 28979_39613)

(1526)

Waku wayi=rna=ngku yi-ngi. neg unreal=1sg.s=2sg.o give-irr ‘Oh no, I was going to give (the song) to you [but I didn’t].’ (MAM: TEN1-2016_002-01: 947694_951142)

10.2.2.2 Phrasal Scope The modal wayi can also exhibit phrasal scope. In (1527), wayi encliticises to a nominal and expresses uncertainty that the state depicted by its host is true, while in (1528) wayi encliticises to an ignorative and serves to enhance the uncertain nature of the Referent depicted by the ignorative. (1527)

Wurra=lku ma-nku ngaka. later=then get-pot later ‘He’ll get it later.’ Warlawarnu=wayi. ready=unreal ‘It might be ready (later).’ (PSM: TEN1-2017_010-01: 2317611_2319015)

(1528)

Karri-nya=rnalu nyajangu=wayi, yakarn yangi, Ruby Plain-ta. be-pst=1pl.excl.s how_many=unreal month one place_name-loc ‘We stayed there for I don’t know how long, (maybe) one month, at Ruby Plains.’ (PSM: TEN1-2016_004-01: 716401_722088)

588 

 10 Auxiliaries, particles and modal constructions

10.2.2.3 Interrogative function Like mayi, wayi can appear in interrogative clauses as in (1529) (and see §10.9.1). Unlike mayi, wayi is not used as a tag morpheme.490 (1529)

Wayi=nta wirriri=ya-ni=rlangu maju? unreal=2pl.s back=go-pst=also indeed ‘Didn’t you lot go back as well?’ (TJN: LC33a)

10.2.3 Dubitative parda The dubitative parda is a third modal marker of epistemic uncertainty. It has functions associated with both clausal (§10.2.3.1) and phrasal (§10.2.3.2) scope. Unlike mayi or wayi, parda is not used as a question word, nor does it function as an extraclausal tag. 10.2.3.1 Clausal scope When operating with clausal scope parda encodes the dubitative modality (1530). As observed for mayi – but unlike wayi – the dubitative parda co-occurs with verbs inflected for a wide range of clausal TAM (present, past, potential, irrealis, anticipatory). In combination with the tense inflections, parda can generate meanings of past (1530) and present (1531) possibility. (1530)

Wanya-ngku parda=yanu ka-nya. evil-erg dub=3pl.o carry-pst ‘The evil spirit might have taken them.’ (PRN: FT1a)

(1531)

a.

b.

Nyampa-jangka=rna ngajirrk=pu-nganta? what-ela=1sg.s sneeze=hit-prs ‘Why am I sneezing?’ Parda=lu ngantany ya-nanta, wurna jangu. dub=3pl.s man go-prs journey assump ‘Maybe some men are coming, like travelling y’know.’491 (PSM: TEN1-2018_021: 2870272_2876063)

490 Wayi and wayila are question words in Walmajarri (Hudson & Richards, 1990). 491 Sneezing is considered a possible sign of the imminent arrival of people, particularly relatives.

10.2 Modality and modal particles 

 589

Certain combinations of parda + prs inflection can refer to events after speech time as in (1532). (1532)

Rangarni=rna ya-nanta ngaju-punta(-kurra)— wiki-kujarra-kurra. morning=1sg.s go-prs 1sg-ppos-all ♂S-dual-all Kuyi parda=pula marda-rnanta. meat dub=3du.s have-prs ‘Tomorrow I am going (to see) my own two sons. Those two might have some meat.’ (PSM: TEN1-2017_027-01: 761788_768687)

In combination with the modal inflections, a much wider array of modal meanings is possible. Thus, consider the various modal meanings attributed to clauses involving parda in combination with the potential (1533), irrealis (1534), (1535) and anticipatory (1537) inflections. [i]  Future possibility (1533) Yalu-kujarra parda=pula wirrirl ya-nku-rni that-dual dub=3du.s back go-pot-hith wurrkal-kariny-ja. green-other-loc ‘Those two might come back next year.’ (PSM: TEN1-2017_026-02: 258980_262579) [ii] Past intention (1534) Yalu-rlu ngantany-ju parda nga-rnngi yuna=karri-nya=wali that-erg man-erg dub eat-irr rotten=be-pst=then ‘That man would have eaten it but the meat was rotten.’ (PSM: TEN1-2018_037-01: 411474_422887) [iii] Frustrated possibility (1535) Parda=rna nga-rnngi, partirri-wangu=wali=rna. dub=1sg.s eat-irr tooth-priv=then=1sg.s ‘I would have eaten it, however I’m toothless.’ (PSM: TEN1-2018_03701: 1040981_1047046)

kuyi. meat

590 

 10 Auxiliaries, particles and modal constructions

[iv] Past obligation492 (1536) Parda=rlipa puurda, ka-ngi pirranginti wirrirl, dub=1pl.incl.s yam carry-irr yesterday back yumpaly=pa=rlipa=wali. tired=ep=1pl.incl.s=then ‘We should have brought the potato back yesterday, but we were (too) tired.’ (PSM: TEN1-2017_026-01: 4130523_4149096). [v] Past intention/Frustrated possibility (1537) Parda ya-nanngirda, nyurnu=karri-nya=wali. dub go-anticip sick=be-pst=then ‘He was going to come, but he got sick.’ ‘He might have come, but he got sick.’ (PSM: TEN1-2018_037-01: 327752_333544) 10.2.3.2 Phrasal scope The dubitative parda can also be used with phrasal scope when encliticised to a phrasal constituent. In its phrasal function, parda encodes the speaker’s uncertainty towards the thematic role of the phrase to which it is encliticised as illustrated in (1538). There is a tendency for parda to be used when the speaker has at least some degree of confidence/belief in a proposition (more so than mayi or wayi) and is often found in speech acts that constitute suggestions. (1538)

Mula=parda laman-ta nya-ngku=rli=rla yalu-ku pujurl-ku. here=dub hollow-loc see-pot=1du.incl.s=3sg.obl that-dat frog-dat ‘Perhaps in this hollow log you and I should look for that frog.’ (MMN: TEN1-2017_016-02: 311079_315827)

In some cases, the phrase to which parda encliticises can be be co-extensive with the entire clause itself, as in (1539). (1539)

Ngaju-ku=parda. 1sg-dat=dub ‘(That) might be mine.’ (TJN: LC32)

492 Importantly, this same clause can be interpreted with both past intentional ‘we were going to’ and past possibility ‘we could have’ readings. An example of parda + irrealis verbal inflection elicited in the context of a past intentional meaning is presented in (1534).

10.2 Modality and modal particles 

 591

By juxtaposing a series of phrases encliticised with parda, a speaker can express a series of possibilities, similar to the English ‘perhaps . . . or . . . or . . . ’ as in (1540). (1540)

Makan-ta=parda, pirranginti-rla=parda, jilan-ta=parda. morning-loc=dub afternoon-loc=dub noon-loc=dub ‘Might be in the morning, or in the afternoon or at lunch time.’ (PRN: TG1_0034-02) [Context: Speaker explaining they don’t know when their parents will be back.]

As with mayi and wayi, parda can also be used to enhance the stance of uncertainty associated with ignoratives as in (1541), and see §10.2.5. The dubitative also co-occurs with the Kriol loan maitbi (1692), see §10.6.1. (1541)

Yala-kujarra-rlu nyampa=parda=yi=pula ka-ngku. that-dual-erg what=dub=1sg.o=3du.s carry-pot ‘Those two will bring me something.’ (PSM: TEN1-2017_008-01: 3516277_3518494)

10.2.4 Hypothetical ngarda The hypothetical modal ngarda is somewhat distinct from the other modals.493 When operating with clausal scope (§10.2.4.1), it may encode possibility/uncertainty of a clausal event but it is unique in having a semi-grammaticalised function as an apprehensional marker – encoding events that are both possible/likely and negatively evaluated. Unlike the other modals, ngarda is not used with phrasal scope. There is only very limited evidence of ngarda appearing in interrogative clauses (§10.2.4.2). 10.2.4.1 Clausal scope As observed with parda, the hypothetical modal ngarda is found with a range of TAM inflections including the potential (1542), hortative (1543), irrealis (1544) and anticipatory (1545), where it conveys possibility/uncertainty as observed for the other modals.

493 The particle ngarda is phonemicised as /ŋaɻa/, although it could arguably also be represented as /ŋara/ for those speakers who consistently render the segment as a tap rather than a flap or continuant. This variation is reflected in Cataldi’s (2011) variable spellings in the dictionary: ngarda and ngarra. These are not separate lexical items. There is likely a dialectal explanation underlying this variation since in Walmajarri a cognate form is rendered with a stop phoneme /ɖ/ realised as [ɖ] while in Jaru it is represented with a tap /r/ (realised as [r] sometimes as a trill [r] or a continuant [ɹ]).

592 

 10 Auxiliaries, particles and modal constructions

(1542)

Ngarda=yi=lu parli=pu-ngku. hyp=1sg.o=3pl.s find=hit-pot ‘Perhaps they will find one for me.’ (MDN: LC22a: 1535933_1538448)

(1543)

Tarn=yirra-rnanta=rla iliri ngarda wirriri ya-nankura cache=place-prs=3sg.obl hide hyp return go-ipfv.hort karrmantawangu. hungry ‘I am putting (the food) in a cache for her in case she comes back hungry.’ (PSM: TEN1-2018_007-03: 65573_84335)

(1544)

Ngarda=ngala ngari ka-ngi. hyp=1pl.incl.o belongings carry-irr ‘Maybe it (a camel) is bringing us our belongings.’ (MDN: LC22b: 831428_835666)

(1545)

Kuli-ngku ngarda=lu=yanu pu-nganngirda. aggressive-erg hyp=3pl.s=3pl.o hit-anticip ‘Those violent ones might be about to fight them.’ (PJA: TT76_2601: 1893652_1897582)

In combinations of ngarda + irrealis, meanings of epistemic possibility or deontic obligation are possible interpretations, as in (1546); however, there are only limited examples of ngarda clauses in the corpus that were directly translated with meanings equivalent to the latter, i.e. ‘x should y’, ‘x ought to y’. Note that the same clause can also have past time reference but in this context the obligative inflection (-Cumarda, §7.4.4) can be used. (1546)

Ngarda=rlipa ya-nngi, wirrpa. hyp=1pl.incl.s go-irr many ‘We might go, all of us.’ ‘We should go, all of us.’ (PSM: TEN1-2018_028-01: 2348472_2352558)

Hypothetical ngarda clauses also appear relatively frequently as apodoses in conditional complex clauses, as in (1547) and (1548). The other modals rarely have this function. (1547)

Kaji Jampit-tu luwa-rnngi, yala ngarda=rna nga-rnngi. condit ♂subsection-erg shoot-irr that hyp=1sg.s eat-irr ‘If Jampit spears it, I might eat it.’ (PRN: TG1_0034-01)

10.2 Modality and modal particles 

(1548)

 593

Kaji=rna kuyi ka-ngku-rni, kupa-ku ngarda=yi=n. condit=1sg.s meat carry-pot-hith cook-pot hyp=1sg.o=2sg.s ‘If I bring some meat, you might cook it for me.’ (PSM: TEN1-2019_007-02: 457412_472869)

An important and distinct feature of ngarda (as opposed to the other modal particles) is its role in apprehensional coordination (§11.2.3). Apprehensional constructions are biclausal coordinated clauses in which a ngarda clause behaves as a ‘lest’ clause, expressing a possible event that is evaluated as both likely and non-desirable. The ngarda clause co-occurs with a primary clause to express one of two types of speech acts: i) warnings/precautions (involving imperative- or directive-initial clauses) (1549) or ii) explanations (involving indicative main clauses) (1550). (1549)

Wakurra=n nga-lku puka, ngarda=n nyurnu=wanti-ju. neg=2sg.s eat-pot rotten hyp=2sg.s sick=fall-pot ‘Don’t eat the rotten (food), you might fall ill.’ ‘Don’t eat it, in case you fall ill.’ (MAM: TEN1-2016_001-01: 329550_346102)

(1550)

Ngu=rna jaku=ya-nanta ngarda=yi=lu seq.aux=1sg.s depart=go-prs hyp=1sg.o=3pl.s kunyarr-u. dog-erg ‘I’m getting out of here, those dogs might bite me.’ ‘I’m getting out of here lest those dogs bite me.’ (PSM: TEN1-2017_010-01: 954694_959466)

paja-rnngi bite-irr

Alternatively, a ngarda clause may be used in isolation with an implied directive as in (1551) and (1552) (and see related discussion of command types in §10.7). (1551)

Ngarda=n wanti-ngi! hyp=2sg.s=3pl.s hit-irr ‘You might fall! (Be careful!)’ (PSM: TEN1-2018_028-01: 3257052_3260800)

(1552)

Ngarda=ngku=lu pi-ngi! hyp=2sg.o=3pl.s hit-irr ‘They might hit you! (Don’t go!)’ (PSM: TEN1-2018_028-01: 3752572_3756064)

The ability for the admonitions in (1551) and (1552) to be interpreted with the illocutionary force of an order is context-dependent, since the irrealis can have past time

594 

 10 Auxiliaries, particles and modal constructions

reference.494 Thus, in the appropriate context, (1551) and (1552) could alternatively be interpreted as ‘you might have fallen (but you didn’t)’ or ‘they might have hit you (but they didn’t)’, respectively. It is likely that prosodic cues differentiate the possible interpretations but this remains to be carefully demonstrated. 10.2.4.2 Interrogative function The modal ngarda only rarely appears in interrogative clauses, as in (1553). Unlike mayi, ngarda has only been found in interrogative clauses where the event is located prior to reference time. (1553)

Kuyi=rna kupa-rni, narn-ngi ngarda=n meat=1sg.s cook-pst eat-irr hyp=2sg.s ‘I cooked some meat, did you eat this meat?’ (PSM: TEN1-2019_007-02: 16664_19015)

kuyi meat

minya? this

10.2.5 Modal particles and ignoratives In the preceding sections it was noted that a subset of modal particles {mayi, wayi, parda} can exhibit phrasal, rather than clausal, scope. A special instantiation of the phrasal scope of modal particles involves their encliticisation to ignoratives. In these contexts, modal particles take on a unique intensifying function, serving to enhance or foreground the speaker’s lack of knowledge encoded by the ignorative.495 The exact interpretation depends on whether the ignorative is functioning as a question word or as an indefinite in a particular clause. For example, the locative marked ignorative wanjirla can be encliticised with =parda, both in indefinite (1554) and interrogative (1555) contexts with meanings roughly equivalent to the English translations ‘somewhere or other/somewhere, I don’t know where’ and ‘where exactly?’. (1554)

Karlarra karri-nya=lu wanji-rla=parda ngurra-ngka. west be-pst=3pl.s where-loc=dub camp-loc ‘They stayed at a camp somewhere or other in the west.’ (TJN: Tjama4)

494 This is in contrast to Western Walmajarri which distinguishes a past irrealis (-ntarla) from a nonpast irrealis (-rla). 495 Cf. also the descriptive accounts of related phenomena in Gurindji (McConvell, 1996b, p. 104), Mudburra (Osgarby, 2018, p. 192–194) and Wanyjirra (Senge, 2015, p. 238).

10.2 Modality and modal particles 

(1555)

 595

Wanji-rla=parda=lu ya-ni? where-loc=dub=3pl.s go-pst ‘Where exactly have they gone?’ ‘Where have they gone? (for I don’t know)’ (YMN: LC23a: 715847_717228)

Declarative and interrogative usages of ignoratives encliticised with mayi are provided in (1556) and (1557), respectively. (1556)

Nyampa=mayi yala kuyi-kariny? what=uncert that meat-other ‘What exactly was that other meat animal? (I don’t know)’ (TJN: LC36a)

(1557)

Billycan minya, warlu karri-nya, kawurru=karri-nya, yirra-ka=rla billycan this fire be-pst heat=be-pst place-imp=3sg.obl nyampa=mayi teabag. something=uncert teabag ‘This billycan, it’s been on the fire, it has heated up, put some something in it, the teabag.’ (PSM: TEN1-2017_027-01: 1018491_1031948)

The attested combinations and semantic interaction of the modal particles and ignoratives are provided in Table 133. The hypothetical ngarda is never used in these contexts and wayi has only been observed in a limited number of cases. Table 133: Ignoratives encliticised with modal particles. Category

Modal marker

Contextual Meaning

=mayi

=parda =wayi

Indefinite

Interrogative

thing nyampa

✓

✓

‘something/ something or other’ ‘whatchamacallit’ (?)

‘what exactly?’

nyarrpa

✓

‘some sort’

—

✓

‘someone/ someone or other’

‘who exactly?’

✓

‘some amount’

‘how many exactly?’

location wanjirla

✓

(‘someplace somewhere’)

‘where exactly?’

time nyangurla

✓

‘some time’

‘when exactly’

entity ngana, ngantu quantity nyajangu

✓

✓

596 

 10 Auxiliaries, particles and modal constructions

Ignoratives encliticised with modal particles are particularly common during lexical elicitation when a speaker is either attempting to recall specific terms (1558) or hypothesising imagined scenarios relevant to the context of a particular elicitation prompt (1559). (1558)

a.

b.

c.

(1559)

Ngalyarr-ngawurr karri-nyanta yangi jaru nyampa=mayi sand-den be-prs one language what=uncert jangu? assump ‘There is a sandhill dweller one (a frog), like some something you know?’ Wurra=rna purda=nya-nganta kuyi-ku. wait=1sg.s listen=see-prs meat-dat ‘Wait while I remember that animal.’ Lirrajapal! sand_frog ‘(It’s called) lirrajapal!’ (TJN: LC36)

Wayi-rnanta, nyampa=mayi, swag, mayi? bind-prs what=uncert swag tag ‘He is tying something or other up – might be a swag?’ (TJN: LC48a: 1755017_757087) [Elicitation context: ‘What about the word wayirnanta?’]

Finally, the modals can also encliticise to the nominal ngurrpa ‘ignorant’ with a similar function of intensifying the speaker’s lack of knowledge, as in (1560) and (1561). (1560)

Ngurrpa=parda maitbi manija kardiya-rlu pulijman-tu. ignorant=hyp maybe manager white_man-erg policeman-erg ‘I dunno, might have been the whitefella policeman who took him (the manager at Billiluna).’ (PSM: TEN1-2016_005-03: 583539_587202)

(1561)

Ngurrpa=wayi karri-ju, wakurra palka mula. ignorant=unreal be-pot neg alive here ‘I don’t know much about it all, it doesn’t grow here (the mangarta tree).’ (TJN: LC28)

10.3 Propositional particles Following Laughren (1982a, p. 159), propositional particles in Ngardi can be defined as a subclass of particles (see §3.5) that serve to denote a speaker’s relation to a proposition at two levels: that of authorship and that of commitment to/belief in the truth

10.3 Propositional particles 

 597

of the proposition. Semantic and morphological items of this nature are sometimes grouped under a wider term, ‘propositional markers’ (Andersen & Fretheim, 2000) or ‘propositional modifiers’ (Dench, 1990, 166ff.) and have been identified in many Australian languages (Dixon, 1980, p. 284). In Ngardi, propositional and modal particles bear many similarities. Both belong to the particle word class and are distinguished from the auxiliary subclass by the fact that they: i) are minimally disyllabic ii) need not host the pronominal clitics iii) can have either clausal or phrasal scope (and can co-occur with the auxiliaries in this latter function) iv) can (in some cases) co-occur with other propositional particles. In addition, a subset of these particles can function as independent interjections (§10.5). Modal and propositional particles themselves can also be partially distinguished on some limited functional criteria (following Laughren, 1982a, p. 131 for Warlpiri). Propositional particles are unlike modal particles in two ways. First, they can express attitudes towards propositions (typically states of belief) and may further index sources/authorship of knowledge (own knowledge, second-hand knowledge, former knowledge). Second, their semantic function may extend to extra-linguistic or wider discourse contexts in which the clause containing the propositional particle is found. An exhaustive list of propositional particles is provided in Table 134 with a brief overview of their function. While these serve as broad characterisations of the functions of each propositional particle, the exact meaning of a propositional particle is only clear in relation to the accompanying ‘proposition’ (Laughren, 1982a, p. 159). In addition, there are a number of Kriol-derived lexemes used in Ngardi that have functions not dissimilar to propositional (and/or modal) particles (e.g. concessive wal < English ‘well’). These are discussed separately in §10.6. Table 134: Propositional particles. Propositional Particle

Gloss

Section

Overview of function

jala kala

delim asrt

§10.3.1 §10.3.2

maju

‘really’

§10.3.3

ngayi

‘indeed’

§10.3.4

nganta

quot

§10.3.5

Restricts interpretation of a proposition Asserts speaker’s own belief in a proposition Intensifies speaker’s commitment to a proposition Indexes speaker’s firsthand knowledge of a proposition Indexes speaker’s secondhand knowledge of a proposition

598 

 10 Auxiliaries, particles and modal constructions

Table 134 (continued) Propositional Particle

Gloss

Section

Overview of function

kulangarra, kulanganta junga

cf

§10.3.6

assump

§10.3.7

wali

‘then’

§10.4.2

Expresses that a proposition was formerly (and erroneously) held to be true Asserts that a proposition is true and based on widespread belief/common knowledge Situates a proposition as occurring (logically or temporally) in the subsequent state of another eventuality

The positioning of a propositional particle within a clause is determined by its scope. When hosting the bound pronouns, propositional particles have clausal scope (i.e. over a proposition expressed by the main predicate and its arguments). When propositional particles occur elsewhere in a clause, they generally have scope over a phrase to which they are adjacent, most typically a preceding phrase.496 This property of variable scope is consistent with modal particles as described in the previous section. Some propositional particles also have functions as interjections or tags where they may constitute independent utterances. Finally, some propositional particles have an ambiguous status between independent words (particles) and semi-dependent clitics (e.g. jala and wali). These issues are discussed in relevant subsections.

10.3.1 Delimitive (=)jala The delimitive particle jala appears both as an independent word and as a clitic. Cataldi provides two related sets of related meanings for jala: ‘indeed, to be sure, certainly’ and ‘it is really, rather, on the contrary, by contrast’ (2011, pp. 34, 51). The first set of glosses can be understood as a single ‘delimiting’ function, which serves to narrow the scope of the proposition from closely related propositions to a more precise proposition. Most of the examples of jala in the corpus have this function. Consider (1562)–(1565). (1562)

And Patricia ita=jala=lu nyantu ka-ngani-nyirra. conj name small=delim=3pl.s recog1 carry-ipfv.pst-narr ‘And they carried Patricia, who was only small then.’ (YMN: LC23b: 28380_33549)

496 Propositional particles which appear clause-finally bear similarities to right-detached ‘complement clauses’ in some languages, e.g. English, as in ‘The mailman came today, I think’. These are sometimes referred to as attitude predicates (Andersen & Fretheim, 2000, p. 5).

10.3 Propositional particles 

 599

(1563)

Wajim=ma-nku wurra=jala=rna warlupaka-rla kaji wash.tr=get-pot wait=delim=1sg.s lunch-loc condit hotfelo=karri-ju. hot.adj=be-pot ‘I will wash only later on, around lunch/noon when it is hot.’ (TJN: LC32b)

(1564)

Wurra Nangala palka, jingka=jala kuja palka-wangu karlarra wait ♀subsection alive child=delim comp alive-priv west hospital-kurra. hospital-all ‘Wait, Nangala is here, it is only the child who is not here who (went) west to hospital!’ (TJN: Cataldi, 2011. p. 248)

(1565)

Lurrij mayi? Waku=jala=rna ngaju. finish tag neg=delim=1sg.s 1sg ‘Is this enough? I have none in fact myself.’ (LC33: 98315_104772)

It is rare, but possible, for the delimitive to appear clause-initially as in (1566). (1566)

Jala warntu, pangkarra, kuriny yirdi, blanket=pa=rnalu delim blanket blanket two word blanket=ep=1pl.excl.s ngarri-rnanta kuriny yirdi. tell-prs two word ‘Indeed warntu and pangkarra are the two words we use; we call ‘blanket’ by these two words.’ (YMN: LC40)

The second function of jala described by Cataldi is as a marker of a contrastive assertion (1567). There are exceedingly few examples in the corpus of this kind but its meaning appears to be consistent with a clitic of the same form in Warlpiri (Laughren et al., 2007, p. 23). Lacking sufficient examples, it is difficult to argue whether this ‘contrastive’ use can be viewed as an extension of the delimitive function or not. (1567)

Jala ma=rlijarra yu nganayi, wartunuma-ku=rlijarra=rla delim top.aux=1du.excl.s yes whatsit ants-dat=1du.excl.s=3sg.obl pangi-rni. dig-pst ‘On the contrary, we two dug, yes, you know, for the flying ants.’ (MMJ: LC3)

600 

 10 Auxiliaries, particles and modal constructions

10.3.2 Assertive kala The assertive particle kala is homophonous with another propositional particle with a contrastive conjunction function (§11.2.4).497 Assertive kala is used to express a speaker’s commitment to the truth value of a proposition, as in (1568).498 (1568)

Karlarra kala=rnalu karri-nya. west asrt=1pl.excl.s be-pst ‘We did in fact stay in the west.’ (PSM: TEN1-2016_004-01: 248207_251403)

The assertive particle can either modify a proposition encoded by an entire clause, in which case it hosts the bound pronouns as in (1568), or else a proposition encoded by a phrase, in which case it appears adjacent to the phrase in question as in (1569). (1569)

Kala yirrwa-rla nganayi. asrt road-loc whatsit ‘(It was) on the road (he did it).’ Aeroplane landed=wanti-nyani-nyirra, yalu-rlu wayitpala-rlu. aeroplane land.pst=fall-ipfv.pst-narr that-erg European-erg ‘That whitefella landed the aeroplane.’ (BSE: TEN1-2017_015-01: 154798_161152)

The propositional particle kala also interacts with other clausal operators. In (1570), for example, the negative operator wakurra falls within the clausal scope of kala, which hosts the bound pronoun. (1570)

a.

b.

Wakurra kala=rnalu wanti-nya-ngurra nyurnu, waku. neg asrt=1pl.excl.s fall-pst-narr sick nothing ‘We (definitely) didn’t fall sick, not at all.’ Onli hawij-ja wanti-nyanta nyurnu. only house-loc fall-prs sick ‘Only (now) in houses are we getting sick.’ (TJN: Manungka_01-035151: 802667_806037)

497 The assertive propositional particle kala is used extensively by Tjama Napanangka in the Cataldi corpus. It is not found in the unprompted speech of Marie Mudgedell or Patrick Smith in 2016–2019 recordings. 498 A formally identically morpheme appears in Warlpiri where it has three functions: i) ‘but’; ii) remote past, usitative; and iii) potential (with aspectual morpheme ka) (Hale et al., 2015, p. 1687, 1703).

10.3 Propositional particles 

 601

Alternatively, kala can fall within the scope of another clausal operator, for example the modal particle mayi, as in (1571). (1571)

Kala mayi=lu mangarri nga-rnanta? asrt uncert=3pl.s veg_food eat-prs ‘Perhaps they are actually eating food?’ (MDN: LC22b: 749257_751885)

10.3.3 maju ‘really’ Cataldi (2011, p. 169) describes maju as an interjection with the semantic content ‘I say, indeed, of course, really’. Examples of its role as an interjection are provided in (§10.5.3). In addition, maju can also be integrated into a clause with a propositional modifying function that enhances or intensifies the strength of an assertion. When exhibiting phrasal scope over NPs, it behaves similarly to the intensifier suffix -nyayirni (§4.4.6) in that it can amplify the degree of an attributive predication, as in (1572) or an equative predication, as in (1573). (1572)

Karlajangka, jurlpu, ngapa-ngawurr, jurlpu maju kirda, bird_sp bird water-den bird really big yawu nga-rnanta. fish eat-prs ‘The karlajangka (lit. ‘one from the west’) is a bird, one that lives around water, a really big bird that eats fish.’ (TJN: LC36b)

(1573)

Wirriji, wirrkirl maju luwa-ku=rlipa. hairstring hair really shoot-pot=1pl.incl.s ‘Wirriji, (It’s) hair really, which we spin.’ (MDN: LC22a: 948131_951184) [Prompt: And what’s the word for hairstring?]

When operating with clausal scope, maju serves to indicate commitment to the truth value of the clausal proposition as in (1574). (1574)

Maju karri-nyankura=rlipa. really be-ipfv.hort=1pl.incl.s ‘Truly, we will stay here.’ (MDN: LC22b: 439778_441961)

The propositional particle maju can also reaffirm the correct truth value of a proposition against an assumed or expected (but false) truth value. Thus, in (1575), the speaker expresses surprise or incredulity at the truth of a proposition: ‘that the five dollars would be (left) with her’ and indicates their contrary assumption: ‘that five dollars wouldn’t be (left) with her’.

602 

 10 Auxiliaries, particles and modal constructions

(1575)

Maju faifdola, nyampa-ku=yi=n=ku=rla really five_dollars what-dat=1sg[lct]=2sg.s=ep=lct ‘Wow five dollars, really! Why have you (left) this money with me?’ (LC56b)

mani? money

10.3.4 ngayi ‘indeed’ Cataldi (2011, p. 36) describes the function of two related forms ngayi ~ ngari as a propositional particle meaning ‘I say, indeed’.499 As indicated by Cataldi’s translations, this particle serves to assert a proposition on the basis of firsthand knowledge as described for kala (§10.3.2). Unlike kala, ngayi does not appear to have any additional function as a contrastive conjunction. The particle ngayi serves to modify propositions about events in which the speaker was involved or at least has clear firsthand knowledge of, as is the case in (1576) and (1577). (1576)

Ngayi=rnalu tarn-tarn=yirra-rni trayila-rla. indeed=1pl.excl.s hang_up-rdp=place-pst trailer-loc ‘Indeed, we loaded up the belongings onto the trailer.’ (PSM: TEN1-2018_038-01: 348522_351013)

(1577)

Ngayi ngaju=rna ya-nanta puurda-kurra. indeed 1sg=1sg.s go-prs yam-all ‘Yeah, I am going for big yams.’ (MDN: LC22a: 1215241_1217128)

When the particle ngayi exhibits scope over a proposition which itself is marked by some form of epistemic uncertainty (e.g. an irrealis-marked verbal event), the use of ngayi serves to assert that the authorship of the assertion rests with the speaker, approximately equivalent to the English assertive ‘I reckon’, as in (1578). (1578)

Ngajarra=rlijarra ya-nku, Napaljarri, ngayi wanja-rnungka-ngi 1du.excl=1du.excl.s go-pot ♀subsection indeed leave-x&move-irr Nakamarra-rlu. ♀subsection-erg ‘Us two will go, Napalajarri, I reckon those two Nakamarras might leave and be off.’ (TJN: 17-8-00e)

499 This particle was not used by speakers recorded between 2015–2019. Additionally, a variant ngari, while listed by Lee Cataldi, is not identified anywhere in the whole corpus. Note that ngari is a separate lexeme in Ngardi meaning ‘belongings, things’.

10.3 Propositional particles 

 603

Similar usages can be found following rhetorical (1579) or performative (1580) reported speech, where a second clause is juxtaposed as a response. Initial ngayi flags the following proposition is the speaker’s own assertion on the topic.500 (1579)

Nyampa=lu kuja nga-rnani-nyirra? ngayi jirrilpaja. what=3pl.s sub eat-ipfv.pst-narr indeed carrot ‘What (then) were they eating? I think bush carrot.’ (TJN: Manungka_01-035151: 38720_43327)

(1580)

a.

b.

Nyarrpa ma-nku=rlipa kuja minya how get-pot=1pl.incl.s thus this ‘How are going to deal with all of this food?’ Ngayi, nga=rlipa kupa-ku warlu-ngka, indeed seq.aux=1pl.incl.s cook-pot fire-loc ‘Yeah I know, let’s cook it on the fire, like this.’ (PPN: Manungka_01-035160: 1667726_1673067)

mangarri? veg_food kuja-rlu. thus-erg

In Warlpiri and Eastern Walmajarri, a propositional particle ngayi can also be used with interrogative intonation to seek confirmation of the truth value of a proposition in the preceding discourse, as in the English ‘isn’t that true?’ (Hudson & Richards, 1990; Laughren, 1982a, pp. 154–155).501 This tag-like function is more commonly expressed via mayi (§10.2.1) in Ngardi or (in Kukatja code-switches): ‘kurlu?’ or ‘ngi?’. However, there is at least one example illustrating a Ngardi speaker using ngayi in this manner: (1581)

Mapirri=ma-ni ngalipa-rlu=wali, ngayi? together=get-pst 1pl.incl-erg=then indeed ‘(We) got it all together, we did it then, yeah?’ (NMN: Nanyuma2)

Finally, there is also at least one example (1582) in which ngayi appears to have an attenuating or limiting function ‘only x’ or ‘just x’ – a function observed for a Warlpiri particle of the same form (ngari ~ ngayi) (Laughren, 1982a, pp. 152–154). Alternatively, this function may actually derive from the enclitic =wali in (1582) (see discussion of =wali in forming degree comparatives in §10.4.2.4). (1582)

Itayi=wali ngayi ka-nya=yi ngati-ngku. small=then only carry-pst=1sg.o M-erg ‘When I was only a little girl, my mother took me.’ (KPA: TEN1-2017_029-02: 23104_27794)

500 This use of ngayi as an extra-clausal interjection to indicate understanding of and/or agreement of a preceding proposition is also reported for ngayi in Warlpiri (Laughren, 1982a, p. 154). 501 In Eastern Walmajarri, ngayi can even be used as an affirmative response (Richards & Hudson, 2012).

604 

 10 Auxiliaries, particles and modal constructions

10.3.5 Quotative nganta Cataldi (2011, p. 213) describes the semantic range of nganta as ‘supposedly, probably, is or was thought, if’. As indicated by these translations, nganta serves to index that a speaker’s belief in a proposition is based on secondary or inferential information and not direct (personal) experience with, or witnessing of, an eventuality (cf. ngayi, §10.3.4). The exact source of knowledge need not be made explicit. The gloss ‘quotative’ is adopted from Simpson’s description of Warlpiri (1983, p. 6) and Dench’s description of Martuthunira (1990, p. 167).502 Particles or clitics with similar functions are also found in Arrernte (Wilkins, 1989, p. 360) and Diyari (Austin, 1981a, p. 179). Examples of nganta exhibiting clausal scope and occupying clause-initial position are provided in (1583) and (1584). (1583)

Nganta kuluwa wanti-nya-nyirra, kaniny-kaniny nga=rnalu quot cold_rain fall-pst-narr inside-rdp seq.aux=1pl.excl.s takurr=ya-ni-nyirra purtku-ngka jayanta. enter=go-pst-narr warm-loc long_time ‘Apparently, (when) cold-weather rain would fall, then we would go right inside (caves) into the warm for a long time.’ (MMN: Manungka_01-035155: 344951_351959)

(1584)

Nganta=lu wilayi pantirri pamarr-ku nawu. quot=3pl.s around dance stone-dat now ‘Apparently they were dancing around for that stone, then.’ (MMN: TEN1-2017_004-03: 297494_302004)

When exhibiting phrasal scope, nganta can co-occur with another propositional particle exhibiting clausal scope. In (1585) for example, nganta ‘apparently’ has phrasal scope over the proposition ‘it was our mother’, while kulanganta ‘mistakenly thought’ (§10.3.6) has clause-level scope indicating a mistaken belief associated with the entire clause: ‘we thought (mistakenly) that we were looking after one who was supposedly our mother’. (1585)

Kulanganta=rli ngati nganta marda-rni-nyirra. cf=1du.incl.s M quot have-pst-narr ‘Mistakenly, we thought we had been looking after our supposed mother (but we weren’t, it was someone else).’ (TJN: LC20a).

502 A formally identical particle in Warlpiri is glossed ‘quotative, suppose’ (Simpson, 1983, p. 6; see also Laughren, 1982a, pp. 137–141).

10.3 Propositional particles 

 605

10.3.6 Counterfactual kulanganta, kulangarra, kulanga, kulangu ‘mistaken belief’ The modal particle kulanganta, kulangarra, kulanga, kulangu encodes a specific type of counterfactual belief in a proposition, that of ‘mistaken belief’.503 The likely historical composition of this particle involves the Warlpiri-origin negative operator kula (§10.8.2) as described by Laughren (1982a, p. 148) and Nash (1986, p. 60).504 The variation in allomorphy may have a dialectal and/or regional component: kulanga and kulangarra are reported in Eastern Walmajarri while kula alone is observed with the same function in Western Walmajarri (see Laughren, 2017a). The quadrisyllabic forms kulanganta, kulangarra both quite plausibly bear a connection to the propositional particles nganta and the hypothetical modal particle ngarda, in a similar fashion to the Warlpiri analysis. The trisyllabic forms are exceedingly rare in the corpus – predominantly appearing in the speech of a single speaker – and may be fused instances of kula with auxiliaries ngu and nga. Whatever their origin, no meaningful differences between the variants have as yet been identified. The function of the mistaken belief propositional particle is to describe a belief (held by the speaker or another referent) in a proposition which is known to be false but that was formerly believed to be true. It can be translated into English as ‘mistakenly thought that x’. As a propositional particle it exhibits both clausal (1586) and phrasal (1587) scope. (1586)

Kulanganta=n ya-nngi, or ya-ni. cf=2sg.s go-irr or go-pst ‘I thought you went (but you didn’t).’ (PSM: TEN1-2018_034-02: 18500_27133)

(1587)

Kulangarra motika, kala meilplein ya-nani-nyirra. cf car asrt mail_plane go-ipfv.pst-narr ‘We thought it was a car, but it was a plane which was moving.’ (YMN: LC23a: 1474036_1477043)

As the translations in (1586) and (1587) indicate, the speaker may use kulangarra to describe a belief held only by themself, a belief that extends to additional referents (typically recovered from the speech or discourse context) or even a belief held by someone other than the speaker (e.g. (1591)). As argued for Warlpiri, ‘it is left open to the addressee to interpret who actually held the belief deemed false by the speaker at the time of speaking’. Furthermore, ‘the implication by the speaker and the interpretation by the addressee can depend very heavily on extrinsic (pragmatic)

503 Cf. the label ‘mistakenly thought’ for a similar particle in Martuthunira (Dench, 1990, p. 168). 504 The general negative kula in Warlpiri is found fused with nga, nganta and ngarra to form a particle with the ‘mistaken belief’ function.

606 

 10 Auxiliaries, particles and modal constructions

information or context and/or contextual information contained in the larger discourse’ (Laughren, 1982a, p. 149). There are no restrictions on compatibility of the counterfactual with clausal TAM features. Further examples are provided illustrating combinations with various verbal inflections: imperfective irrealis (1588), imperfective past (1589), and imperfective past narrative (1590). (1588)

Kulanganta=n ya-nanngi, jalangu. cf=2sg.s go-ipfv.irr today ‘I thought you were heading off today.’ (IPN: TEN1-2018_034-02: 51177_55397)

(1589)

Minya ma=lu ngarrily puurda, ngu=rlipa this top.aux=3pl.s soft yam seq.aux=1pl.incl.s ya-nani-nyirra kulanganta=lu wuruju karri-nyani, wurra=lku go-ipfv.pst-narr cf=3pl.s good be-ipfv.pst wait=then ma=lu ngarrily. top.aux=3pl.s soft ‘These bush potatoes were under-ripe, so we went where we thought they might be good, but they were still not ripe (there).’ (MMN: TEN1-2018_021: 3661331_3678621)

(1590)

Kulangarra=n ma-nani-nyirra mangarri. cf=2sg.s get-ipfv.pst-narr veg_food ‘I thought that you had already got the food.’ (PSM: TEN1-2018_034-01: 727394_3729977)

In some rare cases, the counterfactual can be used in a quotative/semblative function. Syntactically this means it can introduce any kind of clause (including an interjection) but it still asserts that the truth value behind the content of that speech act is counterfactual, as in (1591). (1591)

Kardiya=lu nya-nganta kulanganta, kulanganta white_man=3pl.s see-prs cf cf “oh, nice painting”. Wakurra=lu purda=nya-nganta that’s walya. oh_nice_painting neg=3pl.s listen=see-prs that_is country ‘Whitefellas seem to look at (it), mistakenly (thinking) like, “oh, nice painting”. They don’t understand that that is country (i.e. not just a painting).’ (MAM: TEN1-2016_002-03: 350645_356228)

A distinct feature of (1591) is that the referent who holds the mistaken belief in the proposition ‘that it is just a painting’, is not positioned as necessarily having come to a subsequent knowledge state: ‘that it is not just a painting’. Instead, it is only

10.3 Propositional particles 

 607

the speaker who knows that the proposition is false (i.e. ‘they thought x but they are wrong’ – i.e. ‘I/we know different’). This functionality is also observed of the counterfactual in Warlpiri (see Laughren, 1982a, p. 150).

10.3.7 Assumptive jangu The lexeme jangu has already been identified as a recognitional nominal (§5.2.2.4, ‘you know the one’) which inflects for case and shares the characteristic properties of a demonstrative. A homophonous form jangu also occurs in Ngardi as a polyfunctional, uninflecting particle. As an uninflecting word it functions as a tag word (1592) and a propositional modifer (1593), (1594). Propositional jangu asserts that a proposition is true based on widespread belief or common sense (cf. the quotative nganta, or ‘I say’ ngayi). It can effectively be translated by English expressions ‘as you would know’ or even rhetorical ‘y’know?’. These evidential functions clearly have a semantic connection with the more concrete, referential usage of jangu as a nominal ‘the one you know’ but are distinguished by their clear propositional-modifying roles. This particle has an additional role as a conjunction, discussed separately in §10.3.7.1. (1592)

Pirdiny=karri-ya, jangu? arise=be-imp assump ‘(We say) pirdiny-karriya (get up!), yeah?’ (MMN: TEN1-2019_004-03: 464587_466119)

(1593)

Ailipala ya-ni=rnalu kankarra, jangu early.adj go-pst=1pl.excl.s up.a assump ‘We went early up to Yakka Yakka y’know.’ (TJN: Manungka_02-022027: 2567349_2572365)

(1594)

Wangkajunga jangu=lu=yanu turl=ma-nani-nyirra language_name assump=3pl.s=3pl.o gather=get-ipfv.pst-narr juju-ku. ceremony-dat ‘True/as you know, they gathered the Wangkajunga speakers together (too) for the ceremony.’ (TJN: LC21a: 1201009_1204439)

Yakayaka-rla. place_name-loc

10.3.7.1 Conjunction function As a conjunction, jangu introduces a relation either of overt similitude (x is like y) or approximation ([. . .], for example x). It may conjoin either a nominal (1596) or a much larger discourse unit (1597). In its conjunction role, jangu primarily serves a discourse function of elaboration of exemplification. The conjunction role can be understood as a contextual usage of the propositional function in multiclausal contexts as in (1595).

608 

 10 Auxiliaries, particles and modal constructions

(1595)

Holot pamarr deya, ngalikirdi, jangu=nta pamarr nya-nya. whole_lot stone there grind_stone like=2pl.s stone see-pst ‘All those stones, there, grinding stones, like those stones you lot saw.’ (KPA: TEN1-2017_029-02: 139420_142311)

(1596)

Jangu Gordon Downs-rla, nyanyi karri-nyani ngantany-ngampurr. like place_name-loc before be-ipfv.pst man-paucal ‘Like at Gordon Downs, lots of people had been living there back then.’ (BST: TEN1-2016_003-04: 7604_12187)

(1597)

a.

b.

c.

Iji-jarra, laik a Sam Haslett. gift-assoc, like a name ‘A generous person, someone like Sam Haslett.’ Jangu=rnalu nyanyi ya-nani, wurna jina motika-wangu-rla like=1pl.excl.s before go-ipfv.pst journey foot car-priv-loc waku, kayirra=rnalu ya-nani-nyirra. nothing north=1pl.excl.s go-ipfv.pst-narr ‘Like when we would travel a long time ago, on foot when there weren’t vehicles, no nothing, when we would travel north.’ Kardiya yangi kuriny=pa=pula karri-nyani Palmer Spring-urla white_man one two=ep=3du.s be-ipfv.pst place_name-loc Sam Hazlett, Jimmy Malabi. name name ‘There was one, (no) there were two whitefellas staying at Palm Springs: Sam Hazlett and Jimmy Malabi (they were both generous).’ (PSM: TEN1-2018_007-03: 749634_751631)

During dictionary elicitation jangu was also recruited as part of a formulism for providing definitions for terms as represented in (1598). Contexts like these are arguably grammatical bridging contexts where both demonstrative and particle interpretations of the wordform are possible, proceeding first from an appositional structure ‘x, you know y’ towards a coordinated structure ‘x is like a y that . . . ’. (1598)

[Definiendum] (,) jangu [Definitional nominal/clause]

The definitional term used in construction of the kind schematised in (1598) can be: i) a more generic term (x is a (type of) y) (1599) ii) a single feature/sub-part of the definiendum + description of function (e.g. x is a y which is used for z) (1600), (1601) iii) a well-known synonym (x is like y) (1602).

10.3 Propositional particles 

 609

(1599)

Ngarlurrpu jangu purnu kirda. plant_sp recog2 tree big ‘The ngarlurrpu is a type of big tree.’ (LC28)

(1600)

Kunji jangu purnu minya kayini-mpala karri-nyanta. bauhini recog2 tree this north-along be-prs ‘Kunji is (like) that tree grows here along the north.’ (Cataldi, 2011, p. 138)

(1601)

Wapukuru, kurnti. . . fighting_stick fighting_stick. . . ‘Wapukurru or kurnti (is what we call it, a fighting stick).’ Purnu marrka jangu kuli-ku. wood hard recog2 fight-dat ‘A hard wood, like (one) for fighting.’ (Cataldi, fieldnotes)

(1602)

Nga-lku=rna yarla minya puurda jangu eat-pot=1sg.s yam this yam recog2 ‘I will eat these bush potatoes, puurda, just like yarla.’ (JLI: TT76_2201: 1464039_1468346)

yarla. yam

In the sequence in (1603), Patrick Smith was describing to me where he worked as a stockman and the tracts of country he travelled across. In this extract, he is relating a place name and an area of country to an immediately adjacent tract of country with which he knows I have experience. He acknowledges this shared knowledge state by introducing an explanatory clause with the particle jangu. (1603)

a.

b.

Ngurra, Oak, nga=lu kardiya-rlu ngarri-rnanta, country place_name seq.aux=3pl.s white_man-erg tell-prs ngurra. country ‘The country (there), whitefellas call it Oak, that country.’ Jangu nyuntu ya-ni motika-kurlu Jungurrayi, bat moa-ngka. like 2sg go-pst car-prop ♂subsection but more-loc ‘Just like (where) you went in the vehicle Jungurrayi, but (it’s) further along.’ (PSM: TEN1-2018_038-01: 1060263_ 1075031)

Lastly, the particle jangu is also used in other pragmatic contexts which are neither clear instances of indicating aforementioned knowledge (or shared knowledge states) or clear semblative (x is like y) functions. Consider, for example, the narrative extracts presented in (1604) and (1605).

610 

 10 Auxiliaries, particles and modal constructions

(1604)

a.

b.

(1605)

a.

b.

c.

Wirriri ya-ni=rna ngurrara-kurra. back go-pst=1sg.s homeland-all ‘I went back to that country.’ Jangu=rnalu puliki jilykarra=la-nani yalu-ngka. like=1pl.excl.s cattle brand=pierce-ipfv.pst that-loc ‘Like, we were branding the cattle there.’ (PSM: TEN1-2018_009-01: 517117_522307) Puliki=rnalu majurrum=ma-ni Kururrungku-rla lurrij. cattle=1pl.excl.s muster.tr=get-pst place_name-loc finish ‘We mustered the cattle, finishing up at Kururrungku.’ Jangu=rnalu nganayi katimat=ka-nya drabtim=ma-ni like=1pl.excl.s whatsit cut.tr.compl=carry-pst draft=get-pst puliki, turang-wan. cattle strong_one ‘Like you know, we were whatsit, ‘cutting out’ the cattle, drafting them, all strong ones.’ Wal, wurna=karri-nya nawu=rnalu. well journey=be-pst now=1pl.excl.s ‘Well then, we subsequently headed off.’ (TEN1-2018_009-01: 78533_95753)

In both of the preceding examples, jangu appears to have a discourse-pragmatic function, reminscent of ‘evincive’ usages of the semblative preposition ‘like’ in Aboriginal English (Sharifian & Malcolm, 2003). Evincives are used to ‘allow the speaker to call attention to the current thought in his/her private world without displaying too much detail in the talk’ (Sharifian & Malcolm, 2003).

10.3.8 Certainty junga The particle junga is used to indicate certain belief in a proposition, as in English ‘true, correct, definitely’. In its particle form it can either modify a proposition (1606) or it can be used as an independent exclamative ‘true, that’s correct!’ as per the metalinguistic response in (1607). (1606)

Yala=yi junga=lku jukurr-jangka. that=emph cert=then Dreaming-ela ‘That’s right, it truly is from the Dreaming!’ (MMJ: LC31: 341735_344022)

10.4 Temporal and propositional clitics 

(1607)

 611

Mukurra, junga, my boy. this-all cert my son ‘(We say) mukurra (for ‘this way’), that’s correct, my son.’ (PSM: TEN1-2018_023-02: 629772_630863)

10.4 Temporal and propositional clitics In addition to the range of modal (§10.2) and propositional (§10.3) particles, Ngardi possesses a small class of clitics which serve temporal- or propositional-modifying functions. A subset of these clitics also have discourse-related functions – many of which can be viewed as extensions of their core temporal- or propositional-modifying functions. The inventory of temporal and poropositional clitics is provided in Table 135. They form a heterogenous set of clitics with various semantic and syntactic properties. This may not be an exhaustive list of all enclitics. Table 135: Temporal and propositional clitics. Form

Gloss

Subsection

=lku (=)wali =rra =rlangu =mipa =katu =warru =wu =yi

‘still, then’ ‘then’ ‘again’ ‘also’ restr restrictive compar comparative incre incremental voc vocative emph emphatic

(§10.4.1) (§10.4.2) (§10.4.3) (§10.4.4) (§10.4.5) (§10.4.6) (§10.4.7) (§10.4.8) (§10.4.9)

Clitics in Ngardi all share the following properties: i) they occur outside of verbal and nominal inflection505 ii) they are generally phonologically dependent iii) they are generally unrestricted with respect to their hosts iv) temporal/propositional clitics have scope over the constituent to which they attach. One of the propositional clitics listed above, the ‘then’ clitic = wali, occupies a marginal space between clitichood and wordhood, a feature shared with the modal

505 Unlike Warlpiri, Ngardi possesses no processes of vowel harmony to illustrate a phonological domain which subsumes host + phonologically dependent enclitics. However, there are some limited processes of phonological elision of intervocalic semivowels formed through the attachment of certain clitics (e.g. a=wali > aali).

612 

 10 Auxiliaries, particles and modal constructions

particles and the topic and emphatic auxiliaries (ma, ka).506 Table 136 summarises the types of particles and clitics found in Ngardi and identifies those subtypes which fall somewhere in between. Table 136: Cline of clitics and particles. Clitic

Particle

Pronominal clitics =rna 1sg.s =yi 1sg.o =pulany 3du.o, etc . . .

Modal particles (=)parda dub (=)mayi uncert (=)wayi unreal

Propositional particles kala asrt nganta quot kulanganta cf, etc . . .

Phonological clitic =ku ep ‘epenthetic’ =pa ep ‘epenthetic’

Auxiliaries (=)ma top.aux

Auxiliaries nga, ngu seq.aux ka emph.aux

Temp./Prop. Clitics =lku ‘then, still’ =yi emph =rra ‘again’ =mipa restr =katu compar =warru incre =wu voc

Temp./Prop. clitics (=)wali ‘then’ (=)jangu assump

10.4.1 Subsequent/persistent =lku ‘then, still’ The clitic =lku has a number of logically related functions. First, it has two temporal functions. It can mark both persistent eventualities (i.e. ‘still’ (§10.4.1.1)) and subsequent eventualities (i.e. ‘then’ (§10.4.1.2)).507 Additionally, it has a discoursepragmatic function marking contrastive topics (§10.4.1.3).508 In all functions, =lku can exhibit clausal or phrasal scope. Examples are provided in the relevant subsections. 10.4.1.1 Persistent eventuality function In the majority of examples in the corpus, =lku marks an eventuality as having commenced at some prior timepoint T1 and as continuing at T2, where T2 is synchronous

506 Legate (2008a) similarly argues that some Warlpiri lexemes are better understood as being in a state of flux between independent words and dependent enclitics. 507 Note that I use ‘eventuality’ (Bach, 1981) as a cover term for ‘events’ and ‘states’ since =lku freely marks either verbs or nominals irrespective of aspectual categories of imperfectivity and irrespective of ‘lexical aspect’ (and see Filip, 1999). 508 Temporal and pragmatic functions are also identified for a related clitic =lku in Warlmanpa (Browne, 2020).

10.4 Temporal and propositional clitics 

 613

with reference time.509 Generally speaking, the implication is that the proposition expressed by the eventuality is also true for the interval of time between T1 and T2. Example (1608) exemplifies the clitic =lku in its persistent state function, appearing on both (action) nominal and verbal hosts. An important piece of preceding discourse is provided for context. (1608)

a.

b.

c.

Mu-rlu=yanu ngantany ka-ngani, jayin-kurlu, karlarra this-erg=3pl.o man carry-ipfv.pst chain-prop west Balgo-jangka, Old Balgo-jangka. place_name-ela place_name-ela ‘These ones took the Aboriginal people in chains west from Balgo, from old Balgo. . .’ Jina minya tayimap=pa=lku jayin-ta=lu=yanu foot this tie.tr.compl=ep=still chain-loc=3pl.s=3pl.o ka-ngani. carry-ipfv.pst ‘They were taking them, their feet still tied up in chains.’ Yalu-jangka wurna=lku. that-ela journey=still ‘From there, (they) were still going.’ (PSM: TEN1-2017_007-02: 24339_35035)

Two persistent eventualities are described in (1608): that of being tied up, tayimap; and that of travelling, wurna. Both are explicitly established as having commenced at some prior time point T1 in the discourse. The use of the clitic =lku stipulates that both of the eventualities were also true at T2. The enclitic =lku can have scope over a clause (1609) or over a phrase (1610). (1609)

Karri-jura=lku=rlipa mula. be-hort=still=1pl.incl.s here ‘Let’s keep on staying here (i.e. not leave).’ (PSM: TEN1-2018_029-01: 2523991_2526128)

(1610)

Karri-jura=rlipa mula=lku. be-hort=1pl.incl.s here=still Let’s stay in this place still (i.e. not stay in another place). (PSM: TEN1-2018_029-01: 2571453_2574843)

509 The ‘persistent state’ function of =lku is not found for clitics of the same form in Warlpiri or Warlmanpa (Browne, 2020).

614 

 10 Auxiliaries, particles and modal constructions

An example of =lku appearing on an illocutionary nominal is provided in (1611). Here the resultant meaning is similar to imperfective verbal aspect. (1611)

Wurra=lku kaji warririku! wait=still condit close ‘Keep waiting until he moves closer.’ (MMN: TEN1-2017_026-01: 2644491_2650853)

The enclitic =lku is particularly frequent on temporal nominals, for example mayarra ‘again’, jayanta ‘a long time’, yalawardingki ‘(there) for a long time’. In these contexts, a persistent reading is often possible, as in jayanta=lku ‘still yet (for a long time)’ (1612); however, in other cases it is less clear that a ‘persistent state’ is being described (1613). (1612)

Yala parlaarnpa-rlu kit=ma-ni jayanta-lku. that turtle-erg stuck=get-pst long_time=still ‘(But) the turtle was still grabbing on to him (biting).’ (MMN: TEN1-2017_012-01: 122177_126310)

(1613)

Wirlinyi=wali mayarra=lku=rna jut=pu-ngku. hunting=then again=still=1sg.s set_off=hit-pot ‘I will set off hunting yet again.’ (TJL: TT76_2201: 1192797_1196158)

The enclitic =lku is clearly an outer clitic in that only the bound pronouns may occur outside of it. By contrast, =lku can follow the restrictive enclitic =mipa, as in (1614). (1614)

Nganimpa=rnalu kayirra-ngulu-jangka wangka-nya-ngurra 1pl.excl=1pl.excl.s north-abl-ela speak-pst-narr language_name=mipa=lku. language_name=restr=still ‘Due to (coming) from the north, we were speaking only Ngardi still (but later we learned Kukatja).’ (TJN: Manungka_02-022027: 2807559_2810322)

10.4.1.2 Subsequent eventuality function A second function of =lku is to mark temporal or logical progression, that is, ‘then’. Specifically, subsequent =lku serves to locate the eventualitiy of the host clause as falling in the consequent state of an antecedent eventuality. In this function =lku can mark both eventive predicates and stative predicates – in the latter case, the clitic =lku simply serves to place two states in a relationship of temporal progression. In its subsequent usage (unlike the persistent usage), =lku marks that a proposition did not hold at some preceding timepoint T1 but has only become true at T2. Consider (1615) and (1616).

10.4 Temporal and propositional clitics 

 615

(1615)

Kaji=pula ngapa mirti=nga-rnani-nyirra wati-kujarra-rlu condit=3du.s water bend_down=eat-ipfv.pst-narr man-dual-erg ngapa-kurra=lku=pulany kalkurr=pi-nya. water-all=then=3du.o collapse=hit-pst ‘While the two men were bending down to drink, into the water then, he pushed the two in.’ (TJN: LC21a: 1312260_1315743)

(1616)

Yalu-jangka kakarra nawu, bitumen=pa=lku. that-ela east now bitumen=ep=then ‘Then, (heading) east, and then on to bitumen (i.e. no longer on dirt roads).’ (PSM: TEN1-2018_009-01: 887705_889267)

As a marker of temporal progression the function of =lku is somewhat related to the elative marked demonstrative yalu-jangka ‘after that’ as used in (1616). However, =lku specifically sets up a change in eventualities: namely that some proposition was formerly not true and is later true, rather than simply being a marker of temporal progression ‘then, after that’. 10.4.1.3 Discourse-pragmatic function In addition to its purely temporal and logical functions in marking persistent or subsequent states, =lku has an additional function in signalling a change in topic, as in (1617). A similar discourse usage of a temporal clitic =biyang is reported in Jaminjung (Ritz & Schultze-Berndt, 2015); and compare =rru in Panyjima (Ritz, Dench & Caudal, 2012). (1617) Nya-nya=lu wanji-rla ita-nguniny, minya kirda-nguniny=pa=lku, see-pst=3pl.s where-loc small-dim this big-dim=ep=still ngaakman. frog ‘They looked where that small one (might be). (But) it’s the big frog!’ (MMN: TEN1-2019_005-06: 260258_266202 In the context of the frog story in (1617), the the boy and his friends are looking for a small frog. Instead, they come across the big frog that had been established as a topic in the narrative. Here the antecedent point of reference for =lku is not an overt proposition linked with a state of ‘bigness’ but rather a discourse topic of the ‘big frog being present’ rather than the little frog for whom they are searching. In its constrastive topic usage, =lku can also be sensitive to presuppositions in the wider discourse. Thus, in the following example (1618) which comes from an elicitation session, I asked for the Ngardi expression of the Jaru: lintij=ma-nan ‘he is flaking something’. Implicit in the elicitation context was that Patrick would respond with a distinct Ngardi utterance. However, the Ngardi translation involving the preverb lintij

616 

 10 Auxiliaries, particles and modal constructions

‘flake’ is a clear cognate for the Jaru. Patrick, being well aware of this, subsequently chose to express the Ngardi equivalent by marking the verbal predicate with =lku ‘still x’, affirming that a single linguistic form persists in the two languages. (1618)

TEN: Lindij=man-an? an? flake=get-prs prs ‘He is flaking it.’ (Jaru prompt) PSM: Lintij=pu-nganta yala. flake=hit-prs that ‘He is flaking it.’ til    lintij=pu-nganta=lku. still  flake=hit-prs=still ‘It’s still lintij=pu-nganta (in Ngardi).’ (PSM: TEN1-2018_035-02: 231118_234554)

Finally, the discourse particles walilku (§10.5.7) and nyamulku (1619) are semilexicalised instances of the discourse pragmatic function of =lku marking ‘subsequent states’ in that they encode meanings of ‘finished (now)’ (contrasted against the prior discourse). (1619)

Nyamu=lku=lu=rla karri-nyani. enough=then=3pl.s=3sg.obl be-ipfv.pst ‘That’s enough (time) of them staying with him.’ (PSM: TEN1-2017_024-01: 2113446_2115814)

10.4.2 Subsequent (=)wali ‘then’ The lexeme wali has a variable status as an independent particle and as an enclitic. Where the preceding word is /a/-final, the morphophonemic sequence /a-wa/ is typically realised as a phonetically long vowel [ɐː]. The primary function of =wali is to mark a subsequent temporal eventuality (§10.4.2.1), or a topic shift (§10.4.2.2). Additional functions of =wali include the marking of apodoses in conditional clauses (§10.4.2.3), the marking of quantity/degrees of gradeable predicates (§10.4.2.4), and a discourse-pragmatic function (§10.4.2.5). The discourse- terminating function of wali(lku) when appearing as an independent word is discussed separately amongst the interjections (§10.5.7). Phrases marked by =wali can often appear independent and prosodically separated from the main clause to which they are predicated. Thus, in the following example, ngantany-ju=wali is phonologically separated from the imperative clause whose subject is the unexpressed second person plural referent, made explicit by the postposed ngantany-ju=wali. Example (1620) also provides an instance of the contrastive

 617

10.4 Temporal and propositional clitics 

function available to =wali due to the emphasis placed on the pivot in the agent of the predicate: that is, previously non-Indigenous kardiya had managed the stations, and subsequently Indigenous ngantany were being handed control. (1620)

a.

b.

Yi-nya=nganpa kujarra=wali ngurra-kujarra, give-pst=1pl.excl.o two=then camp-dual ‘He gave the two stations to us, that owner.’ Marda-ka=lu, ngantany-ju=wali. have-imp=pl.s man-erg=then ‘You lot look after them, (all you) Aboriginal mob then.’ (PSM: TEN1-2016_005-03: 692777_)

owner-rlu. owner-erg

10.4.2.1 Subsequent eventuality function The temporal function of =wali may also be extended to one of logical causation. Thus, in (1621), the event ‘being full’ is positioned as a consequent state resulting from the events of the preceding – and juxtaposed – clause, ‘I will eat the meat’. (1621)

a.

b.

Ngu=rna pirdiny=ma-nku, ngu=rna nga-lku seq.aux=1sg.s arise=get-pot seq.aux=1sg.s eat-pot ‘I will take it out, then I will eat the meat.’ Pirda-kurlu=wali=rna ya-nku. full-prop=then=1sg.s go-pot ‘Full then, I will go.’ (TJL: TT762201: 1152038_1154506)

kuyi. meat

In a somewhat lexicalised construction with ngaka ‘later, after’, ngaka=wali has a restrictive meaning ‘it was only later on’, as in (1622). (1622)

Tumaj yirra-rni Bardi Bore kardiya-rlu yalu conj place-pst place_name white_man-erg that ngaka=wali windmill. later=then windmill ‘Because it was only later that the whitefella sunk that windmill Bardi Bore.’ (PSM: TEN1-2018_009-01: 275290_278848)

10.4.2.2 Topic shift function The temporal functions of =wali in marking subsequent eventualities are extended to a topic shift function which may or may not be used with contrastive effect. This is most evident when the constituent to which wali attaches is an argument of a predicate and is positioned as the new topic. This function of =wali can be exemplified by the turn-taking context as in (1623) as well as the novel information presented by the adjunct in (1624). In both contexts, both temporal and contrastive interpretations are

618 

 10 Auxiliaries, particles and modal constructions

clearly available and the interpretation of which sense is primary is determined by the wider discourse-pragmatic context. (1623)

Nyuntu=wali wangka-ya! 2sg=then speak-imp ‘You now (your turn), speak!’ (TJN: Manungka_01-035152: 908724_912769) [Context: Tjama, the speaker, passes the microphone to Maati.]

(1624)

a.

b.

Junga=lku=nganpangku=lu ma-ni motika-kurlu-rlu=wali. recog2=then=1pl.excl.o=3pl.s get-pst car-prop-erg=then ‘(Aftewards) they then took us by vehicle (i.e. before this time travel was by foot and horse).’ Yuwayi, nya-ngku=nta=wali nawu mitwok. yes see-pot=2pl.s=then now meat_work ‘Yes, now you lot will see the abbatoir.’ (PSM: TEN1-2018_038-02: 1331710_1339436)

The topic-shift function of =wali can further be observed in narratives. A speaker may use =wali in order to progress a narrative by presenting the next topic which is sequential in a linear order or novel in the context of the wider discourse. A directly parallel multifunctionality is observed for the Kriol particle nawu < now. (1625)

a.

b.

Yalu-rla=lku=lu marda-rni-nyirra. that-loc=then=3pl.s have-pst-narr ‘Then they looked after him there.’ Wurna-wangu=wali nyantu=wali. journey-priv=then recog1=then ‘They did not travel then.’ (MMN: Manungka_01-035155: 495197_497228) [Narrative context: a male had speared himself as a mourning ritual]

In this function, =wali has also been observed encliticising to an initial modal particle as in (1626). (1626)

Ngarda=wali jurr=yirra-rni kuja-ngku, an yaparti-nya kuyi hyp=then down=place-pst thus-erg conj run-pst meat pila=ma-ni jarrampayi, maitbi. follow=get-pst sand_goanna maybe ‘Perhaps then he might have put it down like so and run and chased that animal, the sand goanna maybe.’ (PSM: TEN1-2018_010-02: 1231360_1235534)

 619

10.4 Temporal and propositional clitics 

A distinct ‘topic-shift’ function of wali is its usage as a clause-initial discourse marker. Here, wali indicates a change in topic but also serves to bring two topics (the preceding and the new) into greater cohesion, to avoid otherwise abrupt topic switches. This function is very similar to usages of Kriol wal described in §10.6.5. Unlike Kriol wal, wali is generally not prosodically separated from the clause and may host the pronominal clitics.510 (1627)

a.

Kuriny ma=rna marda-rnani-nyirra two top.aux=1sg.s have-ipfv.pst-narr ‘(Playing cards) I had (a) two and (a) ten.’

an and

ten. ten

b.

Wali=rna kiji-rni, wakurra=rna=rla yi-nya. then=1sg.s throw-pst neg=1sg.s=3sg.obl give-pst ‘Well, I threw it out, I did not give it to her.’ (TJN: 17-8-00e)

10.4.2.3 Apodosis-marking function Closely related to its role in marking subsequent eventualities, the modal wali can also mark apodoses in conditional clauses (§11.1.2.2) equivalent to the English ‘if x, then y’. Marking of a ‘subsequent clause’ in this way is observed irrespective of whether the apodosis is sequenced after (1628) or before (1629) the protasis. The apodoses are bracketed in these examples. (1628) Kaji=lu wirrpa pirri=ya-nku [warlu=rna ma-nku wirrpa=wali]. condit=3pl.s many arrive=go-pot fire=1sg.s get-pot many=then ‘If lots of people arrive, I will get (much) more wood.’ (MMN: TEN1-201702601: 2090106_2098790) (1629) [Kanyi-ku=rna kunyarr jina-ngku=wali] kaji=yi paja-ku. crush-pot=1sg.s dog foot-erg=then condit=1sg.o bite-pot ‘If he bites me, then I will kick him with my foot.’ (PSM: TEN1-2017_026-02: 555777_560027) In (1628) and (1629), the constituent within the apodosis clause to which =wali attaches is the clause-final element. However, variation in this marking pattern is possible. Deviations from clause-final attachment seem to correlate with an element being in some sense focussed or placed in a contrastive relation to a preceding topic, as in (1630).

510 Additionally, it has not been observed with any ‘mitigation’ or ‘concessive’ functions as observed for Kriol wal (§10.6.5).

620 

 10 Auxiliaries, particles and modal constructions

(1630)

Kaji=yi=n mani yu-ngku ngu=rna=ngku condit=1sg.o=2sg.s money give-pot seq.aux=1sg.s=2sg.o mangarri=wali ma-ninja-nku. veg_food=then get-move&x-pot ‘If you give me some money, I will go get you some food.’ (PSM: TEN1-2017_026-02: 627086_633262)

10.4.2.4 Quantity/degree function There is some evidence that =wali can be used to mark a comparative (cf. English suffix -er) when encliticised to gradeable predicates such as kirda ‘big’ and ita ‘small’ (1631)–(1634). This is set in contrast to the fact that Ngardi lacks any productive inflectional morphology that marks degrees.511 Arguably the usages of =wali on gradeable predicates could simply be considered to be extensions of wali marking contrastive topics or subsequent states. (1631)

Ngaju nga=rna kirda=wali karri-nyani-nyirra. 1sg seq.aux=1sg.s big=then be-ipfv.pst-narr ‘I was somewhat bigger then (as compared to before).’ (BSE: TEN12017_014-01: 67455_69342)

(1632)

Ngunju, wanji=rna yirra-rni? Wakurra ita=wali. tobacco where=1sg.s place-pst neg small=then ‘(My) tobacco, where did I put it? No, (that one) is smaller.’ (KPA: Cataldi, 2011, p. 230)

(1633)

Ngaju=rna kirda=wali=rna wirrirl=ya-ni kakarra-ngurlu. 1sg=1sg.s big=then=1sg.s back=go-pst east-abl ‘I came back as a big woman then (from the east).’ (MMN: fieldnotes: TEN1-2017_029-02: 966220)

(1634)

Ma-nta=lu puliki majurrum kirda-kirda=wali mitwok-ku. get-imp=pl.s cattle muster.tr big-rdp=then meat_work-dat ‘Get the cattle, muster all the bigger ones for the meatwork (i.e. not the smaller ones).’ (PSM: TEN1-2018_038-02: 388984_392908)

511 Two exceptions warrant mention. The first is the rare ‘comparative’ enclitic =katu (§10.4.6). The second is the derivational suffix -kariny which can encode comparative meanings for certain gradeable predicates, e.g. kaninykariny ‘further inside’.

10.4 Temporal and propositional clitics 

 621

10.4.2.5 Discourse-pragmatic function Independent clauses containing =wali also can have an assertive, deontic meaning when combined with the potential inflection. These clauses are, at first glance, similar to multiple clauses where wali marks the apodosis in a conditional clause. In the following examples, protasis clauses can arguably be inferred as suggested in the bracketed translations. (1635)

Nyuntu=wali=n ya-nku. 2sg=then=2sg.s go-pot ‘You should go then.’ ‘(With conditions being as they are), you should go.’ (MMN: TEN1-2017_01901: 134478_136846)

(1636)

Tii ngaju-ku kupa-ku=wali. tea 1sg-dat cook-pot=then ‘(If we are having a break) I will make myself some tea then.’ (MDN: LC22a: 1181511_1183560)

As argued for English ‘then’, =lku appears to be able to link the utterance with which it is associated to a preceding (non-verbalised) proposition in the wider discourse context. For English, it is argued that clause-final ‘then’ has diverged awy from its original temporal meaning and has been assigned a new discourse-pragmatic function (Haselow, 2011; and see Traugott, 1989 on the repurposing of temporal markers more generally). It is quite plausible that a similar process is transpiring with Ngardi =lku. 10.4.3 Iterative =rra ‘again’ A clitic =rra, homophonous with the ‘hence’ verbal directional morpheme -rra, has an iterative function equivalent to English ‘again’.512 It appears most frequently on nominal (1637) and preverbal (1638) hosts.513 (1637) Kaniny=pa=rnalu ya-ni nganayi-kurra Kiyaly-ja=lku=rra. inside=ep=1pl.excl.s go-pst whatsit-all place_name-loc=then=again ‘We went down then to, you know, to Kiyaly again.’ (YMN: LC23a: 1717082_1720383) 512 Recall that the -rra hence morph appears both as a segmentable directional suffix and a portmanteau with certain TAM inflections in the verbal inflectional system (see §7.5). Neither of the directional morphemes -rni (hith) nor -rra (hence) occur productively on preverbs (unlike in neighbouring Warlpiri). 513 Enclitic =rra is possibly also related to certain inflected forms in the directional system (§5.5); e.g. kakarra ‘east’, kakarrara ‘further east, east again’ (where =rra has been dissimilated to -ra). Additionally consider the temporals ngaka ‘later’ and ngakarranyi ‘much later’.

622 

 10 Auxiliaries, particles and modal constructions

(1638)

Yalu-jangka=yi mayarra=lku pirdiny=pa=rra=karri-nya. that-ela=emph again=then arise=ep=again=be-pst ‘After that, he got up again.’ (PRN: FT1_b: 609912_612677)

Preverbs encliticised with =rra need not occur immediately preceding the inflecting verb but may be weak nexus preverbs and be separated from the inflecting verb by the bound pronouns, as in (1639). (1639)

Jungkut=pa=rra=rnalu pi-nya kayilu-mpayi. set_off=ep=again=1pl.excl.s hit-pst north-long_way ‘We set off a long way north again.’ (MNN: LC24b: 137761_141196)

A particularly interesting preverb host of =rra is wirrirl ~ wirirr ‘back’ (1640). In this case, =rra appears to operate with scope over the resultant event ‘to be in the same place’. That is to say, to be ‘back again’ means to be in the same place a second time, not to be enacting a returning event a second time, nor to be in the state of having returned a second time.514 (1640)

La-ni kala=rnalu lurrij, wurna jungkuj wirirr=pa=rra. pierce-pst asrt=1pl.excl.s finish journey set_off back=ep=again ‘We dug and finished up, set out travelling back again.’ (MNN: LC24b: 1184048_1187804)

This is futher clarified by the ability of =rra to be used with wirrirl ~ wirirr in contexts which are reciprocal in nature rather than strict iterations. Consider (1641). (1641)

Kaji=yi=n la-nngi ngaju ngu=rna=ngku condit=1sg.o=2sg.s pierce-irr 1sg seq.aux=1sg.s=2sg.o wirirr=pa=rra la-nngi. back=ep=again pierce-irr ‘If you spear me, I would spear you back (again).’ (MMN: TEN1-2018_02601: 2840087_2848260)

10.4.4 Additive =rlangu ‘also’ The additive enclitic =rlangu is homophonous with the dyadic kinship suffix -rlangu (§5.10.1.1). This enclitic encodes inclusive members of a larger set, approximately equivalent to the English ‘also, as well, along with x’ as in (1642) and (1643). 514 A similar phenomenon is observed in the English ‘back again’. That is, ‘there and back again’ does not imply multiple returning events but simply to be in the same place again.

10.4 Temporal and propositional clitics 

(1642)

Nga-rnani kuyi waku, yawi, nguku=rlangu, waku. eat-ipfv.pst meat neg poor_thing water=also neg ‘The poor thing, he was not eating meat, no water either.’ (NMN: Cataldi, 2011, p. 288)

(1643)

Ka-ngani=rna minya=rlangu. carry-ipfv.pst=1sg.s this=also ‘I brought this one too.’ (PSM: TEN1-2017_013-02: 1245155_1247651)

 623

The additive can also be used to add further specification of a set, equivalent to the English ‘such as, for instance’ (1644). (1644)

Ma-nu medisin=pa=rlangu nga-rnu-ku kungkurr-ku.515 get-inf medicine=ep=also eat-inf-dat sick-dat ‘In the old days (we) would use such things as roots as a type of medicine too, to take for colds.’ (YDN: Cataldi, 2011, 134)

Two NPs, both encliticised with =rlangu, can share the same grammatical relation and are interpreted as ‘x as well as y’ as in (1645). (1645)

Walya=rlangu larru, blanket=pa=rlangu larru ngapa-jangka. ground=also wet blanket=ep=also wet water-ela ‘Both the ground and the blankets are wet from the rain.’ (TJN: Cataldi, 2011, p. 308)

The additive is primarily found on unmarked nominals but it also appears on casemarked nominals (1646) and even inflected verbs (1647). (1646)

Maju, ya-nku=rlipa wurna ngurra-kurra=rlangu. certainly go-pot=1pl.incl.s journey camp-all=also ‘I say, let’s leave and go back home!’ (MDN: LC42_20000926_a: 975068_978342)

(1647)

Kuja=lu ma-nani-nyirra=rlangu. thus=3pl.s get-ipfv.pst-narr=also ‘They used to get it as well.’ (Cataldi, 2011, p. 35)

515 This example is anomalous in that the main verbal predicate is non-finite. It is likely that this is a Warlpiri code-switch since the form -nu is the appropriate Warlpiri past tense inflection.

624 

 10 Auxiliaries, particles and modal constructions

Under the scope of the clausal negator wakurra, =rlangu takes on the negative meaning of ‘not even’, as in (1648). (1648)

Wakurra=rnalu wirlinyi=rlangu ya-nani. neg=1pl.excl.s hunting=also go-ipfv.pst ‘We didn’t even hunt here.’ (PSM: TEN1-2017_013-02: 949136_951371)

10.4.5 Restrictive =mipa The clitic =mipa has a delimiting or restrictive function ‘only x’ or ‘just x’. The restrictive shows some overlap in functionality with the delimitive (=)jala but is distinguished formally by the fact that it is strictly a clitic. There are no constraints on what clausal constituent the restrictive clitic may have logical scope over. It is most frequently found encliticised to NPs in various clausal roles, including subjects (1649), objects (1650), indirect objects (1651), and adjuncts (1652).516 The restrictive serves to constrain the conceivable range of referents which the speaker imagines might be expected in a given thematic role. (1649)

Mani=lu ma-nani-nyirra yirda-ngku grog-purupuru, money=3pl.s get-ipfv.pst-narr other-erg grog-many kardiya-rlu=mipa. white_man-erg=restr ‘Some others would get some money and lots of grog, only the whitefellas got it.’ (PSM: TEN1-2017_007-02: 401437 _406811)

(1650)

Ti=mipa=rnalu nga-rnungka-nya. tea=restr=1pl.excl.s eat-x&move-pst ‘We drank only tea and left.’ (YMN: LC23a: 1495155_1497947)

(1651)

Warlu=rna ma-ni pirranginti ngalijarra-ku=mipa. fire=1sg.s get-pst afternoon 1du.excl-dat=restr ‘Yesterday I got some wood for just you and me (i.e. not for anyone else).’ (PSM: TEN1-2017_026-01: 1460158_1465145)

(1652)

Jipiny ma=rna kuyi nga-rnanta, jalangu=mipa. first top.aux=1sg.s meat eat-prs today=restr ‘I am trying this meat for the first time (I am eating this meat only today).’ (BSE: TEN1-2018_017-03: 1231373_1233335)

516 The corpus lacks examples involving subjuncts.

10.4 Temporal and propositional clitics 

 625

There is at least one example of =mipa encliticising to a verbal predicate (1653) which can be compared with the near minimal pair (1654). An example of a preverb host is provided in (1655). (1653)

Minya lukarrara=rnalu jama-rnanta=mipa. this seed=1pl.excl.s grind-prs=restr ‘(Now) we are only grinding these seeds (i.e. we are not cooking or eating them).’ (MMN: TEN1-2017_013-02: 1523346_1527276)

(1654)

Jalangu=rnalu jama-rnanta mungily=mipa. today=1pl.excl.s grind-prs seed=sembl ‘Today we are grinding only the mungily seeds (i.e. not any other type).’ (MMN: TEN1-2017_013-02: 1567974_1575147)

(1655)

Ya-nani=rnalu wirlinyi=mipa. go-ipfv.pst=1pl.excl.s hunting=restr ‘We went around just hunting (i.e. not doing anything else).’ (PSM: TEN12017_013-01: 1740244_1743083)

10.4.6 Comparative =katu The functions of the comparative enclitic =katu co-vary with the the type of predicate to which it attaches. On non-gradeable predicates, its function appears to be that of specification or intensification: ‘really x, precisely x’ as in (1656) and (1657). In cases where it encliticises to a gradeable predicate, as in (1658) and (1659), it functions as a positive comparative degree marker: ‘more x, better x’. Limited examples predominantly come from a single speaker and her usage of this form may represent a Warlpiri influence on her Ngardi idiolect (amongst her broader multilingual repertoire).517 (1656)

Nganayi=katu? whatsit=compar ‘What particularly?’ (TJN: Cataldi, 2011, p. 119)

(1657)

Shopping=ma-nta maruka-rlu-katu. shopping=get-imp mean-erg=compar ‘Do the shopping, you really mean thing!’ (TJN: Cataldi, 2011, p. 195)

517 Warlpiri has =katu, glossed ‘greater quantity and better quality of something; more; better; best; good (Laughren et al., 2007, p. 317). This enclitic is mentioned but excluded from Bowler’s (2016) analysis of comparatives in Warlpiri due to limited data.

626 

 10 Auxiliaries, particles and modal constructions

(1658)

Kapan=katu yirra-rnu-ku. easy=compar place-inf-dat ‘It is easier to turn on.’ (TJN: Cataldi, 2011, p. 119)

(1659)

Yu-ngka=yi wuruju=katu. give-imp=1sg.o good=compar ‘Give me a better piece! (of food)’ (PSM: TEN1-2017_018-02: 892083_894817)

For (1659), Patrick Smith and Marie Mudgedell translated wuruju-katu with the Aboriginal English ‘more better one’ and conferred that it was more or less equivalent to wuruju-karrikarri, involving the ‘attenuative’ suffix (§4.4.3). 10.4.7 Incremental =warru The clitic =warru has an ‘incremental’ or mild ‘emphatic’ function that can modify verbal predicates (1660), NPs typically depicting states (1661), quantifiers (1662), and temporals (1663). When marking nominals, its semantic function appears to be similar to the nominal intensifier suffix -nyayirni ‘very’ (§4.4.6) but impressionistically appears to be distinguished by a slightly reduced degree of intensification, that is, ‘somewhat (more) x’. (1660)

Kaniny-jarra la-ni, wunpurr=paja-rni=warru=nyanu. inside-assoc pierce-pst wound=bite-pst=incr=refl ‘He speared (himself) inside, he wounded himself rather badly.’ (Maati: Manungka_01-035155: 492155_495226)

(1661)

Juwal=warru=rna karri-nyani lidlbit. long=incr=1sg.s be-ipfv.pst little_bit ‘Then I was a bit taller, somewhat.’ (MMJ: Maati3)

(1662)

Nga=rnalu wurrkaru=warru karri-nyani yalu-ngka. seq.aux=1pl.excl.s few=incr be-ipfv.pst that-loc ‘A fair few of us were staying there.’ (MMN: TEN1-2016_005-01: 161367_164693)

(1663)

Nyanyi=warru, Jungurrayi, yuwayi, nyanyi. before=incr ♂subsection yes before ‘Quite a while ago, Jungurrayi, yeah, a long time ago.’ (PSM: TEN1-2018_038-01: 161656_170175)

10.4 Temporal and propositional clitics 

 627

10.4.8 Vocative =wu The vocative clitic =wu is a positionally restricted clitic typically used in speech acts directed at an addressee over a distance or in speech contexts in which the addressee’s attention is of particular concern to the speaker. The vocative is restricted to a word-final position, irrespective of the morphological makeup of its host—including wordforms to which bound pronouns are encliticised, as in (1664). (1664)

Wurna ya-nku=rlipa=wu! journey go-pot=1pl.incl.s=voc ‘Let’s set off!’ (YDN: LC49b: 1226208_1229483)

Utterances terminating with the vocative clitic also tend to exhibit a characteristic high falling intonational contour with a pitch peak typically occurring on the penultimate syllable of the utterance-final prosodic word (see §2.7.2). The function of the vocative is to highlight the speaker-addressee relation: the function of the clitic is not so much to enhance the content of the message but to attact the attention of the addressee. For this reason, the vocative clitic is frequently found on commands. This includes imperatives (1713), prohibitives (1722), directives (1737), and action nominals used with the illocutionary force of a command (1665). (1665)

Wurna=wu, ya-nku=rlipa! journey=voc go-pot=1pl.incl.s ‘Let’s head off!’ (YDN: Manungka_02-022027: 162849_164945)

The vocative =wu is not constrained to command speech acts but can also appear in declarative contexts (1666). (1666)

Wirrpa=lku=lu=rla luyurr=karri-nyanta=wu, many=then=3pl.s=3sg.obl sad=be-prs=voc ya-nku=lu=rla. wirrirl. go-pot=3pl.s=3sg.obl back ‘Many people are sorry for that one, they will go back for him.’ (DMN: LC42b: 793596_799864)

The vocative clitic also frequently appears on interjections (§10.5) such as the negative wakurra as in (1757), and the exclamative yurru ‘look out’ (1743).

628 

 10 Auxiliaries, particles and modal constructions

10.4.9 Emphatic =yi The function of this clitic is not well understood.518 It appears most frequently on the distal demonstrative yala and may have an emphatic or delimitive function ‘exactly x, definitely x’ as in (1667) and (1668). It bears some formal relationship to the enclitic in Warlpiri =ju/=ji which has two functions: a topic/definite marker on nominals or a ‘phonological extension’ (Nash, 1986, p. 56).519 Emphatic =yi is homophonous with the 1sg.o bound pronoun. (1667)

Wakurra yala=yi yangi. neg that=emph one ‘Not over there in that particular place.’ (PSM: TEN1-2017_013-2: 1376467_1378812)

(1668)

Puurda, puurda, yala=yi yala, yala=yi yala. yam yam that=emph that that=emph that ‘Bush potato, bush potato, that’s the one, that’s it, that’s the one alright.’ (MRN: LC35b: 1373355_1377743)

10.4.10 Presentational =nya Cataldi (2011, p. 236) records an enclitic =nya which she glosses as ‘emphatic, presentational’. There are exceedingly few examples of this enclitic in the corpus. Based on the limited examples, its function appears somewhat similar to the emphatic enclitic =yi. Examples are (1669) and (1670). (1669)

Yalu-rla=nya karri-nya=rnalu. that-loc=prtl be-pst=1pl.excls ‘We stayed right there.’ (NMN, Cataldi, 2011, p. 236)

(1670)

Minya nya-ngka yuka kalkurni yala yuka jala=nya griyinwan. this see-imp grass this_side that grass delim=prtl green_one ‘(Describing ngirili) Look at all this grass along the side here, it’s this one, this green one.’ (Cataldi, 2011, p. 236)

518 Cataldi in fact variously identifies -yi as a ‘locative case’ variant on the demonstrative yala (2011, p. 27), as an emphatic (2011, p. 3), and as a euphonic clitic (2011, p. 3). 519 The variation between =ju and =ji in Warlpiri is dialectal in origin. Despite exhibiting a lenition correspondence with the Ngardi form, a putative historical connection between the Warlpiri and Ngardi forms is still problematic. Unlike many of its Ngumpin neighbours, Ngardi lacks many of the same historical lenition patterns which might explain a correspondence between /c/ and /j/.

10.5 Interjections 

 629

10.5 Interjections Interjections are independent, uninflected words which constitute independent utterances and are assigned their own intonational units. Interjections cover a broad range of speech act functions, including agreement (yuwayi) and hortation (wurna); as well as the expression of various emotions such as alarm (waraa), and compassion/ sadness (yawiyi). Interjections generally do not take any inflection but some may host the emphatic =yi or vocative =wu enclitics. Some interjections are doubly classified as lexemes from other word classes. These include various particles already described (e.g. wakurra neg) but also include a recognitional nominal junga ‘you know the one’; an adverbial preverb lurrij ‘finished, completely’; and illocutionary nominals wurra ‘wait’ and wurna ‘journeying, setting off’ as discussed in §10.7.3. Many of these interjections are widespread in either Ngumpin-Yapa and/or Western Desert languages and are not exclusive to Ngardi. Some are identified by Ngardi speakers as words from other languages which are nevertheless borrowed or code-switched into Ngardi speech contexts – for example wiya ‘no, nothing’ (found in various Western Desert languages). Some interjections described here are/were not used by all Ngardi speakers – some of these details are tracked in the footnotes. Table 137 arranges interjections according to their meaning class. The following subsections provide a brief description and exemplify interjections hitherto undescribed. Cross-referencing is provided for locating relevant discussion/examples of each form. Table 137: Interjections. Meaning Class

Form

Meaning

Agreement

yuwayi, yuwawu, yuwa yu(u), yawu (§10.5.1) wali (§10.4.2) junga (§10.3.7) kujarnawu,(§10.5.2) maju (§10.3.3) ngayi (§10.3.4)

‘yes, alright (affirmation)’ ‘ok, alright, indeed’ ‘true’ ‘that’s it, that’s the one now’ ‘I say, indeed, of course, really’ ‘indeed, truly, yeah’

Disagreement

wakurra (§10.8.1) waku (§10.8.3)

‘no’ ‘no’

Compassion

yawiyi (§10.5.5)

‘poor thing’

Commands/ Hortation

wurna (§10.7.3.1) wurra (§10.7.3.2) yurru (§10.7.3.3)

‘(let’s) go’ ‘go on’, ‘come on’ ‘wait!’ ‘look out!’

Attention

kanti (§10.5.2)

‘hey, here, I say’

630 

 10 Auxiliaries, particles and modal constructions

Table 137 (continued) Meaning Class

Form

Meaning

Interrogation (tags)

mayi? (§10.2.1) wayi? (§10.2.2) ngayi? (§10.3.4) kari? (§10.5.5) jangu? (§10.3.7)

‘see?’ ‘isn’t it?’ ‘really? is that so?’, ‘oh indeed? ‘perhaps?’, ‘maybe?’ ‘perhaps?’, ‘maybe?’ ‘true?’

Terminator/ Farewell

wali(lku), nyamu(lku), lurrij (§10.5.7)

‘that’s it’, ‘that’s all’, ‘finished’

Reactions to behaviours/ events

warriwarri 520 warra! (§10.5.8)

a response to a lewd comment ‘hey!’, ‘ah!’ (cry of pain, surprise, discomfort) ‘oh!’

wuu!

10.5.1 Yuwayi, yuwawu, yuwa, yuu, yawu ‘yes, ok, alright’ An affirmative exclamative with a form similar to yuwayi is widespread in Australian languages (Dixon, 2002, p. 66). In Ngardi, examples of yuwayi can be found throughout the corpus as the generic response to polar questions, as well as indicating agreement either with someone else’s (1671), or one’s own (1672) propositions. (1671) Yuwayi. Yuwayi, ngaju-ku=yi kirda=ma-ni jamirdi-rlu, yes yes 1sg-dat=1sg.o big=get-pst FF-erg ngaju-punta-rlu... 1sg-ppos-erg ‘Yeah, that’s right, my father brought me up. . .’ (BSE: TEN1-2017+014-01: 10437_16380) (1672)

a.

b.

Mula-wana=rlangu kuyu ngu=rna=yanu tarruku, here-perl=also meat seq.aux=1sg.s=3pl.o sacred luwa-rnani-nyirra kuyu-ku. shoot-ipfv.pst-narr meat-dat ‘Right along this place too I used to kill meat, sacred meat for the elders.’ Yuwayi, yala-rla. Jukurr-kurlu ma=rna. . . yes that-loc Dreaming-prop top.aux=1sg.s ‘Yeah that’s it. I have a Dreaming and a Law. . .’ (MAM: TEN1-2016_ 002-01: 461494_465901)

520 This word form has multiple senses: i) genitalia, ii) taboo name (cf. kumunyjayi), iii) urinating, and iv) someone of the other generation level (Cataldi, 2011, p. 323).

10.5 Interjections 

 631

There is significant variation in the phonological form of yuwayi. The trisyllabic forms (yuwayi and yuwawu) are plausibly comprised of the disyllabic form yuwa- and the emphatic=yi (§10.4.9) and vocative =wu (§10.4.8) clitics, respectively. Due to the rare usage of a bare disyllabic root form yuwa, I have chosen to analyse yuwayi and yuwawu as synchronically monomorphemic forms. Impressionistically, the latter is found in more emphatic contexts and can be used with a distinctive intonational contour involving a pitch and intensity peak on the penultimate syllable. I have no explanation for the form exhibiting metathesis (yawu) which is itself homophonous yawu ‘fish’.

10.5.2 Kujarna(wu) ‘that’s it, that’s the one, (you) got it’ The interjection kujarnawu and its Kukatja equivalent tjitjarnawu are both used by Ngardi speakers to indicate agreement. Specifically, such interjections are frequently used to reaffirm or re-identify something with discourse prominence, equivalent to the English ‘yeah that’s the one’ or ‘yeah just like that, (you) got it!’. Both Ngardi and Kukatja forms are compounds derived from a demonstrative (a recognitional demonstrative kuja in Ngardi and a proximal demonstrative tjitja in Kukatja) and the Kriol particle nawu < English ‘now’. I frequently heard this interjection during elicitation as a response to questions on my part as a language learner. This interjection was also used by speaker’s to affirm their own (1673) or each other’s speech (1674). (1673)

a.

b.

(1674)

Tardarr=ya-nkura-rni kaniny-kaniny! enter=go-hort-hith inside-rdp ‘(We say), let him come inside!’ Kujarnawu! that’s_it ‘Yeah that’s the one!’ (PSM: TEN1-2017_026-01: 3061196_3062287)

PSM:

Kupa-rnngi parda=rna. cook-irr dub=1sg.s ‘I should have cooked it.’ (PSM: TEN1-2018_033-03: 1009303_1011439) MMN: Yeah tjitjarnawu. yes that’s_it ‘Yeah, that’s the one!’ (MMN: TEN1-2018_033-03: 1016669_1019096)

10.5.3 Maju ‘certainly’ The lexeme maju was already described as a propositional particle (10.3.3). As an independent interjection it expresses emphatic agreement, as in (1675).

632 

 10 Auxiliaries, particles and modal constructions

(1675)

Maju, wurna=rlipa ya-nku! certainly journey=1pl.incl.s go-pot ‘Yeah! Let’s go!’ (MDN: LC22b: 836780_841858)

10.5.4 Kanti ‘oi, hey!’ 521

The interjection kanti is used to attract someone’s attention or to express a state of surprise. It occurs as an independent utterance but it may also intervene within and between utterances, especially if the speaker notices the addressee is not paying attention. Examples are (1676) and (1677). (1676)

Ngarri-rni=pulany, “Kanti, kuyi, nga-nja=pula, minya-rla.” tell-pst=3du.o hey meat eat-imp=du.s this-loc ‘He said to those two, “Hey! Eat this meat here!”’ (MMJ: MAATI1)

(1677)

Ngu=lu wangka-nyani, “ya-nku=rlipa wirirr ngi? Kanti seq.aux=3pl.s speak-ipfv.pst go-pot=1pl.incl.s back tag hey kakarra nyila.”522 east there“ ‘They were saying, “Let’s go back, yeah?, Hey! (Look) there in the east!”’ (MMN: TEN1-2016_005-01: 477099_486131)

10.5.5 Kari ‘perhaps’ The interjection kari is rare in the corpus and has only been found in interrogative contexts where it functions as a suggestive tag, i.e. ‘perhaps?’, ‘I reckon?’.523 It is postposed to the clause or phrase containing the proposition it is modifying, as illustrated with pinij kari? ‘finished yeah?’ in (1678). In this function, it is equivalent to the more frequent tags ‘mayi?’, ‘kurlu?’ or ‘ngi?’ (see §10.9.2).

521 This exclamative is homophonous with a lexeme kanti 1. ‘stone, (white) chert’; 2. ‘stone knife’; 3. ‘spear tip’ which is used predominantly in Jaru (cf. the place name Kantiyaru stone.prop, place name for ‘Lily Hole Tank’) and Yarlpiri Warlpiri (Laughren et al., 2007, p. 258). It is occasionally used with these senses by Ngardi speakers but these senses are generally covered by the Ngardi and Kukatja word jimarri (Valiquette, 1993, p. 287), and the Ngardi and Walmajarri word jimpirla. 522 The italicised form in this utterance is a Kukatja code switch. 523 The interjection kari was not found in the speech of Ngardi speakers recorded in 2015–2019 nor in Tsunoda’s corpus. A similar form is found in Warlpiri where it is analysed as a propositional particle expressing ‘asserted truth, based upon personal experience’.

10.5 Interjections 

(1678)

Ya-ni=lu jayanta kuja=lu. Pinij kari? go-pst=3pl.s long_time thus=3pl.s finish perhaps ‘So they went for good. The end, I reckon? The end.’ (YDN: Manungka_02-022027)

 633

Pinij. finish

10.5.6 Yawiyi ‘poor thing!’ The exclamative yawiyi is used to express various sentiments of sadness, concern, sympathy, pity and compassion. Many examples occur throughout the corpus and one is provided in (1679). This exclamative can also be used for expressions of endearment and familial affection (1680), or fond nostalgia (1681). (1679)

Ya-ni=rnalu ngurra-kurra wirrirl, yumpaly nawu, habares, go-pst=1pl.excl.s camp-all back tired now have_a_rest yawiyi. poor_thing ‘We went back to camp, tired now, (to) have a rest, poor things.’ (PSM: TEN1-2018_038-01: 1641480_1648162)

(1680)

Warluwurra-ku=rlangu, jaja yawiyi. name-dat=also FF poor_thing ‘Grandmother for Warlawurru too, dear one.’ (NNA: ngurungka 30_4_91)

(1681)

Watimelon nga-rni=lu kinki=karri-nya. watermelon eat-pst=3pl.s full=be-pst Yuwawu, titri nga-rni=lu, yawiyi. yes dates eat-pst=3pl.s poor_thing ‘(With those two whitefellas) They (Aboriginal people) would eat watermelons until they were full up. Yeah, they would eat dates (too). Awh, (I miss that).’ (PSM: TEN1-2017_007-02: 312218_315544)

It appears the lexeme yawiyi is doubly classified as a manner nominal, due to its ability to take ergative case-marking when functioning as a depictive secondary predicate (§9.3) as in (1682).524 This usage is infrequent. (1682)

Ngarda=nyanu kuma-rnngi, yawiyi-rlu. hyp=refl cut-irr poor_thing-erg ‘In mourning, he cut himself.’ (DMN: LC42b: 1280524_1283453)

524 The same is true of a similar interjection/nominal wiyarr in Warlpiri (Laughren et al., 2007, p. 1518).

634 

 10 Auxiliaries, particles and modal constructions

10.5.7 Wali(lku), nyamu(lku), lurrij, pinij ‘that’s it, finished, enough’ The lexeme wali has already been described as a propositional clitic with temporal functions (§10.4.2). As an interjection wali always appears as an independent word and may host the temporal enclitic =lku, (§10.4.1). Its function as an interjection is to terminate a stretch of discourse or finalise an interaction and is equivalent to the English, ‘that’s all, that’s it, ok’. When used with interrogative intonation, it can be used to query an interlocutor as to the end of a stretch of talk or an entire conversational exchange (i.e. walilku? ‘is that all?’). A response in return ‘walilku!’ can effectively serve as a farewell, approximately equivalent to English ‘ok then/alright, see you!’. In its terminative function, wali(lku) is often paired with wuruju ‘good, ok, alright’ (1683) or yuwayi ‘yes, ok, alright’ (1684). (1683)

Ngu=rlipa wuruju-rlu marda-rnanku minya Wirrimanu, seq.aux=1pl.incl.s good-erg have-ipfv.pot this place_name yuwayi. Wali=lku, wuruju. yes term=then good ‘We will keep this Wirrimanu in a good way, yes. That’s it, finished.’ (MMN: TEN1-2017_005-02: 424379_427621)

(1684)

a.

b.

c.

Kunyarr-ja murrku. dog-loc boy ‘The boy with the dog.’ Ka-nya=pula wirrirl. carry-pst=3du.s back ‘Those two took him back.’ Yuwayi, wali=lku. yes term=then ‘Yeah, that’s it.’ (MMN: TEN1-2017_016-02: 385859_388775)

As an interjection, there is variation as to whether the enclitic =lku surfaces. Its presence appears to be optional:525

525 According to Marie Mudgedell, the non-encliticised interjection wali is Walmajarri while walilku is Ngardi. Nevertheless, wali is used just as frequently (if not more so) than walilku and so it is included here. Other Ngardi speakers also made regular use of widespread lexemes nyamu or lurrij. Interestingly, the terminating interjections wali and marndaj are described as highly shibbolethic (former for Jaru, latter for Wanyjirra) for speakers of Wanyjirra by Senge (2015, p. 120) despite speakers using both frequently.

10.5 Interjections 

(1685)

 635

Yuwayi, wali. yes term ‘Alright, that’s it (finished).’ (PSM: TEN1-2016_005-03: 718223_719117)

As an interjection, wali(lku) is sometimes replaced by synonyms nyamu, lurrij (1686) or the Kriol pinij (1678).526 (1686)

Nyamu. Lurrij. finish completely ‘The end, that was it.’ (TJN: LC20a)

10.5.8 Waraa, warraa ‘wow, oh no, alas!’ The exclamative waraa is used as an expression of surprise or shock.527 This intjerection can be used in contexts both where the state of shock is positively assessed (e.g. astonishment as in (1687)) and negatively assessed (e.g. fear/horror as in (1688)). It can also express disappointment or regret (1689). It was also informally observed as an expression of discomfort or pain. (1687)

a.

b.

Parnkaj=ma-ni=lu yala iji. open=get-pst=3pl.s that gift ‘They opened that present.’ Nya-nganta=lu nganayi minya, “Waraa!”. see-prs=3pl.s whastit this wow ‘They look at that whatchamacallit, “Oh wow!”’ (MMN: TEN1-2019_005-06: 34366_42604)

(1688)

Waraa, muwurnmuwurn ma yapartu-nganta, yala lingka. wow heedless top.aux run-prs that snake ‘Oh no! (that kid) is running about unwittingly, (he doesn’t know about) that snake!’ (MMN: TEN1-2018_21-01: 1854874_1861432)

(1689)

Waraa, pampa-ngku=rna kiji-rni, wurra=rna=rla nya-nya. alas blind-erg=1sg.s throw-pst later=1sg.s=3sg.obl see-pst ‘Alas, I threw it out without looking, I should have looked first for it.’ (TJN: LC56b)

526 The lexeme nyamu is widespread across northern Western Desert languages. 527 This interjection could also be rendered orthographically as wardaa, given the inter-speaker variation between a tap/trill, flap and rhotic continuant. Cataldi also includes single instances of emphatic =yi (waraayi) and vocative =wu (waraawu) encliticised forms in her transcripts.

636 

 10 Auxiliaries, particles and modal constructions

10.6 Kriol-origin particles In this section, a number of borrowings from Kriol that function like propositional or modal particles in Ngardi are described. These include maitbi (§10.6.1), yuno (§10.6.2), onli (§10.6.3), nawu (§10.6.4), and wal (§10.6.5). This is not an exhaustive list but includes the most frequent borrowings used in Ngardi speech in the corpus. 10.6.1 Dubitative maitbi ‘maybe, perhaps’ The Kriol-origin lexeme maitbi has functions in Ngardi syntax as a modal particle. In its clausal/sentential usage, maitbi functions entirely analogously to the dubitative particle parda – it conveys epistemic uncertainty about an event. Like parda, it exhibits both clausal (1690) and phrasal (1691) scope. (1690)

Maitbi=rna ya-nku Tanami-kurra, ai dano. maybe=1sg.s go-pot place_name-all 1sg.s don’t_know ‘I might go to Tanami, I don’t know.’ (SGO: T1091_1: 393423_396137)

(1691)

Maitbi nyampa=mayi. maybe what=uncert ‘Could be something or other.’ (YRN: LC33a: 1380012_1381295)

Kriol maitbi operating with clausal scope is even found co-occurring with parda as in (1692) and (1693).528 (1692)

Maitbi parda=n ya-nku warlupaka-rla, wirrirl. maybe dub=2sg.s go-pot lunch-loc back ‘Perhaps you might go back at lunch.’ (JLI: TT75_1301: 1441519_1443818)

(1693)

Maitbi ngu=n wirrirl parda ya-nku yalu-jangka. maybe seq.aux=2sg.s back dub go-pot that-ela ‘Perhaps you’ll go back to Halls Creek from there/after that.’ (JLI: TT75_1301: 1478700_1480488)

528 Compare the co-occurrence of kuyungurla ‘maybe’ with pa-rta- ‘MR1-dub’ (dubitative auxiliary) in Walmajarri (Hudson, 1978, p. 84).

10.6 Kriol-origin particles 

 637

10.6.2 Discourse marker yuno The Kriol-origin lexeme yuno appears in Ngardi speech as a discourse marker. Its function appears to have a number of similarities with the discourse/pragmatic functions of ‘you know’ in varieties of English (Erman, 2001; Ostman, 1981) and can variably be observed as signalling shared or assumed knowledge (either of a proposition or its context) (1694) or as a more general marker of a pragmatically expected event, as in (1695). (1694)

a.

b.

(1695)

That’s kuli, that’s kuli that one. that_is angry that_is angry that one ‘That (word) is like when you’re angry, that’s ‘cheeky’ that one.’ Jarrak=pi-nya=yanu yuno. demand=hit-pst=3pl.o you_know ‘He is yelling at/demanding things from them, like that.’ (PSM: TEN1-2017_010-03: 99098_101786)

Yalu-jangka=lu ya-ni-nyirra wirlinyi pirranginti-nginti yuno. that-ela=3pl.s go-pst-narr hunting afternoon-rdp you_know ‘After that we went hunting later on.’ (TJN: Manungka_01-035151: 1136962_1140283)

Unsurprisingly, yuno was especially frequent in responses to elicitation contexts as a type of metalinguistic comment on a prior utterance. For example, in (1696), Patrick Smith provided a Walmajarri equivalent for the Ngardi expression ‘the sun is rising’, terminating his utterance with yuno. In this case Patrick Smith was not using yuno to confirm prior knowledge of his interlocutor (since I was not expected to have prior knowledge of the Walmajarri translation) but simply as a means of affirming the validity of his prior talk. (1696)

Walmajarri, “kakarni-mpala nyirra=la-nana”, yuno. language_name east-along rise=pierce-prs dm ‘(In) Walmajarri, (they say) something like kakarnimpala nyirra=lanana (for ‘the sun is getting up’) you know.’ (PSM: TEN1-2017_009-02: 443579_447573)

10.6.3 Restrictive onli The Kriol-origin particle onli (< English ‘only’) has been integrated into Ngardi syntax as if it were a propositional particle. It has a restrictive function ‘just x, only x, x to

638 

 10 Auxiliaries, particles and modal constructions

the exclusion of y’ akin to the propositional enclitic -mipa. It typically occurs clauseinitially as in (1697) and (1698) where it hosts (overt) bound pronouns. (1697)

Onli=rna lidlbit purda=nya-ngani Ngardi. restr=1sg.s little_bit listen=see-ipfv.pst language_name ‘I understand only a little Ngardi.’ (THS: TEN1-2017_003-05: 90233_96882)

(1698)

Onli hawij-ja=rnalu wanti-nyanta nyurnu. restr house-loc=1pl.excl.s fall-prs sick ‘We get sick only in whitefella houses.’ (TJN: Manungka_01-035151: 802667_806037)

10.6.4 Contrastive nawu ‘now, now unlike before’ The Kriol-origin particle nawu (occasionally na) (< English ‘now’) appears to have related temporal and dicourse functions (cf. Hudson, 1985, p. 122 who terms it an ‘emphatic particle’ in Fitzroy Valley Kriol).529 In its temporal function, nawu marks an event as occurring simultaneous with a given reference time. In its discourse function, it has a weakly contrastive function, positioning a new event in contrast to a preceding state of affairs. Examples are (1699)–(1701). (1699)

a.

b.

c.

Motika-kariny-ja ya-ni=rnalu nawu. . . car-other=loc go-pst=1pl.excl.s now ‘We came in another vehicle then.’ Yala nawu=rlipa karri-nyanta, minya-rla=rlipa that now=1pl.incl.s be-prs this-loc=1pl.incl.s karri-nyanta niupala-rla Balgo-ngka minya, Wirrimanu, be-prs new.adj-loc place_name-loc this place_name Jaru, Jaru ku=yanu. language_name Jaru dat=3pl.o ‘We are staying there now, here we are staying in New Balgo, this place Wirrimanu (which is) for Jaru people.’ Yuwayi, minya nawu-rla karri-nyanta, wuruju=wali. yes this now-loc be-prs good=then ‘Yeah, at that time then we started living (here), it (was) good then.’ (MMN: TEN1-2017_005-02: 329964_337534)

529 Kriol (r)nawu has also formed a compound with the demonstrative kuja as in kujarnawu ‘that’s it’ as described in §10.5.2. The underlying retroflex nature of the initial apical is represented orthographically in non-initial positions.

10.6 Kriol-origin particles 

(1700)

a.

b.

c.

(1701)

a.

b.

 639

Yalu-jangka=rli ya-ni Kururrungku-jangka, wanji-rla? that-ela=1du.incl.s go-pst place_name-ela where-loc ‘After that you and I went from Kururrungku – where?’ Kulirra, Malarn. south place_name ‘South, (to) Mulan.’ Yalawaringki nawu=rli karri-nyani. that.inhab now=1du.incl.s be-ipfv.pst ‘We stayed there then for a long-time.’ (PSM: TEN1-2016_005: 262046_263759) Wurna nawu nga=lu ya-ni Wirrimanu-kurra, journey now seq.aux=3pl.s go-pst place_name-all yalu-jangka=wu. that-ela=voc ‘Setting off now, they went to Wirrimanu from that place (Old Balgo Mission).’ Ngantany wali=lu ya-ni an jingka-jingka. man then=3pl.s go-pst conj child-rdp ‘Yeah, the men went along with all the kids.’ (MMN: TEN1-2017_005-02: 323753_328101)

The collocation wurna nawu is particularly frequent in signalling a topic shift involving departure, as in (1702). (1702)

Wurna nawu ya-nku=rlipa. journey now go-pot=1pl.incl.s ‘(We’re) off now, let’s go.’ (MDN: LC22b: 1662108_1664813)

10.6.5 Concessive wal The Kriol-origin particle wal (< Eng. ‘well’) is frequently used in Ngardi speech and is used with a range of discourse-pragmatic functions. Frequent in narrative texts, it is used clause-initially with topic- or frame-setting functions.530 It is frequently deployed following pauses between utterances and preempts topic shifts. A plausible English translation is ‘this being the case . . . ’ or ‘well, then’, as in (1703).

530 See Cuenca (2008) for a summary of the extensive literature on pragmatic functions of ‘well’ in English.

640 

 10 Auxiliaries, particles and modal constructions

(1703) a. Ngaju-ku=yi panji karri-nyani nganayi-rla, Flora Valley Station, 1sg-dat=1sg.o in_law be-ipfv.pst whatsit-loc place_name natha Vesty, yangi-kariny Vesty teijin. another name one-other name station ‘My brother-in-law was staying there at Flora Valley Station, another Vesty one, another Vesty Station.’ b. Wal ya-ni=rlijarra ngaju-ku ngati, kayirra, well go-pst=1du.excl.s 1sg-dat M north Flora Valley Station-kurra motika-kurlu. place_name-all car-prop ‘Well then, my mother and I, we went north to Flora Valley Station by car.’ (PSM: TEN1-2018_038-01: 68107_76909) In addition to framing new topics in longer stretches of discourse, wal has an additional function as a ‘concessive’ or ‘mitigation’ marker, used to make concessions/ admissions about propositions as per the English ‘well actually/in fact’ or to soften/ mitigate imminent refusals or objections, ‘well no’. In the following narrative extract (1704), wal acts as an organisational pivot in the discourse. After setting up an expectation of events (the taking of cattle to a slaughterhouse in Derby), the narrative takes an unexpected turn, beginning with a clause with an utterance-initial wal. (1704)

a.

b.

c.

d.

“Ma-nta=lu pik pulok redi mit-wok-ku kuja!” get-imp=pl.s big cattle ready meat_work-dat thus ‘(He said:) “Take the fully-grown cattle, like those ready for the slaughterhouse!”’ “Yuwayi.” yes ‘“Sure thing.”’ Wal wakurra=rnalu puliki yala ka-nya hawij-kurra. concess neg=1pl.excl.s cattle that carry-pst house-all ‘In fact, we didn’t take those bullocks to the slaughterhouse.’ Wi bin go Bindy Yard yalu-kurra nawu=rnalu ya-ni. we pst go place_name that-all now=1pl.excl.s go-pst ‘We went to Bindy Yard, we went there.’ (PSM: TEN1-2018_038-02: 504593_508221)

10.7 Commands Imperatives, directives, prohibitives and admonitives all form a functionally coherent group of constructions in that they all share the basic illocutionary function of ‘moving an addressee to action’ (Bybee & Fleischman, 1995; De Haan, 2006, p. 35),

10.7 Commands 

 641

that is, they are all ‘commands’. These four construction types can be parameterised according to the strength of the illocutionary act, termed ‘directness’, and the semantic polarity of the command, as shown in Table 138.531 Illocutionary nominals are discussed separately in §10.7.3. Table 138: Morphosyntax of types of commands. Positive polarity

Negative polarity

Direct

Imperative (§10.7.1.1)

V-imp

Prohibitive (§10.7.1.2)

V–imp . . . Vinf-priv

Indirect

Directive (§10.7.2.1)

V-pot, V-hort

Admonitive (§10.7.2.2)

wakurra . . . V-pot, wakurra . . . V-irr, wakurra . . . V-hort

A positive direct command is simply encoded via the imperative inflection on the verb. A less direct command can be formed via the use of the potential or hortative inflections, termed a ‘directive’. It is not possible to syntactically negate imperative clauses in Ngardi. The prohibitive (a direct negative command) is formed compositionally via an imperative verb (karri-ya be-imp or wanja-ka leave-imp) and a privative marked non-finite (nominalised) verb depicting the main event. The admonitive (an indirect negative command) is the syntactic negation of potential, irrealis or hortative clause types. Negative commands formed in this way are common in languages which do not permit negation of imperative clauses (Aikhenvald, 2010, p. 170).

10.7.1 Direct commands Direct commands are performative, illocutionary acts which speakers use to compel an addressee to some action/inaction. They include commands (positive direct commands) and prohibitives (negative direct commands). The morphosyntax of direct commands is distinguished from indirect commands by the lack of person morphology in the bound pronouns – only number is registered, since they are always addressed to someone and hence have ‘default’ second person subjects. Additionally, direct commands never occur in subordinate clauses or interrogative clauses – for the obvious reason that they are performative in nature (Palmer, 2001, p. 137).

531 A distinct construction which can be used with the illocutionary effect of a negative command is the ngarda apprehensional construction (§11.2.3), equivalent to such warnings in English as ‘you might fall!’. This construction is not discussed further here since it is not strictly a command itself but rather an expression of a likely and unfavourable event which may or may not be associated with an explicit command.

642 

 10 Auxiliaries, particles and modal constructions

10.7.1.1 Imperatives Imperative constructions exhibit a number of unique syntactic properties which differentiate them not only from declarative clauses but also from indirect commands. Imperative clauses do not permit the inclusion of any of the auxiliaries (ngu ~ nga, ma, ka), nor any of the modal particles exhibiting clausal scope (parda, ngarda, mayi and wayi). Additionally, the bound pronouns always encliticise to the imperative verb form. This occurs irrespective of the placement of the verb within the clause.532 An additional, unique feature of imperatives is that all person marking is absent from the bound pronominal complex – only the number feature of the implied second person addressee is registered. All imperatives are addressed to second persons and are incompatible with any other addressee, that is, they cannot be used for inclusive hortation (cf. English ‘let’s  .  .  . ’). Imperatives can be formed via either the regular imperative inflection (1705) or the directionally inflected imperatives involving both hither (1706) and hence (1707) directional morphology. (1705)

Yut=yirra-ka=lu. sit=place-imp=pl.s ‘Put him down.’ (SCM: TT76_1301)

(1706)

Ka-ngka-rni=pula! carry-imp-hith=du.s ‘You two bring it here!’ (PSM: TEN1-2018_028-01: 5225310_5227753)

(1707)

Nya-ngantarra=lu ngarlkin=ma-nantarra=lu=rla! see-ipfv.imp.hence =pl.s block=get-ipfv.imp.hence=pl.s=3sg.obl ‘You must keep on looking out, you must keep on looking after him!’ (DMN: LC42a: 1286416_1289156)

While imperatives are never marked for tense, they may be marked for imperfective aspect in Ngardi, as in (1707). This feature of imperatives is not peculiar to Ngardi but is found in many of the world’s languages (see Palmer, 2001, p. 137) including Pama-Nyungan languages such as Jaru (Ngumpin-Yapa) (Tsunoda, 1981b, p. 79) and Wangkajunga (Western Desert) (Jones, 2011, p. 191). Imperfective imperative commands are typically only applied to events which can be extended in time, as per nyasee in (1707). Imperfective imperatives convey a slightly less direct illocutionary force ‘because it implies less attention to the result or completion of the action in question’, as argued by Goddard (1985, p. 96) for Yankunytjatjara.

532 Cf. Warlpiri, where pronominal encliticisation is always in second position, irrespective of verbal inflection (Simpson, 1991, p. 204).

10.7 Commands 

 643

Imperative-inflected verbs typically appear clause-initially. However, there are examples that indicate that a range of other constituents (typically subject or object NPs) may precede the imperative verb, as exemplified in (1708)–(1710). See §8.2 for an example of serialised imperatives. (1708)

Nyurrurla ya-nta=lu karla, Walmajarri! 2pl go-imp=pl.s west language_name ‘You mob, go west, you are Walmajarri mob!’ (TJN: LC60a: 263253_267776)

(1709)

Ngantany-ju ma-nta=lu=nyanu mangarri! man-erg get-imp=pl.s=refl veg_food ‘You men, get yourselves some food.’ (MMN: TEN1-2017_022-01: 257111_260518)

(1710)

Jangilany ma-nta=lu=nganpa! fire get-imp=pl.s=1pl.excl.o ‘Get some firewood for us!’ (PRN: FT1_b: 106913_108707)

Imperatives can also appear as main/initial clauses in multi-clause constructions. These may either be coordinate (paratactic) clauses, for example in combination with a ngarda clause functioning as an apprehensional (§11.2.3) as in (1711), or a subordinate clause headed by nyampa functioning as an NP-modifier as in (1712). (1711)

Ya-nta=lu=pulany [ngarda=pula=nyanu pu-ngku]. go-imp=pl.s=3du.o hyp=3du.s=refl hit-pot ‘You lot go for those two, lest they fight each other.’ (FT1a: 1667560_1669331)

(1712)

Yala yirra-ka [naympa=n marda-rnanta taka-ngka]. that place-imp what=2sg.s have-prs hand-loc ‘Show (me) what you are holding in your hand.’ (PSM: TEN1-2019_007-02: 133829_140254)

The vocative clitic=wu (§10.4.8) may also encliticise to imperative forms (rightward of any pronominal number clitics, should they appear) as in (1713) and is often accompanied by a characteristic intonation contour (§2.7.2) (1713)

Wuruly=karri-ya=wu! quiet=be-imp=voc ‘Be quiet!’ (PSM: TEN1-2019_005-01: 40224_44719)

Imperative clauses cannot be negated via the default negative operator wakurra. This syntactic incompatibility is reported in multiple Australian languages (Laughren,

644 

 10 Auxiliaries, particles and modal constructions

2017c) and supports Sadock and Zwicky’s (1985, p. 175) observation that ‘a striking fact about negative imperatives is how frequently they are handled differently from negative declaratives’. Instead, the illocutionary force of ordering someone against a course of (in)action is conveyed via either prohibitives (§10.7.1.2) or admonitives (§10.7.2.2). 10.7.1.2 Prohibitives Prohibitive constructions are comprised of an imperative verb in conjunction with a privative case-marked nominalisation as schematised in (1714). (1714)

V-imp (V= karri-/wanja-)

+ [Vinf/preverb]-priv

The non-finite element may be a non-finite verb or a preverb and is subordinated to one of two possible imperative-inflected verbs: karri-ya be-imp or wanja-ka leave-imp. In prohibitives, the negative operator is therefore a suffix attached to a non-finite verbal element expressing the event being negated. There does not appear to be any semantic difference conveyed by the choice of the finite verb (karri- be or wanja- leave) and there are no strict ordering requirements between the finite and non-finite verb forms, as shown by (1715) and (1716).533 (1715)

Wangka-nyu-wangu wanja-ka!534 speak-inf-priv leave-imp ‘Stop talking!’ (TEN1-2019_005-01: 53932_57190)

(1716)

Karri-ya=wu wangka-nyu-wangu. be-imp=voc speak-inf-priv ‘Stop talking!’ (TEN1-2019_005-01: 97415_102358)

A prohibitive can also be formed with a preverb as the subordinated non-finite constituent as in (1717). (1717)

Iyindij-wangu karri-ya! whinge-priv be-imp ‘Stop complaining!’ (PSM: TEN1-2017_010-01: 74308_76339)

533 Unfortunately, there are no examples with non-singular addressees in the corpus to determine the location of bound pronominal complexes in these constructions. 534 The morphophonemic form /waŋkaɲu-wàŋu/ is typically realised as [wáŋgaɲàːŋu].

10.7 Commands 

 645

In cases where the addressee is the semantic subject of a transitive predicate in the non-finite clause, the non-finite verb + -wangu is typically found further case-marked with the ergative case suffix. (1718)

Yuwayi, jiti-rnu-wangu-rlu wanja-ka jingka. yes tease-inf-priv-erg leave-imp child ‘Stop teasing the child.’ (MDN: LC42a: 176073_179348)

(1719)

Wanja-ka nuku=paja-rnu-wangu-rlu. leave-imp kiss=bite-inf-priv-erg ‘Don’t you kiss him.’ (TJN: LC48a: 672362_675690)

A construction similar to the prohibitive in Ngardi is found in most Ngumpin-Yapa languages (Laughren, 2017a).535 In neighbouring Warlpiri, the choice between an intransitive imperative main verb nyi- sit and a (canonically) transitive main verb yampileave is determined by the transitivity of the nominalised predicate (Laughren, 1982a, pp. 152–155). This broadly holds up in Ngardi too, however, the pair (1715) and (1716) are an obvious exception since both wanja- and karri- appear with the intransitive wangka-. In a subset of Ngumpin-Yapa languages, finite imperative verbs may also be elided and the non-finite verb ‘insubordinated’ while retaining the same illocutionary force, as per the Mudburra and Warlmanpa examples in (1720) and (1721). (1720)

(1721)

Wanyi-nyu-mulu! fall-inf-priv ‘Don’t fall down!’ [Mudburra]

(Osgarby, 2018, p. 228)

Nga-rninja-wangu! eat-inf-priv ‘Don’t drink it!’ [Warlmanpa]

(Browne, 2021, p. 398)

This insubordination strategy is possible – albeit rare – in Ngardi. In (1722), for example, the initial insubordinated construction was provided in isolation, and the second clause was added separately as an independent imperative. (1722)

Yapartu-ngu-wangu=wu, (yut=karri-ya)! run-inf-priv=voc sit=be-imp ‘Stop running around, sit down.’ (PSM: TEN1-2019_005-03: 132695_136965)

535 Note that the semantically related lacking suffix -jirraja is never used in this function.

646 

 10 Auxiliaries, particles and modal constructions

Note, however, that instances of ya-nu-wangu in other contexts were more readily interpreted as referential, derived nominals, for example, ‘one unable to walk’, ‘a lame person’, and not as a command ‘Don’t walk!’. It may be the case that discrete prosodic features of commands may further distinguish such ambiguous uses of putative insubordinate prohibitives.

10.7.2 Indirect commands Indirect commands (directives and admonitions) are constructions used when the speaker wishes to convey an order or request in a more ‘polite’ way. Like direct commands, indirect commands may be deployed to ‘move an addressee to action’ (Bybee & Fleischman, 1995; De Haan, 2006, p. 35) but are differentiated from them by the strength of the illocutionary act – that is, the speaker’s insistence upon the addressee’s compliance is less pronounced than for direct commands. Formally, indirect commands are encoded via the potential, hortative or irrealis (admonitions only) rather than the imperative. Indirect commands take person and number markers in the pronominal complex, unlike direct commands which are not overtly specified for person. 10.7.2.1 Directives I use the term ‘directives’ (from Palmer, 2001, p. 127) to group a set of positive polarity constructions which express the deontic notion of weak obligation. These are less direct than imperatives but nevertheless can be used to direct an addressee to action. Unlike direct commands, they are not performative and need not relate to the present moment (speech time). Directives are encoded via the potential or hortative verbal inflections. Auxiliaries, modal and propositional particles are all generally absent from directive constructions. The formal exponents of directives (potential and hortative verbal inflections) are also used for a much wider array of modal meanings including possibility, uncertainty and intention (as previously detailed in §10.2). The range of functions available to the potential means it is not always possible to clearly distinguish between utterances which are expressing a speaker’s wishes/ desires and utterances which are intended as directives, compelling the addressee to action (Lyons, 1977, p. 53). Directives can be distinguished from other uses of the potential or hortative clauses most clearly in cases involving first or second person subjects where the speaker is performing an illocutionary act of ordering in which they intend to be involved, as in (1723) and (1724). The interpretation can either be as a directive for immediate action subsequent to the request (1723) or it may also be interpreted as one of ongoing deontic obligation holding at time of utterance (1724).

10.7 Commands 

 647

(1723)

Raly=pu-ngku=rli minya kipara. pluck=hit-pot=1du.incl.s this bush_turkey ‘Let’s pluck this bush turkey.’ (MMN: TEN1-2017_002-02: 1154191_1156037)

(1724)

Nya-nganku=rlipa, ngarda nga=ngala pila=ma-ni see-ipfv.pot=1pl.incl.s hyp seq.aux=1pl.incl.o follow=get-pst kalanga. cf ‘Yeah, we must keep on watching for him, he might have followed us.’ (YRN: LC33a: 483076_487583)

The directive functions of the potential can also be applied directly to second persons. This is clearly shown in complex clauses involving a contingent eventuality expressed in the apodosis as in (1725). (1725)

Kaji=ngku kunyarr-u, paja-ku jina-ngku=n condit=2sg.o dog-erg bite-pot foot-erg=2sg.s ‘If the dog bites you, you (should) kick it with your foot.’ (PSM: TEN1-2017_026-01: 3341260_3346763

kanyi-ku. crush-pot

For third person directives, English translations are somewhat cumbersome but, generally speaking, they are interpreted as polite directives and are associated with acts of permission granting or encouragement as in (1726). (1726)

Karri-jura=pula mula ngaju-punta-rla. be-hort=3du.s here 1sg-pposs-loc ‘Let those two stay here at my place.’ (PSM: TEN1-2017_026-01: 2712479_2718586)

10.7.2.2 Admonitives Admonitive constructions are formed via the negative operator wakurra combining with the potential, hortative or irrealis verbal inflections as shown in (1727). (1727)

Wakurra-[bound pronoun(s)] . . . Verb-{pot, hort, irr}

Admonitions are a type of speaker-oriented modality wherein the speaker directly suggests what the addressee should not do in a given context. The warning may or may not be interpreted by the addressee as a command to action – this depends on the temporal location of the event with respect to utterance time. Admonitions can have either a hypothetical, advising function (i.e. ‘x isn’t doing y and x shouldn’t do y’) or a chastising function (i.e. ‘x is doing/did y and x really shouldn’t (have done) y’). The

648 

 10 Auxiliaries, particles and modal constructions

source of obligation implied by the admonition is generally external to the participant (i.e. norms of social acceptability) rather than inherent to the referent’s capacity (i.e. deontic force rather than an expression of root modality). In this sense, all admonitives operate as a sort of permission-withholding expression – ‘x isn’t permitted to y’. The advising function of admonitives is most common with second person subjects where the speaker is directly compelling the addressee to comply to some state of affairs (1728). In clauses involving third person subjects (1729), the function of the admonitive is generally interpreted as having a chastising function or else a means of explaining social norms. (1728)

Wakurra=nta ya-nku-rni rangarni. neg=2pl.s go-pot-hith morning ‘You lot shouldn’t come tomorrow/Don’t come tomorrow.’ (MMN: TEN1-2017_019-01: 353214_357451)

(1729)

a.

b.

Wakurra nga-lku kuyi. neg eat-pot meat ‘He can’t/shouldn’t eat meat.’ Yala ma wirrijaminy-kurlu. that top.aux hairstring_necklace-prop ‘That one is in mourning.’ (MMN: TEN1-2018_034-01: 2255189)

An admonitive clause can also be accompanied by a conditional kaji clause, as in (1730). (1730)

Wakurra=n kuli karri-ju kaji=rna=ngku jingkirdi-ku. neg=2sg.s anger be-pot condit=1sg.s=2sg.o laugh-pot ‘Don’t get angry if I make you laugh.’ (MMN: TEN1-2018_026-01: 2300414_2325491)

As is discussed in §7.4, the potential, and (occasionally) the hortative, are used in contexts where the speaker is referring to events simultaneous or subsequent to reference time (advising function), as in (1731) and (1732), and the irrealis is (optionally) used in contexts prior to reference time (chastising function) as in (1733). (1731)

Wakurra=n pimirdi wirrpa yu-ngankura. neg=2sg.s FZ many give-ipfv.hort ‘You should not give out so many at once, aunty.’ (TJN: LC32b)

10.7 Commands 

 649

(1732)

Wakurra=n ngunyi=pu-nganku-rra yala ita. neg=2sg.s reject=hit-ipfv.pot-hence that small ‘You shouldn’t be kicking out/rejecting that small (frog)!’ (MMN: TEN1-2019_005-06: 277905_281249)

(1733)

Wakurra=n warlu ingki-rnngi. neg=2sg.s fire light-irr ‘You shouldn’t light fires.’ ‘You shouldn’t have lit fires.’ (PSM: TEN1-2018_034-01: 2741886_2745705)

For some speakers in the corpus, the negative operator wakurra is sometimes observed code-switched for the English ‘no’ in admonitives as in (1734). (1734)

Nuu=lu=rla pu-ngku! no=3pl.s=3sg.dat hit-pot ‘Don’t damage it!’ (MDN: LC58)

An additional morphosyntactic construction which has an illocutionary function similar to admonitives are ngarda + irr clauses as described in §10.2.4.1. These can have precautioning speech act functions as in (1735). They are not strictly categorised as commands since they are only pragmatically interepreted as directives to undertake action and are also formally indistinct from the expression of epistemic possibility (with/without negative evaluations). (1735)

Ngarda=n wanti-ngi! hyp=2sg.s=3pl.s hit-irr ‘You might fall! (Be careful!)’ (PSM: TEN1-2018_028-01: 3257052_3260800)

10.7.3 Illocutionary nominals In addition to direct and indirect commands, Ngardi also possesses a small class of ‘illocutionary’ nominals that be used as directives in isolation: wurna ‘let’s go’ (§10.7.3.1), wurra ‘wait/hold on’ (§10.7.3.2), yurru ‘look out!’ (§10.7.3.3).536 All illocutionary nominals can host the vocative enclitic =wu. Unlike syntactic commands, they do not host number enclitic pronouns indicating the number of the addressee.

536 Strictly speaking, these ‘illocutionary nominals’ also meet the criterial definitions of interjections (§10.5) but are described here due to their unique function as directives.

650 

 10 Auxiliaries, particles and modal constructions

10.7.3.1 Wurna The nominal wurna may either be employed as an independent interjection (1736) or it may accompany a potential/hortative-inflected form of the go verb ya- (1737). In this latter context it is not tightly integrated with the verb and may freely occur before or after the inflected verb and be separated by the pronominal complex. In both cases it is addressed to a first person, non-singular audience. (1736)

Wangka-nya=rla, “Yala wurna”. speak-pst=3sg.obl that journey ‘He said to him, “Let’s go there (that country).”’ (PSM: TEN1-2018_00901: 40864_44550)

(1737)

Wurna ya-nku=rlipa=wu! journey go-pot=1pl.incl.s=voc ‘Let’s go!’ (YDN: LC49b: 1226208_1229483)

10.7.3.2 Wurra The illocutionary nominal wurra is a temporal nominal (§5.7) and can also be used as a command when uttered as an independent interjection, as in wurra! ‘wait!’. The illocutionary nominal can serve as an initial directive in complex conditional clasues, as in (1738) and (1739). (1738)

Wurra=lku kaji warririku karri-ju. wait=then condit close be-pot ‘Wait until he (the kangaroo) is closer.’ (MMN: TEN1-2017_026-01: 2644491_2650853)

(1739)

Wurra=lku kaji ngapa kaniny karri-ju. wait=then condit water inside be-pot ‘Wait until the water goes down.’ (MMN: TEN1-2017_001-02: 1759688_1769050)

There are some isolated examples of wurra with an ambiguous status between illocutionary nominal and temporal nominal status. Thus in the following examples it is possible that wurra is integrated into complex verb constructions with imperative (1740) and hortative (1741) inflected forms of the verb karri-. (1740)

Wurra karri-ya, wurra! wait be-imp wait ‘Wait, wait!’ (SCM: TT76_1301: 773438_774494)

10.8 Negation 

(1741)

 651

Wurra karri-jura=rlipa. wait be-hort=1pl.incl.s ‘We can wait and stay here.’ (PRN: FT1_b: 17788_419060)

10.7.3.3 Yurru The illocutionary nominal yurru is used to attract and direct the attention of the addressee typically towards something in the real-world context of the speech act. The use of the nominal yurru is typically accompanied by a gesture, either of the hand, head, eyes or lips to direct the attention and gaze of the addressee. It can also be used in highly localised contexts when attracting someone’s attention to a matter at hand, for example when passing someone something. Marie Mudgedell explains its function as in (1742). (1742) Yurru na jangu, “look out! i coming deya!” “Yurru look_out now like

look_out he’s coming!”’

‘Yurru is like (in English), “Look out there, he is coming!” “Yurru! (we say) he’s coming!”’ (MMN: TEN1-2019_007-01: 420286_424332) Marie Mudgedell further commented that this particle is used when speaking Kukatja, Jaru, (Eastern) Walmajarri and Ngardi. An example of yurru hosting the vocative clitic is provied in (1743). (1743)

Yurru=wu! look_out=voc ‘Look out!’ (MMN: TEN1-2019_007-01: 417198_419125)

10.8 Negation Negation and negative concepts can be encoded in a number of morphosyntactically distinct ways in Ngardi. Table 139 sets out the types of negation (clausal, predicative, individual, exclamative and lexical). While this table is arranged according to the syntactic features of negation, the subsequent discussion is arranged according to forms. This is due to the polyfunctionality of individual lexemes. For example, both wakurra and kulanganta can have either clausal or phrasal scope, and the former is also used as an interjection.

652 

 10 Auxiliaries, particles and modal constructions

Table 139: Morphosyntax of negation in Ngardi. Type of Negation

Semantic subtype

Exponents

Subsection

Clausal

Indicative

wakurra neg kula neg {karri-ya/wanja-ka} + [nfin]-priv wakurra neg + v-pot ngarda hyp + []v-irr / []v-pot kulanganta cf

§10.8.1.1 §10.8.2 §10.7.1.2 §10.7.2.2 §10.2.4 §10.3.6

wakurra neg ‘not x’ -wangu priv ‘without x’ -jirraja lack ‘lacking x’ kulanganta cf ‘mistakenly thought x’

§10.8.1.2 §4.6.8 §4.6.9 §10.3.6

Prohibitive Admonitive Precautionary Mistaken belief Phrasal

Individual Mistaken belief

Predicative

Existential

waku + [. . .]n-dat ‘none of x’

(§10.8.3)

Lexical

Interjection Polar response Preverb

waku, wakurra ‘oh no’ wakurra, waku ‘no’ kanginy ‘fail to recognise’ jakarr ‘try, failed attempt’ waku- ‘speak negatively’

§10.8.1.3 §7.9.1.3 §10.8.5

Verb

§10.8.5

Clausal negation is sensitive to the modal features of the clause. For indicative clauses, wakurra is used in its clausal function (§10.8.1.1). This includes the formation of negative indirect commands referred to as ‘admonitions’, which are discussed separately in (§10.7.2.2). In order to semantically negate imperative clauses, one of two imperative-inflected verbs is used (karri-ya be-imp or wanja-ka leave-imp) as an auxiliary in combination with a privative case-marked predicate (generally a nominalised verb or preverb). This type of clause, termed a ‘prohibitive’, was described in §10.7.1.1. The counterfactual kulanganta (and allomorphs) serves to encode a particular negative belief state towards a proposition, namely one formerly held to be true. It is categorised as a modal particle (§10.3.6) since it reflects a speaker’s (false) beliefs. The privative and lacking suffixes were described in detail elsewhere (§4.6.8, §4.6.9) but some interesting properties of their attachment to action nominals are described in (§10.8.4)

10.8.1 Negative wakurra The negative particle wakurra is polyfunctional. In addition to having both clausal (§10.8.1.1) and phrasal (§10.8.1.2) roles, wakurra is also an interjection (§10.8.1.3) where it functions either as a negative response ‘no’ or as a lamentation ‘oh no!’.

10.8 Negation 

 653

10.8.1.1 Clausal negation The negative particle wakurra is the basic means to negate clausal propositions. This includes clauses containing regular verbal predicates (1744), copular verbal predicates (1745) and even case-marked nominal predicates (1746), (1747). (1744)

Wakurra=rna ya-nku wirlinyi waku, ngarrja=rna marda-rnanta. neg=1sg.s go-pot hunting neg enough=1sg.s have-prs ‘I won’t go hunting, no, I have plenty.’ (PSM: TEN1-2018_021-01: 3829713_3840700)

(1745)

Wakurra rod karri-nyani Rabbit Flat-wana, waku. neg road be-ipfv.pst place_name-perl nothing ‘There wasn’t a road around Rabbit Flat (back then), nothing.’ (PSM: TEN1-2018_009-01: 685616_690823)

(1746)

Wakurra ma=rna Walmajarri-kurlu. neg top.aux=1sg.s language_name-prop ‘I don’t speak/know Walmjarri/I am not Walmajarri (lit. ‘I am not with the Walmajarri (language)’).’ (PSM: TEN1-2018_030-01: 708560_710603)

(1747)

Wakurra kardiya-rla. neg European-loc ‘It wasn’t with whitefellas.’ (YDN: LC49: 1592901_1597173)

The default position of the negative operator wakurra is clause-initial and generally serves as a host to any (overt) bound pronominals – although an overt auxiliary may still appear, as in (1746). Additionally, like modal particles, the clausal negator wakurra can occasionally be displaced by topicalised items as in (1748). (1748)

Nyanung wakurra=rla kakarra ya-nani-nyirra waku. recog1 neg=3sg.obl east go-ipfv.pst-narr no ‘He didn’t go east to him, no.’ (BSE: TEN1-2017_014-01: 187343_190153)

When occurring with clausal scope wakurra has scope over ignoratives as in (1749). (1749)

Wakurra=pula ngana nya-nganta. neg=3du.s who see-prs ‘Those two are not watching anyone.’ (PRN: TG1_0034-01)

The negative particle wakurra combines predictably with all tense/aspect values of verbal clauses. By contrast, there are some idiosyncrasies in the combination of negative wakurra and clausal mood. First, wakurra can not modify an imperative clause.

654 

 10 Auxiliaries, particles and modal constructions

Negative commands can be expressed by admonitive clauses (§10.7.2.2) which involve wakurra in combination with the potential (1750), irrealis (1751) or (rarely) the hortative. Semantically, these can also be interpreted as ‘obligatives’, especially if the subject is non-second person (1752) or refers to past time (1753). Second, wakurra in combination with the irrealis verbal suffix also can be used for negative past hypotheticals (1754). [i] (1750)

Prohibitive/negative obligative Wakurra=n=ku=rla wangka-ju. neg=2sg.s=ep=3sg.obl speak-pot ‘Don’t talk to him./You shouldn’t talk to him.’ (PSM: TEN1-2018_008-01: 1860517_1865231)

[ii] (1751)

Prohibitive/negative obligative Wakurra=n warlu ingki-rnngi. neg=2sg.s fire light-irr ‘Don’t light fires./You shouldn’t light fires.’ (MMN: TEN1-2018_034-01: 2741886_2745705)

[iii] (1752)

Negative obligative (with potential) Wakurra=rlanyanta wangka-ju. neg=3sg.lct speak-pot ‘He shouldn’t speak to her.’ (TEN1-2018_034-01: 2386239_2387766)

[iv] (1753)

Negative past obligative (with irrealis) Wakurra nga-rnngi. neg eat-irr ‘He shouldn’t have eaten it (meat).’ (PSM: TEN1-2018_034-01: 2502765_2504309)

[v] (1754)

Negative past hypothetical Kardiya-rlu wakurra=lu parli=pi-ngi nyantu-kariny-ju ngapa. white_man-erg neg=3pl.s find=hit-irr recog1-other-erg water ‘Whitefellas would not have found water by themselves.’ (MMN: TEN1-2018_026-01: 2548775_2590106)

No examples have been observed where an ignorative has scope over a negative clause, i.e. *‘who didn’t watch someone?’. Negative polarity questions of this kind are instead formulated by the use of a positive polarity question, involving a stative verbal predicate (karri- be) in combination with a privative case-marked nominal or nominalisation (see §10.9.3)

10.8 Negation 

 655

10.8.1.2 Phrasal negation In addition to negating propositions encoded by propositions expressed by clauses, wakurra can also serve to negate a proposition expressed by a phrase, that is, wakurra + NP ‘not x’ as in (1755). (1755)

a.

b.

Yard-ta jangu nganayi-rla. yard-loc like whatsit-loc ‘Like at that you know, at the yard.’ Timba, wakurra tiilyard, timba. timber neg steel_yard timber ‘It was a timber one (then), not a steel yard, a timber one.’ (PSM: TEN1-2018_038-01: 522320 _531095)

Alternatively, these can be considered cases of verbless clausal negation (1756). (1756)

Wakurra minya warlka, minya=yi jungarni. neg this lie this=emph true ‘This is not a lie, this is true.’ (MAM: TEN1-2016_002-1: 310002_313403)

10.8.1.3 Interjection function In addition to its role as a clausal negator, the negative operator wakurra is also used as an interjection to express disagreement (1757), or as a form of lamentation, equivalent to ‘oh no’ in English, as in (1758). In its interjection function, wakurra often appears encliticised with the vocative enclitic =wu. (1757)

Wakurra=wu, nganimpa=nganpa minya ngurra. neg=voc 1pl.excl=1pl.excl.o this country ‘No, this is our country!’ (BSE: TEN1-2017_015-01: 50139_53332)

(1758)

Wakurra, wurna ma. neg journey top.aux ‘Oh no, he (the frog) is going (escaping).’ (MMN: TEN1-2016_016-02: 35115_37402)

10.8.2 Negative kula A subset of Ngardi speakers recorded in the Cataldi corpus make use of another sentential negator: kula, as illustrated in (1759). It is not used by any speakers recorded between 2016–2019. In the data available, kula has not been found in the same range

656 

 10 Auxiliaries, particles and modal constructions

of clause types as wakurra (§10.8.1) and it is not used in the formation of admonitions (§10.7.2.2). Additionally, unlike wakurra, kula is never used as an interjection.537 (1759)

Wakurra, kula=rna=ngku wirrirl ya-nku. neg neg=1sg.s=2sg.o back go-pot ‘No, I will not come back to you.’ (YMN: LC23a: 740336_742298)

Some speakers alternate between the two negators (wakurra ~ kula) in declarative clauses, even within adjacent clauses, as in (1760). It is unclear if there are any attendant semantic differences. (1760)

a.

b.

Wakurra=rnalu purda=nya-ngani nganayi-ku. neg=1pl.excl.s listen=see-ipfv.pst whatsit-dat ‘We did not understand whatchacallem.’ Kula=rnalu=yanu andastan=ma-nani-nyirra=rnalu=yanu neg=1pl.excl.s=3pl.o understand=get-ipfv.pst-narr=1pl.excl.s=3pl.o Kukatja. language_name ‘We did not understand them, the Kukatja.’ (TJN: Manungka_02-022027: 2810322_2815198)

10.8.3 waku ‘no, nothing’ The lexeme waku has three functions. First, as a negative interjection waku can be used both as a response to information or ‘content’ questions (1761) and, less frequently, as a response to polar questions (1772). (1761)

a.

b.

Nyampa=nta win=ya-ni? what=2pl.s win=go-pst ‘What did you lot win?’ Waku. no ‘Nothing.’ (TJN: LC32b: 429404_432551)

537 Wanyjirra is another Ngumpin language with two clausal negators: wagurra and gula, with the former being more common (Senge, 2015, pp. 490–491). For Jaru (Ngumpin), Tsunoda (1981b, p. 204) claims that gula is contrasted with wagurra by the fact that it is typically only used alongside potential verb forms to encode inabilative modalities, i.e. ‘incapable’, ‘cannot’.

10.8 Negation 

(1762)

a.

b.

 657

Nya-nya=nta=pulany jingka-kujarra ya-nu-ngka? see-pst=2pl.s=3du.o child-dual go-inf-loc ‘Did you lot see those two kids going along?’ Waku, wakurra=rnalu=pulany nya-nya. no neg=1pl.excl.s=3du.o see-pst ‘No, we didn’t see those two.’ (PRN: TG1_0034-01)

Second, the lexeme waku may also be used as the main clausal predicate (1781). In this function, waku selects an absolutive subject and a dative indirect object; encoding a person/thing who is not in possession of an entity/item. (1763)

Waku=rna=rla ngaju ngunyju-ku. no=1sg.s=3sg.obl 1sg tobacco-dat ‘I don’t have any tobacco.’ (PSM: TEN1-2018_030-01: 482667_486150).

Lastly, the negative particle waku also appears utterance-finally following negative clauses. For example, the negative clauses in (1744), (1745) and (1748) all terminate with the ‘negative interjection’ waku. In these contexts waku operates as a negative intensifier. It is not uncommon for waku to also undergo phrase-final lengthening for further emphasis. Phrase-final waku is often translated into Kriol/Aboriginal English as najing. In English such translations correspond approximately to: ‘I didn’t x, no (I did not)’, or ‘there isn’t any x, none at all’.538

10.8.4 Negative suffixes: -wangu priv, -jirraja lack Both the privative (-wangu) and lacking (-jirraja) suffixes were described in §4.6.8 and §4.6.9, respectively. Both can indicate the absence of an entity, state or property. The former generally indicates mere absence whereas the latter has an added semantic component of detriment to the would-be possessor. In this section I briefly exemplify special usages of the privative where it is used to negate a predicative nominal (1764), a nominal in a stative construction (1765), and an action nominal (1766). In these examples, the semantic effect is equivalent to clause-level negation. The lacking suffix is not observed in this function.

538 A similar function is reported for wiya in Kukatja (Valiquette, 1993, p. 459), but note that wiya is also the main sentential negator in Kukatja.

658 

 10 Auxiliaries, particles and modal constructions

(1764)

Warririku-wangu warru yala pussycat. near-priv around that pussycat ‘The pussycat wasn’t nearby./The pussycat was nowhere nearby.’ (MMJ: Cataldi, 2011, p.32)

(1765)

Wakurra, minya karri-ju mining-wangu. neg this be-pot minining-priv No, this place won’t have any mining.’ (MMN: TEN1-2017_005-02: 393332_396252)

(1766)

Wirlinyi-wangu ngarda=lu=ngala purangu-rlu jina hunting-priv hyp=3pl.s=1pl.incl.o sun-erg feet parraj=pa-ngi. burn=get-irr ‘(We) won’t go hunting lest the heat burn our feet.’ (PPN: Cataldi, 2011, p. 74).

10.8.5 Lexical items with a negative meaning The following examples illustrate a range of lexical items which encode various negative meanings. Of the following, kanginy- ‘not recognise’ (1767) and jakarr- ‘try to’ (1768) are preverbs and waku- slander (1769) is a simple verb. Note that waku- is homophonous with the negative lexeme described in §10.8.3. Strictly speaking the meaning of this simple verb is not ‘negative’ but rather involves a negative evaluation, i.e. ‘speak bad things’, ‘not speak (well) of’. (1767)

Nyampa-ku=n kanginy=pu-nganta? what-dat=2sg.s not_recognise=hit-prs ‘Why don’t you recognise him?’ (PSM: TEN1-2017_008-01: 3690719_3692710)

(1768)

Jakarr=la-ni ngantany-ju jaji-ku. try=pierce-pst man-erg kangaroo-dat ‘A man tried to spear a kangaroo (unsuccessfully).’ (IPN: TEN1-2018_02601: 1214368_1219477)

(1769)

Nyuntu=ngku=lu waku-rnanta. 2sg=2sg.o=3pl.s slander-prs ‘They are saying bad things about/slandering you.’ (MDN: LC48b: 1301298_1305309)

Interestingly, the preverb jakarr appears in one additional context which constitutes the marking of a modal category not generally distinguished in Ngardi – namely

10.9 Interrogation 

 659

inability/incapacity (a participient-oriented modality). In its canonical usage, this adverbial preverb qualifies an event as having been attempted but not successfully completed, as in (1768). In another function, it may be used to encode the inability of a clausal subject to complete the event described in the clausal predicate, as in (1770). (1770)

Murti ngawu=karri-nya, jakarr ma=rna wurna=ya-nanta. knee bad=be-pst try top.aux=1sg.s journey=go-prs ‘My knee is injured, I am unable to travel around.’ (PSM: TEN1-2018_001-01: 2051618_2056848)

Thus, in (1770), the intended event being depicted is not a sequence of attempting and failing to walk around but rather refers to the speaker’s state of incapacity (at time of utterance) with respect to the event expressed by wurna=ya-. This is the only lexeme within the verbal complex (verb or preverb) which has been observed to have an apparent modal function (i.e. inabilitive) in Ngardi.

10.9 Interrogation Ngardi has polar (yes/no) questions (§10.9.1), tag questions (§10.9.2) and information questions (§10.9.3). Ngardi lacks interrogative verbs or any derivational morphology which may derive a verb from an interrogative element.539

10.9.1 Polar Questions Polar (yes/no) questions are morphosyntactically identical to declaratives but can be distinguished by rising intonation – for example, the initial question in (1771), which is affirmed with an identical clause by the interlocutor but with declarative intonation.540 (1771)

TJN:

YRN:

Jakarr=pa=nyurra warri-rni? try=ep=2pl.o search-pst ‘Did he try and look around for you lot?’ Jakarr=pa=nganpa warri-rni. try=ep=1pl.excl.o search-pst ‘He did try and look for us (but couldn’t find us).’ (LC33a: 318860_321635)

539 Cf. Kukatja which has a ‘question verb’ nyaalparriwa meaning ‘do what?’ (Valiquette, 1993, p. 461). 540 Cf. Walmajarri where interrogatives are distinguished from indicatives by the auxiliary nga vs pa (Hudson, 1978, p. 78).

660 

 10 Auxiliaries, particles and modal constructions

Alternatively, polar questions can include one of two modal particles which also encode epistemic uncertainty: unreal wayi (1772), (1773) (see §10.2.2) and dubitative mayi (1518) (see §10.2.1) – although the latter appears more often as a tag question. Clausal TAM does not affect the formation of interrogatives. (1772)

YDN:

TJN:

YDN:

(1773)

Ya-ni=rnalu Walajaparnta nga-rnungka-nya. go-pst=1pl.excl.s place_name eat-x&move-pst ‘We went to Walajaparnta, ate and kept going.’ Kala Ngantalarra-kurra wayi=nta ya-ni? contr place_name-all unreal=2pl.s go-pst ‘But what about Ngantalarra, did you lot go there?’ Waku. neg ‘No.’ (LC33a: 450576 _456309)

Wayi=n marda-rnanta? unreal=2sg.s have-prs ‘Do you have any (food)?’ (MMN: TEN1-2017_026-02: 929061_931035)

Since questions can be formed via intonation alone, it is also possible for questions to be formed from clauses with non-realis modalities, for example hypothetical ngarda, as in (1774). Within the corpus parda dub has only been found in questions when encliticised to ignoratives (e.g. nyampa-ku=parda). (1774)

Ya-ni=lu ngarda=lu? go-pst=3pl.s hyp=3pl.s ‘They might have gone?’ (PSM: TEN1-2018_028-01: 1493038_1497542)

10.9.2 Tag questions Tag questions are formed by post-posing a ‘question tag’ to a clause, as per mayi in (1775) and (1776). In addition to their positioning outside the clause as semiindependent utterances, tag words are further assigned a distinct rising intonation. The function of tag questions is to call for the (dis)confirmation of the assertion posed in the preceding speech. (1775)

Tajitaji, mayi? plant_sp tag ‘Tajitaji perhaps?’ (YDN: cards 17-8-00)

10.9 Interrogation 

(1776)

 661

Mula=rlipa karri-ju tirrip mayi? here=1pl.incl.s be-pot overnight tag ‘Are we going to sleep overnight here, do you think?’ (LC33b)

The range of tag words used by Ngardi speakers shows significant variation across the corpus. The full list of tag words was provided in Table 137. Generally speaking, mayi appears to be the default lexeme for forming tag questions for most speakers, although some Ngardi-Warlpiri speakers recorded in the 1990s also used kari (§10.5.5). As extraclausal constituents, tag particles are frequently code-switched in the data. In the speech of Ngardi speakers recorded in 2016–2019, mayi was not used as the default tag question. Instead, in unprompted speech, it has been replaced by ngi? (1777) or kurlu? (1778) – forms of a Kukatja or Wangkajunga origin (Jones, 2011, p. 311) – which were also mixed in the speech of speakers recorded 15 years earlier (see (1779)).541 (1777)

Marlayi, purnu ngi? tree_sp tree tag ‘Marlayi is a type of tree, yeah?’ (MMN: TEN1-2018_004-02: 2881419_2889453)

(1778)

Knee, murtu, kurlu? knee knee tag ‘Knee is murtu, yeah?’ (MMN: TEN1-2018_001-01: 2040699_2043619)

(1779)

Wanja-rnungka-nganta=nganpa ngi, mayi? leave-x&move-prs=1pl.excl.o tag tag ‘He/she went and left us, yeah?’ (NMN: Nanyuma2).

Walmajarri has a modal particle ngayi meaning either ‘isn’t that true?/Is that ok?’ or ‘yes, mm’, where the latter affirmative sense is only found in the Eastern dialect (Richards & Hudson, 2012). In Ngardi the form ngayi is also found (see §10.3.4) but is best described as a type of assertive propositional particule, used to affirm the truth value of a proposition expressed by a clausal event ‘indeed x, yes x’, similar to the secondary sense of ngayi reported for Walmajarri. It does, however, function similarly to Walmajarri ngayi when appearing in declarative clauses used with interrogative intent, as in (1780).

541 Of additional note is the presence of the Kriol lexeme yuno in Ngardi speech to either overtly check an interlocutor’s understanding (as a direct question) or else as something of a discourse marker (Erman, 2001; Schourup, 1985). The former usage is in fact rare and brought about through elicitation contexts whereas the latter is used more frequently in natural speech (see §10.6.2).

662 

 10 Auxiliaries, particles and modal constructions

(1780)

Nyampa=pula kupa-rnanta ngayi? what=3du.s cook-prs indeed ‘What are those two cooking again?’ (I’m asking because I’ve forgotten) (YMN: LC33b: 567254_569048)

10.9.3 Information questions Information questions are formed via the use of an ignorative (see §5.3). Some examples are repeated here with a range of ignoratives (1781)–(1783). (1781)

Ngana pawu-karra ya-nanta? who shout-ssub go-prs ‘Who is going along shouting?’ (PRN: TG1_0034-01)

(1782)

Nganayi yala jingkirdi=ma-nanta, jingkirdi-nyanta? whatsit that laugh=get-prs laugh-prs ‘Who is that laughing?’ (PSM: TEN1-2017_009-01: 81976_86168)

(1783)

Nyarrpa karri-ju=rlipa, jalangu? how be-pot=1pl.incl.s today ‘What will we be doing today?’ (PSM: TEN1-2017_008-01: 2616250_2622085)

Information questions can also take the form of verbless clauses, as in (1784) and (1785). (1784)

Nyampa=n nyuntu? what=2sg.s 2sg ‘What do you have?’ (LC56b)

(1785)

Nyampa jalangu? what today ‘What is today?’ (PSM: TEN1-2017_008-01: 2613280_2615137)

The use of case morphology on ignoratives further generates a rich inventory of question words which function to query various other knowledge categories or ‘epistememes’ such as reason (nyampa-jangka what-ela (513) or nyampa-ku what-dat (1787)), kind (nyampa-piya what-sembl), as well as more specific meanings such as wanji-wana where-perl ‘along which direction’, as in (1789). Further examples of ignoratives in combination with demonstratives with recognitional or evocative functions were provided in §5.3.

10.9 Interrogation 

(1786)

Minya ma shirt, yuruly=pu-nganta, nyampa-jangka-rlu? this top.aux shirt remove=hit-prs what-ela-erg ‘This man is taking off his shirt, what for?’ (PSM: TEN1-2018_035-02: 327490_330625)

(1787)

Nyampa-ku=n=ku=rla kurrku pangi-rni? what-dat=2sg.s=ep=3sg.obl hole dig-pst ‘Why are you digging holes?’ (YMN: LC23b: 514148_515911)

(1788)

Nyampa-piya=lu pu-nganta kuyu yalu-ngka? what-sembl=3pl.s hit-prs meat that-loc ‘What sort of meat are they hunting there?’ (MMN: TEN1-2016_005-02: 65627_68977)

(1789)

Wanji-wana nyurrurla yala ya-ni? where-perl 2pl that go-pst ‘Along which route did you lot go?’ (YMN: LC23a: 1285487_1288238)

 663

It is not possible to place information questions under the scope of the standard clausal negator wakurra (§10.8.1.1). Instead, a positive polarity information question must be formed using a stative verbal predicate (karri- be) in combination with a privative case marked nominal (or nominalisation) as in (1790). (1790)

Nyampa-jangka=n mangkirrkirr-wangu karri-nya? what-ela=2sg.s ear-priv be-pst ‘How come you can’t hear me?’ (PSM: TEN1-2018_001-02: 642121_645244)

It is also possible for ignoratives functioning as question words to be involved in inclusory constructions. On the basis of Cataldi’s translations, the ignorative ngana is used as a question word to query which member of the subject set ‘you two ’ in (1791) whereas nyampa in (1792) modifies an unexpressed subject nominal to query ‘which two’ of a larger set of individuals. (1791)

Ngana=npula ya-nku? who=2du.s go-pot ‘Which of you two will go?’ (PRN: TG1_0034-01)

(1792)

Nyampa=pula pi-nya? what=3du.s hit-pst ‘Which two were fighting?’ (LC56b)

11 Complex syntax This chapter describes syntax beyond the simple clause. Complex syntactic phenomena in Ngardi includes both subordination and coordination. The discussion of complex clause types is divided according to a major morphosyntactic division between finite complex clauses (comprising subordinate clauses (§11.1) and coordinate clauses (§11.2)) and non-finite complex clauses (§11.3). Subordinate clauses (both finite and non-finite) function either as relative clauses (modifying NPs) or as adverbial clauses (modifying entire clauses).542 Subordinating and coordinating morphology in Ngardi is nearly entirely characterised by sets of forms which are either analysed as polyfunctional (with roles in addition to subordinating or coordinating functions) or as homophonous with other lexemes. Finite subordinators543 take the form of particles (e.g. kuja, kaji, nyampa, etc.), a number of which are homophonous with demonstratives and are only distinguished by their inability to be inflected and constraints on their clausal positioning. Finite coordinators also take the form of uninflecting particles and are variously polyfunctional (serving as null auxiliary hosts for bound pronouns (e.g. ngu ~ nga) or are propositional modifying particles (e.g. kala) in simple, independent clauses. Finally, non-finite subordinators are polyfunctional case suffixes which were introduced in §4.5. Finite and non-finite (§11.3) clauses are easily distinguished from one another on a number of criteria. Finite subordinate clauses, unlike non-finite subordinate clauses, have a finite verbal predicate and their arguments are expressed by enclitic bound pronouns. Furthermore, finite subordinate clauses do not require an argument in the subordinate clause to be co-referent with an argument of the main clause (although this is also true of ‘obviative’ non-finite clauses, §11.3.6.2). Non-finite clauses exhibit none of these features. Instead, their tense and argument features are all interpreted relative to features of the main clause via various case suffixes and their subjects are generally required to be co-referent with an argument in the main clause. Finite subordinate clauses are slightly more difficult to distinguish from finite coordinated clauses (§11.2) or indeed from adjacent finite clauses.544 For at least one type, conditional kaji clauses, the difference is clear: the presence of the subordina542 Similar dual functions of subordinate clause are identified for most Australian languages. Hale (1976) refers to these as NP-relative and T-relative interpretations, respectively, while McGregor (1988, p. 37) makes reference to ‘adnominal’ and ‘adsentential’ modification. 543 A number of grammars of related Australian languages refer to subordinators as ‘complementisers’. Since there is very little evidence of subordinate clauses functioning as syntactic complements in Ngardi (although cf. kuja §11.1.3.2), I refer to them instead simply as ‘subordinators’. 544 Similar problems have been identified in other Australian languages, including Wambaya (Nordlinger, 1998) and Kuuk Thaayorre (Gaby, 2017, p. 552). https://doi.org/10.1515/9783110752434-011

11.1 Finite subordinate clauses 

 665

tor itself (i.e. kaji) can be taken as evidence of subordination as opposed to coordination. This is because kaji clauses cannot appear without an accompanying main clause. However, all other finite subordinators are homophonous with demonstratives (§11.1.1) and, under one view, these ‘subordinate’ clauses are indistinguishable from two adjacent finite clauses. Both clauses are finite and neither can be argued to be syntactically embedded or a constituent of the other. For Gurindji, McConvell (2006, p. 114) provides evidence for subordination by the fact that multiple embedded clauses can be identified – and the anaphoric reference of the demonstrative/subordinator is obligatorily linked to preceding embedded clauses. For Ngardi, I have no such examples. Instead, finite subordinate clauses (like coordinated clauses) are largely indistinguishable from main clauses (see similar comments by Senge (2015, p. 598) for Wanyjirra). However, one can point to various characteristics of subordinators themselves as evidence for subordinate status. The features of these lexemes (lack of inflection, constructionalised position, obligatory hosting of bound pronouns, incompatibility with auxiliaries) are not typical of demonstratives in any of their usual other functions. Furthermore, finite subordinate clauses generally form a single intonational unit with their main clauses – a feature not observed for independent clauses.545 Therefore, I follow McConvell (2006) in treating these ‘demonstrative-initial’ finite clauses as involving ‘subordination’ but acknowledge that the arguments for subordination are less convincing than for Gurindji.

11.1 Finite subordinate clauses Finite subordinate clauses in Ngardi are clauses with finite verb inflection which are logically and/or temporally related to a main clause via one of three independent subordinators, as shown in Table 140.546 Two subordinators (kuja and nyampa) are homophonous with a demonstrative and an ignorative, respectively. This relationship will be briefly addressed in §11.1.1. All finite subordinators: – are uninflecting (and are classified as a type of particle) – obligatorily host the bound pronouns – occur in initial or second position of the subordinate clause.

545 The status of ‘subordinate’ clauses in Wambaya (Nordlinger, 1998, p. 231) similarly hinges on intonational criteria alone. 546 The term ‘subordinate’ is used predominantly from a semantic or morphological perspective and in keeping with the tradition of terminology used in the description of Ngumpin-Yapa languages. While non-finite subordinate clauses show some evidence of syntactic embedding (they can appear within a main clause), finite ‘subordinate’ clauses are not strictly embedded.

666 

 11 Complex syntax

Table 140: Finite subordinators and their homophones. Subordinator

Function

Homophone

kaji (§11.1.2) condit

Temporal subordination ‘if, until, when’

–

kuja* (§11.1.3) sub

Relative and adverbial subordination ‘which, who, while’ Complement quotative ‘that’

Evocative ‘thus’

§5.2.2.1

nyampa (§11.1.4) ‘what’

Relative subordination ‘which, who’

Ignorative ‘what’

§5.3.1

In addition, finite subordinators cannot co-occur with any of the auxiliaries (ma, ngu/ nga). An example of a subordinator kuja in initial position hosting a bound pronoun is provided in (1793) and an example of a subordinator kaji in second position is provided in (1794). (1793) Kuja=rna ita=karri-nyani, nyanyi, ngu=rna Sturt Creek-ja sub=1sg.s small=be-ipfv.pst before seq.aux=1sg.s place_name-loc karri-nyani, teijin-ta. be-ipfv.pst station-loc ‘When I was small, a long time ago, I was staying at Sturt Creek Station.’ (MMN: TEN1-2018_038-03: 18071_31724) (1794) Ngaju=rna nya-ngankura ngurra-ngulu [ngapa kaji wanti-ju]. 1sg=1sg.s see-ipfv.hort camp-abl water condit fall-pot ‘I might watch from the camp if it starts raining.’ (PJA: TT76_2601: 718134_722704) Examples (1793) and (1794) also illustrate a second property of finite subordinate clause – they are generally unrestricted in their linear order and can appear before (1793) or after (1794) their accompanying main clauses. This is, in fact, the motivation for Hale’s (1976) use of the term ‘adjoined relative clause’ for similar clause types in Warlpiri which transparently reflects the peripheral placement of these clause types. However, it will be shown that these generalisations do not hold in all cases and that there is some vari-

11.1 Finite subordinate clauses 

 667

ation based on the specific subordinate clause type. In fact, it is necessary to assess the status of different subordinate clauses independently and on a construction-specific basis (see similar comments by Nordlinger, 2006; McGregor, 1988b; Gaby, 2017, p. 538).

11.1.1 Relation between subordinators and demonstratives As shown in Table 140, kaji is the only subordinator that is not homophonous with another lexeme. The two other subordinators are homophonous with demonstratives/ignoratives. The grammaticalisation of demonstratives into subordinators in Ngumpin-Yapa languages has been explored in some detail by McConvell (2006) who identifies two types of (finite) subordinator systems (complementisers in his terms): i) paradigm type: choice among a set of demonstratives in initial position, followed in some cases by a separate or adjoined ‘complementiser’ ii) specialised type: a single ‘complementiser’ descended from one demonstrative in initial or second position. McConvell (2006, p. 128) adds the following property of type (ii): the ‘complementiser’ occupies either first or second position in the subordinate clause, rather than ‘first position’ in the paradigm stage. McConvell (2006, p. 109) argues that, broadly speaking, western Ngumpin languages have a paradigm type system while all Eastern Ngumpin and Yapa languages follow a specialised type system. Ngardi, however, has a blended system of the paradigm and specialised types. Like specialised systems, Ngardi has a dedicated subordinator, kaji, which may appear in initial or second positions (although kaji has not been reconstructed to an historical demonstrative). In addition to kaji, Ngardi also has at least one demonstrative kuja and the ignorative nyampa, which have subordinating functions, typical of McConvell’s ‘paradigm type’.

11.1.2 Kaji clauses The subordinator kaji is utilised in two semantic types of subordinate clause: temporal subordinate clauses (x when y), (x until y), §11.1.2.1; and conditional subordinate clauses (if x, then y), §11.1.2.2.547 Some older speakers also used kaji as a type of nominal (NP) relativiser ‘who, which, what’, §11.1.2.3. The subordinator kaji is unique among the finite subordinators in that it overhwlemingly occurs initially within the subordinate clause.

547 The form kaji is only found in one other Ngumpin-Yapa language, Walmajarri (Hudson & Richards, 1990). It is likely related to katji ‘when, if, general subordinator’ in Kukatja (Valiquette, 1993, p. 460) and a clitic =kaji ‘when, while’ in Wangkajunga (Jones, 2011, p. 270).

668 

 11 Complex syntax

Temporal and conditional interpretations of kaji clauses co-vary with three morphosyntactic factors: i) tense/mood of the subordinate clause, ii) sequencing of the subordinate clause before or after the main clause, and iii) co-reference of the subject of the subordinate clause with the main clause. First, when events depicted by the subordinate clause are situated prior to reference time, the irrealis inflection signals a conditional clauses and past tense morphology signals a temporal clause. In non-past contexts the potential inflection is used and both conditional and temporal readings are available. Strictly speaking, the irrealis can be interpreted with any time reference, but it is the only inflection available for constructing past conditionals. Tensed clauses headed by kaji are interpreted as temporal subordinate clauses only. These facts are summarised in Table 141. Table 141: Interaction of TAM and interpretation of kaji clauses. Tense/Mood of sub. clause

Temporal reference of sub. clause

Interpretation of kaji clause

Example

irr pot

Past, non-past Non-past

(1809) (1807)

prs pst pst-narr

nil (habitual) Past Past

Conditional Temporal, Conditional Temporal Temporal Temporal

(1804) (1800) (1806)

The overlap in encoding of temporal and conditional relations by a single subordinator is not unusual for Australian languages (e.g. Wambaya (Nordlinger, 2006, p. 9)) nor cross-linguistically (Traugott et al., 1986). In Ngardi, this functional overlap is further complicated by kaji sharing a temporal relativising function with kuja as shown in (1795) and (1796). Unlike kaji, kuja cannot be used to form conditionals. (1795) Kaji=lu ngaju-ku kapulu nyin=ma-ni, ya-ni=lu condit=3pl.s 1sg-dat eZ lose=get-pst go-pst=3pl.s Karntawarra-kurra. place_name-all ‘So when they lost my older sister (i.e. she passed away), they went to Karntawarra.’ (MMN: Manungka01-035167) (1796) Ngaju=rna=yanu purda=nya-ngani-nyirra juju kuja=lu 1sg=1sg.s=3pl.o listen=see-ipfv.pst-narr ceremony sub=3pl.s yunpa-rnani-nyirra. sing-ipfv.pst-narr ‘I was listening to them while they were performing ceremony.’ (PSM: TEN12018_027-01: 1707614_1715230).

11.1 Finite subordinate clauses 

 669

In temporal kaji clauses, the subordinate clause may precede (1797) or follow (1798) the clause it is modifying. This variability generally co-varies with ‘when’ and ‘until’ temporal interpretations (discussed in §11.1.2.1). By contrast, conditional interpretations of kaji clauses typically are only possible when the subordinate clause precedes the main clause (1799) (the example in (1794) is an exception to the norm). This reflects a typologically common clausal sequencing in conditional clauses in the world’s languages (Diessel, 2005).548 (1797) Kaji ya-nku-rni, pamarr yu-ngka-rra=rla. condit go-pot-hith stone give-imp-hence=3sg.obl ‘When he comes, give the money to him.’ (PSM: TEN1-2018_034-01: 732736_738100)548 (1798) La-nku=rna kaji wirrpa karri-ju. pierce-pot=1sg.s condit many be-pot ‘I will dig until there is plenty (of bush potatoes).’ (JLI: TT76_2201: 1418550_1425505) (1799) Kaji=rna marda-rnanngi kurlarda, ma=rna=ngku condit=1sg.s have-ipfv.irr spear top.aux=1sg.s=2sg.o la-nngi. pierce-irr ‘If I’d had a spear, I would have speared you.’ (PSM: TEN1-2018_034-02: 621709_623375) Lastly, the co-reference of the subordinate clause subject constrains likely interpretations of kaji clauses. Conditional interpretations pattern with subject-controlled clauses (as in (1799)), whereas the temporal interpretation patterns with non-subjectcontrolled clauses (as in (1798)). This is a statistical correspondence rather than a strict constraint. 11.1.2.1 Temporal subordinate clause The subordinator kaji can index one of two temporal relations between events p and q: i) a simultaneous relation where p and q co-occur, i.e. ‘when’ ii) a prespan-end relation (Hollenbach, 1975) between p and q where p holds up to a point at which q commences and at which point p ceases, i.e. ‘until’.

548 As discussed above, kaji clauses involving potential inflection are formally indistinguishable from conditional clauses. Thus (1797) can equally be interpreted as ‘if he comes, give the money to him’.

670 

 11 Complex syntax

As introduced in §11.1.2, the two interpretations are not morphosyntactically distinguished; however, there is a tendency for ‘when’ clauses to involve variable clausal ordering (either before or after the main clause) whereas ‘until’ clauses have only been observed sequenced after the main clause. Further examples illustrating these two functions are (1800) and (1801), respectively. (1800) Kaji=lu ruwarra=ma-ni ya-ni=rnalu mayarra=lku wirlinyi. condit=3pl.s drag=get-pst go-pst=1pl.excl.s again=then hunting ‘While they pulled the truck out, we went off hunting again.’ (TJN: LC19a) (1801)

Jarra=nya-nganta=rnalu kaji pirriri=ya-nku. wait=see-prs=1pl.excl.s condit arrive=go-pot We are waiting until she arrives. (PSM: TEN1-2017_019-01: 1436900_1444587)

Virtually any TAM feature can appear in a main clause and take a subordinate kaji clause in a temporal adverbial function. Main clause commands with kaji subordinate clauses are shown in (1802) and (1803). (1802)

Jina=marda-ka=pulany kaji=pula mangarri nga-lku lurrij. feet=have-imp=3du.o condit=3du.s veg_food eat-pot finish ‘Wait for those two (women), until the two of them have finished eating.’ (PSM: TEN1-2018-027-01: 2423409_2435204)

(1803)

Karri-ya=wu wurra kaji=rlijarra karlarra kuja nya-nku be-imp=voc wait condit=1du.excl.s west thus see-pot minya, karlarra. this west ‘Stay put while we two look around here in the west.’ (TJN: LC21a: 435512_440821)

Present tense kaji clauses are typically interpreted as having unspecified temporal reference. In effect, they are statements of habit or customary practise as indicated by the translations for (1804). (1804) Jangu=rnalu ma-nanta kaji=rnalu assump=1pl.excl.s get-prs condit=1pl.excl.s tatakurr=karri-nyanta wurna-jangka. tired=be-prs journey-ela ‘Truly, we collect it (the snake vine) when we are tired from travelling.’ ‘Truly, we get it (the snake vine) if we are tired from travelling.’ (TJN: Cataldi, 2011, p. 242)

11.1 Finite subordinate clauses 

 671

While there are no strict constraints on argument co-reference for kaji clauses in temporal functions, temporal interpretations are found more frequently in non-subject construal situations. In fact, kaji clauses most commonly lack any co-referent argument in the main clause (examples were (1798), (1800) and (1801)). A temporal kaji clause controlled by an indirect object is shown in (1805). (1805) Jina=marda-ku=rli=rla kaji ya-nku warriri=lku. foot=have-pot=1du.incl.s=3sg.obl condit go-pot close=then ‘Let’s wait for itx, until itx comes closer.’ (MMN: TEN1-2017_026-01: 2516068_2526169) An additional possibility is for a non-subject argument of the subordinate clause to be co-referent with a main clause subject (e.g. the indirect object in (1797)). Lastly, there are some instances where the main clause accompanying the kaji clause contains a demonstrative which itself is co-referent with the entire proposition depicted in the kaji clause as in (1806). (1806) Kaji=lu=yanu yu-ngani-nyirra yala ngaringka, [yala=yi condit=3pl.s=3pl.o give-ipfv.pst-narr that woman that=emph nyinpa=lku]. forever=then ‘[When they used to give the woman to them]x, that x was forever.’ (MAM: TEN1-2016_002-01: 783338_795061) 11.1.2.2 Conditional subordinate clause The subordinator kaji may also introduce a type of conditional ‘if . . . then’ subordinate clause. Conditional clauses necessarily involve hypothetical events and require potential (1807) or irrealis (1808), (1809) inflection on the verb of the subordinate clause. As discussed in §11.1.2, the potential inflection is used for non-past conditionals while the irrealis may be used with any time reference but most often encodes past conditionals. (1807) Kaji=yi=n money yu-ngku ngu=rna=ngku condit=1sg.o=2sg.s money give-pot seq.aux=1sg.s=2sg.o mangarri=wali ma-ninja-nku. veg_food=then get-move&x-pot ‘If you give me some money, I will go and get you some food.’ (PSM: TEN12017_026-02: 627086_633262)

672 

 11 Complex syntax

(1808) Kaji=rna marda-rnngi kuturu, ngu=rna=ngku pi-ngi. condit=1sg.s have-irr nulla_nulla seq.aux=1sg.s=2sg.o hit-irr ‘If I’d had a nulla nulla, I would have attacked you.’ (MMN: TEN1-2018_02601: 2940536_2958740) (1809) Kaji=yi=n=ku=rla ya-nngi, kuyi=rna=ngku yi-nngi. condit=1sg[lct]=2sg.s=ep=lct go-irr meat=1sg.s=2sg.o give-irr ‘If you had come to me, I would have given you money.’ (MMN: TEN12018_026-01: 2680199_2693946) Subordinate kaji clauses with potential inflection have both conditional and temporal meanings available, as in (1810). The clausal sequencing of the protasis before the apodosis is typical of the conditional interpretation however. (1810) Kaji=lu wirrpa pirri=ya-nku warlu=rna ma-nku condit=3pl.s many arrive=go-pot fire=1sg.s get-pot wirrpa=wali. many=then ‘If more people arrive, I will get more wood.’ ‘When more people arrive, I will get more wood.’ (MMN: TEN1-2017_026-01: 2090106_2098790) Moreover, if a speaker uses a modal particle in the apodosis, e.g. parda (1811), (1812), or ngarda (1813), then generally only conditional interpretations are observed. In the absence of these modal markers, the potential inflections alone could be interpreted with either intentional (‘will’) or possible (‘might’) modal meanings. (1811)

Kaji=rlipa ya-nku parda=rna=ngku milki=yirra-ku condit=1pl.incl.s go-pot dub=1sg.s=2sg.o show=place-pot yala purnu parntarr-a. that tree hot-loc ‘If we go, I might show you that tree on the hot ground.’ (TJN: Cataldi, 2011, p. 328)

(1812)

Kaji=yi pu-ngku ngantany-ju mu-rlu, ngaju ma=rna condit=1sg.o hit-pot man-erg this-erg 1sg top.aux=1sg.s pu-ngku-kariny=parda. hit-pot-recip=dub ‘If this man hits me, I might hit him back.’ (PSM: TEN1-2019_005-01: 1289675_1294319)

11.1 Finite subordinate clauses 

 673

(1813) Kaji=rna kuyi ka-ngku-rni kupa-ku ngarda=yi=n. condit=1sg.s meat carry-pot-hith cook-pot hyp=1sg.o=2sg.s ‘If you bring the meat here, you might cook it for me.’ (PSM: TEN1-2019_007-02: 488530_495759) Lastly, a conditional subordinate clause can even be formed with a verbless clause, as in (1814).549 (1814) Kaji=n=ku=rla ruju-ku ya-nta kaniny kiliki-kurra. condit=2sg.s=ep=3sg.obl play-dat go-imp down creek-all ‘If you want to play go down to the creek! (lit. ‘If you (are) for play, go down to the creek!’)’ (MMN: TEN1-2018_026-01: 2018291_2032456) In conditional clauses, the protasis (the clause expressing the condition) is always encoded in the subordinate clause headed by kaji, and the apodosis (the main clause expressing the conclusion/outcome) lacks any specific morphological marking. This is consistent with typological observations that conditionals are typically marked in the protasis (Comrie, 1986, p. 87). The partial exceptions to this are the possibilities for either the sequential auxiliary ngu= (1842) or the ‘subsequent clitic’ =wali (735) – or both (1807) – to optionally mark the apodosis. In both cases, such optional marking is used to emphasise the subsequent nature of an apodosis in a conditional construction and is analogous to the optional use of ‘then’ as a ‘correlative proform’ in English conditionals, for example, ‘if I remember, then I will come’ (Bhatt & Pancheva, 2007). 11.1.2.3 General subordinator (some speakers only) Cataldi’s corpus contains examples of kaji functioning as a generic subordinator heading a range of additional subordinate clauses. This is reflected in Cataldi’s gloss of kaji, covering the English relativisers ‘who’, ‘which’ and ‘that’ and is similar to how a subordinator of the same form is used in the Western Desert language Kukatja (Valiquette, 1993, p. 460). The variation observed in the data may represent variation in different speakers’ multilingual repertoires – especially since most nominal relativisation in Ngardi is achieved via the subordinator kuja (§11.1.3), which has clear cognates with other demonstrative/subordinators in other Ngumpin-Yapa languages. Example (1815) illustrates kaji with a general subordinating function. The potential Kukatja influence on this type of usage is perhaps reinforced in this example by the presence of a code-switched Kukatja enclitic =lpi.

549 Example (1814) is somewhat anomalous, however, in that the purpose subordinate clause involving the nominal ruju would generally not be expected to be cross-referenced in the bound pronoun.

674 

 11 Complex syntax

(1815) Yalu-rla nawu kujun-ta=lpi, kaji nguku-ngku parntik=pi-nya. that-loc now cushion-loc=?550 sub water-erg flatten=hit-pst ‘There now (on the) cushions, which the water soaked.’ (MRN: Cataldi, 2011, p. 97) Finally, there are some apparent non-subordinating usages of kaji. In (1816) and (1817), for instance, kaji seemingly functions as a null host for bound pronouns. I am unable to provide any motivation for such examples. (1816) a.

Ngaju=rna=rla kuja ya-nku karlarra-purda=wu. 1sg=1sg.s=3sg.obl thus go-pot west-wards=voc ‘I will go into the west, therefore, for it.’ b. Kaji=rna parli=pu-ngku ngamarna=rlangu la-nku=rna. sub=1sg.s find=hit-pot yam_sp=also pierce-pot=1sg.s ‘I will find things like yams, and I will dig for them.’ (TJN: LC22a: 1194418_1198255)

(1817)

a.

Nyarrpa ma-nku=rlipa nguku? how get-pot=1pl.incl.s water ‘What are going to do about this water?’ b. Ngayi kaji=ngala munga=karri-ju. indeed sub=1pl.incl.o dark=be-pot ‘Indeed, it’s going to get dark on us.’ (PPN: Manungka_01-035160: 1543221_1546332)

11.1.3 Kuja clauses The subordinator kuja can introduce a wide range of semantically diverse finite subordinate clauses. Finite subordinate clauses headed by kuja may function as: i) nominal and temporal relativisers (§11.1.3.1) a. relative clauses modifying an NP b. subordinate clauses modifying a clause (with adverbial functions) ii) complement quotatives (§11.1.3.2) iii) insubordinate clauses (§11.1.3.3).

550 This is likely a borrowing of Kukatja =lpi ‘finally, end up doing’ (Valiquette, 1993, p. 459).

11.1 Finite subordinate clauses 

 675

11.1.3.1 Nominal and temporal relativisation Kuja clauses operating as adjuncts can serve both adverbial and subordinate clause functions.551 Semantically, these serve to provide information about nominal reference (similar to a ‘restrictive relative clause’) or adverbial specification (as a type of temporal subordinate clauses). In many cases it is not possible to identify one usage over the other since there are no formal properties that distinguish them. The semantic interpretations available to a single kuja clause are exemplified in (1818). (1818) Jingka-ngku nya-nganta ngantany kuja paja-rnanta purnu. child-erg see-prs man sub bite-prs tree ‘A child watches the man that is cutting the tree.’ ‘A child watches the man while he cuts the tree.’ (MMN: TEN1-2018_023-02: 6997224_7004753) In its adverbial interpretation, kuja typically introduces a temporal adverbial clause which is simultaneous to the main clause, as in (1818). In some rare instances, kuja may introduce a clause in an explanatory relation to the main clause (‘because x’) as in (1819). (1819) Ngarri-rnanta=rna=ngku nyuntu kuja=yi=n pi-nya yala tell-prs=1sg.s=2sg.o 2sg sub=1sg.o=2sg.s hit-pst that kunyarr ngaju-ku. dog 1sg-dat ‘I am scolding you because you hit my dog.’ ‘I am scolding you, you being the one who hit my dog.’ (MMN: TEN1-2017_00102: 2932878_2938682) In both nominal-relativising and adverbial subordinating functions, kuja clauses can either follow (1819) or precede (1820) the main clause. (1820) Kuja=yi=n jangku=pi-nya, yalu=rna=ngku marda-rnanta. sub=1sg.o=2sg.s promise=hit-pst that=1sg.s=2sg.o have-prs ‘That which you promised me, I have for you here.’ (DMN: LC42b: 358280_362274) Subjects of kuja clauses can be co-referent with any constituent of a main clause, including transitive subjects (1821), intransitive subjects (1822), objects (1823), indirect objects (1825), as well as locative (1826) and allative (1827) subjuncts. Subordinate kuja clauses need not be adjacent to the main clause argument with which they show co-reference.

551 These are termed T-relative and NP-relatives clauses by Hale (1976, p. 79).

676 

 11 Complex syntax

(1821) Ngaringka-rlu marda-rnani jingka kuja kupa-rnani mangarri. woman-erg have-ipfv.pst child sub cook-ipfv.pst veg_food ‘The woman was holding a child while cooking food.’ (MMN: TEN1-2018_02701: 2483130_2503146) (1822) Kuja=rna ita=karri-nyani, nyanyi, ngu=rna sub=1sg.s small=be-ipfv.pst before seq.aux=1sg.s Jurt Krik-ja karri-nyani, station-ta. place_name-loc be-ipfv.pst station-loc ‘When I was little, a long time ago, I was staying at Sturt Creek Station.’ (MMN: TEN1-20187_038-03: 18071_31724) (1823) Ngaju-ngku=rna parli=pu-ngku kuja wirrkirl=mipa pit=karri-nyanta. 1sg-erg=1sg.s find=hit-pot sub head=restr emerge=be-prs ‘I can find (game) when just a head is sticking out of the grass.’ (MMN: TEN12018_026-01: 3304857_3322876) (1824) Wangka-nyanta=rna=ngku kuja=n luk=karri-nyanta. speak-prs=1sg.s=2sg.o sub=2sg.s lie_down=be-prs ‘I am talking to you while you are lying down.’ (MMN: TEN1-2018_025-01: 3792712_3800375) (1825) Yi-nya=rna=rla mangarri yalu-ku kuja jalangu give-pst=1sg.s=3sg.obl veg_food that-dat sub today pirriri=ya-ni. arrive=go-pst ‘I gave the food to the girl who just arrived.’ (MMN: TEN1-2019_007-02: 414337_420361) (1826) Ngaringka yala yut=karri-nyanta=rlanyanta kuja kupa-rnani woman that sit=be-prs=3sg.lct sub cook-ipfv.pst mangarri yalu-ngku. veg_food that-erg ‘That woman is sitting with the one who cooked the food.’ (MMN: TEN12019_007-02: 230712_247880) (1827) Kunyarr-ta=rna=rlanyanta pula-nya kuja ma-ni kuyi. dog-loc=1sg.s=3sg.lct shout-pst sub get-pst meat ‘I shouted at the dog that took the meat.’ (MMN: TEN1-2019007-02: 268662_282316)

11.1 Finite subordinate clauses 

 677

In (1828), a kuja clause is modifying an adjunct in the allative rather than a main clause argument. (1828) Ka-ngani=lu=yanu wurna nawu Sturt Creek-kurra, carry-ipfv.pst=3pl.s=3pl.o journey now place_name-all kuja=lu=yanu kampa-rnani wel-ta. sub=3pl.s=3pl.o cook-ipfv.pst well-loc ‘They took them, on the move then, to Sturt Creek, where they burnt them in the well.’ (MMN: TEN1-2019_001-04: 243862_250829) 11.1.3.2 Complement ‘quotative’ The subordinator kuja can introduce a special type of subordinate clause which functions as a complement to the predicate wangka- ‘speak, talk’.552 In this context alone, kuja can appropriately be termed a ‘complementiser’. (1829) Wangka-nya=yi kapulu ngaju-ku kuja ya-nani-nyirra karla-ngurlu. speak-pst=1sg.o eZ 1sg-dat sub go-ipfv.pst-narr west-abl ‘My sister said to me that she came from the west.’ (PRN: FT1_b: 535088_538345) 11.1.3.3 Insubordinate clauses A distinct type of kuja clause is one where it seemingly appears without any accompanying main clause. I term these ‘insubordinate’ kuja clauses. Analogous functions of similar subordinators have been identified in Wanyjirra (Senge, 2015, p. 489) and Jaru (Tsunoda, 1981b, pp. 165–171).553 It is often difficult to ascertain whether these clause types are simply non-case-marked instantiations of the adverbial demonstrative kuja (i.e. absolutive) or true insubordinate clauses. As argued by Senge (2015, p. 489) and Tsunoda (1981b, pp. 165–171) for equivalent clauses in Wanyjirra and Jaru, the main function of these constructions is to pragmatically link a sub-event to preceding subevents in a wider discourse context. (1830) a. Ya-ni=rnalu mu-kurra nawu, Wirrimanu-kurra. go-pst=1pl.excl.s this-all now place_name-all ‘Then we came here, to Wirrimanu.’

552 Similar functions are reported for Wanyjirra guya (Senge, 2015, p. 607) and Warlpiri kuja (Legate, 2008a). 553 For Wanyjirra, Senge argues that guya ‘shows no syntactic dependency between clauses but pragmatic dependency, i.e. linking a sub-event to preceding sub-events in a larger context’ (Senge, 2015, p. 489). Tsunoda analyses Jaru guwa ~ guja in these contexts as something akin to discourse pivots marking the boundary between old and new information (Tsunoda, 1981b, pp. 165–171).

678 

 11 Complex syntax

b. Kuja=rna lurrij=karri-nyani skul-ku. sub=1sg.s finish=be-ipfv.pst school-dat ‘Then I was finishing off school.’ (MMN: TEN1-2018_038-03: 800115_ 809403) (1831)

a. Puuj-a nyanyi. bush-loc before ‘(In the bush) in the olden days.’ b. Yuwa, jina kuja=rnalu ya-nani. yes foot sub=1pl.excl.s go-ipfv.pst ‘Yes, we went by foot.’ (YDN: LC24a: 845253_848080)

Insubordinate usages (1830) and (1831) can be contrasted with a clause-initial usage of the adverbial demonstrative kuja ‘thus’ in (1832) which, from context, was clearly intended as an evocative ‘like this/like so’. Further systematic investigation is needed to determine whether there are further properties that distinguish adverbial demonstrative kuja from subordinator kuja or indeed if this distinction is poorly motivated. (1832) Kuja=rna juwantarra=karri-nyanta. thus=1sg.s legs_straight=be-prs ‘I am sitting stretched out like so.’ (DMN: LC48a: 1888228_1890557)

11.1.4 Nyampa clauses Of all the ignoratives (§5.3), nyampa alone has a subordinating function. However, as for kuja (§11.1.3), it is difficult to definitively assert whether nyampa is strictly functioning as a subordinator or is merely an ignorative appearing in a semi-constructionalised context involving closely juxtaposed finite clauses. Alternative translations for nyampa are presented in (1833) and (1834) to reflect this uncertainty in analysis. Additionally, not all speakers appear to use the ignorative in these subordinator-like contexts. Across the corpus, kuja is by far the default strategy for forming relative clauses with NP-modifying functions. (1833) Nyampa=n marda-rnanta jirri=yirra-ka! what=2sg.s have-prs show=place-imp ‘Show me what you are holding!’ ‘You are holding something. Show me!’ (PSM: TEN1-2019_007-02: 110336_116259)

11.2 Finite coordinated clauses 

 679

(1834) Jalparr nyampa=rnalu ma-ni-nyirra kulini-mpala. tomato what=1pl.excl.s get-pst-narr south-along ‘Bush tomatoes, which we used to get in the south.’ ‘Some sort of bush potatoes, we used to get them in the south.’ (TSM: TT75_1301) [Answer in response to ‘What bush tucker were you getting?’] Where nyampa is case-inflected, it is unambiguously interpreted as an ignorative nominal. Thus, compare (1833) and (1834) with (1835). (1835) Ngaju ma=rna=ngku palka=marda-rnanta nyampa-ku=yi=n 1sg top.aux=1sg.s=2sg.o alive=have-prs what-dat=1sg.o=2sg.s ngarri-rnanta. tell-prs ‘I have iti for you, you were asking me for iti.’ (DMN: LC42b: 338332_341378) * ‘I have that which you were asking me for.’

11.2 Finite coordinated clauses Ngardi possesses a number of finite complex clause types which share a close semantic or logical relationship but are in a paratactic rather than a hypotactic relationship. Coordinated clauses include those without an overt coordinator, known as ‘asyndetic coordination’ (§11.2.1) or those for which a semantic relation is specified via a coordinator.554 Coordinators in Ngardi are polyfunctional lexemes with both main clause functions and coordinator-specific functions as detailed in Table 142. In addition to the polyfunctional lexemes in Table 142, a number of English/Kriol lexemes have been borrowed into the speech of Ngardi speakers with coordinative functions. These include an (11.2.6), tumaji (§11.2.7), and bat (§11.2.8). Finite coordinate clauses (unlike their subordinate equivalents) are generally subject to strict linear orderings. In the description of the coordinated clause types I will refer to ‘primary’ and ‘secondary’ clauses in contrast to ‘main’ and ‘subordinate’ used in the description of subordination. A general property of coordinated clauses is that any overt marking of coordination occurs within the secondary clause and the ordering of the clauses is strictly iconic of the temporal or logical relations they depict.

554 Although they are somewhat similar to what are termed ‘correlative’ clauses in descriptions of Indo-Aryan languages (see Nordlinger, 2006, p. 11).

680 

 11 Complex syntax

Table 142: Finite coordinating morphology. Word form

Gloss

Main clause function

Coordinator function

ngu, nga

seq.aux

Null auxiliary

§10.1.1

ngarda

hyp

Hypothetical (modal particle)

§10.2.4

kala

asrt

§10.3.2

(=)wali

‘then’

Assertive (propositional particle) Various (generally marks a subsequent eventuality)

§10.4.2

Sequential ‘then, so that’ Apprehensional ‘lest’ Contrastive ‘but (in fact)’ Subsequent ‘then’

§11.2.1 §11.2.3 §11.2.4 §11.2.5

11.2.1 Asyndetic coordination It is possible to coordinate two clauses in Ngardi without any overt coordinator, simply by juxtaposing two clauses as in (1836). (1836) Jamuny pirri=ya-ni, ya-ni=rlipa. ready arrive=go-pst go-pst=1pl.incl.s ‘As soon as he arrived, we left.’ (PSM: TEN1-2018_034-01: 892036_895816) Such juxtaposition of finite clauses, known cross-linguistically as ‘asyndetic coordination’, is particularly frequent in narratives in the Ngardi corpus and is used to link a series of related clauses. Multiple asyndetic clauses appearing in close succession and contained under a single intonational contour (1837) are often used as a stylistic device in Ngardi stories, as in (1837). These clauses are often accompanied by dramatic lengthening of the final vowel of the penultimate or final constituent. Interestingly, the same discourse effect can be achieved by coordination of a series of insubordinated non-finite clauses (as described in §11.3.8). (1837) Ya-nani=rnalu ya-nani=rnalu ya-nani=rnalu, go-ipfv.pst=1pl.excl.s go-ipfv.pst=1pl.excl.s go-ipfv.pst=1pl.excl.s ya-nani=rnalu. go-ipfv.pst=1pl.excl.s ‘We went and went and went and went.’ (BTI: TEN1-2017_011-03: 388769_394516) Note that juxtposition is also the default means of coordinating phrase-level constituents, such as appositional NPs (although cf. the conjunction manu, §4.2.4.1).

11.2 Finite coordinated clauses 

 681

11.2.2 Sequential coordination ngu ~ nga The sequential auxiliary ngu ~ nga was introduced in §10.1.1. In this section, its role in marking a sequential relationship between two clauses is described in further detail. The sequential functions of ngu ~ nga span both temporal (x then y) and purposive (x so that y, x in order that y) senses. In its purposive role, ngu ~ nga clauses are semantically equivalent to dative-marked non-finite subordinate clauses (§11.3.5.1) and express the purpose, or objective of the event depicted by the primary clause. This can be neatly demonstrated in (1838) where Tsunoda elicited a Ngardi finite subordinate clause (aʹ) from a Nyininy non-finite subordinate clause (a).555 (1838) a. Ngaju nga=rna ya-nan ngaba-wu ma-nu-wu. 1sg cat=1sg go-prs water-dat do-inf-dat ‘I am going (off) to drink water.’ [Nyininy prompt] (TTT: TT76_2601: 852105_865774). aʹ. Ngapa-kurra=rna ya-nanta ngu=rna nga-lku, parrajarda-rlu. water-all=1sg.s go-prs seq.aux=1sg.s eat-pot thirsty-erg ‘I am going to the water, so that I may drink thirstily.’ [Ngardi] (PJA: TT76_2601: 865874_871433) In (1839), Patrick Smith reflects on the possibility to use either a dative-marked nonfinite subordinate clause (§11.3.5.1), marked []x, or a finite subordinate clause with sequential ngu, marked []y, to express an event depicting the purpose of the main clause. (1839) Ngapa=rna ma-nku [yarli-rnu-ku]x or [ngu=rna=nyanu water=1sg.s get-pot wash-inf-dat seq.aux=1sg.s=refl yarli-ku]y. wash-pot ‘I will get water so that I can wash (myself).’ (PSM: TEN1-2017_026-02: 830725_834719) In Ngarinyman (McConvell, 1995, p. 11), a cognate ‘auxiliary form’ ngu ~ nga is analysed as a subordinator rather than a coordinator. Such an analysis is ill-suited to Ngardi since clauses headed by ngu ~ nga can appear independent of a main clause and do so without any implied main clause interpretation (as clearly demonstrated in §10.1.1.2). Clause-coordinating functions of ngu ~ nga are, however, clearly apparent

555 A direct transliteration of the Nyininy in (a.) into Ngardi would be ‘Ngaju nga=rna ya-nanta ngapa-ku ma-nu-ku’.

682 

 11 Complex syntax

in both narratives and procedural texts where a speaker details a specific sequence of events or procedures, as in (1840).556 (1840) Ngu-rna kupa-ku, turn=kupa-ku. seq.aux=1sg.s cook-pot, cover=cook-pot Ngu=rna pirdiny=ma-nku, ngu=rna nga-lku kuyi. seq.aux=1sg.s arise=get-pot seq.aux=1sg.s eat-pot meat ‘I will cook it and bake it (cover it in hot ashes/sand). Then I will take it out (of the fire), and then I will eat the meat.’ (TJL: TT76_2201: 1143802_1150334) The coordination of clauses via the sequential auxiliary is set in opposition to serialised verb constructions (§8.2) where closely related and sequential events are not formally marked for their temporal succession but are instead integrated into a single clause with a single set of bound pronouns. Unlike serial verb constructions, coordinated ngu ~ nga clauses can encode a wide range of logico-temporal relationships between clauses, as illustrated in Table 143. Interpretations are organised according to the TAM value of the secondary clause. I include the unique clause types created by the kaji ‘conditional’ which may occur in the main clause preceding a ngu ~ nga coordinated clause. Examples illustrating the various speech act functions of coordinated ngu ~ nga clauses follow. Table 143: Interaction of TAM and interpretation of sequential auxiliary ngu, nga in coordinated clauses. TAM in clause type

Interpretation

Examples

Speech act function

Main

Secondary

pst prs

pot pot

Logical Logical

(1841) (1838), (1842)

pot pst imp

pot pst pot

Logical Temporal Logical

(1843) (1844) (1846), (1847), (1848)

Reasoned intention (pst) Reasoned intention, Capacity (non-pst)557 Reasoned intention (fut) Past sequence (narration) Reasoned command

kaji + irr, pot

(Any)

Logical

(1849)

Conditional, Hypothetical

556 Note that turn=kupa-ku forms a serialised verb construction with the preceding verbal predicate and lacks its own pronominal arguments. 557 The ‘non-past’ categorisation is needed to encompass ‘intentional usages’ of the present tense (typically with motion verbs) which may described events sequenced after speech time, as in (1838).

11.2 Finite coordinated clauses 

 683

[i] Reasoned intention (Past) (1841) Ngaju=rna ya-nurni, ngu=rna=nyurrarla wangka-ju 1sg=1sg.s go-pst.hith seq.aux=1sg.s=2pl.lct speak-pot nyurrurla-rla. 2pl-loc ‘I came here so that I can speak with all of you.’ (PSM: TEN1-2017_026-02: 853132_862350) [ii] Capacity (Non-past) (1842) Ngaju=rna marda-rnanta rararr-kaji, ngu=rna=ngkupula ka-ngku. 1sg=1sg.s have-prs run-nmlz seq.aux=1sg.s=2du.o carry-pot ‘I have a car so I can take you two.’ (PSM: TEN1-2017_027-01: 893717_901638) [iii] Reasoned intention (Potential) (1843) Ka-ngku=rna=ngkupula ngu=yi=npula jirri=marda-ku ngurra. carry-pot=1sg.s=2du.o seq.aux=1sg.o=2du.s show=have-pot country ‘I will take you two so that you two can show me the country.’ (MMN: TEN12017_027-01: 948822_958105) [iv] Past sequence (Past) (1844) Jamuny=pa=rna pirdiny=karri-nya ngu=rna parli=pi-nya.561 ready=ep=1sg.s arise=be-pst seq.aux=1sg.s find=hit-pst ‘I had just woken up when he found it.’ (MMN: TEN1-2018_023-02: 7407420_7419309) (1845) Ya-ni, nga pirriri=ya-ni. go-pst seq.aux arrive=go-pst ‘He went and then he arrived.’ (TJN: Manungka_01-035160: 711903_716037) [v] Reasoned command (Imperative)558 (1846) Warlu yirra-ka ngu=rlipa kupa-ku kuyi. fire place-imp seq.aux=1pl.incl.s cook-pot meat ‘Make a fire so that we can cook the meat.’ (MMN: TEN1-2017_027-01: 866462_871062)

558 Cf. temporal function of jamuny ‘soon’ in juxtaposed clauses, e.g. (1836).

684 

 11 Complex syntax

(1847) Yalu-rla yirra-ka ingkarna, ngu=rlipa kupa-ku. that-loc place-imp kindling seq.aux=1pl.incl.s cook-pot ‘Put that kindling down there, so that we can cook.’ (PSM: TEN1-2018_00401: 473976_479665) (1848) Yunpa-ka=rla jingka-ku ngu pukany=karri-ju. sing-imp=3sg.obl child-dat seq.aux sleep=be-pot ‘Sing to that child so he can go to sleep.’ (MMN: TEN1-2017_003-02: 1746418_1757907) A special type of ngu ~ nga clause is one functioning as an apodosis of a conditional clause introduced by kaji (§11.1.2), as in (1849). Note that ngu here is not strictly a coordinator, since kaji can link two clauses (as a subordinator) independently without any additional marking of the apodosis. Here ngu ~ nga is simply an optional means to further mark the subsequent nature of apodoses in conditional clauses. (1849) Kaji kawurr=karri-ju, yala ngapa, ngu=rna kupa-ku condit hot=be-pot that water seq.aux=1sg.s cook-pot ngurlukaji. tea ‘If it (the water) is hot, then I will make the tea.’ (MMN: TEN1-2017_027-01: 1047193_1060480) Lastly, an example of a ngu clause sequenced after a finite subordinate clause is provided in (1850). While Ngardi lacks evidence of multiple levels of syntactic embedding, this example indicates how ‘semantic’ embedding can be achieved in Ngardi. (1850) Ngantany=pa=rna nya-nganta yala [kuja=rla wurru=ka-nganta] man=ep=1sg.s see-prs that sub=3sg.obl sneak=carry-prs [ngu la-nku]. seq.aux pierce-pot ‘I am watching that man who is sneaking up on that (animal) so he can spear it.’ (PJA: TT76_2601: 1111853_1117880)

11.2.3 Apprehensional coordination ngarda The hypothetical modal particle ngarda was described in §10.2.4. In independent simple clauses, it encodes dubitative modality (epistemic uncertainty). Additionally, a ngarda clause can also co-occur with another main clause in a coordinate structure where it has an apprehensional function, that is, ‘lest x’, ‘for fear that x’ (Vuillermet,

11.2 Finite coordinated clauses 

 685

2018).559 A wide range of clause types may serve as primary clauses in apprehensional constructions, including imperatives (1851), admonitives (1852), indicatives (1853), and (intentional) potential (1854). The primary clause and the ngarda coordinator are bolded in the following examples. (1851)

Ngarda mayawun-tu luwa-rnanngirda tuwa yala nyit=pa-nta. hyp wind-erg shoot-anticip door that close=get-imp ‘Shut the door, lest the wind slam it just now!’ (PSM: TEN1-2018_037-01: 125568_136481)

(1852) Wakurra=n ya-nku ngarda=ngku=lu pi-ngi. neg=2sg.s go-pot hyp=2sg.o=3pl.s hit-irr ‘You shouldn’t go lest they hit you.’ (PSM: TEN1-2018-028-01: 3756745_3765943) (1853) Ngu=lu=yanu wirrin=ma-ni yard-ta=wali ngantany seq.aux=3pl.s=3pl.o bind=get-pst yard-loc=then man jayin-kurlu jina ngarda=lu yaparti-ngi. chain-prop foot hyp=3pl.s run-irr ‘Then they tied the men up by their feet in the yard (along the posts) with chains lest they run away.’ (PSM: TEN1-2017_007-02: 72766_82623) (1854) Ngu=rna karri-ju ngaju-kariny ngarda=rna pi-ngi ngawu. seq.aux=1sg.s be-pot 1sg-other hyp=1sg.s hit-irr bad ‘I will stay by myself in case I get bad (meat).’ (TJN: Manungka_01-035159: 664905_668806) Apprehensional clauses serve a wide range of speech act functions largely dependent on the TAM of the primary clause. With imperatives and prohibitives, the speech act function is that of a warning: i.e. ‘don’t x lest y’. In the case of indicatives (past and present inflections) and intentives (potential inflections), the speech act function of apprehensional ngarda clauses is to provide reasoned motivation for one’s actions: ‘x (did/does/will do) y in case of z’.

559 Analyses of ‘apprehensional morphology’ in Ngumpin-Yapa languages vary significantly for different languages. Senge (2015), for instance, treats ngarra ‘admonitive’ as a subordinator in Wanyjirra and accounts for its appearance in independent clauses as ‘insubordination’, which is not unusual in Australian languages (Evans, 2007). Senge goes on to say that its single-clause use is somewhat ‘conventionalised’ and has arguably been reanalysed as an auxiliary used with future tense verbs expressing admonitive mood (Senge, 2015, p. 642). Similar phenomena are observed in neighbouring languages, e.g. Jaru (Tsunoda, 1981b, p. 204) and Bilinarra (Meakins & Nordlinger, 2014, p. 420).

686 

 11 Complex syntax

As described already for independent ngarda clauses, apprehensional clauses typically co-occur with potential or irrealis inflections. The following is an example with a hortative inflection with no observable difference in meaning from a potential. (1855) Tarn=yirra-rnanta=rla iliri ngarda wirriri ya-nankura cache=place-prs=3sg.obl hide hyp return go-ipfv.hort karrmantawangu. hungry ‘I am putting (the food) in a cache for her in case she comes back hungry.’ (PSM: TEN1-2018_007-03: 65573_84335) A different semantic relationship between primary and secondary clauses involving ngarda can be observed in contexts where the primary clause involves a fear predicate (1856). Here the semantic fuction of ngarda cannot be interpreted as ‘lest, for fear that’ since the ‘fear’ component is semantically encoded by the main clausal predicate. Instead ngarda here simply expresses a hypothetical (and feared) event and is coordinated with the primary clause without any specified semantic component. (1856) Rayin=pa=rna karrarda=karri-nyanta, ngarda=yi=lu pi-ngi. afraid=ep=1sg.s afraid=be-prs hyp=1sg.o=3pl.s hit-irr ‘Fearful, I am afraid, (that) they might attack me.’ (DMN: LC42b: 1471175_1475516) ≠ ‘I am afraid in case they attack.’ Sometimes a feared event will be expressed in a ngarda clause in addition to overt marking of a feared nominal already having been specified in the main clause via an evitative case-marked NP, as in (1857). (1857) Ya-nku=rna yalu-rlamarra, ngarda wangka-nyanku. go-pot=1sg.s that-evit hyp speak-ipfv.pot ‘I am going to go on account of that one, she might talk.’ (MMN: TEN-2017_01601: 1168486_1172039)

11.2.4 Contrastive coordination kala The coordinator kala is homophonous with the assertive propositional particle kala ‘assertive’ as described in §10.3.2. In its coordinating function it appears to have a meaning similar to English ‘but’ or ‘however’ as in (1858)–(1861).560 A contrastive kala clause is typically sequenced after a main clause against which the proposition is being

560 A Warlpiri propositional particle kala shares the same function (see Laughren, 1982a, p. 149).

11.2 Finite coordinated clauses 

 687

evaluated as being either in a state of contrast or constituting an unpredicted outcome. Unlike other coordinators, kala may also co-occur with the topic auxiliary ma, a feature not shared with any of the other coordinators. (1858) Ngu=rna yangi kala ma=yi lankarr=ya-ni yangi. seq.aux=1sg.s one contr top.aux=1sg.o split=go-pst one ‘I have one, but the other one broke on me.’ (TJL: TT76_2201: 1940014_1943160) (1859) Jangu ma dokta-rlu jakarr jala fiksim=ma-nanta, like top.aux doctor-erg try delim fix.tr=get-prs kala ma waku. contr top.aux neg ‘As in to say, the doctor did indeed try to fix him, but to no avail.’ (PPN: LC09) (1860) Jingka marda-ka=yanu yumpardi kala ma=rnalu child have-imp=3pl.o one_place contr top.aux=1pl.excl.s ya-nku wirlinyi nganimpa. go-pot hunting 1pl.excl ‘Keep all the kids (safe) in this one place! But (as for us), we will go out hunting.’ (MDN: LC22b: 309247_311441) (1861) Ngarrka=ma-ni=rna=ngku, kala kuja=n kuli. recognise=get-pst=1sg.s=2sg.o contr thus=2sg.s angry ‘I watched you, but you were so angry.’ (DMN: LC42b) Contrastive kala clauses do not require a preceding primary clause in all cases. In the following dialogic example (1862), the contrastive kala clause is set in juxtaposition to a proposition made by the speaker’s interlocutor. (1862) YRN:

TJN:

YRN:

YRN:

Waku=rna=yanu jingka-ku kamparri. neg=1sg.s=3pl.o child-dat first ‘I didn’t have any kids back then.’ Kala nyanyi jingka-kurlu kamparri. contr long_time child-prop first ‘But I thought you had kids back then.’ Ah, ah ‘Ah.’ Kujanguny=pa=n karri-nya-ngurra, yalu-piya, manga-piya. such_kind=ep=2sg.s be-pst-narr that-sembl girl-sembl ‘(You’re right) You were about the size of that girl.’ (YRN: LC33a: 1026040_1029511)

688 

 11 Complex syntax

11.2.5 Subsequent coordination =wali The polyfunctional enclitic =wali was introduced in (§10.4.2). In complex coordinated clauses, =wali marks a relation of temporal or logical subsequence as in (1863) and (1864). It thus has a function similar to the ngu ~ nga coordinator. Subsequent =wali is unusual in being the only coordinator that is a clitic rather than an independent word.561 (1863) Yalu-rlu ngantany-ju parda nga-rnngi, yuna=karri-nya=wali kuyi. that-erg man-erg dub eat-irr rotten=be-pst=then meat ‘That man was going to eat, the meat was rotten, however.’ (PSM: TEN12018_037-01: 411474_422887) (1864) Warlu=rna ma-ni, yut=karri-nya=wali yumpaly. fire=1sg.s get-pst sit=be-pst=then tired ‘I got firewood, then I sat down, tired.’ (PSM: TEN1-2017_026-01: 2169541_2174603) The ‘subsequent’ semantics of the enclitic =wali, means it is also frequently found marking apodoses of conditional clauses. Examples were (1807) and (1810).

11.2.6 an ‘and’ The Kriol coordinator en, an ‘and’ is commonly used to link two clauses in a logical or sequential function in a manner similar to ngu ~ nga (§11.2.1).562 The coordinator an does not attract the bound pronouns. Examples are (1865) and (1866). (1865) Jurr-jurr=yirra-rni, an yaparti-nya=yanu. descend-rdp=place-pst and run-pst=3pl.o ‘He put down all his belonging and ran after them (the kangaroos).’ (PSM: TEN1-2018_010-02: 1209046_1223184)

561 The use of clitics to relate two clauses via a temporal relationship (particularly succession) is also found in complex clauses in Wangkajunga (Jones, 2011, 285ff.). 562 Speakers also use en ~ an when resuming lapsed thoughts, or simply after long pauses in narratives. Despite not having a syntactic coordinating function in these contexts, en ~ an arguably serves a discourse-pragmatic function linking a clause with a preceding discourse topic rather than a specific clausal event.

11.2 Finite coordinated clauses 

 689

(1866) Yirra-rnani-nyirra an wajim=ma-nani-nyirra kuja-rla. place-ipfv.pst-narr and watch.tr=get-ipfv.pst-narr thus-loc ‘He put it (the salt) out and then watched it in just such a place.’ (TJN: Manungka_01-035161: 496051_499580) In addition to its role as a clausal coordinator, en ~ an can combine NPs into a complex NP as in (1867). This usage replaces the Ngardi inclusive conjunction manu (§4.2.4.1). Note that manu, unlike Kriol en and an, cannot be used as a clausal coordinator. (1867) Kuja=lu wuruly=ya-ni yala, Japangardi-kujarra an sub=3pl.s quiet=go-pst that ♂subsection-dual and ngantany wirrpa=lu. man many=3pl.s ‘(They didn’t understand white people) when they went quietly then, those two Japangardi and all the men.’ (MMN: TEN1-2019_001-04: 393636_405658)

11.2.7 tumaji ‘because’ The English/Kriol origin particle tumaj(i) ‘because’ (< ‘too much’) is used by some Ngardi speakers to introduce an explanatory clause ‘because of y’ (1868)–(1870).563 Unlike the Kriol coordinator an, tumaj(i) can host bound pronouns as shown by the overt subject enclitic in (1869). (1868) Ngaju ma=rna wirrirl=ya-nku Kururrungku-kurra=wali 1sg top.aux=1sg.s back=go-pot place_name-all=then tumaji Nakarra-rlu parli=pi-nya. because ♀subsection-erg find=hit-pst ‘I will go back to Kururrungku then because Nakarra (my wife) had found me.’ (PSM: TEN1-2018_009-01: 509265_511511) (1869) Jangu timana ngaju-ku wakurra=lu=yanu pilaka=ka-nya like horse 1sg-dat neg=3pl.s=3pl.o shoulder=carry-pst tumaji=lu=nawu pakjampa. because=3pl.s=now buck_jumper ‘Like my horses, they didn’t ride them, because they were wild buck jumpers.’ (PSM: TEN1-2016_005-03: 387079_388890)

563 The English/Kriol tumaj ‘lots, excessive’ is used as a nominal or an adverbial preverb with semantics akin to the English ‘too much’. It typically modifies NPs but is also observed modifying verbal events.

690 

 11 Complex syntax

(1870) Yalu-ngka=rnalu yumpaly luk=karri-nya, two night. . . that-loc=1pl.excl.s tired lie_down=be-pst two night tumaj yumpaly=ma-ni bitumen-tu mariyawu. because tired=get-pst bitumen-erg long_way ‘In that place, tired, we stopped for two nights. . . because that bitumen had made us tired, a long way.’ (PSM: TEN1-2018_009-01: 782235 _792127)

11.2.8 bat ‘but’ The English/Kriol coordinator bat ‘but’ introduces a contrastive assertion, similar to kala (§11.2.4). Examples are (1871) and (1872). In the latter, bat is used in conjunction with =wali to mark a contrastive subsequent clause (cf. §11.2.5). (1871)

a.

Ngurrpa=parda, maitbi manija kardiya-rlu pulijman-tu. ignorant=dub maybe manager white_man-erg policeman-erg ‘Dunno, maybe the policemen (took) the manager away.’ b. Bat parli=pi-nya=rli yalu-ngka=lku kardiya yala. but find=hit-pst=1du.incl.s that-loc=still white_man that ‘But you and I found that whitefella still there.’ (PSM: TEN1-2016_00503: 588120_590802)

(1872)

Ngaju ma=rna kupa-rnngi, bat nyanyi ma 1sg top.aux=1g.s cook-irr but before top.aux kupa-rni ngaringka-rlu=wali. cook-pst woman-erg=then ‘I was going to cook, but the woman had cooked it already.’ (MMN, PSM: TEN1: 2018_033-03: 1106755_1109367)

As observed for the coordinator an, bat is also used in a phrase-linking function as in (1873). (1873)

Yala ngu=rna ya-nku yangi rodhawij, bat hawij-kurlu, that seq.aux=1sg.s go-pot one roadhouse but house-prop pub-kurlu. pub-prop ‘I will go to the one roadhouse, but (the one) with a house and pub.’ (PSM: TEN1-2018_009-01: 849872_853790)

11.3 Non-finite subordinate clauses 

 691

11.3 Non-finite subordinate clauses Non-finite subordinate clauses are characterised by the following properties: i) they contain a predicate that is either a non-finite verb (simple or complex), a preverb or certain nominals ii) they exhibit case marking in subordinating functions iii) they lack bound pronouns (and lack an auxiliary) iv) they are (generally) positioned on the periphery of a main clause. Example (1874) illustrates the basic structure of the non-finite subordinate clause. (1874) Nya-nya=rna=pulany [jantura-kurra la-nu-kurra]. see-pst=1sg.s=3du.o bustard-all pierce-inf-all ‘I saw those two spearing a bush turkey.’ (PRN: TG1_0034-01) In (1874), the subject of the subordinate clause janturakurra lanukurra is interpreted as co-referent with the object of the main clause (i.e. ‘those two’) due to the control properties of the allative case in its subordinating function. The allative case further specifies a tense dependency between the two clauses such that the event time of the subordinate clause is concurrent with that of the main clause. The non-finite subordinate clause lacks its own bound pronouns but permits the expression of a non-subject argument jantura. All constituents of the non-finite subordinate clause are marked by the allative subordinator.564 Non-finite subordinate clause structures of this (and similar) type(s) are found in a large number of Australian languages, although such structures show variability with respect to the degree to which they can be considered ‘syntactically embedded’ (see Nordlinger, 2006).565 In Ngardi, non-finite subordinate clauses are ‘reduced’ in that their predicates carry none of their own TAM marking.566 Ngardi subordinate clauses are ‘dependent’ in that they rely on main clauses for the interpretation of

564 This type of subordinate case marking across all non-finite clause constituents is also observed in Yulparija (Burridge, 1996, p. 26). 565 The nature of the syntactic relationship or linkage between subordinate and main clause has been the topic of much discussion in Australian languages – a 1988 edited volume dealt specifically with this issue on a large number of languages (Austin, 1988). Specific chapters dealing with ‘subordinated’ non-finite clauses are found for Manyjilyjarra (Clendon, 1988), Warlpiri (Simpson, 1988) and Martuthunira (Dench, 1988). 566 While some descriptions of Australian languages treat non-finite (infinitive) verb forms as ‘nominalised’ (Austin, 1998; Blake, 1987, p. 141), Nordlinger (2006) argues that non-finite verbs in Wambaya could be said to have the status of ‘converbs’ (see Haspelmath, 1995). Simpson (1988) considers non-finite verbs in Warlpiri as sitting on the edge of the ‘spectrum’ of the nominal class (Simpson, 1988, p. 207). See also Nordlinger (2002) for a discussion of the status of non-finite verbs and their status as ‘nominalised’ in Australian languages more generally.

692 

 11 Complex syntax

their subjects and they can be considered ‘embedded’ in that they can (in some cases) appear within clauses rather than at their peripheries.567 The most common function of non-finite clauses in Ngardi is to act as what Simpson (1983) terms ‘circumstantials’ – adjunct clauses providing the reason, time, purpose, condition, etc. for the action denoted by the main verb. An illustration of the possibility for the variable ordering of subordinate clauses is provided by Patrick Smith’s reflection on the possible word orderings in (1875). Nevertheless, there is a general tendency for non-finite subordinate clauses to appear at the periphery of main clauses – a property they share with finite subordinate clauses (§11.1). (1875) a. Wirriri=ya-ni=lu, paja-rnu-jangka purnu-jangka. back=go-pst=3pl.s bite-inf-ela wood-ela OR b. Purnu-jangka wirriri=ya-ni=lu paja-rnu-jangka. wood-ela back=go-pst=3pl.s bite-inf-ela ‘They came back after chopping the wood.’ (PSM: TEN1-2018_037-01: 1120922_1124405) The ordering of the subordinate clause in (1875), illustrates that constituents of a subordinate clause can be rendered discontinuous. Interestingly, the fact that constituents of subordinate clause can be discontinuous is further reinforced by the ability for left-adjoined constituents of the subordinate clauses to even host the pronominal arguments of the main clause, as in (1876). (1876)

Mangarri-rla=rna ngaringka nya-nganta kupa-rnu-rla. veg_food-loc=1sg.s woman see-prs cook-inf-loc ‘I am watching that woman while she is cooking bread.’ (PJA: TT76_2601: 979727_985012)

In the following sections, I first discuss the different types of predicates found in non-finite clauses (§11.3.1). Subcategories of subordinating case are given in §11.3.2 which divide the set of subordinating case suffixes into temporal subordinators (T-subordinators) and concord subordinators (C-subordinators) following Dench and Evans (1998). Individual case suffixes in subordinating functions will be described in sections on prior subordinators (§11.3.3), simultaneous subordinators (§11.3.4), subsequent subordinators (§11.3.5), and concord subordinators (§11.3.6). The chapter concludes with a discussion of observed combinations of temporal and concord subordinators (§11.3.7) as well as insubordinate usages of non-finite clauses (§11.3.8).

567 In the Western Desert language Wangkajunga, certain temporal subordinate clauses can also be found embedded in main clauses (Jones, 2011, p. 284).

11.3 Non-finite subordinate clauses 

 693

11.3.1 Types of predicates in non-finite subordinate clause The predicate in a non-finite subordinate clause can be an infinitive verb, as illustrated previously in (1875), a preverb as in (1877), or a complex verb as in (1878). (1877) Ngana pawu-karra ya-nanta? who shout-ssub go-prs ‘Who is going along shouting?’ (PRN: TG1_0034-01) (1878) Purda=nya-ngu-jangka ngu=rna nya-ngani-nyirra ngaju=rlangu. listen=see-inf-ela seq.aux=1sg.s see-ipfv.pst-narr 1sg=also ‘Having matured (lit. ‘having understood/learned’), I too saw it (the Dreaming ceremony).’ (PSM: TEN1-2016_002-01: 644686_648488) Additionally, it is also possible for a nominal to constitute the sole predicating element of a type of non-finite subordinate clause. This was initially raised as a possibility in the description of ‘secondary predicates’ (§9.3). An example of an ‘action nominal’ functioning as a predicate in a subordinate clause (of which it is the sole constituent) is provided in (1879). (1879) Karli=rna paja-rni, wirlinyi-rlu, ya-ni nawu. boomerang=1sg.s bite-pst hunting-erg go-pst now ‘I cut out a boomerang while out on a hunting trip, then I went.’ (TJL: TT76_2201: 1786234_1790780) The possibility of case appearing on nominals in a putative ‘subordinating function’ is what Hale (1982) refers to as ‘autonomous case’ or a ‘vague predicational’ use of subordinated nominals, specifically stating ‘ . . . the interpretation of an autonomous complementised nominal involves an abstract, semantically unspecified verb in logical form’ (1982, p. 286). The peculiar status of these nominals as somehow subordinated is particularly clear when the nominal does not depict an event (e.g. ngapa ‘water’ instead of wirlinyi ‘hunting’). In these cases, a verbal meaning has to be inferred from context, as in (1880). (1880) Warlu=rna=rla yirra-ku ngapa-jangka-rlu wali lurrij. fire=1sg.s=3sg.o place-pot water-ela-erg indeed finish ‘I will make the fire for it after drinking some water, then (I’m) done.’ (JLI: TT76_2201: 1569223_1573710) The fact that the elative in (1880) is used in subordinating function (specifically prior time) is clearly apparent from the fact that a semantic reading of the argument as a

694 

 11 Complex syntax

main clause argument without an implied subordinated verb is non-logical: ‘I will make a fire from water’ or ‘I, from water, will make a fire’. The fact that the intended verbal event was ‘drinking water’ was reinforced by the preceding discourse in which Jack Lightning specified the drinking of water with the eat verb nga-. Constructions such as (1880) are frequent in Ngardi speech and appear in the full range of subordinate clause types, including allative (1881) and dative (1882) purpose clauses. (1881) Ya-nta-rni mu-kurra kuli-kurra. go-imp-hith this-all anger-all ‘Come here to fight!’ (PSM: TEN1-2018_021-01: 618438_622687) (1882) Yirdi-ku=rna=rla wangka-nya. word-dat=1sg.s=3sg.obl speak-pst ‘I said to her to let (you) know.’ (TJN: Cataldi, 2011, p. 355) The possibility for structures (1880)–(1882) raises the question as to whether all non-finite clauses in Ngardi are truly subordinate clauses or whether they can be considered full nominalisations occupying adjunct positions in main clauses. Blake argues that where nominalisation is ‘complete’ (as Simpson argues is the case in Warlpiri for instance), there is not a true subordinate clause but instead ‘the functional equivalent of a subordinate clause’ (Blake, 1977, p. 110).568 Furthermore, the ability of case in ‘subordinating functions’ to appear on nominals also raises the question of what the strict distinction is between case in main clause functions and case in subordinating functions. In some cases, there is clear evidence in support of a strict division. The dative in a main clause function is registered in the bound pronoun (subject to animacy and argument co-occurrence constraints), whereas datives in non-finite subordinate clauses are never cross-referenced (cf. the widespread dative versus purposive distinction in Australian languages; see Dixon, 1976; Hale, 1982). However, this criterion cannot be consistently applied to any other case suffix. Moreover, there are multiple situations in which a case-marked nominal is not registered in the bound pronoun but this does not mean it is subordinating in function. A single example is the presence of the ergative marking ‘instrumental’ secondary predicates. Instead, it can only be said that subordinators tend to appear on non-finite verbs or preverbs and case markers only tend to be restricted to nominals. Hale (1982, p. 256) thus reflects: ‘The terminological distinction which I have drawn between cases and

568 Blake (1977, p. 110) also argues that more ‘nominalised’ arguments are more likely to be treated as core arguments rather than adjuncts: Form:

Finite > Non-finite > Partial Nominalisation > Full Nominalisation

Function:

Outer Adjunct > Inner Adjunct > Complement > Core Function.

11.3 Non-finite subordinate clauses 

 695

complementisers does not correspond to a sharp and precise classification in Warlpiri syntax or morphology.’569 However, in Ngardi – as in Warlpiri – it is not possible to account for the range of argument control and tense sequencing properties discussed in the following subsections if case suffixes are not permitted to introduce (at least the functional equivalent) of a reduced clause with its own non-finite predicate. That is, the functions of case go well beyond marking simple argument relations of a main clause predicate to its argument or semantic/local relations of adjuncts. In this description, I follow recent grammatical descriptions such as Meakins and Nordlinger (2014) and Senge (2015) that adhere to the more general Australianist principle of avoiding positing excessive homophony and allowing single case suffixes to behave polyfunctionally (see Dench & Evans, 1988). Throughout this section, cases in subordinating function are glossed identically to case-markers appearing in main clauses. My use of the term ‘subordinator’ or ‘subordinating case’ is used simply to refer to ‘cases in subordinating functions’.

11.3.2 Types of subordinating case Case suffixes serve to relate a non-finite subordinate clause to the main clause by providing information about i) control relations, ii) relative (or dependent) tense, or iii) various logical/semantic inter-clausal relations. Cases in this function are found marking the predicate of the subordinate clause but may also appear on subordinated arguments. Nearly all of the suffixes with subordinating roles have primary roles as grammatical or spatial or derivational case suffixes in main clauses (e.g. locative -rla, allative -kurra, elative -jangka, associative warnu, dative -ku, evitative -ngkamarra, privative -wangu). Ngardi also possesses a small number of cases that are exclusively subordinating (simultaneous -karra, preparative -kungarnti, and obviative -rlakarni).570 Cases in subordinating function are typically analysed in terms of two primary grammatical functions: i) their control relation, and ii) their relative tense value. ‘Control’ refers to the ability for a case suffix to index the subject of a non-finite subordinate clause as either co-referential with the main clause subject or with some other argument. ‘Relative tense’ – also termed ‘dependent tense’ (Simpson, 1983) – refers to the tem-

569 Simpson (1988, p. 266) makes a near-identical point: ‘Rather than being different types of element, they [case suffixes] consist of clusters of properties, whose nature, or presence or absence, determines whether or not a particular instance of a suffix acts more like a case-marker or more like a complementiser.’ 570 This can be contrasted with the Ngumpin language Bilinarra, for example, which has been analysed as a traditional ‘switch-reference’ system that is based on a binary distinction (‘same subject’ [locative] vs ‘different object’ [allative]) (Meakins & Nordlinger, 2014, p. 426). Even in Bilinarra, however, three additional cases are used in subordinating functions.

696 

 11 Complex syntax

poral reference of a subordinate clause and is evaluated with respect to the main clause. Typically, a subordinating case suffix can mark a clause as occurring before (prior tense), concurrently with (simultaneous tense), or after (subsequent tense) the event depicted in the main clause. It will be shown in the following subsections that certain subordinators primarily encode semantic relationships between subordinate and main clauses (such as cause, purpose and reason) and only encode temporal relations as a secondary function. Again, this is consistent with the general obscuring of functional and formal distinctions between main clause usages of case and subordinating usages of case in Ngardi. A classification of subordinators according to their control and tense sequencing features is provided in Table 144. The formalisation ‘+’ indicates that a case form is specified for this feature. Some case suffixes lack a specification for certain features and these are marked ‘ – ’. Table 144: Cases with subordinating functions in Ngardi. Case suffix

T-subordinating feature Prior

sequential elative associative ssub allative privative dative preparative evitative

-rla -rla -ngkarla -jangka -warnu -karra -kurra -wangu -ku -kungarnti -ngkamarra

ergative obviative572

-rlu -rla(ka)rni

locative

+ + + +

–

Simult.

Subsequent

Subject + – + – – +

+ +?

+ + –

C-subordinating feature

– + + +

– – – – +574

Object/Other – – – + – – – – +

Ngardi’s system of marking subordination with a wide range of case suffixes can be distinguished from languages with classical ‘switch-reference’ systems (Jacobsen, 1967), where there is a binary distinction between a ‘same subject’ marker and a ‘different subject’ marker. Austin’s (1981b, p. 331) landmark survey of switchreference in Australian languages noted the following prevalent patterns: either the locative as ‘same subject marker’ and the allative as a ‘different subject marker’ or the locative marker as different subject. While Ngardi’s system of subordination

571 Ergative case-marking of non-finite subordinate clauses in agreement with the main-clause subject is discussed in detail by Simpson for Warlpiri (Simpson, 1983, 434ff.). 572 The obviative is defined by the fact that it is not controlled by any main-clause arguments.

11.3 Non-finite subordinate clauses 

 697

happens to include these two widespread properties, the features of non-finite subordination extend well beyond these two case markers and their control functions. Included in Jacobsen’s (1967) original definition of switch reference is the typological observation that the morphological marker of switch reference ‘may or may not include other meanings in addition’ (Jacobsen, 1967, p. 240). This expanded definition i particularly pertinent to most of the Ngumpin-Yapa languages. Nearly all of these languages share a substantial number of subordinating suffixes and these suffixes vary with respect to their overt specification of their control feature, their tense value as well as any accompanying semantic specification.573 In Ngardi, subordinating suffixes may be unspecified for relative tense but specified for control (C-subordinators) and similarly, unspecified for control but specified for relative tense (T-subordinators). Acknowledgement of such possibilities is particularly important for an analysis of the subordinating case -rla (the locative) in Ngardi. In the Ngumpin languages – Jaru, Wanyjirra, Gurindji and Bilinarra – the same form is analysed fairly straightforwardly as a ‘simultaneous’ T-subordinator.574 In Ngardi however, the locative subordinator is found marking both ‘simultaneous’ (1883) and ‘prior’ (1884) relative tense: [i] Simultaneous tense, subject control (1883) Ngarlarri-nyu-rla ya-nanta. laugh-inf-loc go-prs ‘He is going along laughing.’ (PSM: TEN1-2017_026-01: 1071389_1075319) [ii] Prior tense, subject control (1884) Paja-rnu-rla, wa, ka-ngku=rra kita-kurra paja-ku bite-inf-loc ? carry-pot=again outside-all bite-pot kurrkurr=ma-nku. trim=get-pot ‘Having cut it, she will take it again into an open place and she will refine it and trim it.’ (MDN: Cataldi, 2011, p. 146) A further complication is that when encoding simultaneous tense, the -rla subordinator may be controlled by a main clause subject, as in (1883), or a main clause object, as

573 For example, in Jaru, both the locative and allative suffixes can be used for simultaneous relative tense and control by an absolutive main clause argument – however, the allative appears more commonly when the referent of the subordinate clause is moving rather than stationary. 574 A major difference between Ngumpin languages, though, is that in Bilinarra and Gurindji -rla is ‘subject-controlled’ whereas in Wanyjirra and Jaru -rla has no specified control feature.

698 

 11 Complex syntax

in (1876). However, when encoding prior tense, it tends to be only subject-controlled (1884). The polyfunctionality of -rla as a subordinator also means it exhibits functional overlap with the allative subordinator -kurra (§11.3.4.1) (used for simultaneous object control) on the one hand, and -karra (§11.3.4.2) (used for simultaneous same subject in intransitive sentences) on the other. Thus consider the equivalent pairs presented in (1885) and (1886). (1885) a. Ngaringka-rlu nya-nganta pukany-kurra. woman-erg see-prs sleep-all OR b. Ngaringka-rlu nya-nganta pukany-ja. woman-erg see-prs sleep-loc ‘A woman sees (a man) sleeping.’ (PSM: TEN1-2018_029-01: 1176121_1181787) (1886) a.

Yunpa-rnu-rla ya-nanta. sing-inf-loc go-prs

OR b. Yunpa-rnu-karra ya-nanta. sing-inf-ssub go-prs ‘He is going along singing.’ (MMN: TEN1-2017_026-01: 1071389_1075319) Arguably, some of these facts may reflect a subordinating case system in flux – either through variation in the various linguistic repertoires of different speakers or perhaps even diachronic changes in progress. On the basis of limited data for individual speakers it is difficult to speculate further. Briefly, it is worth noting that the subordinator -rla exhibits some of the same polyfunctionality in Warlpiri. In the Warlpiri literature, there has been a tendency to analyse two homophonous ‘complementising case suffixes’: a -rla locative and a -rla sequential (Simpson, 1983). The locative, when attached to nominalised verbs, signals simultaneous tense and may either be subject-controlled or non-subject controlled, whereas sequential -rla signals that the event of the subordinate clause occurred prior to the event of the main clause and is strictly subject-controlled.575 Ngardi differs from Warlpiri in that it also possesses a subordinator -ngkarla (§11.3.3.2) that has identical control and tense features with what is described as the ‘sequential’ -rla subordinator in Warlpiri. This subordinator is generally only used 575 With respect to this analysis, Hale (1982, p. 257–275) has also suggested that the relationship between spatial cases and their relative tense functions in Warlpiri has to do with a single concept of ‘central coincidence’ and ‘terminal coincidence’. The locative case is – semantically speaking – a ‘central coincidence’ suffix and is thus expected to have a ‘simultaneous’ relative tense function. This is distinct from -rla sequential which does not behave semantically like its homophonous locative counterpart.

11.3 Non-finite subordinate clauses 

 699

with prior time reference and exhibits subject control. Morphologically it comprises the locative -rla and an initial element -ngka (which also happens to be an allomorph of the locative). Due to these similarities, the gloss seq_loc was chosen. The following sections deal first with T-subordinators and are subdivided according to relative tense values: prior (§11.3.3), simultaneous (§11.3.4), and subsequent (§11.3.5). Strictly speaking, the privative case is unspecified for relative tense but when occurring by itself it has a default interpretation of simultaneous tense and is described therein. C-subordinators are described in §11.3.6. As noted above -rla is the only suffix that functions both as a T-subordinator and a C-subordinator. Possibilities for combining T-subordinator and C-subordinators are described in §11.3.7. A limitation of the following sections is the lack of examples evincing control properties of subordinators with a full range of clause types. This topic has occupied significant research interest in Warlpiri and it is likely that similar complexities may be revealed with further research in Ngardi.

11.3.3  Prior subordinators Subordinators with prior time reference in Ngardi are {-ngkarla, -rla, -jangka and -warnu} which show formal identity with locative, elative and associative case suffixes. One unique aspect of the Ngardi subordinating system is the form -ngkarla, termed seq_loc here, which is not attested in related Ngumpin or Yapa languages. Within the Ngardi system, it appears to exist alongside sequential -rla as a means of marking prior tense.576 11.3.3.1 Locative -rla The locative has two subordinating functions in Ngardi. It can be used as a ‘simultaneous tense’ subordinator (§11.3.4.1) or as a variant of sequential locative -ngkarla (§11.3.3.2) as a subordinator encoding prior tense.577 The interchangeability between -rla and -ngkarla is demonstrated in (1887) where two subordinate clauses are both interpreted with prior tense but one takes the locative and the other takes the sequential locative.

576 Note that -ngkarla is equivalent to Warlpiri -rla (sequential) which attaches directly to the Warlpiri infinitive -nja (Cataldi, 2011, p. 7). 577 This prior tense usage of -rla is also found for all conjugation classes in Eastern Walmajarri. Other varieties of Walmajarri utilise -ngurla in most conjugation classes as the prior-tense subordinator (cf. Hudson & Richards, 1990, p. 387).

700 

 11 Complex syntax

(1887) Kuyi=lu kupa-rnu-rla ma-nu-ngkarla tarn=yirra-rni.578 meat=3pl.s cook-inf-loc get-inf-seq_loc hang=place-pst ‘Having got the meat and having cooked it, they hung it up.’ (NMN: LC21b: 651696_654634) The locative as a prior-tense subordinator is found controlled by main clause subjects (transitive (1888) or intransitive (1889)). For those complex clauses where both finite and non-finite predicates are transitive, co-referentiality of objects is also observed. It is unclear if this is a strict constraint on the locative in its ‘prior’ function, or a limitation of the available data. (1888) Kuyu ngu=lu=nyanu luwa-rnu-rla yu-ngani-nyirra. meat seq.aux=3pl.s=refl shoot-inf-loc give-ipfv.pst-narr ‘They used to kill the meat and share it amongst each other. (lit. ‘having killed the meat, they would share it amongst themselves’).’ (MAM: TEN1-2016_00201: 323367_325463) (1889) Ngaka yalu karri-nyu-rla wangka-nya. later that be-inf-loc speak-pst ‘Later that one, after sitting a while, spoke.’ (MDN: Cataldi, 2011, p. 343) The subject control properties of the prior locative can be compared with the locative functioning as a simultaneous tense subordinator (§11.3.4.1) where it may be controlled by subjects and non-subjects alike. 11.3.3.2 Sequential locative -ngkarla The ‘sequential locative’ -ngkarla is used with prior time reference and marks a subject-controlled subordinate clause as illustrated in (1890). (1890) Marda-rni=pula, pu-ngu-ngkarla=pula=rla yi-nya. have-pst=3du.s hit-inf-seq_loc=3du.s=3sg.obl give-pst ‘The two looked after her and – (those two) having killed (the meat) – the two gave it to her.’ (TJN: LC21a: 323566_327165)

578 On initial inspection, this example provides suggestive evidence that -ngkarla is an allomorph of locative -rla, used with disyllabic stems, somewhat similar to the conditioning of the locative in its main-clause use. However, there are examples of locative -rla attaching directly to disyllabic stems (e.g. ma-nu-rla) and examples of sequential locative -ngkarla attaching to trisyllabic and longer stems (e.g. wangka-nyu-ngkarla). Therefore, even if -ngkarla has a diachronic origin as a regular allomorph, it is no longer found in a predictable distribution.

11.3 Non-finite subordinate clauses 

 701

The sequential locative can be argued to comprise the locative -rla and an initial element -ngka.579 Potentially this may actually represent a series of two locative suffixes, the inner suffix also being a locative allomorph. There is some limited evidence for this hypothesis. First, -ngkarla is primarily attested with inflected monosyllabic verb roots, e.g. ka- carry as in (1891). Infinitive verb stems formed from monosyllabic roots are invariably disyllabic and this is the precise environment in which the -ngka allomorph of the locative is to be expected. Second, there may be a functional explanation for why double locative marking might have developed. The locative suffix -rla in Ngardi marks both prior and simultaneous events and thus -rla has lost its ability to unambiguously signal relative tense. The addition of a form -ngkarla may represent a change in progress of a system restructuring to distinguish a prior tense subordinator (-ngkarla) from a generic subordinator (-rla). (1891)

Ka-ngu-ngkarla kujarra pulaa, juka tiyilip=pa=pula carry-inf-seq.loc two flour sugar tea_leaf=ep=3du.s ma-ninja-nani-nyirra, ngunyju.580 get-move&x-ipfv.pst-narr tobacco ‘The two of them, having brought flour, sugar and tea, went and got tobacco.’ (YMN: LC23a: 224261_231524)

Problematically, however, there is one example (1892), where the sequential locative was provided as a simultaneous tense marker – displaying identical polyfunctionality to the simple locative. I am unable to provide an account for this. . (1892)

Nga-rnu-ngkarla ya-nanta motika-rla581 eat-inf-seq.loc go-prs car-loc ‘He is going, eating in the vehicle.’ (PSM: TEN1-2018_033-03: 1970388_1972774)

11.3.3.3 Elative -jangka The elative subordinator -jangka is functionally similar to the (sequential) locative subordinator in encoding prior tense. The elative subordinator appears in subordinated clauses exhibiting object control (1893) and subject (1894) control. When con-

579 There is at least one example of yaninkurla in the corpus (TJN: LC05). This form appears to have a similar function to yanungkarla (go-inf-seq.loc) but perhaps shows some Warlpiri influence, since it uses a verb stem ya-ni that is tensed in Ngardi (go-pst) but non-finite in neighbouring Warlpiri. 580 In this example kujarra is missing expected ergative (-rlu) marking as a transitive subject. 581 In fact, on this occasion, the plain locative subordinator was not accepted, i.e. *nga-rnu-rla.

702 

 11 Complex syntax

trolled by transitive subjects the elative T-subordinator is followed by the ergative C-subordinator -rlu582: (1893) Kunyarr-u=nyanu ngalya=ma-nanta yala paja-rnu-jangka taka. dog-erg=refl lick=get-prs that bite-inf-ela hand ‘The dog licked his paw, his paw having been bitten.’ (MNN: TEN1-2017_01201: 148367_154564) (1894) Wakurra=rna=rla purda=nya-nganta, nyanyikarnu=warru neg=1sg.s=3sg.obl listen=see-prs before=emph nya-ngu-jangka-rlu. see-inf-ela-erg ‘I don’t remember it (that place), it’s a long time since I have seen it.’ (MRN: LC35b: 1346220_1351139) The elative subordinator also differs from the (sequential) locative subordinator semantically, in that – as per its relational usage in main clauses – it typically encodes the reason, motivation or causal mechanism of the main clause event. This can be observed in subordinate clauses involving non-finite verbs (1895), (1896), preverbs (1897), Kriol verbal loans (1898), and complex verbs (1899). (1895) Pu-ngu-jangka lu-ngani wali. hit-inf-ela cry-ipfv.pst indeed ‘Because of the assault, he was crying.’ (PSM: TEN1-2017_009-01: 837976_840042) (1896) Yarli-rnu-jangka=wu walyka=karri-ju. wash-inf-ela=voc cool=be-pot ‘I will cool down by washing.’ (PSM: TEN1-2018_037-01: 1203669_1207500) (1897) Minya ma ngayirrkuta karri-nya yapart-jangka. this top.aux short_breath be-pst run-ela ‘This one, he was short of breath from running.’ (PSM: TEN1-2018_021-02: 805222_814231)

582 There is an unexplained irregularity involving this subordinating suffix when occurring with the stem karri-nyu be-inf in that the expected final -u is often not realised and the subordinated form is rendered as karrinyjangka.

11.3 Non-finite subordinate clauses 

 703

(1898) Tayimap-jangka rurriny=ma-nta! tie.tr.compl-ela untie=get-imp ‘Untie him from his fastened/bound state!’ (PSM: TEN1-2018_035-02: 1143511_1146599) (1899) Ngaka=lu jangu nguku-kurra wali ya-nani, later=3pl.s like water-all indeed go-ipfv.pst kaji=lu=nganpa yitaki=ma-ni jina jingka condit=3pl.s=1pl.excl.o track=get-pst foot child ruju=karri-nyu-jangka Ngariyirli-rla. play=be-inf-ela place_name-loc ‘Later, they went for water, and so they followed our tracks from where we children had been playing at Ngariyirli.’ (PPN: Bye bye 10-12_2_91) An example of an elative-marked subordinate clause with multiple subordinate-clause constituents is provided in (1900). (1900) Jarrapany ma=rna lirrij=pa-rnanta, karrka, jangu shoulder top.aux=1sg.s sore=get-prs sore recog2 parrparrpa pangi-rnu-jangka yarla-jangka walya-jangka. pain dig-inf-ela yam-ela ground-ela ‘I have a sore shoulder, painful yeah, hurting, from digging yams out of the ground.’ (TJN: Cataldi, 2011, p. 258). 11.3.3.4 Associative2 -warnu The associative2 -warnu is found in just a few contexts as a subordinator – seemingly reflecting its rarity in main clauses. The suffix -warnu is distinct among the subordinators in that it is a derivational suffix (§4.3.3) rather than a relational or derivational case in finite clauses.583 Like the elative subordinator (§11.3.3.3), it encodes prior relative tense and is found co-occurring with the ergative C-subordinator -rlu marking subject control. (1901) a.

Wurra=lku luk=karri-nya. later=then lie_down=be-pst ‘Later she lay down.’

583 Warlpiri, in contrast, has a cognate case suffix -warnu which has a straightforward polyfunctionality as a relational case-marker in main clauses (termed ‘elative’ (Nash, 1986)) and a subordinating function similar to that found in Ngardi. In Warlpiri, -jangka-marked nominals are found used in concert with -warnu-marked non-finite verbs as a means of expressing prior events (Hale, 1982, p. 274).

704 

 11 Complex syntax

b. Luk-warnu-rlu nga-rni mangarri. lie-assoc2-erg eat-pst veg_food ‘Having lain down, she ate some food.’ (MDN: Cataldi, 2011, p. 161)

11.3.4 Simultaneous subordinators The simultaneous subordinators are the locative (§11.3.4.1), the simultaneous (§11.3.4.2), the allative (§11.3.4.3) and the privative (§11.3.4.4). Strictly speaking, the privative is unspecified for relative tense but in the absence of further case marking, it has a default interpretation of simultaneous relative tense. 11.3.4.1 Locative -rla In addition to its role as a prior tense, subject-controlled subordinator (§11.3.3.1), the locative has a discrete function as a simultaneous tense subordinator. In its simultaneous tense function, the locative has no overt control specification and can be construed with subjects (1902), (1903) and non-subjects (1904) alike.584 [i] Subject control, Simultaneous tense (1902) Wangka-nyu-rla ya-nanta. speak-inf-loc go-prs ‘He is going along talking.’ (MMN: TEN1-2017_026-01: 719025_721097) (1903) Luwa-rnu-rla karli yut=karri-nyanta ngantany. shoot-inf-loc boomerang sit=be-prs man ‘A man is sitting down, refining/smoothing a boomerang.’ [ii] Non-subject control, Simultaneous tense (1904) Purda=nya-nganta=rna wangka-nyu-rla. listen=see-prs=1sg.s speak-inf-loc ‘I am listening to what is being said.’ (MDN: LC42b: 1202063_1207051) Interestingly, non-subject control uses of the simultaneous locative subordinator were not always accepted by Patrick Smith and Marie Mudgedell, for whom the allative (§11.3.4.3) was the preferred simultaneous tense subordinator. Thus consider (1905), in which I prompted with a Jaru simultaneous locative subordinator.

584 See (320) for an example of -rla exhibiting object control involving a double object main-clause predicate.

11.3 Non-finite subordinate clauses 

 705

(1905) TEN: Ngumbirr-u mawun nya-ngan ngaba-ngga nga-rnu-ngga. woman-erg man see-prs water-loc eat-inf-loc ‘A woman sees a man drinking water.’ [Jaru prompt] PSM: Ngaringka-rlu nya-nganta ngantany—ngapa-kurra nga-rnu-kurra. woman-erg see-prs man water-all eat-inf-all ‘A woman sees a man drinking water.’ [Ngardi] (PSM: TEN1-2018_02701: 1188648_1197703) This variation seems to reflect that the simultaneous subordinators in Ngardi are caught between two systems. Arguably there is one system in which the locative can signal simultaneous tense and different subject (i.e. as in Jaru) and another system in which the allative has this function (i.e. as in Warlpiri). Across speakers and the corpus as a whole, both appear to be used variably. Unfortunately, there is insufficient data for every speaker to determine whether individuals consistently follow one pattern or another. In its simultaneous function, the subordinator -rla appears structurally and semantically very similar to simple locative marking of temporals – effectively encoding the time at which the main clause event occurred, as in (1906). (1906) Kanjurrpa, ngintaka jangu=rlipa pangi-rnani-nyirra reptile_sp reptile_sp recog2=1pl.incl.s dig-ipfv.pst-narr wirlinyi-rla kayirra jangu. hunting-loc north assump ‘The kanjurrpa or ngintaka is the sand monitor, which we used to dig up when hunting in the north, you know.’ (YDN: Cataldi, 2011, p. 105) In other contexts, the locative subordinator appears semantically very similar to simple locative marking of locationals, encoding a relative position in which something occurred, as in (1907). (1907) Motika-rla=lu karri-nyu-rla ya-nanyarra ngapa-kariny car-loc=3pl.s be-inf-loc go-prst.hence water-other ngapa-kariny-kurra. water-other-all ‘Sitting in the car, they are going from waterhole to waterhole.’ (PRN: TG0034-01) Hale (1982, pp. 257–275) argues that the uniting semantic component of the locational and temporal functions of the locative, as well as the subordinating- and main clause functions of the locative, is a notion of ‘central coincidence’.

706 

 11 Complex syntax

11.3.4.2 Simultaneous subordinator -karra The simultaneous subordinator (ssub) -karra is homophonous with the resultative suffix -karra ‘until x’ as described in §9.3.2. The simultaneous subordinator -karra marks a subordinate clause as temporally simultaneous with the main clause and exhibiting subject (transitive or intransitive) control. A subordinator of this form and function is found in at least Walmajarri (Hudson, 1978, p. 36), Warlpiri (Simpson, 1988, p. 207) and Warlmanpa (Nash, 1979, p. 3).585 The simultaneous subordinator can be controlled by both intransitive (1908)–(1910) and transitive (1911)–(1913) main clause subjects. In the case of the latter, the subordinate clause is obligatorily marked both with -karra and the ergative C-subordinator -rlu (1908) And ya-ni=pula wurna nawu, lu-ngu-karra. and go-pst=3du.s journey now cry-inf-ssub ‘And the two went then, crying along the way.’ (TJN: LC03) (1909) Nyampa-jangka=n ikirt-karra ya-nanta? what-ela=2sg.s sway-ssub go-prs ‘Why are you getting around limping?’ (PSM: TEN1-2017_008-1: 3855018_ 3857033) (1910)

Ngawiji kilaki=pula yut-parra karri-nyani. FM FF=3du.s sit-ssub be-ipfv.pst ‘My grandfather and grandmother were sitting there.’ (MNN: LC24b: 643582_654148)

(1911)

Nya-ngani-nyirra=rna kuja-rlu ya-nu-karra-rlu. see-ipfv.pst-narr=1sg.s thus-erg go-inf-ssub-erg ‘So I saw it while I was travelling.’ (MDN: LC22b: 930878_934415)

585 A suffix of a similar form is actually reported in a number of other Ngumpin languages but the specifics of its function – and its subsequent analysis – differ substantially. In Jaru, for instance, Tsunoda suggests that -karra has an ‘aspectual meaning’ and appears affixed to adverbs, preverbs, verbids and demonstratives. In its attachment to verbids and preverbs it behaves very much like a subordinating suffix but in other contexts it appears to have some type of stem-forming suffix as it may take further derivational morphology such as -kaji/-waji (Tsunoda, 1981b, p. 242). Meakins and Nordlinger (2014) similarly treat -karra in Bilinarra as an aspectual marker – specifically a ‘continuative’ – and do not treat it as part of the subordinating system. In contrast to -karra in Ngardi, the continuative in Bilinarra is not found attached to nominalised verbs. Nevertheless, aside from this distributional difference, there does not appear to be any further basis to distinguish it from a same subject, simultaneous subordinator (Meakins & Nordlinger, 2014, p. 342).

11.3 Non-finite subordinate clauses 

 707

(1912)

Yapart-karra-rlu jirinti ngardak=ma-nanta. run-ssub-erg path manufacture=get-prs ‘Running back and forth, it (a goanna) creates a path.’ (TJN: Cataldi, 2011, p. 78)

(1913)

Inkil-karra-rlu=yi yi-nya. unwilling-ssub-erg=1sg.o give-pst ‘He gave it to me unwillingly.’ (MMN: TEN1-2018_019-01: 131785_132975)

Note that there are a number of instances of a morpheme -karra that appear in lexicalised combinations with preverbs. Examples include nyinkarra=wanti ‘to swim’, ngurnngurtkarra=ya-nanta ‘it is thundering’, and possibly the Kriol borrowing jartkarra-kiji- ‘to start (a fire)’. A similar ‘fusion’ of the ‘continuative suffix’ -garra has been observed with coverbs in Bilinarra (Meakins & Nordlinger, 2014, p. 186). 11.3.4.3 Allative -kurra The allative suffix -kurra marks simultaneous tense in non-finite subordinate clauses and exhibits object control (1914)–(1916).586 The example in (1916) demonstrates a single action nominal jaru ‘talk’ functioning as the non-finite predicate of the subordinate clause. (1914)

Wakurra=yi=n purda=nya-nganta ngati wangka-nyu-kurra. neg=1sg.o=2sg.s listen=see-prs M speak-inf-all ‘You never listen to my mother while she is talking.’ (MDN: LC42b: 189144_193740)

(1915)

Yalu-jangka, ya-nku=lu ngaringka-rlu=lu parli=pu-ngku luwanja, that-ela go-pot=3pl.s woman-erg=3pl.s find=hit-pot coolamon jarranti-nyu-kurra. stand-inf-all ‘Then the women will go and find trees standing with wood suitable for coolamons (lit. ‘find coolamons while they are standing’).’ (MDN: LC22: 1352027_1357664)

(1916)

Ngali=lu=ngali nya-nganta wirrpa-ngku mu-rlu jaru-kurra. 1du.incl=3pl.s=1du.incl.o see-prs many-erg this-erg talk-all ‘That big group there is watching us two talking.’ (PJA: TT76_2601: 1528163_1533521)

586 In Warlpiri, the related -kurra subordinator is found controlled by both dative and absolutive object arguments – but not subjects (Simpson, 1983, p. 168).

708 

 11 Complex syntax

Due to the regular ellipsis of full NPs and the fact that third singular objects lack overt registration in the bound pronoun, it is entirely possible for an allative subordinate clause to be the only overt constituent of a complex clause referring to the object relation. Thus, compare the overt object in (1914) with the unexpressed object in (1917). (1917) Ngaju=rna purda=nya-nganta iyindij=wangka-nyu-kurra. 1sg=1sg.s listen=see-prs complain=speak-inf-all ‘I hear someone complaining.’ (PSM: TEN-2018_029-01: 1760813_1766247) Lastly, (1918) illustrates the use of the allative in subordinating function attaching to multiple constituents of a non-finite clause. (1918)

Turrku=ma-ni=rna ngirnti-ngulu parntarrngarna nga-rnu-kurra grab=get-pst=1sg.s tail-abl kangaroo eat-inf-all waparl-kurra wurrkal-kurra jungurru-kurra. turn-all grass-all depart-all ‘I grabbed the kangaroo by the tail while it was eating grass with its back turned, not looking.’ (PRN: TG0034-01)

11.3.4.4 Privative -wangu The privative case -wangu has an interesting status as a subordinator since it lacks both any relative tense feature and any argument control features and instead simply encodes a semantic relation of ‘without x–ing’.587 When occurring as the sole subordinating case, it has a default interpretation of simultaneous tense and subject control (1919). However, the fact that the privative can be followed by the locative-marking prior relative tense (1920) and the ergative-marking subject control (1921) indicate that the privative is underspecified as a subordinator for both relative tense and control. (1919)

Waku ma=rnalu ngarnturl=pu-ngu-wangu wangka-nyanta. neg top.aux=1pl.excl.s clear=hit-inf-priv speak-prs ‘No, we are speaking imperfectly.’ (MDN: Cataldi, 2011, p. 220)

(1920) Ya-nu-wangu-rla=rna yirra-rni kamparri nyunpula-wangu-rla go-inf-priv-loc=1sg.s place-pst first 2du-priv-loc nyanyi. before ‘Without you two having arrived, I had already put it down sometime before.’ (Cataldi, 2011, p. 312) 587 In Warlpiri a similar function of a cognate suffix -wangu is described as a ‘negative complementiser’ (Nash, 1986, p. 32).

11.3 Non-finite subordinate clauses 

(1921)

 709

Iji-wangu-rlu nga-lku, ngaju-kariny-ju=lku. gift-priv-erg eat-pot 1sg-other-erg=then ‘I will eat it without giving it away, (I will eat) all by myself.’ (TJL: TT76_2201: 1961893_1964790)

Privative-marked non-finite predicates also appear in a special type of negative command termed ‘prohibitive’, as described in §10.7.1.2.

11.3.5 Subsequent subordinators The subsequent subordinators are the dative -ku (§11.3.5.1), preparative -kungarnti (§11.3.5.2), and evitative -ngkamarra (§11.3.5.3). Strictly speaking, all of the subsequent subordinators are unspecified for their control properties but are nevertheless typically construed with co-referent main clause subjects (transitive and intransitive). 11.3.5.1 Dative -ku The dative case in its subordinating function encodes subsequent relative tense. Dative-marked subordinate clauses need not be controlled by any argument of the main clause but they are typically construed with subjects (transitive and intransitive). Dative-marked subordinate clauses typically indicate the purpose of the event expressed in the main clause, as in (1922). (1922)

Nyantu=lu wokabat=ya-nani jangu nya-ngu-ku. recog1=3pl.s walk_around=go-ipfv.pst recog2 see-inf-dat ‘Those ones used to go walkabout to look around.’ (KAP: LC43b: 495112_498365)

The dative case in subordinating function is also found on arguments of the subordinate clause as in (1923). (1923)

Ngaju=rna ya-nku ngurra-kurra jilan-ku 1sg=1sg.s go-pot camp-all lunch-dat ‘I’m going home to eat lunch.’ (PRN: TG1_0034-01)

nga-rnu-ku eat-inf-dat

Simpson suggests that Warlpiri structures similar to (1923), which use a dative on a nominalised verb as a purposive, are no different ‘from a nominal with -ku used as a purposive’ (1983, p. 181). Nash (1986) chooses to formalise a distinction between a dative -ku and a purposive -ku on the basis that it is only the former that is ever cross-referenced in the bound pronoun. These facts are equally true of Ngardi. Even when a non-finite subordinate clause shares the same morphological case marker

710 

 11 Complex syntax

with a main clause argument, they are clearly different syntactic elements. Thus, when attempting to elicit ‘I am speaking to you, (while you are) lying down’ in (1924), it was accepted as grammatical only if luk=karri-nyu-ku was interpreted as an independent-purpose clause, while the second person object bound pronoun was considered a distinct dative argument of the verb. (1924)

Wangka-nyanta=rna=ngku luk=karri-nyu-ku. speak-prs=1sg.s=2sg.o lie_down=be-inf-dat ‘I am speaking to you for the purpose of sleep (solicitation for sex).’ (TEN: TEN1-2018_025-01: 3816815_3821273) *‘[I am speaking to youi] lying downi.’588 *‘[I am speaking] so you lie down.’589

The restrictions on interpretations of (1924) provide some evidence that when an overt dative NP does appear in a clause it may variously be positioned in the main clause or the subordinate clause. Thus, in (1925), a dative NP nyuntu-ku is prosodically separated from the main clause and could arguably constitute a subject of the subordinate clause ‘for you to eat’, or an independent benefactive argument of the main clause ‘for you’. Irrespective of how the overt dative NP is parsed, the 2sg.o bound pronoun in the main clause can only refer to an argument of the main clause, be it a prosodically detached dative NP or an elided dative NP.590 (1925)

Minya=rna=ngku kuyi ka-ngani, nyuntu-ku nga-rnu-ku. this=1sg.s=2sg.o meat carry-ipfv.pst 2sg-dat eat-inf-dat ‘I brought this meat for you, for you to eat.’ ‘I brought this meat for you, for eating.’ (MMN: TEN1-2018_017-03: 103577_107438)

Importantly there are some restricted instances of dative subordinate clauses which operate as complements to a small set of predicates (typically predicates expressing mental states or cogitation). Examples are strong nexus preverbs jirniny ‘dislike’ (1926) and inkilji ‘unwilling’ (1927). (1926)

Jirniny ma=rna karri-nyanta wurna-ku. dislike top.aux=1sg.s be-prs journey-dat ‘I don’t feel like travelling.’ (PSM: TEN1-2018_035-02: 1221089_1226128)

588 This reading requires allative case-marking on the non-finite subordinate clause. 589 This reading is ungrammatical because the cross-referencing bound pronouns are incompatible with a subordinate clause involving a monovalent main verb, i.e. ‘I am speaking [for you to lie down]’. 590 Importantly, a dative NP expressing a benefactive dative argument never co-occurs with an overt and co-referential subject argument of a non-finite subordinate clause.

11.3 Non-finite subordinate clauses 

(1927)

 711

Inkilji yirra-rnu-ku tungkurl-ku. unwilling place-inf-dat neck_strap-dat ‘(The horse) is unwilling to have the halter put on.’ (PSM: TEN1-2018_01901: 41406_43809)

Dative subordinate clause can also serve as complements to verbless clauses as in (1928) (see §9.2.1.3 for further discussion). The subject of these subordinate clauses is frequently left unspecified. (1928) Minya kuyi jalangu-ku nga-rnu-ku. this meat today-dat eat-inf-dat ‘This meat is ready for eating.’ (MMN: TEN1-2018_017-03: 103577_107438) 11.3.5.2 Preparative -kungarnti The preparative is one of the few suffixes in Ngardi that has an exclusively subordinating function.591 It signals subsequent relative tense and indicates that the event depicted by the non-finite predicate is imminent and somewhat unavoidable. Like the other subsequent subordinators, it has no strict control relation with arguments of the main clause and can be variously construed with absolutive subjects (1929), absolutive objects (1930) and transitive subjects (1931), (1932), or even none of the above. Unlike the dative, when the preparative is controlled by a transitive subject, the non-finite subordinate clause takes further ergative case-marking. (1929)

Tititi=pu-nganta wanti-nyu-kungarnti. shiver=hit-prs fall-inf-prep ‘It’s wobbling, about to fall.’ (MDN: Cataldi, 2011, p. 299)

(1930) Puliki=rnalu=yanu yard-ta yirra-rni mayarra=lku cattle=1pl.excl.s=3pl.o yard-loc place-pst again=then jilykarra-kungarnti. branding-prep ‘We put them (back) in the yard again, ready for branding.’ (TJL: TT75_1301: 1307832_1312958)

591 Warlpiri has a suffix of the same form and similar function labelled ‘preparative purposive’ (Nash, 1986, p. 299).

712 

 11 Complex syntax

(1931)

Kupa-rni=lu hawij lurrij, wurna-kungarnti-rlu kardiya-rlu cook-pst=3pl.s house finish journey-prep-erg white_man-erg kuliny-ju aggressive-erg ‘They burnt that house to the ground; those mean whitefellas did it before they left.’ (BSE: TEN1-2017_015-01: 727877_731180)

(1932)

Para-nganta la-nu-kungarnti-rlu, kuyi la-ni=lu, para-nganta follow-prs pierce-inf-prep-erg meat pierce-pst=3pl.s follow-prs jina. foot ‘He is following it, ready to spear it, they speared the animal, he follows it on foot.’ (LC48_a: 905850_918078)

The preparative appears on action nominals (1931), non-finite verb forms (1932) as well as regular nominals (1933). (1933)

Parr=pi-nya=lu=nyanu purlapa-kungarnti-rlu, apply=hit-pst=3pl.s=refl male_ceremony-prep-erg purr=pu-nganta=lu=nyanu juju-kungarnti-rlu. apply=hit-prs=3pl.s=refl ceremony-prep-erg ‘They put kapok on each other, ready for the ceremony, they are putting kapok on each other ready for the sacred performance.’ (MDN: LC48b)

11.3.5.3 Evitative -ngkamarra The evitative case in subordinating function signals subsequent relative tense and indicates that the event depicted by the non-finite predicate is feared and to be avoided. Like the other subsequent subordinators, the evitative has no strict control relationships with main clause arguments but is often construed with objects of main clauses. (1934) Ya-nanta ma=rna kayirra-purda, nya-ngu-ngkamarra go-prs top.aux=1sg.s north-wards see-inf-evit ngaringka-rlamarra. woman-evit ‘I am going north, in case the woman sees me/for fear of the woman seeing me.’ (PSM: TEN1-2018_34-01: 1109870_1118879) There are very few examples of evitative case-marked subordinate clauses in Ngardi and the semantic content of an evitative non-finite subordinate clause is instead often achieved through ‘apprehensional coordinate clauses’, as described in §11.2.3.

11.3 Non-finite subordinate clauses 

 713

11.3.6 Concord subordinators Concord or ‘C-subordinators’ are case suffixes that relate subjects of subordinate clauses with arguments of main clauses. There are only two C-subordinators: the ergative (§11.3.6.1) and the obviative (§11.3.6.2).592 11.3.6.1 Ergative -rlu The ergative C-subordinator marks same-subject identity with a transitive main clause subject. This control relation applies irrespective of the transitivity of the subordinate clause predicate. As a C-subordinator, it lacks any of its own relative tense features and may co-occur with T-subordinators (see §11.3.7). If the ergative subordinator appears without an accompanying T-subordinator (1935), the temporal relation between the two clauses is generally interpreted as being simultaneous to that of the main clause. (1935)

Mayarra=lku=yi=nta ngawurr-ngawurr=ma-nanta jaru-ngku again=then=1sg.o=2pl.s hot-rdp=get-prs talk-erg wangka-nyu-rlu. speak-inf-erg ‘You are making me angry by talking.’ (MDN: LC42b: 1228669_1233955)

The ergative subordinator may also occur with both the elative -jangka (1936) and the associative -warnu (1937). In both contexts, the ergative marks subject control while the allative -jangka and the associative2 -warnu both mark prior relative tense. (1936)

Wakurra=rna=rla purda=nya-nganta, nyanyikarnu=warru neg=1sg.s=3sg.obl listen=see-prs olden_days=incr nya-ngu-jangka-rlu. see-inf-ela-erg ‘I don’t remember it, it’s a long time since I have seen it.’ (MRN: LC35b: 1346220_1351139)

(1937)

Ngurra-ngka=rna ya-nku kujarra nga-rnankura luk-warnu-rlu. camp-loc=1sg.s go-pot two eat-ipfv.hort lie-assoc2-erg ‘I will go to camp and eat them both (bush yams), having lain down.’ (JLI: TT76_2201: 1485062_1487349)

592 In Wanyjirra, the dative -ku is a productive C-subordinator (in addition to also being a T-subordinator) and can be found following a T-subordinating locative -rla (Senge, 2015, p. 133).

714 

 11 Complex syntax

The ergative is also found following the ssub -karra subordinator (1938). The -karra subordinator encodes the temporal relation of simultaneous tense while the ergative C-subordinator marks the clause as being controlled by the subject of the transitive main clause. (1938) Nya-ngani-nyirra=rna kuja-rlu ya-nu-karra-rlu. see-ipfv.pst-narr=1sg.s thus-erg go-inf-ssub-erg ‘I saw it while I was travelling.’ (MDN: LC22b: 930878_934415) The ability of the ergative C-subordinator to follow certain T-subordinators in nonfinite clauses is functionally parallel to the ergative case used to mark depictive secondary predicates that are predicated of transitive subjects (see §9.3.1). In fact, the only clear criterion on which these structures are distinguished from each other is the presence or absence of a non-finite verb form or preverb. 11.3.6.2 Obviative -rlakarni Cataldi (2011) lists a suffix -rlakarni ‘elsewhere doing’ which she equates to the Warlpiri obviative subordinator -ngkarni/-rlarni (Nash, 1986, p. 262). As per Cataldi’s observations, I have labelled this suffix ‘obviative’ following the Warlpiri literature. This subordinator serves to mark simultaneous tense and non-identity of the subordinate-clause subject with any of the arguments of the main clause, as illustrated in (1939) (1939)

Purdangirli, wirlinyi-rlakarni. behind hunting-obv ‘(They stayed) behind while the others were hunting.’ (PPN: Cataldi, 2011, p. 276)

Examples (1940) and (1941) illustrate the obviative following the locative subordinator. The obviative marks that the subject of the non-finite subordinate clause is not co-referent with any of the main clause arguments; while the locative marks simultaneous tense. (1940) Jirrminy-ja-rlakarni, ngaju=rna luk=wanti-nya mula-ngka. straight-loc-obv 1sg=1sg.s lie_down=fall-pst here-loc ‘While the others were passing through, I lay down (slept) here.’ (PSM: TEN12019_029-01: 1932293_1936983)

11.3 Non-finite subordinate clauses 

(1941)

 715

Wakurra, yala-rla=rna turn=ma-ni, nyuntu-ngka neg that-loc=1sg.s cease=get-pst 2sg-loc Balgo-ngka-rlakarni.593 place_name-loc-obv ‘I put it (the fire) out there while you were elsewhere [sic] at Balgo.’ (MNN: MAATI3)

The Warlpiri suffix -ngkarni/-rlarni is used in what Hale terms ‘pure obviative clauses’ whose subjects are not co-referent with any of the core arguments of the main clause and encodes simultaneous relative tense. In Warlpiri, the subjects of these pure obviative clauses may appear in the main clause as non-core dative arguments and will similarly be marked with a dative case form in the subordinate clause (see Hale, 1983, 295ff.). Interestingly, a related form -ngkarniny/-larniny appears in Wanyjirra and similarly marks simultaneous tense and non-identity with any NP of the main clause.594 Crucially, it differs from Warlpiri in a number of other syntactic aspects, including the marking of other subordinate clause arguments. Senge (2015, p. 629) tentatively suggests that -ngkarniny/-larniny can be analysed as a sequence of -rla + -rniny, the former being the locative subordinator and the latter being a C-subordinator – which she labels a pure obviator (see Senge, 2015, 627ff. for a complete discussion).595 The case for multimorphemic status is certainly no stronger for Ngardi – the major limitation being the presence of only two tokens in the corpus. Cross-linguistically, it is nevertheless interesting that Warlpiri, Ngardi and Wanyjirra all possess a subordinator built off the locative that serves to mark simultaneous tense and subordinate subject non-identity with main clause arguments.

11.3.7 T- and C-subordinators in combination Throughout the preceding sections it was observed that an ergative C-subordinator may follow another T-subordinator, including the privative (technically unspecied for 593 I have no explanation as to why the obviative does not also appear on nyuntu-ngka. 594 Note that this suffix – evident in data collected in the 1970s – was no longer being used by Wanyjirra speakers at the time of Senge’s fieldwork (2009–2012). 595 The -rni element may be related to the ‘direction towards’ clitic -rni. Haiman and Munro (1981, p. xiv) flag deictic markers (specifically towards and away suffixes) as being one of a number of common diachronic origins of switch reference markers – although the specific grammaticalisation path in this context is unclear. Senge (2015, p. 629) notes that it is unclear whether the Wanyjirra form -arniny is etymologically related to Warlpiri -rni of -ngkarni-rarni (and compare Ngardi -rlakarni). Alternatively, Senge notes that -arniny is formally similar to the perlative suffix -rniny which follows demonstratives in Wanyjirra. Compare also a highly restrictive allative suffix -rniny which appears in Gurindji and Bilinarra but has a very restricted distribution, namely on the demonstratives guya ‘thus’ and murla ‘this’ in the former and just guya in the latter (Meakins & Nordlinger, 2014, p. 176).

716 

 11 Complex syntax

relative tense). This pattern reflects a more general function of the ergative in marking ‘transitive subject construal’ – introduced in the discussion of ‘secondary predicates’ (§9.3.1). Unlike the ergative, the obviative C-subordinator does not appear to combine productively with the T-subordinators. This may, however, be a limitation of the data. Table 145 summarises all attested T- and C-subordinator sequences. Table 145: Co-occurrence of T-subordinators and the Ergative. Subordinator sequence

Examples

seq_loc-erg loc-erg ela-erg assoc2-erg ssub-erg dat-erg prep-erg all-erg priv-erg

N N Y (1936) Y (1901) Y (1911) N Y (1931) N Y (1921)

The only T-subordinators that cannot be followed by the ergative are the sequential locative, locative, dative and allative subordinators. All of these subordinators can be construed with a range of argument types including ergative subjects, despite not needing any further case marking. Note that the dative subordinator was unique in permitting the subject of the subordinated clause to remain unspecified. The set of subject co-reference possibilities (within the core arguments of A, S and O) for this set of subordinators, which I term ‘specified’, is presented in Table 146. Table 146: Subject co-reference possibilities of ‘specified’ subordinators (no ergative marking).

Locative (Prior) Locative (Simult) Sequential Locative Dative Allative

Transitive subject (A)

Intransitive subject (S)

✓ (1888) ✓ (1906)

✓ (1889) ✓ (1902) ✓ (1892) ✓ (1922)

✓ (1839)596

Transitive object (O) ✓ (1876) ✓

596 Note that, strictly speaking, dative subordinate clauses are frequently employed without overt specification of the subject. It is possible for subjects of dative subordinate clauses to be construed with referents not specified in the main clause and represent a general nominalised purpose, e.g. yarli-rnu-ku ‘for washing’ etc.

11.3 Non-finite subordinate clauses 

 717

All other T-subordinators are ‘unspecified’, which is to say they require ergative case-marking when exhibiting subject control. When lacking ergative marking, they can freely be construed with any unmarked (that is to say absolutive) main clause argument. To recall a single example, the preparative -kungarnti can be construed of intransitive subjects (1929) or transitive objects (1930).

11.3.8 Insubordinate usages of subordinating case A final usage of subordinating case in Ngardi is in certain stylistically-marked constructions involving the repetition of subordinator-marked non-finite verbs in succession to denote extreme duration or repetition of an event. Interestingly, these forms are typically not introduced by a main clause predicate and are instead syntactically ‘insubordinate’. Chains of dative insubordinate clauses are the most frequent in the corpus. Examples are (1942) and (1943). (1942) Luwa-rnu-ku, luwa-rnu-ku, luwa-rnu-ku, luwa-rnu-ku, shoot-inf-dat shoot-inf-dat shoot-inf-dat shoot-inf-dat luwa-rnu-ku nyamu. shoot-inf-dat finish ‘(They) were striking and striking and striking and striking and striking (it). . . finish.’ (PPN: LC09) (1943)

Ngurra-ngka yala jarnti-rnu-ku, jarnti-rnu-ku, jarnti-rnu-ku camp-loc that incise-inf-dat incise-inf-dat incise-inf-dat jarnti-rnu-ku, jarnti-rnu-ku, jarnti-rnu-ku. incise-inf-dat incise-inf-dat incise--inf-dat ‘At home they were carving and carving and carving and carving and carving and carving.’ (MDN: LC22a: 1427833_1432820)

Insubordinate clauses also involve the rare imperfective infinitive as shown in (1944). (1944) Kankarla kuma-rnani. top cut-ipfv.pst ‘He cut it all up on top.’ nga-rnanu-ku nga-rnanu-ku nga-rnanu-ku. eat-ipfv.inf-dat eat-ipfv.inf-dat eat-ipfv.inf-dat ‘(Then we) were eating and eating and eating.’ (YDN: LC49b: 1589525_1593045)

Appendix: Narrative texts Text 1: Two boomerangs Author: Recording Date: Description:

Jack Langgamarru 30 September, 1976 Jack Langgamarru describes a hunting trip and the collection of wood for the manufacture of boomerangs.

1.

Ya-ni=rna wirlinyi. go-pst=1sg.s hunting ‘I went hunting.’

2.

Karli=rna paja-rni. boomerang=1sg.s bite-pst ‘I cut a boomerang (out).’

3.

Karli=rna paja-rni, wirlinyi-rlu, ya-ni nawu. boomerang=1sg.s bite-pst hunting-erg go-pst now ‘I cut out a boomerang, while I was out hunting.’

4.

Karli=rna paja-rni, nyurnin. boomerang=1sg.s bite-pst special ‘I cut a suitable piece of (wood for making a) boomerang.’

5.

Karli=rna paja-rni, ya-ni=rna. boomerang=1sg.s bite-pst go-pst=1sg.s ‘I cut a boomerang, then I went.’

6.

Jarany=pa=rna pi-nya. roughtail=ep=1sg.g hit-pst ‘I killed a roughtail lizard.’

7.

Ya-ni=rna wirlinyi. go-pst=1sg.s hunting ‘I’d gone hunting.’

8.

Ngu=rna yut=wanti-nya. seq.aux=1sg.s sit=fall-pst ‘Then I sat down.’

https://doi.org/10.1515/9783110752434-012

720 

9.

 Appendix: Narrative texts

Ngapa=rna nga-rni. water=1sg.s eat-pst ‘I drank (some) water.’

10.

Ngapa wirlinyi-jangka-rlu, ngu=rna palman-tu nga-rni ngapa. water hunting-ela-erg seq.aux=1sg.s passing-erg eat-pst water ‘Having been hunting, I drank some water as I was passing through.’

11.

Ngu=rna nga-rni. seq.aux eat-pst ‘I drank it (the water).’

12.

Ya-ni=rna. go-pst=1sg.s ‘I went.’

13.

Ngapa=lku=rna nga-rnungka-nya palman-tu. water=then=1sg.s eat-x&move-pst passing-erg ‘On my way through, I drank water and kept moving.’

14.

Ya-ni=rna, ngurra-kurra, ngurra-kurra go-pst=1sg.s camp-all camp-all ‘I went to the camp, to my own camp.’

15.

Ngu=rna ya-ni. seq.aux=1sg.s go-pst ‘Then I went.’

16.

Ya-ni=rna yirra-rni=rna. go-pst=1sg.s place-pst=1sg.s ‘I went, I put it (the lizard) down.’

17.

Jarany yala karli=rna=pulany yirra-rni ngurra-ngka roughtail that boomerang=1sg.s=3du.o place-pst camp-loc ngaju-punta-rla. 1sg-ppos-loc ‘I put the roughtail lizard and the boomerang down at my camp.’

ngaju-punta-kurra. 1sg-ppos-all

Text 1: Two boomerangs 

18.

Nga=rna–, yirra-rni=rna=pulany ngurra-ngka ngaju-punta-rla. seq.aux=1sg.s place-pst=1sg.s=3du.o camp-loc 1sg-ppos-loc ngurra-ngka. camp-loc ‘And then I–, I put the two down at my camp, at the camp.’

19.

Karli-kujarra kuriny=pa=rna=pulany yirra-rni. boomerang-dual two=ep=1sg.s=3du.o place-pst ‘I put the two boomerangs down.’

20.

Kuja=rna=pulany nyurnin paja-rni. sub=1sg.s=3du.o special bite-pst ‘Those two special ones which I had cut out.’

21.

Ya-ni=rna. go-pst=1sg.s ‘I went.’

22.

Kuyi=rna yirra-rni. meat=1sg.s place-pst ‘I put that meat (down).’

23.

Ngu=rna yirra-rni ngurra-ngka ngaju-punta-rla. seq.aux=1sg.s place-pst camp-loc 1sg-pposs-loc ‘I put (the wood for the boomerang) down at my camp.’

24.

Paja-rni=rna. bite-pst=1sg.s ‘I chopped it.’

25.

Kuja=rna yungkut=paja-rni. sub=1sg.s tear=bite-pst ‘Then I cut it in half.’

26.

Nawu ma-ni=rna yangi. now get-pst=1sg.s one ‘Then I got another one.’

27.

Wuruju=ma-ni=rna yangi wuruju=wali. good=get-pst=1sg.s one good=then ‘I made that other one good, it’s a good one.’

 721

722 

 Appendix: Narrative texts

28.

Wuruju=wali=rna yirra-rnurra. good=then=1sg.s place-pst.hence ‘It was good and I put it away.’

29.

Kunyjuru-rla warlu-ngka=rna waa=pi-nya. smoke-loc fire-loc=1sg.s straighten=hit-pst ‘I put it in the smoke in the fire and straightened it.’

30.

Maja-rni=rna, wuruju=ma-ni yirra-rni. refine-pst=1sg.s good=get-pst place-pst ‘I straightened it, set it and made it true.’

31.

Ngu=rna yirra-rni wuruju, wuruju=ma-ni=rna karli yala. seq.aux place-pst good good=get-pst=1sg.s boomerang that ‘I fashioned it well, I made that one into a good boomerang.’

32.

Wuruju=ma-ni=rna. good=get-pst=1sg.s ‘I made it well.’

33.

Wuruju=ma-ni=rna=pulany, karli ‘good=get-pst=1sg.s=3du.o boomerang ‘I fixed up two, those boomerangs.’

34.

Ngu=rna yangi kala ma=yi lankarr=ya-ni yangi. seq.aux=1sg.s one contr top.aux=1sg.o split=go-pst one ‘I have one, but the other one split on me.’

35.

Kuyi-pardu=rna kupa-rni jangu jarany. meat-dim=1sg.s cook-pst recog2 roughtail ‘I cooked that meat, that roughtail lizard.’

36.

Jarany, ma=rna kupa-rni. roughtail top.aux=1sg.s cook-pst ‘That roughtail lizard, I cooked it.’

37.

Lurrij=pa=rna finish=ep=1sg.s ‘I ate it all.’

nga-rni. eat-pst

yala. that

Text 1: Two boomerangs 

38.

Iji-wangu-rlu nga-lku, ngaju-kariny-ju=lku. gift-priv-erg eat-pot 1sg-other-erg=still ‘I will eat it all by myself, without sharing (I had thought).’

39.

Yangi-pardu-pardu. one-dim-rdp ‘It’s only one.’

40.

Yangi-pardu=rna nga-rni lurrij. one-dim=1sg.s eat-pst finish ‘I ate just one, that’s it.’

41.

Dat pinij. that finish ‘That’s all.’

42.

Ya-ni=rna. go-pst=1sg.s ‘I went.’

43.

Pirranginti=karri-nya=yi. afternoon=be-pst=1sg.o ‘It was getting dark on me/late afternoon arrived.’

44.

Ya-ni=rna pukany=karri-nya, pukany nawu. go-pst=1sg.s sleep=be-pst sleep now ‘I went and slept, I drifted off to sleep then.’

45.

Warruru-rla=wali. night-loc=then ‘It was night time now.’

46.

Warurru-rla. night-loc ‘It was night time.’

47.

Luk=karri-nya=rna. lie_down=be-pst=1sg.s ‘I was lying down.’

 723

724 

 Appendix: Narrative texts

48.

Luk=karri-nya=rna rangarni karri-nya. lie_down=be-pst=1sg.s morning be-pst ‘I was lying down, then morning came.’

49.

Kakarrara=yirla raakarra ya-ni. east=1sg.lct sunrise go-pst ‘Sunrise broke in the east on me.’

50.

Raakarra ya-ni=yirla kakarra. sunrise go-pst=1sg.lct east ‘The sunrise came upon me from the east.’

51.

Purangu pirdiny-parra=karri-nya. sun arise-ssub=be-pst ‘That’s when the sun got up.’

52.

Pinij. Finish ‘That’s it!’

Text 2: Frog story: Frog where are you? 

 725

Text 2: Frog story: Frog where are you? Author: Recording Date: Description:

Marie Mudgedell 30 May 2017 Marie Mudgedell composed this story as a response to the picture storybook: Frog, where are you? (Mayer, 1969)

1.

Minya=lu karri-nyanta, marru-ngka, murrku, kunyarr, pujurl. this=3pl.s be-prs house-loc boy dog frog ‘This lot are at home in the house, the boy, the dog and the frog.’

2.

Ngu=pula pukany=karri-nya, kunyarr seq.aux=3du.s sleep=be-pst dog ‘Then the dog and the boy went to sleep.’

3.

Pujurl murlukun-ta. frog jar-loc ‘The frog is in the jar.’

4.

Wakurra wurna ma. neg journey top.aux ‘Oh no, the frog is escaping.’

5.

Ya-ni yalu-jangka murlukun-jangka. go-pst that-ela jar-ela ‘He got out of that jar.’

6.

Rangarni-kariny=pa=pula pirdiny=karri-nya. morning-other=ep=3du.s arise=be-pst ‘Next morning those two wake up.’

7.

Nya-nya=pula murlukun. see-pst=3du.s jar ‘Those two saw the jar.’

8.

“Wakurra, wanji-rla=parda ya-ni.” neg where-loc=dub go-pst ‘“Oh no, he has gone somewhere.”’

9.

Yala pujurl. that frog ‘That frog.’

murrku. boy

726 

 Appendix: Narrative texts

10. Nya-nya=pula puuj–, waku. see-pst=3du.s bush neg ‘Those two looked in the [bush], nothing.’ 11.

Murlukunpa parnti=ma-nan yalu-ngku kunyarr-u, waku. jar smell=get-cust that-erg dog-erg nothing ‘That dog sniffs the jar, nothing.’

12. Ya-ni=pula nya-ngani, wakurra=pula parli=pi-nya. go-pst=3du.s see-ipfv.pst neg=3du.s find=hit-pst ‘Those two went, they were looking for him, but they did not find him.’ 13. Kunyarr wanti-nya. dog fall-pst ‘The dog fell.’ 14.

Murrku-ngku ma-ni yala kunyarr ngamurlu=marda-rnani. boy-erg get-pst that dog embrace=have-ipfv.pst ‘The boy picked up that dog and held him in an embrace.’

15. Nya-ngani=pula. see-ipfv.pst=3du.s ‘Those two were looking.’ 16. Ya-ni=pula– warri-rnani-nyirra. go-pst=3du.s search-ipfv.pst-narr ‘Those two went – (they) were looking around.’ 17.

Warri-rnani=rla jakarr purnu-kariny, purnu-kariny. search-ipfv.pst=3sg.obl try tree-other tree-other ‘(Those two) are searching for him unsuccessfully, (going) from tree to tree.’

18.

Parli=pi-nya=pula kurrku nawu. find=hit-pst=3du.s hole now ‘Then the two of them found a hole.’

19.

Yala murrku ma-rnani, “Mula ngarda=n kurrku-ngka?” wangka-nyani. that boy talk-ipfv.pst that hyp=2sg.s hole-loc speak-ipfv.pst ‘That boy said, “Are you there in the hole?”’

Text 2: Frog story: Frog where are you? 

20. ‘Mula ngarda=n kurrku-ngka tardarr=ya-ni? Wakurra.’ this hyp=2sg.s hole-loc enter=go-pst neg ‘Did you go into this hole? – No.’ 21. Parli=pi-nya punypuny kurrku-jangka. find=hit-pst mouse hole-ela ‘He found the mouse from inside that hole.’ 22. Punypuny ma-rni, ‘Wakurra=yi=n parnki=ma-nankura!’ mouse speak-pst neg=1sg.o=2sg.s wake=get-ipfv.hort ‘The mouse said, “Do not wake me up!”’ 23. ‘Pukany=pa=rna luk=wanti-nya.’ sleep=ep=1sg.s lie_down=fall-pst ‘I am having a sleep.’ 24. Kunyarr-tu parli=pi-nya kirangu, purnu-ngka kankarni. dog-erg find=hit-pst honey, tree-loc up.b ‘The dog found the honey up in the tree.’ 25. Wakurra wanti-nya yala neg fall-pst that ‘That honey fell down.’

kirangu. honey

26. Murrku waru=ya-ni purnu-ngka kankarra. boy climb=go-pst tree-loc up.a ‘The boy climbed up the tree.’ 27.

Nya-ngani=rla pujurl-ku wilayi – wakurra. see-ipfv.pst=3sg.obl frog-dat around neg ‘He looks around for the frog – but nothing.’

28. Yalu-jangka laman-jangka pirriri=ya-ni nganayi, kurrkurr. that-ela hollow-ela emerge=go-pst whatsit owl ‘From that hollow, a whatsit emerged, an owl.’ 29. Murrku wanti-nya xxx, wanti-nya boy fall-pst xxx fall-pst ‘The boy fell, he fell down.’

kaniny. down

 727

728 

 Appendix: Narrative texts

30. Rayin=karri-nya=rlanyanta yalu-ngka, kurrkurr-ta. afraid=be-pst=3sg.lct that-loc owl-loc ‘He was afraid of that one, the owl.’ 31. An minya-rlu, nguriny-ju yalu kuliny-ju=lu pila=ma-nani conj this-erg fly-erg that aggressive-erg=3pl.s follow=get-ipfv.pst kunyarr, – kirangu-jangka-rlu. dog honey-ela-erg ‘And those ones, those aggravated bees, chased after the dog on account of that honey.’ 32. Kurrkurr-tu pila=ma-nani murrku. owl-erg follow=get-ipfv.pst boy ‘The owl was chasing after the boy.’ 33. “Wakurra=yi=n ngaju, ngaju-kuny-kurra ngurra-kurra neg=1sg.o=2sg.s 1sg 1sg-poss-all camp-all ‘“Do not come into my home (and disturb me)!”’

ya-nanku.” go-ipfv.pot

34. “Parnki=ma-nanta=yi=n pukany-ja, wurna=ya-nta!” wake=get-prs=1sg.o=2sg.s sleep-loc journey=go-imp ‘“You are waking me while I am sleeping, go away!”’ 35. Kurrkurr-tu nya-ngani yalu murrku. boobook-erg see-ipfv.pst that boy ‘The owl was (still) watching that boy.’ 36. Kunyarr yala=lku – paja-rni yalu-ngku nguriny-ju, kuliny-ju dog that=then bite-pst that-erg fly-erg aggressive-erg ‘That dog then, he was bitten by the enraged bees.’ 37.

Yalu-jangka, ya-ni parli=pi-nya reindeer. that-ela go-pst find=hit-pst reindeer ‘From there, he (the boy) went and found a reindeer.’

38. Reindeer-rlu=pulany pila=ma-ni, “Ya-nta=pula!” reindeer-erg=3du.o follow=get-pst go-imp=du.s ‘The reindeer chased those two, “You two go away!”’ 39.

“Nyampa-ku=npula mu-kurra ya-nanta?” what-dat=2du.s here-all go-prs ‘“Why are you two coming here?’”

Text 2: Frog story: Frog where are you? 

 729

40. Ngu=pulany puju=ma-ni— ngu=pulany warrit=kiji-rni, seq.aux=3du.o push=get-pst seq.aux=3du.o uncover=throw-pst ngapa-kurra kaniny. water-all down ‘He pushed those two away, chucked them off down into the water.’ 41. Purluk=wanti-nya=pula ngapa-ngka. submerge=fall-pst=3du.s water-loc ‘The two fell into the water.’ 42. Pirdiny=karrinya=pula. arise=bepst=3du.s ‘Those two got up.’ 43. Pilaka=ka-nya murrku-ngku, kunyarr, kankarra. shoulder=carry-pst boy-erg dog up.a ‘The boy carries the dog around his shoulders, up top.’ 44. “Mula=parda laman-ta nya-ngku=rli=rla, yalu-ku, pujurl-ku.” here=dub hollow-loc see-pot=1du.incl.s=3sg.obl that-dat frog-dat ‘“Perhaps here in the log you and I should look for him, for the frog.”’ 45. Nya-ngani=pula wilayi. see-ipfv.pst=3du.s around ‘Those two looked around.’ 46. Parli=pi-nya=pulany kuriny, pujurl-kujarra. find=hit-pst=3du.o two frog-dual ‘(He) found two, two frogs.’ 47.

Yalu-jangka=lu wirrpa=wali pirri=ya-ni, that-ela=3pl.s many=then arrive=go-pst ‘After that, out came lots of frogs.’

pujurl. frog

48. Yala=lku=pula yut=karri-nyani, murrku, kunyarr, ngurrku=wali. that=then=3du.s sit=be-ipfv.pst boy dog happy=then ‘Those two were sitting down, the boy, the dog, they were happy.’ 49. Ngurrku=karri-nya=pula. happy=be-pst=3du.s ‘They were happy.’

730 

 Appendix: Narrative texts

50. Ma-rni=lu yala wirrpa purruk– say-pst=3pl.s that many frog ‘All the frogs said . . .’

pujurl. frog

51. “Yangi ka-ngka=pula yala pujurl wirrirl nyuntu-kuny ngurra-kurra.” one carry-imp=du.s that frog back 2sg-poss camp-all ‘“You two, take that one, the frog, back with you to your place.”’ 52. ‘Kuja=npula marda-rnanta naya yalu-ngka marru-ngka.’ sub=2du.s have-prs always that-loc house-loc ‘Because you two always look after him there in that house.’ 53. Ya-ni wali=lu wirrirl. go-pst then=3pl.s back ‘Then they all went back.’ 54. Ngurrku=pula karri-nyani. happy=3du.s be-ipfv.pst ‘Those two were happy.’ 55. Kunyarr-tu murrku, ka-nya=pula wirrirl. dog-erg boy carry-pst=3du.s back ‘The dog, and boy, took him back (to the house).’ 56. Yuwayi, wali=lku. yes indeed=then ‘Yeah, that’s it.’

Text 3: Mungkururrpa 

 731

Text 3: Mungkururrpa Author: Recording Date: Description:

Patrick ‘Jupiter’ Smith 4 May 2018 This is an historical account of the formation of Mongrel Downs/ Tanami Downs Station (Mungkururrpa), by Patrick Smith. Patrick recounts his involvement in the creation of the station and his connection to the Mahood family.

1.

Ol kardiya yangi, yirdi=rla Joe Mahood. old white_man one name=3sg.obl name ‘There was this one old whitefella, his name – Joe Mahood.’

2.

En yangi-kariny, Bill Wilson wurna ngurra-kurra-rlu and one-other name travel camp-all-erg milyine, mani-kurlu wirrpa-kurlu. millionaire money-prop many-prop ‘And another one, Bill Wilson, (they) were heading to those homelands, a millionaire, with lots of money.’

3.

Wangka-nyani=pula–. speak-ipfv.pst=3du.s ‘Those two had got to talking–.’

4.

Ya-ni=lu abalein-kurlu nya-nya=lu kantri Mongrel Downs, go-pst=3pl.s airplane-prop see-pst=3pl.s country place_name “Oh ngapa-kurlu, wuruju ngurra puliki-ku.” oh water-prop good camp cattle-dat ‘They had gone by aeroplane and seen the country around Mongrel Downs, “Oh (this country) has water; it is good country for cattle!”’

5.

Wangka-nya=rla, “yala wurna!” speak-pst=3sg.obl that travel ‘He said to him, “Let’s go there!”’

6.

Bill, nganayi-ku, Joe Mahood-ku, “Wal nyuntu kardiya manija, name whatsit-dat name-dat well 2sg white_man manager Mongrel Down ngurra yala, Mongrel Down.” place_name camp that place_name ‘Bill (said) to him, to Joe Mahood, “Well, you should be the manager (for) that Mongrel Downs country, Mongrel Downs.”’

732 

 Appendix: Narrative texts

7.

Wal puliki nawu=rnalu well cattle now=1pl.excl.s ‘Well, then we took the cattle.’

ka-nya. carry-pst

8.

Wangka-nya, “Puliki ka-ngka=lu drovem kakarra! speak-pst cattle carry-imp=pl.s drive_cattle east wiyinastap yangka, dat far.” weaner_stop recog2 that far ‘He said, ‘“You lot, take the cattle, drive them east, to you know, that weaning (yard), that far.”’

9.

Sambala drei kao an sam yang bul, yang Micky, sambala bul. some dry cow and some young bull, young ‘Micky’ some bull ‘(There were) some dry cows and some young Micky bulls and some older bulls.’597

10. Ka-nya=rnalu kakarra. carry-pst=1pl.excls east ‘We took them all east.’ 11.

Puliki=rnalu majurrum=ma-ni, Kururrungku-rla lurrij. cattle=1pl.excls muster.tr=get-pst place_name-loc finish ‘We mustered the cattle, finishing up at Kururrungku (Billiluna).’

12. Jangu=rnalu nganayi katimat=ka-nya drabtim=ma-ni, puliki recog2=1pl.excl.s whatsit cut.tr_out=carry-pst draft.tr=get-pst cattle turangwan. strong_one ‘We held the cut, we drafted them; all good, strong cattle.’ 13. Wal wurna karri-nya nawu=rnalu, kayirra-ngulu. well journey be-pst now=1pl.excl.s north-abl Well then, we set off, from the north. 14. Ya-ni=rnalu kankarni dreikamp=pa=rnalu wajim=ma-ni go-pst=1pl.excl.s up.b dry_camp=ep=1pl.excl.s watch.tr=get-pst Mangkalarla. place_nameloc ‘We moved up to the dry camp at and watched them there at Mangkala.’

597 ‘Dry cows’ are non-lactating female cattle, while Micky (bulls) are castrated males which have become ‘scrub bulls’ (i.e. they are not held in yards).

Text 3: Mungkururrpa 

 733

15. Minya kayili, Blue Hill-ngulu karlarra. this north place_name-abl west ‘There in the north, west from Blue Hill.’ 16. Yalu-jangka=rnalu ka-nya puliki Pantri Bore-rla. that-ela=1pl.excl.s carry-pst cattle place_name-loc ‘From there, we took the cattle to Pantri Bore.’ 17.

Yalu-ngku=rnalu ngurra karri-nya, wajim=ma-ni=rnalu puliki. that-erg=1pl.excl.s camp be-pst watch.tr=1pl.excl.s cattle ‘We stayed overnight, and we watched the cattle.’

18. Yalu-jangka=rnalu ya-ni New Mission, minya Balgo. that-ela=1pl.excl.s go-pst place_name this place_name ‘After that we went to New Mission, this place here, Balgo.’ 19. Kakarra minya ma warririku. east this top.aux closeby ‘There in the east, close up.’ 20. Warririku xxx ma-ni=lu wajim=ma-ni.598 close xxx get-pst=3pl.s watch.tr=get-pst ‘Closeby, xxx they watched the (cattle).’ 21. Yalu-jangka rangarni-rangarni, wurna, kakarra! that-ela morning-rdp journey east ‘Then, early next morning, (we) set off, east!’ 22. Kakarra, kankarra, Loner Spring. east up.a place_name ‘East, up, (to) Loner Spring.’ 23. Minya kakani=rnalu wajim=ma-ni yalu-ngka, warririku. this east=1pl.excls watch.tr=get-pst that-loc close ‘There (just nearby?) in the east we watched them, close by.’

598 In this text, and others, ‘xxx’ marks a small section of speech that was unclear and/or a sudden break off in speech.

734 

 Appendix: Narrative texts

24. Yalu-jangka wurna=lku rangarni-rangarni, pirriya-puru, pirriya, that-ela journey=then morning-rdp cold-while cold wakurra ita, kirda. neg small big ‘From there travelling early in the morning, in the cold, not just a little cold – absolutely freezing.’ 25. Yalu-jangka=rnalu ya-ni puliki-kurlu kakarra Mintipungu-rla. that-ela=1pl.excl.s go-pst cattle-prop east place_name-loc ‘From there, we went with the cattle east to Mintipungu.’ 26. Kayini-mpala kakarra=rnalu wajim=ma-ni. north-along east=1pl.excl.s watch.tr=get-pst ‘Along the northside in the east, we watched them.’ 27.

Yalu-jangka=rnalu ya-ni, rangarni-rangarni, kakarra Pujikat Boa that-ela=1pl.excl.s go-pst morning-rdp east place_name ‘After that we went, early in the morning, east (to) Pussycat Bore.’

28. Yalu-jangka Pujikat Boa-ngka=rnalu luk=karri-nya. that-ela place_name-loc=1pl.excl.s lie_down=be-pst ‘After that we lay down at Pussycat Bore.’ 29. Wajim=ma-ni=rnalu mayarra. watch.tr=get-pst=1pl.excl.s again ‘All the while, we were watching them.’ 30. Yalu-jangka=rnalu ya-ni, kakarra, that-ela=1pl.excl.s go-pst east ‘After that we moved east, to a spring.’

spring-urla. spring-loc

31. Nga=rnalu wajim=ma-ni. seq.aux=1pl.excl.s watch.tr=get-pst ‘We watched them there.’ 32. Ngurra karri-nya wajim=ma-ni yalu-ngka=lku. country be-pst watch.tr=get-pst that-loc=then ‘(We) stayed on that country and watched (the cattle) in that place.’ 33. Yalu-jangka=rnalu ya-ni Bloodwood. that-ela=1pl.excl.s go-pst place_name ‘After that we went (to) Bloodwood (Bore).’

Text 3: Mungkururrpa 

 735

34. Lidel bit warririku karri-nyanta nawu, ngurra. little bit close be-prs now country ‘Getting a bit closer now, to the country.’ 35. Yalu-jangka=rnalu ya-ni Bloodwood-ta wajim=ma-ni that-ela=1pl.excl.s go-pst place_name-loc watch.tr=get-pst ngurra=rnalu karri-nya. country=1pl.excl.s be-pst ‘After that we went to Bloodwood (Bore), we watched them and stayed overnight.’ 36. Yalu-jangka=rnalu ya-ni kakarra nya-nya=rnalu, that-ela=1pl.excl.s go-pst east see-pst=1pl.excl.s ngurra=wali, Territory Border=rnalu nya-nya. country=then territory border=1pl.excl.s see-pst ‘After that we went east, we saw it, country, we saw the (Northern) Territory border.’ 37.

‘Maitbi warririku nawu’, ngaju=rna yangfala wangka-nya. maybe close now 1sg=1sg.s young.adj speak-pst ‘“Perhaps (we are) close now”, I said, myself just a young fellow.’

38. Yuwayi, kakarra=rnalu ya-ni, nganayi-kurra, Lake Alec, yes east=1pl.excl.s go-pst whatsit-all place_name kayini-mpala yala=rnalu wajim=ma-ni. north-along that=1pl.excl.s watch.tr=get-pst ‘Yeah, we went east, to whatsit, to Lake Alec, along the north, we watched (them) there.’ 39. Yalu-jangka=rnalu ya-ni rangarni-rangarni kakarra. that-ela=1pl.excl.s go-pst morning-rdp east ‘From there we went east in the early morning.’ 40. Ngurra-kariny karri-nya Pamij-mob. country-other be-pst ? mob ‘We stopped at another country, Pamij mob(?).’ 41. Rangarni-rangarni wurna=lku loadimap ngari. morning-rdp journey=then load.tr.compl belongings ‘Early in the morning, travelling again, loading up all the belongings.’ 42. Timana purlpurl— kakarra. horse cover east ‘Covering up the horses and then (heading) east.’

736 

 Appendix: Narrative texts

43. Wanji-rla=rnalu luk=karri-nya. where-loc=1pl.excl.s lie_down=be-pst ‘We lay down somewhere.’ 44. xxx-ngurlu kakarra Bardi Bore—, yuwayi Bardi Bore-rla. place_name-abl east place_name yes place_name-loc karlarra manja-ngka pilayin-ta. west mulga-loc plain-loc ‘From xxx, east to Bardi Bore, yeah, Bardi Bore. There in the west, on the plains in mulga scrub country.’ 45. Yalu-jangka=rnalu ya-ni, wakurra Bardi Bore karri-nyanta that-ela=1pl.excl.s go-pst neg place_name be-prs windmil waku. windmill neg ‘From there we went, at that time, Bardi Bore wasn’t even a windmill, nothing.’ 46. Tumaj yirra-rni Bardi Bore kardiya-rlu yalu ngaka=wali conj place-pst place_name white_man-erg that later=then windmil windmill ‘Because it wasn’t until later that the whitefella sunk that bore.’ 47.

Wal Bardi Bore nawu=rnalu ngarri-rni. well place_name now=1pl.excl.s tell-pst Well, we (all) call it Bardi Bore now.

48. Yalu-jangka=rnalu ya-ni Bardi Bore-jangka kakarra. that-ela=1pl.exl.s go-pst place_name-ela east ‘Then we went from Bardi Bore east.’599 49. Warririku nawu. close now ‘(Getting) close now.’ 50. Kakarra=rnalu ya-ni wal pirdayita ngurrara-kurra nawu east=1pl.excl.s go-pst well xxx country-all now ‘We went east, (back) to homeland country, now, (with) the cattle.’

599  This is possibly ‘Beady Bore’ in the Tanami.

puliki. cattle

Text 3: Mungkururrpa 

 737

51. Lurrij, kardiya-rlu=nganpa manija-rlu mitim=ma-ni finish white_man-erg=1pl.excl.o manager-erg meet.tr=get-pst Joe Mahood-tu. name-erg ‘Finish, the whitefella, the manager greeted us, Joe Mahood.’ 52. Tobacco-kurlu-rlu, mangarri-kurlu-rlu, rajin-kurlu-rlu, Toyota-kurlu-rlu, tobacco-prop-erg veg_food-prop-erg rations-prop-erg Toyota-prop-erg Kururrungku-jangka Toyota yi-nya=lu wurna-rlu yalu-ngku, place_name-ela Toyota give-pst=3pl.s journey-erg that-erg Bill Wilson-tu. name-erg ‘With tobacco, food, rations, Toyota–they sent a Toyota from Kururrungku, that travelling one, Bill Wilson.’ 53. Puliki=rnalu=yanu yupa-rni nawu ngapa-ngka=wali cattle=1pl.excl.s=3pl.o send-pst now water-loc=then ‘We sent the cattle to water, to the trough.’

trof-kurla. trough-loc

54. Rangarni-rangarni=rnalu pirdiny=karri-nyani. morning-rdp=1pl.excl.s arise=be-ipfv.pst ‘Early next morning we got up.’ 55. Jangu rangarni-rangarni=rnalu pirdiny=karri-nyani timana-kurlu wirrirl like morning-rdp=1pl.excl.s arise=be-ipfv.pst horse-prop back jina kuja ngarda=lu wirrirl=ya-nngi puliki ngurrura-kurra foot sub hyp=3pl.s back=go-irr cattle country-all Kururrungku-kurra. place_name-all ‘Yeah, early next morning we were up and heading back on foot with horses in case the cattle happened to head back to their own country – to Billiluna.’ 56. Ya-ni=rnalu wilayi, wakurra=lu wirrirl=ya-ni wuruju ma=lu. go-pst=1pl.excl.s around neg=3pl.s back=go-pst good top.aux=3pl.s puliki minya. cattle this ‘We looped around, (but) they hadn’t gone back, the cattle had settled there, all ok.’ 57.

Tumaj=lu mangarri wuruju ngu nga-rni. conj=3pl.s veg_food good seq.aux eat-pst ‘Since they had good feed there, they were grazing.’

738 

 Appendix: Narrative texts

58. Parrayari-purupuru jalyirr. gidgee-many leaf ‘Lots of pararri foliage.’ 59. Mayarra manja=lu ma-nani. again scrub=3pl.s get-ipfv.pst ‘They were getting mulga scrub again.’ 60. Wurrkal wirrpa ma. green many top.aux ‘Lots of green pastures/growth.’ 61. Ngurra-kurra wirrirl. camp-all back ‘Back to camp then!’ 62. Karri-nyani=rnalu wirrirl be-ipfv.pst=1pl.excl.s back ‘We went back to that campsite.’

yalu-ngka=lku. that-loc=then

63. Patik nawu=rnalu yirra-rnani, hawij-ta. paddock now=1pl.excl.s place-ipfv.pst house-loc ‘We subsequently built a yard, by the house.’ 64. Hawij-wangu-rla, wurra=lku=rnalu. house-priv-loc wait=still=1pl.excl.s ‘(Actually), no house then, we were still there.’ 65. Yirra-rnani patik, lurrij=pa=rnalu yirra-rni. place-ipfv.pst paddock finish=ep=1pl.excl.s place-pst ‘(We) put the paddock in, we finished erecting (the paddock).’ 66. Yalu-jangka pirri=ya-ni nganayi nawu, kardiya, Frank Fitler, that-ela arrive=go-pst whatist now white_man name yangi rod tren. one road train ‘After that, whatsisname arrived, a whitefella, Frank Fitler, (driving) one road train.’ 67.

Hawij jungarni nyampa-piya. house true what-sembl ‘Yeah, something a bit like a house.’

Text 3: Mungkururrpa 

 739

68. Yangka wurdumardi hawij nyampa-piya. like automatic house what-sembl ‘You know, a prefabricated house, something like that. 69. Ka-nya=yi yirra-rni yalu-ngka. carry-pst=emph place-pst that-loc ‘He brought it and put it there.’ 70. Wal jangu wakurra=rnalu mayarra nganayi Mijin-kurlu-rlu well recog2 neg=1pl.excl.s again whatist place_name-prop-erg pangi-rnani walya taka-ngku=rnalu kurrupardu baa-kurlu-rlu, dig-ipfv.pst ground hand-erg=1pl.excl.s boomerang bar-prop-erg patik-ku-purupuru, yart-ku. paddock-dat-many yard-dat ‘Well like, we didn’t dig the site like we did with the Mission perviously, we dug the ground by hand with boomerangs, and crowbars for all of the paddocks too, and the yard.’ 71. Yuwayi, karri-nya yalu-rlu yalu-ngka=lku. yeah be-pst that-erg that-loc=then ‘Yeah, that one stayed there then.’ 72. Werak ngantu=wali nga=lu werak=pa-nani work recog=then seq.aux=3pl.s work=get-ipfv.pst puliki=rnalu marda-rnani, wan yia. cattle=1pl.excl.s have-ipfv.pst one year ‘There they worked there then, we worked and looked after the cattle, for one year.’ 73. Nganayi, yalu-rla=lku=rnalu whatsit that-loc=then=1pl.excls ‘Y’know, we kept working there.’

werak=ma-nani. work=get-ipfv.pst

74. Kardiya Joe Mahood mangarri wuruju. white_man name veg_food good ‘The kardiya Joe Mahood had good food.’ 75. An mijij=ka=wali pirri=ya-ni nyanungu-ku, wayif, conj missus-loc=then arrive=go-pst recog-dat wife ‘And then he came in with his wife, Marie Mahood.’

Marie Mahood. name

740 

 Appendix: Narrative texts

76. An nyanungu-ku daughter, Kim Mahood and Bobby Mahood, Bob. conj recog-dat daughter name conj name name ‘And his daughter, Kim Mahood and Bobby Mahood, Bob.’ 77.

Nyanungu-ku san Kim Hood-ku, bratha ita. recog-dat son name-dat brother small ‘His son, little brother for Kim Mahood.’

78. Karri-nya=rnalu yalu-ngka=lku. be-pst=1pl.excl.s that-loc=then ‘We were still staying there.’ 79. Puliki wirrpa karri-nya=wali. cattle many be-pst=then ‘There were lots of cattle at that time.’ 80. Wal ngantany, alidei-jangka, mani=yanu nga-– pawungurru=wali well man holiday-ela money=3pl.o seq.aux many=then yangi-yangi ngarrja=lku. one-rdp more=then ‘Well, Aboriginal men and women, from holidays, each and every (one) then had more money at that time.’ 81. Ya-nu ma-nani kalarra, ngantany. go-pst get-ipfv.pst west man ‘Aboriginal men and women went west and would get it.’ 82. Balgo-ngka, Molly Brown wurra=wali wirrpa-lu werak=ma-nani. place_name-loc name still=then many=3pl.s work=get-ipfv.pst ‘Molly Brown in Balgo, they were all still working.’ 83. Minya Malarn-ta, mei mayit, yangi-kariny Rexon. this place_name-loc my mate one-other name ‘That one in Malarn, my mate, another one, Rexon.’ 84. Ngaju-ku mayit-kariny yangi, pinarri, Harry Mahood. 1sg-dat mate-other one knowledge name ‘Another mate of mine, I know him, Harry Mahood.’ 85. Yala-rla=lu Mongrel Down-ta karri-nyani yalu-wardingki nawu. that-loc=3pl.s place_name-loc be-ipfv.pst that-denizen now ‘They were staying there at Mongrel Downs for a long time.’

Text 3: Mungkururrpa 

 741

86. Ngaju ma=rna wirrirl=ya-nku Kururrungku-kurra=wali. 1sg top.aux=1sg.s back=go-pot place_name-all=then ‘I will go back to Kururrungku then.’ 87.

Tumaji Nakarra-rlu parli=pi-nya. conj ♀ subsection-erg find=hit-pst ‘Because Nakarra was looking for me.’

88. Wirriri ya-ni=rna ngurrara-kurra—. back go-pst=1sg.s country-all ‘I went back to that country.’ 89. Jangu=rnalu puliki jilykarra=la-nani recog2=1pl.excl.s cattle brand=pierce-ipfv.pst ‘You know, we were branding cattle in that place.’

yalu-ngka. that-loc

90. Niyutayim, puliki-wangu-rla, puuj purangu-kurlu-rlu, wakurra new_time cattle-priv-loc bush hot-prop-erg neg karri-nyani waku. be-ipfv.pst neg ‘At that time without cattle, hot time, there was no yard, nothing.’ 91. Karri-nyani=rnalu lurrij, puliki wirrpa karri-nya=wali. be-ipfv.pst=1pl.excl.s finish cattle many be-pst=then ‘We stopped there, finish, there were lots of cattle then.’ 92. Jalangu puliki wirrpa nawu, tumaj. today cattle many now too_much ‘Now there are loads of cattle now, so many.’ 93. Ngalipa-kuny, nganimpa-kuny, Kururrungku-jangka. 1pl.incl-poss 1pl.excl-poss place_name-ela ‘All of ours, our mob’s, from Kururrungku (Billiluna).’ 94. Jampala nawu=lu yi-nya kayirra, nganayi, Bob Shepherd. some now=3pl.s give-pst north whatsit name ‘Some now they sent north, (to) whatsisname, Bob Shepherd.’ 95. Puliki nga=lu=rla yi-nya, cattle seq.aux=3pl.s=3sg.obl give-pst ‘They sent the cattle to him, one half.’

half-kariny. half-other

yard yard

742 

 Appendix: Narrative texts

96.

Yuwayi, wuruju. yes good ‘Yeah, good.’

97.

Wakurra=rnalu alidei=wali wirrirl neg=1pl.excl.s holiday=then back ‘We didn’t go back then on holiday.’

98.

Ngaju ma=rna ngaringka-kurlu karri-nya=wali Nakarra-kurlu. ∮+subsection-prop 1sg top.aux=1sg.s woman-prop be-pst=then ‘I stayed then with Nakarra.’

99.

Ya-ni=rlijarra Nguriny-kurra-wali. go-pst=1du.excl.s place_name-all=then ‘The two of us went to Nguriny (Carranya Station) then.’

wirrirl back

ya-ni. go-pst

100. Nguriny, jaku=rlijarra wirrirl ya-ni, Kururrungku-kurra. place_name depart=1du.excl.s back go-pst place_name-all ‘Carranya Station, we two set off and went back, to Kururrungku (Billiluna Station).’ 101.

Yala wurlkaman Brenna Brown nya-nya Kururrungku-ku=wali. that old_woman name see-pst place_name-dat=then ‘That old woman, Brenna Brown looked after Kururrungku.’

102.

Ol man langa Joe Mahoood manija. oldman loc.prep name manager ‘Old man with manager Joe Mahood.’

103.

Pinij. finish ‘All finished.’

104. Nyurnu=karri-nya=pula. sick=be-pst=3du.s ‘Those two passed away.’ 105.

Olgaman xxx Kim Mahood-ku, ngati-nyanu. old_woman xxx name-dat M-anaph ‘Old woman for Kim Mahood, her mother.’

Text 3: Mungkururrpa 

 743

106. Ngaka yalu-jangka pathawan, yawiyi, ol Joe Mahood later that-ela father_one sorry old name ‘Later, sometime after, the father, poor thing, old Joe Mahood.’ 107.

Yuwayi, pinarri=rna=yanu ngaju holot-ku pamili-ku, manija-ku, yes knowledge=1sg.s=3pl.o 1sg all-dat family-dat manager-dat wayif-pa-ku. wife-ep-dat ‘Yeah, I know all of the family, the manager, the wife.’

108. Dota an son-ku=rna=yanu daughter conj son-dat=1sg.s=3pl.o ‘I know their daughter and son.’ 109.

Yuwayi, wuruju. yes good ‘Yeah, that’s all.’

pinarri. knowledge

References Aikhenvald, Alexandra. (2006). Serial verb constructions in a typological perspective. In Alexandra Aikhenvald & Robert Dixon (Eds.), Serial verb constructions: A cross-linguistic typology (pp. 1–87). Oxford: Oxford University Press. Aikhenvald, Alexandra. (2010). Imperatives and commands. Oxford: Oxford University Press. Aikhenvald, Alexandra & Dixon, Robert (Eds.). (2006). Serial verb constructions: A cross-linguistic typology. Oxford: Oxford University Press. Alpher, Barry. (1990). Some Proto-Pama-Nyungan paradigms: A verb in the hand is worth two in the phylum. In Geoffrey O’Grady & Darrell Tryon (Eds.), Studies in Comparative Pama-Nyungan (pp. 155–171). Canberra: Pacific Linguistics. Amberber, Mengistu, Baker, Brett & Harvey, Mark. (2010). Complex predicates: Cross-linguistic perspectives on event structure. Cambridge/New York: Cambridge University Press. Andersen, Gisle & Fretheim, Thornstein. (2000). Pragmatic markers and propositional attitude. Amsterdam: John Benjamins. Andrews, Daisy. (1996). I’m gone for good. In Kimberley Language and Resource Centre (Ed.), Moola Bulla: In the shadow of the mountain (pp. 148–157). Broome: Magabala Books. Aronoff, Mark. (1992). Morphology now. Albany: SUNY Press. Austin, Peter. (1981a). A grammar of Diyari, South Australia. Cambridge/New York: Cambridge University Press. Austin, Peter. (1981b). Switch-reference in Australia. Language, 57(2), 309–334. https://doi. org/10.2307/413693 Austin, Peter. (1982). Transitivity and cognate objects in Australian languages. In Paul Hopper & Sandra Thompson (Eds.), Syntax and Semantics (pp. 37–47). New York: Academic Press. Austin, Peter. (1988). Complex sentence constructions in Australian languages. Amsterdam: John Benjamins. Austin, Peter. (1989). A reference grammar of the Mantharta languages, Western Australia. Unpublished manuscript. La Trobe University. Austin, Peter. (1995). Double case marking in Kanyara and Mantharta languages, Western Australia. In Frans Plank (Ed.), Double case: Agreement by Suffixaufnahme (pp. 363–379). Oxford: Oxford University Press. Austin, Peter. (1997). Causative and applicative constructions in Australian Aboriginal languages. In Kazuo Matsumura & Hayasi Syobo (Eds.), The dative and related phenomena (pp. 165–225). Tokyo: Hituzi Syobo. Austin, Peter. (1998). Argument coding and clause linkage in Australian Aboriginal languages. Unpublished manuscript. La Trobe University. Austin, Peter & Bresnan, Joan. (1996). Non-configurationality in Australian Aboriginal languages. Natural Language & Linguistic Theory, 14(2), 215–268. Bach, Emmon. (1981). On time, tense, and aspect: An essay in English metaphysics. In Peter Cole (Ed.), Radical Pragmatics (pp. 63–81). New York: Academic Press. Baker, Brett. (2014). Word structure in Australian languages. In Harold Koch & Rachel Nordlinger (Eds.), The languages and linguistics of Australia: A comprehensive guide (pp. 139–213). Berlin/Boston: De Gruyter Mouton. Baker, Brett. (2005). The domain of phonological processes. In Ilana Mushin (Ed.), Proceedings of the 2004 Conference of the Australian Linguistics Society. University of Sydney. http://hdl. handle.net/2123/112 Baker, Brett & Harvey, Mark. (2003). Word structure in Australian languages. Australian Journal of Linguistics, 23(1), 3–33. https://doi.org/10.1515/9783110752434-013

746 

 References

Baker, Brett & Harvey, Mark. (2010). Complex predicate formation. In Mengistu. Amberber, Brett Baker, & Mark Harvey (Eds.), Complex predicates: Cross-linguistic perspectives on event structure (pp. 13–47). Cambridge/New York: Cambridge University Press. Baker, Mark. (1996). The polysynthesis parameter. New York: Oxford University Press. Bárány, András, Bond, Oliver & Nikolaeva, Irina. (2019). Prominent internal possessors. Oxford: Oxford University Press. https://www.soas.ac.uk/linguistics/prominent-possessors/workshop/ Barry, William. (1997). Another R-tickle. Journal of the International Phonetic Association, 27(1–2), 35–45. https://doi.org/10.1017/S0025100300005405 Bavin, Edith, & Shopen, Tim. (1985). Warlpiri and English: Language in contact. In Michael Clyne (Ed.), Australia, meeting place of languages (pp. 81–94). Canberra: Pacific Linguistics. Beard, John Stanley & Webb, Martyn Jack. (1974). Vegetation survey of Western Australia: Great Sandy Desert. Perth: University of Western Australia Press. Belkadi, Aicha. (2016). Associated motion constructions in African languages, Africana Linguistica, 22, 43–70, Berndt, Catherine. (1965). Woman and the ‘secret life’. In Catherine Berndt & Ronald Berndt (Eds.), Aboriginal man in Australia (pp. 238–282). Sydney: Angus and Robertson. Berndt, Ronald. (1970). Traditional morality as expressed through the medium of an Australian Aboriginal religion. In Ronald Berndt (Ed.), Australian Aboriginal anthropology: Modern studies in the social anthropology of the Australian Aborigines (pp. 216–247). Nedlands: Western Australia University Press. Berndt, Ronald. (1972). The Walmadjeri and Gugadja. In Marco Bicchieri (Ed.), Hunters and gatherers today: A socioeconomic study of eleven such cultures in the twentieth century (pp. 177–217). New York: Holt, Rinehart & Winston. Berndt, Ronald & Berndt, Catherine. (1988). The world of the first Australians: Aboriginal traditional life: Past and present. Canberra: Aboriginal Studies Press. Berndt, Ronald & Berndt, Catherine. (1960). Survey of the Balgo Hills area (Southern Kimberley, Western Australia. Unpublished report. Berndt, Ronald & Berndt, Catherine. (1987). End of an era: Aboriginal labour in the Northern Territory. Canberra: Australian Institute of Aboriginal Studies. Bhat, Dharbe. (2004). Pronouns. Oxford: Oxford University Press. Bhatt, Rajesh & Pancheva, Roumyana. (2006). Conditionals. In Martin Everaert & Hank van Riemsdijk (Eds), The Blackwell companion to syntax (pp. 638–687). Malden/Oxford: John Wiley & Sons. https://doi.org/10.1002/9780470996591.ch16 Binnick, Robert. (2012). Oxford handbook of tense and aspect. Cambridge: Cambridge University Press. Bird, Rebecca., Bird, Douglas, Codding, Brian, Parker, Christopher & Jones, James. (2008). The ‘fire stick farming’ hypothesis: Australian Aboriginal foraging strategies, biodiversity, and anthropogenic fire mosaics. PNAS, 105(39), 14796–14801. Bittner, Maria & Hale, Kenneth. (1995). Remarks on definiteness in Warlpiri. In Emmon Bach, Eloise Jelinek, Angelika Kratzer & Barbara Partee (Eds) Quantification in Natural Languages (Vol. 64, pp. 81–105). Dordrecht: Springer. https://doi.org/10.1007/978-94-017-2817-1_5 Blake, Barry. (1977). Case marking in Australian languages. Canberra: Australian Institute of Aboriginal Studies. Blake, Barry. (1979a). A Kalkatungu grammar. Canberra: Pacific Linguistics. Blake, Barry. (1979b). Pitta-Pitta. In Robert Dixon & Barry Blake (Eds), Handbook of Australian languages (Vol. 1, pp. 183–242). Canberra: Australian National University Press. Blake, Barry. (1983). Structure and word order in Kalkatungu: The anatomy of a flat language. Australian Journal of Linguistics, 3(2), 143–175. https://doi.org/10.1080/07268608308599307 Blake, Barry. (1985). Case markers, case and grammatical relations: An addendum to Goddard. Australian Journal of Linguistics, 5(1), 81–84.

References 

 747

Blake, Barry. (1987). Australian Aboriginal Grammar. London: Croom Helm. Blake, Barry. (1990). Relational Grammar. New York: Routledge. Blake, David & Yeates, Anthony. (1976). 1976, Stansmore, Western Australia (1:250 000 Geological Series Explanatory Notes, p. 20). Bureau of Mineral Resources and Geological Survey of Western Australia. Boersma, Paul & Weenik, David. (2019). Praat: Doing phonetics by computer (Version 6.1.06). Computer software. http://www.praat.org/ Bolger, Audrey. (1987). Wangkajunga women: Stories from the desert. Aboriginal History, 11, 102–115. Bolton, B. & Latz, Peter. (1978). The western hare-wallaby Lagorchestes hirsutus in the Tanami Desert. Australian Wildlife Research, 5, 285–293. Bolton, Geoffrey. (1958). Alexander Forrest, his life and times. Melbourne: Melbourne University Press. Botne, Robert. (2012). Remoteness distinctions. In Robert Binnick (Ed.), Oxford handbook of tense and aspect (pp. 536–562). Oxford: Oxford University Press. Bowe, Heather. (1990). Categories, constituents, and constituent order in Pitjantjatjara: An Aboriginal language of Australia. London: Routledge. Bowern, Claire. (2006). Australian complex predicates. Annual Meeting of the Berkeley Linguistics Society, 32(2), 17–33. Bowern, Claire. (2012). A grammar of Bardi. Berlin/Boston: De Gruyter Mouton. Bowern, Claire. (2014). Complex predicates in Australian languages. In Harold Koch & Rachel Nordlinger (Eds.), The languages and linguistics of Australia: A comprehensive guide (pp. 263–294). Berlin/Boston: De Gruyter Mouton. Bowern, Claire & Atkinson, Quentin. (2012). Computational phylogenetics and the internal structure of Pama-Nyungan. Language, 88(4), 817–845. https://doi.org/10.1353/lan.2012.0081 Bowler, Margit. (2016). The status of degrees in Warlpiri. Proceedings of The Semantics of African, Asian and Austronesian Languages, 2, 1–17. Breen, Gavin & Dobson, Veronica. (2005). Central Arrernte. Journal of the International Phonetic Association, 35(2), 249–254. Bresnan, Joan. (1982). Control and complementation. Linguistic Inquiry, 13(3), 343–434. Browne, Mitchell. (2021). A grammatical description of Warlmanpa, a Ngumpin-Yapa language spoken around Tennant Creek (Northern Territory). PhD dissertation. Brisbane: University of Queensland. Browne, Mitchell. (2020). Contrast and retroactive implicatures: An analysis of =lku ‘now, then’ in Warlpiri and Warlmanpa. Australian Journal of Linguistics. Burgman, Albert. (2003). Warnman sketch grammar. South Headland: Wangka Maya Pilbara Aboriginal Language Centre. Burridge, Kate. (1996). Yulparija sketch grammar. In William McGregor (Ed.), Studies in Kimberley languages in honour of Howard Coate (pp. 15–69). Munich: Lincom Europa. Burrows, Neil. (2018). Feral animals in the semi-arid and arid regions of Australia: Origins, impacts and control. In Hans Lambers (Ed.), On the ecology of Australia’s arid zone (pp. 331–373). Cham: Springer International. https://doi.org/10.1007/978-3-319-93943-8_13 Burrows, Neil, Burbidge, Andrew, Fuller, Phillip, & Behn, Graeme. (2006). Evidence of altered fire regimes in the Western Desert region of Australia. Conservation Science Western Australia, 5(3), 272–284. Burrows, Neil & Christensen, P. E. S. (1990). A survey of Aboriginal fire patterns in the Western Desert of Australia. In Stephen Nodvin & Thomas Waldrop (Eds.), Fire and the environment: Ecological and cultural perspectives (pp. 20–24). Asheville: U.S. Dept of Agriculture and Forest Service. Butcher, Andrew. (1995). The phonetics of neutralisation: The case of Australian coronals. Studies in general and English phonetics: Essays in honour of Professor J. D. O’Connor. London: Routledge, 10–38.

748 

 References

Butcher, Andrew. (1999). What speakers of Australian Aboriginal languages do with their velums and why: The phonetics of the nasal/oral contrast. Proceedings of the International Congress of Phonetic Sciences 1999. ICPhS99, San Francisco. Butcher, Andrew & Harrington, John. (2003). An acoustic and articulatory analysis of focus and the word/morpheme boundary distinction in Warlpiri. In Sallyanne Palethorpe & Marija Tabain (Eds.), Proceedings of the 6th International Seminar on Speech Production (pp. 19–24). Sydney: Macquarie University. Butcher, Andrew & Loakes, Debbie. (2008). Enhancing the left edge: The phonetics of pre-stopped sonorants in Australian languages. The Journal of the Acoustical Society of America, 124, 2527. https://doi.org/10.1121/1.4782973 Butt, Miriam. (1995). Complex predicates in Urdu. Stanford: CSLI Publications. Butt, Miriam. (2010). Light verb jungle: Still hacking away. In Mengistu Amberber, Brett Baker & Mark Harvey (Eds.), Complex predicates: cross-linguistic perspectives on event structure (pp. 48–78). Cambridge/New York: Cambridge University Press. Butt, Miriam & Geuder, Willem. (2001). On the (semi) lexical status of light verbs. In Henk van Riemsdijk (Ed.), Semi-lexical categories: The function of content words and the content of function words (pp. 323–369). Berlin: De Gruyter Mouton. Bybee, Joan. (1985). Morphology. A study of the relation between meaning and form. Amsterdam: Benjamins. Bybee, Joan. (1998). ‘Irrealis’ as a grammatical category. Anthropological Linguistics, 40, 257–271. Bybee, Joan & Fleischman, Suzanne. (1995). Modality in grammar and discourse: An introductory essay. In J.oan Bybee & Suzanne Fleischman (Eds), Modality in grammar and discourse (pp. 1–14). Amsterdam: John Benjamins. Bybee, Joan, Pagliuca, William, Perkins, Revere, Traugott, Elizabeth & Heine, Bernd. (1991). Approaches to grammaticalization. Amsterdam: John Benjamins. Bybee, Joan, Perkins, Revere & Pagliuca, William. (1994). The evolution of grammar: Tense, aspect and modality in the languages of the world. Chicago: University of Chicago Press. Byrne, Francis. (1989). A hard road. Brother Frank Nissl, 1888–1980: A life of service to the Aborigines of the Kimberleys. Nedlands: Tara House. Cane, Scott. (1984). Desert camps. A case study of stone artifacts and Aboriginal behaviour in the Western Desert. PhD dissertation. Canberra: Australian National University. Cane, Scott. (1987). Australian Aboriginal subsistence in the Western Desert. Human Ecology, 15, 391–493. Cane, Scott. (2006). Countrymen: An ethnography of the Ngurrurpa Native Title claim. A connection report to Ngaanyatjarra Council and the Office of Native Title, Department of Premier and Cabinet. Government of Western Australia. Cane, Scott. (2016). People and policy in the development and destruction of Yagga Yagga outstation, Western Australia. In Nicholas Peterson & Fred Myers (Eds), Experiments in self-determination: Histories of the outstation movement in Australia (pp. 253–278). Canberra: ANU Press. Cane, Scott & Novak, V. (1981). Archaeological survey of EP 134 West of Stansmore Range, Western Australia. https://trove.nla.gov.au/version/41562207. Capell, Arthur. (1940a). The classification of languages in north and north-west Australia. Oceania, 10(3), 241–272. Capell, Arthur. (1940b). The classification of languages in north and north-west Australia (continued). Oceania, 10(4), 404–433. Capell, Arthur. (1956). A new approach to Australian linguistics (Handbook of Australian Languages, Part 1). Oceania Linguistic Monographs 1. University of Sydney. Capell, Arthur. (1962) Some linguistic types in Australia. Oceania Linguistic Monographs 7. University of Sydney.

References 

 749

Capell, Arthur. (1963) Linguistic survey of Australia: Prepared for the Australian Institute of Aboriginal Studies. University of Sydney. Carnegie, David. (1898). Spinifex and sand: A narrative of five years pioneering and exploration in Western Australia. London: C. Arthur Pearson. Carty, John. (2011). Abstraction, economics and the social life of style in Balgo Art. PhD dissertation. Canberra: Australian National University. Carty, John, Davenport, Carly & Lafontaine, Monique. (2010). Yiwarra Kuju: The Canning Stock Route. Canberra: National Museum of Australia Press. Cataldi, Lee. (2011). A dictionary of Ngardi. Balgo: Balgo School. Chappell, Hillary & McGregor, William. (1995). The grammar of inalienability: A typological perspective on body part terms and the part-whole relation. Berlin: De Gruyter Mouton. Clendon, Mark. (1988). Some features of Manjiljarra nominalised relative clauses. In Peter Austin (Ed.), Complex sentence constructions in Australian languages (pp. 193–204). Amsterdam: John Benjamins. Comrie, Bernard. (1985). Tense. Cambridge: Cambridge University Press. Comrie, Bernard. (1986). Conditionals: A typology. In Elizabeth Traugott, Alice Meulen, Judy Reilly & Charles Ferguson (Eds), On conditionals (pp. 77–100). Cambridge: Cambridge University Press. Comrie, Bernard. (2003) When agreement gets trigger-happy. Transactions of the Philological Society. 101(2), 323–337. https://doi.org/10.1111/1467-968X.00121 Comrie, Bernard, Haspelmath, Martin & Bickel, Balthasar. (2008). Leipzig glossing rules. https:// www.eva.mpg.de/lingua/resources/glossing-rules.php Corbett, Greville. (2000). Number. Cambridge: Cambridge University Press. Corbett, Greville. (2006). Agreement. Cambridge: Cambridge University Press. Corbett, Greville. (2017). Heads – a canonical approach. Workshop on Prominent Internal Possessors, SOAS, University of London. https://www.soas.ac.uk/linguistics /prominentpossessors/workshop/ Cristofaro, Sonia. (2006). The organization of reference grammars: A typologist user’s point of view. In Felix Ameka, Alan Dench & Nicholas Evans (Eds), Catching language: The standing challenge of grammar writing (pp. 137–170). Berlin: De Gruyter Mouton. Crowley, Terry. (2002). Serial verbs in Oceanic: A descriptive typology. Oxford: Oxford University Press. Cuenca, Maria-Josep. (2008). Pragmatic markers in contrast: The case of well. Journal of Pragmatics, 40(8), 1373–1391. https://doi.org/10.1016/j.pragma.2008.02.013 Dahl, Östen. (2000). The grammar of future time reference in European languages. In Östen Dahl (Ed.) Tense and aspect in the languages of Europe (Vol. 6, pp. 309–328). Cambridge: De Gruyter Mouton. De Haan, Ferdinand. (2006). Typological approaches to modality. In William Frawley (Ed.), The expression of modality (pp. 27–70). Cambridge: Mouton De Gruyter. Dé Ishtar, Zohl. (2005). Holding Yalyulyu: White culture and black women’s law. Melbourne: Spinifex Press. Deegan, Bonnie, Sturt, Barbara, Ryder, Doris, Butcher, May, Brumby, Stan, Long, Gracie, Badngarri, Nora, Lannigan, Jack, Blythe, Joe & Wightman, Glenn. (2010). Jaru plants and animals: Aboriginal flora and fauna knowledge from the south-east Kimberley and western Top End, north Australia. Halls Creek/Palmerston: Kimberley Language Resource Centre and Dept. of Natural Resources, Environment, the Arts and Sport. Dench, Alan. (1988). Complex sentences in Martuthunira. In Peter Austin (Ed.), Complex sentence constructions in Australian languages (pp. 97–139). Amsterdam: John Benjamins. Dench, Alan. (1991). Panyjima. In Robert Dixon & Barry Blake (Eds.), Handbook of Australian languages 4 (pp. 124–243). Oxford: Oxford University Press.

750 

 References

Dench, Alan. (1995) Martuthunira: A language of the Pilbara region of Western Australia. Canberra: Pacific Linguistics. Dench, Alan. (2006). Case marking strategies in subordinate clauses in Pilbara languages – Some diachronic speculations. Australian Journal of Linguistics, 26(1), 81–105. https://doi. org/10.1080/07268600500531644 Dench, Alan. (2007). Demonstrative paradigm splitting in the Pilbara languages of Western Australia. In Joseph Salmons & Shannon Dubenion-Smith Historical linguistics 2005: Selected papers from the 17th International Conference on Historical Linguistics (pp. 223–237). Amsterdam: John Benjamins. Dench, Alan & Evans, Nicholas. (1988). Multiple case-marking in Australian languages. Australian Journal of Linguistics, 8(1), 1–47. Derrick, Donald & Gick, Bryan. (2011). Individual variation in English flaps and taps: A case of categorical phonetics. Canadian Journal of Linguistics, 56(3), 307–319. https://doi. org/10.1017/S0008413100002024 Diessel, Holger. (2001) The ordering and distribution of main and adverbial clauses. Language, 77, 343–365. Diessel, Holger. (1999). Demonstratives: Form, function and grammaticalization. Amsterdam: John Benjamins. Diessel, Holger. (2005). Competing motivations for the ordering of main and adverbial clauses. Linguistics, 43(3), 449–470. Dixon, Robert. (Ed.). (1976). Grammatical categories in Australian languages. Canberra: Australian Institute of Aboriginal Studies. Dixon, Robert. (1980). The languages of Australia. Cambridge: Cambridge University Press. Dixon, Robert. (1994). Ergativity. Cambridge: Cambridge University Press. Dixon, Robert. (1997). The rise and fall of languages. Cambridge: Cambridge University Press. Dixon, Robert. (2002). Australian languages: Their nature and development. Cambridge: Cambridge University Press. Dixon, Robert. (2011). Serial verb constructions in Dyirbal. Anthropological Linguistics, 53(3), 185–214. Dixon, Robert & Blake, Barry. (1983) Handbook of Australian Languages. Canberra: Australian National University Press Dixon, Sally (2010). Juwaliny: Dialectal variation and ethnolinguistic identity in the Great Sandy Desert. In Brett Baker, Rod Gardner, Mark Harvy & Ilana Mushin (Eds.) Indigenous language and social identity: Papers in honour of Michael Walsh (pp. 67–78). Canberra: Australian National University. Dousset, Laurent, Koch, Harold & McConvell, Patrick. (2015). AustKin database on Australian Aboriginal kinship systems and systems of social organisation. http://www.austkin.net/index. php?loc=people Dowty, David. (1989). Thematic proto-roles and argument selection. Language, 67(3), 547–619. Dowty, David. (2000). The dual analysis of adjuncts/complements in categorial grammar. ZAS Papers in Linguistics, 17, 53–78. Dryer, Matthew. (1986). Primary objects, secondary objects, and antidative. Language, 62(4), 808–845. https://doi.org/10.2307/415173 Ennever, Thomas. (2014). Stop lenition in Gurindji: An acoustic phonetic study. Honours thesis. Brisbane: University of Queensland. Ennever, Thomas (2018). Nominal and pronominal morphology of Ngardi: A Ngumpin-Yapa language of Western Australia. MPhil thesis. Brisbane: University of Queensland. Ennever, Thomas, Meakins, Felicity & Round, Erich. (2017). A replicable acoustic measure of lenition and the nature of variability in Gurindji stops. Laboratory Phonology: Journal of the Association for Laboratory Phonology, 8(1). https://doi.org/10.5334/labphon.18

References 

 751

Erman, Britt. (2001). Pragmatic markers revisited with a focus on you know in adult and adolescent talk. Journal of Pragmatics, 33(9), 1337–1359. https://doi.org/10.1016/S0378-2166(00)00066-7 Evans, Nicholas. (1989). Morphological ergativity and fixed-transitivity phenomena in Australian languages. Paper presented at conference on non-standard case, Tucson, Arizona. Evans, Nicholas. (1995). A grammar of Kayardild: With historical-comparative notes on Tangkic (Vol. 15). Berlin: Walter de Gruyter. Evans, Nicholas. (2003). Bininj Gun-wok: A pandialectal grammar of Mayali, Kunwinjku and Kune. Pacific Linguistics. http://hdl.handle.net/1885/53188 Evans, Nicholas. (2007). Insubordination and its uses. In Irina Nicolaeva (Ed.), Finiteness: Theoretical and empirical foundations (pp. 366–431). Oxford: Oxford University Press. Evans, Nicholas & Watanabe, Honoré. (2016). Insubordination. Amsterdam: John Benjamins. https:// benjamins.com/catalog/tsl.115 Fauconnier, Stefanie & Verstraete, Jean-Christophe. (2010). Distinguishing animacy effects for agents: A case study of Australian languages. Australian Journal of Linguistics, 30, 183–207. https://doi.org/10.1080/07268601003678619 Filip, Hana. (1999). Aspect, eventuality types and nominal reference. London: Routledge. Fletcher, Janet & Butcher, Andrew. (2014). Sound patterns of Australian languages. In Harold Koch & Rachel Nordlinger (Eds.), The languages and linguistics of Australia: A comprehensive guide (pp. 91–138). Berlin/Boston: De Gruyter Mouton. Forker, Diana. (2010). Nonlocal uses of local cases in the Tsezic languages. Linguistics, 48(5), 1083–1109. Gaby, Alice. (2017). A Grammar of Kuuk Thaayorre (Vol. 74). Berlin/Boston: De Gruyter Mouton. Ganter, Regina. (2016). Balgo (1940–1965). In Regina Ganter (Ed.), German missionaries in Australia – A web-directory of intercultural encounters. http://missionaries.griffith.edu.au/ Gard, Ronele & Gard, Eric. (1990). Canning Stock Route. A traveller’s guide for a journey through history. Wembley Downs: Western Desert Guides. Gasser, Emily & Bowern, Claire. (2014). Revisiting phonotactic generalizations in Australian languages. In John. Kingston, Claire Moore-Cantwell, Joe Pater & Robert Staubs (Eds.), Supplemental Proceedings of Phonology 2013. Linguistic Society of America. Goddard, Cliff. (1982). Case systems and case marking in Australian languages: A new interpretation. Australian Journal of Linguistics, 2(2), 167–196. https://doi. org/10.1080/07268608208599290 Goddard, Cliff. (1985). A grammar of Yankunytjatjara. Alice Springs: Institute for Aboriginal Development. Goddard, Cliff. (1988). Verb serialisation and the circumstantial construction in Yankunytjatjara. In Peter Austin (Ed.), Complex sentence constructions in Australian languages (pp. 177–192). Amsterdam: John Benjamins. Green, Tom. (1988). Ngardily wordlist (with preliminary grammatical notes and text). https://aiatsis. gov.au/sites/default/files/catalogue_resources/0034_Access.zip Guillaume, Antoine. (2009). Les suffixes verbaux de ‘mouvement associé’ en cavineña. Faits de Langues, 34(2), 181–204. Guillaume, Antoine. (2016). Associated motion in South America: Typological and areal perspectives. Linguistic Typology, 20(1), 81–177. https://doi.org/10.1515/lingty-2016-0003 Guillaume, Antoine. (2017). Associated Motion: Australia, South America and beyond. Paper presented at the 12th Conference of the Association for Linguistic Typology, Canberra. Guillaume, Antoine & Koch, Harold. (In press) Introduction: Associated Motion as a grammatical category in linguistic typology. In Antoine Guillaume & Harold Koch (Eds.), Associated Motion, Berlin/Boston: De Gruyter Mouton.

752 

 References

Haiman, John & Munro, Pamela. (1981). Switch-reference and universal grammar. Amsterdam: John Benjamins. Hale, Kenneth. (1973). Person marking in Walbiri. In Steven Anderson & Paul Kiparsky (Eds), Festschrift for Morris Halle (pp. 308–344). New York: Holt, Rinehart & Winston. Hale, Kenneth. (1976). The adjoined relative clause in Australia. In Robert Dixon (Ed.), Grammatical categories in Australian languages (pp. 78–105). Canberra: Australian Institute of Aboriginal Studies. Hale, Kenneth. (1977). Elementary remarks on Walbiri orthography, phonology and allomorphy. Unpublished manuscript. Cambridge: Massachusetts Institute of Technology. Hale, Kenneth. (1981). Preliminary remarks on the grammar of part-whole relations in Warlpiri. In Jim Hollyman & Andrew Pawley (Eds.), Studies in Pacific languages & cultures in Honour of Bruce Briggs (pp. 334–344). Linguistic Society of New Zealand. Hale, Kenneth. (1982). Some essential features of Warlpiri verbal clauses. In Stephen Swartz (Ed.), Papers in Warlpiri grammar: In memory of Lothar Jagst (pp. 217–315). Summer Institute of Linguistics. Hale, Kenneth. (1983). Warlpiri and the grammar of non-configurational languages. Natural Language & Linguistic Theory, 1(1), 5–47. https://doi.org/10.1007/BF00210374 Hale, Kenneth, Laughren, Mary & Simpson, Jane. (1995). Warlpiri syntax. In Joachim Jacobs, Arnim von Stechow, Wolfgang Sternefeld, & Theo Venneman (Eds.), Syntax. An international handbook of contemporary research (Vol. 2, pp. 1430–1451). Walter de Gruyter. Hale Kenneth, Laughren, Mary & Simpson, Jane (2015) Warlpiri. In Tibor Kiss & Artemis Alexiadou (Eds.), Syntax – Theory and and analysis (Vol. 3, pp. 1677–1709). De Gruyter Hamilton, Amanda & Ennever, Thomas. (2017, December 13). Possibilities a-bound: A typological study of Kriol verb borrowings into Ngumpin-Yapa languages. 12th Conference of the Association for Linguistic Typology, Canberra, ACT. Hamilton, Phillip. (1996). Phonetic constraints and markedness in the phonotactics of Australian Aboriginal languages. PhD dissertation. Toronto: University of Toronto. Hansen, Kenneth & Hansen, Lesly. (1969). Pintupi phonology. Oceanic Linguistics, 8(2), 153–170. Hansen, Kenneth & Hansen, Lesly. (1978). The core of Pintupi grammar. Institute for Aboriginal Development. Harvey, Mark. (2008). The origin of conjugation markers in Australian languages. In Claire Bowern, Bethwyn Evans & Luisa Miceli (Eds.), Morphology and language history: In honour of Harold Koch (pp. 123–138). Amsterdam: John Benjamins. Haselow, Alexander. (2011). Discourse marker and modal particle: The functions of utterance-final then in spoken English. Journal of Pragmatics, 43(14), 3603–3623. https://doi.org/10.1016/ j.pragma.2011.09.002 Haspelmath, Martin. (1995). The converb as a cross-linguistically valid category. In Martin Haspelmath & Ekkehard König (Eds.), Converbs in crosslinguistic comparison (pp. 1–53). Berlin: De Gruyter Mouton. Haspelmath, Martin. (2014). Arguments and adjuncts as language-particular syntactic categories and as comparative concepts. Linguistic Discovery, 12(2), 3–11. Hayes, Bruce. (1995). Metrical stress theory: Principles and case studies. Chicago: University of Chicago Press. Heath, Jeffrey. (1984). Functional grammar of Nunggubuyu. Canberra: Australian Institute of Aboriginal Studies. Heath, Jeffrey. (1986). Syntactic and lexical aspects of nonconfigurationality in Nunggubuyu (Australia). Natural Language & Linguistic Theory, 4(3), 375–408.

References 

 753

Henderson, John. (2002). The word in Eastern/Central Arrernte. In Robert Dixon & Alexandra Aikhenvald (Eds.), Word: A cross-linguistic typology (pp. 100–124). Cambridge: Cambridge University Press. Hengeveld, Kees. (1992). Parts of speech. In Michael Fortescue, Peter Harder & Lars Kristoffersen (Eds.), Layered structure and reference in a functional perspective (pp. 29–55). Amsterdam: John Benjamins. Hercus, Luise. (1994). A grammar of the Arabana-Wangkangurru language, Lake Eyre Basin, South Australia. Canberra: Pacific Linguistics. https://doi.org/10.15144/PL-C128 Hercus, Luise. (1999). A grammar of the Wirangu language from the west coast of South Australia. Canberra: Australian National University. http://hdl.handle.net/1885/146617 Himmelmann, Nikolaus. (1996). Demonstratives in narrative discourse: A taxonomy of universal uses. Amsterdam: Benjamins. Hobson, John. (1990). Current distribution of Central Australian languages [Map]. Alice Srpings: Institute for Aboriginal Development. Hoffman, Dorothea. (2015a). Complex predication and serialization in the Daly River languages (and beyond). Presented at the School of Languages and Comparative Cultural Studies Research Seminar, University of Queensland, Brisbane. Hoffman, Dorothea. (2015b). Serialisation in complex predicates in MalakMalak. Unpublished manuscript. Holcombe, Sarah. (2004). The politico-historical construction of the Pintupi Luritja and the concept of tribe. Oceania, 74(4), 257–275. Hollenbach, Bruce. (1975). Discourse structure, interpropositional relations, and translation. Notes on Translation, 56, 2–21. Honeyman, Tom. (2005). Topic and focus in Ngardi. Honours thesis. Sydney: University of Sydney. Hopper, Paul & Thompson, Sandra. (1980). Transitivity in grammar and discourse. Language, 1, 251–299. Hudson, Joyce. (1978). The core of Walmatjari grammar. Canberra: Australian Institute of Aboriginal Studies. Hudson, Joyce. (1983). Transitivity and aspect in the Kriol verb. In Peter Mühlhäusler, Papers in pidgin and creole linguistics (pp. 161–176). Canberra: Pacific Linguistics. Hudson, Joyce. (1985). Grammatical and semantic aspects of Fitzroy Valley Kriol. Darwin: Summer Institute of Linguistics. Hudson, Joyce & Richards, Eirlys. (1969). The phonology of Walmatjari. Oceanic Linguistics, 8(2), 171–189. Hudson, Joyce & Richards, Eirlys. (1984). The Walmatjari: An introduction to the language and culture (Vol. 1). Darwin: Summer Institute of Linguistics. Hudson, Joyce & Richards, Eiryls. (1990). Walmajarri–English Dictionary. Darwin: Summer Institute of Linguistics. Ise, Megumi. (1999). Grammatical sketch of the Malngin Language. Master of Arts thesis. Sapporo: Hokkaido University. Jacobsen, William. (1967). Switch-reference. In Dell Hymes & William Biddle (Eds), Studies in southwestern ethnolinguistics (pp. 238–263). Hague/Paris: De Gruyter Mouton. Jagst, Lothar. (1975). Ngardilpa (Warlpiri) phonology. Pacific Linguistics, A-39(21). Jagst, Lothar. (1982). A tentative description of Ngardilpa (Warlpiri) verbs. In Stephen Swartz (Ed.), Papers in Warlpiri grammar: In memory of Lothar Jagst (Vol. 6, pp. 21–57). SIL-AAB. Jelinek, Eloise. (1984). Empty categories, case and configurationality. Natural Language & Linguistic Theory, 2(1), 39–76. Jepson, Kathleen. (2019). Prosody, prominence & segments in Djambarrpuyŋgu. PhD dissertation. Melbourne: University of Melbourne.

754 

 References

Jespersen, Otto. (1924). The philosophy of grammar. London: Routledge. Jolly, Lesly. (1989). Aghu Tharrnggala: A language of the Princess Charlotte Bay Region of Cape York Peninsula. Honours thesis. Brisbane: University of Queensland. Jones, Barbara. (2011). A grammar of Wangkajunga: A language of the Great Sandy Desert of North Western Australia. Canberra: Pacific Linguistics. Kearney, William. (1985). Warlpiri, Kukatja and Ngarti land claim. (Report by the Aboriginal Land Commissioner, Mr Justice Kearney, to the Minister of Aboriginal Affairs and to the Administrator of the Northern Territory No. 22). Australian Government Publishing Service. Keen, Sandra. (1983). Yukulta. In Robert Dixon & Barry Blake (Eds.), Handbook of Australian languages (Vol. 3, pp. 190–301). Canberra: ANU Press. Koch, Harold. (1984). The category of ‘associated motion’ in Kaytej. Language in Central Australia, 1, 23–24. Koch, Harold. (2014). The reconstruction of inflectional classes in morphology. In Robert Pensalfini, Myfany Turpin & Diana Guillemin, Language description informed by theory (Vol. 147, pp. 153–190). Amsterdam: John Benjamins. Koenig, Jean-Pierre, Mauner, Gail & Bienvenue, Breton. (2003). Arguments for adjuncts. Cognition, 89(2), 67–103. https://doi.org/10.1016/S0010-0277(03)00082-9 Koning, Robin. (2001). Kukatja mourning rites and inculturation. Mission, 8(2), 209–246. Koptjevskaja-Tamm, Maria. (1993). Nominalizations. London: Routledge. Langacker, Ronald. (1990). Concept, image and symbol: The cognitive basis of grammar. Berlin: De Gruyter Mouton Laughren, Mary. (1978). Directional terminology in Warlpiri. Working Papers in Language and Linguistics, 8, 1–16. Laughren, Mary. (1982a). A preliminary description of propositional particles in Warlpiri. In Stephen Swartz (Ed.), Papers in Warlpiri grammar: In memory of Lothar Jagst (pp. 129–164). SIL-AAB. Laughren, Mary. (1982b). Warlpiri kinship structure. In Jeffrey Heath, Francesca Merlan & Alan Rumsey (Eds), Oceania Linguistic Monographs (pp. 72–85). Sydney: University of Sydney. Laughren, Mary. (1988). Towards a lexical representation of Warlpiri verbs. In Wendy Wilkins (Ed.), Thematic relations (pp. 215–242). San Diego: Academic Press. Laughren, Mary. (1989). The configurationality parameter and Warlpiri. In László Marácz & Pieter Muysken (Eds.), Configurationality: The typology of asymmetries (pp. 319–353). Dordrecht: Foris. Laughren, Mary. (1992). Secondary predication as a diagnostic of underlying structure in Pama-Nyungan languages. In Iggy Roca (Ed.), Thematic structure: Its role in grammar (pp. 199–246). Dordrecht: Foris. Laughren, Mary. (2000). Le développement de l’opposition phonologique entre l’articulation occlusive et le battement dans une langue australienne (le warlpiri). In Jean-Philippe Dalbera (Ed.), Hommages à Zvonimir Junković. Université de Nice. Laughren, Mary. (2002). Syntactic constraints in a ‘free word order’ language. In Mengistu Amberber & Peter Collins (Eds.), Language universals and variation (pp. 83–130). London: Amblex. Laughren, Mary. (2010). Warlpiri verbs of change and causation: The thematic core. In Mengistu Amberber, Brett Baker & Mark Harvey (Eds.), Complex predicates: Cross-linguistic perspectives on event structure (pp. 167–236). Cambridge: Cambridge University Press. Laughren, Mary. (2016a). Decomposing the Warlpiri verbal complex. Mirndi-Ngumpin Language Contact Workshop, St Lucia. Laughren, Mary. (2016b). Possession in Kuku-Thaypan through a comparative lens. In Jean-Christophe Verstraete & Diane Hafner (Eds), Land and language in Cape York Peninsula and the Gulf Country (pp. 179–198). Amsterdam: John Benjamins.

References 

 755

Laughren, Mary. (2016c). Serial verbs in Waanyi and its neighbours. In Peter Austin, Harold Koch & Jane Simpson (Eds.), Language, land & song: Studies in honour of Luise Hercus (pp. 172–193). London: EL Publishing. Laughren, Mary. (2017a). A comparative study of clausal negation in Ngumpin-Yapa languages. Ngumpin-Yapa Workshop, University of Queensland, St Lucia. Laughren, M. (2017b). A comparison of TAM categories and forms in Ngumpin-Yapa languages. Australian Linguistics Society Conference, University of Sydney. Laughren, Mary. (2017c). Negation: How it imposes limits on constituent order even in ‘non-configurational’ languages. Australian Linguistics Society Annual Conference, Sydney. Laughren, Mary. (2017d). Reconstructing a morphologically complex genitive pattern in Australian languages. Unpublished manuscript. Laughren, Mary. (2017e). The Ergative in Warlpiri: A case study. In Jessica Coon, Diane Massam & Lisa Travis (Eds.), The Oxford Handbook of Ergativity. https://doi.org/10.1093/ oxfordhb/9780198739371.013.39 Laughren, Mary, Hale, Kenneth & Warlpiri Lexicography Group. (2007). Warlpiri-English encyclopaedic dictionary. University of Queensland Laughren, Mary, Hoogenraad, Robert, Hale, Kenneth & Granites, Robin (1996). A learner’s guide to Warlpiri, tape course for beginners. Alice Springs: IAD Press. Laughren, Mary & Ingram, John. (1999). The stop-flap contrast in Western Warlpiri. Australian Linguistic Society, University of Western Australia. Legate, Julie (2002). Warlpiri: Theoretical implications. PhD dissertation. Cambridge: Massachusetts Institute of Technology. Legate, Julie. (2003). The morpho-semantics of Warlpiri counterfactual conditionals. Linguistic Inquiry, 34(1), 155–162. Legate, Julie. (2008a). Warlpiri and the theory of second position clitics. Natural Language & Linguistic Theory, 26(1), 3–60. Legate, Julie. (2008b). Morphological and abstract case. Linguistic Inquiry, 39(1), 55–101. https:// doi.org/10.1162/ling.2008.39.1.55 Lempert, William. (2018). Palya futures: The social life of Kimberley Aboriginal Media. PhD dissertation. Boulder: University of Colorado. Levin, Beth. (1993). English verb classes and alternations: A preliminary investigation. Chicago: University of Chicago Press. Liberman, Mark & Prince, Alan. (1977). On stress and linguistic rhythm. Linguistic Inquiry, 8, 249–336. Lichtenberk, Frank. (2000). Inclusory pronominals. Oceanic Linguistics, 39, 1–32. Lichtenberk, Frank. (2002). The possessive-benefactive connection. Oceanic Linguistics, 41, 439–474. Loakes, Debbie., Butcher, Andrew., Fletcher, Janet & Stoakes, Hywel. (2008). Phonetically prestopped laterals in Australian languages: A preliminary investigation of Warlpiri. Proceedings of Interspeech 2008 incorporating ASSTA 2008, 90–93. Long, Jeremy. (1989). Leaving the desert: Actors and sufferers in the Aboriginal exodus from the Western Desert. Aboriginal History, 13(1/2), 9–43. Louagie, Dana & Verstraete, Jean-Christophe. (2015). Personal pronouns with determining functions in Australian languages. Studies in Language, 39(1), 159–198. https://doi.org/10.1075/ sl.39.1.06lou Louagie, Dana & Verstraete, Jean-Christophe. (2016). Noun phrase constituency in Australian languages: A typological study. Linguistic Typology, 20(1), 25–80. https://doi.org/10.1515/ lingty-2016-0002 Lyons, John. (1977). Semantics (Vol. 1). Cambridge: Cambridge University Press.

756 

 References

Macklin-Cordes, Jayden & Round, Erich. (2015). High-definition phonotactics reflect linguistic pasts. Proceedings of the 6th Conference on Quantitative Investigations in Theoretical Linguistics. Tübingen: University of Tübingen. https://doi.org/10.15496/publikation-8609 Malchukov, Andrej, Haspelmath, Martin & Comrie, Bernard. (2010). Ditransitive constructions: A typological overview. In Andrej Malchukov, Martin Haspelmath & Bernard Comrie (Eds.), Studies in ditransitive constructions: A comparative handbook (pp. 1–64). Berlin/Boston: De Gruyter Mouton. Mayer, Mercer. (2003). Frog, where are you? A boy, a dog, and a frog. New York: Dial Books. Marantz, Alec. (1981). On the nature of grammatical relations. PhD dissertation. Cambridge: Massachusetts Institute of Technology. Marsh, James (1976). The grammar of Mantjiltjara. MA thesis. Tempe: Arizona State University. Matthews, Peter. (1972). Inflectional morphology: A theoretical study based on aspects of Latin verb conjugation. Cambridge: Cambridge University Press. Matthews, Peter. (1991). Morphology (2nd ed.). Cambridge: Cambridge University Press. McConvell, Patrick. (1980). Hierarchical variation in pronominal clitic attachment in the Eastern Ngumbin languages. In Bruce Rigsby & Peter Sutton (Eds.) Papers in Australian Linguistics No 13: contributions to Australian linguistics (pp. 31–117). Canberra: Pacific Linguistics. McConvell, Patrick. (1982). Neutralisation and degrees of respect in Gurindji. In Jeffrey Heath, Francesca Merlan & Alan Rumsey (Eds), Languages of kinship in Aboriginal Australia (pp. 86–106). Sydney: University of Sydney. McConvell, Patrick. (1988). Mix-im-up: Aboriginal code-switching old and new. In Monica Heller (Ed.), Codeswitching: Anthropological and sociolinguistic perspectives (pp. 97–150). Berlin: De Gruyter Mouton. McConvell, Patrick. (1995). Subordinate clauses in Gurindji. Unpublished manuscript. Brisbane: University of Queensland. McConvell, Patrick. (1996a). Backtracking to Babel: The chronology of Pama-Nyungan expansion in Australia. Archaeology in Oceania, 31, 125–144. McConvell, Patrick. (1996b). Gurindji grammar. Unpublished manuscript. McConvell, Patrick. (1996c). The functions of split-Wackernagel clitic systems: Pronominal clitics in the Ngumpin languages. In Aaron Halpern & Arnold Zwicky Approaching Second: Second Position Clitics and Related Phenomena (pp. 299–332). Stanford: CSLI Publications McConvell, Patrick. (1997). Long-lost relations: Pama-Nyungan and Northern kinship. In Patrick McConvell & Nicholas Evans (Eds.), Archaeology and Linguistics: Aboriginal Australia in Global Perspective, 207–236. McConvell, Patrick. (2006). Grammaticalization of demonstratives as subordinate complementizers in Ngumpin-Yapa. Australian Journal of Linguistics, 26(1), 107–137. https://doi. org/10.1080/07268600500531669 McConvell, Patrick. (2009). Loanwords in Gurindji, a Pama-Nyungan language of Australia. In Martin Haspelmath & Uri Tadmor (Eds.), Loanwords in the world’s languages: A comparative handbook (pp. 790–821). Berlin/Boston: De Gruyter Mouton. https://doi.org/10.1515/9783110218442 McConvell, Patrick. (2010). Mood swings: Imperative verbs attract pronominal enclitics in Ngumpin-Yapa (Australian) and Southern European languages. In Rachel Hendry & Jennifer Hawkins (Eds), Grammatical change: Theory and description (pp. 123–156). Canberra: Pacific Linguistics. McConvell, Patrick & Bowern, Claire. (2011). The prehistory and internal relationships of Australian languages. Language and Linguistics Compass, 5(1), 19–32. McConvell, Patrick & Laughren, Mary. (1996). Where did the Western Desert language come from? The Third Australian Linguistics Institute: Where did the Western Desert Language come from?, Canberra: Australian National University.

References 

 757

McConvell, Patrick & Laughren, Mary. (2004). The Ngumpin-Yapa subgroup. In Claire Bowern & Harold Koch (Eds), Australian languages: Classification and the comparative method (Vol. 249, pp. 169–196). Amsterdam: John Benjamins Publishing. McConvell, Patrick & Simpson, Jane (2012). Fictive notion down under: The locative-allative case alternation in some Australian Indigenous languages. In Diana Santos, Krister Lindén & Wanjiku Ng’ang’a (Eds), Shall we play the festschrift game?: Essays on the occasion of Lauri Carlson’s 60th birthday (pp. 159–180). Berlin: Springer. McCoy, Brian. (2008). Holding men: Kanyirninpa and the health of Aboriginal men. Canberra: Canberra: Aboriginal Studies Press. McGregor, William. (1988a) Handbook of Kimberley Languages. Canberra: Pacific Linguistics McGregor, William. (1988b). Mood and subordination in Kuniyanti. In Peter Austin (Ed.), Complex Sentence Constructions in Australian Languages (pp. 37–67). Amsterdam: John Benjamins. McGregor, William. (1990). A functional grammar of Gooniyandi. Amsterdam: John Benjamins. McGregor, William. (1997). Functions of noun phrase discontinuity in Gooniyandi. Functions of Language, 4, 83–114. McGregor, William. (2002). Verb classification in Australian languages. Berlin: De Gruyter Mouton. McGregor, William. (2004). The Languages of the Kimberley, Western Australia. London: Routledge Curzon. https://doi.org/10.4324/9780203434710 McGregor, William & Verstraete, Jean-Christophe. (2010). Optional ergative marking and its implications for linguistic theory. Lingua, 120(7), 1607–1609. https://doi.org/10.1016/j. lingua.2009.05.009 Meakins, Felicity. (2009). The case of the shifty ergative marker: A pragmatic shift in the ergative marker in one Australian mixed language. In Jóhanna Barðdal, & Shobhana Chelliah (Eds.), The role of semantic, pragmatic, and discourse factors in the development of case (pp. 59–91). Amsterdam: John Benjamins. Meakins, Felicity. (2011). Case-marking in contact: The development and function of case morphology in Gurindji Kriol (Vol. 39). Amsterdam: John Benjamins. https://doi.org/10.1075/cll.39 Meakins, Felicity. (2014). Nominals as adjuncts or arguments: Further evidence from language mixing. In Robert Pensalfini, Myfany Turpin & Diane Guillemin (Eds), Language description informed by theory (pp. 283–315). Amsterdam: John Benjamins. Meakins, Felicity, McConvell, Patrick, Charola, Erika, McNair, Norm, McNair, Helen & Campbell, Lauren. (2013). Gurindji to English dictionary. Batchelor: Batchelor Press. Meakins, Felicity & Nordlinger, Rachel. (2014). A grammar of Bilinarra: An Australian Aboriginal language of the Northern Territory. (Vol. 640). Berlin/Boston: De Gruyter Mouton. Meakins, Felicity & Nordlinger, Rachel. (2017). Possessor dissension in the Ngumpin-Yapa languages of northern Australia. Linguistic Typology, 21(1), 143–176. https://doi.org/10.1515/lingty-2017-0004 Meakins, Felicity & O’Shannessy, Carmel. (2016). No fixed address: The grammaticalisation of the Gurindji locative as a progressive suffix. In Felicity Meakins & Carmel O’Shannessy, Loss and renewal: Australian languages since colonisation (pp. 367–396). De Gruyter Mouton. https:// doi.org/10.1515/9781614518792-018 Meggit, Mervyn. (1962). Desert people: A study of the Warlpiri Aborigines of Central Australia. Sydney: Angus & Robertson. Mel’čuk, Igor. (1998). Kurs obščej morfologii (General morphology). Moscow: Jazyki Russkoj Kul’tury, 2. Merlan, Francesca. (1994). A grammar of Wardaman: A language of the Northern Territory of Australia (Vol. 11). Berlin: De Gruyter Mouton. Merlan, Francesca. (2016). Correlation of textual and spatial reference: This and that. In Jean-Christophe Verstraete & Diane Hafner (Eds), Land and language in Cape York Peninsula and the Gulf Country (pp. 199–217). Amsterdam: John Benjamins. https://doi.org/10.1075/ clu.18.10mer

758 

 References

Mithun, Marianne. (1984). The evolution of noun incorporation. Language, 60(4), 874–894. Mithun, Marianne (2000). Valency-changing derivation in Central Alaskan Yup’ik. In Robert. Dixon & Alexandra Aikhenvald (Eds.), Changing valency: Case studies in transitivity (pp. 84–114). Cambridge: Cambridge University Press. Moora, Mark. (2015). Tjawa Tjawa. Pilbara and Kimberley Aboriginal Media. Moyle, Robert. (1997). Balgo: The musical life of a desert community. Perth: Callaway International Resource Centre for Music Education. Munt, Simon, Roberts, Amy & Gorman, Alice. (2018). An investigation of human responses to climatic fluctuations at Allen’s Cave, South Australia, from ca. 40,000 to 5,000 BP, by a technological analysis of stone artefacts. Australian Archaeology, 84(1), 67–83. https://doi. org/10.1080/03122417.2018.1458446 Murray, Robert & Vennemann, Theo. (1983). Sound change and syllable structure in Germanic phonology. Language, 59(3), 514–528. JSTOR. https://doi.org/10.2307/413901 Mushin, Ilana. (1995). Epistememes in Australian languages. Australian Journal of Linguistics, 15, 1–31. Mushin, Ilana & Simpson, Jane (2008). Free to bound to free? Interactions between pragmatics and syntax in the development of Australian pronominal systems. Language, 84(3), 566–596. https://doi.org/10.1353/lan.0.0048 Myers, Fred. (n.d.). What is the business of ‘Balgo Business’: A contemporary religious movement. Napaljarri, Peggy Rockman & Cataldi, Lee. (1994). Yimikirli: Warlpiri Dreamings and histories. San Francisco: HarperCollins. Napanangka, Tjama, Napangarti, Payi Payi, Napanangka, Martinagli Mudegell., Nampitjin, Kuninyi Rita, Napurrula, Nanyuma Rosie & Nampitjin, Millie Skeen. (1997). Yarrtji: Six women’s stories from the Great Sandy Desert (S. Peter & P. Lofts, Eds.). Canberra: Aboriginal Studies Press. Nash, David. (1979). Grammatical preface to vocabulary of the Warlmanpa language. Cambridge: Massachusetts Institute of Technology. http://www.anu.edu.au/linguistics/nash/aust/wpa/ wpa-vocab.intro.html Nash, David. (1982). Warlpiri preverbs and verb roots. In Stephen Swartz (Ed.), Papers in Warlpiri grammar: In memory of Lothar Jagst (Vol. 6, pp. 165–216). Darwin: SIL-AAB. Nash, David. (1986). Topics in Warlpiri grammar. New York: Garland. Nash, David. (1988). Draft grammar of Mudburra (and Kuwaarrangu). Unpublished manuscript. Nash, David. (1991). Warlpiri fire management. In David Pearson (Ed.), Management of Spinifex Deserts for Nature Conservation. Proceedings of a Workshop held at the Department of Conservation and Land Management (p. 12). Department of Conservation and Land Management Nash, David. (1992). An Australian kinship affix *-rti. Australian Journal of Linguistics, 12(1), 123–144. https://doi.org/10.1080/07268609208599473 Nash, David. (1996). Pronominal clitic variation in the Yapa languages. In William McGregor (Ed.), Studies in Kimberley languages: In honour of Howard Coate (pp. 117–138). Munich: Lincom Europa. Nash, David. (2008). Warlpiri verb roots in comparative perspective. In Claire Bowern, Bethwyn Evans & Luisa Micelli (Eds.), Morphology and language history: In honour of Harold Koch (pp. 221–234). Amsterdam: John Benjamins. Nichols, Johanna. (1986). Head-marking and dependent-marking grammar. Language, 62(1), 56–119. https://www.jstor.org/stable/415601 Nordlinger, Rachel. (1998). A grammar of Wambaya, Northern Territory (Australia). Canberra: Pacific Linguistics. Nordlinger, Rachel. (2006). Spearing the emu drinking: Subordination and the adjoined relative clause in Wambaya. Australian Journal of Linguistics, 26(1), 5–29. https://doi. org/10.1080/07268600500531610

References 

 759

Nordlinger, Rachel. (2014). Constituency and grammatical relations in Australian languages. In Harold Koch & Rachel Nordlinger (Eds.), The languages and linguistics of Australia: A comprehensive guide (pp. 215–261). Berlin/Boston: De Gruyter Mouton. Nordlinger, Rachel. (2017). The languages of the Daly River region (Northern Australia). In Nicholas Evans, Michael Fortescue & Marianne Mithun (Eds), The Oxford handbook of polysynthesis (pp. 782–807). Oxford: Oxford University Press. Nordlinger, Rachel. (2002). Non-finite subordinate verbs in Australian Aboriginal languages: Are nominalised verbs really nominalised? In Cynthia Allen (Ed.), Proceedings of the 2001 Conference of the Australian Linguistic Society (pp. 1–10). Oates, Lynette. (1975). The 1973 supplement to a revised linguistic survey of Australia (Vol. 1). Armidale: Armidale Christian Book Centre. Oates, William & Oates, Lynette. (1970). A revised linguistic survey of Australia (Vol. 12). Canberra: Australian Institute of Aboriginal Studies. O’Connell, James & Allen, Jim. (2012). The restaurant at the end of the universe: Modelling the colonisation of Sahul. Australian Archaeology, 74, 5–17. JSTOR. O’Grady, Geoffrey, Voegelin, Carl & Voegelin, Florence (1966). Languages of the world: Indo-Pacific fascicle six. Anthropological Linguistics, 8(2), 1–197. Ohala, John (1983). The origin of sound patterns in vocal tract constraints. In Peter MacNeilage The production of speech (pp. 175–187). New York: Springer-Verlag. Ohala, John (1990). The phonetics and phonology of aspects of assimilation. In John Kingston & Mary Beckman (Eds.), Papers in Laboratory Phonology I: Between the grammar and the physics of speech (pp. 258–275). Cambridge: Cambridge University Press. Olney, Howard (1992). Tanami Downs Land Claim. (Report by the Aboriginal Land Commissioner to the Minister for Aboriginal and Torres Strait Islander Affairs and to the Administrator of the Northern Territory. No. 42). Osgarby, David. (2018). Verbal morphology and syntax of Mudburra: An Australian Aboriginal language of the Northern Territory. MPhil thesis. Brisbane: University of Queensland. Osgarby, David. (In press) Complex predication and serialization. In Claire Bowern (Ed.), Oxford Handbook of Australian Languages. Oxford: Oxford University Press. O’Shannessy, Carmel. (2005). Light Warlpiri – A new language. Australian Journal of Linguistics, 25(1), 31–57. O’Shannessy, Carmel & Meakins, Felicity. (2012). Typological constraints on verb integration in two Australian mixed languages. Journal of Language Contact, 2, 216–246. https://doi. org/10.1163/19552629-006001001 Östman, Jan-Ola. (1981). You know: A discourse-functional study. Amsterdam: John Benjamins. https://benjamins.com/catalog/pb.ii.7 Palmer, Frank. (2001). Mood and modality (2nd ed.). Cambridge: Cambridge University Press. Payi Payi & Ors on behalf of the Ngururrpa People v the State of Western Australia, WAD357/06 (Federal Court of Australia October 18, 2007). Payne, Doris & Barshi, Immanuel. (1999). External possession: what, where, how and why. In Doris Payne & Immanuel Barshi (Eds.), External Posession (pp. 3–31). Amsterdam: John Benjamins. Peile, Anthony & Bindon, Peter. (1997). Body and soul: An Aboriginal view. Carlisle: Hesperian Press. Pensalfini, Robert. (2004). Towards a typology of configurationality. Natural Language & Linguistic Theory, 22(2), 359–408. https://doi.org/10.1023/B:NALA.0000015794.02583.00 Pentland, Christina. (2004). Stress in Warlpiri: Stress domains and word-level prosody. MPhil thesis. Brisbane: University of Queensland. Pentland, Christina & Ingram, John. (2003). Connected speech processes and prosodic constituency in Warlpiri. In Sallyane Palethorpe & Marija Tabain (Eds.), Proceedings of the 6th International Seminar on Speech Production (pp. 213–218). Macquarie Centre for Cognitive Science.

760 

 References

Pentland, Chritina & Laughren, Mary. (2004). Distinguishing prosodic word and phonological word in Warlpiri: Prosodic constituency in morphologically complex words. In Ilana Mushin (Ed.), Proceedings of the 2004 Conference of the Australian Linguistic Society. The University of Sydney. Perlmutter, David. (1980). Relational grammar. In Edith Moravcsik & Jessica Wirth (Eds.), Syntax and semantics: Current approaches to syntax (Vol. 13, pp. 195–229). New York: Academic Press. Perlmutter, David. (1983). Studies in relational grammar 1. Chicago: Chicago University Press. Perlmutter, David & Rosen, Carol. (1984). Studies in relational grammar 2. Chicago: Chicago University Press. Peterson, Nicolas & Myers, Fred. (2016). Experiments in self-determination: Histories of the outstation movement in Australia. Canberra: ANU Press. Pillans, Brad. (2018). Seeing red: Some aspects of the geological and climatic history of the Australian arid zone. In H. Lambers (Ed.), On the ecology of Australia’s arid zone (pp. 5–43). Cham: Springer International. Plank, Frans. (1995). Double case: Agreement by Suffixaufnahme. Oxford: Oxford University Press. Platt, John. (1972). An outline grammar of the Gugada dialect: South Australia. Canberra: Australian Institute of Aboriginal Studies. Poirier, Sylvie. (2005). A world of relationships. Itineraries, dreams and events in the Australian Western Desert. Toronto: University of Toronto Press. Radcliffe-Brown, Alfred. (1930). Social organisation of Australian tribes. Oceania Monographs, 1, 3–124. Rapold, Christian. (2010). Beneficiary and other roles of the dative in Tashelhiyt. In Fernando Zúñiga & Seppo Kittilä (Eds.) Benefactives and Malefactives: Typological perspectives and case studies (pp. 351–376). Amsterdam: John Benjamins. Reichenbach, Hans. (1947). Elements of symbolic logic. New York: Macmillan. Reid, Nicholas. (2000). Complex verb collocations in Ngan’gityemerri: A nonderivational strategy for encoding valency alternations. In R. M. W. Dixon & A. Y. Aikhenvald (Eds.), Changing valency: Case studies in transitivity (pp. 333–359). Cambridge: Cambridge University Press. Richards, Eirlys & Hudson, Joyce. (2012). Interactive Walmajarri – English dictionary: With English – Walmajarri finderlist (2nd ed.). Darwin: Australian Society for Indigenous Languages. Ritz, Marie-Eve, Dench, Alan & Caudal, Patrick. (2012). Now or then? The clitic -rru in Panyjima: Temporal properties in discourse. Australian Journal of Linguistics, 32(1), 41–72. https://doi. org/10.1080/07268602.2012.657753 Ritz, Marie-Eve & Schultze-Berndt, Eva. (2015). Time for a change? The semantics and pragmatics of marking temporal progression in an Australian language. Lingua, 166, 1–21. https://doi. org/10.1016/j.lingua.2015.07.007 Round, Erich. (2009). Kayardild morphology, phonology and morphosyntax. PhD dissertation. New Haven: Yale University. Rumsey, Alan. (2000). Bunuba. In Robert Dixon & Barry Blake (Eds.) The Handbook of Australian languages (Vol. 5, pp. 1–153). Cambridge: Cambridge University Press. Ryan, Judith. (1989). Mythscapes: Aboriginal art of the desert from the National Gallery of Victoria; with an essay by Geoffrey Bardon. Melbourne: National Gallery of Victoria. Sadock, Jerrold & Zwicky, Arnold. (1985). Speech act distinctions in syntax. Language Typology and Syntactic Description, 1, 155–196. Sasse, Hans-Jürgen. (2006). Theticity. In Georg Bossong & Bernard Comrie (Eds.), Pragmatic organization of discourse in the languages of Europe (Vol. 8, pp. 255–308). Walter de Gruyter. Schiffrin, Deborah. (1981). Tense variation in narrative. Language, 57(1), 45–62. Schourup, Lawrence. (1985). Common discourse particles in English conversation. New York: Garland.

References 

 761

Schultze-Berndt, Eva. (2006). Towards a semanto-pragmatic account of optional ergativity in Jaminjung/Ngaliwurru. Paper presented at the 2nd European Australianists Meeting, Somló, Hungary. Schultze-Berndt, Eva. (2000). Simple and complex predicates in Jaminjung: A study of event categorisation in an Australian language. PhD dissertation. Nijmegen: Radboud University Nijmegen. Schultze-Berndt, Eva. (2004). Depictive secondary predicates in crosslinguistic perspective. Linguistic Typology, 8(1), 59–131. Schultze-Berndt, Eva. (2003). Preverbs as an open word class in Northern Australian languages: Synchronic and diachronic correlates. In Geert Booij & Jaap van Marle (Eds), Yearbook of morphology 2003 (pp. 145–177). Dordrecht: Kluwer. Schultze-Berndt, E. (2006). Secondary predicates in Australian languages. In Martin Everaert & Henk van Riemsdijk (Eds.), The Blackwell companion to syntax (pp. 180–208). Malden/Oxford: John Wiley & Sons. Schultze-Berndt, Eva. (2008). Discontinuous noun phrases as an iconic strategy of marking thetic clauses. Paper presented at the International Conference on Syntax of the World’s Languages III, Berlin. Schultze-Berndt, Eva & Himmelmann, Nikolaus. (2004). Depictive secondary predicates in crosslinguistic perspective. Linguistic Typology, 8(1), 59–131. Schultze-Berndt, Eva & Simard, Candide. (2012). Constraints on noun phrase discontinuity in an Australian language: The role of prosody and information structure. Linguistics, 50(5), 1015–1058. Schweiger, Fritz. (2000). Compound case markers in Australian languages. Oceanic Linguistics, 39(2), 256–284. Ségéral, Philippe & Scheer, Tobias. (2008). Positional factors in lenition and fortition. In Joacquim de Carvalho, Tobias Scheer & Philippe Ségéral (Eds.), Lenition and fortition (pp. 131–172). Berlin: De Gruyter Mouton. Selkirk, Elizabeth. (1980). The role of prosodic categories in English word stress. Linguistic Inquiry, 11, 563–605. Senge, Chikako. (2008). A sketch grammar of Wanyjirra. Postgraduate diploma. Melbourne: University of Melbourne. Senge, Chikako. (2015). A grammar of Wanyjirra, a language of Northern Australia. PhD dissertation. Canberra: Australian National University. http://hdl.handle.net/1885/109341 Sharifian, Farzad & Malcolm, Ian. (2003). The pragmatic marker like in English teen talk: Australian Aboriginal usage. Pragmatics & Cognition, 11(2), 327–344. Sharp, Janet. (2004). Nyangumarta: A language of the Pilbara region of Western Australia. Pacific Linguistics. http://hdl.handle.net/1885/146175 Shimada, Masaharu, Nagano, Aiko & Okubo, Tatsuhiro. (2015). Priscianic formation in compounding and upgrading as stem-formation. Tsukuba English Studies, 34, 1–22. Silverstein, Michael. (1976). Hierarchy of features and ergativity. In Robert Dixon (Ed.), Grammatical categories in Australian languages (pp. 112–172). New Jersey: Humanities Press. Simpson, Jane (1983). Aspects of Warlpiri morphology and syntax. PhD dissertation. Cambridge: Massachusetts Institute of Technology. http://hdl.handle.net/1721.1/15468 Simpson, Jane (1988). Case and complementiser suffixes in Warlpiri. In Peter Austin (Ed.), Complex sentence constructions in Australian languages (Vol. 15, p. 289). Amsterdam: John Benjamins. Simpson, Jane (1991). Warlpiri morpho-syntax: A lexicalist approach (Vol. 23). Dordrecht: Kluwer Academic. Simpson, Jane (2001). Preferred word order and the grammatization of associated path. In Miriam Butt & Tracy King (Eds), Time over matter: Diachronic perspectives on morphosyntax (pp. 173–208). Stanford: CSLI Publications.

762 

 References

Simpson, Jane (2005). Depictives in English and Warlpiri. In Nikolaus Himmelmann & Eva Schultze-Berndt (Eds.), Secondary predication and adverbial modification. The typology of depictives (pp. 69–106). Oxford: Oxford University Press. Simpson, Jane (2007). Expressing pragmatic constraints on word order in Warlpiri. In Annie Zaenan, Jane Simpson, Tracy Holloway Jane Grimshaw & Joan Maling (Eds.), Architectures, rules, and preferences: Variations on themes: A festschrift for Joan Bresnan (pp. 403–427). Stanford: CSLI Publications. Simpson, Jane (2014). Pama-Nyungan. In Rochelle. Lieber & P.avol Štekauer The Oxford handbook of derivational morphology (pp. 651–668). Oxford: Oxford University Press. https://doi. org/10.1093/oxfordhb/9780199641642.013.0035 Simpson, Jane & Mushin, Ilana. (2008). Clause-initial position in four Australian languages. In Ilana Mushin & Brett Baker (Eds), Discourse and grammar in Australian languages (pp. 25–57). Amsterdam: John Benjamins. Singer, Ruth. (2001). The inclusory construction in Australian languages. Honours thesis. Melbourne: University of Melbourne. Smith, Carlota. (1997). Viewpoint aspect. In Carlota Smith (Ed.), The parameter of aspect (pp. 61–95). Dordrecht: Springer Netherlands. https://doi.org/10.1007/978-94-011-5606-6_4 Smith, Mike. (2006). Characterizing Late Pleistocene and Holocene stone artefact assemblages from Puritjarra rock shelter: A long sequence from the Australian Desert. Records of the Australian Museum, 58(3), 371–410. Smith, Pamela. (2000). Station camps: Legislation, labour relations and rations on pastoral leases in the Kimberley region, Western Australia. Aboriginal History, 24, 75–97. JSTOR. Smith, Pamela. (2016). Purrkuji: Massacre on Sturt Creek. Report of the History, Archaeological Survey and Forensic Investigation. Report to the Descendants of those Killed in the Sturt Creek Massacre, Relevant Tjurabalan Native Title owners. Unpublished manuscript https://doi. org/10.13140/RG.2.2.18819.76321 Smith, Pamela, Raven, Mark, Walshe, Keryn, Fitzpatrick, Robert & Pate, Donald. (2017). Scientific evidence for the identification of an Aboriginal massacre at the Sturt Creek sites on the Kimberley frontier of north-western Australia. Forensic Science International, 279, 258–267. Stevens, Kenneth. (1998). Acoustic phonetics. Cambridge: Massachusetts Institute of Technology Press. Stoakes, Hywel, Fletcher, Janet & Butcher, Andrew. (2019). Nasal coarticulation in Bininj Kunwok: An aerodynamic analysis. Journal of the International Phonetic Association, 1–28. https://doi. org/10.1017/S0025100318000282 Stokes, Bronwyn. (1982). A Description of Nyigina. A language of the West Kimberley, Western Australia. PhD dissertation. Canberra: Australian National University. http://hdl.handle. net/1885/110250 Sturt on behalf of the Jaru Native Title Claim v State of Western Australia, No. 1923 (Federal Court of Australia December 6, 2018). Sutton, Peter. (2011) The politics of suffering. Melbourne: Melbourne University Press Swartz, Stephen. (1991). Constraints on zero anaphora and word order in Warlpiri narrative text (Vol. 1). Darwin: Summer Institute of Linguistics. Tabain, Marija, Fletcher, Janet & Butcher, Andrew. (2014). Lexical stress in Pitjantjatjara. Journal of Phonetics, 42(1), 52–56. Talmy, Leonard. (1983). How language structures space. In Herbert Pick & Lidna Acredolo (Eds.), Spatial orientation: Theory, research, and application (pp. 225–282). Springer US. Talmy, Leonard. (1985). Lexicalization patterns: Semantic structure in lexical forms. Language Typology and Syntactic Description, 3(99), 36–149.

References 

 763

Talmy, Leonard. (1996). Fictive motion in language and ‘ception.’ In Paul Bloom (Ed.), Language and space (pp. 211–276). Cambridge: Massachusetts Institute of Technology Press. Terry, Michael. (1925). Across unknown Australia: A thrilling account of exploration in the Northern Territory of Australia. London: Herbert Jenkins. Terry, Michael. (1926). A surgical operation as performed by the Boonarra tribe of Northern Australia, and a short vocabulary of the languages of some North Australian tribes. Man, 26(129), 193–194. https://doi.org/10.2307/2787434 Terry, Michael. (1927). Through a land of promise: With gun, car and camera in the heart of Northern Australia. London: Herbert Jenkins. Terry, Michael. (1931). Hidden wealth and hiding people. London: Putnam. Terry, Michael. (1987). The last explorer: The autobiography of Michael Terry, FRGS, FRGSA. Canberra: Australian National University Press. Tex on behalf of the Lappi Lappi and Ngulupi Claim Group v State of Western Australia, No. 1591 (Federal Court of Australia October 24, 2018). Thomson, Donald. (1962). The Bindubi expedition. The Geographical Journal, 128(1), 1–18. Thomson, Donald. (1975). Bindubi country. Melbourne: Nelson. Tindale, Norman. (1954). Field journal of Norman B. Tindale: 18th expedition under the auspices of Board for Anthropological Research. Unpublished manuscript. Tindale, Norman. (1956). A journal kept on the anthropological expedition to Central Australia, Haast Bluff, 1956. Unpublished manuscript. Tindale, Norman. (1974). Aboriginal tribes of Australia: Their terrain, environmental controls, distribution, limits, and proper names. Berkeley: University of California Press. Traugott, Elizabeth. (1978). On the expression of spatio-temporal relations in language. In Joseph Greenberg, Charles Ferguson & Edith Moravcsit (Eds.), Universals of Human Language (Vol. 3, pp. 369–400). Stanford: Stanford University Press. Traugott, Elizabeth. (1989). On the rise of epistemic meanings in English: An example of subjectification in semantic change. Language, 65(1), 31–55. JSTOR. https://doi.org/10.2307/414841 Traugott, Elizabeth, Meulen, Alice, Reilly, Judy & Ferguson, Charles. (1986). On conditionals. Cambridge Core. https://doi.org/10.1017/CBO9780511753466 Tsunoda, Tasaku. (1981a). Split case-marking patterns in verb-types and tense/aspect/mood. Linguistics, 19(5–6), 389–438. Tsunoda, Tasaku. (1981b). The Djaru Language of Kimberley, Western Australia. Canberra: Pacific Linguistics. Tsunoda, Tasaku. (1985). Remarks on transitivity. Journal of Linguistics, 21(2), 385–396. JSTOR. Tsunoda, Tasaku. (1995). The possession cline in Japanese and other languages. In Hilary Chappell & William McGregor (Eds.), The grammar of inalienability: A typological perspective on body part terms and the part-whole relation (565–630). Berlin: De Gruyter Mouton. https://doi. org/10.1515/9783110822137.565 Tsunoda, Tasaku. (2006). A comparative word list of Jaru: Its dialects and surrounding languages. Unpublished manuscript. Tsunoda, Tasaku. (2011). A grammar of Warrongo. Berlin: De Gruyter Mouton. Turpin, Myfany & Meakins, Felicity. (2019). Songs from the stations. Sydney: Sydney University Press. Vaarzon-Morel, Petronella & Gabrys, Kasia. (2009). Fire on the horizon: Contemporary Aboriginal burning issues in the Tanami Desert, Central Australia. GeoJournal, 74(5), 465–476. JSTOR. Valiquette, Hillaire. (Ed.) (1993). A basic Kukatja to English dictionary. Balgo: Luurnpa Catholic School. Veth, Peter. (2000). Origins of the Western Desert language: Convergence in linguistic and archaeological space and time models. Archaeology in Oceania, 35(1), 11–19. https://doi. org/10.1002/j.1834-4453.2000.tb00447.x

764 

 References

Veth, Peter, McDonald, Jo & de Koning, Sarah. (2018). Archaeology and rock art of the north-west arid zone with a focus on animals. In Hans Lambers (Ed.), On the ecology of Australia’s arid zone (pp. 283–305). Cham: Springer International. https://doi.org/10.1007/978-3-31993943-8_11 Veth, Peter, Smith, Mike, Bowler, Jim, Fitzsimmons, Kathryn, Williams, Alan & Hiscock, Peter. (2009). Excavations At Parnkupirti, Lake Gregory, Great Sandy Desert: OSL ages for occupation before the Last Glacial Maximum. Australian Archaeology, 69(1), 1–10. Vincent, Nigel. (2013). Conative. Linguistic Typology, 17(2), 269–289. https://doi.org/10.1515/ lity-2013-0012 Vuillermet, Marine. (2018). Grammatical fear morphemes in Ese Ejja: Making the case for a morphosemantic apprehensional domain. Studies in Language. International Journal Sponsored by the Foundation ‘Foundations of Language’, 42(1), 256–293. https://doi. org/10.1075/sl.00010.vui Wafer, Jim. (1980). Preliminary observations on Ngari. Unpublished manuscript. Walshe, Keryn. (1994). Taphonomic analysis of the vertebrate assemblage from Allen’s Cave; implications for Australian arid zone archaeology. PhD dissertation. Canberra: Australian National University. Watson, Christine. (1996). Kuruwarri: The generative force. MPhil thesis. Canberra: Australian National University. Watson, Christine. (1997). Re-embodying sand drawing and re-evaluating the status of the camp: The practice and iconography of women’s public sand drawing in Balgo. The Australian Journal of Anthropology, 8(2), 104–124. https://doi.org/10.1111/j.1835-9310.1997.tb00180.x Watson, Christine. (2003). Piercing the ground. Freemantle: Freemantle Art Centre Press. Westerlund, Torbjörn. (2015). A grammatical sketch of Ngarla (Ngayarta, Pama-Nyungan). Canberra: Pacific Linguistics Press. Wierzbicka, Anna. (1977). The ignorative: the semantics of speech acts. International review of Slavic linguistics, 2, 251–312. Wilkins, David. (1989). Mparntwe Arrernte (Aranda): Studies in the structure and semantics of grammar. PhD dissertation. Australian National University. http://hdl.handle.net/1885/9908 Wilkins, David. (1991). The semantics, pragmatics and diachronic development of ‘associated motion’ in Mparntwe Arrernte. Buffalo Papers in Linguistics, 1, 207–257. Wilkins, David. (1997). The verbalization of motion events in Arrernte (Central Australia). In Eve Clark (Ed.), The proceedings of the Twenty-Eighth Annual Child Language Research Forum (pp. 295–308). Stanford: CSLI Publications. Wilkins, David. (2006). Towards an Arrernte grammar of space. In Stephen Levinson & David Wilkins (Eds.), Grammars of space: Explorations in cognitive diversity (pp. 24–62). Cambridge: Cambridge University Press. Wilkins, David & Hill, Deborah. (2009). When ‘go’ means ‘come’: Questioning the basicness of basic motion verbs. Cognitive Linguistics, 6(2–3), 209–260. https://doi.org/10.1515/ cogl.1995.6.2-3.209 Wilson, Stephen. (1999). Coverbs and complex predicates in Wagiman. Stanford: CSLI Publications. Wohlgemuth, Jan. (2009). A typology of verbal borrowing. Berlin: De Gruyter Mouton. Wrigley, Matthew. (1991). Ngarti wordlist. Halls Creek: Kimberley Language Resource Centre. Young, Elspeth., Marchevsky, Alberto & Wick, Mark. (1981). Balgo business at Yuendumu. Canberra: Australian Institute of Aboriginal Studies. Zucker, Margaret. (2005). From patrons to partners and the separated children of the Kimberley: A history of the Catholic Church in the Kimberley, WA (2nd ed.). Freemantle: University of Notre Dame.

Index ablative case 8–10, 156, 166, 167, 190, 194, 196, 197, 215, 227, 230, 237, 240, 245, 268, 272, 273, 305, 318, 326, 458, 470, 509, 514–516 – See also elative case Aboriginal English 1, 140, 156, 201, 272, 610, 657 absolutive case 9–11, 88, 116, 123, 134, 152, 162–167, 170, 171, 176, 183, 184, 187, 232, 233, 237, 240–245, 247, 250, 251, 298, 301, 313, 314, 323, 324, 331, 436, 453, 502, 503, 506, 509, 511, 512, 513, 521, 524, 525, 526, 527, 530–542, 546, 550, 551, 555, 657, 677, 697, 707, 711, 717 – See also object – See also primary object – See also secondary object abstract case 163 accompaniment 21, 159, 165, 166, 180, 199, 202, 209, 218, 316, 318, 319, 474, 509, 518, 543 – accompaniment suffix 10, 159 – See also proprietive case – See also locative case accusative case 11, 162, 163, 304, 306, 308 – See also object action nominals 9, 106, 109, 126, 173, 210, 284, 285, 286, 382, 436, 437, 510, 513, 552, 627, 652, 712 – See also manner nominals adjective 107, 108, 285 – adjectival suffix 137, 156, 160 adjunct 107, 113, 132, 165, 166, 175, 179, 181, 184–186, 216, 218, 310, 318, 325, 435, 484, 501, 504, 506, 508, 509, 526, 548, 557, 617, 677, 692, 694 admonitives 190, 381, 416, 571, 641, 648, 654, 685 – See also commands adnominal 107, 123, 134, 140, 162, 166, 167, 179, 190, 191, 192, 202, 204, 208, 212, 214, 215, 216, 218, 234, 241, 313, 556, 557, 559, 564, 565, 566, 664 adverb 285, 435 adverbial preverbs 79, 82, 100, 109, 114, 115, 147, 148, 272, 280, 346, 424–427, 434, 435, 525, 532, 546, 629, 659, 689 – See also preverbs https://doi.org/10.1515/9783110752434-014

afterthought 131, 207 agentive 10, 21, 22, 107, 108, 127, 144, 383, 487 agreement 10, 110, 120, 130, 131, 140, 148, 162, 164, 166, 173, 195, 200, 216, 239, 286, 500, 513, 550, 562, 603, 629, 631, 696 alienable possession 456, 556, 563 – See also possession allative case 8–11, 160, 164–167, 176, 185–188, 192, 214, 215, 217, 227, 237, 240–245, 259, 261, 298, 299, 305, 316–318, 320, 326, 467, 470, 502, 508– 510, 514–516, 527, 529, 530, 532, 537, 557, 558, 675, 677, 691, 694–698, 704, 705, 707, 708, 710, 713, 715, 716 – subordinating function 691, 704, 707, 708 – See also subjunct – See also lative case allophony 53, 63, 64, 72 – of consonants 8, 51 – of vowels 8, 72, 73 animacy 154, 175, 318, 325, 498, 514, 536, 694 anticipatory inflection 390 – in related languages 416, 417 – See also TAM applicative 269, 501, 507, 544, 745 apposition 10, 12, 126, 134, 135, 142, 148, 548, 566 apprehension 380, 582 – See also commands – See also irrealis argument structure 11, 343, 393, 428, 429, 434, 435, 439, 467, 477, 479, 492, 498–501, 515–519, 544–548, 550–559, 561, 564–566, 617, 664, 671, 675, 677, 691–695, 697, 708–710, 716, 717 Arrernte 51, 63, 252, 393, 404, 529, 543, 604 ascriptives 107, 108, 110, 127, 133, 192, 433, 437, 516, 518, 520, 521 aspect (verbal) 115, 341, 349, 350, 353, 355–358, 361–363, 367, 369, 386, 403, 406, 415, 418, 423, 424, 441, 472, 568, 577, 578, 612, 614, 642, 653 – aspectual distinction 355–358, 363, 364, 405, 415 – See also imperfective – See also TAM assertive 600, 680

766 

 Index

assimilation 8, 52, 57, 78, 326, 406, 411 associated motion 12, 79, 115, 116, 341, 342, 347–351, 354, 358, 393, 402–412, 415, 416, 422–424 – prior 79, 404–406, – subsequent 404 – See also verbs > directional inflections – See also TAM associative suffix 107, 109, 113, 114, 138, 140–146, 194, 262, 265, 270, 383, 425 – subordinating function 695, 696, 703, 704, 713 assumptive 607 asyndetic coordination 12, 679 attenuative 10, 21, 157, 626 auxiliary 118, 567–569, 612, 646 avoidance 30, 181, 189, 323, 484, 514 Balgo 1–8, 12–14, 17–20, 22, 24–30, 36, 38–40, 151, 170, 199, 292, 363, 443, 613, 638, 639, 715 benefactives 165, 186, 204, 208, 242, 301, 309, 316, 332, 423, 504, 565, 710 – See also dative case Bilinarra 6, 7, 34, 47, 61, 64, 76, 85, 114, 129, 143, 146, 151, 152, 163, 169, 170, 188, 190, 199, 200, 228, 235, 237, 241, 245, 251, 261, 299, 301, 309, 318, 325, 333, 336, 347, 354, 356, 382, 392, 425, 445, 514, 515, 526, 562, 576, 685, 695, 697, 706, 707, 715 – See also Ngumpin Yapa (subgroup) Billiluna 1, 12, 14, 20, 21–25, 27, 29, 34, 39, 40, 207, 263, 415, 525, 596 bimoraic stems 49, 100, 223, 226 borrowing 49, 50, 141, 149, 188, 210, 219, 221, 228, 235, 275, 305, 359, 441, 443, 454, 491, 674, 707 bound pronouns 8–11, 79, 106, 110, 112, 116–121, 123, 132, 154, 163–165, 175, 187, 222, 236, 297–300, 304, 307–316, 318, 322, 323, 325, 326, 329, 330, 332, 333, 335, 338–340, 379, 423, 428, 430, 434, 498, 499, 501–503, 508, 512, 514, 515, 530, 535, 557, 561, 564, 566–569, 571, 574, 576, 577, 583, 598, 600, 614, 622, 627, 638, 641–642 – dual number replacement 333 – locational 316 – object/oblique, 315

– metathesis of 117, 299, 301, 305, 306, 326, 339, 340 – positioning 310, 311 – reciprocal 323 – reflexive 323 – subject 313 cardinal directions 253 – morphology of 258–261 case-marking 9, 11, 109, 114, 134, 141, 161–166, 249 case allomorphy 161, 168, 169, 229 case alternations 538, 540–543 case-stacking 204, 208, 212–218 case in subordinating function 11, 114, 116, 123, 159, 161, 167, 180, 184, 188, 191, 196, 342, 382, 403, 673, 675, 678, 691–696, 699, 703, 708, 709, 712 change of locative relation 430, 432, 445, 446, 452, 462, 463, 465, 470, 471, 476, 484, 525, 529 clause types – conative 165, 306, 504, 506, 516, 531, 532, 540–542, 546 – copula constructions 516, 520 – ditransitive 170, 172, 175, 315, 321, 502, 506, 532, 534–538, 548, 756 – extended intransitive 165, 170, 460, 477, 503, 504, 518, 527–529, 531 – impersonal 322, 402, 521 – intransitive 162, 165, 170, 171, 174, 199, 313, 314, 318, 343, 404, 412, 424, 429, 436, 444, 447, 452, 453, 460, 477, 503, 504, 506–508, 510, 518, 520, 521, 523, 525, 527–529, 542–544, 548 – quasi-transitive 165, 171, 501, 525, 526 – semi-ditransitive 175, 532 – transitive 109, 111, 117, 162, 164, 165, 170–174, 195, 200, 213, 218, 226, 232, 239, 249–251, 268, 285, 286, 313–315, 319, 324, 343, 404, 423–425, 429, 436, 441, 443, 452–454, 462, 476, 477, 489, 501, 506–508, 513, 521, 525, 526, 530, 531, 536, 537, 540, 542–546, 548, 550, 552, 554, 564, 645, 675, 700–702, 706, 709, 711, 713, 714, 716, 717 clause types – conative 165, 306, 504, 506, 516, 531, 532, 540–542, 546

Index 

– copula constructions 516, 520 – ditransitive 170, 172, 175, 315, 321, 502, 506, 532, 534–538, 548 – extended intransitive 165, 170, 460, 477, 503, 504, 518, 527–529, 531 – impersonal 322, 402, 521 – intransitive 162, 165, 170, 171, 174, 199, 313, 314, 318, 343, 404, 412, 424, 429, 436, 444, 447, 452, 453, 477, 506–508, 510, 520, 521, 523, 525, 527, 528, 542–544, 548, 554, 645, 675, 698 – quasi-transitive 165, 171, 501, 525, 526 – semi-ditransitive 175, 532 – transitive 109, 111, 117, 162, 164, 165, 170–174, 195, 200, 213, 218, 226, 232, 239, 249–251, 268, 285, 286, 313–315, 319, 324, 343, 404, 423–425, 429, 436, 441, 443, 452–454, 462, 489, 506–508, 513, 521, 525, 526, 530, 531, 536, 537, 540, 542–546, 548, 550, 552, 554, 564, 645, 675, 700–702 clitics 9, 120, 121, 611, 612 – See also Wackernagel clitics – See also bound pronouns code-switching 218, 228 – See also loan – See also borrowing cognate object 500, 526, 533, 745 commands 9, 119, 379, 383, 516, 567, 627, 629, 640–649 – See also imperative – See also admonitive – See also prohibitives comparatives 603, 620, 625, 626 – See also degrees complex NPs 9, 131 conditionals 667–673, 682 conjunction 607 consonants 8, 47 – clusters 88 – See also plosives – See also rhotics – See also liquids – See also nasals – See also semi-vowels coordination 12, 134, 593, 664, 679–682, 684, 686, 688 – apprehensional 684 – contrastive 686

 767

– NP coordination 132 – sequential 681 – subsequent 688 customary inflection 386, 387 – in related languages 388, 389 – See also present – See also presentative dative case 165–167, 175, 176, 178, 237, 239, 240, 242, 245, 274, 507, 519, 520, 709, 711, 716 – See also benefactive – See also indirect object – subordinating function 694, 709–711 degrees 148, 601, 603, 620, 625, 626 – See also comparative delimitive 598 ‘denizen’ suffix 70, 150, 151, 227, 256 – See also inhabitant suffix demonstratives 12, 106, 111, 128, 129, 131, 219, 223, 224, 226, 230, 231, 233–235, 246, 391, 392, 516, 517, 575, 662, 664, 665, 667, 706, 715 – adverbial 230 – distal series 228 – proximal series 226 depictives 109, 114, 121, 127, 171, 173, 199, 203, 209, 213, 285, 286, 296, 513, 552, 554, 633, 714 – See also secondary predication derivation 79 – derivational case 10, 122, 123, 140, 142, 148, 161, 162, 190–211, 214, 426, 695, 703 – derivational suffixes 10, 107, 108, 122, 123, 127, 140–152 diminutive 10, 155 directives 373, 396, 593, 641, 646, 647, 650 double object construction – See secondary object dual suffix 111, 154, 219, 233, 413 – See also number – See also bound pronouns > dual number replacement dubitative 119, 582, 583, 586, 588, 590, 591, 636 elative case 10, 166, 167, 193, 237, 240, 245, 274, 701 – See also ablative case – subordinating function 695, 701–703

768 

 Index

epenthesis 87 – bound pronoun 340 ergative case 10, 11, 78, 109, 111, 130, 162–167, 169, 171, 216, 237, 240, 242, 245, 247, 249, 511, 513, 696, 713, 716 – See also subject – See also bound pronouns > subject – on free pronouns 247, 249 – subordinating function 694, 696, 701–703, 706 evitative case 10, 166, 167, 169, 189, 237, 712 – subordinating function, 712, 713 ‘expert’ suffix, , 152, 153 external possession 133, 135, 165, 316, 504, 552, 556, 559, 562, 566 – See possession flaps 62, 68–71, 76, 591, 635, – See also rhoticf – See also retroflexion free pronoun 9, 11, 106, 110, 122, 128, 129, 147, 162–164, 173, 191, 203, 210, 219, 223, 234, 247–252, 299, 300, 335, 514, 538, 549, 561 grammatical relations 11, 116, 162, 166, 170, 180, 297, 317, 318, 321, 322, 324, 338, 343, 393, 419, 420, 422, 424, 428, 429, 434, 435, 439, 444, 445, 452, 460, 462, 467, 477, 479, 492, 498–566 – hierarchies of 333, 336 – non-canonical arguments, 511 – See also argument structure – See also secondary predication – See also adjunct – See also object – See also subject – See also subjunct ‘group’ suffix, 10, 149 – See also kinship > kinship group suffix Gurindji 7, 34, 38, 39, 47, 52, 54, 61, 126, 141, 143, 146, 151–153, 163, 176, 200, 205, 228, 235, 237, 241, 245, 251, 261, 292, 299, 303, 305, 306, 309, 330, 333, 336, 347, 354, 360, 392, 396, 416, 441, 514, 515, 562, 567, 571, 572, 576, 594, 665, 697, 715 – See also Ngumpin-Yapa (subgroup)

hence suffix 2, 8, 10, 12, 46, 61, 104, 116, 118, 157, 177, 214, 258, 266, 342, 349, 350, 352, 357, 361, 388, 391–401, 403, 409, 418, 474, 510, 513, 516, 573, 621, 641, 642, 649, 669, 705, 722 – See also hither suffix – See also verbs > directional inflections hierarchies 117, 299, 326, 332–336, 338, 500, 514, 541 – of bound pronouns 333 – of grammatical relations 333, 336 hither suffix 8, 12, 349, 350, 357, 361, 388, 393, 396, 398–400, 402, 417, 474, 642 – See also hence suffix – See also verbs > directional inflections hortative 349, 355, 359, 360, 371, 373, 375–379, 392, 398–400, 416, 417, 423, 572, 578, 591, 641, 646–648, 650, 654, 686 ideophones 9, 120 ignoratives 9, 111, 128, 236, 246, 595, 596 illocutionary 373, 379, 385, 516, 614, 627, 629, 640–642, 644–646, 649–651 imperative 8, 9, 104, 135, 212, 300, 348–350, 353, 355, 360, 361, 363, 379, 383, 392, 394, 396–398, 415, 454, 571, 572, 578, 593, 616, 641–646, 650, 652, 653 – See also commands imperfective 21, 350, 355, 356, 358–363, 366, 367, 369–371, 373–375, 377, 380–384, 390, 392, 396–400, 404, 406–408, 415, 423, 577, 583, 606, 614, 642, 717 – See also TAM inclusory constructions 132, 133 incremental suffix 626 indirect object 110, 116, 165, 175, 176, 178, 187, 215, 268, 298, 306, 314–317, 321, 323, 325, 332, 457, 498, 499, 501–508, 510, 511, 514, 515, 519, 520, 527, 528, 531, 532, 534, 536, 543, 546, 547, 557, 624, 657, 671, 675 – See also dative case inhabitant suffix 151 – See also denizen suffix insubordination 383, 645, 685 interjections 9, 81, 119, 120, 629 internal possession 134, 165, 167, 178, 179, 203, 207, 316, 506, 556–558, 561–564

Index 

interrogatives 111, 112, 118, 177, 191, 195, 233, 236, 238, 241–246, 310, 568, 571, 578, 581, 582, 585–588, 591, 594, 595, 603, 632, 634, 641, 654, 659, 661–663, 665, 667, 678, 679, 764 intervocalic (plosive) 46, 47, 61, 71 intonation 9, 102, 104 irrealis 380, 416, 748 – See also potential – See also TAM irregular verb 343 iterative function 137, 281, 369, 418, 623 – iterative verbal clitic 621, 622 Jaru 3, 4, 6–9, 21, 22, 24, 26, 28, 30, 32–34, 36, 39, 40, 45, 47, 50, 64, 70, 71, 76, 85, 86, 113, 114, 128, 131, 135, 143, 146, 150, 152, 155, 156, 163, 169, 181, 188, 190, 200, 205, 219, 223, 226–228, 230, 235, 241, 247, 248, 250–252, 254–256, 275, 288, 289, 294, 299, 301, 303, 305–307, 309, 318, 328, 331, 333, 340, 347, 354, 355, 359, 360, 362, 368, 369, 373, 375–377, 382, 385, 387–389, 391, 392, 396, 398, 405, 414, 415, 435, 437, 438, 453, 454, 468, 478, 481, 482, 484, 486, 502, 509, 515, 522, 532, 543, 558, 561, 564, 567, 571, 591, 615, 616, 632, 634, 638, 642, 651, 656, 677, 685, 697, 704–706, 749, 762, 763 – See also Ngumpin-Yapa (subgroup) kinship 122, 152, 287, 292–294, 296, 555, 556, 622 – ‘group suffix’ 10, 152 – bereft 296 – dyadic suffix 293, 294, 622 – marriage partners 287 – suffixes 10, 152 Kriol 1, 30, 90–92, 106, 144, 145, 149, 159, 206, 207, 272, 274, 286, 341, 374, 419–421, 436, 440–444, 454, 460, 471, 477, 487, 553, 567, 572, 591, 597, 618, 619, 631, 635–639, 657, 661, 679, 688–690, 702, 707 Kukatja 3, 5, 8, 15, 18, 19, 21, 22, 24–29, 32, 33, 38–40, 88, 136, 143, 145, 149, 163, 189, 199, 200, 210, 219, 221, 255, 262, 288, 304, 305, 323, 355, 375, 401, 416–418, 420, 515, 562, 603, 614, 631, 632, 651, 656, 657, 659, 661, 667, 673, 674 – See also Western Desert (languages)

 769

‘lacking’ suffix 166, 167, 210, 599 – See also Privative case laterals – See liquids lative case 10, 166, 167, 188, 240 – See also allative case lenition 9, 52, 54, 60–62, 78, 188, 349, 628 – See also plosives like suffix 10, 149 liquids 63, 71, 80, 88, 90–93 loan 51, 131, 143, 156, 168, 169, 173, 183, 420, 440–444, 450, 454, 461, 477, 572, 591 – See also borrowing – See also code-switching local case 194, 514 locationals 267 locative case 10, 50, 165–167, 169, 180, 181, 237, 240, 245, 274, 699, 704, 716 – See also subjunct – Subordinating function 696–699, 700, 704, 705 loose nexus preverbs – See weak nexus preverbs Malngin 6, 7, 34, 36, 47, 152, 156, 188, 200, 226, 251, 354 – See also Ngumpin-Yapa (subgroup) manner nominals 106, 109, 122, 137, 173, 286, 549 – See also action nominals Manyjilyjarra 25, 131, 189, 312, 406, 416, 417, 691 – See also Western Desert (languages) metathesis 8, 74, 75, 77, 255, 631 – See also bound pronoun > metathesis of missionaries 19, 22 modal particle 79, 82, 117–120, 310, 311, 364, 371, 372, 391, 567–569, 572, 578, 580–582, 584, 586, 593–598, 601, 605, 618, 636, 642, 652, 653, 660, 661, 672, 680, 684 mood – dubitative 588 – hypothetical 591, 592 – uncertain 582 – unrealised 586 – See TAM – See irrealis – See obligative

770 

 Index

Mudburra 2, 6, 7, 34, 47, 135, 151, 163, 190, 200, 205, 235, 309, 313, 333, 347, 354, 359, 368, 369, 392, 393, 396, 441, 514, 515, 529, 562, 566, 567, 571, 572, 576, 594, 645 – See also Ngumpin-Yapa (subgroup) narrative inflection 79, 349, 354, 355, 357–359, 362, 367, 373, 374, 378, 392, 415, 423, 583, 606, 639, 640 – See also TAM nasals 47, 48, 53, 63, 64, 88, 169, 340 negation 651–653, 655–658 Ngarinyman 6, 7, 681 – See also Ngumpin-Yapa (subgroup) Ngumpin-Yapa (subgroup) 1, 6, 7, 44, 151, 230, 237, 298, 416, 420, 421, 562 – See also Bilinarra – See also Gurindji – See also Jaru – See also Malngin – See also Mudburra – See also Ngarinyman – See also Nyininy – See also Walmajarri – See also Warlmanpa – See also Warlpiri nominals 9, 11, 82, 88, 106, 122, 128, 129, 141, 148, 163, 171, 203, 219, 421, 435, 437, 439 – action nominals 9, 106, 109, 126, 173, 210, 284–286, 382, 436, 437, 510, 513, 552, 627, 652, 712 – illocutionary nominals 649 – manner nominals 106, 109, 122, 137, 173, 286, 549 non-configurational properties 11, 125, 127 Northern Territory 1, 20, 27, 30, 39 noun phrase 9–11, 122, 125, 127, 128, 130, 131, 133, 141, 153, 161, 164, 173, 185, 192, 207, 212, 548, 555, 557 – complex NPs 9, 131 – discontinuous NPs 126 NP-relative clause 664, 675 – See also relative clause number 4, 10, 153, 216, 300, 304, 348, 426, 442 – dual number replacement 233, 248, 326, 328, 333, 335, 338, 535

– number suffix 110, 122, 123, 143, 153, 202, 219, 248 – See also dual suffix – See also paucal suffix Nyininy 2–4, 6, 8, 9, 20, 34, 36, 71, 250, 252, 305, 387–389, 392, 415, 416, 681 – See also Ngumpin-Yapa (subgroup) object 21, 12, 78, 110, 112, 116–118, 134, 145, 158, 162, 164, 165, 171, 175, 176, 181, 184, 185, 187, 188, 203, 208–210, 215, 222, 223, 233, 259, 261, 268, 270, 271, 297–299, 301–309, 314–317, 321, 323–334, 336–340, 365, 403, 404, 423, 436, 443, 456, 457, 462, 467, 473, 476, 481–486, 489, 498–507, 511, 512, 514, 515, 519, 520, 525–528, 530, 532–537, 542, 543, 546, 547, 550, 551, 553–555, 557–562, 564, 566, 643, 657, 671, 691, 695, 697, 698, 701, 704, 707, 708, 710, 716 – See also absolutive case – See also primary object – See also secondary object – See also indirect object obligative 21, 349, 355, 360, 363, 364, 371, 374, 375, 391, 416, 417, 578, 592, 654 – in related languages 416, 417 – See also TAM obviative 664, 695, 696, 713–715, 716, 213 – See also subordination > non-finite orthography 2, 41, 44–48, 71 other suffix 10, 146 Pama-Nyungan 1, 3, 5, 9, 18, 51, 75, 96, 125, 186, 250, 353, 354, 404, 417, 420, 499, 642 particles – modal 9, 81, 119, 583, 594, 612 past tense 7, 28, 359, 363, 589, 590, 668, 682, 683 paucal suffix 21, 10, 92, 117, 153, 154, 190, 225, 271, 332, 335, 547, 608 – See also number perfective 350, 355, 358, 363–365, 369, 374, 375, 381, 392, 406, 415 perlative case 10, 166, 167, 197, 216, 245 personal possessive 166, 206 phonation – creaky voice 65, 72, 120 – voicing 2, 9, 46, 52–54, 61, 62

Index 

phoneme – alternations 8, 76 – contrasts 8, 47 – inventory 8, 44 phonotactics 8, 84, 86, 88 Pintupi 1, 3, 15, 16, 18, 20, 24, 30, 39, 45, 71, 130, 144, 149, 163, 190, 199, 200, 210, 219, 247, 262, 304, 312, 323, 355, 365, 375, 416, 417, 420, 515, 558 – See also Western Desert (languages) plosives 9, 90, 349 – voicing 2, 9, 46, 52–54, 61, 62 – See also lenition polar question 585, 630, 656, 660 possession 133–135, 165, 167, 174, 178, 179, 199, 201, 203–205, 207, 212, 218, 313, 316, 322, 446, 450, 456–458, 466, 467, 498, 504, 506, 518, 547, 552, 555–559, 561–564, 566, 657 – alienable possession 456, 556, 563 – external possession 133, 135, 165, 316, 504, 552, 556, 559, 562, 566 – internal possession 134, 165, 167, 178, 179, 203, 207, 316, 506, 556–558, 561–564 – prominent external possession 134, 165, 178, 179, 316, 562–564 possessive case 128, 166, 167, 203, 555 potential 8, 9, 94, 190, 274, 277, 343, 348–350, 355, 360, 371–375, 377, 379, 380, 388, 392, 398, 399, 407, 408, 415–417, 423, 572, 574, 578, 579, 584, 586, 588, 589, 591, 600, 621, 641, 646–648, 650, 654, 656, 668, 669, 671–673, 685, 686 – See also irrealis – See also TAM predication – See also secondary predication – ascriptive 517 – equative 517 – identifications 517 – non-verbal, 516 preparative 213,695, 696, 709, 711, 717 – See also dative case > subordinating function present 353, 384, 385, 388, 670 – historic use 9, 12, 14, 248, 349, 386, 401, 405 – See also TAM presentative inflection 349, 350, 360, 387, 392, 401, 402, 416 – See also verbs > directional inflections

 771

preverbs 21, 9, 11, 51, 76, 79, 81–83, 87, 88, 91, 98–101, 106, 109, 113–116, 137, 140, 148, 161, 184, 285, 341, 346, 357, 391, 412, 419–421, 424–426, 428–431, 433–435, 437, 440, 442, 445–448, 451, 452, 454, 455, 462, 465, 467, 468, 471, 473, 475, 477, 478, 489, 490, 494–496, 510, 513, 516, 525, 544, 545, 549, 552, 621, 622, 658, 694, 702, 706, 707, 710 – See also adverbial preverbs – See also loose nexus preverbs – See also strong nexus preverbs primary object 165, 171, 315, 511, 512, 534, 535 – See also object – See also secondary object privative case 166, 167, 208, 209, 708, 709 – See also lacking suffix – subordinating function 708, 709, 715, 716 productive verbs 419, 439, 445, 446, 470, 485, 489 progressive 366, 370, 385, 388, 393, 402, 472 prohibitives 210, 417, 641, 644, 645, 652, 709 – See also commands prominent internal possession 134, 165, 178, 179, 316, 562–564 pronominal enclitics – See bound pronouns proprietive case 10, 166, 167, 199, 237, 240, 538 – See also accompaniment – See also locative case quantifiers 9, 109, 110, 129, 153, 219, 221 – See also number quotative 604 reduplication 10, 93, 122, 135–139, 154, 342, 418, 430, 431 relative clause 118, 235, 664, 666, 674, 675, 678 – See also NP-relative clause – See also T-relative clause restrictive 121, 611, 614, 617, 624, 637, 675, 715 retroflexion 9, 45–47, 49, 52, 56, 61, 64, 65, 70, 71, 75, 76, 84–86, 199, 298, 307, 638 rhotics 74–76, 97, 635 – See liquids

772 

 Index

second position 117, 310, 567, 570, 578, 582, 642, 665–667 – See also Wackernagel clitics secondary object 171, 315, 500, 511–513, 534, 535 – See also object – See also primary object – See also indirect object secondary predicates 498, 548, 761 – depictives 109, 114, 121, 127, 171, 173, 199, 203, 209, 213, 285, 286, 296, 513, 552, 554, 633, 714 semblative case 190, 191, 193, 201, 203, 207, semi-vowels, 65, 74, 80, 85, 88, 97, 347, 349, 611 serial verbs 368, 419, 421–423, 643, 682 sequential locative 696, 698, 699–701 – See also locative > subordinating function skin name 292 sonority 90, spatial case 10, 112, 122, 166, 197, 198, 213, 215, 217, 226, 227, 240, 243, 268, 273, 306, 435, 456, 465, 509, 551, 698 stress 9, 44, 58, 93–97, 99–101, 428, 430, 434 – acoustic correlates of 94 subject 116, 164, 185, 315, 324, 325, 498, 501, 508, 518, 527, 529, 531, 534 subjunct 116, 165, 180, 181, 185, 186, 243, 309, 316–322, 326, 338, 498, 507, 509, 529, 530, 532, 536, 537, 540, 543, 547, 561 – See also locative case – See also allative case subordination 12, 378, 382, 407, 533, 664, 665, 666, 679, 696 – finite 12, 664, 665, 679–690 – non-finite, 691–717 subordinators 9, 118, 691–717 subsection 92, 107, 116, 129, 131, 132, 134, 141, 145, 148, 152, 157, 180, 181, 194, 203, 249, 287, 288, 292, 293, 296, 320, 386, 401, 446, 447, 504, 517, 528, 539, 557, 558, 565, 592, 599, 602, 609, 626, 689 syllable 44, 52, 57, 67, 71, 74, 76, 80–84, 87, 93–96, 98–102, 139, 152, 188, 340, 347, 349, 354, 355, 418, 627, 631 – deletion 8, 75 – structure 81, 82, 98

Tag questions 102, 103, 119, 581, 582, 585, 586, 588, 596, 599, 603, 607, 632, 659–661 – See also interrogatives TAM 8, 12, 115, 342, 348–350, 353–355, 357, 360–362, 382, 391, 392, 403, 405, 407, 408, 411, 413, 415, 418–420, 422, 423, 516, 521, 568, 577, 582, 583, 586, 588, 591, 606, 621, 660, 668, 670, 682, 685, 691 – in a regional perspective 415–418 – See also aspect (verbal) – See also anticipatory – See also mood – See also narrative inflection – See also obligative – See also present – See also imperfective – See also irrealis – See also verbs > directional inflections temporals 273, 278, 284, 365 strong nexus preverbs 81, 114, 357, 425–430, 442, 710, – See also preverbs T-relative clause, 664, 675, 676 – See also relative clause valency 343, 346, 495, 500, 501, 521–523, 538, 539, 544–546 – alternation 538 verbalisation 437 verbs 341–494 – complex verb construction 341, 424–439, 544, 545 – directional inflections 360, 391–402 – inflecting 11, 12, 100, 115, 135, 141, 257, 285, 341, 343, 344, 346–348, 357, 403, 418, 419, 421, 424, 425, 428–430, 434–439, 441, 442, 444–446, 454, 465, 477, 479, 489, 495, 496, 501, 520, 541, 544, 545, 622 – Kriol loans 341, 419, 421, 440–444, 454 – parasitic inflections 359, 360 – serial 419, 421, 422 – See also associated motion – See also TAM verbs of contact/force 476, 532 verbs of locomotion 393, 452, 458, 470, 471, 474, 525, 529 vocative 9, 102, 104, 105, 117, 120, 121, 170, 611, 627, 629, 631, 635, 643, 649, 651, 655

Index 

Wackernagel clitic 297, 310 – See also bound pronouns Walmajarri 1–3, 5–9, 16, 21, 22, 24, 25, 28, 30, 32, 34, 38, 39, 47, 61, 70, 71, 85, 86, 90, 130, 131, 146, 159, 163, 188–190, 200, 219, 224, 226, 228, 230, 235, 241, 252, 272, 288, 299, 301, 305–309, 322, 347, 354, 356, 359, 369, 380, 387, 388, 392, 396, 415, 416, 421, 430, 433, 436, 438, 448, 451, 477, 479, 483, 486, 509, 515, 543, 548, 568, 569, 571, 574, 588, 591, 594, 603, 605, 632, 634, 636, 637, 643, 651, 653, 659, 661, 667, 699, 706 – See also Ngumpin-Yapa (subgroup) Wangkajunga 1, 3, 24, 30, 45, 75, 120, 129, 143, 144, 149, 159, 181, 186, 189, 190, 194, 201, 210, 219, 230, 235, 285, 364, 375, 378, 406, 416, 417, 420, 421, 435, 437, 445, 448, 453, 477, 496, 515, 543, 607, 642, 661, 667, 688, 692 – See also Western Desert (languages) Wanparta 5 Wanyjirra 7, 36, 45, 47, 52, 70, 76, 85, 97, 114, 125, 126, 128–130, 135, 146, 147, 156, 163, 169, 183, 188, 190, 191, 193, 199, 200, 205, 211, 219, 226, 228, 241, 245, 251, 294, 299, 301, 305–309, 325, 333, 347, 354, 360, 375, 376, 382, 391, 392, 405, 416, 435, 438, 442, 514, 515, 526, 540, 548, 571, 572, 576, 582, 594, 634, 656, 665, 677, 685, 697, 713, 715 Waringari 5 Warlmanpa 3, 6, 47, 145, 151, 200, 227, 309, 347, 348, 354, 391, 441, 514, 543, 612, 613, 645, 706 – See also Ngumpin-Yapa (subgroup) Warlpiri 2–6, 8–11, 19, 21, 25, 28–30, 32, 33, 36–39, 45–47, 49, 63, 65, 70, 71, 79, 80, 83, 85, 86, 94, 98–100, 104, 106, 113, 114, 117, 125–127, 130, 131, 135, 136, 143, 145, 149–152, 155, 157–163, 174, 181, 186–190, 193–195, 199, 200, 203, 207, 215, 216, 218–222, 226, 228–230, 235, 237, 241, 248–255, 260, 261, 277, 285, 287–289, 292, 294, 295, 301–303, 306, 309–311, 320, 325, 330, 331, 333, 346–348, 370, 380, 391, 392, 394, 400, 401, 403,

 773

405–407, 410–412, 415, 416, 418, 421, 435, 437, 438, 441, 454, 486, 491, 494, 496, 497, 500, 504, 505, 513–517, 524, 528, 531, 532, 540, 541, 543, 551, 552, 554, 568, 577, 581, 582, 597, 599, 600, 603–605, 607, 611–613, 621, 623, 625, 628, 632, 633, 642, 645, 661, 666, 677, 686, 691, 694–696, 698, 699, 701, 703, 705–709, 711, 714, 715 – See also Ngumpin-Yapa (subgroup) Warnayaka 5, 38 warnings 190, 381, 416, 571, 641, 648, 654, 685 Warnman 151, 566 Warrmarla 5 weak nexus preverbs 79,–82, 100, 113–115, 425–417, 429–433, 552, 662, – See also preverbs Western Australia 1, 12, 19, 29, 40 Western Desert (languages) 18, 29, 45, 71, 88, 129, 130, 143, 149, 151, 163, 181, 189, 194, 200, 210, 219, 234, 235, 247, 262, 288, 304, 312, 323, 348, 355, 360, 364, 375, 406, 407, 412, 415–418, 420, 421, 435, 438, 448, 453, 477, 496, 514, 515, 619, 635, 642, 673, 692 – See also Kukatja – See also Manyjilyjarra – See also Pintupi – See also Wangkajunga – See also Warnman – See also Yankunytjatjara – See also Yulparija word classes 9, 106 word order 11, 115, 125, 310, 311, 567, 692 Yagga Yagga 3, 13, 27, 40, 142, 151, 182, 192, 316 Yankunytjatjara 130, 144, 210, 285, 406, 412, 421, 514, 642 – See also Western Desert (languages) Yulparija 3, 6, 143, 190, 194, 219, 375, 416, 691, 747 – See also Western Desert (languages) zero-derivation, 438