Complete Dissertation

VU Research Portal

A Linguistic History of Awyu-Dumut Wester, R.

2014

document version Publisher's PDF, also known as Version of record

Link to publication in VU Research Portal

citation for published version (APA) Wester, R. (2014). A Linguistic History of Awyu-Dumut: morphological study and reconstruction of a Papuan language family.

General rights Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal ?

Take down policy If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

E-mail address: [email protected]

Download date: 29. Sep. 2021 A Linguistic History of Awyu-Dumut morphological study and reconstruction of a Papuan language family c 2014, Ruth Wester Cover: artwork from Eastern Highlands Province, Papua New Guinea, 1984 Cover design: Flip Wester sr. and Ridderprint BV Typeset in LATEX Printed and bound by Ridderprint BV, Ridderkerk ISBN: 978-90-5335-793-4 VRIJEUNIVERSITEIT

A Linguistic History of Awyu-Dumut morphological study and reconstruction of a Papuan language family

ACADEMISCHPROEFSCHRIFT

ter verkrijging van de graad Doctor aan de Vrije Universiteit Amsterdam, op gezag van de rector magniﬁcus prof.dr. F.A. van der Duyn Schouten, in het openbaar te verdedigen ten overstaan van de promotiecommissie van de Faculteit der Letteren op donderdag 20 maart 2014 om 15.45 uur in de aula van de universiteit, De Boelelaan 1105

door

Ruth Wester

geboren te Ukarumpa, Papoea Nieuw Guinea promotor: prof. dr. L.J. de Vries copromotor: dr. G.P. Reesink Acknowledgments

The writing of a PhD thesis can be compared to going on a long journey, in which the traveller amasses numerous experiences and learns about herself in the process. I can honestly say that writing “A Linguistic History of Awyu-Dumut Languages” has been a pleasant journey, no doubt largely due not only to the exciting research topic, but also to all those who travelled with me, supporting and encouraging me along the way. I would like to begin by acknowledging the excellent work done by Petrus Drabbe, a Catholic priest of the Missionaries of the Sacred Heart in New Guinea between 1935 and 1960. Each page of his clear and detailed grammar descriptions communicates his love of languages and his deep insight into how they work. I could not have wished for better data, and hope this book does his life’s work at least some justice. My promotor Lourens de Vries is the one who got me started on this journey, and I cannot thank him enough. After Lourens secured the funding by the Nether- lands Research Council (NWO) for the project “The Awyu-Dumut language family in its cultural and linguistic context”, I joined him and Wilco van den Heuvel as the third member of the ‘Awyu-Dumut family’ at the VU University in Amsterdam. Lourens, thank you for your contagious enthusiasm, your ability to always inspire, your encouragement whenever I was stuck, and your availability and willingness to answer questions whenever I felt like asking them. Thanks also for graciously let- ting me disagree with you on the origins of Awyu-Dumut switch reference. I have learned much from you, not just about linguistics but also about life, and will miss our collaboration. Ever since I can remember, wantok is the term used by those with the surname ‘Wester’ to refer to those with the surname ‘Reesink.’ Ger, you have been an excellent wantok the past four years, and going back to the origin of this Tok Pisin word, I would like to thank you for ‘speaking the same language’ as I. The endless hours you selflessly put in as co-promotor, as well as your wisdom, insight and eye for detail, will not quickly be forgotten. Thank you for coming along on this journey. The highlight of my PhD journey was without a doubt my three-month stay at the Australian National University in Canberra, funded by the Australian Nether- lands Research Collaboration (ANRC) and the VU University Amsterdam. The researchers at the ANU College of Asia and the Pacific provided a welcoming and sti- mulating research environment. I would like to thank Nick Evans, Andrew Pawley, Malcolm Ross, Harold Koch, Mark Donahue, Bethwyn Evans, Gwendolyn Hyslop, Alexander François, Julia Miller, Christian Döhler, Darja Hoenigman, Fanny Cottet, Niko Kobepa, Aung Si, Charlotte van Tongeren and Matthew Carroll for their interest in and comments on my research, and for sharing their own research with me. A special thank you goes to my housemates at Wongoola Close for making my time in Australia unforgettable. However, most of my time was not spent in Australia but in Amsterdam at the VU. Bertie Kaal and Agata Cybulska, thank you for being such great roommates, putting up with my sighs, exclamations and inexplicable stories about Drabbe and Awyu-Dumut languages, and for enjoying so many coffees (or ‘wiener melanges’) with me. Good luck on finishing your own PhD journeys. Wilco, thank you for your continual support, your ability to listen, your gentle encouragement along the way and for always believing my research was going well and meaning something. Thank you also for your meticulous – and fast – reading of the entire draft version just before your summer holidays. I have thoroughly enjoyed working together and can only hope I will have colleagues like you in the future. I would like to thank my reading committee for taking the time to read and comment on the thesis. An additional thank you to Laura Robinson for commenting on draft versions of chapters 4 and 6 at an early stage, giving me much-needed confidence, and mercietjes to Michael Dunn for helping me with the phylogenetics section. I would also like to thank my friend and fellow linguist Cindy Groff Heiner for editing my thesis. Throughout my journey I have been supported by my wider community of friends and family. I am grateful to my parents for an upbringing that instilled a deep appreciation of diversity in me, and to my dad for suggesting I study linguistics when I was 16 years old. Furthermore, I would like to thank my eldest brother Flip for making it look so difficult to do a PhD, as that made it much eas- ier. And thank you Thomas for your timely suggestion that I use LATEX instead of Word to format this book, saving me numerous frustrations and opening up a whole new world. A big thank you also goes to all my friends and family members who cheered me on, both to those who were truly intrigued and to those who still cannot remember the names ‘Awyu-Dumut’ and ‘Drabbe’ after four years. Lastly, I would like to thank Gerben Dekker. Your belief in me and your unwavering love kept and keep me going. You are the very best travel companion, and I look forward to the many journeys ahead that we will share.

Oslo, January 2014 Ruth Wester Contents

Introduction1

1 Setting the Scene3 1.1 Location and Sources of Data ...... 3 1.2 Previous Study of Awyu-Dumut Languages...... 10 1.3 Awyu-Dumut Languages are Constructs...... 14 1.4 Methodology: the Comparative Method...... 16 1.5 The Structure of the Book...... 19 1.6 A Note on Glossing...... 20

2 Proto Phonology and Phylogenetics 21 2.1 Data...... 22 2.2 Orthography and Morphophonemics...... 22 2.3 Sound Correspondences...... 24 2.4 PAD Consonant Reconstruction ...... 33 2.5 Note on Korowai ...... 33 2.6 Sound Changes and Subgrouping...... 34 2.7 Note on Awyu-Dumut Vowels...... 36 2.8 Phylogenetics ...... 40 2.9 Summary...... 47

3 Nouns and Adjectives 49 3.1 Adjectives ...... 49 3.2 Compound Nouns ...... 50 3.3 Possession ...... 53 3.4 Plurality ...... 53 3.5 Kinship Terms...... 54 3.6 Coordination of Nouns...... 57 3.7 Summary...... 64

4 Awyu-Dumut Pronouns 65 4.1 Reconstruction of Awyu-Dumut Personal Pronouns...... 65 4.2 Awyu-Dumut Possessive Pronouns...... 72 4.3 Awyu-Dumut Emphatic Pronouns ...... 74 4.4 Summary...... 75

5 Subject Person-Number Marking 77 5.1 First Person Singular Subject Marker...... 78 5.2 Non-ﬁrst Person Singular Subject Marker...... 80 5.3 First Person Plural Subject Marker ...... 83 5.4 Non-ﬁrst Person Plural Subject Marker...... 84 5.5 Summary...... 85

6 Mood 87 6.1 Verb Stems and Mood...... 88 6.2 Irrealis Mood...... 91 6.3 Realis Mood...... 93 6.4 Reconstruction...... 102

7 Tense 105 7.1 Past Tense ...... 106 7.2 Future Tense...... 113 7.3 Summary...... 116

8 Aspect 117 8.1 Position Verbs...... 117 8.2 Iterative Verb Stems...... 121 8.3 Connective Verbs ...... 123 8.4 Reconstruction...... 126

9 Negation 127 9.1 Awyu Negation...... 127 9.2 Dumut Negation ...... 133 9.3 Kombai Negation...... 139 9.4 Reconstruction...... 140

10 Deictics and Demonstratives 141 10.1 Awyu-Dumut Basic Deictics...... 141 10.2 Awyu-Dumut Demonstratives...... 147 10.3 Awyu-Dumut Textual Deixis...... 149 10.4 Summary...... 153

11 Clause Linkage 155 11.1 Coordination...... 155 11.2 Clause Chaining...... 157 11.3 Subordination...... 158 11.4 Tail-head Linkage...... 171 11.5 Awyu-Dumut Switch Reference Systems ...... 173 11.6 Summary...... 188

Conclusion 191

Appendix A: Word List 195

Bibliography 211

Samenvatting (Dutch Summary) 219 List of Tables

1.1 The comparative method...... 16

2.1 Awyu-Dumut phonemes and orthography ...... 22 2.2 Awyu-Dumut common morphophonemic changes...... 23 2.3 Sound correspondences ...... 24 2.4 Reflexes of *p ...... 26 2.5 Initial /p/ in four languages...... 26 2.6 Reflexes of *t...... 27 2.7 Reflexes of *k...... 28 2.8 Reflexes of *mb ...... 29 2.9 Reflexes of *nd...... 29 2.10 Reflexes of *ŋg...... 30 2.11 Reflexes of *m...... 31 2.12 Reflexes of *n ...... 31 2.13 Reflexes of *r...... 32 2.14 Reflexes of initial *w and *y ...... 32 2.15 PAD consonant inventory...... 33 2.16 Shiaxa epenthetic vowel examples ...... 35 2.17 Vowels in PA, PD and Kombai...... 37 2.18 Examples of PD *u >Mandobo /ö/...... 38 2.19 PD *ü >YWB and DWB /i/...... 38 2.20 PD *a >MAN /o/ ...... 38 2.21 PD *a to MAN /e/...... 39 2.22 PA *ü to /u/ in PSA, SHI, YEN ...... 39 2.23 PA *u to Shiaxa /o/...... 39 2.24 PA *e to YEN, PSA, AXU /i/...... 40 2.25 Lexicostatistic results of Healey and Voorhoeve...... 40 2.26 Coded lexical items showing various cognate patterns...... 41 2.27 Cognate sets from Table 2.26 expressed in a binary matrix...... 43

3.1 Dumut and Kombai intensiﬁers...... 50 3.2 Awyu-Dumut endocentric nouns...... 52 3.3 Awyu-Dumut coordinate compound nouns...... 52 3.4 Awyu-Dumut kinship terms...... 55 3.5 Awyu-Dumut kinship plural markers ...... 56

4.1 Greater Awyu and Awyu-Dumut pronouns...... 66 4.2 Awyu-Dumut and Trans New Guinea proto pronouns...... 71 4.3 Awyu-Dumut possessive pronouns...... 73 4.4 PAD possessive pronouns compared to PAD and PTNG personal pronouns 74

5.1 Awyu-Dumut first person singular subject markers ...... 78 5.2 Shiaxa epenthetic vowel examples ...... 79 5.3 Awyu-Dumut non-first person singular subject markers...... 81 5.4 Future NON1SG Forms...... 82 5.5 Two analyses of realis mood in Awyu languages...... 82 5.6 Awyu-Dumut first person plural subject markers...... 83 5.7 Awyu-Dumut non-first person plural subject markers...... 85

6.1 Awyu primary and secondary stems...... 89 6.2 Additional Pisa and Shiaxa verb stems...... 90 6.3 Awyu-Dumut semi-ﬁnite irrealis forms ...... 92 6.4 Dumut realis paradigms...... 95 6.5 Comparison of -t and -ken forms in Yonggom Wambon texts ...... 99 6.6 Awyu realis paradigms...... 101 6.7 Kombai realis paradigms...... 102

7.1 Dumut past paradigms...... 107 7.2 Shiaxa and Yenimu hodiernal past paradigms ...... 108 7.3 Shiaxa person-number markers...... 109 7.4 Awyu hesternal past paradigms...... 109 7.5 Awyu distant past paradigms...... 110 7.6 Aghu distant past paradigm...... 111 7.7 Awyu historical past paradigms...... 112 7.8 Awyu past tenses ...... 112 7.9 Dumut future tense paradigms ...... 114 7.10 Digul Wambon immediate future paradigm ...... 115

8.1 Durative markers and verbs in Awyu-Dumut languages ...... 120 8.2 Awyu-Dumut iterative verb stems ...... 121 8.3 Awyu-Dumut connective verbs...... 123

9.1 Aghu negated paradigms ...... 129 9.2 Pisa negated semi-finite paradigms...... 130 9.3 Pisa negated finite paradigms...... 130 9.4 Dumut negation strategies...... 133 9.5 Mandobo negated future tense paradigm ...... 135 9.6 Digul Wambon negated irrealis semi-finite paradigm ...... 138 9.7 Digul Wambon negated irrealis semi-finite ...... 138 9.8 Dumut negation strategies...... 139

10.1 Awyu-Dumut deictics ...... 142 10.2 Reconstructed Awyu-Dumut deictics...... 146 10.3 Awyu-Dumut demonstratives...... 147

11.1 Temporality and conceptual non-close markers found on SS verbs . . . . 178 List of Figures

1.1 Approximate location of the Awyu-Dumut language family...... 4 1.2 Location of Awyu-Dumut languages...... 5 1.3 Greater Awyu language family ...... 5 1.4 Portrait of Petrus Drabbe...... 10

2.1 NeigborNet network graph of Awyu-Dumut languages...... 44 2.2 Bayesian rooted tree based on 430-item word list...... 46 2.3 Bayesian rooted tree based on Swadesh subset of word list ...... 47

4.1 Greater Awyu language family ...... 67

7.1 Awyu-Dumut verb structure...... 105 Abbreviations

ADJ Adjective REAL Realis CAUS Causative SEQ Sequence marker ERG Ergative marker SG Singular CONN Connective SIM Simultaneity marker COORD Coordinator SS Same subject COP Copula SUBJ Subject marker DIST.PST Distant past SUPP Support verb DS Different subject TOP Topic marker DUR Durative ERG Ergative marker AXU Aghu language F Final DWB Digul Wambon language FOC Focus marker MAN Mandobo language FUT Future KOM Kombai language HAB Habitual KOR Korowai language HIST Historical past KYD Komyandaret language I Initial PA Proto Awyu IMP Imperative PAD Proto Awyu-Dumut IRR Irrealis PD Proto Dumut IT Iterative PSA Pisa language INTENS Intensiﬁer PTNG Proto Trans New Guinea LIG Ligature SHI Shiaxa language LOC Locative TSA Tsaukwambo language M Medial WNG Wanggom language NEG Negator YEN Yenimu language NON1 Non-ﬁrst person YWB Yonggom Wambon language NON.CLOSE Conceptually non-close PST Past EB elder brother PL Plural EZ elder sister PN Person-number F father POSS Possessive FF father’s father FM father’s mother WM wife’s mother M mother YB younger brother MB mother’s brother YZ MF mother’s father younger sister MM mother’s mother ZS sister’s son

Introduction

All that is has a history, a story about where or what it came from, and how it became what it is. Each language, spoken by generation upon generation, has a history. This book presents the linguistic history of a Papuan language family called Awyu-Dumut. It focuses on the history of the morphology within the Awyu-Dumut language family, bound morphology being one of the most stable elements within a language. Through bottom-up reconstruction of proto morphology, I establish where Awyu-Dumut morphology came from and what it might have looked like originally. However, just knowing the source or origin of a language does not tell its complete history: each language travels a long path after splitting off from its proto language. Therefore, I also trace diachronic changes in morphology, illustrating how Awyu-Dumut languages have developed over time, becoming what they are today. While focusing on the shared histories that bind Awyu-Dumut languages together, this book will also shed light on the morphological diversity found within Awyu- Dumut languages, telling each language’s own story. This book aims to show that, given high-quality data, it is proﬁtable and possible to use a bottom-up reconstructive approach to Papuan language families, reconstructing entire language systems of lower-level families. My hope is that the picture of the Awyu-Dumut linguistic history presented in this book will prove to be a useful piece in the greater puzzle of unraveling the past of Papuan people, and will inspire anyone involved in this effort. Without the years of intensive linguistic work on the Papuan languages of New Guinea undertaken by Petrus Drabbe between 1935 and 1960, not a word of this book could have been written. The current work makes the rich content of Drabbe’s grammars available to a larger audience who are unable to access or read his Dutch publications. Unlike Drabbe’s work, this book is not a synchronic grammar description and hence does not offer as exhaustive a discussion of Awyu-Dumut morphology as a grammar would. Rather, this is a comparative work, offering new perspectives on each Awyu-Dumut language through contrasting it with its sister languages.

Setting the Scene

1.1 Location and Sources of Data

Awyu-Dumut languages are spoken in the southwestern part of Papua, Indonesia, from the upper part of the 528-kilometer-long Digul River to its estuary in the Ara- fura sea, between the Digul and Mapi Rivers and from the border area near the Fly River east of the Digul River all the way to the southwest of the Wildeman River. Figure 1.1 indicates the general location of the Awyu-Dumut language family, whereas Figure 1.2 contains a map of the area where Awyu-Dumut languages are spoken, with the location of the clan lands of the primary informants for each language indicated by dots. Andrew Pawley notes that “to apply the comparative method thoroughly takes a long time and needs reliable descriptive data” (Pawley 2005a:4). For Awyu-Dumut languages a gold mine of quality descriptive data is available due to the efforts of Dutch missionary linguists Petrus Drabbe and Lourens de Vries, while time is what I brought to the equation. For six of the nine Awyu-Dumut languages a complete grammar with texts is available, whereas for another three Awyu-Dumut languages Drabbe wrote a grammar sketch. Furthermore, a 430-item word list is available for all nine Awyu-Dumut languages and can be found in Appendix A. Recent ﬁeldwork by HongTae Jang and Sung-Kyu Choi resulted in completed 430-item word lists for Kombai and Digul Wambon, whereas before only 200-item word lists were available for these languages. Although a wealth of information is available on Awyu-Dumut languages, the data does have its limits. First, the Awyu languages are less well-documented than the Dumut languages; a grammar with texts is available for only one of the four Awyu languages, namely Aghu. The other three Awyu languages, namely Pisa, Shiaxa and Yenimu, are described in a 50-page grammar sketch by Drabbe, where Yenimu in particular is often left out of the discussion. Secondly, the work of both Drabbe and de Vries focuses more on morphology and less on phonology, resulting in limited phonological descriptions for the Awyu-Dumut languages. Drabbe admits that he writes less about phonology than some might wish but also states that 4 1. Setting the Scene

Figure 1.1: Approximate location of the Awyu-Dumut language family (Wolters Atlas 1950)

“according to our humble opinion, one can also go too far in one’s appreciation of exact phonological data, or the lack thereof,”1 whereas de Vries concentrated on the morphological description of Digul Wambon, while the phonological introduction to the grammar was written by his collaborator R. Wiersma. As the focus of the current comparative study is also on morphology, the less-than-ideal phonological descriptions of Awyu-Dumut languages are not problematic. Thirdly, most of the texts included in the grammars are myths. As most texts come from one genre, finding examples of all types of constructions in the texts is not always easy. Future tense verb forms, for example, are infrequent in myths, which always take place in the past. The grammars can be compact at times, with few examples given of specific linguistic phenomena. However, as the reader will see, the data are of a sufficient quality and quantity to allow for the construction of a sizeable proto morphology. The following sections briefly present all languages relevant to the present study, indicating where they are spoken and what descriptive materials are available for them. To help the reader place the different languages, a schematic representation of the Awyu-Dumut language family, drawn up and presented in de Vries et al.(2012), is given in Figure 1.3.

1“Men kan naar ons bescheiden inzicht ook te ver gaan in de waardering van exacte fonologische gegevens, of van het ontbreken daarvan” (Drabbe 1957:iii). 1.1. Location and Sources of Data 5

Figure 1.2: Map of Awyu-Dumut language constructs (prepared by Jaap Fokkema, cartogra- phy department VU University)

Greater Awyu

Becking-Dawi Awyu-Dumut

KORTSAKYD

Dumut Awyu Ndeiram

MANYWBDWB SHIYENPSAAXU KOMWNGTYN

Figure 1.3: Greater Awyu language family 6 1. Setting the Scene

1.1.1 Shiaxa (SHI) The source for Shiaxa is Drabbe’s grammar sketch Twee dialecten van de Awju-taal from 1950. Voorhoeve(1975:375) notes that Shiaxa is spoken along the Shiaxa river north of the Digul River’s delta. Drabbe states that his informants were of the Aboghoj clan, who live in Gimikja (Drabbe 1950:93).

1.1.2 Yenimu (YEN) Like Shiaxa, all data on Yenimu are found in Drabbe’s grammar sketch Twee dialecten van de Awju-taal. As the title of the sketch implies, it is about two languages, those languages being Pisa and Shiaxa. Only when Yenimu differs from Shiaxa does Drabbe mention it. Thus Yenimu is similar to Shiaxa in many ways but will be treated as a separate language in the current study.2 Drabbe notes that his Yenimu informants belonged to the Jaso clan living at Kunubi.

1.1.3 Pisa (PSA) Pisa is the third language for which data was distilled from Drabbe’s 1950 Twee dialecten van de Awju-taal. Additional data were available from an unpublished grammar sketch written by Drabbe in 1947 entitled Spraakkunst van het Pisa-dialect der Awjoe-taal, which can be found at the KITLV (Koninklijk Instituut voor Taal-, Land- en Volkenkunde) library in Leiden. The Pisa language described by Drabbe was spoken along (the tributaries of) the Wildeman and Kampung Rivers. Drabbe’s informants were from the Wefoe and Ikwero clans, who had both moved from their locations along the Wildeman River to live near the government post at Masing (Drabbe 1947:1).

1.1.4 Aghu (AXU) Aghu, also known as ‘Djair’, was spoken between the Digul and Mapi Rivers in Drabbe’s time. Drabbe’s informants came from the Ghoghonasafo clan, who lived along the Eba River (Drabbe 1957:1). A detailed grammar containing ten texts published by Martinus Nijhoff in 1957 is the source of Aghu data.

1.1.5 Mandobo (MAN) Drabbe’s 1959 grammar published by Martinus Nijhoff entitled Kaeti en Wambon is the source of data for Mandobo. Kaeti is another name for Mandobo and means ‘real people’ (Drabbe 1959:4), but I choose to use the stream-based language name ‘Man- dobo’ introduced by Boelaars(1970) in order to avoid confusion with the neighboring Ok language Kati.

2In his proto phonology, Healey(1970) also views Yenimu as a separate language, whereas Voor- hoeve always sees it as a subdialect of Shiaxa. 1.1. Location and Sources of Data 7

Mandobo has three very similar dialects: Kambon on the lower Dumut River, Rungwanjap in the middle of the Dumut River and Wambon (not to be confused with Yonggom Wambon or Digul Wambon) on the upper part of the Dumut River. The dialect described by Drabbe is that of Rungwanjap, and his main informant was a 50-year-old man from the Omba clan, who had their homeland along the Marek, a tributary of the Dumut river. Drabbe’s Mandobo description consists of twelve myths with extensive explanatory notes both about Mandobo language and culture.

1.1.6 Yonggom Wambon (YWB) Yonggom Wambon, simply called ‘Wambon’ by Drabbe, owes its name to its Digul Wambon neighbors, who according to Lourens de Vries (p.c.) call their southern neighbors ‘Yonggom’. Drabbe described Yonggom Wambon as spoken in the little village of Waniktit, which lies along the Wanik, a tiny tributary of the Kao River. The second part of Drabbe’s (1959) published grammar Kaeti en Wambon, from page 115 onwards, contains all Yonggom Wambon data used in this book.

1.1.7 Digul Wambon (DWB) Digul Wambon as described by de Vries & Wiersma(1992) is spoken along the Digul River, north of Yonggom Wambon and into the foothills of the central mountain ranges. The main informant of Lourens de Vries belonged to the Keriŋgere clan, who live downstream from the village of Manggelum, near the Mbonop maelstrom and a village named Sawagit (Lourens de Vries, p.c.). I supplement the 1992 grammar by Lourens de Vries and Robinia Wiersma with an unpublished 2008 grammar of Digul Wambon by HongTae Jang.3 Jang had two main informants; the ﬁrst was the same as de Vries’ informant, namely Ahitup Keriŋgerey from the Keriŋgere clan, whereas his second informant, Obaja Wandawon, came from the upstream Digul area around Klofkamp. De Vries had a second informant, Yohanes Yaŋginop, while compiling a 200-item word list. Yohanes came from the Yaŋginop clan but grew up on Malimbisip clan lands. Jang compared the 430-item word list collected by Drabbe for the other languages with the 200 word list gathered by de Vries and elicited all missing items.

1.1.8 Kombai (KOM) Kombai was the second language described in de Vries’ dissertation (de Vries 1989). In 1993, a Kombai grammar based on de Vries’ dissertation was published by Pa- ciﬁc Linguistics (de Vries 1993). In his grammar, de Vries notes that “the Kombai live in a swampy rainforest area, hot and very humid. The terrain is hilly. The soil

3Jang(2008:1-2) refers to the language as Kenon Wambon, noting that there are three dialects of Wambon. He also notes that he describes the same variety as de Vries & Wiersma(1992). His ’Kenyam Wambon’ might correspond to the Wambon variety described by Drabbe(1959). 8 1. Setting the Scene is very poor and the area extremely sparsely populated. Kombai speakers number around 4000” (de Vries 1993:1). Lourens de Vries informed me that the Kom- bai grammar was based on language data gathered from informants who belong to clans who have settled near the village of Wanggemalo. His main informant was Khane Renakhoremba from the Renakhoremba clan, who live at the headwaters of the Kawo River, which ends in the Digul River (de Vries, p.c.). In 2013, the 430-item word list available for other Awyu-Dumut languages was kindly completed by Sung-Kyu Choi for Kombai using several informants from the villages of Yaniruma and Wanggemalo, namely Niko Dendemoku, Banio Kabukaruba and Yusuf Yafumano.

1.1.9 Waŋgom (WGM) The existence of a language called ‘Waŋgom’ was ﬁrst recorded by Drabbe, who described the Waŋgom people as northern neighbors of the Mandobo living on the Digul-side of the Erimop watershed. Interestingly, Drabbe notes that Mandobo speakers can more or less understand Waŋgom speakers (Drabbe 1959:5). De Vries later reported that Waŋgom is a dialect close to Kombai, which Kombai speakers told him they could understand. Intelligibility judgments recorded by both Hughes (2009) and Versteeg(1983) also show that Kombai and Wa ŋgom are closely related; Versteeg gives a lexical similarity percentage of 61%. Furthermore, de Vries et al. (2012) show that morphologically, Waŋgom is also very close to Kombai. The morphological data on Waŋgom are limited to a few paradigms written down in Baas’ ﬁeldnotes (Baas 1981), which mostly concern Tsaukwambo.

1.1.10 Tayan (TYN) Tayan, like Waŋgom, is a dialect that is close to Kombai. There is reference to a language called Tayan by both de Vries(1993) and Baas(1981), who note that both Kombai speakers and Citak speakers call the speakers of the dialect that borders the southwestern border of Kombai and Citak, Tayan speakers. De Vries et al.(2012) classify Kombai, Tayan and Waŋgom together as a dialect continuum and as a subgroup of the Awyu-Dumut language family. Unfortunately, no linguistic data what- soever are available for Tayan.

1.1.11 Korowai (KOR) The Korowai people live in small clans on their clan lands. Concerning the location of these clan lands, Rupert Stasch, an anthropologist who worked with the Korowai in the 1990’s and 2000’s, writes that “after emerging from the highland mountain chains, New Guinea’s southern rivers cross a ninety-thousand-square- mile lowland plain. The Korowai lands lie in the northwest corner of this plain, near the mountains and far from the coast, in the upper watersheds of the Eilanden and the Ndeiram Kabur Rivers” (Stasch 2009:05). A grammar on the language of 1.1. Location and Sources of Data 9 the Korowai people was published by Oxford University Press in 1997 (van Enk & de Vries 1997). The grammar is based on ﬁeldwork undertaken by both authors in the late 1980’s. Their main informants were Labulun from the Sendekh clan, who live in the upper area of the Mabul River, and Fenelun from the Molonggai clan, whose clan lands are where the Khelame stream joins the Becking River.

1.1.12 Tsaukwambo (TSA) Tsaukwambo is spoken on the Lower Dawi River between the Digul River and the Becking River, both in villages and on clan lands. Field notes of Peter Baas, a Dutch Reformed missionary who spent a year living in Kawagit in 1981, are the source of language data for Tsaukwambo. Baas gave his notes, written in Bahasa Indonesia, to Lourens de Vries, who presented a summary of their contents in a 2012 article (de Vries 2012a). He writes that “[t]he notebook consists of 235 pages with miscellaneous notes on cultural practices, survey and patrol reports, maps, four Tsaukambo texts, transcribed in IPA and with an interlinear Indonesian translation, a word list with 110 basic vocabulary items, a few shorter word lists with terms from speciﬁc semantic domains, such as body parts, names of sago species, kinship nouns and numerals. The ﬁrst 129 pages are concerned with cultural aspects. The pages 129- 178 contain texts, word lists and notes on the Tsaukambo language. The only other source on Tsaukambo is an Upper Digul Survey by the Summer Institute of Linguis- tics (SIL) linguist Hughes(2009) that gives basic survey information on Tsaukambo and contains a list of 239 Tsaukambo words” (de Vries 2012a:166).

1.1.13 Komyandaret (KYD) In his Tsaukwambo field notes, Baas(1981:90) mentions a small speech community of 300-500 people living further north along the Dawi River. According to his Tsau- kwambo informants, the language spoken by Komyandaret speakers is almost the same as their own, and Baas talks about Komyandaret as a variety of Tsaukwambo. The SIL survey by Hughes(2009) confirms the similarity between Tsaukwambo and Komyandaret, citing a lexical similarity percentage of 60%. Tsaukwambo and Komyandaret are classified together with Korowai as forming the Becking-Dawi dialect continuum (de Vries et al. 2012). Almost no linguistic data on Komyandaret exist, except for a 239-item word list in the 2009 SIL survey.

1.1.14 Other possible Awyu-Dumut languages There are several other possible Awyu-Dumut languages for which too little data are available to determine their afﬁliation(s). Voorhoeve classiﬁes Sawuy as an Awyu- Dumut language based on its pronouns (Voorhoeve 1971, 2005), Healey(1970:998) mentions a possible Awyu-Dumut language named Airo-Sumaghage, while sociolinguistic surveys carried out by SIL over the past decade present a plethora of likely Awyu-Dumut languages. Of all these languages, most data are available for Sawuy; 10 1. Setting the Scene a word list and some grammatical notes on Sawuy were published by Voorhoeve (1971). As for Airo-Sumaghahe, Healey refers to a mention of it in a genetic study published as Simmons et al.(1967), but no language data are available. The SIL language surveys (Jang(2003), Susanto(2004), Hughes(2009), Kriens & Lebold(2010), Kriens et al.(2013)), although they contain valuable sociolinguistic data, do not contain any morphological data and are hence not a source of data for the current study. Furthermore, it is not always easy to relate the languages listed in these surveys to the languages described in previous published work, making it unclear which language variety is being discussed. Therefore I limit myself to the languages described by Drabbe and de Vries.

1.2 Previous Study of Awyu-Dumut Languages

1.2.1 Petrus Drabbe Petrus Drabbe (*04-06-1887, † 27-10-1970) was a Catholic missionary of the order of the Sacred Heart who did descriptive linguistic work in West Papua between 1935 and 1960. Before his arrival at the mission station on the Mimika coast, he had already spent 20 years on the Moluccan Tanimbar Islands doing ethnographic, linguistic and missionary work. Although he was not formally trained as a (field) linguist – there was no such educa- tion when he was young – he had a natural talent for language description and took great interest in any language he came across. That he took pleasure in the scientific study of languages is clear from a quote in a radio address he gave in 1962, where he says “I have always done my work with pleasure, because every language is a fascinating miracle and it is an intense pleasure to ana- Figure 1.4: Petrus Drabbe lyze such a miracle, as one makes new and unexpected Source: Boelaars(1995) discoveries again and again.”4 Drabbe spent 25 years doing linguistic fieldwork in New Guinea, being freed from other missionary work from 1939 onwards. Every few years he went to a new area to study other languages. The following list gives an indication of which languages Drabbe worked in during which years, as he listed them himself in his 1962 radio address.

• Kamoro, from 1935 until 1938 • Sampan, 1938

4“Ik heb mijn werk altijd met veel plezier gedaan, omdat iedere taal een boeiend wonder is en het een intens genoegen is zulk een wonder te ontleden, doordat men telkens nieuwe en onverwachte ontdekkingen doet.” (Drabbe 1962:27) 1.2. Previous Study of Awyu-Dumut Languages 11

• Jaqai, 1939 • Awyu (Pisa, Shiaxa, Yenimu), from 1940 until 1942 • 3 languages on Frederik Hendrik Island, from 1940 until 1942 • Kati and another language in the Muyu area, from 1942 until 1945 • Boazi, Jelmek and Maklew, 1946 • furlough in the Netherlands, from early 1947 until early 1950 • Ekagi and Moni, 1950-1951 • Marind, 1952-1954 • Awyu (Aghu) and Dumut languages (Mandobo, Wambon), from early 1954 until 1956 • Three Asmat dialects, from 1956 until the end of 1958 • A fourth dialect of Asmat at Kepi, and Tamagario, 1959 • After repeated illnesses, Petrus Drabbe was repatriated to the Netherlands in 1960 at the age of 73, where he lived for another 10 years.

Thus, Drabbe first worked on three Awyu languages, Shiaxa, Yenimu and Pisa, for which he wrote a grammar sketch published in 1950, while only a decade later did he encounter a fourth Awyu language, Aghu, as well as the Dumut languages, for which he wrote more elaborate grammars. According to a 1954 publication in An- thropos, Drabbe thought on his first encounter with Mandobo – a Dumut language – that it was related to the Ok languages he had studied in the Muyu area during World War II. Culturally, the Mandobo and other Dumut speakers are closer to the Ok language speakers than to Awyu language speakers. An example of a cultural difference between Awyu speakers and Mandobo speakers that Drabbe notes is the difference in dress. Awyu speakers wear ‘schaamschorten’, while Dumut and Ok speakers wear ‘penisdoppen.’5 Only careful comparison of Mandobo to both Kati (an Ok language) and Awyu languages revealed that Mandobo was an Awyu- Dumut language, an indication that Drabbe did do some classification and comparison in the midst of his descriptive efforts. Petrus Drabbe’s observations and analyses are sharp and clear. More than half a century after their publication, Drabbe’s grammars are still very useful because of his straight-forward way of describing the phenomena he found in Papuan languages and because he included many glossed examples and texts. His down-to-

5‘Schaamschorten’ are a type of skirt made of fibers, while it is unclear what Drabbe means exactly by ‘penisdoppen’; perhaps they are akin to what the Korowai wear, described by van Enk & de Vries(1997:44) as “a leaf wrapping to cover the glans of the penis.” The exact words Drabbe uses are as follows: “we hadden het vooropgezette idee, dat Kaeti en Kati zeker verwant zouden zijn, wat ook tot op zekere hoogte waar is, maar we dachten er geen ogenblik aan, dat het Kaeti wel eens meer verwant zou kunnen zijn met het Awju, wat achteraf toch blijkt het geval te zijn. Ons vooropgestelde idee kwam hier vandaan dat etnografisch Kati’s en Kaeti’s duidelijk bij elkaar horen; zo is o.a. de dracht (rotangordel met penisdop, en bij feestelijke gelegenheden de peniskoker) dezelfde, en verschilt geheel van die der Awju’s (schaamschortdragers)” (Drabbe 1959:5). 12 1. Setting the Scene earth personality might also have contributed to Drabbe’s lasting success as a linguist, as described in an obituary by Professor Gonda, who worked together with Drabbe in the Netherlands. Professor Gonda describes Drabbe as follows: Father Drabbe was a likeable, generous man, thoroughly real, focused on facts, taking life as it came, and also wary of any form of interpretation outside his linguistic studies. The intellectual and emotional parts of his person went together harmoniously, according to those who knew him better than I ever did. He was never exuberant, but he could take pleasure in everyday, innocent things, and knew, if necessary, how to express his sympathies and annoyances, but always in a personal manner. With great sobriety he was able to see the relative importance of things, and in his scientific study he was aware of the limits of his own ability, and of the limited importance of every human labor.6 Drabbe himself ascribes his linguistic abilities and insights as perhaps coming from God when he says: “It seemed that Our Dear Lord gave me a special sort of ability for the work to which He called me... I must admit honestly that I have often called upon the Holy Spirit to do a little Pentecost miracle in me, for the sake of the Papuans.”7

1.2.2 Alan Healey and Bert Voorhoeve Ten years after Drabbe left Papua, Alan Healey published a proto phonology of Awyu-Dumut languages, taking word lists collected and published by Drabbe as his source of data. Alan Healey was a member of SIL and defended a PhD thesis at the Australian National University ANU in 1964 on a comparative study of Ok languages. By the time Healey published his Awyu-Dumut proto phonology in 1970, Bert Voorhoeve, a Dutch linguist, had already hypothesized that Awyu- Dumut languages were part of a larger language family consisting of Asmat, Ok and Awyu-Dumut languages, which he called the Central South New Guinea phylum (Voorhoeve 1968). In 1970, Voorhoeve, together with Ken McElhanon, proposed the existence of a Trans New Guinea family, at that point in time consisting of Voorhoeve’s Central South New Guinea phylum and McElhanon’s reconstructed Finnistere-Huon phylum (Voorhoeve & McElhanon 1970). Throughout the

6“Pater Drabbe was een aimabel, hartelijk man, door en door reëel, faktisch ingesteld, het leven nemend zoals het was, en ook buiten zijn taalkundige studiën afkerig van enige vorm van interpre- tatie. De intellectuele en emotionele zijden van zijn persoon gingen, ook volgens hen die hem veel beter gekend hebben dan ikzelf, op harmonische wijze samen. Uitbundig was hij nooit, maar hij kon veel plezier hebben in gewone, onschuldige dingen, en wist, als het nodig was, van zijn sympa- thieën en van ergernis wel degelijk, maar dan op zeer persoonlijke wijze, blijk te geven. Hij was er in grote nuchterheid in geslaagd de relatieve belangrijkheid der dingen te zien en was zich ook bij zijn wetenschappelijke werk wel degelijk van de grenzen van zijn eigen kunnen en van de beperkte belangrijkheid van ieder mensenwerk bewust” (Gonda & Anceaux 1970:450-461). 7“Maar het leek, dat O.L. Heer me een speciaal soort knobbel heeft gegeven voor het werk waar- toe Hij mij bestemde...ik moet eerlijk bekennen, dat ik heel vaak de H. Geest heb aangeroepen, dat hij ten bate van de Papoea’s een klein Pinksterwondertje met mij zou doen” (Drabbe 1962:20). 1.2. Previous Study of Awyu-Dumut Languages 13 chequered career of the Trans New Guinea family hypothesis, as clearly sketched by Andrew Pawley(2005b), it has never been doubted that Awyu-Dumut languages are a part of the Trans New Guinea family, whatever other languages may or may not belong to it. Years after Healey published his proto phonology of Awyu-Dumut languages, Bert Voorhoeve published a second proto phonology of Awyu-Dumut languages in 2001, incorporating data on Digul Wambon, Kombai and Korowai that had become available through Lourens de Vries’ ﬁeldwork (see Section 1.2.3). Both Healey’s and Voorhoeve’s proto phonologies will be further discussed in Chapter2. Due to their ground-breaking work on Awyu-Dumut proto phonology, I am able to focus on the proto morphology in the current study.

1.2.3 Lourens de Vries; Awyu-Dumut group at VU Twenty years after Petrus Drabbe left New Guinea, Lourens de Vries arrived. From 1981 until 1992, he lived in the Awyu-Dumut area, working as a linguistic missionary for the Dutch Reformed Mission (Zending Gereformeerde Kerken: ZGK). From 1981 until 1983, he lived in the village of Manggelum and worked on Digul Wambon. From early 1984 he and his family lived in Wanggemalo where he studied Kombai and from where he also made frequent trips to the Korowai village of Yaniruma as well as to Korowai clan lands. Lourens de Vries’ doctoral thesis, defended in 1989, concerns Digul Wambon and Kombai, while in 1997 he co-authored a book on Korowai grammar (van Enk & de Vries 1997). After his return to the Netherlands in 1993, Lourens de Vries always kept an interest in Awyu-Dumut languages, publishing on them over the next 20 years (most importantly, de Vries(1994, 2005, 2006, 2010)). From 2009, Lourens de Vries has been the project leader of a 4-year research project funded by the Netherlands Organisation for Scientiﬁc Research (NWO) entitled The Awyu-Dumut language family of Papuan languages in its cultural and linguistic context at the VU University Amsterdam, in the context of which the current research was undertaken. Besides Lourens de Vries and myself, the Awyu-Dumut research group at the VU has a third member, Wilco van den Heuvel, who studied the relationship between Awyu-Dumut languages and their immediate neigbors. In a paper containing preliminary results (de Vries et al. 2012), the three members of the Awyu-Dumut research group conclude that Kombai and Korowai are both part of larger dialect chains and that the dialect chain of which Korowai is a part, the Becking-Dawi chain, is only distantly related to the other Awyu-Dumut languages. The dialect chain to which Kombai belongs, the Ndeiram chain, forms a subgroup of the Awyu-Dumut language family. These preliminary results led to the establish- ment of the family tree of Awyu-Dumut languages represented in Figure 1.3 above. 14 1. Setting the Scene

1.3 Awyu-Dumut Languages are Constructs

Awyu-Dumut speakers in pre-contact days did not have or use names for their languages; when asked about the name of the language they spoke and what group or ‘tribe’ they belonged to by missionaries such as Drabbe, they replied with terms meaning ‘our sound’ or ‘sound of real people’, and for the group to which they belonged they used terms meaning ‘(real) people’ or ‘humans’. For example, the language (and group) name ‘Aghu’ means ‘person’, as does ‘Kombai’, which is an Indonesianized version of xoba ‘person’. At times, languages and groups were named after the river or stream where they were spoken by the ﬁrst western person who had need for a name. For Shiaxa and Yenimu, Drabbe notes that “one of the dialects here treated is spoken along the Shiaxa, and here and there we will give some differences which occur along the Yenimu, a branch of the Shiaxa. Because the groups which live along these little rivers have no common name, we simply speak of the Shiaxa dialect and the Yenimu dialect.”8 In a similar manner, ‘Korowai’ is an Indonesianized version of xolufo ‘upstream’, referring to the clans who live ‘upstream’. The names ‘Mandobo’ and ‘Digul Wambon’ are also river-based, whereas the origins of the names ‘Pisa’ and ‘Yonggom Wambon’ remain unknown; ‘Pisa’ is a name the speakers thereof used for themselves, whereas ‘Yonggom’ is a name given to Yonggom Wambon speakers by Digul Wambon speakers, according to Lourens de Vries (p.c.). Drabbe admits freely that he is the one giving the Awyu-Dumut languages names, stating that “because of lack of a better name, or rather because of lack of an actual name, we speak of the Kaeti language.”9 In clan societies such as those in which the Awyu-Dumut speakers lived when Drabbe and de Vries worked on their grammar descriptions, clan membership and clan lands are the focus of identity, rather than language. Awyu-Dumut speakers do not necessarily view the language they speak as setting them apart from the next clan or group that speak another language. De Vries notes that in Awyu-Dumut society, “[l]anguage transcends clan lands; people of many different clan lands may share a language. But among the Awyu-Dumut, language as a clan transcending entity is never the focus of identity construction. Not language but clan afﬁliations determine who you are” (de Vries 2012b:10). People within one clan may speak different languages, and conversely, people who speak the same language may belong to very different clans. Thus the speakers of one language do not necessarily have a common sense of identity; indeed, they might be enemies. Hence, language is not a strong marker of (group) identity in Awyu-Dumut societies. At the same time,

8“Een der beide dialecten welke we hier behandelen wordt gesproken aan de Sjiagha, en we zullen hier en daar enige afwijkingen geven die aan de Jenimu voorkomen, een tak van de Sjiagha. Daar de groepen die aan de riviertjes wonen geen algemene naam hebben, spreken we eenvoudig van Sjiagha- en Jenimu-dialect.” (Drabbe 1950:93) 9“Daarom spreken ook wij bij gebrek aan beter, of liever bij gebrek aan een werkelijke naam, van Kaeti taal.” (Drabbe 1959:4) Note that Kaeti is the term Drabbe uses for Mandobo; Mandobo is the stream along which this language is spoken, and was used as a language name by Boelaars(1970). Here Mandobo is preferred to avoid confusion with a neighboring Ok language called Kati. 1.3. Awyu-Dumut Languages are Constructs 15 language is used as a tool to express one’s relationships, in a society in which it is important to whom you are connected. Foley notes that “like other cultural arti- facts, language is a trade item” in New Guinea (Foley 1986:24). Likewise, de Vries notes about Awyu-Dumut society that “[w]hen two clans or two individuals with different languages or dialects have a relationship in the context of trade, marriage, or otherwise, the partners not only exchange goods to symbolise their relationship but they also exchange elements of language. By borrowing elements from the language of partners and relations, sociocentric conceptions of identity are expressed in language”(de Vries 2012b:12). Thus the mix of languages that a group or individual speaks reflects their identity. Awyu-Dumut ideas about language were quite different from the language ideologies held by the missionaries, linguists and government officials who encountered them, whose language ideologies originated in nation states with national languages where language is related to culture, poli- tics and national identity. When naming languages that before had no names, they assumed that these languages were bounded entities that were expressions of ho- mogeneous cultures, and that the people who spoke a language formed a cultural unit. Quite the opposite was the case in the New Guinea context, where “commu- nities must be viewed as foci in areal networks of cultural and linguistic patterns. Each community constructs its identity by drawing on the available pool of cultural and linguistic traits” (Foley 1986:26). Another way in which Awyu-Dumut languages did not fit the language ideologies of the scholars who described them is that they are not bounded, discrete entities. Rather, Awyu-Dumut languages form dialect chains or continua. For example, a clan speaking dialect (or variant) A of a language may understand another clan speaking dialect B, who in turn understand clans speaking dialect C and D, but speakers of dialect A do not understand speakers of dialect D and vice versa. Thus A and D are then defined as different languages, but are part of the same dialect chain. This means that Awyu-Dumut languages have unclear, fuzzy boundaries, as it is unclear which dialects or variants should or should not be grouped as being part of a language. Recent SIL surveyors of the Awyu languages note the following about dialect chaining in these languages: “The Awyu people [...] speak a number of closely related languages. In addition, there is a great deal of language ‘chaining’ from one village to the next, which makes it difficult to define the boundaries between these languages” (Kriens and Lebold 2010:5). In a setting of dialect chaining, where language boundaries are unclear, it is imperative to note where one’s informants come from in order to make clear which variety is being described. Both Drabbe and de Vries are strong on this point, clearly indicating in their grammars which clan(s) their informants were from and where those clans lived, so that the variant they describe can be well located on a map (see Figure 1.2). Awyu-Dumut languages are constructs, defined by the missionary linguists who described them rather than by the speakers of these languages. The linguistic comparisons presented in this book are based on the grammars written by Drabbe and de Vries, which are their interpretations that can best be viewed as momentary snap- shots, concrete pictures of a larger reality. The reader should keep in mind through- 16 1. Setting the Scene out that any statement involving the phrase ‘in language X...’ is in fact an abstrac- tion of ‘in what Drabbe/de Vries defined as language X based on interaction with a limited number of speakers.’

1.4 Methodology: the Comparative Method

The comparative method of historical linguistics, developed in the 19th century in the study of the geographically widespread Indo-European language family, is employed in this book to reconstruct the proto morphology of a Papuan language family. Applying the comparative method, summarized as a set of instructions by Durie & Ross(1996) (see Table 1.1), the historical linguist can determine what the linguistic residues of an earlier (proto) language are within a group of related languages. That is, the comparative method allows the linguist to ﬁlter out linguistic similarities that languages share because they descend from a common ancestor and history.

Table 1.1: The linguistic comparative method summarized as a set of instructions (Durie & Ross 1996)

1. Determine on the strength of diagnostic evidence that a set of languages are genetically related, that is, that they constitute a ‘family’; 2. Collect putative cognate sets for the family (both morphological paradigms and lexical items); 3. Work out the sound correspondences from the cognate sets, putting ‘irregular’ cognate sets on one side; 4. Reconstruct the proto language of the family as follows: a) Reconstruct the proto phonology from the sound correspondences worked out in (3), using conventional wisdom regarding the directions of sound changes; b) Reconstruct proto morphemes (both morphological paradigms and lexical items) from the cognate sets collected in (2), using the proto phonology reconstructed in (4a); 5. Establish innovations (phonological, lexical, semantic, morphological, morphosyntactic) shared by groups of languages within the family relative to the reconstructed proto language; 6. Tabulate the innovations established in (5) to arrive at an internal classiﬁcation of the family, a ‘family tree’; 7. Construct an etymological dictionary, tracing borrowings, semantic change etc. for the lexicon of the family (or of one language of the family).

The comparative method is closely linked to the family tree metaphor, which holds that languages split off from a common ancestor and then follow their own paths, diverging into separate languages that retain traces of the original proto language. The tree model assumes that there is no, or very little, contact between languages once they have split off from their common ancestor. However, within the 1.4. Methodology: the Comparative Method 17

Awyu-Dumut language family there is extensive contact – and borrowing of linguistic material – between languages. This interaction is fueled by a cultural language ideology that does not see language as a focus of group identity to be protected but rather as something that can be shared or exchanged. This leads to a situation in which individuals express their identities – which are largely determined by their relationships to those around them – by incorporating (part of) the language(s) of those with whom they are in contact (de Vries 2007). Thus a linguistic feature that is shared between Awyu-Dumut languages has a high potential of having its origin in language contact, rather than in a common linguistic ancestor. Nevertheless, a rigorous application of the comparative method does allow one to distinguish between inherited and borrowed elements. However, it can never be ruled out altogether that an element is not borrowed, and hence a claim for genealogical affiliation is always a probabilistic one (Foley 2000:359). On the basis of Awyu-Dumut lexical data, the second and third steps as well as part of the fourth step listed by Durie & Ross(1996), namely the collecting of lexical cognate sets, working out of sound correspondences and the reconstruction of a proto phonology, were carried out by Alan Healey(1970), Bert Voorhoeve(2001) and by myself in Chapter2 of this book. The applicability of the comparative method in Papuan lexical data is at times questioned because core vocabulary has been known to be borrowed across languages in the Papuan setting (Foley 2000:392) and therefore, in theory, any item that appears cognate between two Papuan languages could have been borrowed. Nonetheless, if two languages (or a group of languages) share a large part of their vocabularies, and if regular sound correspondences can be traced between the languages, then the languages in question are very likely to be genealogically related. Not only were the lexical data employed to reach a proto phonology, but phylogenetic computational methods were applied to the lexical data to arrive at a subgrouping, or internal classification, of the Awyu-Dumut language family (Section 2.8). Thus the lexical data found in Drabbe’s and de Vries’ grammars were useful, but as Antoine Meillet already noted, “one can initially establish vocabulary resemblances between two or several languages as an indication of where to do further research, [but] this cannot furnish a definitive demonstration; vocabulary can only orient the research, and proof comes from elsewhere” (Antoine Meillet, quoted by Nichols 1996). In Awyu-Dumut languages, this ‘elsewhere’ is bound morphology, the reconstruction of which will prove beyond a doubt that Awyu-Dumut languages share a common ancestor. The current study focuses on the reconstruction of bound morphology. Concern- ing the use of bound morphology in historical Papuan linguistics, Foley notes that

For non-Austronesian languages, the vast majority of which have no doc- umentation older than 50 years, it is problematic to sift what is true, genetically inherited material from what is borrowed from other languages, especially the borrowings from related contiguous languages or from languages now defunct. Consequently, comparative linguistics in Papuan languages must proceed with care and the utmost rigor. It would 18 1. Setting the Scene

appear that bound morphological forms are the most resistant to borrowing [again, however, not entirely immune], so that bound morphological forms that appear cognate are the most reliable guide to genetic relationships between Papuan languages. Although not unchallenged, this working hypothesis seems the most trustworthy, albeit conservative, way to proceed (Foley 2000:359).

Although bound morphology is more resistant to borrowing than are lexical items, it is not immune to being copied, especially not in the cultural environment in which Papuan languages operate, an environment which encourages linguistic borrowing. Papuan languages might have bound morphology in common for one of three reasons. First, the shared bound morphology can be a reﬂection of genealogical relationship; each language has inherited the morphology from their proto language. Second, shared bound morphology can be due to borrowing, even though borrowing of bound morphology is much rarer than lexical borrowing. Third, shared bound morphology, especially when only the function but not the form is shared, can be ascribed to a shared areal feature that has spread through intensive language contact. Thus, to use bound morphology as an effective tool to trace genealogical relationship, cognate shared morphology must be distinguished from copied shared morphology. The following three constraints facilitate the making of this distinction:

1. the paradigmaticity constraint (Ross 2005:50) 2. the form-function correlation constraint (Foley 2005:110) 3. the systematicity constraint (Foley 2005:141)

Constraint (1) requires (proposed) related languages to share two or more forms in a particular paradigm and leads to entire paradigms, rather than individual forms, being compared. Individual forms are more likely to be borrowed than is an entire paradigm, and thus if multiple similarities or cognates are found in one paradigm, a genealogical explanation becomes more likely. Constraint (2) requires the forms and meanings of putative cognates in shared bound morphology to be identical or else relatable in a non-ad hoc manner. When discussing the use of structural similarities to establish genealogical relationship, Pawley notes that “in general, structural resemblances do not constitute strong evidence for genetic relationship unless they are also associated with cognate morphemes” (Pawley et al. 2005:75). The form-function correlation constraint rules out the consideration of cases of metatypy as reﬂecting a common ancestry. Metatypy is a term coined by Malcolm Ross and can be described as a process wherein “the morphosyntactic constructions of one of the languages of a bilingual speech community are restructured on the model of the constructions of the speakers’ other language” (Ross 2007:116). In such instances, only the function (‘type’) of a grammatical construction enters the language, rather than the grammatical construction plus the morphological form that expresses it. Once again, it must be emphasized that the borrowing of the function and form of a grammatical construction is rare. 1.5. The Structure of the Book 19

Furthermore, the form-function constraint also prevents the researcher from comparing morphemes that only share a function in a broad sense. Foley(2005:109- 110) cites an example of Laycock and Z’graggen who reconstructed a 1SG pronoun *wun with 1SG, 2SG and 3SG pronoun reflexes in the daughter languages. The form- function constraint prohibits such reconstructions and other mass comparisons like it. In addition to entire paradigms being compared under strict form-function correlation conditions, the overall grammatical systems of languages must be compared and must show similarities in bound morphology in multiple, independent paradigms in order to meet Constraint (3). When these three constraints are met, a genealogical relationship is far more likely to account for the observed correspondences in bound morphology than an explanation in terms of either chance or language contact. A strict application of the form-function correlation constraint forces the researcher to carry out diachronic trajectory research: when the forms and/or functions of bound morphology have changed, the form-function constraint stipulates that the innovation paths or the trajectories of change and diversification that a morpheme has followed be described and explained in a principled way. This approach allows the researcher to discover the historical processes that have been at work in a language, shaping it into what it is today. Tracing diachronic changes tells a much more detailed story about a language (or language family) than would be possible by merely establishing its genealogical relatedness. I agree with Foley that “we must not be side-tracked into believing that isolating the genetic affiliation, the ‘Platonic essence’, of a language will be the key to understanding its nature” (Foley 1998:505) and that “simple labels like ‘Austronesian’ or ‘Papuan’ advance our understanding of the complex linguistic history of New Guinea but little. Seeing them as labels for complex assemblages of traits which are negotiable and contestable seems to me to offer a more fruitful avenue of investigation” (Foley 1998:515-516).

1.5 The Structure of the Book

The structure of this book follows that of a standard grammar description. Chapter 2 focuses on (proto) phonology, summarizing the work done by Healey (1970) and Voorhoeve(2001) and presenting a computational phylogenetic analysis of Awyu-Dumut lexical data. Chapters 3 and 4 cover non-verbal morphology: Chapter 3 is about nouns and adjectival morphology, while Chapter 4 is on pronouns. In Chapters 6 through 9 Awyu-Dumut verbal morphology is covered, with separate chapters dedicated to tense, aspect, mood, subject person-number marking and negation. Chapter 10 covers deictics and demonstratives and the grammaticalization path they follow. Chapter 11 moves on to morphology on the clausal level, describing the diachronic development of switch reference and clause chaining in Awyu-Dumut languages. The book concludes with a short summary and discussion of the ﬁndings. 20 1. Setting the Scene

Each chapter is divided into a synchronic and a diachronic portion. The synchronic portion offers descriptions of the grammatical phenomenon discussed in that chapter for each Awyu-Dumut language, often discussing the phenomenon within each subgroup. These synchronic descriptions summarize Drabbe’s and de Vries’ work, while also containing my own reanalyses of their data. Reconstructed proto morphemes and diachronic (grammaticalization) pathways are then offered in the diachronic portion of each chapter, taking a bottom-up reconstructive approach that ﬁrst reconstructs Proto Awyu and Proto Dumut before reconstructing Proto Awyu-Dumut.

1.6 A Note on Glossing

The examples I use to illustrate and support my analyses come from the grammars by Petrus Drabbe and Lourens de Vries. The language concerned is indicated for each example, and a citation is given, which can help the reader locate the original example in the grammar. It must be noted that the glosses reﬂect my own analyses and may, therefore, differ from how they are represented in the grammar or grammar sketch of that particular language. Any reanalyses of the data presented in Drabbe’s and de Vries’ grammars are thoroughly explained in the text. The Leipzig Glossing Rules are followed in the glosses in this book.10 Further- more, any morphophonemic changes that occur are not ﬁltered out of the examples. Thus, for example, if the realis marker -t changes to -l intervocalically in Digul Wambon, it is written as /l/ in the example, not as -t. For a list of the most common morphophonemic rules that occur in Awyu-Dumut languages, see Section 2.2 in the chapter on Awyu-Dumut phonology.

10http://www.eva.mpg.de/lingua/resources/glossing-rules.php 2

Proto Phonology and Phylogenetics

Within the last half century, two major publications appeared concerning the proto phonology of Awyu-Dumut language family. Alan Healey was the first to reconstruct a proto phonology of the Awyu-Dumut family in 1970. Bert Voorhoeve incorporated extra data on Awyu-Dumut languages gathered since 1970 in his “Awyu- Dumut Proto Phonology II” in his 2001 publication. Both researchers used a combination of the comparative method and lexicostatistics to arrive at their conclusions. The current study of Awyu-Dumut proto phonology employs a strict application of the comparative method, in addition to using phylogenetic computational methods, to arrive at a proto phonology and a subgrouping of the Awyu-Dumut language family. The work by Holton et al.(2012) and Holton & Robinson(2012) on the proto phonology and subgrouping of the Alor-Pantar family of Papuan languages greatly inspired the work presented in this chapter. In a proper application of the comparative method, sound correspondences are established first before an attempt at subgrouping is made, as set out by Durie and Ross (1996) and summarized in Table 1.1. Both Healey and Voorhoeve begin by presenting their subgroups without extensively giving sound correspondences; they focus immediately on establishing sound changes. In the current chapter, consonantal sound correspondences are first established by studying a set of 123 putative cognate sets that can be reconstructed to the proto Awyu-Dumut level. Only then are regular sound changes reconstructed. In Section 2.7 on Awyu-Dumut vowels, it is concluded that the comparative method cannot be applied rigorously to the Awyu-Dumut vowels but that some sound changes within the vowel system can nonetheless be traced. Phylogenetic computational methods are employed in Sec- tion 2.8 to arrive at a subgrouping of the Awyu-Dumut language family. The conclusions drawn in this chapter facilitate and enable the reconstruction of Awyu-Dumut proto morphology. 22 2. Proto Phonology and Phylogenetics

2.1 Data

The 430-item word list contained in Drabbe(1959:162-181) forms the basis of the proto phonological work done by Healey(1970), Voorhoeve(2001) and myself. Hea- ley restricted himself to lexical items that also appear on the Swadesh 200 list, thus minimizing the chance that he was dealing with borrowed words. The word list to which Healey had access contained six Awyu-Dumut languages. By the time Voor- hoeve wrote his Awyu Dumut proto phonology II, he had received data on two further Awyu-Dumut languages, Digul Wambon and Kombai, from Lourens de Vries. From communication with Lourens de Vries, it is known that he gave Voorhoeve a word list containing 200 lexical items, most of which also occurred in Drabbe’s word list. Thus Voorhoeve had less data for Digul Wambon and Kombai than he did for the other Awyu-Dumut languages. Voorhoeve leaves Yenimu out of his analysis because of its similarity to Shiaxa, and he supplements the Pisa and Mandobo word lists collected by Drabbe with word lists he himself gathered in the villages of Ketu (Pisa) and Kwem (Mandobo). The data for the current study are (nearly) complete word lists of 430 items for all Awyu-Dumut languages: Shiaxa, Yenimu, Pisa, Aghu, Mandobo, Yonggom Wambon, Digul Wambon, Kombai and Korowai. The full Digul Wambon and Kom- bai word lists were collected in 2012 and 2013 by HongTae Jang and Sun-Kyu Chi, respectively, and I am grateful to them for sharing their data. The Korowai data were taken from the lexicon contained in van Enk & de Vries(1997). Drabbe’s 430- item word list is the source of data for the other six Awyu-Dumut languages.

2.2 Orthography and Morphophonemics

Appendix A contains the lexical database on which I base my analysis, including the words that could be reconstructed for Proto Awyu, Proto Dumut and Proto Awyu- Dumut. An item was reconstructed for Proto Awyu-Dumut only if it had a reﬂex in both an Awyu and a Dumut language. In the appendix, the original orthographies of the various sources are maintained. However, the orthography I use in my writing is slightly different; it is presented in Table 2.1 in the column entitled ‘grapheme.’ Note that although Drabbe, de Vries and Voorhoeve all claim that Dumut voiced plosives are always prenasalized, they are not consistent in writing them as such, an inconsistency that has made its way into my work as well (when the sources write a voiced plosive as not being prenasalized, I also do so).

Table 2.1: Awyu-Dumut phonemes and their representation

IPA Source Grapheme Languages p p p SHI, YEN, PSA, AXU, MAN, YWB, DWB, KOR t t t SHI, YEN, PSA, AXU, MAN, YWB, DWB, KOR k k k SHI, YEN, PSA, AXU, MAN, YWB, DWB, KOR continued on next page 2.2. Orthography and Morphophonemics 23

Table 2.1: continued

IPA Source Grapheme Languages mb mb, b mb, b MAN, YWB, DWB, KOM, KOR nd nd, d nd, d MAN, YWB, DWB, KOM, KOR ŋg ŋg, ngg, g ŋg, g MAN, YWB, DWB, KOM, KOR b b b SHI, YEN, PSA, AXU,KOR d d d SHI, YEN, PSA, AXU,KOR g g g SHI, YEN, PSA, AXU,KOR F f, w f SHI, YEN, PSA, AXU, KOM, KOR B v, w v,w MAN, YWB, DWB x kh, gh x SHI, YEN, PSA, AXU, DWB, KOM, KOR G ch x SHI, YEN, PSA, AXU, DWB m m m SHI, YEN, PSA, AXU, MAN, YWB, DWB, KOM, KOR n n n SHI, YEN, PSA, AXU, MAN, YWB, DWB, KOM, KOR l l l DWB, KOM, KOR r r r SHI, YEN, PSA, AXU, MAN, YWB, DWB, KOM s s s SHI, YEN, PSA, AXU, YWB, DWB, KOR w w w SHI, YEN, PSA, AXU, MAN, YWB, DWB, KOM, KOR j j y SHI, YEN, PSA, AXU, MAN, YWB, DWB, KOM, KOR

i i i SHI, YEN, PSA, AXU, MAN, YWB, DWB, KOM, KOR y ü ü AXU, MAN, KOM, KOR u u u SHI, YEN, PSA, AXU, YWB, DWB, KOR e e, é e SHI, YEN, PSA, AXU, MAN, YWB, DWB, KOM, KOR ø ö ö MAN o o o SHI, YEN, PSA, AXU, MAN, YWB, DWB, KOM, KOR E è E SHI, YEN, PSA, AXU, KOR a a a SHI, YEN, PSA, AXU, MAN, YWB, DWB, KOR

Table 2.2: Awyu-Dumut common morphophonemic changes

rule example languages

p → {w,f,v}/V_V ep+e=ewe (YWB) MAN, YWB, DWB, KOM there+ TOP =that t → {r,l}/V_V mbet+o=mbelo (DWB) MAN, YWB, DWB, KOM SEQ + CONN =and then k → {g}/V_V towe+karigiap=towe garigiap MAN, YWB, DWB (MAN) very+afraid=very afraid mb → ∅/p_ mba-t-ep-mbo=mbalepo (DWB) MAN, YWB, DWB stay-REAL-1SG-PST =I stayed ∅ → n/V_V na+a=nana (KOM) MAN, YWB, DWB, AXU, 1.SG.POSS +house=my house PSA, SHI, YEN, KOM ∅ → n/V_V˜ ax˜ı+e=axine (AXU) AXU, PSA, SHI, YEN, KOM go.FUT + NON1SG =he will go V˜ → Vŋg/_k kumã+ki=kuma ŋgi (AXU) AXU, PSA, SHI sick+be=to be sick

The most common morphophonemic changes that occur in between words and due 24 2. Proto Phonology and Phylogenetics to morphological clustering are given in Table 2.2.1 Dumut languages and Kombai share a fricativization of voiceless plosives when they occur intervocalically. Fur- thermore, two vowels may not follow each other in any Awyu-Dumut language, and a transitional nasal is inserted whenever two vowels appear after one another. The last two morphophonemic changes listed in Table 2.2 concern nasal vowels found in Awyu languages and Kombai; the nasality of a ﬁnal vowel is often trans- ferred to the sound that follows it, or if a vowel follows a nasal vowel, an /n/ occurs between the two vowels. In his phonological descriptions Drabbe further notes that there is extensive vowel harmony in Awyu-Dumut languages.

2.3 Sound Correspondences

In this section, I describe the consonant correspondences found in 123 cognate sets of the Awyu-Dumut 430-item word list that were reconstructible to the Proto Awyu- Dumut level. Only those items that had a reflex in both an Awyu language and a Dumut language were reconstructed to Proto Awyu-Dumut. Awyu-Dumut consonants behave differently depending on their context; therefore, a distinction is made between consonants that occur initially, medially and finally. Sometimes the vowels that follow a consonant also influence its realization in the various Awyu- Dumut languages. Taking these conditioning factors into account, there are 27 consonant correspondences that can be identified in the dataset. Table 2.3 lists the set of consonant correspondences found in the Awyu-Dumut languages, as well as the reconstructed Proto Awyu-Dumut phoneme for each correspondence set. The environment column (Env.) indicates the condition in which the correspondence occurs: initial (I), medial (M), or final (F). Sometimes the environment is conditioned by a certain vowel; such vowels will be included in parentheses in the environment column. A zero (∅) indicates that the Proto Awyu-Dumut sound has been dropped in that particular language, while a slash (/) indicates that there is more than one reflex of a sound. A question mark (?) indicates that there is insufficient data to determine a reflex for that language. In tables 2.4-2.14 below containing lexical items that illustrate these sound correspondences, a blank means that the data is missing for that language, whereas a dash (–) indicates that the language has a non-cognate, different lexical realization for the term in question.

Table 2.3: Awyu-Dumut sound correspondences

PAD Env. SHIYENPSAAXUMANYWBDWBKOMKOR *p I f f f f ∅ ∅ h f f *p M f f f f ∅ w w f ? *p F ∅ ∅ ∅ ∅ p p p f, ∅ p *t I t t t, s t, s t t, s t, s r ? continued on next page

1 Minor morphophonemic changes in examples will be clariﬁed in the text or in footnotes and are not given in Table 2.2. 2.3. Sound Correspondences 25

Table 2.3: continued

PAD Env. SHIYENPSAAXUMANYWBDWBKOMKOR *t M t t t, s t, s t, r t, s, r t, s, l r ? *t F t, ∅ ∅ ∅ ∅ t t t l, ∅ l *k I (i,e,u) k k k k k k k x x *k I (a, o) x x x x k k k x x *k M (i,e,u) k k k k g g x x x *k M (a,o) x x x x g g x x x *k F x, ∅ ∅ ∅ ∅ k k k x, ∅ ? *mb I b b b b mb mb mb b b *mb M b b b b mb mb mb b b *nd I d d d d nd nd nd d d *nd M d d d d nd nd nd d d *ŋg I ŋg g g, ŋg g, ŋg ŋg ŋg ŋg g ? *ŋg M ŋg g g, ŋg g, ŋg ŋg ŋg ŋg ŋg ? *m I m m m m m m m m m *m M m m m m m m m m ? *m F ∅ ∅ n n n, m m m m m *n I n n n n n n n n n *n M n n n n n n n n ? *n F ∅ ∅ n n n n n ∅, n n *r I r ? r ∅ r r l l l *r M r r r r r r l r,l? *w I w w w w w w w w f *y I y y y ∅ ∅, y y ∅ ∅, y y

2.3.1 Voiceless plosives

*p Proto Awyu-Dumut *p is found in initial, medial and final position, as shown in Table 2.4. Initial *p has a reflex /f/ (IPA F) in Shiaxa, Yenimu, Pisa, Aghu and Kombai. In Yonggom Wambon and Mandobo, initial *p is dropped, while in Digul Wambon, initial *p changes into /h/. It must be noted that Shiaxa, Yenimu, Pisa and Aghu do have words that begin with a /p/, as shown in Table 2.5, but these words cannot be reconstructed to a Proto Awyu-Dumut form with an initial *p, as there are no cognates that have a reflex of *p in any of the other languages. The initial /p/ in these words is hence not a reflex of Proto Awyu-Dumut *p but rather, as already noted by (Healey 1970:1000), reflects a history of borrowing in Proto Awyu. In the whole dataset, medial *p occurs only once, namely in kip(V)i ‘wind’, so it can only be reconstructed tentatively. This medial *p changes to /f/ in the same languages in which initial *p changes to /f/. In Mandobo, medial *p is dropped, while in Yonggom Wambon and Digul Wambon, medial *p changes to /B/, alternatively written as /w/ or /v/ in the sources. Final *p is again more frequent than medial *p. It is dropped altogether in Shiaxa, Yenimu, Pisa and Aghu, while it does not change in Mandobo, Yonggom Wambon 26 2. Proto Phonology and Phylogenetics

Table 2.4: Reﬂexes of *p

I name tongue to see M wind

PAD *p *pi *paŋgat *peta *p *kip(V)i SHI f fi fagE fete- f kifi YEN f fi fagE fite- f kifi PSA f fi fagE feto- f kifi AXU f fi fage ete- f kifi MAN ø üp oŋgat itigio- ∅ kiou YWB ø ip – eto- w kiwui DWB h hit hatgat hetak- w kiwin KOM f fi faŋga fera- f xifei KOR f fi – – ?

F eye fruit vagina take

PAD *p *ketop *rop *atop *rap SHI ø kero ro ato re- YEN ø kiro ro ato re- PSA ø krimogo jindero ato ra- AXU ø kiomogo – ato a- MAN p kerop rop atop – YWB p kerop rop atop rap- DWB p kelop lop atop lap- KOM ø, f xoro lo aro lofa- KOR p – op – –

Table 2.5: Initial /p/ in four languages

PADSHIYENPSAAXUMANYWBDWBKOM

long – pere piri kowE pi ŋguruop kolamop ŋgolo ŋgurup old – pusu patu semebe posyü wandop woŋgopon sinop muno shoot – piemo- pemo- t˜ı- teen- taem- tamya- rabi-

wound – kEwi kefi pEso peso kogu mbom hitop ruruko and Digul Wambon. In Kombai, final /f/ is often dropped, but it does occur sometimes; de Vries 1993 notes that it is underlying in many instances and only appears when the next word or morpheme starts with a vowel (hence /f/ might not always show up in the word list, although it sometimes does). Healey does not think that *p occurred initially and medially; rather he posits a *w, noting that final *p and this *w were in complementary distribution (Healey 1970:999). However, together with Voorhoeve, I see no need to reconstruct this *w, although Proto Awyu-Dumut *p most likely did have a fricativized allophone. 2.3. Sound Correspondences 27

*t and /s/ Proto Awyu-Dumut *t occurs in initial, medial and final position (Table 2.6). Word- finally, it is dropped in Shiaxa, Yenimu, Pisa and Aghu, while Kombai sometimes drops final /t/ and sometimes retains it as an optional /l/. Initially and medially, Proto Awyu-Dumut *t goes to /r/ in Kombai. Pisa, Aghu, Yonggom Wambon and Digul Wambon have reflexes /t/ and /s/ initially; it is not entirely clear under what conditions *t becomes /s/. The same is true for medial *t, which sometimes but not always becomes /s/ in Pisa and Aghu, /r/ in Yonggom Wambon and Mandobo, and /l/ in Digul Wambon.

Table 2.6: Reﬂexes of *t

I armpit cold ear M sweat leech beard vagina

PAD *t *tVŋgo(top) *tarü *turu(top) *t *koten *teten *matüt *atop SHI t togo toru toro t,s kote tisi – ato YEN t togo toru turu t kuti – – ato PSA t,s toxõ taru suru˜ t,s kas˜ı – masi ato AXU t,s bodo toŋgõ tü suketo t,s kes˜ı sisi masü ato MAN t taŋgo top – – t,r koten teren matit atop YWB t,s taŋgo top sarui turutop t,r koten teren matit atop DWB t,s taŋgotop salon silutop t,l – selen – atop KOM r rogo – ruro r kware – maru aro KOR ? ––– ? ––––

F banana bird louse

PAD *t *tüt *yet *ŋgut SHI ø tu yi go YEN ø tu yi gu PSA ø su yi aŋgu AXU ø sjü yi aŋgu MAN t tyut et ŋgut YWB t tit yet ŋgut DWB t sit yet ŋgut KOM ø, l rü(l) el gu KOR ? ––– 28 2. Proto Phonology and Phylogenetics

Table 2.7: Reﬂexes of *k

I (a,o) spirit skin to walk I (i,e,u) ﬂower eye to die

PAD *k *kVkVi *kat *ka/*ko- *k *ket *kerop *kum-

SHI x xogoi xa xo- k kete kEro ku- YEN x xoxEi xa xo- k ki kiro ku- PSA x xoxoi xa xo- k ki kriomogo ku- AXU x xoxi xa xo- k ki kiomogo ku- MAN k kugu kota ko- k ket kerop kün- YWB k kagui kotae ko- k ket kerop kim- DWB k kaxui kat/kotai ko-, ka- k ketek kerop kim- KOM x xwai xa xa- x xe xoro xumo- KOR x xayan xal xai- x xel – –

M (a, o) ﬁsh to order M (i) knife F urine stairs

PAD *k *rakaV *roka- *k *waki *k *yetuk *parik SHI x axae roxo- ? – ø yiti ti furi YEN x axae roxo- ? – ø yitu – PSA x axae ro- k waki ø yi su aﬁri AXU x axe o- k woki ø isi – MAN g rogoa – g wegi k erok werik YWB g ragae rogo- g wagi k yetok – DWB x laxai loxo- x waxi k etok halik KOM x – luxa- x waxi ø – fali KOR ? –– x ﬁx ? ––

*k Proto Awyu-Dumut *k occurs in initial, medial and final position, as illustrated in Table 2.7. Kombai and Korowai always have /x/, a fricativized form of *k, as a reflex of *k. Word-finally, Kombai drops *k. Digul Wambon has reflex /x/ for medial *k, while for initial and final *k it has reflex /k/. In Mandobo and Yonggom Wambon, medial *k becomes /g/, while as in Digul Wambon, initial and final *k remain /k/. Shiaxa, Yenimu, Pisa and Aghu have different reflexes initially and medially depending on whether *k occurs before a high vowel or a low vowel. When *k comes before a low vowel /a/ or /o/, these four languages have a reflex /x/, while before high vowels (/e/, /i/, /u/, /ü/), *k does not change. Medially, there is only one cognate set (‘knife’) in which *k precedes a high vowel, but in analogy to the behavior of word-initial *k before high vowels, it can be hypothesized that medial *k behaves similarly. Word-finally, *k is dropped in Shiaxa, Yenimu, Pisa and Aghu. Healey did think there was a word-final /k/ or /x/ in these languages, but only because he viewed the /k/, /g/ and /x/ that Drabbe writes after verb stems in these languages as part of the verb stem. However, from Drabbe’s description, it becomes clear that 2.3. Sound Correspondences 29 he includes these consonants after the verb stems in his lexicon to indicate how the realis marker is realized with that particular verb, namely as -k, -g, -x or -ox (Drabbe 1957:10). Similarly, Drabbe writes -d or -nd after Shiaxa verb stems in order to indicate whether the first person realis marker is realized as -d or -nd in that particular verb (Drabbe 1950:109). These consonants hence are not part of the verb stem, and I agree with Voorhoeve(2001:365) that *k (or a reflex /x/) does not occur word-finally in Shiaxa, Yenimu, Pisa or Aghu. The same misunderstanding probably led Healey to propose a *x for Proto Awyu-Dumut (where Voorhoeve and I have *k), stating that a /k/ found in Mandobo or Yongom Wambon was a reflex of that *x.

2.3.2 Voiced plosives Voiced stops *mb, *nd and *ŋg do not occur word-ﬁnally in Awyu-Dumut languages, as shown in Tables 2.8, 2.9 and 2.10.

Table 2.8: Reﬂexes of *mb

I sit chest arm M head tail testicle

PAD *mb *mba- *bema *be(do) *mb *ka(i)mb(i)a *wambut *wambura SHI b bo- – bedo b xeiba – – YEN b be- – bido b xeiba wobu – PSA b ba- b˜ı bidE b xaibã wobu woburo AXU b ba- besame bido b xabã wobugo wobio MAN mb mba- mbeman – mb kembian wombüt wambirap YWB mb mba- mbemit – mb – wambit – DWB mb mba- – – mb ambat wambit wambilin KOM mb mba- mbema mbei mb xambia wambu wamblo KOR b bai- – – b xabean – –

Table 2.9: Reﬂexes of *nd

I heart sago M meat bird thigh

PAD *nd *ndüma/*dVbo(p) *ndü *nd *kandü *pVndo *midi SHI d dibo do d kodo fodo midi YEN d dibo du d kodo fodo midi PSA d dubaro du d kadu – midi AXU d dübo dü d kudu – midi bigi MAN nd ndümarop ndu nd kandö wando – YWB nd ndimndop ndun nd – andoi – DWB nd dimlop ndu nd kandu – mindin KOM nd ndümo ndoü nd xundo fondo kinomindi KOR d debop daü ? ––– 30 2. Proto Phonology and Phylogenetics

Table 2.10: Reﬂexes of *ŋg

I blood M mouth penis tongue

PAD *ŋg *ŋgom *ŋg *maŋgot *teŋget *paŋgat SHI g goŋ g – tege fagE YEN g go g – tigi fagE PSA g go g – sigi fagE AXU g gõ g – segi fage MAN ŋg ŋgom ŋg maŋgot teŋget oŋgat YWB ŋg ŋgom ŋg maŋgot teŋget – DWB ŋg ŋgom ŋg maŋgotop teŋget hatgat KOM ŋg gom ŋg moŋgoro reŋge faŋgat KOR ? – ? –––

Shiaxa, Yenimu, Pisa, Aghu and Korowai have non-prenasalized reflexes of *mb, *nd and *ŋg in most instances. For Korowai, only reflexes for initial and medial *mb and initial *nd were found. Mandobo, Yonggom Wambon, Digul Wambon and Kombai retain the prenasalization of voiced stops. For Kombai, this is not apparent from the word list, but de Vries(1993) notes that voiced stops are always prenasalized in Kombai, though he does not always represent them as such. Healey does not think that Proto Awyu-Dumut voiced plosives were prenasalized, but rather thinks that the prenasalized reflexes in Mandobo and Yonggom Wambon are the result of the presence of consonant clusters in these languages (Hea- ley 1970:1000). However, I agree with Voorhoeve that these plosives were nasalized in Proto Awyu-Dumut, while the prenasalization of these sounds disappeared in Shiaxa, Yenimu, Pisa and Aghu (Voorhoeve 2001:371).

2.3.3 Nasals Nasals occur initially, medially and finally in Proto Awyu-Dumut. In Pisa and Aghu, a final nasal is often realized as nasalization on the preceding vowel. In Mandobo, sometimes final *m becomes final /n/. Table 2.11 gives examples of *m in Proto Awyu-Dumut and its daughter languages, while Table 2.12 illustrates the reflexes of Proto Awyu-Dumut *n in its daughter languages. 2.3. Sound Correspondences 31

Table 2.11: Reﬂexes of *m

I to drink beard star tendon to come

PAD *m *mi- *matu(t) *minVp *met *me- SHI m mi- – – me mode- YEN m mi- – – me mede- PSA m mi- masirõ mi – – AXU m mi- masü mi me – MAN m emi- matit minap temet me- YWB m mi- matit – met me- DWB m ami- – minop metmet – KOM m mi- maru mi melo me- KOR m mi- – – – –

M widow ripe warm F blood far off to die

PAD *m *ran *yamu *mamV *m *ŋgom *kamam *kum- kamo SHI m xamose – – ø goŋ – ku- YEN m xamsE –– ø go – ku- PSA m xamE –– ø, V˜ go xamã ku- AXU m – yomu – ø, V˜ gõ – ku- MAN m ran omu – n, m ŋgom koman kün- kamok YWB m – amui mamin m ŋgom kamam kim- DWB m – – mamin m ŋgom kommop kim- KOM m kamo yamu mamü m gom – xumo- KOR ? ––– ? –––

Table 2.12: Reﬂexes of *n

I mother EZ M be sleepy extinguish F woman sweat ﬁre

PAD *n *ni *nani *n *kunu ke *n *ran *koten SHI n wini ni n kero kono ge- buni- ø – kote YEN n wini neni n kiru kunu gi- buni- ø – kuti PSA n ni – n kunu gi- buni- ø, V˜ rã kasi AXU n – eni n kunuŋ gi- bunü- ø, V˜ rã kesi MAN n nou neni n kinum eren ge- – n ran koten YWB n noi nani n kinum kok ke- – n ran koten DWB n ni non n kinimxexe- – n lan – KOM n nani nani n xunu-ge – ø, n la(n) kware KOR n ni – ? –– ? –– 32 2. Proto Phonology and Phylogenetics

2.3.4 Liquids and glides

Table 2.13: Reﬂexes of *r

I woman ulcer M ear enough eye stairs

PAD *r ran rü(n) *r *turu(top) *kare *kerop *parik SHI r – ruŋ r toro yaxaro kEro furi YEN ? –– r turu kiro – PSA r rã ru r suru xarini krimogo aﬁri AXU ø ã üne ø – xaini kiomogo – MAN r ran orün r – kare kerop werik YWB r ran – r turutop kare kerop – DWB l lan – l silutop kit kelop halik KOM l la(n) lü r, l ruro – xoro fali KOR l lal – ? ––––

Table 2.14: Reﬂexes of initial *w and *y

initial *w tail knife testicle

PAD *w *wambut *waki *wambura(p) SHI ?––– YEN w wobu – – PSA w wobu waki wobu-ro AXU w wobugo woki wobio MAN w wombüt wegi wambirap YWB w wambit wagi – DWB w wabit waxi wambilin KOM w wabu waxi wablo KOR f – ﬁx –

initial *y bird garden urine

PAD *y *yet *yaküp, *yatim *yetuk SHI y yi – yiti ti- YEN y yi – yitu PSA y yi – yi su- AXU ø i – isi MAN ø, y et yogüp erok YWB y yet yagip yetok DWB ø et – etok KOM ø, y el yarimo – KOR y – yasim –

Proto Awyu-Dumut *r occurs word-initially and word-finally (Table 2.13). In Aghu, *r is always dropped. Digul Wambon and Korowai have /l/ as a reflex of *r. Kombai 2.4. PAD Consonant Reconstruction 33 has both /l/ and /r/ as distinct sounds; a minimal pair is ro ‘hole’ versus lo ‘sound’ (de Vries 1993:6). For initial *r, it is not possible to determine what Yenimu has as a reflex, whereas for Korowai, it is unknown what the reflex is of Proto Awyu-Dumut final *r. There are two glides in Proto Awyu-Dumut that only occur word-initially, *w and *y (see Table 2.14). They do not occur frequently, but cognate sets with initial *w or *y are found in nearly all Awyu-Dumut languages, clearly showing that these two sounds belong in the Proto Awyu-Dumut consonant inventory. The initial *w cannot be seen as a reflex of *p (remember that medial /w/ in Digul Wambon and Yonggom Wambon is a reflex of medial *p).

2.4 PAD Consonant Reconstruction

Having established regular sound correspondences, the Proto Awyu-Dumut consonant inventory can be reconstructed (see Table 2.15). It consists of eleven consonants. All consonants occur initially, the glides *w and *y exclusively occur initially. Voiced plosives and *r did not occur word-ﬁnally in Proto Awyu-Dumut, while voiceless plosives and nasals had no restriction on their distribution, occurring initially, medially and ﬁnally.

Table 2.15: PAD consonant inventory

labial alveolar velar voiceless plosive p t k voiced plosive mb nd ŋg nasal m n glide w y liquid r

This consonant inventory differs from the one presented by Voorhoeve(2001) in that it has no *s, while Healey(1970:999) also reconstructs an *f and and *x for Proto Awyu-Dumut. I propose that /f/, /s/ and /x/ in Proto Awyu-Dumut’s daughter languages are fricativized reﬂexes of Proto Awyu-Dumut *p, *t and *k.

2.5 Note on Korowai

For Korowai, it is not known in many instances whether it even has a reflex of a certain Proto Awyu-Dumut phoneme in a certain environment. For example, it cannot be determined whether Korowai has a reflex for Proto Aywu-Dumut initial *t. Although all Proto Awyu-Dumut phonemes except *ŋg have a reflex in Korowai, for 8 out the 27 consonant environments it cannot be determined whether Korowai has a reflex of a Proto Awyu-Dumut consonant in that environment. Thus it cannot be ascertained whether Korowai has a reflex for Proto Awyu-Dumut medial *p, 34 2. Proto Phonology and Phylogenetics initial *t, medial *t, initial and medial *ŋg, medial *m, medial *n and medial *r. In comparison, a reflex is found in all other examined languages in all environments; only Shiaxa and Yenimu each have one environment in which a reflex could not be determined (initial *w and initial *r, respectively). Furthermore, the reflexes that are found in Korowai are always only found in one cognate set, whereas reflexes in other languages are nearly always supported by occurrence in two or more cognate sets. Thus the phonological evidence is too weak to prove without a doubt that Ko- rowai is an Awyu-Dumut language, even though the correspondences that can be traced between Korowai, Proto Awyu-Dumut and the other languages are too numerous to be ascribed to mere chance or borrowing. What is clear is that Korowai is the most divergent language of all the languages examined and can only be considered distantly related to the other languages. For these reasons, Korowai will not be a part of the further discussion of Awyu-Dumut languages or the reconstruction of Proto Awyu Dumut in this book.

2.6 Sound Changes and Subgrouping

With a reconstructed Proto Awyu-Dumut consonant inventory, it is now possible to trace the sound changes between Proto Awyu-Dumut and its daughter languages. From the sound correspondences illustrated in Section 2.3 and summarized in Table 2.3, the following 20 regular sound changes can be distilled:

1. PAD initial and medial *p >/f/ in SHI, YEN, PSA, AXU and KOM 2. PAD initial *p dropped in MAN and YWB 3. PAD initial *p >/h/ in DWB 4. PAD medial *p >B in YWB and DWB 5. PAD medial *p dropped in MAN 6. PAD final *p dropped in SHI, YEN, PSA and AXU 7. PAD final *p >(underlying) /f/ in KOM, or dropped 8. PAD final *t dropped in SHI, YEN, PSA, AXU 9. PAD final *t >(underlying) /l/ in KOM, or dropped 10. PAD initial *k before low vowels >/x/ in SHI, YEN, PSA, AXU and KOM 11. PAD medial *k before low vowels >/x/ in SHI, YEN, PSA, AXU, DWB and KOM 12. PAD medial *k before high vowels >/x/ in DWB and KOM 13. PAD medial *k >/g/ in MAN and YWB 14. PAD final *k dropped in SHI, YEN, PSA, AXU 15. PAD final *k >(underlying) /x/ in KOM, or dropped 16. PAD *mb, *nd and *ŋg denasalized in SHI, YEN, PSA and AXU 2.6. Sound Changes and Subgrouping 35

17. PAD ﬁnal *m and *n dropped in SHI and YEN 18. PAD ﬁnal *m and *n dropped or realized as nasalized vowel in PSA and AXU 19. PAD initial *r >/l/ in DWB and KOM 20. PAD initial *r dropped in AXU

These sound changes can be used to make a preliminary subgrouping of Awyu- Dumut languages. Shiaxa, Yenimu, Pisa and Aghu share a number of innovations or sound changes. They drop final consonants *p, *t, *k, *m and *n and denasalize voiced stops *mb, *nd and *ŋg. Mandobo, Yonggom Wambon and Digul Wambon keep the final consonants and the prenasalization of the voiced stops. Thus there are two subgroups that can be established on the basis of final consonant deletion and nasalization of voiced stops; these two subgroups correspond to the Awyu and Dumut subgroups established by Healey(1970) and Voorhoeve(2001). A side note must be made concerning the final consonant deletion in Awyu languages. In some instances, Shiaxa has an extra vowel at the end of a word that the other three Awyu languages do not have; some examples are given in Table 2.16. The Shiaxa epenthetic vowel is hypothesized to be added to a word or morpheme in order to reach a CV.CV syllable structure; a CV.CV syllable structure is found in all Awyu languages. The consonant that precedes the extra vowel in Shi- axa corresponds to the final vowel found in Dumut languages. Hence Shiaxa in these instances retains the Proto Awyu-Dumut final consonant. In order for Shiaxa to be able to do so, Proto Awyu must have had underlying final consonants, much like Kombai. These underlying final consonants only surfaced when followed by a vowel.

Table 2.16: Shiaxa epenthetic vowel examples

breadfruit ﬂower hair

SHI yuwato kete moxo YEN yowo ki mo PSA yawo ki rõ AXU xã ki mu PA *yowo(t) *ke(t) *mu(x)/*rõ

MAN rawot ket ron YWB rawot ket ron DWB - mutmut lon, muk PD *rawot *ket *ron/*muk

KOM aluwo xe lo

PAD *rawot/*yawot *ket *ron/*muk

Thus Kombai and Proto Awyu are similar in that they both have underlying consonants, which are realized only when followed by a vowel. At the same time, Kombai 36 2. Proto Phonology and Phylogenetics and Proto Awyu also drop final consonants completely; that is, in many instances there is no proof that an underlying consonant is indeed present in a Kombai or Proto Awyu word. However, this commonality is not enough reason to subgroup Kombai with the Awyu languages; final consonant deletion can very well have been an independent innovation in Kombai, rather than an innovation it shares with the Awyu languages. The fact that, synchronically, the Dumut language Mandobo is starting to delete its final consonants (Drabbe 1959:5), suggests that final consonant deletion is an independent innovation that can take place in all Awyu-Dumut languages. On the basis of prenasalization of voiced stops, Awyu languages form a separate subgroup from Kombai and Proto Dumut; both Kombai and Proto Dumut retain the prenasalization of /b/, /d/ and /g/ found in Proto Awyu-Dumut, while the drop- ping of prenasalization is an innovation Awyu languages share. Kombai can also be subgrouped together with Digul Wambon, a Dumut language, on the basis of fricativization of *k to /x/ and of Proto Awyu-Dumut medial *r to /l/ change. However, both fricativization and a change of /r/ to /l/ are typologically quite frequent and are therefore a weak criterion for subgrouping languages together. Furthermore, fricativization of *k to /x/ also occurs in Awyu languages, only in fewer contexts than in Kombai and Digul Wambon. Based on the 20 regular sound changes presented above, it is unclear whether Kombai should be subgrouped with Awyu languages or with Dumut languages, and thus additional methods must be employed to determine its place in the Awyu- Dumut language family. The processes reflected in most of the sound changes, namely final consonant deletion, fricativization and prenasalization, are cross- linguistically common and not strong criteria to support the subgrouping of languages. Thus the use of phylogenetic methods below, as well as the reconstruction of proto morphology in the remainder of this book, will offer additional arguments that solidify the subgrouping of Awyu and Dumut languages.

2.7 Note on Awyu-Dumut Vowels

Two tentative subgroups of Awyu-Dumut were established on the basis of shared phonological innovations or the lack thereof. For vowels, it is not possible to repli- cate what was done for consonants, namely to establish what the Proto Awyu- Dumut vowels are based on cognate sets. The main reason for this is that Awyu- Dumut vowels show much variation, and there are few cognate sets in which the Awyu-Dumut languages display a similar pattern of vowels. This variation may exist because Awyu-Dumut vowels are more prone to change than Awyu-Dumut consonants, and because extensive vowel harmony occurs. There might also be ir- regularities in the recorded data, as Drabbe honestly admits that he finds it difficult to distinguish between different vowels: “It is often very difficult to hear the difference between the various vowels, especially between /a/ and /e/, /e/ and /i/, 2.7. Note on Awyu-Dumut Vowels 37

/i/ and /ü/, /o/ and /u/, /u/ and /ü/ and /ü/ and /ö/.”2 For these reasons, the reconstruction of any Awyu-Dumut vowels must be approached with caution. It can never be as rigorous as the reconstruction of the Awyu-Dumut consonants. That being said, it is possible to trace some vowel sound changes between Proto Awyu and its daughter languages, and between Proto Dumut and its daughter languages. The vowel inventories of Proto Awyu and Proto Dumut are based on which vowels occur in their daughter languages; Proto Awyu has seven vowels and Proto Dumut six, as shown in Table 2.17. The Proto Awyu-Dumut vowel inventory is the same as the Proto Awyu vowel inventory; Proto Awyu and Proto Awyu-Dumut both make a distinction between /e/ and /E/ that was lost in Proto Dumut and Kombai.

Table 2.17: Vowels in PA, PD and Kombai

front rounded central back front PA *i *ü *u close PD *i *ü *u KOM i ü u

PA *e *o close mid PD *e *o KOM e o

open PA *E mid

PA *a open PD *a KOM a

The vowel inventory for Proto Awyu presented here is the same as the one presented by Healey(1970:1002), but in the Proto Dumut vowel inventory Healey has two further vowels: *ö and *O. Healey reconstructs *O, a sound that occurs in no Awyu-Dumut daughter languages, when Mandobo has /a/ and Yonggom Wambon has /o/.3 It is true that Mandobo and Kombai have /a/ in two cognate sets (‘breast’ and ‘faeces’) where all the other languages have /o/. This could reflect a common innovation of Mandobo and Kombai, or they might be the only languages that retained the original sound; the two cognate sets provide insufficient evidence to decide what the best scenario is. However, reconstructing a vowel that has no reflex in a daughter language is not a good solution either; it is best in those instances to reconstruct two forms, so *am/*om for ‘breast’ and *a/*o for ‘faeces’. The other

2“Het is dikwijls zeer moeilijk het verschil te horen tussen de verschillende klinkers, en dat geldt vooral voor /a/ en /e/, /e/ en /i/, /i/ en /ü/, /o/ en /u/, /u/ en /ü/ en /ü/ en /ö/” (Drabbe 1959:7). 3Recall that Healey did not have access to Digul Wambon data. Note also that Voorhoeve left vowels out of his analysis. 38 2. Proto Phonology and Phylogenetics vowel that Healey includes in his Proto Dumut vowel inventory is /ö/, a vowel only found in Mandobo. Basing himself on just two languages, Healey concluded that his *ö changed to /u/ in Yonggom Wambon. However, with the inclusion of Digul Wambon data, we see that Digul Wambon, like Yonggom Wambon, has /u/ where Mandobo has /ö/, so it is more likely that Proto Dumut *u changed to /ö/ in Mandobo, as illustrated by the cognate sets given in Table 2.18. The occurrence of /ö/ is then a Mandobo innovation, which should not be reconstructed in Proto Dumut.

Table 2.18: Examples of PD *u >Mandobo /ö/

climb meat thunder earthquake fat

MAN törö- kadö komöt ondön tögö YWB turu- - kumut indum tugui DWB matulo- kadu kumut - tuxut

PD *turu- *kadu *kumut - *tuku(t)

Other sound changes in Proto Dumut vowels are Proto Dumut *ü into /i/ in Digul Wambon and Yonggom Wambon (Table 2.19), Proto Dumut *a into Mandobo /o/, usually before a syllable containing /a/, /o/ or /u/ (Table 2.20), and Proto Dumut *a into Mandobo /e/, usually before a syllable containing /e/ or /i/ (Table 2.21).

Table 2.19: PD *ü >YWB and DWB /i/

heart tail thorn hit to urinate

MAN ndümarop wombüt orün ü- erok tü- YWB ndimndop wambit arin in- jetok ti- DWB dimlop wabit alin in- etoksi-

PD *ndüm(a)rop *wambüt *arün *ü(n)- *etok tü-

Table 2.20: PD *a >MAN /o/

break rope feminine ﬁsh tail hear

MAN roŋgwamo- roŋgu rogoa wombüt ndot- YWB ra ŋgamo- raŋgui ragae wambit ndat- DWB lap ŋgamo- laŋgui lakhai wabit ndat-

PD *raŋgwamo- *raŋgu(i) *ragaV *wambüt *ndat- 2.7. Note on Awyu-Dumut Vowels 39

Table 2.21: PD *a to MAN /e/

knife sun torch young woman

MAN wegi teet eŋgot meŋgeet YWB wagi sat jaŋgot maŋgat DWB waghi sat jaŋgot lan maŋgat

PD *waki *tat *(j)aŋgot *(lan) maŋgat

Within the Awyu subgroup, there are three sound changes that tend to occur in the vowels. Aghu is the only language that has retained *ü; in the three other Awyu languages, it becomes /u/ (Table 2.22). When *u is reconstructed for Proto Awyu, Shiaxa has /o/, while Yenimu sometimes has /o/ and sometimes has /u/ (Table 2.23). Lastly, Proto Awyu *e frequently changes into /i/ in Aghu, Pisa and Yenimu (Table 2.24 on the next page).

Table 2.22: PA *ü to /u/ in PSA, SHI, YEN

banana to chop down to thunder younger brother

SHI tu ru- xo ru- kuda YEN tu ru- xu ru- kuda PSA su ru- xou ru- kude AXU sjü ü- ghü ü- küda

PA *tü *rü- *xü rü- *küda

Table 2.23: PA *u to Shiaxa /o/

enemy daughter louse to ascend to sleep voice

SHI kowE otobra go oto- kono ré- rogho YEN kufE otoba gu oto- kunu ré- ro PSA waki gu˜ subrã aŋgu su- kunu˜ ri- ru AXU kufe subã aŋgu osu- kumun’i-g u

PA *kufE *(o)tub(r)an *gu *otu- *kunun ré- *ru

These are the vowel sound changes that can be established for Proto Awyu and Proto Dumut with some certainty. In the reconstruction of lexical data, I often revert to reconstructing multiple forms with different vowels, or if there is a choice between more than two vowels, I reconstruct *V. It should again be emphasized that my reconstructed vowels are not rigorously supported by the comparative method, as Awyu-Dumut vowel data are messy and seemingly inconsistent in the picture they present. 40 2. Proto Phonology and Phylogenetics

Table 2.24: PA *e to YEN, PSA, AXU /i/

eyelashes ﬂower penis saliva tears

SHI kero-moxo kete tege xate efe ken-ogho YEN kiro-mo ki tigi xatipa efe kino PSA kiro-bi ki sigi mase kin-E AXU kio-b˜ı ki segi xasi kin’oxo

PA *kero- *kete *tege *xate ken-oxo mo/kero-bi

2.8 Phylogenetics

Healey and Voorhoeve used lexicostatistic computational methods to calculate the degree of similarity between Awyu-Dumut languages, leading to a subgrouping of Awyu and Dumut languages; their results are presented in Table 2.25. However, lexicostatistic methods are infamous for their subjectivity, relying more often than not on similarity judgments rather than on the comparative method to determine which lexical items are cognate. The lexicostatistical method can always be applied and yield results, whatever the quality of the data and the cognacy judgments is. As a check on Healey’s and Voorhoeve’s lexicostatistic methods, I apply phylogenetic computational methods to Awyu-Dumut lexical data.

Table 2.25: Lexicostatistic results of Healey and Voorhoeve

SHIYENPSAAXUMANYWBDWB

YEN 83/- PSA 51/52 53/– AXU 53/55 55/– 64/68 MAN 32/32 34/– 34/33 36/33 YWB 32/32 32/– 31/35 33/34 55/62 DWB –/34 –/36 –/34 –/51 –/50 KOM –/30 –/32 –/30 –/40 –/38 –/36

The left number is the percentage calculated by Healey, and the right number is the percentage calculated by Voorhoeve. Only Healey looked at Yenimu, while Voorhoeve had access to Digul Wambon and Kombai data that Healey did not have. Sources: Healey(1970:1014), Voorhoeve(2001:362)

Phylogenetic methods have been adapted from biology, where they are used to construct evolutionary trees of biological species. In the last ten years, phylogenetic methods have been used in linguistics to model the spread of Austronesian in Ocea- nia (Gray & Greenhill 2005, Gray et al. 2009), the origins of Indo-European (Gray & Atkinson 2003) and Bantu languages (Holden & Gray 2006), as well as the prehistory of Papuan languages (Dunn et al. 2005, Reesink et al. 2009), amongst many other applications. Nichols & Warnow(2008) offer a clear, useful description and evaluation of the use of phylogenetic methods in linguistics. I will use phylogenetic 2.8. Phylogenetics 41 methods to (a) comment on the treelike (or non-treelike) nature of the Awyu-Dumut data and (b) to determine the relationships between Awyu-Dumut languages, and especially to determine where Kombai fits in the Awyu-Dumut family tree, something that could not be determined using the sound changes established using the comparative method. These analyses are based on the 430-item lexicon described in detail in Section 2.1 and found in Appendix A. Phylogenetic methods are a powerful addition to the comparative method in determining the evolutionary histories of languages. Whereas the comparative method sees language evolution as entirely treelike, phylogenetic computational methods view language change as variation in different character states; the various character states (be they lexical items, morphological items, phonological sound correspondences, etc.) can each tell a different story. The comparative method assumes that the development of separate languages occurs in a linear, hierarchical fashion and aims to produce one specific family tree, leaving no room for alternatives. As a result, trees in the comparative method exclude any horizontal transfer. However, lexical evidence shows that multiple cognacy patterns occur within one language family, and that there are therefore multiple conflicting trees that can account for the data. It is also clear that horizontal transfers between different languages or subgroups of a language family occur. The seven lexical items in Table 2.26 each present a different cognacy pattern; not one word is coded exactly like another word. Some languages tend to go together, like Yonggom Wambon and Digul Wambon (in 5 out of 7 words), but they do not always do so, and in the case of dry, Digul Wambon groups with the Awyu languages rather than with Yonggom Wambon. On the basis of Table 2.26, seven different trees can be drawn to capture the development of the eight Awyu-Dumut languages. For phylogenetic models of language evolution, the fact that there are multiple evolutionary histories within one language family is not problematic, as will become clear below where I present the results obtained by using (a) a split decomposition network analysis and (b) a stochastic Bayesian tree model.

Table 2.26: Coded lexical items showing various cognate patterns

bone breadfruit chin dry ﬂower hair torch

SHI boge 1 yuato 1 gando-boge 1 kera 1 kete 1 noxo 1 xeino 1 YEN begi 1 yowo 1 tere-begi 2 soxo 2 ki 1 mo 2 sia 2 PSA bagi 1 yawo 1 makabi 3 se 2 ki 1 rõ 3 sia 2 AXU bigi 1 xã 2 te-bigi 2 so 2 ki 1 xõ 4 xasisia 3

MAN mirap 2 raot 3 ŋgandöm 1 tomap 3 ket 1 ron 3 eŋgot 4 YWB mit 2 rawot 3 maŋgor-ip 4 kerewet 4 ket 1 ron 3 yaŋgot 5 DWB mit 2 lawot 3 maŋgot 4 sok 2 ketek 2 lon, mux 3, 2 yaŋgot 5

KOM ﬁa 3 aluwo 3 gadu 1 roxe 2 xe 1 lo 3 kwaiyeria 6 42 2. Proto Phonology and Phylogenetics

2.8.1 Coding of the lexical data All 430 lexical items (except known borrowings) in the Awyu-Dumut dataset were carefully coded for cognacy following the consonantal sound correspondences and changes established using the comparative method. Table 2.26 shows the coding of seven lexical items, while Appendix A contains the coding for all lexical items in the 430-item word list.4 If a word has the same reflex in all Awyu-Dumut languages, all languages receive the same character value, indicating that the word is cognate throughout the language family. When a language has a different lexical item from another language, which might arise either through innovation or borrowing, that lexical item receives another character value, resulting in a new cognate class (also called ‘character state’). Thus for the word ‘bone’, Shiaxa, Yenimu, Pisa and Aghu form cognate class one, marked by character value (1), Mandobo, Yonggom Wambon and Digul Wambon form cognate class two, marked by character value (2), while Kombai forms a cognate class by itself, marked by (3). If a lexical item cannot be established as a cognate reflex using established sound changes, it gets a new character value. For example, the final /ek/ in Digul Wambon ketek ‘flower’ cannot be traced as originating in a regular sound change, and hence ketek is assigned to a new cognate class. On the other hand, Shiaxa kete ‘flower’ is cognate with other reflexes of the word ‘flower’ because Shiaxa is known to regularly add epenthetic vowels at the end of words, ‘retaining’ the final consonant, as it does for yuato ‘breadfruit.’ This coding method leaves room for multiple realizations of a word within one language. Digul Wambon has both lon and mux for ‘hair’, and these two realizations are each assigned to distinct cognate classes. This coding process resulted in a 8 x 430 matrix (8 x 430 = 3440 character states). The data was then converted to binary coding, with a distinct binary character for each combination of lexical item and cognate class. Table 2.27 contains the binary code for three lexical items from Table 2.26. Taking ‘breadfruit’ as an example, we see that each cognate class gets a separate column, and for each language it is noted whether it has a reflex in that cognate class (1) or not (0). Thus the first column asks “does this language have a reflex of yawol for ‘breadfruit’?” while the second column asks “does this language have a reflex of xã for ‘breadfruit’?” and the third column asks “does this language have a reflex of rawot for ‘breadfruit’?” and so forth. The resulting binary matrix (8x2280=18240 character states) forms the input for the phylogenetic analyses presented in the next two sections.

4The Awyu-Dumut 430-item word list is much longer than word lists usually employed for phylogenetic methods. It contains both basic and non-basic vocabulary. Below, a Bayesian phylogenetic analysis is also applied to a sub-list of basic vocabulary items within the 430-item word list, namely lexical items which also occur on the Swadesh 100 word list. 2.8. Phylogenetics 43

Table 2.27: Cognate sets from Table 2.26 expressed in a binary matrix

meaning breadfruit dry hair cognate set 1 2 3 1 2 3 4 1 2 3 4

SHI 1 0 0 1 0 0 0 1 0 0 0 YEN 1 0 0 0 1 0 0 0 1 0 0 PSA 1 0 0 0 1 0 0 0 0 1 0 AXU 0 1 0 0 1 0 0 0 0 0 1 MAN 0 0 1 0 0 1 0 0 0 1 0 YWB 0 0 1 0 0 0 1 0 0 1 0 DWB 0 0 1 0 1 0 0 0 1 1 0 KOM 0 0 1 0 1 0 0 0 0 1 0

2.8.2 NeighborNet network graph A network graph shows what patterns exist in the data and what the possible relationships between the different languages are. Such a graph is not acyclic and hence does not appear as a tree showing an explicit evolutionary scenario. Rather, it shows how the data do not fit a tree exactly (Nichols & Warnow 2008:763-764). A network graph partitions languages into groups according to whether or not they share a particular character state (Holton & Robinson 2012:134). When there are conflicting signals in the data, these are represented as multiple branches connecting the languages, or as Nichols & Warnow(2008) put it: “the internal nodes of [a network] graph do not represent ancestors of the given languages, but are introduced in order to make possible the representation of the conflict between the different splits that are produced in the data analysis” (Nichols & Warnow 2008:764). Parallel lines in a network graph may represent contact events, borrowing or homoplasy (Nichols & Warnow 2008:764). The greater the ‘webbing’ in the center of the graph, the less treelike the data is. The computer program SplitsTree (Huson & Bryant 2006) was used to generate a network graph of the Awyu-Dumut lexical data using the NeighborNet algorithm (Huson & Bryant 2006). I followed Gray et al.(2010) and Holton & Robinson(2012) in using gene content distances as the distance metric in the NeighborNet analysis. Figure 2.1 contains the network graph thus produced. 44 2. Proto Phonology and Phylogenetics

Figure 2.1: NeigborNet network graph of Awyu-Dumut languages

The network graph in Figure 2.1 suggests that there are several subgroups in the Awyu-Dumut family. Shiaxa, Yenimu, Pisa and Aghu group together on one end, while Mandobo, Yonggom Wambon and Digul Wambon are together on the other end. Kombai is part of neither subgroup, though closer to the Dumut subgroup than to the Awyu subgroup. Furthermore, Mandobo and Yonggom Wambon seem to form a further subgroup in the Dumut group, while the Awyu group can be divided into Shiaxa and Yenimu versus Pisa and Aghu. The parallel lines throughout the graph indicate instances of contact between Awyu-Dumut languages, while the length of the branches indicate each language’s differentiation from the proto language. To get a clearer picture of the degree of treelike signal present in the data, I calculated a delta score using the SplitsTree program. For a detailed explanation and discussion of delta scores, see Gray et al.(2010:3925-3928). The average delta score for the Awyu-Dumut languages is 0.26. Comparing this delta score to the delta scores calculated for much larger language families (.22 for Indo-European, but .41 for Austronesian, Gray et al. 2010), the Awyu-Dumut data seem to emit a modera- tely treelike signal.

2.8.3 Bayesian analysis Bayesian phylogenetic methods calculate a statistical tree that best ﬁts the data. For data that are not entirely treelike, a Bayesian analysis generates multiple trees and calculates the most likely tree, with percentage measures indicating how likely each split in a tree is. Nichols & Warnow(2008:774) provide a clear explanation of how Bayesian methods work: 2.8. Phylogenetics 45

Bayesian methods used in phylogenetics utilize the following basic al- gorithmic strategy. The algorithm begins with an initial model tree (i.e., a rooted tree with initial values for each of the associated parameters of evolution). Then, the algorithm follows a ‘random walk’ through ‘model tree space’, at each point computing the probability of the observed se- quences being produced by the given model tree. If this probability is higher than the previously computed probability, the move to the new model tree is accepted; if it is lower, the move is accepted with some lower probability. After a ‘burn-in’ period, the random walk is supposed to be in the stationary distribution, and then the algorithm randomly samples from the model tree space that it visits. This collection of model trees is then used to produce a probability distribution on the space of model trees. A standard output of a Bayesian analysis is a consensus tree (usually the majority consensus tree) of the sampled trees. Sometimes, however, the tree appearing the most frequently (called the ‘maximum posterior probability tree’) is returned.

It is important to note that the statistical tree generated by a Bayesian analysis is different from a family tree reconstructed using the comparative method. Whereas the comparative method leads to the reconstruction of one absolute tree, a Bayesian analysis generates the tree that is statistically most likely and that best fits the data. Bayesian methods do not produce a single tree but a probability distribution on the set of trees that they generate. The comparative method is essentialistic and absolute, while Bayesian methods are probabilistic, taking into account the non- essentialist nature of language. A Bayesian analysis using a Markov Chain Monte Carlo (MCMC) algorithm was performed on the Awyu-Dumut lexical dataset with several different models, using the MrBayes program (version 3.2.1, Ronquist & Huelsenbeck(2003)). Each model was run for 750,000 generations with a sampling rate of 150 and a burn-in of 25 percent, which discarded the first quarter of all samples. The number of chains used was four and the number of runs done on the data was two. The best perfor- ming model was a gamma distribution model which sampled across the substitu- tion model space.5 Korowai was used as an outgroup to root the Bayesian tree. Figure 2.2 shows the maximum clade credibility tree thus generated. The first thing to note is that all branches and subbranches have 100% probability, meaning that all of the optimal trees make these divisions. The Bayesian tree, like the Neigh- borNet network graph, supports the subgrouping of Awyu languages versus Du- mut languages. It further shows that Kombai is closer to the Dumut languages than

5This model was compared to a gamma distribution model in which all substitu- tions had the same rate. The computer program Tracer (Rambaut & Drummond 2009, http://tree.bio.ed.ac.uk/software/tracer/) was used to calculate Bayes Factors to determine which model performed best (see Kass & Raftery(1995) for a discussion of Bayes Factors). The different models performed almost equally well, but the gamma distribution model with no a priori substitu- tion model did slightly better. The Tracer program was also used to test the effective sample size of the various analyses to ascertain that the samples were not auto-correlated. 46 2. Proto Phonology and Phylogenetics to the Awyu languages, forming a subgroup with them, as also concluded by Voor- hoeve using lexicostatistic methods. Within the Dumut languages, Mandobo and Yonggom Wambon form a subgroup, whereas the Awyu subgroup consists of two further subgroups: Shiaxa and Yenimu versus Aghu and Pisa.

Figure 2.2: Bayesian rooted tree based on 430-item word list

In order to see whether a different result would be achieved if only traditional ‘basic vocabulary’ were considered, the same model was run on a subset of the 430-item word list containing basic vocabulary items also found in the Swadesh 100 word list; this subset consisted of 91 lexical items. This analysis arrived at a very similar Bayesian tree supporting the same subgrouping as the tree based on the 430-item word list, as shown in Figure 2.3. This can be interpreted as meaning that this tree is robust, and that the full 430-item word-list containing non-basic vocabulary emits the same phylogenetic signal as the subset consisting of only basic vocabulary. 2.9. Summary 47

Figure 2.3: Bayesian rooted tree based on Swadesh subset of word list

2.9 Summary

This chapter conﬁrmed Awyu-Dumut as a language family using the comparative method, before turning to phylogenetic methods to establish the subgroups of the language family. The application of the comparative method to the lexical data in this chapter is more transparent - and perhaps more rigorous - than the application of the comparative method by Healey(1970) and Voorhoeve(2001). Sound correspondences and changes were established before the subgroups of the language family were established, an approach not taken by Healey and Voorhoeve. Further- more, the current analysis was based on a complete 430-item word list for all eight Awyu-Dumut languages and Korowai. Healey based his analysis on around 200 lexical items in six Awyu-Dumut languages. Although Voorhoeve had more data than Healey, he only had around 200 words for both Digul Wambon and Kombai, and did not include Korowai at all. Extra data kindly collected within the last year by HongTae Jang and Song-Kyu Chi for Digul Wambon and Kombai provided me with a complete dataset on which to base my analyses. Phylogenetic methods showed that a statistical tree with high probabilities for the placement of all Awyu-Dumut languages can be generated. Awyu and Dumut have been shown by multiple methods to be two subgroups, while the Bayesian analysis indicated that Kombai is best classiﬁed as forming a subgroup with the Du- mut languages. In the reconstructive efforts of the next chapters, Proto Awyu and Proto Dumut proto morphemes will be reconstructed before Kombai data is incorporated to reconstruct a Proto Awyu-Dumut form (thus no reconstruction is done 48 2. Proto Phonology and Phylogenetics for the potential sub-node that would link Kombai and Dumut languages together). Korowai will not be taken into consideration in these reconstructions, as its position within the Awyu-Dumut language family is deemed too uncertain, and because previously published research has shown that Korowai has little bound morphology in common with the other Awyu-Dumut languages (de Vries et al. 2012). The conclusion of this book will discuss whether the subgroups generated by the phylogenetic methods applied to the lexicon are also supported by the shared innovations and retentions found in Awyu-Dumut morphology. 3

Nouns and Adjectives

Nouns and adjectives are in the minority in Awyu-Dumut languages; in frequency and complexity they are greatly superseded by verbs. The nouns and adjectives that do occur in Awyu-Dumut languages have very little morphology. Some possessive and number morphology does exist, as described in Sections 3.3 and 3.4. In addition, Awyu-Dumut kinship terms have morphology not found on other nouns (Section 3.5). This chapter begins with a description of Awyu-Dumut adjectives (Section 3.1) and compound nouns (Section 3.2) and concludes with a discussion of nominal coordinators (Section 3.6).

3.1 Adjectives

All Awyu-Dumut languages have adjectives that function as modifiers of nouns, though they are infrequent. The adjectives that one is most likely to find are words that specify dimension, value and age, thus words for ‘small, large, good, bad, old, new’ and so forth. These types of adjectives are likely to occur in any adjectival inventory, however small it is (Dixon 1982:46). An Awyu-Dumut adjective follows the noun it modifies but need not follow the noun directly. In the three Dumut languages as well as in Kombai, adjectives can be reduplicated to yield plural adjectives; a plural adjective may occur when the noun that the adjective modifies has a plural referent. Adjectives can be either partly or entirely reduplicated. In (1), there are multiple men who came, while (2) talks about more than one pig, which is indicated by the second adjective ‘big’, rather than the first adjective ‘young’.

(1) Kagup mba-mbari mi-gin-in. human adult-adult come-REAL-NON1PL ‘The adult humans came.’ Yonggom Wambon (Drabbe 1959:118)

(2) U mun mbo ndigio-gen-on do ko-goneni ke-gen. pig young that fatten-REAL-NON1PL CONN big-big be-REAL[NON1SG] ‘When they fatten the young pigs, they become big.’ Mandobo (Drabbe 1959:22) 50 3. Nouns and Adjectives

In the same languages where we find reduplication, namely in Dumut languages and Kombai, we also find intensifiers that, by and large, mean ‘very’ and thus emphasize that the noun has the quality expressed by the adjective to a high degree. The Mandobo intensifier is mep (3). (3) u mep koneni togümo-r-o-an. pig INTENS big buy-REAL-1SG-PST ‘I bought a very large pig.’ Mandobo (Drabbe 1959:107) The intensifiers found in Awyu-Dumut languages are listed in Table 3.1. The intensifiers follow the adjective, except the Yonggom Wambon intensifier monop and the Mandobo intensifier mep, which precede the adjective.

Table 3.1: Dumut and Kombai intensiﬁers

language intensiﬁer(s) Yonggom Wambon monop Digul Wambon mbalin Mandobo arap, eren, kariagap, mep Kombai rabo

Awyu-Dumut adjectives occur independently and have no morphology that marks them as being adjectives, except in Yonggom Wambon, Aghu and Kombai, which have optional adjectival markers that follow the adjective. The Yonggom Wambon adjective markers are -op, -matan and -mban (Drabbe 1959:118), the Aghu adjective marker is -axa (4), while the Kombai adjective marker is -xe. (4) tadi-axa large-ADJ ‘large’ Aghu (Drabbe 1957:6) To summarize, all Awyu-Dumut languages have specific adjectives that modify nouns. In Kombai, Aghu and Yonggom Wambon, an optional adjectival marker may occur. In Dumut languages and Kombai, adjectives can be reduplicated when the noun they modify is plural. Dumut and Kombai adjectives can be modified by intensifiers. The appearance of adjectives in Awyu-Dumut texts is very infrequent.

3.2 Compound Nouns

A compound noun consists of two or more nouns, and its meaning is either more speciﬁc or entirely different than the combined meanings of the words that make up the compound. Shopen(2007) distinguishes three different kinds of compound nouns:

1. Endocentric compounds denote a subclass of items referred to by one of their elements; this element can be treated as the head of the 3.2. Compound Nouns 51

compound [...] The semantic relationship between the components of endocentric compounds can be of a genitive or part-whole type (soap-dish). Or one component may modify the other (e.g. black- bird). 2. Exocentric compounds denote something which is different from either of their components (e.g. egghead); they cannot be reduced to any one of their components. 3. Coordinate compounds consist of two juxtaposed nouns which refer to a unitary concept, e.g. Bengali chele-mee (boy-girl) ‘children’, San- skrit mata-pitarau (mother-father) ‘parents’, Russian hleb-solj (bread- salt) ‘Russian hospitality’; they can be considered a kind of exocentric compound since their meanings equal that of neither component. (Shopen 2007:30-31)

In Awyu-Dumut languages both endocentric and coordinate compounds occur. Drabbe refers to endocentric compounds as ‘compounds of the Germanic ‘tree trunk’ type, as found in all known Papuan languages,’1 noting that the modifying noun comes first and receives stress. Table 3.2 gives examples of endocentric nouns from all Awyu-Dumut languages,2 where stress is indicated on the first noun of the compound by a grave accent ( `). The modifying noun does not always come first in an Awyu-Dumut endocentric compound noun; it may come second in Mandobo and Aghu, and then still receives stress. For Mandobo, Drabbe notes that the modifier only comes second when the compound means ‘a type of X’, where X is specified by the head noun and what type it is by the modifying noun, as in in tomèt, in komütì, in ndambèt, which are all different types of trees (where in=tree). Similarly, in Aghu, a general noun such as ‘bird’ can come first, and the type of bird is then specified by the second noun. Thus kuso is the general name for marsupials, rats and mice, while kuso-baxì means ‘kangaroo’, kuso-maxikò is a ‘field mouse’, and kuso-yowò refers to a ‘rat’ in Aghu. Coordinate compounds are not as frequent as endocentric compounds in Awyu- Dumut languages, and examples of coordinate compounds are only found in the grammar descriptions of Yonggom Wambon, Mandobo and Kombai. For Yonggom Wambon and Mandobo, Drabbe clearly distinguishes coordinate nouns from endocentric nouns, noting that in coordinate compounds, both nouns receive stress (Drabbe 1959:12, 117). The two nouns that make up the coordinate compound noun in Yonggom Wambon can be joined by the connective -o. Most coordinate compounds fall within the semantic domains of family or edibles, as becomes clear from the list of coordinate compound nouns given in Table 3.3.

1“samenstelling van het Germaanse type ‘boomstam’, die men vindt in alle bekende Papoea- talen”(Drabbe 1959:12) 2Yenimu is not included in Table 3.2, as the little data available on Yenimu does not contain endocentric compound nouns. 3The anthropologist Stasch notes that, in Korowai, the dyad ‘cassowary-pig’ is metaphorically used to refer to the relationship between uncle (mother’s brother) and nephew (sister’s son). He 52 3. Nouns and Adjectives

Table 3.2: Awyu-Dumut endocentric nouns

word meaning literal meaning

YWB Kawòn mogot mouth of Kao river Kao-river mouth yùr andui yur tree trunk yurtree-tree trunk itìt wagot cassowary egg casssowary-egg DWB el-ap nest bird-house kelop-ok tear eye-water MAN ndù-mirop place of sago preparation sago-place mbümàn-mirap spine back-bone ìn-mogüp tree top tree-top taŋgòn-top armpit cavity armpit-hole AXU woküè-õ cassowary feathers cassowary-hair naŋgì-maga dogs teeth dog-teeth wì-ba pigs kennel pig-kennel PSA uì wobu pigs tail pig-tail karuwè mugo cassowary egg cassowary-egg SHI jì ro tree fruit tree-fruit KOM wòdei-bei riverbank river-bank

Sources: Yonggom Wambon (Drabbe 1959:117); Digul Wambon (Jang 2008:70-71); Mandobo (Drabbe 1959:12); Aghu (Drabbe 1957:4); Pisa (Drabbe 1950:97); Shiaxa (Drabbe 1950:97); Kombai (de Vries 1993)

Table 3.3: Awyu-Dumut coordinate compound nouns

word meaning literal meaning

YWB yanoi yanati parents his mother-his father yasarip yakopari couple his wife-her husband itir-o-kurup big game animals cassowary-forest pig3 tenor-o-kayok vegetables genumu plant-climbing plant ndun-o-kom vegetarian food sago-? tir-o-yagip garden produce banana-garden ŋgin-o-kerop face head-eye MAN tere-ŋgoromut insects grasshopper-spider itit-taon big wild animals cassowary-forest pig ndu-yog-anen vegetarian food sago-garden.produce-eat raŋgen-mom family sisters son-mothers brother KOM momo-laŋge family mothers brother-sisters son

Sources: Yonggom Wambon (Drabbe 1959:117); Mandobo (Drabbe 1959:12-13); Kombai (de Vries 1987)

notes that “part of what the statement says about the uncle-nephew relationship is that this bond stands out markedly against the broader run of human relations, in the strange and valuable manner that pig and cassowary stand out against the broader run of animals. Korowai do not recognize pigs as belonging to the taxonomic class “mammal” and they do not consider cassowaries “birds”. The animals are too much bigger than other members of those classes and too morphologically and be- haviorally different from them. They are each one of a kind. Instead, the two are paired together as “pig-cassowary,” based on their common taxonomic oddity, their anthropomorphization, and their extraordinary value as game. The aphorism says in effect that uncle-nephew relations are the valu- 3.3. Possession 53

3.3 Possession

The default way to express possession in Awyu-Dumut languages is to juxtapose the possessor and the possessee (with the possessor ﬁrst), and to place prosodic stress on the possessee. In addition to juxtaposing possessor and possessee, all Awyu- Dumut languages can make use of possessive pronouns to encode possession. The possessive pronoun is placed between the possessor and the possessee, much as in Dutch ‘Ruth haar proefschrift’ (literally: Ruth her dissertation, so Ruth’s dissertation). In Awyu-Dumut languages there need not be an explicit reference to the possessor, although there can be. Possessive pronouns and their exact usage are discussed in further detail in Section 4.2. Pisa can use possessive pronouns to express possession, but it more frequently makes use of a possessive marker na, ‘of’. This na4 appears between the two nouns that stand in a possessive relation to each other, for example nabo na xãy, literally ‘father of house’, which means ‘father’s house’. The possessive marker na is also found in Mandobo.5 For example, in Mandobo, both u gatèt (juxtaposed, with stress on the possessee) and u na gatet mean ‘the pig’s leftover food’. Drabbe claims that possessive na also occurs in Aghu. However, as it only occurs in Aghu before a vowel-initial possessee and is then shortened to n, it is better analyzed as a transitional nasal; both vowel-initial and consonant-initial possessees are juxtaposed to their possessor in Aghu. There are no occurrences of a possessive marker na in Yonggom Wambon, Digul Wambon or Kombai. Example (5) from Mandobo contains the three possessive constructions found in Awyu-Dumut languages: juxtaposition (u gater), possessive pronoun (yeŋgine n’anemo) and possessive marker na (anemo na gatet).

(5) U gater ete komo ŋgun-owon ge na-ŋgaŋgo, yengine pig food.rests CONN put burn.magically-1PL be say-CAUS 3PL.POSS n-anemo na gatet komo ŋgun-on. tr.nasal-son.in.law POSS food.rests put burn.magically-NON1PL ‘Thinking/saying ‘let us burn the pig’s food rests’, they burn their son-in- law’s food rests (instead).’ Mandobo (Drabbe 1959:29)

3.4 Plurality

It is rare for the plurality of a noun to be overtly expressed in Awyu-Dumut languages. Instead the listener must rely on context to determine whether a certain noun is singular or plural. However, if the context does not contain enough clues, able category of oddballs of kinship experience” (Stasch 2009:118). 4It must be noted that the 1SG possessive pronoun is na in all Awyu-Dumut languages except Mandobo. As possessive pronouns also occur in possessive contexts, 1SG.POSS na is a likely source of the possessive marker na. 5Mandobo has ne and n’ as allomorphs of na. 54 3. Nouns and Adjectives the speaker may reduplicate a noun to indicate its plurality. As Drabbe notes for Mandobo, “as an indication of plurality the noun is sometimes reduplicated, but only if otherwise the meaning is not clear. One at times hears küap-küap ‘men’, and tarap-tarap ‘houses’; however, most often one must determine from the context whether a singular or a plural meaning is meant.”6 Reduplication of nouns is attested in all Awyu-Dumut languages except Shiaxa, Yenimu and Kombai, whose grammars contain no data on reduplication of nouns. In Awyu languages, in addition to reduplication, the third person plural pronoun yoxo can be used to indicate that the noun that it follows has a plural referent. The verb conjugation is then also plural (6).

(6) Wi yoxo temko-x-enã pig 3PL go.away-REAL-NON1PL ‘The pigs ran away.’ Aghu (Drabbe 1957:5)

As in Aghu, yoxo can indicate plurality in Pisa and Shiaxa, for example Pisa éna fa yoxo means ‘his friends’, while Shiaxa chopoxoy yoxo means ‘the commanders’. The plural pronoun yoxo was only found to follow subject nouns, not object nouns. In addition to reduplication and the use of the third person pronoun in Awyu languages, the plurality of a noun can be expressed by reduplicating the adjective that modiﬁes it (as described in Section 3.1) or by using an iterative verb stem (as described in Section 8.2).

3.5 Kinship Terms

Kinship terms are frequent in Awyu-Dumut languages because the use of personal names is avoided. Stasch explains for the Korowai:7

Although Korowai have proper names, speakers rarely call people by name in their presence [...] Instead, Korowai call each other by kinship terms and kinship-based expressions emphasizing people’s locations in social bonds. (Stasch 2009:77)

There are three interesting things to note about kinship terms in Awyu-Dumut languages. First, some Awyu-Dumut languages have separate forms of address and reference, while others do not. Second, kinship terms in all Awyu-Dumut languages have their own plural marker. And third, Awyu-Dumut kinship terms tend to co- occur with possessive pronouns, in certain cases merging with the possessive pronoun. Table 3.4 contains all known Awyu-Dumut kinship terms.

6“als aanduiding van het meervoud worden sommige substantieven wel eens verdubbeld, maar alleen als de bedoeling anders niet duidelijk is; zo hoort men wel eens küap-küap ‘mensen’, en tarap- tarap ‘huizen’; meestal echter moet uit de samenhang worden opgemaakt of een enkelvoud of een meervoud wordt bedoelt” (Drabbe 1959:16). 7Although Korowai is only distantly related linguistically to Awyu-Dumut languages, as shown in Chapter2, culturally they are quite similar. 3.5. Kinship Terms 55

Table 3.4: Awyu-Dumut kinship terms

DWBYWB YWB MAN MAN (address) (reference) (address) (reference)

F mbap natiop nati nombap neti M ni, nayap enanop noi néyap nauwi EB (na)net anerop (a)net anérop (a)net YB nana(n) nan nanan nan EZ onop aniop (n)ani anéyap ani YZ yanaŋgot randui mbiat nanan nan, embiat MM/FM asek nomben ayèyop ayè MF/FF embayop mbaeop mbae ambayop ambai WM konép konép MB mom mom mom ZS raŋgen raŋgen

AXUPSASHIYENKOM

F eto bo ayè ayè are M api ni wini wini yeni EB exo kapù yaxà yaxò nai YB küda kudè kudà kudà wabü EZ eni de ni néni nani YZ küda kudè kudà némé wabü MM/FM itime wi anè anè ama MF/FF anisi si nimu y’enéti ndare WM as˜ı xuni MB omõ mõ momo ZS ag˜ı laŋge

Sources: Digul Wambon (Hughes 2009), (de Vries & Wiersma 1992), (Jang 2008:84, 129); Yonggom Wambon (Drabbe 1959:117, 120, 168); Mandobo (Boelaars 1970), (Drabbe 1959); Aghu (Drabbe 1957:7, 48); Pisa (Drabbe 1950:135); Shiaxa (Drabbe 1950:135); Yenimu (Drabbe 1950:135); Kombai (de Vries 1987)

For Yonggom Wambon and Mandobo, separate address forms and reference forms of kinship terms were found in Drabbe(1959:120) and Boelaars(1970). The address form is used when speaking to the family member directly, while the reference form is used when talking about a speciﬁc family member in conversation; an English equivalent would be the use of the two terms ‘dad’ and ‘(my) father’. For Yonggom Wambon, the element -op is added to the reference form to yield an address form.8 Most Mandobo address forms are also a combination of the reference form and - op or -ap. For Digul Wambon, it can be hypothesized that nayap ‘mother’ and embayop ‘grandfather’ are address forms, while all other Digul Wambon kinship terms listed in Table 3.5 are reference forms. Thus the three Dumut languages, Yonggom Wambon, Mandobo and Digul Wambon, each have kinship address forms ending in -op or -ap, as well as separate reference kinship terms. For Kombai, de Vries (1987:108) states that he did not ﬁnd different forms for terms of address and of

8Except in the case of ‘younger sister’, where the reference form is mbiat but the addressee form is not mbialop, but randui. Randui literally means ‘little girl’ (Drabbe 1959:120). 56 3. Nouns and Adjectives reference. The same can be said of the Awyu languages described by Drabbe: the grammars do not contain examples of an opposition between kinship reference and kinship address terms. Thus Dumut languages make a distinction between address and reference kinship terms, while Awyu languages and Kombai do not. In addition to having an elaborate system of reference and address kinship terms, most Awyu-Dumut languages have a separate plural marker used only with kinship terms. A speciﬁc kinship plural marker occurs in the three Dumut languages as well as in Aghu and Pisa. Table 3.5 lists the plural sufﬁxes used with kinship terms in these languages, while (7) contains two plural kinship terms from Yong- gom Wambon.

(7) Ngurŋgoron ŋga raga-t te: ŋgo nan-ŋgui Ngurŋgoron ERG say-REAL[NON1SG] CONN: 2SG.POSS younger.brother-PL e, ŋgo nen-ŋgui e Matiram ŋga ŋgotonde e-no TOP 2SG.POSS elder.brother-PL TOP Matiram ERG kill eat-NON.CLOSE mir e top me agumo... bones TOP ditch come throw.in... ‘Ngurŋgoron said: your younger brothers, your older brothers, Matiram killed them, ate them and put the bones in a ditch...’ Yonggom Wambon (Drabbe 1959:153-154)

The plural marking on Awyu-Dumut kinship terms, the only nouns with a plural marker, might have originated in a cultural practice of taboo and avoidance. Stasch notes that Korowai mothers-in-law and their sons-in-law stand in a taboo relationship and out of politeness avoid speaking of each other in the singular, pre- ferring plural forms instead (Stasch 2009:86). The kinship plural marker can be reconstructed as *-gi for Proto Awyu, *-ŋgu(i) for Proto Dumut and *-ŋgV for Proto Awyu-Dumut.

Table 3.5: Awyu-Dumut kinship plural markers

Yonggom Wambon -ŋgui Digul Wambon -ŋguye Mandobo -ŋgu Aghu -gi Pisa -gi

Concerning kinship terms and possession, Drabbe(1950:101) already notes that “in some Papuan languages, personal or possessive pronouns are tightly interwoven with kinship terms.”9 Again, an anthropological explanation can be given: in Awyu- Dumut languages, kinship terms express a dyadic relationship. They do not refer to one person, as such, but to the relationship in which that person stands to the

9“In sommige Papoea-talen worden persoonlijke of bezittelijke voornaamwoorden heel nauw verbonden met verwantschapsnamen” (Drabbe 1950:101). 3.6. Coordination of Nouns 57 speaker or another person (Stasch 2009:72). Awyu-Dumut kinship terms are not inherently possessed, but it is rare for a kinship term to occur without the ‘possessor’ of that particular kin being named. Thus Drabbe notes that Aghu eto ‘father’ is realized as neto when it concerns the speaker’s own father (neto means ‘my father’), while geto means ‘your father’. In neto, eto has combined with the 1SG possessive pronoun na, while geto is ga (2SG.POSS) plus eto. Aghu neto has the same form as Yonggom Wambon and Mandobo nati/neti. The initial /n/ or /nV/ in Mandobo and Yonggom Wambon neti/nati, as well as in many other instances, should be analyzed as expressing possessiveness rather than as being part of the kinship term. This reanalysis of n(V) would explain why Drabbe talks about ‘elder sister’ as ani in his texts but lists nani for the same lexical item in his wordlist. The difference between Digul Wambon mbap and Mandobo nombap is then also easily explained, as well as the difference between nanan and nan for ‘younger brother’ in Dumut languages. The initial /n/ could have the 1SG possessive pronoun as its source, but Drabbe notes that, in Pisa, na is a general possessive marker, as discussed in Section 3.3 on possession. Drabbe also notes that kinship terms which begin with /n/ or /na/ are not always possessed by the person speaking. They rather express an absolute, deﬁnite form of the kinship term, for example nabo and namun ‘the Father’ and ‘the Son’ cited by Drabbe as terms for God the Father and God the Son of the Christian Trinity.10 Another indication that the initial /n/ does not express (or no longer expresses) 1SG meaning is that, in Yonggom Wambon, na nati is used to say ‘my father’, while ŋgo nati means ‘your father’; nati is also used when the possessor is not the speaker himself. What is clear is that the initial /n/ found in many Awyu- Dumut kinship terms expresses possession, and that it probably has its origin in another possessive marking strategy.

3.6 Coordination of Nouns

According to a typological study of noun coordination done by Stassen(2000), there are two strategies that languages can employ to coordinate nouns, either by using an element meaning ‘and’ or an element meaning ‘with’. Stassen notes that, typologically, pure WITH languages are rare; rather, elements meaning ‘with’ tend to develop into coordinators with meaning ‘and’ in the world’s languages. Awyu-Dumut languages all have a reﬂex of a comitative marker kup that functions as a nominal coordinator, and thus can be classiﬁed as WITH languages, albeit with a grammaticalized ‘with’ element. Dumut languages, in addition to having a grammaticalized comitative marker, also have a grammaticalized focus marker te that functions as a nominal coordinator. What distinguishes the coordinating function of the comitative marker and focus marker from their original functions is that they are repeated after each element of the enumeration when they have a coordinating function. Thus ‘X kup Y kup’

10Note, however, that it must have been Drabbe who, as a Catholic missionary, introduced these terms into Awyu-Dumut languages; they can hardly have been indigenous. 58 3. Nouns and Adjectives means ‘X and Y’, while ‘X kup Y’ means ‘X with Y’ in Awyu-Dumut languages. The doubling of a coordinator is common in verb-ﬁnal languages (Shopen 2007:46). In Section 3.6.1, the comitative marker kup is discussed, whereas focus marker te and its development into a nominal coordinator are discussed in Section 3.6.2. Section 3.6.3 describes a third nominal coordinator found in Digul Wambon and Kombai, namely =o. It will be shown that, for Proto Awyu-Dumut, a nominal coordinator *kup can be reconstructed, which has the reﬂex *ku in Proto Awyu and *kup in Proto Dumut. The focus marker te is also hypothesized to have already developed into a nominal coordinator in Proto Dumut.

3.6.1 Comitative kup as nominal coordinator

In Awyu-Dumut languages, the element *-kup11 functions as a coordinator of nouns, occurring after all members of a coordination. In Mandobo, köp does not occur as a coordinator, but only as a comitative marker (16). Neither does Drabbe give an example of Yenimu ka functioning as a coordinator. Examples (8-13) illustrate the coordinating role of kup (or its cognate) in Yonggom Wambon, Digul Wambon, Aghu, Pisa, Shiaxa and Kombai, respectively.

(8) Ran kup, kagup kup kimbarukmo Wambon woman COORD man COORD swim.IT Wambon.area mendi-r-in-an. come-REAL-NON1PL-PST ‘Women and men came to the Wambon area swimming.’ Yonggom Wambon (Drabbe 1959:146)

(9) Belanda-n=o kav=e loti-ŋgup susu-ŋgup Dutch-tr.nasal=CONN man=TOP bread-COORD milk-COORD ene-mo-knd-e eat.IT-do-REAL-NON1PL ‘Dutch men eat bread and milk.’ Digul Wambon (de Vries & Wiersma 1992:71)

(10) Ã go xobas˜ı go su˜ ba-x-enã. woman COORD men COORD feast prepare-REAL-NON1PL ‘Women and men prepare the feast.’ Aghu (Drabbe 1957:80)

11ku in Pisa (Drabbe 1950:107), xu in Kombai (de Vries 1993:46-48), ko in Aghu (Drabbe 1957:6) and Shiaxa (Drabbe 1950:107), ka in Yenimu (Drabbe 1950:107) and köp in Mandobo (Drabbe 1959:55). Shiaxa ko and Mandobo köp can be established as cognate to -kup (found in Digul Wambon and Yonggom Wambon) on the basis of regular sound correspondences, namely

1. PA *u >Shiaxa /o/ 2. PD *u >Mandobo /ö/

Aghu ko and Yenimu ka cannot be traced to kup on the basis of regular sound correspondences but are considered cognates based on their form (kV) and function (as nominal coordinators). 3.6. Coordination of Nouns 59

(11) nu-ku gu-ku I-COORD you-COORD ‘I and you’ Pisa (Drabbe 1950:107)

(12) waye ko wawini ko his.father COORD his.mother COORD ‘his father and his mother’ Shiaxa (Drabbe 1950:107)

(13) Ola-xu xe-wabü Fiabo-xu xuroxuro Ola-COORD 3SG.POSS-younger.brother Fiabo-COORD each.other unafa-no hit-[REAL].NON1PL ‘Ola and his younger brother Fiabo hit each other.’ Kombai (de Vries 1993:48)

The element *kup, when it does not occur in a coordination environment, can be described as a comitative marker that has the interpretation ‘with’ (14-17) or ‘also’.

(14) Wano nde-ma-l=e nexep-kup ka-wa fellow come-do-REAL[NON1SG]=TOP he-COMIT go-1PL ma-l-ewa-ta. do-REAL-1PL-AFFIRM ‘When that fellow comes, we will go with him.’ Digul Wambon (Jang 2008:82)

(15) No kup pa-gen 1SG COMIT sit-REAL[NON1SG] ‘He was with me.’ Yonggom Wambon (Drabbe 1959:120)

(16) Ene taget köp. 3SG shell.money COMIT ‘He is with shell money (he has money).’ Mandobo (Drabbe 1959:55)

(17) Nani-xu bo-xu-g-i. my.mother-COMIT DUR-ill.be-REAL-NON1SG ‘My mother is also ill.’ Kombai (de Vries 1993:46)

Stassen notes that a comitative marker can develop into a coordinator in verb-ﬁnal languages by doubling:

The creation of a coordinate NP structure involves the ‘doubling’ of the comitative marker, so that this marker becomes associated with both NP’s instead of just the second NP [...] The double occurrence of the comitative marker signals the equality in rank, and hence essentially coordinate status, of the syntactic relation between the two NP’s. (Stassen 2000:31-32)

This is clearly what happens in Awyu-Dumut languages as well: the coordinator kup originated as a comitative marker, and this comitative marker then doubled after 60 3. Nouns and Adjectives each NP. Schachter & Shopen(2007) also ﬁnd that ‘and’ and ‘with’ are expressed by the same word in many languages and that, historically, ‘and’ develops from ‘with’ (Schachter & Shopen 2007:46-47). The same is found by Mithun(1988:339-343) in an interesting chapter on the grammaticalization of coordination. She notes that an adverbial with the meaning ‘also’ or ‘with’ that develops into a coordinator will only coordinate nominals, not appearing with conjoined verbs or clauses (Mithun 1988:342). This seems to hold for Awyu-Dumut languages; kup does not occur as a coordinator after verbs or clauses.

3.6.2 Focus marker te as nominal coordinator

In addition to *-kup, Dumut languages have a focus element *te,12 which functions as a coordinator when it is repeated after each noun in a coordination. The deﬁnition of ‘focus marker’ used in this book is as follows:

Focus: the part of a proposition that makes the proposition into an assertion and that is emphasized by the speaker so the hearer may know it is informative.

This deﬁnition is based on Lambrecht’s and Halliday’s deﬁnitions of focus, which reads as follows: “Focus is the element of information whereby the presupposition and the assertion differ from each other. The focus is that portion of a proposition which cannot be taken for granted at the time of speech. It is the unpredictable or pragmatically non-recoverable element in an utterance. The focus is what makes an utterance into an assertion” (Lambrecht 1994:207) and “[i]nformation focus is one kind of emphasis, whereby the speaker marks out a part (which may be the whole) of a message block as that which he wishes to be interpreted as informative” (Halliday 1967:204f). The element te in Dumut languages, analyzed byde Vries & Wiersma(1992) as a focus marker and by Drabbe(1959) as an emphatic marker, is here analyzed as a focus marker that places emphasis on what precedes it, as illustrated by (18, 19) and (20). The background to this Mandobo example is that Ngou (the ‘he’ in the sentence) had seen that someone or something was eating nibung fruits; he had even seen tracks. In the last clause, it is revealed that it is a pig who was eating the fruit.

(18) Wemin ge-gen do, kinum reŋgi-ro ö Night be-REAL[NON1SG] CONN sleep lie-NON.CLOSE dream itigio-gen: yomorop u te mbe ŋ-gen. see-REAL[NON1SG] nibung.fruit pig FOC DUR eat-REAL.[NON1SG] ‘When it is night he sleeps and in a dream he sees: a pig is eating the nibung fruit.’ Mandobo (Drabbe 1959:15)

12Allomorphs nde, ndi in Digul Wambon, de, do in Mandobo and Yonggom Wambon. 3.6. Coordination of Nouns 61

In the Yonggom Wambon clause chain presented in (19), Kori first thinks he sees his cuscus and shoots it. However, his arrow goes straight through what he thought was the cuscus. He climbs the tree and finds out that his son Naerom had put bow scrapings there that looked like a cuscus. The bow scrapings provide us (and Kori) with the reason why his arrow flew straight through the ‘cuscus.’13

(19) Kori me etaga-t te yan amun e mbon dok, Kori come see-REAL[NON1SG] CONN14 his cuscus TOP be CAUS taemba-t te keroke te turu etaga-t shoot-REAL[NON1SG] CONN go.through CONN go.up see-REAL[NON1SG] te Naerom ŋga sinam mon de uru-r-an. CONN Naerom ERG bow scrapings FOC put-REAL[NON1SG]-PST ‘Kori came and saw that his cuscus was (there), therefore he shot it (but) when it (the arrow) went through he went up (into the tree) and saw that Nearom put bow scrapings.’ Yonggom Wambon (Drabbe 1959:147)

Besides marking the noun phrase that provides the ‘clue’, te can function as a cor- rective focus marker as well, as shown by Digul Wambon example (20).

(20) Ndu-n=e taxi-mo-knd-e. Woyo, laxai-nde sago-tr.nasal=TOP buy-do-REAL-NON1PL no ﬁsh-FOC taxi-mo-knd-e. buy-do-REAL-NON1PL ‘They buy sago. No, they buy ﬁsh.’ Digul Wambon (de Vries & Wiersma 1992:67)

When Dumut focus marker te is repeated after two or more subsequent nouns, it functions as a nominal coordinator, as is illustrated by (21, 22) and (23) from Man- dobo, Yonggom Wambon and Digul Wambon, respectively.

(21) Rumo te, roŋgu te mbetin de ö göp tima-gen two COORD female COORD male COORD he himself keep-REAL[NON1SG] ‘There are two, a female and a male, he keeps himself.’ Mandobo (Drabbe 1959:21)

(22) Kagup te, rarumun de, munotit te i-ŋgin-in mbon. man COORD woman COORD children COORD eat-REAL-NON1PL DUR ‘Men and women and children are eating.’ Yonggom Wambon (Drabbe 1959:145)

(23) oi=nde aŋgai=nde pig=COORD dog=COORD ‘a pig and a dog’ Digul Wambon (de Vries 1986:16)

13The further background is that Naerom had come earlier and shot and eaten his father’s cuscus. Later in the story, Kori has Naerom shot and eaten. 14The focus marker te also occurs after verbal clauses and then functions as a connective, as is further explained in Chapter 11 on Clause Linkage. 62 3. Nouns and Adjectives

The development of Dumut focus marker te into a coordinator can be explained by the fact that coordinated nouns, like focus markers, are salient in Digul Wambon discourse, as well as in other Papuan languages. Reesink(1987:177) notes for Usan that “coordination of terms is unusual.15 In those instances where coordination does occur, it concerns terms that are more salient than others.” Thus coordinators and focus markers only occur when salient information is given, and this similarity allowed Dumut focus marker te to start functioning as a nominal coordinator. How- ever, there are two issues that make the analysis of te as a nominal coordinator problematic. First, the data supporting the analysis of te as a coordinating conjunction is very slim; there is only one example of te functioning as a nominal coordinator from each Dumut language, and these examples were given in (21-23). Secondly, it is unclear how a nominal coordinator te would differ from the nominal coordinator kup. In other words, Awyu-Dumut languages did not have a need to develop te into a nominal coordinator, because they already had one, which they used frequently. Hence the analysis of te as a nominal coordinator should be considered hypothetical, not proven. Both Drabbe and de Vries ascribe a third function (besides focus marker and (coordinating) connective) to Dumut te, namely that of a copula or a predicator. However, this ‘copular function’ can be seen as an extension of the focus meaning of te; the focus marker te can have a copular reading but should not be analyzed or glossed as a copula. In expressions of the type ‘X is Y’, focus marker te occurs optionally, as illustrated by Jang(2008) for Digul Wambon (24).

(24) Ewo enow-e hiwin=ndi). That tree-TOP big(=FOC) ‘That tree is big.’ Digul Wambon (Jang 2008:64)

This ends the discussion of focus marker te occurring after nouns. In Chapter 11, the occurrence of focus marker te after verbal clauses will be discussed; it will be shown that te functions as a subordinating conjunction after verbal clauses.

3.6.3 Connective =o as nominal coordinator In Digul Wambon, Yonggom Wambon and Kombai, a connective =o is found that functions as a coordinator if repeated after each element of the coordination. For Digul Wambon, de Vries & Wiersma(1992) note that =o is a non-exhaustive enumer- ator that may mean either ‘and’ or ‘or’ (de Vries & Wiersma 1992:72). However, only examples where =o has the meaning ‘and’ are found in Digul Wambon (25-27).

(25) aŋgay=o kav=o dog=COORD man=COORD ‘the dog and the man’ Digul Wambon (de Vries & Wiersma 1992:72)

15The term ‘term’ is here used in the Functional Grammar sense, as an element that is used to refer to things, rather than to predicate. 3.6. Coordination of Nouns 63

(26) Koma-t-po sanip=o oi=o die-REAL[NON1SG]-PST cassowary=COORD pig=COORD ande-l-e-mbo. eat-REAL-NON1PL-PST ‘She died and cassowaries and pigs ate (her).’ Digul Wambon (Jang 2008:131)

(27) Na-mbap=o na-yop=o nuk=o noxop ilumtaxemo 1SG.POSS-father=COORD my-mother=COORD 1SG=COORD 1PL three ka-t-ewa-mbo. go-REAL-1PL-PST ‘My father, my mother and I, we three went.’ Digul Wambon (Jang 2008:108)

The Yonggom Wambon nominal coordinator o can have a conjunctive ‘and’ (28) or a disjunctive ‘or’ reading (29) and is repeated after each noun in the enumeration (Drabbe 1959:145).16

(28) Taemba-r-in de nan ip tamburum rogo-na-tir o Shoot-REAL-NON1PL CONN 1SG.POSS name just say-IMP-PROH CONN nde-t; mun mberon o, ran mbari o, kagup say-REAL[NON1SG] child little COORD women grown COORD men pari o mimir atik kaendi-y-iw-e o nde-t. grown COORD back bite break-LIG-1SG[FUT] CONN say-REAL[NON1SG] ‘They shoot him and he says: “you may not just speak my name, (because then) I will break with my teeth the back of small children, grown women and grown men” he says.’ Yonggom Wambon (Drabbe 1959:149)

(29) Ui o itir o ragae o ra mando-nan-in=e pig COORD cassowary COORD ﬁsh COORD take come-NON1PL-FUT=TOP tagimo-y-ip. buy-LIG-1SG[FUT] ‘If they bring pig or cassowary or ﬁsh, I shall buy it.’ Yonggom Wambon (Drabbe 1959:145)

The Kombai nominal coordinator -o can mean either ‘and’ or ‘or’ (de Vries 1993:50). Kombai -o optionally harmonizes with the first vowel of the noun that follows -o. De Vries(1993:12) notes that when the noun has a front vowel, -o may appear as /e/, when the noun has a back vowel in the first syllable, -o does not change, and when the noun has a central vowel in the first syllable, -o may harmonize to /a/.

16In Yonggom Wambon, o also occurs as a connective between a verb of speaking and that which was said, as happens twice in (28). In Digul Wambon, =o is a coordinating connective that occurs on nearly all medial verbs. 64 3. Nouns and Adjectives

(30) Ay-o duwoy-o rül-o el-o nage foro pig-COORD ﬁsh-COORD banana-COORD bird-COORD you bring ma-non-a nu ade. come[REAL]-NON1PL-CONN 1SG eat[IRR.1SG] ‘You have to bring pig, ﬁsh, bananas and/or birds in order that I (may) eat.’ Kombai (de Vries 1993:50)

3.7 Summary

Although Awyu-Dumut languages have little nominal morphology, some proto morphology can be reconstructed. In Section 3.1, it was pointed out that all Awyu- Dumut languages have adjectives that follow the noun, while some Awyu-Dumut languages make use of intensiﬁers or adjectival markers. Three ways of forming compound nouns that Awyu-Dumut languages share were described in Section 3.2. Section 3.3 pointed out that juxtaposition is the most common way to mark possession in Awyu-Dumut languages, although all Awyu-Dumut languages also use possessive pronouns, which will be reconstructed in Chapter4. The plurality of nouns was then discussed in Section 3.4, where it was found that reduplication is a pluralization strategy shared by all Awyu-Dumut languages. Kinship terms in Awyu-Dumut languages have a separate plural marker, which was reconstructed as *-gi for Proto Awyu, *-ŋgu(i) for Proto Dumut and *-ŋgV for Proto Awyu-Dumut. Concerning kinship terms, it was also noted in Section 3.5 that they tend to occur with a possessive element n-, and that several Awyu-Dumut languages have address kinship terms as well as reference kinship terms. This chapter concluded with a discussion of the coordination of nouns, and two nominal coordinators were reconstructed: *kup for Proto Dumut and Proto Awyu-Dumut (*ku for Proto Awyu), and *te for Proto Dumut. The nominal coordinator kup originates as a comitative marker, while nominal coordinator te is a grammaticalized focus marker. Awyu- Dumut languages have polysyndetic nominal coordination; the coordinating conjunction is repeated after each element of the coordination. 4

Awyu-Dumut Pronouns

Pronouns have fulfilled a central role in historical Papuan linguistics as conducted since the 1960’s. Wurm(1975:191-217) noted the existence of a certain set of pronouns as a characteristic of Trans New Guinea (TNG) languages. However, the classification methods employed by Wurm and his colleagues, as well as their proposal of a Trans New Guinea family, has received much critique, which can be summarized in a quote from Foley(1986:13): “At this stage, these [groups proposed by Wurm] remain hypotheses and conjectures, because detailed comparative work according to the rigorous methods of comparative linguistics ... has yet to prove these links.” Ross’ publication of a careful reconstructive study of Papuan pronouns in 2005 gave a new impetus to Papuan historical linguistics (Ross 2005). Ross was able to use pronouns as a primary, albeit preliminary, diagnostic in classifying Papuan languages, resulting in a revitalization of the Trans New Guinea hypothesis. The detailed, bottom-up reconstructions of Awyu-Dumut personal pronouns presented in this chapter meet all criteria for ‘a rigorous application of the comparative method’ and confirm that Awyu-Dumut languages belong to the larger Trans New Guinea family, based on the TNG pronouns reconstructed by Ross(2005). Section 4.1 presents Awyu-Dumut personal pronouns and their reconstructions, taking into consideration all Awyu-Dumut and Greater Awyu languages for which pronominal data are available. The possessive pronouns discussed in Section 4.2 shed further light on the diachronic development of Awyu-Dumut pronouns. The chapter ends with a short discussion on the use of pronouns in Awyu-Dumut languages, introducing frequently used Awyu-Dumut emphatic pronouns.

4.1 Reconstruction of Awyu-Dumut Personal Pronouns

Awyu, Dumut, Ndeiram and Becking-Dawi languages all have six personal pronouns, making a three-way distinction in person and a two-way distinction in num- 66 4. Awyu-Dumut Pronouns ber.1 The pronouns found in each language, as well as the proto forms of all six personal pronouns for Proto Awyu, Proto Dumut and Proto Awyu-Dumut reconstructed in the rest of this chapter are given in Table 4.1. A ﬁrst glance at Table 4.1 reveals many similarities between the pronouns of different branches of the Greater Awyu language family which, to refresh the reader’s memory, can be schematically represented as in Figure 4.1.

Table 4.1: Greater Awyu and Awyu-Dumut pronouns

1SG 2SG 3SG 1PL 2PL 3PL Dumut MAN nöp ŋgöp ege noŋgüp neŋgip yeŋgip YWB nup ŋgup yup naŋgup gaŋgup yaŋgup DWB nuk ŋgup nexep noxop ŋgoxop yaxop Awyu SHI no go ewe/ege noxo goxo yoxo YEN nu gu ewi/egi nugu gugu yoxo PSA nu gu eki nugu gugu yoxo AXU nu gu efe/eke nügu gügu yoxo Ndeiram KOM nu(f) gu(f) xe nagu nage ya WNG nuP ŋguP - naŋgu naŋgiP - Becking- Dawi KOR nup gup yup noxup gexenép yexenép TSA nu gu yu nau/nahu gau yau KYD nonu gonu ye nau gagu -

PD *nup *ŋgup *yup, *eke *nakup *ŋgakup, *yakup *nakip PA *nu *gu *eke *nüku *güku *yaku PAD *nup *ŋgup *yup, *eke *nakup *ŋgakup, *yakup *nakip

Sources: Mandobo, Yonggom Wambon (Drabbe 1959); Digul Wambon (de Vries & Wiersma 1992); Shiaxa, Pisa, Yenimu (Drabbe 1950); Aghu (Drabbe 1957); Kombai (de Vries 1993); Wanggom (Hughes 2009); Korowai (van Enk & de Vries 1997); Tsaukwambo (Baas 1981); Komyandaret (Hughes 2009)

1Awyu-Dumut pronouns have a wide range of functions, appearing both in subject, object and indirect object positions. Genitive meaning is expressed using possessive pronouns, discussed in Section 4.2. 4.1. Reconstruction of Awyu-Dumut Personal Pronouns 67

Greater Awyu

Becking-Dawi Awyu-Dumut

KORTSAKYD

Dumut Awyu Ndeiram

MANYWBDWB SHIYENPSAAXU KOMWNGTYN

Figure 4.1: Greater Awyu language family

The following sound changes, as described in Chapter2 on Awyu-Dumut Proto Phonology, are relevant when reconstructing pronouns. They correctly predict, among other things, that Proto Awyu pronouns do not have prenasalized stops or ﬁnal consonants, while Proto Dumut pronouns do.

1. PAD ﬁnal *p dropped in PA, underlying /f/ in KOM2 2. PAD medial *k before low vowels >/x/ in PA and KOM 3. PD medial *k >/x/ in DWB 4. PD medial *k >/g/ in MAN and YWB 5. PAD *mb, *nd and *ŋg denasalized in PA 6. PD *u >Mandobo /ö/ 7. PD *a >Mandobo /e/ before /i/ or /e/ 8. PD *a >Mandobo /o/ before /a/, /o/ or /u/ 9. PA */ü/ >/u/ in all Awyu languages except Aghu 10. PA *u >Shiaxa /o/ 11. PA *e >/i/ in SHI, YEN, PSA (not always)

In the rest of this chapter, each personal pronoun will be reconstructed for Proto Awyu, Proto Dumut, and Proto Awyu-Dumut. The pronouns from the Becking- Dawi languages, although not descended from Proto Awyu-Dumut, will prove to be useful in determining which form to reconstruct for Proto Awyu-Dumut. Pronouns will not be reconstructed for Greater Awyu because no proto phonology and regular sound correspondences have yet been established for Greater Awyu.

2Kombai is here treated as a direct descendent of Proto Awyu-Dumut, whereas Proto Ndeiram is a node between Kombai and Proto Awyu-Dumut. However, Proto Ndeiram cannot be reconstructed due to paucity of data, and hence Kombai is for the time being used as a proxy for Proto Ndeiram. 68 4. Awyu-Dumut Pronouns

4.1.1 First person singular

MANYWBDWBAXUPSASHIYENKOMWNGKORTSAKYD 1SG nöp nup nuk nu nu no nu nu(f) nuP nup nu nonu

The 1SG pronoun is reconstructed as *nu in Proto Awyu and *nup in Proto Dumut and Proto Awyu-Dumut. In Proto Awyu, the Proto Aywu-Dumut final consonant /p/ was dropped. The final /p/ has also disappeared in the Becking-Dawi languages of Tsaukwambo and Komyandaret. The final consonant /p/ is underlying in Kombai (sound rule1), where it is realized as /f/ whenever the pronoun is followed by a vowel-initial morpheme, for example the focus marker -a: nuf-a ‘I- FOC’(de Vries 1993). The Digul Wambon first person pronoun does not end in /p/ but in /k/. Drabbe does note that final /p/ is often not realized in Dumut speech (Drabbe 1959:6); Digul Wambon could have dropped the /p/ and replaced it with /k/, though it is unclear where the /k/ would have come from. The reconstructed vowel /u/ is straightforward; /u/ occurs in all languages except Shiaxa and Man- dobo. The different vowels in Shiaxa and Mandobo can be explained by regular sound changes (6) and (10) as found in Section 4.1; Proto Dumut *u always changes to /ö/ in Mandobo, while Proto Awyu *u always changes to /o/ in Shiaxa.

4.1.2 Second person singular

MANYWBDWBAXUPSASHIYENKOMWNGKORTSAKYD 2SG ŋgöp ŋgup ŋgup gu gu go gu gu(f) ŋguP gup gu gonu

The Proto Awyu-Dumut 2SG form is *ŋgup, as it is in Proto Dumut, while in Proto Awyu the 2SG pronoun is *gu. In Shiaxa and Mandobo, we ﬁnd the same vowel change as in the 1SG pronouns, namely *u to /o/ and /ö/ respectively. The Proto Awyu-Dumut and Proto Dumut 2SG forms have an initial prenasalized stop /ŋg/ while Proto Awyu has /g/, due to prenasalized stops having become regular stops in Proto Awyu (sound change5). Regarding the 2 SG form, it is interesting to note Drabbe’s statement that in Mandobo, ŋgöp is also sometimes found as a 3SG pronoun, though it clearly is the 2SG pronoun. Drabbe ascribes this to contamination from verbal person-number marking, where no distinction is made between second and third person (Drabbe 1959:10).

4.1.3 Third person singular

MANYWBDWBAXUPSASHIYENKOMWNGKORTSAKYD 3SG ege yup nexep efe/eke eki ewe/ege ewi/egi xe - yup yu ye 4.1. Reconstruction of Awyu-Dumut Personal Pronouns 69

The 3SG pronoun is the most diverse of all Awyu-Dumut pronouns across multiple languages and hence complicated to reconstruct. The first thing to note is that the Becking-Dawi languages and Yonggom Wambon share the form yup, reflecting a shared retention from a proto-form *yup, which can be reconstructed for Proto Awyu-Dumut and which can be related to Proto Trans New Guinea *[y]a/*ua. The other Awyu-Dumut 3SG pronouns all reflect a common form eke or ege, while three of the four Awyu languages also have a form ewe or efe in common. These forms could well have originated in an Awyu-Dumut deictic element ep or e ‘there’. When discussing Papuan third person pronouns, Ross notes that “it is common to find that third person forms are actually members of demonstrative sets, and this often means that they are not comparable and that third person reconstructions cannot be made” (Ross forthcoming:5) and also that “the reconstruction of third person forms is less easy because they are readily replaced by demonstratives” (Ross forthcoming:3). The 3SG pronoun forms with /k/, /g/ or /x/ are a combination of the deictic element e(p) or ewe (‘there’), found in all Awyu-Dumut languages, and an element ke, a topic marker found in Awyu languages. For Aghu pronouns, Drabbe notes that the 3SG eke consists of the 3SG efe plus ke. The Aghu form efe never occurs by itself but can occur with ke or with te, which is an emphatic or focus marker found in both Awyu and Dumut languages (Drabbe 1957:6). Similarly, Shiaxa ewe only co- occurs with a focus marker te, while Shiaxa ege never combines with te but is better analyzed as ke fused with ewe. In the Pisa 3SG form eki, -ki is discernable, which is cognate to the topic marker ke found in Shiaxa and Aghu. Thus the Awyu 3SG pronouns that contain a /k/, /g/ or /x/ are a combination of e(p) and ke, while the forms with a /w/ or /f/ only co-occur with a focus marker te. As was noted by Ross, demonstrative-based 3SG pronouns are difficult to reconstruct. For Proto Awyu, it is clear that the 3SG pronoun was based on a deictic because only reflexes of such a pronoun are found in Awyu languages. Therefore for Proto Awyu, **eke3 is reconstructed; the alternative form is *ewe, but as it only occurs as a subject pronoun together with focus marker te and does not occur at all in Pisa, *eke is chosen as the more likely proto-form. The second vowel is reconstructed as an *e, since /i/ in Pisa, Shiaxa and Yenimu is a reflex of *e (sound change 11). What is clear is that Digul Wambon, Mandobo and Kombai did not independently and individually develop their 3SG forms nexep4, ege and xe after they split off from Proto Awyu-Dumut and Proto Dumut because no topic marker ke cognate to Awyu ke is found in these languages. It is unclear where the Mandobo, Digul Wambon and Kombai 3SG pronouns came from; one possibility is that Digul Wambon, Mandobo and Kombai copied the eke form from Awyu languages, but it is just as likely that in Proto Dumut and Proto Awyu-Dumut, yup was already being

3A double asterisk reﬂects the fact that there were multiple reconstructions to choose from but that one of them was chosen as more likely than the other(s). 4The initial /n/ and ﬁnal /p/ in the Digul Wambon 3SG form are thought to have been added to reach uniformity with the rest of the Digul Wambon pronoun paradigm. 70 4. Awyu-Dumut Pronouns replaced by a deictic-based pronoun. Therefore both *yup and *eke are reconstructed as competing 3SG pronouns in both Proto Awyu-Dumut and Proto Dumut.

4.1.4 First person plural

MANYWBDWBAXUPSASHIYENKOMWNGKORTSAKYD 1PLnoŋgüp naŋgup noxop nügu nugu noxo nugu nagu naŋgu noxup na(h)u nau

The first person plural pronoun is reconstructed as *nüku in Proto Awyu and *nakup in Proto Dumut and Proto Awyu-Dumut. The middle consonant is reconstructed as *k. For Dumut languages, regular sound changes (3) and (4) state that Proto Dumut medial *k turns into /(ŋ)g/ in Mandobo and Yonggom Wambon and into /x/ in Digul Wambon. For Awyu languages, it cannot be established with certainty that Proto Awyu medial *k changes to /g/ before high vowels, but nonetheless it is not unlikely that the medial /g/ in Awyu pronouns reflects an original *k that became voiced intervocalically, and hence the Proto Awyu form is reconstructed with a *k. In the case of Shiaxa, which has goxo as its 2PL form, the regular sound change Proto Awyu *k >/x/ before low vowels applies (sound change2). The /h/ in Tsau- kwambo is also a reflex of /x/ and ultimately of *k. De Vries(2012a:18) notes that where Korowai has a velar fricative /x/, Tsaukwambo has either a corresponding ‘h with friction’ or complete deletion. The first vowel of the Dumut 1PL pronoun is *a, while in Proto Awyu it is *ü. In Proto Dumut, *a is reconstructed because Mandobo /o/ reflects PD *a (sound change8), which means that two Dumut languages have a reflex of Proto Dumut *a in the 1PL form. Although Digul Wambon often has an /o/ where Mandobo and Yonggom Wambon have /a/, this correspondence is not regular. In Proto Awyu, the vowel is *ü as this is the vowel found in Aghu; in all other Awyu languages, Proto Awyu *ü always changes to /u/ (sound change9). In Kombai and Wanggom, the vowel in the 1PL pronominal form is /a/, and as Proto Dumut also has *a, the Proto Awyu-Dumut 1PL pronoun is reconstructed with *a rather than with *ü. The second vowel in the 1PL pronoun is *u in all three proto languages.

4.1.5 Second person plural

MANYWBDWBAXUPSASHIYENKOMWNGKORTSAKYD 2PLneŋgip gaŋgup ŋgoxop gügu gugu goxo gugu nage naŋgiP gexenép gau gagu

The Proto Awyu 2PL pronoun is güku, while for both Proto Dumut and Proto Awyu- Dumut, two forms are reconstructed, namely *nakip and *ŋgakup. The form with initial /n/, *nakip, is reconstructed to account for the 2PL pronouns found in Kom- bai, Wanggom and Mandobo, which all start with /n/. This initial /n/ is seen as a 4.1. Reconstruction of Awyu-Dumut Personal Pronouns 71 shared retention, rather than a shared innovation, between Mandobo and the two Ndeiram languages. At the same time, *ŋgakup is reconstructed in Proto Dumut to account for the Digul Wambon and Yonggom Wambon 2PL pronouns. A *ŋg-initial proto-form is also reconstructed for Proto Awyu-Dumut because a reﬂex of *ŋg is found in Proto Awyu *g as well as in the Proto Dumut *ŋgakup form. Furthermore, the three Becking-Dawi languages also have initial /g/. Note that Proto Trans New Guinea (see Table 4.2 below) also has two alternatives for 2PL: ŋgi and nja, which re- ﬂects a similar situation as the one found in Proto Awyu-Dumut and Proto Dumut.

4.1.6 Third person plural

MANYWBDWBAXUPSASHIYENKOMWNGKORTSAKYD 3PL yeŋgip yaŋgup yaxop yoxo yoxo yoxo yoxo ya - yexenép yau -

The third person plural pronoun is *yaku in Proto Awyu and *yakup in Proto Dumut and Proto Awyu-Dumut. Although PA *u to /o/ is only established as a regular sound change between Proto Awyu and Shiaxa, it is hypothesized to have taken place in all Awyu languages in the second syllable of the third person plural pronoun. This vowel change then led to the fricativization of the medial *k before an /o/ (sound change2). Note that although the Kombai 3 PL pronoun consists of a single syllable ya, the Proto Awyu-Dumut form is reconstructed as *yakup; Kombai is thought to have dropped the second syllable -kup, a reﬂex of which is found in all other Awyu-Dumut languages as well as in the Becking-Dawi languages.

4.1.7 Reconstructed Awyu-Dumut pronouns Table 4.2 contains all pronouns reconstructed in the above paragraphs and contrasts them with the Proto Trans New Guinea (PTNG) pronouns as reconstructed by Ross (2005:29).

Table 4.2: Awyu-Dumut and Trans New Guinea proto pronouns

1SG 2SG 3SG 1PL 2PL 3PL

PA *nu *gu **eke *nüku *güku *yaku PD *nup *ŋgup *yup, *eke *nakup *ŋgakup, *yakup *nakip PAD *nup *ŋgup *yup, *eke *nakup *ŋgakup, *yakup *nakip PTNG *na *ŋga *[y]a/*ua *ni/*nu *ŋgi/*nja *i

The ﬁrst difference between PTNG and Proto Awyu-Dumut pronouns is that the singular forms in Proto Awyu-Dumut all have /u/, while in PTNG /a/ occurs. There are two further differences between the Awyu-Dumut forms and the Trans 72 4. Awyu-Dumut Pronouns

New Guinea forms: all Proto Awyu-Dumut forms end in /p/ (Proto Awyu forms dropped this ﬁnal consonant again) and the plural forms are bisyllabic. The /p/ was clearly added; Drabbe notes that it is often not pronounced in Mandobo and Yonggom Wambon, indicating that /p/ is not always added to pronominal forms. A possible origin of this /p/ is the deictic element ep ‘there,’ also found as the basis for the Proto Awyu 3SG pronoun. The second syllable in Awyu-Dumut plural forms can be analyzed as a plural marker -ku added to the PTNG singular forms na, ŋga and ya. Lastly, a similarity between Proto Awyu-Dumut and Proto Trans New Guinea is that both proto languages have two forms for 2PL, one starting with /ŋg/, the other with /n/. This comparison shows on the basis of its pronouns that the Awyu-Dumut language family is part of the larger Trans New Guinea family; Awyu-Dumut pronouns can be traced to Proto Trans New Guinea pronouns.

4.2 Awyu-Dumut Possessive Pronouns

Besides personal pronouns, Awyu-Dumut languages also have possessive pronouns. The possessive pronouns occur before the possessed noun and are often found with kinship terms, which are inherently possessed. For Digul Wambon and Kombai (de Vries & Wiersma 1992; de Vries 1993) as well as for Korowai (van Enk & de Vries 1997), de Vries describes the possessive pronouns as prefixes, while Drabbe treats them as separate words. As possessive pronouns are often bound forms in Papuan languages, a reanalysis of Drabbe’s possessive pronouns as bound and not free is a more likely reanalysis than a reanalysis of de Vries’ possessive pronouns as free. Hence, all Awyu-Dumut possessive pronouns are analyzed as bound. Bound pronominal forms often represent an older diachronic stratum than free forms (Ross forthcoming:4), and thus Awyu-Dumut possessive pronouns might reflect historically older forms than Awyu-Dumut personal pronouns. The Awyu-Dumut possessive pronouns are presented in Table 4.3. Aghu and Kombai each have only two possessive pronouns, while all other languages have six possessive pronouns. Aghu and Kombai employ personal pronouns as possessives where they have no specific possessive pronouns. For Pisa, Shiaxa, Yenimu, Aghu and Mandobo, Drabbe notes that the possessive pronoun is followed by the possessive postposition na or ne. In Pisa and Mandobo, this na has become part of the possessive pronoun. For Shiaxa, Yenimu and Aghu, Drabbe notes that ne, shortened to n’, only occurs when the possessed noun begins with a vowel. It therefore makes more sense in these languages to analyze this n’ as a transitional nasal between vowels rather than as a possessive marker. Thus only in Pisa and Mandobo is the possessive marker ne or na found, as was also illustrated in Section 3.3 on possession.5 All other languages employ only possessive pronouns to mark possession.

5Note, however, that the -no in Kombai 3PL.POSS pronoun might also be a reﬂex of a possessive postposition na. 4.2. Awyu-Dumut Possessive Pronouns 73

Table 4.3: Awyu-Dumut possessive pronouns

1SG 2SG 3SG 1PL 2PL 3PL Dumut MAN ne-ne- ŋgo-ne- e-ne- noŋgü-ne- neŋgi-ne- yeŋgi-ne- YWB na- ŋgo- ya- naŋgo- ŋgaŋgo- yaŋgo- DWB na- ŋga- nexo- noxo- ŋgoxo- yaxo- Awyu SHI na- ga- wa- naxa- gaxa- yaxa- YEN na- ga- wa- niga- giga- yaxa- PSA na- ga- ena-, ewa- nuna- guna- yoxona- AXU na- ga- - - - - Ndeiram KOM na- - - - - yano- Becking- Dawi KOR nV- gV- yV- noxu- gexené- yexené-

PA *na- *ga- *wa- *nVka- *gVka- *yaxa- PD *na- *ŋga- *ya-, *e- *nakV- *ŋgakV- *yakV- PAD *na- *ŋga- *ya-, *wa- *nVkV- *ŋgVkV- *yaka-

Sources: Mandobo, Yonggom Wambon (Drabbe 1959); Digul Wambon (de Vries & Wiersma 1992); Shiaxa, Pisa, Yenimu (Drabbe 1950); Aghu (Drabbe 1957); Kombai (de Vries 1993); Korowai (van Enk & de Vries 1997)

The reconstruction of Proto Awyu, Proto Dumut and Proto Awyu-Dumut possessive pronouns is straightforward, although the vowels are uncertain, and yields reconstructions that can be compared with Proto Awyu-Dumut personal pronouns as well as with PTNG pronouns. The addition of the possessive marker ne/na in Mandobo and Pisa is hypothesized to have occurred after these languages split off from their mother languages and hence ne/na is not reconstructed in Proto Awyu, Proto Dumut or Proto Awyu-Dumut as part of the possessive pronoun. The reconstruction of the 3SG possessive pronoun is much like the reconstruction of the 3SG personal pronoun described in Section 4.1.3; Korowai and Yonggom Wambon reflect the older form *ya-, while Mandobo, Digul Wambon and Awyu languages reflect the demonstrative-based form *wa-. When the reconstructed Awyu-Dumut possessive pronouns are compared to Proto Trans New Guinea and Proto Awyu-Dumut personal pronouns, given in Table 4.4, they are clearly more similar to Proto Trans New Guinea personal pronouns than to Proto Awyu-Dumut personal pronouns. The Proto Awyu-Dumut singular possessive pronouns have a vowel /a/ also found in PTNG pronouns, while Proto Awyu-Dumut personal pronouns have /u/. The Proto Awyu-Dumut personal pronouns also have a final /p/ not present in the Proto Awyu-Dumut possessive pronouns or in PTNG pronouns. In other words, the personal pronouns underwent a sound change from /a/ to /u/ and added a /p/ to the end of the forms. Hence, possessive pronouns reflect older pronominal forms, 74 4. Awyu-Dumut Pronouns

Table 4.4: PAD possessive pronouns compared to PAD and PTNG personal pronouns

1SG 2SG 3SG 1PL 2PL 3PL

PADPOSS *na *ŋga *ya, *wa *nVkV *ŋgVkV *yaka PAD *nup *ŋgup **yup *nakup *ŋgakup, *yakup *nakip PTNG *na *ŋga *[y]a/*ua *ni/*nu *ŋgi/*nja *i and as such they form a clear link between PTNG and Proto Awyu-Dumut personal pronouns. This conclusion is in line with Ross’ observation that “in the majority of micro groups, bound and free forms are obviously cognate with each other, and one can infer that bound forms are the outcome of the clitization and afﬁxation of earlier free forms. Often, free forms have additional material sufﬁxed to them” (Ross 2005:26).

4.3 Awyu-Dumut Emphatic Pronouns

In many Papuan languages, the verb is the head of the clause and the only obligatory constituent, making the realization of other types of words infrequent. Nominals in the clause are optional modiﬁers of the verbal head. Often, there is no nominal at all in the clause, or at most one nominal, as speakers tend to distribute nominals across multiple clauses. Due to this pragmatic discourse preference for verb-only clauses, “it is not surprising that there is little room for nouns and anaphoric pronouns to track given, active referents. Most speakers of Papuan languages avoid nouns and anaphoric use of independent pronouns to an extreme extent in referent tracking, especially in the tracking of active or given subject and object referents” (de Vries 2006:813). Rather, referents are tracked through subject person number markers on the verb and by switch reference strategies. In the rare case that a pronoun is used to refer to the agent of an action, the pronoun is marked with pragmatic marker(s) te and/or ke. The pragmatic marker singles the pronoun out as being unusual and therefore worthy of the hearer’s attention. Another strategy besides using pragmatic markers to emphasize pronouns is found in Dumut languages, which can all add some form of -ot to the possessive pronouns to yield emphatic pronouns. Mandobo adds -anot to its possessive pronouns, resulting in emphatic pronouns nanot, ŋgoanot, eanot, noŋgünot, neŋginot and yeŋginot. In Yonggom Wambon, wot combines with possessive pronouns, yielding emphatic pronouns nawot, ŋgowot, yawot, naŋgowot, ŋgaŋgowot and yaŋgowot. In Digul Wambon, possessive pronouns function as emphatic pronouns when followed by -votke: novotke, ŋgovotke, nexovotke, noxovotke, ŋgoxovotke and yaxovotke. When the initial w, v and an are considered ligatures, as they are by Drabbe, and Digul Wambon ke is analyzed as the verb ke ‘to be’, a morpheme ot is left over. For Digul Wambon, Jang notes that “probably this [-otke] has been derived from ot ‘stomach/wish’, sufﬁxed by -ke. Often representing the very heart of someone in many Papuan languages, 4.4. Summary 75

‘stomach’ in Wambon may also signify the very self of the corresponding pronoun.” In Digul Wambon, Mandobo and Yonggom Wambon, as well as in Awyu languages, ot, or or o is indeed the word for ‘insides’, ‘intestines’ and ‘faeces’. Drabbe gives an alternative explanation of the origin of emphatic -ot in Mandobo: he states that it is the subject marker -o followed by the focus marker te, which has shortened to t (Drabbe 1959:10). Drabbe’s explanation seems the more likely one, as adding the pragmatic marker te is the most common way to form emphatic pronouns in Dumut languages. Also, if the word for ‘intestine’ was the source of -ot in emphatic pronouns, a similar form would be expected in Awyu languages. The subject marker -o only occurs in Dumut languages.

4.4 Summary

This chapter presented a bottom-up reconstruction of Awyu-Dumut personal and possessive pronouns. The bound possessive pronouns are older than the personal pronouns and provide the evidence that, in Awyu-Dumut personal pronouns, a ﬁ- nal /p/ was added. Furthermore, it was shown that the 3SG personal pronoun in the majority of Awyu-Dumut languages is derived from the demonstrative element ep ‘there’, while a more archaic form yup in Korowai and Yonggom Wambon reﬂects *ya, the 3SG form reconstructed for PTNG. Lastly, it was shown that personal pronouns are not frequently used in Awyu-Dumut languages and that, when they are used, they appear in their emphatic form.

Subject Person-Number Marking

In Awyu-Dumut languages, both the person and the number of the subject are marked by a suffix on the verb. The subject is often not expressed explicitly in Awyu-Dumut languages; one knows from the context and the person-number marking on the verb who or what the subject of a clause is. It is then all the more surprising that Awyu-Dumut languages make no distinction between second and third person but only a distinction between first person and non-first person. This way, the Mandobo clause

(31) ande-n-en eat[IRR.]-tr.nasal-FUT[non1.SG] could mean either ‘you (SG) will eat’, ‘she will eat’, ‘he will eat’ or ‘it will eat’. The correct meaning can only be determined from the context in which the clause appears. According to Foley, the conflation of second and third person in the non-singular is common in (highland) Papuan languages (Foley 1986:72). However, Cysouw (2003) has found that in many Papuan languages that were described as having 2/3 person homophony (a conflation of second and third person), a dual form also exists. The only Papuan languages where he found a complete 2/3 homophony and no dual form are Kombai, Digul Wambon and ‘Awyu’ (an old term for Awyu-Dumut languages excluding Digul Wambon), which are all Awyu-Dumut languages, and Moroari, a Trans-Fly language (Cysouw 2003:131-132). Cysouw found a complete 2/3 homophony only in two other languages worldwide besides these Papuan languages. Thus the conflation of second and third person in both the singular and the non-singular found in Awyu-Dumut languages is typologically rare. There are four different subject person-number markers in Awyu-Dumut languages, which make a two-way distinction in both person (first person versus non- first person) and number (singular versus plural). The four different person-number markers are: (a) first person singular, (b) non-first person singular, (c) first person plural and (d) non-first person plural. Subject markers occur in Awyu-Dumut semi- finite and finite verbs; semi-finite verbs consist of a verb stem, a mood marker and 78 5. Subject Person-Number Marking a subject marker, while finite verbs consist of a verb stem, a mood marker, a subject marker and a tense marker. In Dumut languages, subject person-number markers always appear at the end of the verb, after the mood and tense morphemes. In Awyu languages, the subject markers most often occur right after the mood marker but they occur before the tense marker if there is one. Furthermore, in different tenses and moods, Awyu-Dumut subject person-number markers vary slightly in form. The subject person-number markers found in each mood and tense will be presented for each language in this chapter, although only one subject marker will be reconstructed for multiple moods and tenses for Proto Awyu, Proto Dumut and Proto Awyu-Dumut. This chapter will present descriptions and reconstructions of 1SG (Section 5.1), NON1SG (Section 5.2), 1PL (Section 5.3) and NON1PL (Section 5.4) subject markers.

5.1 First Person Singular Subject Marker

The ﬁrst person singular subject person-number marker is -ep, or a variant thereof, in all Awyu-Dumut languages. It occurs in all verbal tense and mood forms, as illustrated in Table 5.1.

Table 5.1: Awyu-Dumut ﬁrst person singular subject markers

Irrealis Future Realis Dumut Awyu Awyu Awyu Awyu past hodier- hester- distant histori- nal nal cal

MAN -p -p -ep -op YWB -p -p -ep -ep DWB -p -ep -ep -ep

SHI -Ewe -we - -owe -ewe -ewe -we YEN - -fi -i -iki -fi -fi -efi PSA -∅ -∅ -i -i -ri AXU -∅ -∅ -e -e -e

KOM -f -f -e(f)

Sources: Mandobo (Drabbe 1959:10); Yonggom Wambon (Drabbe 1959:127-129); Digul Wambon (de Vries & Wiersma 1992:23, 31); Shiaxa, Pisa, Yenimu (Drabbe 1950:111-116; 123); Aghu (Drabbe 1957:10-14); Kombai (de Vries 1993:22-25)

The initial /e/ of -ep is dropped when it occurs after another vowel; the elision of /e/ is a common morphophonemic change in Awyu-Dumut languages. In verbs realized in realis mood or in past tense, the subject person-number marker follows a realis marker or a past tense marker that consists of or ends in a consonant, and /e/ is not dropped. In most Awyu-Dumut languages, irrealis and future verb stems end in a vowel, the subject person-number marker directly follows the verb stem in irrealis, and future forms and hence the initial /e/ of ep is not found in these verb realizations. Digul Wambon (32) and Shiaxa (33) form an exception to this rule: in Digul Wambon -ep follows the future marker -mat, while in Shiaxa, verb stems 5.1. First Person Singular Subject Marker 79 that end in a consonant prevent the initial /e/ of the 1SG subject marker from being dropped. (32) ande-p-mal-ep eat-1SG-FUT-1SG ‘I will eat’ Digul Wambon (de Vries & Wiersma 1992:27) (33) ader-Ewe hear.II-[IRR]1SG ‘I want to/let me hear’ Shiaxa (Drabbe 1950:123) What is perhaps more remarkable about the Shiaxa first person marker is its final vowel, rather than its initial vowel. All Shiaxa and most Yenimu realizations of the 1SG subject marker end in a vowel /e/ or /i/. It is not uncommon for Shiaxa to have an extra vowel that the other Awyu languages do not have; some examples are given in Table 5.2. The Shiaxa epenthetic vowel is hypothesized to be added to a word or morpheme in order to reach a CV.CV syllable structure; a CV.CV syllable structure is found in all Awyu languages. As noted in Chapter 2 on phonology, Proto Awyu had underlying consonants, like Kombai, which only surfaced when followed by a vowel. Some Shiaxa reflexes of Proto Awyu words retained these underlying consonants, allowing them to resurface in certain instances when an epenthetic vowel is added to them. Aghu, Pisa and Yenimu deleted the underlying Proto Awyu consonants. When Shiaxa has added an epenthetic vowel, the Proto Awyu form is reconstructed with an underlying consonant, indicated by placing parentheses around it, as in Table 5.2.

Table 5.2: Shiaxa epenthetic vowel examples

breadfruit ﬂower hair

SHI yuwato kete moxo YEN yowo ki mo PSA yawo ki rõ AXU xã ki mu PA *yowo(t) *ke(t) *mu(x)/*rõ

MAN rawot ket ron YWB rawot ket ron DWB - mutmut lon, muk PD *rawot *ket *ron/*muk

KOM aluwo xe lo

PAD *rawot/*yawot *ket *ron/*muk

The Proto Awyu 1SG form is reconstructed as *-e(fe),1 where the regular sound

1The 1SG subject marker in Shiaxa and Yenimu is much like the 3SG pronoun in these languages, which, as was shown in Chapter4 on pronouns, consist of the deictic element ep ‘there’ plus a vowel /e/. This similarity is striking enough to be noteworthy, although probably coincidental. 80 5. Subject Person-Number Marking change *e to /i/ occurred in Yenimu and Pisa. The consonant is reconstructed as *f.2 Yenimu has a 1SG form -iki in the hodiernal past, which is not a reflex of Proto Awyu *-e(fe) but whose origin is unknown. The Kombai language gives further evidence which supports the reconstruction of the Proto Awyu-Dumut 1SG subject marker as *-ep. Kombai has no past tense, only a realis mood. In the realis mood, the final /f/ (a fricativized reflex of /p/) of the 1SG subject marker is only realized when the verb is followed by a word or morpheme starting with a vowel. The following example from a Kombai story contains three occurrences of a 1SG subject marker in a semi-finite realis form, twice realized as -e at the end of the verb, and once realized as -ef in badefa when followed by the connective -a:

(34) Deyalu nu hutan fa-d-e. Xo ba-d-ef-a hari Jumat Recover 1SG jungle go-REAL-1SG go stay-REAL-1SG-CONN day Friday x-en-a nu sakit-ma-d-e. be-NON1SG-CONN 1SG ill-do-REAL-1SG ‘I recovered and went to the jungle. I stayed there and when it was Friday, I became ill.’ Kombai (de Vries 1993:105)

In irrealis mood and future tense, the Kombai 1SG subject marker consists only of /f/ because the initial /e/ is elided after stem-ﬁnal vowels. There are no attested occurrences of 1SG -f in semi-ﬁnite irrealis verb forms, but there are examples of 1SG -f in future tense forms. In future tense forms, the person-number marker is always followed by the vowel-initial future tense marker -e, as illustrated by emonemalefe in (35).

(35) Mo-ro umo-ma-no lü menana marora mene thus.do-CONN tell-HAB-3PL word this and this umo-n-e-male-f-e. tell-tr.nasal-FUT-DUR-1SG-FUT ‘Thus they used to tell and this is what I shall be telling about.’ Kombai (de Vries 1993:114)

To conclude, the Proto Awyu-Dumut and Proto Dumut 1SG subject person-number marker is reconstructed as *-ep, which was often realized as *-p if it occurred after a vowel. In Proto Awyu, the ﬁnal *p had fricativized to *f when an epenthetic vowel followed to retain a CV.CV pattern. The reconstructed Proto Awyu form is thus *-e(fe).

5.2 Non-ﬁrst Person Singular Subject Marker

The non-ﬁrst person singular subject marker is often not realized overtly in Awyu- Dumut languages, as expressed by -∅ in Table 5.3. However, sometimes it is expressed as either -e or -n. 5.2. Non-ﬁrst Person Singular Subject Marker 81

Table 5.3: Awyu-Dumut non-ﬁrst person singular subject markers

Irrealis Future Realis Dumut Awyu Awyu Awyu Awyu past hodier- hester- distant histori- nal nal cal

MAN -n -n -∅ -∅ YWB -n -n -∅ -∅ DWB -∅ -∅ -∅ SHI -en -ne -∅ -rea -ere -ere -re YEN -ni -∅ -i -∅ -i -a PSA nas. stem nas.stem -i -i -i AXU nas. stem -n -e -e -i -∅ KOM -n -n -∅, -n, -e

(a) The Shiaxa past forms of the NON1SG subject marker are different from those found in any other Awyu-Dumut language. They appear to consist of -e plus re, a demonstrative marker. This addition of -re is a Shiaxa innovation. Sources: Mandobo (Drabbe 1959:10); Yonggom Wambon (Drabbe 1959:127-129); Digul Wambon (de Vries & Wiersma 1992:23, 31); Shiaxa, Pisa, Yenimu (Drabbe 1950:111-116; 123); Aghu (Drabbe 1957:10-14); Kombai (de Vries 1993:22-25)

The reconstructed Proto Awyu-Dumut form *-en best explains both occurrences of -e and -n as NON1SG subject markers in Awyu-Dumut languages. Shiaxa irrealis verbs are the only verb forms in Awyu-Dumut languages that have an exact reflex of *-en; reflexes consisting of just -n can be explained by the phonological rule stating that the vowel /e/ elides when it follows another vowel, while reflexes of *-en consisting of just -e can be explained by the regular sound change stating that Aghu, Pisa and Yenimu may not have syllables that end in a consonant. Thus a reflex -n is often found in irrealis and future forms where the subject marker follows vowel-final verb stems that cause the elision of /e/, while -e3 or -en occurs in realis and past forms where the subject marker is preceded by a consonant- final mood or tense marker.4 In Dumut realis and past forms, as well as in some Kombai forms, the NON1SG subject marker is left off altogether. In Kombai realis forms, this happens when there is no mood marker5 and the verb is clause-final or is followed by a consonant-initial word or morpheme. In Kombai realis forms, -n occurs as the NON1SG subject marker when no realis mood marker -g occurs and when it is followed by a vowel-initial word or morpheme, as illustrated by xumoleina in (36).

(36) Xumolei-n-a ifama-no. die-NON1SG-CONN bury-NON1PL ‘He died and they buried him.’ Kombai (de Vries 1993:108)

De Vries(1993) analyzes the /n/ in xumoleina as a transitional nasal, but a compari-

3/e/ changes to /i/ in Yenimu and Pisa due to a regular sound change. 4As discussed above, Shiaxa has irrealis verb stems that do end in consonants, and hence -en, rather than -n, is found in Shiaxa irrealis forms. 5In Kombai, the realis mood marker -g only occurs after verb stems that end in a nasal vowel. 82 5. Subject Person-Number Marking son with other Awyu-Dumut languages makes it clear that it is a realization of the NON1SG marker. In Kombai realis forms where the realis mood marker -g occurs, the NON1SG subject person-number marker is realized as -e (37). (37) u-g-e.˜ kill-REAL-NON1SG ‘He killed’ Kombai (de Vries 1993:23) In Dumut, Awyu and Kombai irrealis and future forms, -n is the most common realization of the NON1SG subject marker. In ﬁnite future forms, the subject marker is followed by a future tense marker -en, -in, -e or -i in Mandobo, Yonggom Wambon, Aghu and Kombai, respectively, as illustrated in Table 5.4.

Table 5.4: Future NON1SG Forms

NON1SG future form Gloss Meaning

MAN ande-n-en eat-NON1SG-FUT he/she/it will eat YWB mayo-n-in come.down-NON1SG-FUT he/she/it will come down AXU atü-n-e go.up-NON1SG-FUT he/she/it will go up KOM ai-n-i go-NON1SG-FUT he/she/it will go

The Pisa future tense form ends in a nasal vowel (thus mariõ means ‘he/she/it will come down’); the -n of the subject marker merges with the verb stem because there is no future tense marker in Pisa. This results in a ﬁnal nasal vowel. In Shiaxa and Yenimu, -e (in Yenimu -i) is added to the subject person-number marker. However, this -e is not a future tense marker because it does not occur in other person-number forms. The origin of this -e (or -i) is unclear. Essential to the analysis of Awyu NON1SG subject markers is the analysis of -k, rather than -ke, as the Awyu realis marker used in non-ﬁrst person verb forms. For a detailed account of this mood marker, see Chapter 5 on Mood. Here, however, it must be explained why the Awyu realis mood marker is -k and not -ke, in other words, why -e is analyzed as belonging to the subject marker (Analysis II in Table 5.5) rather than to the mood marker (Analysis I in Table 5.5).

Table 5.5: Two analyses of realis mood in Awyu languages

Analysis I Analysis II

KOM u-ge˜ u-g-e˜ kill-REAL.NON1[SG] kill-REAL.NON1-NON1SG AXU da-xe da-x-e hear-REAL.NON1[SG] hear-REAL.NON1-NON1SG PSA ada-ki ada-k-i bind-REAL.NON1[SG] bind-REAL.NON1-NON1SG

There are two reasons why the realis mood marker is analyzed as not containing an /e/. First, /e/ also occurs as a part of the NON1SG subject marker in irrealis forms, 5.3. First Person Plural Subject Marker 83 that is, in Shiaxa irrealis forms, such as ader-en ‘he/she/it wants to hear’. Therefore, it must be posited that -e (or reﬂex -i) is always part of the subject person-number marker, rather than of a realis mood marker. Secondly, in some Awyu past tenses, -e (or reﬂex -i) occurs as (part of) the subject person-number marker following a tense marker rather than the mood marker -k, for example in the Yenimu hesternal past, given in (38).6 (38) oto-d-i rise-HEST-NON1SG ‘he/she/it rose yesterday’ Yenimu (Drabbe 1950:114)

Thus /e/ belongs to the NON1SG subject marker, rather than to the realis mood marker -k, although it must be noted that synchronically these two morphemes are best analyzed as fused morphemes. Hence, *-en is reconstructed as the Proto Awyu-Dumut, Proto Awyu and Proto Dumut NON1SG subject person-number marker, which due to morphophonemic rules of vowel elision and ﬁnal consonant deletion is thought to at times have been realized as *-e or *-n in the proto languages, just as in their daughter languages.

5.3 First Person Plural Subject Marker

The ﬁrst person plural marker in Awyu-Dumut languages consists of the 1SG form -ep combined with a plural marker -an, leading to the form -ewan or a reﬂex thereof in all Awyu-Dumut languages. The 1PL subject markers as found in various tenses and moods are given in Table 5.6.

Table 5.6: Awyu-Dumut ﬁrst person plural subject markers

Irrealis Future Realis Dumut Awyu Awyu Awyu Awyu past hodier- hester- distant histori- nal nal cal

MAN -won -won -ewon -un YWB -wan -wan -ewan -ewan DWB -va -eva -eva -eva

SHI -ewan -wa -a -owa -owa -owa -wa YEN - -fa -a -a -fa -fa -efa PSA -a -a -a -a -ya AXU -oã -oã -oã -oã -oã -oã

KOM -fo -fo -efo

For Proto Dumut, *-ewan is reconstructed. In the irrealis and future forms, the initial /e/ is elided, except in Digul Wambon future forms because there the 1PL subject

6A full Yenimu hesternal past paradigm can be found in Table 7.4 in chapter 7 on tense. 84 5. Subject Person-Number Marking person-number marker follows the Digul Wambon future tense marker -mat. In Digul Wambon, furthermore, the final /n/ is dropped. The Mandobo past tense 1PL form -un, not relatable to *-ewan, will be discussed below. The /w/ in *-ewan is the /p/ of 1SG -ep, which fricativizes between vowels and is written by de Vries in Digul Wambon as /v/. Furthermore, in Mandobo, due to regular sound change rules, PD *a has changed to /o/. The Proto Awyu 1PL form is reconstructed as *-efan. The final /n/ is dropped or realized as nasalization of the final vowel. The /f/ is written by Drabbe as /w/ in some daughter languages of Proto Awyu; in Aghu it has been dropped altogether. In Shiaxa and Aghu, the initial /e/ has changed to /o/; for Aghu, Drabbe notes that this vowel is short and that oã sounds like a diphthong. In Pisa, little is left of PA *-efan except the /a/ of the plural *-an; realizations consisting only of /a/ are also found in Shiaxa and Yenimu. The Mandobo 1PL marker -un, found only in Mandobo past tense verbs, cannot be related to *-ewan. However, it is not a Mandobo idiosyncracy, because -un is also found as a 1PL subject person-number marker in Korowai (van Enk & de Vries 1997:98). More specifically, -un only appears in the hodiernal past in Korowai, while in other Korowai verbs, -le and -te are the 1PL subject markers used. The fact that -un is found in remote branches of the Greater Awyu family means that it is most likely an older, archaic form.7 In Proto Awyu-Dumut and Proto Dumut, -un must have been a past tense marker existing alongside *-epan/*-ewan in order to still be realized in Mandobo past tense. Thus, *-un is reconstructed in both Proto Dumut and Proto Awyu-Dumut as the 1PL subject marker in past tense. In all Awyu-Dumut languages except Mandobo, *-epan, the newer 1PL subject person-number marker, has spread through the entire language system.

5.4 Non-ﬁrst Person Plural Subject Marker

Table 5.7 contains realizations of NON1PL subject person-number markers from all Awyu-Dumut languages in different tenses and moods. The Proto Awyu-Dumut NON1PL subject marker is reconstructed as *-enan and is a combination of the NON1SG subject marker *-en plus a plural marker -an, much like the 1PL subject marker reconstructed above is a combination of 1SG *-ep and the plural marker -an. It is clear that the Awyu languages have NON1PL reﬂexes of *-enan; the ﬁnal /n/ is dropped or realized as a nasal vowel, while the initial /e/ often elides. In Pisa, the initial /e/ has changed to /i/, as it has in the Pisa NON1SG subject form; in Yenimu, both /e/ and /i/ occur. In the realis forms of Shiaxa and Yenimu, the initial /e/ has been replaced by /o/, as also occurs in Shiaxa and Aghu 1PL subject markers. For Proto Awyu the NON1PL subject person-number marker is therefore reconstructed as *-enan.

7Note the similarity to the situation of the 3SG personal pronoun: yup was the older form found in Yonggom Wambon and Korowai, while in all other Awyu-Dumut languages, demonstrative-derived ewe or ege had replaced yup. 5.5. Summary 85

Table 5.7: Awyu-Dumut non-ﬁrst person plural subject markers

Irrealis Future Realis Dumut Awyu Awyu Awyu Awyu past hodier- hester- distant histori- nal nal cal

MAN -non -non -on -in YWB -nan -nan -in -in DWB -na -e -e

SHI -enan -na -ona -ena -ena -ena -na YEN -na -ona -ina -na -ena -ena PSA -nã -nã -inã -ina -ina AXU -enã -enã -enã -enã -ena -nã

KOM -no -no -eno

Kombai also has reflexes of *-enan; the final nasal has been dropped and the final /a/ has changed to /o/, a common sound change in Kombai. The same sound change occurs between PD and Mandobo, where the irrealis, future and realis NON1PL subject markers have /o/ rather than /a/. The Dumut languages have reflexes of *-enan in the irrealis and future verb forms, where the initial /e/ has elided. However, the Mandobo, Yonggom Wambon and Digul Wambon realis and past tense NON1PL markers are not clear reflexes of *-enan; at best they could be seen as shortened forms. Nonetheless, *-enan is reconstructed as the Proto Dumut NON1PL form, as reflexes are found in the Dumut languages, and *-enan is reconstructed for Proto Awyu and Proto Awyu-Dumut.

5.5 Summary

The four subject person number-markers reconstructed for Proto Awyu-Dumut are *-ep (1SG), *-en (NON1SG), *-epan (1PL) and *-enan. For Proto Awyu the same forms are reconstructed as *-e(fe), *-en, *-efan and *-enan, while the corresponding Proto Dumut forms are *-ep, *-n, *-ewan and *-enan. The four reconstructed subject markers occurred in all moods and tenses in the three proto languages, although Proto Dumut, like its daughter languages, did not overtly mark NON1SG subject. There is quite some variation in the surface forms of the subject markers due to morphophonemic changes, which are thought to have also occurred in the proto languages. The initial /e/ found in all four subject person- number forms elides when it follows a vowel, while in PA, the ﬁnal /n/ was often dropped or realized as nasalization on the vowel preceding it. Some forms found in the daughter languages of PA and PD did not express the reconstructed forms exactly. The most noteworthy of these is the Mandobo past tense 1PL subject marker *-un, which could not be related to Proto Dumut *-ewan. 86 5. Subject Person-Number Marking

This form -un was also found as a 1PL marker in Korowai’s past tense. Hence *-un was reconstructed as a 1PL form used in past tense in Proto Dumut and Proto Awyu-Dumut, and *-epan is thought to be a newer form that replaced *-un. Often subject person-number markers are derived from personal pronouns; they are personal pronouns that have cliticized to the verb. This is not the case in Awyu- Dumut subject person-number markers, which are quite distinct from the personal pronouns. One parallel that can be drawn is between the 1SG subject marker *-ep and the 3SG personal pronoun ewe found in multiple Awyu-Dumut languages. It could be said that both of these forms are newer forms based on the deictic element ep ‘there’. For the 3SG pronoun, an older form yup occurs and is reconstructed, and for the 1PL subject marker, an older form -un is found and reconstructed. A comparison with more Papuan languages might reveal whether indeed the 1SG subject person-number marker *-ep is a newer form only found in Awyu-Dumut languages. Awyu-Dumut subject person-number markers in most cases follow the mood marker and interact with it, as became apparent in the discussion in Section 5.2 on whether the mood marker found in Awyu languages should be analyzed as -k or -ke. The next chapter focuses on this and other mood markers. 6

Mood

Mood describes a speaker’s attitude towards a situation, while tense locates a particular situation in time. A main modal distinction is between realis and irrealis mood, which can be deﬁned as follows:

The realis portrays situations as actualized, as having occurred or actually happening, knowable through direct perception. The irrealis portrays situations as purely within the realm of thought, only knowable through imagination. (Mithun 1999:173)

The opposition between realis and irrealis mood1 is at the heart of the Awyu-Dumut verb system, which is mood-driven. With regard to mood-driven languages, Foley & Van Valin(1984) note that in these languages mood is a more basic category than tense and that

The likelihood of the proposition along the realis-irrealis dimension is ﬁrst established, and then the temporal orientation of this modiﬁed proposition is established with respect to the time of the speech act. (Foley & Van Valin 1984:216)

Thus, in mood-driven languages, a verb form is first specified for what mood it is in, and then for tense. This is certainly true in Awyu-Dumut languages, where mood is always expressed, while tense is optionally expressed. Mood and tense interact with each other; events that occur in the past will largely be categorized as realis, whereas events that occur in the future are typically in irrealis mood. Hence mood and tense are sometimes realized as one morphological category, though they can also be represented separately (Foley 1986:159). In Awyu- Dumut languages, mood and tense are expressed as two separate morphological categories. Awyu-Dumut finite verb forms contain a mood marker and a tense marker,

1The terms ‘mood’ (Palmer 2001), ‘mode’ (Payne 1998), ‘status’ (Foley & Van Valin 1984) and ‘modality’ are all used when the distinction between realis and irrealis is discussed. Following Palmer(2001), I will use the term ‘mood’ to refer to realis and irrealis, and the term ‘modality’ to refer to constructions such as the adhortative, the conditional and the imperative. 88 6. Mood while semi-finite verb forms only contain a mood marker. In Awyu languages and Kombai, a distinction in mood is also made in the verb stems of non-finite, semi- finite and finite verb forms, while Dumut languages do not have specific realis and irrealis verb stems. After discussing the role of mood in non-finite verb forms (Section 6.1), this chapter explores Awyu-Dumut irrealis mood (Section 6.2) before turning to realis mood (Section 6.3). Mood realizations in each Awyu-Dumut language will be discussed before mood morphemes for Proto Awyu, Proto Dumut and Proto Awyu-Dumut are reconstructed (Section 6.4).

6.1 Verb Stems and Mood

Realis and irrealis mood are distinguished on verb stems in Awyu languages and Kombai but not in Dumut languages. All Awyu-Dumut languages have primary and secondary verb stems. The secondary stems are derived from the primary stems, sometimes regularly, at other times irregularly. Some Awyu-Dumut verbs have just one stem. In glosses, primary stems will be marked with I, while secondary stems will be glossed with II. In Awyu languages, primary verb stems are only used in realis conditions, while Awyu secondary stems are restricted to appearing in irrealis forms. Thus Awyu primary and secondary stems express a distinction in mood; mood is marked on the verb stem. Kombai primary and secondary stems express the same difference in meaning: primary stems occur in realis forms, while Kombai secondary stems occur only in irrealis or future verb forms. There are some regular ways in which Awyu secondary stems are derived from primary stems. In Aghu, Pisa and Shiaxa, secondary stems can be formed by adding a prefix a-, a suffix -me (if the primary stem ends in a nasal vowel) or a suffix -fV to the primary verb stem. Table 6.1 contains examples of Awyu verbs that employ one of these three derivational strategies.2 Shiaxa and Pisa share further ways of deriving secondary stems, by adding the suffix -rV or -xo(y), as illustrated by the examples in Table 6.2. Besides these verbs where the derivation of secondary stems from primary stems is transparent and straightforward, many Awyu secondary stems are derived irregularly and can only be recognized as secondary stems by the context in which they appear. Many Dumut verbs have primary verb stems and derived secondary verb stems. For Mandobo, Drabbe notes that there are many verbs with only a primary stem, as well as quite a few verbs with a primary and a secondary stem (Drabbe 1959:12).

2Drabbe represents all Aghu secondary stems as ending in a nasal vowel and then claims that the stem is denasalized when these secondary stems are used in future and adhortative/optative verb forms, except in non-ﬁrst person singular verb forms. I reanalyze the secondary verbs as ending in a normal vowel, just like the secondary stems in the other Awyu languages. The nasalization that occurs in NON1SG verb forms reﬂects the realization of the NON1SG subject person-number marker -n. 6.1. Verb Stems and Mood 89

Table 6.1: Awyu primary and secondary stems

primary stem secondary stem meaning prefix a- AXU i- ai- to mention mi- ami- to drink si- asi- to dig ü- agü- to fell ku-˜ akume- to die oxo ki- oxo akime- to bathe da- ade- to hear xo- axi- to go PSA do- ado- to bake fo- afo- to marry su- atu- to rise SHI na- ade- to eat ri- ari- to call roxo- aroxo- to say suffix -me AXU ku-˜ kume- to put in ifi-˜ ifime- to bind PSA tu-˜ atumo- to beat sago SHI tigi- tigimo- to build suffix -fV AXU ada- adfe- to bind agu- agufe- to seek da- dafi- to come PSA de- defi- to come ra- rafi- to take agu- agufo- to search SHI re- refe- to take

Sources: Aghu (Drabbe 1957:12-14); Pisa (Drabbe 1950:116-117); Shiaxa (Drabbe 1950:119-121)

If a Mandobo verb has a secondary stem, the primary stem is only used in semi- finite realis -ken forms, while the secondary stem is used in all other verb forms. In Yonggom Wambon, as in Mandobo, not every verb has a secondary stem; Drabbe notes that only verbs whose primary stem ends in a consonant tend to have a secondary stem. Yonggom Wambon secondary stems can be used in finite past tense forms, semi-finite realis -t forms, semi-finite irrealis forms, finite future forms and non-finite SS forms. As in Mandobo, the Yonggom Wambon primary stem is then largely used with semi-finite realis -ken forms (Drabbe 1959:124-131). Whereas Mandobo and Yonggom Wambon verbs have a primary stem and sometimes one secondary stem, most Digul Wambon verbs have past and future secondary stems in addition to a primary stem. The past secondary stem is often formed by adding -a to the primary stem and is used for finite past tense forms and semi- finite realis -t forms. Future secondary stems are often formed by adding -o to the 90 6. Mood

Table 6.2: Additional Pisa and Shiaxa verb stems

primary stem secondary stem meaning sufﬁx -xo PSA doto- adotoxoy- to get up edo- edoxo- to give ogo- ogoxoy- to go away ko- koxoy- to go towards river ro- roxo- to say efeto- efetoxo- to see, to know SHI ku- akoxo- to dig dato- datoxo- to get up ede- edoxo- to give axo- axoxo- to go away oku- okoxo- to go towards river fete- fetoxo- to see bo- baxa- to sit sufﬁx -rV PSA gi- giro- to become xaxa- xaxaro- to swell ni- diro- to say ri- masero- to stand SHI da- adere- to hear fe- fero- to place wiE- wuru- to hit

Sources: Pisa (Drabbe 1950:117), Shiaxa (Drabbe 1950:119-121) primary stem and are used in finite future forms and semi-finite irrealis forms. The primary stem is then used with semi-finite realis -ke(nd) forms, as in Mandobo and Yonggom Wambon. Digul Wambon verbs do not always have two secondary stems; some verbs have one secondary stem that is then used for all verb forms except semi-finite realis -ke(nd) forms. For example, the verb en- ‘to eat’ has a secondary stem ande- that can be used in both finite past tense forms and finite future tense forms. Thus in Dumut languages a distinction in mood is not marked by different verb stems, although it is likely that, at one point in time, Dumut verb stems did mark this distinction: Dumut primary stems are still never used with irrealis verb forms and hence could be analyzed as realis verb stems. However, the secondary verb stem occurs with either realis or irrealis meaning, so no specific irrealis verb stem exists. Proto Awyu-Dumut did mark a distinction in mood on its verb stems, a distinction that was carried on in Proto Awyu and Kombai but became blurred in Proto Du- mut. As not all Dumut verbs have secondary stems, and because they do not mark a distinction in mood, Dumut verb stems will not be glossed with I or II. Awyu verbs stems, where the distinction between primary and secondary verbs carries meaning, 6.2. Irrealis Mood 91 will be glossed with I or II.

6.2 Irrealis Mood

Irrealis verb forms express wishes, doubts, questions and events that might happen in the future. Awyu-Dumut languages have two types of irrealis verb forms: (1) semi-finite irrealis forms consisting of a (secondary) stem and a person-number marker that express adhortative, optative or intentional meaning, and (2) finite future tense forms encoding actions that will occur in the future. This section focuses on the semi-finite irrealis forms, while the finite future forms will be discussed in Chapter7 on Tense. All Awyu-Dumut languages have a semi-finite irrealis paradigm, called the ‘zero form paradigm’ by Drabbe(1950, 1957, 1959) because they consist of only a verb stem, a subject person-number marker and no further morphological markers. Ir- realis mood is not expressed overtly by a separate morpheme in Awyu-Dumut languages. Thus a typical semi-finite irrealis form is wagaemo-p ‘do.good-1SG’(Drabbe 1959:128, Yonggom Wambon) which can mean ‘I want to do good’ (intentional), ‘let me do good’ (adhortative), or ‘that I may do good’ (optative). Table 6.3 contains a semi-finite irrealis paradigm for all Awyu-Dumut languages except Yenimu.3 Digul Wambon and Kombai only have first person semi-finite irrealis forms. De Vries & Wiersma(1992:31) note that, in Digul Wambon, NON1 forms were elicited but were never used outside elicitation; they do not give the elicited forms. For Kombai, de Vries(1993:25) states that finite NON1 future tense forms are also used with intentional meaning. Thus Awyu-Dumut semi-finite irrealis forms consist of a verb stem (in Awyu languages and Kombai a secondary irrealis verb stem), a non-overt irrealis mood marker and a subject person-number suffix. The non-overt irrealis mood is indicated in the glosses as IRR between square brackets before the person-number marker. The example given in (39) contains three Mandobo semi-finite irrealis forms.

(39) Orat ke-n do kotoma-on do light become-[IRR]NON1SG CONN pull-[IRR]1PL CONN moto-n. come.out-[IRR]NON1SG ‘When it has become morning, let us pull it (the canoe) and it will come out.’ Mandobo (Drabbe 1959:37)

Example (40) illustrates the use of a semi-ﬁnite irrealis form in Aghu.

3When describing the optative mood for Awyu languages, Drabbe(1950) does not mention Ye- nimu at all, so nothing can be said about a possible semi-ﬁnite irrealis paradigm in Yenimu. Drabbe does note that Pisa does not have a speciﬁc optative modality; that is, the same forms are used to express optative, intentional or adhortative meaning as are used to express future tense. However, it would be speculative to state that the same is true of the future tense paradigm found in Yenimu. 92 6. Mood

Table 6.3: Awyu-Dumut semi-ﬁnite irrealis forms

1SGNON1SG 1PLNON1PL

AXU ade ade˜ ad-oã ad-enã hearII[IRR.1SG] hearII[IRR].NON1SG hearII-[IRR]1PL hearII- [IRR]NON1PL PSA adi ad˜ı adi-a adi-nã hearII[IRR.1SG] hearII[IRR].NON1SG hearII-[IRR]1PL hearII- [IRR]NON1PL SHI ader-ewe ader-en ader-ewan ader-enan hearII-[IRR]1SG hearII- hearII-[IRR]1PL hearII- [IRR]NON1SG [IRR]NON1PL

MAN ande-p ande-n ande-won ande-non eat-[IRR]1SG eat-[IRR]NON1SG eat-[IRR]1PL eat-[IRR]NON1PL YWB tami-p tami-n tami-wan tami-nan make.canoe- make.canoe- make.canoe- make.canoe- [IRR]1SG [IRR]NON1SG [IRR]1PL [IRR]NON1PL DWB ka-p ka-va go-[IRR]1SG go-[IRR]1PL

KOM ade-(f) ade-fo hearII-[IRR]1SG hearII-[IRR]1PL

Sources: Shiaxa (Drabbe 1950:58); Pisa (Drabbe 1950:118); Aghu (Drabbe 1957:14); Mandobo (Drabbe 1959:23); Yonggom Wambon (Drabbe 1959:128); Digul Wambon (de Vries & Wiersma 1992:31); Kombai (de Vries 1993:25)

(40) Xati bagidi, xati mi-di-k xo-x-enã. Dü ni again next.day again come.down-SS-CONN go-REAL-NON1PL sago for kiam-enã xo-x-enã. search.II-[IRR]NON1PL go-REAL-NON1PL ‘Again the next day, again they descend (from the house) and go. In order to look for sago they go.’ Aghu (Drabbe 1957:63)

A common use of semi-ﬁnite irrealis forms in Digul Wambon is in quotative constructions, which express intention or volitionality. The intention is expressed as silent speech or thought directed by the agent to himself (de Vries & Wiersma 1992:32), as in the next two examples:

(41) Oi taximo-p ne-mbel=o ka-t-mbo. pig buy[IRR]-1SG say-SEQ=CONN go-REALNON1SG-PST. ‘I want to buy a pig’ he said and went (In order to/with the intention of buy(ing) a pig he went).’ Digul Wambon (de Vries & Wiersma 1992:32) 6.3. Realis Mood 93

(42) Xitulop nombo-n=e nux=eve mba-mba-mo ka-p ne-mbel=o Thursday this-tr.nasal=TOP I=TOP hunt-hunt-do go-1SG say-SEQ=CONN mba-mba-mo ka-l-ep-mbo. hunt-hunt-do go-REAL-1SG-PST ‘That Thursday ‘I want to hunt’ I said, and I did so (that Thursday I wanted to go hunting)’ Digul Wambon (de Vries & Wiersma 1992:92)

To summarize, all Awyu-Dumut languages have semi-finite irrealis forms that express adhortative, intentional or optative meaning and that consist of a verb stem (a secondary verb stem in Awyu languages and Kombai) and a person-number marker. Proto Awyu, Proto Dumut and Proto Awyu-Dumut all had semi-finite irrealis forms consisting of a verb stem and a person-number marker. In Proto Awyu and Proto Awyu-Dumut, specific irrealis secondary verb stems existed. The reconstruction of Awyu-Dumut semi-finite realis verb forms in the next section will prove to be less straightforward than the reconstruction of Awyu-Dumut semi-finite irrealis verb forms.

6.3 Realis Mood

6.3.1 Introducing realis markers in Awyu-Dumut through reanalysis In the Awyu-Dumut grammar descriptions written by Drabbe(1950, 1957, 1959), de Vries & Wiersma(1992) and de Vries(1993), there is no mention of realis mood markers. However, there are two verbal markers in Dumut languages as well as two verbal markers in the Awyu languages and in Kombai, as described by Drabbe and de Vries, that are better reanalyzed as realis markers. In the Dumut languages the reanalysis concerns the morphemes -t and -ken, which immediately follow the verb stem and which Drabbe and de Vries analyze as ‘neutral-present tense markers’, that is, as tense markers that can express both present and past meaning. De Vries & Wiersma(1992)also call -t in Digul Wambon a ‘non-future tense marker’. In the Awyu languages, the markers are -d and -k, analyzed by Drabbe as either part of the person-number marker (-d) or as a ligature between the verb stem and the person- number marker (-k). Kombai also has -d and -g (an allomorph of -k); de Vries(1993) analyzes -d as a non-future tense marker but seems to analyze Kombai -g as part of the NON1 subject person-number marker. An analysis of -t, -ken, -d and -k as realis markers is logical for a number of reasons. When Dumut -t is analyzed as either a present-neutral or a non-future tense marker, there are two Dumut past tense markers. A Dumut past tense consists of a verb stem, the morpheme -t, a person-number marker, and past tense marker -an (Yonggom Wambon, Mandobo) or -mbo (Digul Wambon). A way around this problem would be to state that -t...-an/-mbo is a discontinuous tense marker that expresses past tense. The co-existence of a past tense marker and a non-future or 94 6. Mood present-neutral tense marker in Dumut languages would be odd, and is an analysis best avoided by reanalyzing the ‘present-neutral tense marker’ -t as a realis marker. The introduction of a realis marker in Dumut languages leads to a symmetric verbal system: a past tense form corresponds to a future tense form, while a realis form corresponds to an irrealis form. The future and past forms are then fully finite, while the realis and irrealis forms are semi-finite because they do not express tense. A reanalysis of Awyu -d and -k yields a similar symmetric system in Awyu languages, although Awyu languages have multiple past tenses and not all Awyu languages have a future tense marker. For Awyu languages, there are additional reasons to reanalyze -d and -k as realis markers. Drabbe analyzes -d as part of the first person person-number markers. There are two problems with this analysis. First, if -d is analyzed as being part of the first person person-number marker, then when a tense marker occurs after -d, the person-number marker is discontinuous, as Drabbe notes for Aghu: “the tense marker -k we find in between the two parts of the person-number markers.”4 Second, -d only occurs in verbs with realis and past tense meaning, which, if it is part of the person-number marker, would mean that there are two sets of person- number markers. It makes sense to analyze the phoneme that constitutes the only difference between these two supposed sets of person-number markers as a marker of realis mood. The same argument applies to -k in non-first person verb forms; it only occurs in realis contexts and hence is better analyzed as a realis marker than as part of the subject person-number marker. For Aghu, Drabbe does not analyze -k as part of the person-number marker but as a ligature between the verb stem and the person-number marker. A ligature occurs to ease pronunciation, and the ligatures that occur elsewhere in Awyu-Dumut languages are n, y and w, but not k; k is not a prototypical transitional sound. An analysis of -k as a realis marker takes care of this odd ‘ligature’ explanation by Drabbe. Thus Awyu languages have two realis markers, one that occurs in first person forms and one that occurs in non-first person forms. Kombai has the same realis markers as the Awyu languages. Dumut languages have two realis markers, -t and -ken, which both occur throughout the paradigm, that is, with all person-number markers. One problem which remains after this reanalysis is that the two realis markers in the Dumut languages appear to have the same function. In the paragraphs on realis mood in each individual Dumut language below, it will be shown that -t and -ken do have differentiated functions.

6.3.2 Dumut realis forms Table 6.4 contains realis -t and -ken paradigms from Mandobo, Yonggom Wambon and Digul Wambon. The realis marker -t also occurs in past tense forms, which are discussed in Chapter7 on Tense.

4“De tijdwijzer k vinden we tussen de beide delen van de subjectwijzers” (Drabbe 1957:11). 6.3. Realis Mood 95

Table 6.4: Dumut realis paradigms

1SGNON1SG 1PLNON1PL

MAN -t mende-r-ep mende-t menda-r-ewon menda-r-on come-REAL-1SG come-REAL[NON1SG] come-REAL-1PL come-REAL-NON1PL -ken ti-gin-ep ti-gen ti-gin-iwon ti-gin-on build-REAL-1SG build-REAL[NON1SG] build-REAL-1PL build-REAL-NON1PL YWB -t etaga-r-ep etaga-t etaga-r-ewan etaga-r-in see-REAL-1SG see-REAL[NON1SG] see-REAL-1PL see-REAL-NON1PL -ken ti-gen-ep ti-gen ti-gen-ewan ti-gin-in build-REAL-1SG build-REAL[NON1SG] build-REAL-1PL build-REAL-NON1PL DWB -t aktu-l-ep aktu-t aktu-l-ewa aktu-l-e wrap-REAL-1SG wrap-REAL[NON1SG] wrap-REAL-1PL wrap-REAL-NON1PL -ken ndat-kend-ep ndat-ke ndat-kend-eva ndat-kend-e hear-REAL-1SG hear-REALNON1SG hear-REAL-1PL hear-REAL-NON1PL

Sources: Mandobo (Drabbe 1959:11, 32); Yonggom Wambon (Drabbe 1959:128-129); Digul Wambon (de Vries & Wiersma 1992:25); (Jang 2008:33)

Digul Wambon has several allomorphs for -ken: -kend, -knd, -ke and -xe. The -ke and -xe allomorphs only occur in NON1SG verb forms; -xe occurs when the verb stem ends in a vowel, -ke when it is consonant-final. De Vries & Wiersma(1992) and Jang (2008) describe the -kend and -knd forms as ending in an /e/ (thus as -kende and - knde), but it makes more sense, especially in the NON1PL form, to analyze this -e as belonging to the subject person-number marker that follows the mood marker. The final -d then found in Digul Wambon 1SG, 1PL and NON1PL forms is absent in both Yonggom Wambon and Mandobo; unfortunately it is at this point not known (1) how this -d became attached to the Digul Wambon -ken form or (2) what its function is. Besides the Digul Wambon allomorphs of -ken, some other phonological changes also occur in the paradigms in Table 6.4. The realis marker -t changes to /r/ intervocalically in Mandobo and Yonggom Wambon, and to /l/ in Digul Wambon. In Mandobo and Yonggom Wambon, the /k/ of -ken changes to /g/ intervocalically, while vowel harmonization also occurs, for example in Yonggom Wambon ti-gin-in ‘they are/were building’. These changes can be accounted for by the morphophonemic sound changes listed in Chapter2 (Section 2.2). All Dumut semi-finite realis forms are used to express actions that are currently happening or that happened in the recent past. In the three Dumut languages, -t and -ken forms have various other functions as well as different distribution frequencies. The following paragraphs offer a synchronic description of -t and -ken forms in the three Dumut languages. 96 6. Mood

Mandobo semi-finite realis forms In Mandobo, semi-finite realis -ken forms are far more frequent than semi-finite realis -t forms; in all Mandobo texts (Drabbe 1959:10-102), there are only six occurrences of -t forms, while there are hundreds of -ken forms. There are so few -t forms because they only occur in what Drabbe(1959:32,104-105) describes as conditional constructions. These constructions have a -t form in the protasis and a semi-finite irrealis form or an imperative form in the apodosis. So in (43), ŋgwandet is the -t form and the apodosis is formed by the semi-finite irrealis keaon, whereas in (44) mendaron ends the protasis and noa forms the apodosis. The occurrence of Mandobo realis -t forms is further discussed in Sections 11.3.3 and 11.3.6.

(43) Murup ŋgwande-t kea-on. rain stop-REAL[NON1SG] go[IRR]-1PL. ‘If it stops raining, let us go.’ Mandobo (Drabbe 1959:32)

(44) Mbe kee-r-in-an ane ro menda-r-on, that bring-REAL-NON1PL-PST eat with come-REAL-NON1PL noa. give[IRR.NON1SG] ‘if you (PL) brought something with you to eat, give (it to me).’ Mandobo (Drabbe 1959:31)

Thus the difference between -t and -ken forms in Mandobo is that -t forms occur speciﬁcally in clauses with a conditional interpretation, while -ken forms occur in all other contexts where a Mandobo speaker wants to express realis meaning. The Mandobo sentences in (45) contain ten -ken forms occurring with perception verbs (itigio-gen, ndot-ken), before the connective doro in a tail-head linkage construction (ra-ŋgen), with the sequence marker -aro, and in its usual position, clause-ﬁnal position (ko-gen, ra-ŋgen, taga-ŋgen, ra-ŋgen).

(45) Teen-gen do ko-gen. Ko itigio-gen shoot-REAL[NON1SG] CONN go-REAL[NON1SG] go see-REAL[NON1SG] do, kümo ra-ŋgen. Kümo ra-ŋgen doro, CONN die lie-REAL[NON1SG] die lie-REAL[NON1SG] CONN ndot-ken do, a wüop mbo u mun mbo hear-REAL[NON1SG] CONN inside center stay pig young.one DUR taga-ŋgen. Ko küap tagamo, me-re make.sound-REAL[NON1SG] go human tell come-NON.CLOSE itigio-gen do, u mun mbo mangor o namo see-REAL[NON1SG] CONN pig young.one stay mouth CONN through meri-gin-aro ra-ŋgen. come.down-REAL[NON1SG]-SEQ lie-REAL[NON1SG] 6.3. Realis Mood 97

‘When he (Ngou) shot (it), it (the pig) went. He (Ngou) goes and sees (that) (the pig) is dead. (The pig) is dead and Ngou hears a piglet make sound inside the pig (in the pig’s belly). He goes and tells the humans, he comes and sees (that) the piglet has come out (of the pig) through the mouth and is lying down.’ Mandobo (Drabbe 1959:17)

Yonggom Wambon semi-finite realis forms From Drabbe’s description of Yonggom Wambon, it appears that semi-finite realis -t and -ken forms have very similar functions. He notes that “in general, the same thing can be said about the use of -t forms as about the use of -ken forms.”5 However, languages hardly ever have two very different forms with exactly the same function. The question that therefore needs to be answered is what distinguishes -t forms from -ken forms in Yonggom Wambon. A detailed study of Drabbe’s analysis of Yonggom Wambon reveals two possible differences between -t and -ken forms. First, the sequential marker -a may only occur after -t forms, and such sequential forms will always be followed by a -ken form or a past finite -ran verb form (Drabbe 1959:134). According to Drabbe, a second instance in which only -ken forms but no -t forms may occur is after a switch in subject has occurred in a sentence (Drabbe 1959:133). However, both of these claims by Drabbe can be disproved by examples from Yonggom Wambon texts. In (46) and (47), a -t form, rather than Drabbe’s prescribed -ken or -ran form follows the verb form to which the sequence marker -a is attached.6

(46) Mbanep mende-r-a mbarukrawa-t. crocodile come-REAL[NON1SG]-SEQ sleep.with-REAL[NON1SG] ‘The crocodile comes and sleeps with (the daughter).’ Yonggom Wambon (Drabbe 1959:149)

(47) Mberema-r-a turo-ro enop jojomara ko take-REAL[NON1SG]-SEQ go.up-NON-CLOSE tree tree.type go andawa-t. bind-REAL[NON1SG]. ‘He takes (them) and goes up and binds one tree to another.’ Yonggom Wambon (Drabbe 1959:153)

5“In ’t algemeen kunnen we over het gebruik der -t vormen hetzelfde zeggen als over dat der -ken vormen” (Drabbe 1959:129). 6The Yonggom Wambon sequence marker is -a, whereas in Mandobo it is -ra or -aro. The Yong- gom Wambon sequence marker -a cannot be reanalyzed as actually being -ra, as -a occurs after NON1PL forms in which it is not directly preceded by an /r/:

(1) ra ko mbukma-r-in-a and-r-in-an. hold go stay-REAL-1PL-SEQ eat-REAL-1PL-PST ‘They take it, stay (sit down) and eat (it).’ Yonggom Wambon (Drabbe 1959:149) 98 6. Mood

The second difference claimed by Drabbe – that realis -t forms would not occur immediately after a change of subject – is invalidated by (48) and (49), where a semi- ﬁnite -t form occurs in a sentence after a change of subject has occurred.

(48) irow=e soma-t mari ŋgamburu ke-t. stone=TOP throw-REAL[NON1SG] come.down thump be-REAL[NON1SG] “He throws the stone and it comes down with a thump.” Yonggom Wambon (Drabbe 1959:154)

(49) Ri-r-in kande ok rira-t... cut.down-REAL-NON1PL fall.down river go.down-REAL[NON1SG]... ‘They cut down (the tree), it falls down, it goes down the river...’ Yonggom Wambon (Drabbe 1959:154)

Thus, the differences in function between -t and -ken forms proposed by Drabbe are not supported by his own Yonggom Wambon textual data. A further look at Yonggom Wambon texts will reveal some differences between Yonggom Wambon semi-finite -t and -ken forms. A quick glance at Yonggom Wambon texts reveals that -t forms are far more frequent than -ken forms. A count of all semi-finite verb forms in Yonggom Wambon texts (Drabbe 1959:145-157) results in 25 -ken forms and 258 -t forms.7 Thus, in these texts, 91% of all semi-finite realis forms are -t forms, whereas only 9% are -ken forms. Although this difference in frequency is significant, it does not reveal when Yonggom Wambon speakers use a -ken form rather than a -t form. Hypothetically, Yonggom Wambon -t and -ken forms could still be used interchangeably. However, there are circumstances in which -ken forms are more likely to be used than -t forms. When looking at the 25 -ken forms found in the Yonggom Wambon texts, one finds that a high percentage of them, 60% to be precise, are formed with position verbs mba- ‘sit’, jaŋ- ‘lie’ and ri- ‘stand,’ which are used in progressive or durative constructions in Yonggom Wambon. Furthermore, almost half (40%) of all -ken forms found occur in the final clause of a sentence rather than in sentence-medial clauses.8 However, these are only tendencies of Yonggom Wambon -ken forms, as they do also occur non-finally and with verbs other than position verbs. Furthermore, these tendencies towards occurring finally and with position verbs do not distinguish -ken forms from -t forms, as 27% of all -t forms occur finally and 10% of all -t forms are made with position verbs. These results are given in Table 6.5. To conclude, -t forms are far more frequent than -ken forms in Yonggom Wambon, while -ken forms tend to occur with position verbs and clause-finally. Hypotheti- cally, the two forms could be used interchangeably because no individual identifying difference between the two forms was found. Only when the sequence marker

7Fifty-five of these -t forms occur with the sequence marker -a. 8Overall, sentence-medial clauses with semi-finites are far more frequent than sentence-final clauses with semi-finites. Of 283 clauses containing semi-finite verbs in Yonggom Wambon, only 80 were sentence-final clauses. Of semi-finite realis -t forms, only 27% occurred in sentence-final clauses. 6.3. Realis Mood 99

Table 6.5: Comparison of -t and -ken forms in Yonggom Wambon texts

-ken forms -t forms total 25/283 = 9% 258/283 = 91% of which are ﬁnal 10/25 = 40% 70/258 = 27% of which are position verbs 15/25 = 60% 26/258 = 10%

-a attaches to a semi-ﬁnite form must it be a -t form. Knowing the exact difference in meaning or function between Yonggom Wambon -t and -ken forms is, based on the current data, beyond our grasp.

Digul Wambon semi-finite realis forms Digul Wambon, like Yonggom Wambon and Mandobo, has two realis mood markers, -ke(nd) and -t. The allomorphs of Digul Wambon -ke(nd) were discussed above: in NON1SG the mood marker is realized as -ke or, if the verb stem preceding the mood marker ends in a vowel, as -xe. In 1SG, 1PL and NON1PL verb forms, the mood marker is realized as -kend or, if shortening occurs, as -knd. The mood marker -t is realized as -t except when it occurs intervocalically; it is then realized as -l. When Digul Wambon realis semi-finites occur clause-medially, they are obligatorily followed by the coordinating connective =o. A study of Digul Wambon texts (de Vries and Wiersma 1992:83-96 and Jang 2008:125-131) reveals a difference between the two Digul Wambon semi-finite realis verb forms. The -ke(nd) forms occur in constructions with connective verbs that express aspectual meaning,9 while -t forms do not.10 Two aspectual constructions exist with which -ke(nd) forms often co-occur. The first construction is with the non- finite form of the verb kit ‘to finish’, indicating that an ongoing action has finished. In 50, the measuring of string is the action which is completed before another action.

(50) yat-ke-l=o awaiktop=ka lavi-lo light-be-REAL[NON1SG]=CONN early.morning=ERG descend-NON.CLOSE kono sifal=e hano-mo-knd-eva-n=o and measure.string=TOP measure-do-REAL-1PL-tr.nasal=CONN kit-mbel=o mbup-ma-l-eva-mbo. ﬁnish-SEQ=CONN split-do-REAL-1PL-PST ‘When it became light we descended and we ﬁnished measuring string and we split (the logs).’ Digul Wambon (de Vries & Wiersma 1992:88)

9See also Section 8.3. 10Jang(2008) recognizes that -ke(nd) forms occur with an ongoing or progressive meaning: “The majority of the frequent uses of the ﬁnal verb present form [= Jang’s analysis of ke(nd) forms] in medial clauses are understood to be pragmatically analogous to the progressive aspect of ﬁnal verbs” (Jang 2008:60). De Vries & Wiersma(1992) only note that -ke(nd) forms are present narrative verb forms. 100 6. Mood

The second aspectual construction is formed with the NON1SG -ke(nd) of the verb ko ‘to go,’ and has means ‘until’.11 A verb conjugated with -ke(nd) precedes koxeno and expresses an ongoing action. In (51), the sawing is continued until something else happens.

(51) Osak osax-a waxot han lap-kend-eva-n=o koxeno Again again-CONN month one saw-REAL-1PL-tr.nasal=CONN until taxemo miŋgu-n=eve taxem-ka lav=o nda-mbel=o three week-tr.nasal=TOP three-ERG saw=CONN not.be-SEQ=CONN yat-ke-l=o nux=e hitulov=e ndave-l-ep-mbo. light-be-REAL[NON1SG]=CONN I-TOP Thursday=TOP return-REAL-1SG-PST ‘Again one month we sawed until three weeks had gone, the sawing ﬁnished and when it became light on Thursday I returned.’ Digul Wambon (de Vries & Wiersma 1992:92)

Of 25 -ke(nd) forms found in Digul Wambon texts, 16 occur in either of these two aspectual ‘until’ or ‘ﬁnish’ construction.12 Thus Digul Wambon -ke(nd) forms, like Yonggom Wambon -ken forms, tend to occur in aspectual contexts, a tendency that is not shared by Digul Wambon -t forms. Rather, Digul Wambon -t forms occur when there is a switch in subject, as illustrated in (52), where the subject changes after loxolevo from ‘I’ to ‘they’ and after akmoxeleno from ‘they’ to ‘we’.

(52) Et-mbel=o ap nda-no loxo-l-ev=o kav=e Leave-SEQ=CONN house come-SIM speak-REAL-1SG=CONN people=TOP nuk akmoxe-l-en=o kono oy=e talemo hali-mbel=o me follow-REAL-NON1PL=CONN next pig=TOP cut carry-SEQ=CONN lap-ndave-l-eva-mbo. take-come-REAL-1PL-PST ‘I left and came to the house speaking and people followed me and we cut up the pig and carried it and came (back).’ Digul Wambon (de Vries & Wiersma 1992:84-85)

On the other hand, -ke(nd) forms do not necessarily mark a switch in subject. Thus the main difference between Digul Wambon -t and -ke(nd) forms is not their frequency (besides 25 -ke(nd) forms, 22 -t forms were found in Digul Wambon texts), but in whether or not they mark a switch in subject. In addition, Digul Wambon kend- forms tend to occur in continuous or progressive contexts.

11The stem of the verb ‘to go’, namely ko, can also have the meaning ‘until’. 12The -ke(nd) forms that were followed by a time clause such as ‘4 weeks’ or ‘afternoon’ were also counted as occurring in an ‘until’ construction. It must also be noted that -ke(nd) verbs are not the only verb forms that can precede kitmbelo or koxeno; non-ﬁnite SS forms and fully ﬁnite past tense forms also occur in that position. 6.3. Realis Mood 101

6.3.3 Awyu realis forms The realis verb form found in all four Awyu languages consists of a verb stem followed by a realis marker and a person-number marker. In these forms, no tense marker occurs, and the realis forms can be used to express both events occurring presently (Aghu, Pisa, Shiaxa, Yenimu) and those occurring in the recent past (Aghu, Pisa). All Awyu languages have two realis mood markers, either of which follows the verb stem; -d in ﬁrst person forms and -k in non-ﬁrst person forms.

Table 6.6: Awyu realis paradigms

1SGNON 1SG 1PLNON 1PL

AXU da-d-e da-k-e da-d-oã da-k-enã come.I-REAL.1-1SG come.I-REAL.NON1- come.I-REAL.1-1PL come.I-REAL.NON1- NON1SG NON1PL PSA de-d-i de-x-i de-d-a de-x-enã come.I-REAL.1-1SG come.I-REAL.NON1- come.I-REAL.1-1PL come.I-REAL.NON1- NON1SG NON1PL SHI da-d da-g da-d-a da-gox-ona hear.I-REAL.1 hear.I-REAL.NON1 hear.I-REAL.1-1PL hear.I-REAL.NON1- NON1PL YEN wu-d-i wu wu-d-a wu-gox-ona come.I-REAL.1-1SG come.I come.I-REAL.1-1PL come.I-REAL.NON1- NON1PL

Sources: Aghu (Drabbe 1957:10); Shiaxa, Pisa, Yenimu (Drabbe 1950:111-112)

Table 6.6 contains a realis paradigm from each of the four Awyu languages, clearly showing the alternation between realis marker -d in first person forms and realis marker -k in non-first person forms. The realis marker -k has several allomorphs. In Aghu, -g, -x and -ox are allomorphs of -k; Drabbe states that there is no rule governing which allomorph is chosen (Drabbe 1957:10). For Pisa, Drabbe notes that allomorph -g is used when the verb stem ends in /e/ or /o/, allomorph ŋg occurs after a nasal vowel, while -k appears after /i/ or /a/ (Drabbe 1950:112). In Shiaxa and Yenimu, -kox or -gox are found in the NON1PL forms; -kox is not an exact allomorph of -k, but rather is -k plus -ox. However, it is unclear how this form came about; it will be glossed as an allomorph of -k. Unlike for Dumut languages, little can be said about the distribution or specific function of semi-finite realis forms in Awyu languages, as a complete grammar is available only for Aghu.13

6.3.4 Kombai realis forms Kombai marks realis mood like Awyu languages: -d is the mood marker in ﬁrst person forms, while -k (allomorph -g intervocalically) is the non-ﬁrst person mood marker. However, -g only occurs as a mood marker with verbs whose stems end in

13The function of Aghu semi-ﬁnites is discussed in Chapter 11 on Clause Linkage. 102 6. Mood a nasal vowel; the realis marker does not appear in non-ﬁrst person forms when the verb stem does not end in a nasal vowel. These two situations are illustrated by two paradigms in Table 6.7.

Table 6.7: Kombai realis paradigms

non-nasal stem nasal stem

1SG xa-d-e(f) u-d-e(f)˜ go.I-REAL.1-1SG kill.I-REAL.1-1SG NON1SG xa u-˜ g-e go.I[REAL.NON1SG] kill.I-REAL.NON1-NON1SG PL xa-d-efo u-d-efo˜ go.I-REAL.1-1PL kill.I-REAL.1-1PL NON1PL xa-no u-˜ g-eno go.i-[REAL]NON1PL kill.I-REAL.NON1-NON1PL

Source: de Vries(1993:22-23)

The non-first person forms for the verb xa ‘to go’ are formed by adding a person- number marker to the verb stem; no mood marker occurs.14 In the non-first person forms of the verb u˜ ‘to kill’, the non-first person realis marker -g appears. Note that the analysis of -d and -g as mood markers differs from the analysis given in de Vries (1993), where -d and -g are analyzed as being part of the person-number markers.

6.4 Reconstruction

Proto Dumut Dumut languages have two realis mood markers, -t and -ken, which occur in separate paradigms. In Digul Wambon, -ken occurs mainly in aspectual contexts with completive kitmbelo ‘ﬁnish’ and progressive koxeno ‘until’ constructions. In Yonggom Wambon, 60% of all -ken forms occur with position verbs, which in Awyu-Dumut languages express progressive meaning (see Chapter8 on Aspect). Interestingly, Aghu has a durative marker -ke to express ongoing action; thus bey-oa-ŋge15 means ‘we are beating sago’ in Aghu (Drabbe 1957:21). The origin of Dumut -ken is likely to be this progressive marker -ke, now only found in Aghu; this marker, in turn, possibly derived from the auxiliary verb *ke ‘to be’. In Proto Dumut, *-ken forms then only occurred in progressive contexts. Thus *-ken is reconstructed as a realis marker in Proto Dumut. Proto Dumut *-ken occurred in constructions that express progressive meaning and is thought to have originated from a progressive marker -ke derived from *ke ‘to be’. Realis

14Kombai irrealis non-first person forms are also formed by adding a person-number marker to a verb stem. The non-first person realis forms can be distinguished from irrealis non-first person forms because different stems are used. 15The /k/ of -ke has become ŋg because it occurs intervocalically and after a nasal vowel. 6.4. Reconstruction 103

*-t is reconstructed as the older Proto Dumut realis marker; the *-ken form entered Proto Dumut later and started competing with realis *-t forms. The reason to reconstruct *-t as the older form is because it has been marginalized in Mandobo and restricted in Digul Wambon. Linguistic archaisms tend to survive in marginalized or less basic functions, while the new forms that replace them are more widespread (Koch 1996:219). In Mandobo, -ken forms pushed -t forms aside to the point where Mandobo -t forms only occur in conditional clauses. In Yonggom Wambon, -t forms remained frequent while very few -ken forms occur. The -ken forms that do occur in Yonggom Wambon do not all appear in a progressive context, unlike in Proto Dumut. In Digul Wambon, -t and -ken forms are about equally frequent. To summarize, Proto Dumut had two realis markers, *-t and *-ken. Proto Du- mut *-ken originated as a progressive marker and only occurred in progressive contexts. In Proto Dumut’s daughter languages, especially in Mandobo, *-ken forms expanded in function and frequency, pushing aside the more archaic *-t forms.

Proto Awyu

All four Awyu languages have -d as a realis marker in 1SG and 1PL verb forms, and realis marker -k in NON1SG and NON1PL verb forms. Hence for Proto Awyu *-d is reconstructed as REAL.1 marker and *-k as REAL.NON1 marker. The only difference in meaning and function of Proto Awyu *-d and *-k mood markers is the subject person-number marker with which they co-occur. Proto Awyu might have started to use two different mood markers in 1SG and NON1SG contexts because final consonant deletion leveled out the difference between 1SG and NON1SG subject person-number markers. Thus Proto Awyu Dumut *-ep 1SG and *-en eventually both became *-e in Proto Awyu because of the deletion of final /p/ and /n/. Before final consonant deletion took place in Proto Awyu, mood markers were not needed to distinguish between first and non-first person, and the proto language at that stage might have had only one mood marker. This one mood marker is most likely to have been *-d, which is still found in both first and non-first person in several Awyu past tense paradigms (see Chapter7 on Tense). The mood marker *-k then later entered Proto Awyu, replacing *-d in non-first person forms.

Proto Awyu-Dumut For Proto Dumut, two realis markers were reconstructed, *-t and *-ken, where *-t is the older realis marker and *-ken has its origin in a progressive aspectual marker -ke. For Proto Awyu, *-d REAL.1 and *-k REAL.NON1 were reconstructed; *-d was hypothesized to be the older form, while *-k came to be used as a realis marker once the distinction between ﬁrst and non-ﬁrst person became unclear in Proto Awyu. Given these reconstructed situations in the daughter languages, what can be said about mood in Proto Awyu-Dumut? In both Proto Dumut and Proto Awyu, *-t or *-d was reconstructed as the older form, while *-ken and *-k entered these proto languages at a later stage. Proto Awyu 104 6. Mood

*-k and Proto Dumut *-ken are false cognates. Although both express realis mood, they differ in form, function and origin.16 These differences are hard to explain, and thus, following the form-function constraint introduced in Chapter1, neither *-ken nor *-k is reconstructed for Proto Awyu-Dumut. Rather, Proto Awyu-Dumut had one realis marker *-t,17 which occurred in both first and non-first person forms. In Proto Awyu and Kombai, the Proto Awyu-Dumut realis marker *-t was replaced in NON1 forms by *-k when 1SG and NON1SG endings became the same due to final consonant deletion; in Proto Awyu, *-k was then pressed into use as a non-first person realis marker in all instances, while in Kombai, -k only came to be used as a realis marker if the verb stem ended in a nasal vowel. Thus Kombai illustrates an intermediate step between Proto Awyu-Dumut and Proto Awyu, in which the realis marker *-k has not spread to all NON1 verb forms. The fact that Kombai and Proto Awyu share the innovation of a realis marker -k supports a subgrouping of Kombai with the Awyu languages rather than with the Dumut languages, a hypothesis to which I will return in the conclusion of this book.

16Proto Dumut *-ken has its origins in a progressive marker -ke and has aspectual functions, while Proto Awyu *-k marks non-ﬁrst person, besides expressing realis meaning. 17*-t rather than *-d is reconstructed for Proto Awyu-Dumut because /t/ is likely to have changed to /d/ intervocalically in Proto Awyu. It must be noted however that /t/ to /d/ is not a regular sound change occurring (intervocalically) between Proto Awyu-Dumut and Proto Awyu. 7

Tense

Awyu-Dumut languages are mood-driven, as demonstrated in the chapter on Mood. Awyu-Dumut verbs are first marked for mood and then optionally for tense. Thus tense is only marked on the verb if mood is expressed as well; if the verb is in realis mood, a past tense marker may occur, whereas if the verb is in irrealis mood, a future tense marker may attach to the verb. Figure 7.1 depicts the interaction between mood and tense in Awyu-Dumut languages. The verbs that are marked for both mood and tense are finite verb forms, those only marked for mood are semi-finite verb forms, whereas non-finite verb forms mark neither tense, mood, nor subject person-number, but may contain a temporality marker.1 Both semi-finite and finite verbs do mark subject person-number. In semi-finite verb forms, the subject person-number marker always follows the mood marker. In Dumut and Kombai finite verbs, the tense marker comes after the subject person-number marker, whereas in the majority of Awyu finite verbs, the tense marker precedes the subject person- number marker.

Verbs

+Mood –Mood

–person-number(PN)

+REAL +IRR +temporality -temporality +TNS,+PN –TNS,+PN +TNS,+PN –TNS,+PN (past) (future)

Figure 7.1: Awyu-Dumut verb structure

Semi-ﬁnite verbs, which do not contain a tense marker but which are marked for mood, are far more frequent than ﬁnite forms, which express both tense and mood.

1As was pointed out in Chapter6 on mood, Awyu non-ﬁnite verbs do express mood, although not by a separate mood marker. 106 7. Tense

Indeed, in the Aghu text “Two Orphan Girls” (Drabbe 1957:55-62), there is one occurrence of a historical past form at the very end of the story, which places the whole story, a myth, in the historical past. In the same story, there are a few occurrences of distant past forms but only with the verb e- ‘to sit’ in durative contexts (Drabbe 1957:40).2 On the other hand, there are hundreds of semi-finite realis forms in this story. Five of the nine other Aghu stories and myths recorded by Drabbe also have a historical past form at the end of the story but few or no other finite past tense forms, while semi-finite forms are abundant. The two tenses that occur in Awyu-Dumut languages are past tense and future tense. There is no present tense; rather, actions that occur at or near the moment of speaking are expressed using a semi-finite realis form, as described in Chapter6 on Mood. The Dumut morphemes -t and -ken were originally analyzed as ‘present- neutral’ tense forms by both Drabbe(1959) and de Vries & Wiersma(1992) but were re-analyzed as realis mood markers in Section 6.3.1. In Awyu languages multiple past tenses are found, while Dumut languages have one past tense. A future tense is found in most but not all Awyu-Dumut languages. In Section 7.2, future tense is discussed after past tenses are treated in Section 7.1. Reconstructions are presented at the end of each section.

7.1 Past Tense

Foley notes that “tense systems in Papuan languages are generally more elaborate than in European languages. In almost all cases more than one past tense is distinguished” (Foley 1986:159). This is true for Awyu languages, where a distinction can be made between the recent past of today, yesterday’s past, the distant past and the historical past. Dumut languages have only one past tense, while Kombai has no past tense at all. Both Awyu and Dumut past tenses are expressed through tense markers.

7.1.1 Dumut Past Tense All three Dumut languages have one past tense, which is formed by sufﬁxing a past tense marker to a semi-ﬁnite realis -t form after the subject person-number marker. The Dumut past tense is used to refer to any action or process that took place before the time of speaking. The past tense marker in Yonggom Wambon and Mandobo is -an, while Digul Wambon employs -mbo as a past tense marker. A past tense paradigm for each of these three languages is presented in Table 7.1. A few changes that occur in these paradigms due to morphophonemic rules need to be explained. When the realis marker -t occurs intervocalically, it changes to /r/ in Yonggom Wambon and Mandobo, and to /l/ in Digul Wambon. In Digul Wambon the sounds /p/ and /mb/ combine to yield /p/, which then does not

2For an explanation of how position verbs are used in durative contexts in Awyu-Dumut languages, see Section 8.1. 7.1. Past Tense 107

Table 7.1: Dumut past paradigms

Mandobo Yonggom Wambon Digul Wambon

1SG roroan etagarewan andelepo ro-t-op-an etaga-t-ep-an ande-t-ep-mbo put-REAL-1SG-PST see-REAL-1SG-PST eat-REAL-1SG-PST NON1SG roran etagaran andetmbo ro-t-an etaga-t-an ande-t-mbo put-REAL[NON1SG]-PST see-REAL[NON1SG]-PST eat-REAL[NON1SG]-PST 1PL rorunan etagarewanan andelevambo ro-t-un-an etaga-t-ewan-an ande-t-eva-mbo put-REAL-1PL-PST see-REAL-1PL-PST eat-REAL-1PL-PST NON1PL rorinan etagarinan andelembo ro-t-in-an etaga-t-in-an ande-t-e-mbo put-REAL-NON1PL-PST see-REAL-NON1PL-PST eat-REAL-NON1PL-PST

Sources: Mandobo (Drabbe 1959:10); Yonggom Wambon (Drabbe 1959:130); Digul Wambon (de Vries & Wiersma 1992:26) fricativize, unlike /p/ in the Yonggom Wambon 1SG form, which fricativizes to /w/. The /p/ that would be expected in the Mandobo 1SG form disappears between the /a/ and the /o/, as Drabbe notes: “between /a/ and /o/ the /p/ is dropped, and /w/ does not occur as a ligature”3 in Mandobo. For Proto Dumut, *-an is reconstructed as the past tense marker, as it occurs in both Mandobo and Yonggom Wambon. It is likely that Digul Wambon introduced -mbo as a past tense marker after losing Proto Dumut *-an. Digul Wambon -mbo might have its origin in the durative marker -mbo, which in turn is derived from the locative verb mba ‘to sit, to stay’.

7.1.2 Awyu Past Tenses Awyu languages can have up to four past tenses. The four past tenses found in Awyu languages are the hodiernal and hesternal pasts, respectively used to express actions that occurred today or yesterday, and the distant and historical pasts. Not all Awyu languages have all four tenses. Aghu, for example, distinguishes only two past tenses, the distant and the historical, and uses semi-ﬁnite realis forms to express events that occurred recently, as described in Section 6.3.3. Pisa also has just two past tenses; hesternal past and distant past. Shiaxa and Yenimu each have four past tenses. The semantic use of the past tenses is not exact. Rather, it is based on how much time the speaker feels has passed. Drabbe observes the following about the use of past tenses in Aghu:

The boundary between the use of both past tenses is of course not exactly marked off, all the more because the Papuan, and especially the Aghu Awyu who has no wet and dry season to measure time by, has a very bad

3“Tussen /a/ en /o/ vervalt echter de /p/, en treedt geen /w/ op als verbinding” (Drabbe 1959:6). 108 7. Tense

sense of time. In many cases both past tenses might be used, depending on how much time one feels has passed.4

The following paragraphs give a description of each of the four past tenses found in Awyu languages.

Hodiernal past The hodiernal past form is used to express actions that took place on the same day as when the utterance was spoken. Only Shiaxa and Yenimu have specific hodiernal past forms; Aghu and Pisa use semi-finite realis forms, consisting of a verb stem, a mood marker and a person-number marker, to express hodiernal actions. Semi- finite realis paradigms for all four Awyu languages can be found in Table 6.6 in Chapter6, while Table 7.2 gives hodiernal past paradigms for Shiaxa and Yenimu.

Table 7.2: Shiaxa and Yenimu hodiernal past paradigms

Shiaxa Yenimu

1SG dadowe atidiki da-d-owe ati-di-k-i come.I-REAL.1-1SG bite.I-REAL.1-HOD-1SG NON1SG dagore atiki da-k-ore ati-k-k-i come.I-REAL.NON1-NON1SG bite.I-REAL.NON1-HOD-NON1SG 1PL dadowa atidika da-d-owa ati-di-k-a come.I-REAL.1-1PL bite.I-REAL.1-HOD-1PL NON1PL dagena atikina da-k-ena ati-k-k-ina come.I-REAL.NON1-NON1PL bite.I-REALNON1-HOD-NON1PL

Sources: Shiaxa (Drabbe 1950:113); Yenimu (Drabbe 1950:113-114)

Shiaxa forms its hodiernal past by using a specific set of person-number markers, different from the one it uses for semi-finite realis verb forms. The hodiernal past person-number markers are also used in all other Shiaxa past tenses. The two different person-number sets Shiaxa uses are listed in Table 7.3. The initial vowels of the singular person-number markers vary according to the final vowel of the stem: if a stem ends in /a/, then the singular marker start with /o/; otherwise, the initial vowel of the singular person-number marker is the same as the final vowel of the verb stem. In Yenimu, a hodiernal marker -k occurs throughout the paradigm; in first person forms it comes after the mood marker -d and in non-first person forms it combines

4“De grens tussen het gebruik der beide praeterita is natuurlijk niet nauwkeurig afgebakend, temeer daar de Papoea, en vooral de Aghu-Awyu die geen natte en droge seizoenen kent om er de tijd mee te meten, een zeer slecht begrip van tijd heeft. In veel gevallen zal men zowel het ene als het andere praeteritum gebruiken, al naar gelang men de verstreken tijd aanvoelt” (Drabbe 1957:12). 7.1. Past Tense 109

Table 7.3: Shiaxa person-number markers

ﬁnite past semi-ﬁnite forms realis forms

1SG -(V)we – NON1SG -(V)re – 1PL -owa -a NON1PL -ena -ona with the mood marker -k. The subject person-number markers in the Yenimu hodiernal past paradigm do differ from the subject person-number markers it uses in other paradigms, but this variation in form does not express a (distinctive) variation in meaning.

Hesternal past Awyu speakers use a hesternal past verb form when talking about an action that happened on the day before the time of speaking. Each Awyu language has its own way of forming the hesternal past. Table 7.4 gives hesternal past paradigms for Pisa, Shiaxa and Yenimu.

Table 7.4: Awyu hesternal past paradigms

Pisa Shiaxa Yenimu

1SG dedira dagodewe otodiﬁ de-d-i-ra da-go-d-ewe oto-d-iﬁ come.I-REAL-1SG-HEST hear.I-HEST-REAL-1SG ascend.I-REAL-1SG NON1SG dexira dagodere otodi de-k-i-ra da-go-d-ere oto-d-i come.I-REAL-1SG-HEST hear.I-HEST-REAL- ascend.I-REAL-NON1SG NON1SG 1PL dedara dagodowa otodifa de-d-a-ra da-go-d-owa oto-d-ifa come.I-REAL-1PL-HEST hear.I-HEST-REAL-1PL ascend.I-REAL-1PL NON1PL dexenãra dagodenã otodina de-k-enã-ra da-go-d-enã oto-d-ina come.I-REAL-NON1PL- hear.I-HEST-REAL-NON1PL ascend.I-REAL-NON1PL HEST

Sources: Pisa (Drabbe 1950:114); Shiaxa (Drabbe 1950:114); Yenimu (Drabbe 1950:114)

In Pisa, a tense marker -ra is added after the semi-finite realis form. This is the only Awyu past tense marker that occurs at the end of the verb form; all other Awyu past tense markers occur in between the mood marker and the subject person-number marker. In Dumut languages, past tense markers always occur verb-finally. In Shiaxa, a hesternal tense marker -go is positioned in between the verb stem and the mood marker -d, which in this paradigm occurs in both the first person 110 7. Tense and non-first person forms. The vowel in Shiaxa’s hesternal mood marker is not always -o but rather depends on the vowel which occurs in the stem. In Yenimu, the hesternal paradigm is distinguished from other Yenimu past paradigms by the occurrence of realis marker -d throughout the paradigm. Aghu uses its semi-finite realis paradigm to also express hesternal meaning, just as it uses these forms to express hodiernal meaning; the Aghu semi-finite realis paradigm is not included in Table 7.4 but can be found in Table 6.6.

Distant past All four Awyu languages have the possibility of expressing the distant past for events that did not take place recently nor a very long time ago. Table 7.5 contains a distant past paradigm from Aghu, Pisa, Shiaxa and Yenimu.

Table 7.5: Awyu distant past paradigms

Aghu Pisa

1SG fidke dedaxari fi-d-(a)k-e de-d-aka-ri come.I-REAL.1-DIST-1SG come.I-REAL.1-DIST-1SG NON1SG fiaki daki fi-ak-i de-ak-i come.I-DIST-NON1SG come.I-DIST-NON1SG 1PL fidkoã dedaxaya fi-d-(a)k-oã de-d-aka-ya come.I-REAL.1-DIST-1PL come.I-REAL.1-DIST-1PL NON1PL fiakenã dakinã fi-ak-enã de-ak-inã come.I-DIST-NON1PL come.I-DIST-NON1PL Shiaxa Yenimu

1SG atimakewe tagamaﬁ ati-mak-ewe taga-ma-ﬁ bite.I-DIST-1SG tell.I-DIST-1SG NON1SG atimakere tafamaki ati-mak-ere tafa-mak-i bite.I-DIST-NON1SG tell.I-DIST-NON1SG 1PL atimakowa tagamafa ati-mak-owa taga-ma-fa bite.I-DIST-1PL tell.I-DIST-1PL NON1PL atimakena tagamakina ati-mak-ena taga-mak-ina bite.I-DIST-NON1PL tell.I-DIST-NON1PL

Sources: Aghu (Drabbe 1959:10); Pisa (Drabbe 1950:115); Shiaxa (Drabbe 1950:115); Yenimu (Drabbe 1950:115)

In the Pisa and Aghu paradigms, the realis marker -d occurs in ﬁrst person forms, while realis -d is absent in the Shiaxa and Yenimu distant past paradigms. In Aghu 7.1. Past Tense 111 and Pisa, the distant past marker is -ak, with variants -k in Aghu and -axa in Pisa. Drabbe does not analyze the /a/ in Aghu as being part of the tense marker but writes it separately, without a gloss. A comparison with Pisa, however, justiﬁes analyzing Aghu’s distant past marker as -ak. The initial /a/ only occurs in NON1 forms, and then only if the verb stem ends in /i/ or /ü/. If the verb stem ends in an /a/ or an /e/, the /a/ of -ak replaces that vowel. Thus, the distant past paradigm for the Aghu verb ede ‘to give’ is

Table 7.6: Aghu distant past paradigm

SGPL

1 ededke ededkoã ede-d-(a)k-e ede-d-(a)k-oã give-REAL.1-DIST-1SG give-REAL.1-DIST-1PL NON1 edaki edakenã ede-ak-i ede-ak-enã give-DIST-NON1SG give-DIST-NON1PL

In Shiaxa, the distant past marker is -mak, while Drabbe claims that the Yenimu distant past marker is -ma (Drabbe 1950:115). However, in NON1 verb forms, a /k/ follows Yenimu -ma, and it makes more sense to analyze Yenimu as having a distant past marker -mak, like Shiaxa, of which the ﬁnal /k/ was dropped in ﬁrst person verb forms.

Historical past Aghu, Shiaxa and Yenimu each have a historical past to express events that took place a long time ago. All three languages have a tense marker that occurs in between the mood marker -d(i), found throughout the paradigm, and the person- number marker. The tense marker in Aghu is -ia, in Shiaxa it is -ra and in Yenimu it is -r. Table 7.7 contains three historical past paradigms, one for each of the Awyu languages with a historical past.

Comparison and reconstruction of Awyu past tenses Great diversity exists within and between the different Awyu past tenses. Not one tense is formed the same way in any of the four languages. In some paradigms, -d occurs throughout as a realis marker (Yenimu hesternal past, all historical pasts); in another paradigm it is absent altogether (Shiaxa distant past). Three distinct past tense markers occur throughout the paradigms: -(a)k, -ra and -ma(k). We ﬁnd -(a)k as a hesternal past marker (Yenimu) and as a distant past marker (Aghu, Pisa). The marker -ra is a hodiernal past tense marker in Pisa, but the historical past tense marker in Shiaxa and Yenimu. The marker -ma is only found as a distant past marker in Shiaxa and Yenimu, while the variant -mak occurs throughout the Shiaxa distant 112 7. Tense

Table 7.7: Awyu historical past paradigms

Aghu Shiaxa Yenimu

1SG dadia atidirawe atidreﬁ da-d-ia ati-di-ra-we ati-d-r-eﬁ come.I-REAL-HIST[1SG] bite.I-REAL-HIST-1SG bite.I-REAL-HIST-1SG NON1SG dadia atidirare atidra da-d-ia ati-di-ra-re ati-d-r-a come.I-REAL- bite.I-REAL-REAL- bite.I-REAL-REAL- HIST[NON1SG] NON1SG NON1SG 1PL dadiaoã atidirawa atidrefa da-d-ia-oã ati-di-ra-wa ati-d-r-efa come.I-REAL-HIST-1PL bite.I-REAL-HIST-1PL bite.I-REAL-HIST-1PL NON1PL dedianã atidirana atidrena da-d-ia-nã ati-di-ra-na ati-d-r-ena come.I-REAL-HIST- bite.I-REAL-HIST- bite.I-REAL-HIST- NON1PL NON1PL NON1PL

Sources: Aghu (Drabbe 1957:11); Shiaxa (Drabbe 1950:116); Yenimu (Drabbe 1950:116)

Table 7.8: Awyu past tenses

Aghu Pisa Shiaxa Yenimu

hodiernal past 1SG realis form realis form realis form+PN1 di+k+PN hodiernal past NON1SG realis form realis form realis form+PN1 k+k+PN hodiernal past 1PL realis form realis form realis form+PN1 di+k+PN hodiernal past NON1PL realis form realis form realis form+PN1 k+k+PN

hesternal past 1SG realis form d+PN+ra g(V)+d+PN1 d+PN hesternal past NON1SG realis form k+PN+ra g(V)+d+PN1 d+PN hesternal past 1PL realis form d+PN+ra g(V)+d+PN1 d+PN hesternal past NON1PL realis form k+PN+ra g(V)+d+PN1 d+PN

distant past 1SG d+(a)k+PN d+axa+PN mak+PN1 ma+PN distant past NON1SG (a)k+PN ak+PN mak+PN1 mak+PN distant past 1PL d+(a)k+PN d+axa+PN mak+PN1 ma+PN distant past NON1PL (a)k+PN ak+PN mak+PN1 mak+PN

historical past 1SG d+ia – d+ra+PN1 d+r+PN historical past NON1SG d+ia – d+ra+PN1 d+r+PN historical past 1PL d+ia+PN – d+ra+PN1 d+r+PN historical past NON1PL d+ia+PN – d+ra+PN1 d+r+PN past paradigm and in NON1 forms in the Yenimu distant past paradigm. Diversity is also found in the order of the morphemes; sometimes the tense marker precedes the mood marker (Shiaxa hesternal past, Shiaxa distant past, Yenimu distant past), while in other paradigms it follows the person-number marker (Pisa hesternal past). In all other cases the tense marker is placed in between the mood marker and the person-number marker. Table 7.8 summarizes the past tenses found in Aghu, Pisa, 7.2. Future Tense 113

Shiaxa and Yenimu. Concerning the use of Awyu past tenses, little is known because only the Aghu grammar description contains texts. It is therefore unclear in what contexts or with what frequency Awyu past tenses are used. For Aghu, it is clear that distant past forms are very infrequent (only two occurrences in all the texts), while historical past forms only occur at the beginning or the end of a story. Thus it can be hypothesized that the past tenses are not used very frequently in the other Awyu languages either. For these two reasons (great diversity and paucity of data) I do not consider it possible to reconstruct past tense(s) for Proto Awyu, except to say that Proto Awyu most probably had at least one past tense.

7.2 Future Tense

If a language has a distinction between realis and irrealis forms, as Awyu-Dumut languages do, that does not necessarily mean that the language also has a speciﬁc future tense. As Comrie notes:

Some languages have a basic modal distinction between realis and irrealis, where realis refers to situations that have actually taken place or are actually taking place, while irrealis is used for more hypothetical situations, including situations that represent inductive generalisations, and also predictions, including also predictions about the future [...] Since future time reference in these languages is subsumed under irrealis, while present time reference (in the absence of any other modal value) is subsumed under realis, it is indeed the case that present and future time reference will have different grammatical realisations, but without it being the case that these languages have a distinct future tense [...] We are therefore left with the problem of ﬁnding a language in which there is a separate grammatical form used for future time reference, but where the use of this form cannot be treated as a special use of a grammatical category with basically non-tense meaning. (Comrie 1985:45-46)

In two Dumut languages (Mandobo and Yonggom Wambon), one Awyu language (Aghu) and in Kombai, such a future tense is found. These languages have a specific future tense marker that is attached to the semi-finite irrealis form. In the three Awyu languages without a future tense marker (Pisa, Shiaxa and Yenimu), future meaning is expressed by the semi-finite irrealis forms. The future tense marker in Yonggom Wambon is -in, the Mandobo future marker is -en, while Aghu has a future tense marker -E and Kombai expresses future tense meaning by adding the future tense marker -i, which de Vries(1993) states is in free allomorphic variation with -e. Table 7.9 contains future tense paradigms for these four languages. In Yonggom Wambon and Mandobo, the future tense marker comes before the 1SG subject person-number marker. It then loses its final /n/ before the 1SG marker -ep, 114 7. Tense

Table 7.9: Dumut future tense paradigms

Mandobo Yonggom Wambon

1SG andeyep mayoyip ande-y-e-ep mayo-y-i-ep eat[IRR]-LIG-FUT-1SG descend[IRR]-LIG-FUT-1SG I will eat I will descend NON1SG andenen mayonin ande-n-en mayo-n-in eat[IRR]-NON1SG-FUT descend[IRR]-NON1SG-FUT you/he/she/it will eat you/he/she/it will descend 1PL andewonen mayowanin ande-won-en mayo-wan-in eat[IRR]-1PL-FUT descend[IRR]-1PL-FUT we will eat we will descend NON1PL andenonen mayonanin ande-non-en mayo-nan-in eat[IRR]-NON1PL-FUT descend[IRR]-NON1PL-FUT they will eat they will descend Aghu Kombai

1SG adeyE aiﬁ ade-y-E ai-f-i hear.II[IRR.1SG]-LIG-FUT go.II[IRR]-1SG-FUT I will hear I will go NON1SG adenE aini ade-n-E ai-n-i hear.II[IRR]-NON1SG-FUT go.II[IRR]-NON1SG-FUT you/he/she/it will hear you/he/she/it will go 1PL adoanE aifoni ad-oan-E ai-fon-i hear.II[IRR]-1PL-FUT go.II[IRR]-1PL-FUT we will hear we will go NON1PL adenanE ainoni ad-enan-E ai-non-i hear.II[IRR]-NON1PL-FUT go.II[IRR]-NON1PL-FUT they will hear they will go

Sources: Mandobo (Drabbe 1959:23); Yonggom Wambon (Drabbe 1959:128); Aghu (Drabbe 1957:14); Kombai (de Vries 1993:23) and a ligature /y/ connects the verb stem and the future tense marker. In all other instances, the future tense marker follows the subject person-number marker. Digul Wambon,Pisa, Shiaxa and Yenimu do not have a future tense marker that can be added to irrealis semi-finite forms. Drabbe does note that Pisa and Shiaxa can add the affirmative postposition ni after semi-finite irrealis forms when they want to express indicative future meaning. Drabbe even notes that this ni is never left out when future meaning is expressed in Shiaxa, although he leaves it out in his Shiaxa 7.2. Future Tense 115 future paradigm. Although Digul Wambon does not have a future tense marker, it does have a construction that expresses immediate future meaning. By placing a realis conjugated form of the verb ma ‘to do’ after a semi-finite irrealis form, immediate future tense is realized in Digul Wambon, as illustrated by the paradigm given in Table 7.10. Table 7.10: Digul Wambon immediate future paradigm

1SG atkip malep atki-ep ma-t-ep wrap[IRR]-1SG do-REAL-1SG I will wrap soon NON1SG atki mat atki ma-t wrap[IRR.NON1SG] do-REAL[NON1SG] you/he/she/it will wrap soon 1PL atkiwa ma(le)wa5 atki-ewa ma-t-ewa wrap[IRR]-1PL do-REAL-1PL we will wrap soon NON1PL atkina mat atki-na ma-t wrap[IRR]-NON1PL do-REAL[NON1PL]

Source: Digul Wambon (Jang 2008:34)

Jang(2008:34) further notes that if ma ‘to do’ is conjugated with the realis marker ke(nd) instead of with the realis marker -t, the construction has a remote future meaning. De Vries & Wiersma(1992:27) have a different analysis than Jang(2008), claim- ing that -mat is Digul Wambon’s future tense marker and that in 1SG and 1PL forms the person-number markers appear twice, before and after ma(t). Thus de Vries & Wiersma(1992) analyze atkip malep as in (53). (53) atki-p mal-ep wrap[IRR]-1SG FUT-1SG ‘I will wrap’ Jang’s analysis is more elegant and has more credibility because Kombai also uses the verb ma to express immediate future meaning, as in (54). In Kombai, a realis form of the verb ma ‘to do’ follows an inﬁnitive verb, marked by the inﬁnitive marker -ni, to express what the speaker is about to do or say. (54) Doü ade-ni ma-d-e Sago eat-INF do-REAL-1SG ‘I am about to eat sago’ Kombai, de Vries(1993:17)

5Jang(2008) places le in the 1PL form malewa in parentheses, while de Vries & Wiersma(1992) also note that the /l/ does not appear in the 1PL form. This is an idiosyncracy for which the current data offer no explanation. 116 7. Tense

Reconstruction of Awyu-Dumut future tense Yonggom Wambon, Mandobo, Aghu and Kombai respectively have future tense markers -in, -en, E and -i/-e. These are all reflexes of the same proto future tense marker, which is reconstructed as *-e in Proto Awyu and as *-en/*-in in Proto Dumut and Proto Awyu-Dumut. Kombai and Proto Awyu dropped the final consonant /n/. Whether the vowel in the Proto Dumut and Proto Awyu-Dumut future tense marker should be *e or *i cannot be determined because no regular sound changes were found which could help answer this question. The Awyu-Dumut languages that do not have a reflex of Proto Awyu-Dumut/Proto Dumut *-en/*-in or Proto Awyu *-e are thought to have dropped future tense marking altogether, which is not unlikely if one considers how infrequent finite verb forms are in Awyu-Dumut languages. The use of the support verb ma ‘to do’ to express immediate future tense in Kombai and Digul Wambon is a shared retention, a construction that was lost - or not attested - in all other Awyu-Dumut languages.

7.3 Summary

Awyu-Dumut languages, which are mood-driven, all have distinct verb forms in which tense is marked. These tensed verb forms occur very infrequently, which perhaps explains why the Proto Awyu-Dumut future tense marker does not appear in all Awyu-Dumut languages, and why Digul Wambon lost Proto Dumut past tense marker *-an. At the same time, there is great freedom for Awyu-Dumut languages to vary in their realizations of tense, resulting in a plethora of past tenses. Indeed, the diversity within Awyu-Dumut past tenses is so great that no Proto Awyu or Proto Awyu-Dumut past tense(s) can be reconstructed. 8

Aspect

While tense is concerned with the sequence of events in real time and mood marks the speaker’s attitude towards the situation, aspect describes the internal temporal or phasal structure of actions or states. Thus, amongst other things, the aspect of a verb can express whether an action is ongoing, whether it has been completed, if it occurs over a long period of time or whether it happens once or multiple times. Within Awyu-Dumut languages, aspectual structures are employed to express whether an action is ongoing (progressive) or completed, whether an action takes place habitually, and whether or not an action is repeated and/or iterative. A continuous or progressive aspect implies an ongoing, dynamic process. The habitual aspect on the other hand describes a situation that is characteristic of an extended period of time. Section 8.1 describes how Awyu-Dumut locative verbs can express durative meaning, Section 8.2 discusses how iterative verb forms express habitual meaning, while Section 8.3 introduces connective verbs used to either encode completed action or bounded ongoing action.

8.1 Position Verbs

The Proto Awyu-Dumut position verb *mba ‘to sit/to stay’ has reflexes in all eight Awyu-Dumut daughter languages, and in all Awyu-Dumut languages except Shi- axa, this verb can express durative meaning when joined to another verb. More specifically, if a conjugated form of the verb mba ‘to sit/to stay’ follows a non-finite verb form in Digul Wambon (55), Yonggom Wambon (56), Aghu (57) or Kombai, that non-finite verb expresses an ongoing action.

(55) Yaxop hetax=o mba-knd-e. 3PL look=CONN stay-REAL-NON1PL ‘They are looking.’ Digul Wambon (de Vries & Wiersma 1992:33) 118 8. Aspect

(56) Waepmo-no mboge-nan rakonmo ok kimbarukmo-nan-in. travel-SIM sit-[IRR]NON1PL capsize river swim-NON1PL-FUT (while) you(pl) will be traveling, you will capsize and will swim (across) the river.’ Yonggom Wambon (Drabbe 1959:149)

(57) Gobümü-du-k emu xasu ax˜ı si-k-enã. Si-di-k cut.off-SS-CONN then nibung.leaf hut build-REAL-NON1PL build-SS-CONN ba-de-k syü e-ŋg-enã. E-ne-k okem’ ba-de-k sit-SS-CONN banana eat-REAL-NON1PL eat-SS-CONN forever sit-SS-CONN syü enigomo-do-k ba-dia-na. banana eat.it-SS-CONN sit-DIST.PST-NON1PL ‘They cut of nibung leaves and built a hut. They built it, stayed there and ate bananas. They ate, stayed forever eating bananas.’ Aghu (Drabbe 1957:62)

In Aghu example (57), the non-finite forms of the verb ‘to sit’ (badek) do not have durative meaning, while the fully finite distant past form of ba, badiana, does result in a durative reading of the preceding verb. Besides ba ‘to sit’, two other position verbs, namely e ‘to stand’ and i ‘to lie down’ may also function in a durative construction in Aghu. In Pisa, position verbs baxamo- ‘to sit’ or ri(mo)- ‘to lie down’ can occur after another verb, giving it a durative interpretation. From the description by Drabbe (1950:126) it is not clear, however, whether or not in Pisa the first verb is non-finite and whether or not the position verb is inflected. As examples, Drabbe gives radi rimo- ‘to continue holding’ and bu baxamo- ‘to fast for an extended period of time’. In Mandobo, the conjugated form of the verb mba does not express durative aspect, but a durative marker mbe, clearly derived from the verb mba, occurs before a conjugated verb, giving it a durative reading (58).

(58) Wemin ge-gen do, kinum raŋgi-ro ö Night be-REAL[NON1SG] CONN sleep lie.down-NON.CLOSE dream itigio-gen: yomorop u te mbe ŋ-gen. see-REAL[NON1SG] fruit pig FOC DUR eat-REAL[NON1SG] ‘When it was night, he lie down asleep and in a dream he saw: a pig was eating the fruit’ Mandobo (Drabbe 1959:14)

Like Mandobo, Aghu has a durative marker before the verb, namely bu,1 derived from the Aghu position verb ba (59). Kombai similarly has a pre-verbal durative marker bo (60).2

1The durative marker bu is shortened to b’ before a vowel-initial verb. The /u/ of bu often changes due to vowel harmony. 2The durative marker is written separately from the verb by Drabbe and attached to the verb by de Vries. Therefore they imply different levels of grammaticalization. What is certain is that the durative marker in Awyu-Dumut languages is not part of a serial verb construction or simply a non- ﬁnite form of the verb ‘to sit’, as the marker always has a slightly different form than the stem of the verb ‘to sit’. 8.1. Position Verbs 119

(59) Büsyaghatigi yoxo mida-de-k oxo bi Büsyaghatigi they downstream.come-SS-CONN river DUR ki-ŋg-enã. Oxo bi ki-ŋg-enã, mi-di-k bathe-REAL-NON1PL river DUR bathe-REAL-NON1PL come.down-SS-CONN fe u-g-e. one stab-REAL-NON1SG ‘The Büsyaghatigi clan, they come downstream and are bathing in the river. They are bathing in the river, he comes down and stabs one.’ Aghu (Drabbe 1957:58)

(60) Nu doü bo-ne-d-e. I sago DUR-eat-REAL-1SG ‘I am eating sago’ Kombai (de Vries 1993:28)

Pisa has a pre-verbal durative marker ri, derived from the position verb ri(mo) ‘to lie down’. Thus ri mari means ‘to descend over an extended period of time’ and ri n˜ı means ‘to eat lots, to keep eating’ (Drabbe 1950:126). Yonggom Wambon and Digul Wambon have a durative marker that occurs after the verb rather than before it, as in the other Awyu-Dumut languages. The Yong- gom Wambon post-verbal durative marker mbon (61) and Digul Wambon durative marker mbo (62) are both derived from the verb mba ‘to sit’.

(61) Kagup te, raramun de, munotit t’ i-ŋgin-in mbon men COORD women COORD children COORD eat-REAL-NON1SG DUR ‘Men and women and children are eating’ Yonggom Wambon (Drabbe 1959:145)

(62) Yu si-knd-ep-o3 String.bag make-REAL-1SG-DUR ‘I am making a string bag’ Digul Wambon (de Vries & Wiersma 1992:33)

So far, Shiaxa has not been discussed, although it does also have a position verb with an aspectual function. Drabbe(1950:125) notes that baxamo in Shiaxa encodes habitual meaning. Emed-baxemo, for example, means ‘to be used to doing’. Com- rie explains how habitual meaning can be expressed by a position verb or another element which expresses location:

Of the languages examined where habitual meaning is expressed by means of a locative, it is always the case that progressive meaning is also expressed as a locative, indeed it is usually the case that the same locative construction is used for both meanings [...] the locative expression of progressive meaning is basic, and only if a language has this possibility

3In this Digul Wambon example, the morphophonemic rule /p/+/mb/=p applies, and thus the durative marker -mbo is expressed as -o in this example. Note that the past tense marker in Digul Wambon is also -mbo. However, the past tense marker only attaches to realis -t forms, not to realis -kend forms. 120 8. Aspect

can it further extend the same form to habitual meaning, and this extension is rather an extension of the earlier progressive to become the only imperfective form. (Comrie 1976:103)

Thus Shiaxa baxemo is hypothesized to have, at one point, expressed a durative meaning as well. Indeed, it might be the case that baxemo still has a durative reading in Shiaxa in addition to its habitual reading, but that Drabbe did not note it in his short grammar sketch. The use of a position verb to express both durative and habitual meaning is not found in other Awyu-Dumut languages. Rather the habitual tends to be formed with iterative verb stems in other Awyu-Dumut languages, as will be illustrated in the next section. To summarize, there are two ways in which durative meaning is expressed in most Awyu-Dumut languages:

1. a conjugated form of the position verb mba- ’to sit’ (and in Awyu also i- or ri- ‘to lie down’) following a non-ﬁnite SS verb form 2. the use of a separate durative marker before or after the verb; the durative marker is always derived from a position verb.

Table 8.1 contains the durative markers found in each Awyu-Dumut language, as well as the stems of the position verbs that can express durative meaning.4

Table 8.1: Durative markers and verbs in Awyu-Dumut languages

durative marker position verb expressing durative meaning

YWB ...mbon mba/mbage to sit DWB -mbo mba to sit MAN mbe... -

AXU bu..., i... ba to sit i to lie down, -e to stand PSA ri... ba to sit , ri to lie down KOM boba to sit

The durative markers are derived from position verbs, most often from the position verb mba- ‘to sit, stay’. In many languages, there is a similarity between durative aspect and position adverbs or verbs (Comrie 1976:98-102). Examples in Dutch are ik ben aan het schrijven, literally ‘I am at the writing’, where aan is a preposition, and hij staat kofﬁe te drinken, literally ‘he stands coffee to drink’, where a position verb ‘to stand’ combines with the inﬁnitive form of the verb ‘to drink’. Thus, although all Awyu-Dumut languages have position verbs that express durative meaning, this similarity may not be best explained by a common linguistic history but rather by general typological tendencies in languages. Nevertheless, the forms of the position verbs are clearly cognate, and clearly stem from a shared linguistic ancestor.

4The [...] preceding or following the durative markers in Table 8.1 indicate whether the durative marker precedes or follows the verb. 8.2. Iterative Verb Stems 121

8.2 Iterative Verb Stems

Iterative or repeated actions can express a habit, an action that occurs regularly over an extended period of time. Habitual aspect might express an action that is repeated regularly, as in The bird used to chirp at ﬁve every morning, or a general truth that held for a bounded period of time, such as The city of Oslo used to be called Christiania. In Awyu-Dumut languages, iterative verb stems are used to express both repeated and general habitual aspect. Iterative verb stems are formed in all Awyu-Dumut languages through partial or full reduplication of the basic or suppletive verb stem. In most cases, the auxiliary verb mo ‘to do’ is then added to the reduplicated stem. Table 8.2 contains some iterative verb stems from each Awyu-Dumut language.

Table 8.2: Awyu-Dumut iterative verb stems

verb stem meaning

YWB ti-/titimo- to build/to build repeatedly mbikmo-/mbikgimbikmo- to give a shot/to give multiple shots ip-/iwipmo- to twist rope/to twist rope continuously DWB ndat-/ndatndatmo- to listen/to listen usually kunomo-/kunomomo- to kiss/to kiss repeatedly/habitually MAN rü-/rürüomo- to go down/to go down several times karemo-/kageremo- to treat well/to always treat well tömo-/tötömo- to pull X from ground/to pull several X from ground AXU xo-/xoxomo- to go/to go repeatedly ﬁo-/ﬁogomo- to bake/to bake repeatedly or thoroughly poni-/poponi- to burst/to burst over a period of time PSA n˜ı-/nin˜ı- to eat/to eat repeatedly kit˜ı-/kikitimo- to plant/to plant repeatedly SHI ri-/ririmi- to call/to call repeatedly mode-/modedeme- to come/to come repeatedly wako-/wakowakomo- to have pity/to have pity repeatedly KOM xaxe-/xaxexaxema- to listen/to be obedient ne-/nenema- to eat/to eat repeatedly or usually

Sources: Yonggom Wambon (Drabbe 1959:126-127); Digul Wambon (de Vries & Wiersma 1992:15); Mandobo (Drabbe 1959:34,57,86); Aghu (Drabbe 1957:22); Pisa (Drabbe 1950:125); Shiaxa (Drabbe 1950:126); Kombai (de Vries 1993:15)

In Digul Wambon (63) and Kombai (64), the use of a verb formed with an iterative verb stem implies habitual meaning.

(63) Evo-n-amil=e nexo-ni-n=o luk that-tr.nasal-child=TOP his-mother-tr.nasal=CONN word ndatndat-mo-xe. listen.IT-do-REAL[NON1SG] ‘That child usually obeys his mother.’ Digul Wambon (de Vries & Wiersma 1992:33) 122 8. Aspect

(64) Ya doü nene-ma-no. They sago eat.IT-do-[REAL]NON1PL ‘They usually eat sago/they are sago-eaters.’ Kombai (de Vries 1993:16)

In Yonggom Wambon and in the Awyu languages, besides an iterative stem, additional morphology is needed to form a habitual construction. In Yonggom Wambon, the habitual marker -op5 is added to the iterative stem before mo-, and this construction is followed by a conjugated form of the verb mba- ‘to sit, stay’, so that

(65) kogo-y-op mo na-mbon go.IT-LIG-HAB do IMP-sit Yonggom Wambon (Drabbe 1959:142) means ‘continue to always go (there)’, formed with the iterative stem of ko- ‘to go’ and the imperative form of mba- ‘to sit, stay’. A Yonggom Wambon habitual construction with an iterative, reduplicated verb stem is exempliﬁed in (66).

(66) Ndun e yugup ra me enene-y-op-mo mbage-t. ...sago TOP he take come eat.IT-LIG-HAB-do sit-REAL[NON1SG] ‘He takes the sago, comes (home) and always he (is the one who) eats it.’ Yonggom Wambon (Drabbe 1959:154-155)

Awyu languages do not have a habitual marker like Yonggom Wambon. Rather, they make use of several other strategies to form habituals. In Aghu, the support verb -sumo is added to iterative stems in order to yield a form with habitual meaning. Thus Aghu xoxomoxe means ‘he kept going away’, while xoxomsumoxe means ‘he habitually went away’ (Drabbe 1957:23). For Shiaxa, Drabbe notes about iterative verb stems that “many such derived verb stems can be used to express a habitual process also.”6 Drabbe does not state that Pisa iterative verb stems express a habitual process. He does state that they imply an ongoing process, or at times that either the subject or the object of the iterative verb is plural. That an iterative verb indicates a plural subject is also found in Kombai; in (67) the iterative verb gege ‘to inspect’ has a plural subject, namely multiple males involved in ‘inspecting’ a village ﬁght (de Vries 1993:28).

(67) Yademo-ra xuro ugino-no˜ lu mene come.together-CONN each.other hit-[REAL]NON1PL quarrel this gege-mo-ra irabi irabi rimo-fo-nene. inspect.IT-do-CONN hand hand shake-[IRR]1PL-QUOTE.PL ‘They said: ‘let us have a meeting and then discuss this quarrel that they want to hit each other, and then make peace.’ Kombai (de Vries 1993:27)

5The Yonggom Wambon habitual marker -op can also occur with other verb forms to yield a habitual meaning; Drabbe does not describe an iterative stem as being required in the Yonggom Wambon habitual construction (Drabbe 1959:141-142). 6“Veel van de aldus afgeleide secundaire stammen gebruikt men ook wel om een habitueel proces aan te duiden” (Drabbe 1950:127). 8.3. Connective Verbs 123

In summary, it can be said that all Awyu-Dumut languages except Pisa use iterative verb stems to express habitual meaning but that an Awyu-Dumut iterative verb stem need not necessarily express a habitual process.

8.3 Connective Verbs

Connective verbs are Awyu-Dumut verbs that have a specialized function in connecting clauses. Certain connective verbs imply aspectual meaning, either completive meaning or bounded continuative meaning. Although examples of connective verbs in other Papuan languages can be cited,7 a detailed study into the typology of connective verbs in Papuan languages is lacking. Two categories of connective verbs exist in Awyu-Dumut languages: (1) verbs with meaning ‘to finish’ and (2) motion verbs. The connective motion verbs can be translated as ‘until’ and imply that the verb preceding the motion verb has a prolonged duration, which ends when the action expressed by the verb following the motion verb begins. The connective ‘finish’ verbs express a completive aspectual meaning. It is not always easy to determine whether a connective verb still functions as a verb or has grammaticalized into a connective (de Vries 1986:48). Most often the connective verb appears in 3SG form and therefore seems petrified, as in (69) below, but at times connective verbs can also be conjugated in 1SG or any other form, as in (74) below. I therefore gloss connective verbs as conjugated verbs; they are used periphrastically. Table 8.3 contains the motion verbs and ‘finish’ verbs that are used as connective verbs expressing aspectual meaning in five Awyu-Dumut languages.

Table 8.3: Awyu-Dumut connective verbs

motion meaning ﬁnish verb meaning verb

YWB ko to go mbumo, ndoimo, oro to finish, to not do, to put down DWB ko to go kit(mo) to finish MAN - etamo, ndamo to finish, to not do AXU - eme, buomo, mu˜ to do thus, to continue, to not want KOM xa to go lei to lie down

Connective motion verbs were found in Yonggom Wambon, Digul Wambon and Kombai grammars, but were not attested in the other Awyu-Dumut languages. Ex- amples of how the verb ‘to go’ can express the aspectual meaning ‘until’ in these three languages is illustrated in Section 8.3.1. The connective ‘ﬁnish’ verbs found in ﬁve different Awyu-Dumut languages are discussed in Section 8.3.2.

7For example, Reesink(1987:83-84) notes that several Usan verbs act as connective verbs. 124 8. Aspect

8.3.1 Connective motion verb ‘to go’ In Yonggom Wambon, Digul Wambon and Kombai, the verb ku/ko/xa ‘to go’ can function as a connective verb with meaning ‘until’. In Yonggom Wambon, the most common form of this connective verb is kura, which is a NON1SG realis form with a sequence marker -a attached (68). In Digul Wambon, koxeno is the most commonly used form of the connective verb ko ‘to go’ (69). In Kombai, the motion verb xa combines with a marker nege ‘until’ when functioning as a connective verb expressing aspectual meaning (70). (68) Ngop tare-r-in ku-r-a arrow scrape-REAL-NON1PL go-REAL[NON1SG]-SEQ wagae-ke-r-an. smooth-be-REAL[NON1SG]-PST ‘They scraped an arrow until it was smooth.’ Yonggom Wambon (Drabbe 1959:135) (69) Nda-knd-eva-n-o ko-xe-n-o come-REAL-1PL-tr.nasal-CONN go-REAL[NON1SG]-tr.nasal-CONN kutipke-lo Kui mata-l-eva-mbo. dark.become-NON.CLOSE Kouh arrive-REAL-1PL-PST ‘We traveled until the night fell and arrived in Kouh’ Digul Wambon (de Vries 1986:48) (70) Dunorof-a ox-a b-adiya-non-a xa-nege food-CONN drink-CONN DUR-give-NON1PL-CONN go-until fa-n-a gone-n-a... put.aside-NON1SG-CONN overfull.be-NON1SG-CONN... ‘They gave him food and drink until he put it aside (because) he was overfull...’ Kombai (de Vries 1993:127)

8.3.2 Connective verb ‘to finish’ When a verb with the meaning ‘to finish’ or ‘to stop’ follows another verb in Awyu- Dumut languages, it indicates that the action expressed by the first verb has ended. The ‘finish’ verb might be conjugated as a non-finite, semi-finite or finite verb, and may follow non-finite, semi-finite or finite verbs. The ‘finish’ verb acts as a connective verb, indicating that the action in a second clause takes place after the action in a first clause; ‘finish’ verbs, like connective motion verbs, have not been attested sentence-finally. A sentence that contains a ‘finish’ verb can be translated as ‘after (subject) finished X, Y happened’. Drabbe describes connective finish verbs as emphasizing anteriority (Drabbe 1959:136), and de Vries(1993:29) notes that the Kombai verb lei ‘to lie down’ expresses both completion and posteriority of the next event, as in (71). In the Man- dobo and Yonggom Wambon verbs ndamo and ndoimo the negative element nda/ndoi is present, and these verbs literally mean ‘to not do’. When they function as connective verbs they express completed action, as in (72). 8.3. Connective Verbs 125

(71) Gana fa-d-ef-a lei-n-a bush.knife take-REAL-1SG-CONN lie-[REAL]NON1SG-CONN xi-d-ef-a... run-REAL-1SG-CONN... ‘After I had taken the bush knife, I ran and...’ Kombai (de Vries 1993:29) (72) Kou mba-gen do, noŋgun-irandop u meto ane just sit-REAL[NON1SG] CONN 1PL.POSS-tuber pig TOP eat ndamo-gen ne-gen-on. not.do-REAL[NON1SG] say-REAL-NON1PL ‘While you just sit there, that pig is eating completely (all) our tubers’ they said.’ Mandobo (Drabbe 1959:26) In (73) and (74) the other Yonggom Wambon and Mandobo verbs for ‘to finish’ are used, namely etamo and mbumo. (73) Roa etamo-gen-ep to togümo-r-an. work finish-REAL-1SG CONN pay-REAL[NON1SG]-PST ‘After I finished working he paid (me)’ Mandobo (Drabbe 1959:101) (74) Ande-r-ew-a mbumo-gon-ep te me-gen. eat-REAL-1SG-SEQ finish-REAL-1SG CONN come-REAL[NON1SG] ‘After I finished eating, he came.’ Yonggom Wambon (Drabbe 1959:136) Example (75) contains multiple instances of the Digul Wambon verb kit(mo) ‘to finish’,8 while (76) illustrates the use of connective verb eme ‘to do thus’ in Aghu. (75) Ndano la-l-eva-mbo-n=o yat-ke-lo next sleep-REAL-1PL-PST-tr.nasal=CONN light-be-NON.CLOSE wesat-ke-lo kono enov-andil=e li-no day-be-NON.CLOSE and tree-trunk=TOP cut-SIM nda-ndaxa-mo-knd-eva-n=o kit-mo-mbel=o odo sini=ŋga put-put-do-REAL-1PL-tr.nasal=CONN finish-do-SEQ=CONN and stick=ERG uxumo ndatkap-mo lap-ko-tulo mbait=ka put.under lift.up-do take-go-ascend hill=ERG halo-knd-eva-n=o kit-mbel=o epka lavo put.down-REAL-1PL-tr.nasal=CONN finish-SEQ=CONN there bring ŋguliŋge-no lap-ko hala-l-eva-mbo. roll-SIM take-go put.down-REAL-1PL-PST ‘Next we slept until daybreak and after we finished cutting down some tree trunks, we put sticks under them and we lifted the sticks and brought the tree trunks uphill and after we finished putting them down, we rolled them (from) there and put them down.’ Digul Wambon (de Vries & Wiersma 1992:87-88) 8In Digul Wambon, kit means ‘enough’ or ‘full’ (de Vries, p.c.) 126 8. Aspect

(76) Xasu-ax˜ı si-d eme-de-k widi küto-d-oã. leaf-hut build-SS ﬁnish-SS-CONN river descend-REAL-1PL ‘After we ﬁnished building the leaf hut, we went down to the river.’ Aghu (Drabbe 1957:36)

8.4 Reconstruction

This chapter looked at how Awyu-Dumut languages express durative, habitual and completive aspects. Durative meaning is expressed in all Awyu-Dumut languages except Shiaxa by a conjugated position verb and/or a durative marker derived from the position verb mba ‘to sit/stay’. The use of position verbs to express durative meaning is common in the languages of the world (Comrie 1976). Concerning connective verbs, it can be said that the indication of completed action lies within the semantics of the verb ‘to ﬁnish’, while the verb ‘to go’ implies moving, ongoing action. Similarly, a habitual reading lies within the semantics of iterative verbs; any action that is repeated several times might become a habit. Therefore the fact that position verbs, motion verbs and connective verbs have aspectual meanings in all Awyu-Dumut languages cannot be used to proof that these languages are genealogically related, as all Awyu-Dumut languages might have developed the aspectual meanings independently from the same verbs. In other words, the similarities in expression of aspectual meaning in Awyu-Dumut languages might be reﬂections of general typological tendencies. Nevertheless, as it has become clear in other parts of the morphology that Awyu-Dumut languages are indeed related, it is not unlikely that these changes and broadening of meaning of position, location and connective verbs already took place in Proto Awyu-Dumut, and hence were inherited by Proto Awyu, Proto Dumut and their daughter languages. 9

Negation

This chapter explores the development of double negation in Awyu languages, as well as the abundance of negation strategies found in Dumut languages. In Proto Awyu, negated verbal clauses were emphasized, and the emphatic marker turned into a second negation marker, resulting in double negation in Proto Awyu and Awyu languages. Dumut languages do not have double negation; they employ a number of other negation strategies. Section 9.1 offers a description of negation in Awyu languages, culminating with a description in Section 9.1.4 of how the Jes- persen cycle is visible in Awyu negation. The ways in which various negation strategies function in Dumut languages is shown in Section 9.2. The negation constructions of Kombai (Section 9.3), as well as of Korowai and Tsaukwambo, will shed light on the reconstruction of Proto Awyu-Dumut negation in Section 9.4.

9.1 Awyu Negation

All Awyu languages have double verbal negation, but each Awyu language has a slightly different double negation construction. A diachronic path in the development of double negation in Awyu languages can be traced, and will be discussed in Section 9.1.4. The descriptions of negation in the Awyu grammars are much less detailed than those in Dumut grammars. For one Awyu language, Yenimu, no information at all is available on how its nouns and verbs are negated. Descriptions of negation strategies in Aghu, Pisa and Shiaxa are given in the following paragraphs.

9.1.1 Aghu negation Aghu has double and sometimes triple negation. The three negative elements are fede, de and oxo. The element oxo is a copula, or as Drabbe calls it, a predicative element, and in negative constructions is glossed as NEG.COP to show that it functions more as a third negator than as a copula. Besides the negator de, there is also an emphatic marker de with copular function in Aghu, but Drabbe warns his readers that the emphatic de should not be confused with the negative de (Drabbe 1957:8). How- 128 9. Negation ever, given that according to Croft(1991:5) negative markers often derive from emphatic markers, it is not unlikely that diachronically the negator de derived from emphatic marker de. In the following section, non-verbal predicate negation in Aghu is examined ﬁrst before Aghu verbal negation is described. non-verbal predicate negation in Aghu Nominal predicates are negated in Aghu by placing de oxo after the noun (77 and 78). The morpheme oxo always shortens to xo if it follows a vowel-ﬁnal morpheme such as de.

(77) Nu de xo. I NEG NEG.COP ‘It was not me.’ Aghu (Drabbe 1957:9)

(78) Xofe te Eba n’ axu de xo. man TOP Eba.river POSS human.being NEG NEG.COP ‘That man is not a man from the Eba river.’ Aghu (Drabbe 1957:9)

Existential clauses are negated not with de oxo but with fede oxo (79 and 80).

(79) Büshü tadiga fede xo. house big NEG NEG.COP ‘There are no big houses.’ Aghu (Drabbe 1957:9)

(80) Dü fede xo. sago NEG NEG.COP ‘There is no sago.’ Aghu (Drabbe 1957:9) verbal predicate negation in Aghu Aghu semi-finite realis verbs, finite past tense verbs and finite future tense verbs are all negated by de oxo following the verb, while a negator fede may optionally precede the verb. Table 9.1 contains a negated semi-finite realis paradigm, a negated distant past tense paradigm and a negated future tense paradigm. In negated verb forms, Drabbe writes the negator de attached to the verb, whereas when de follows a noun, it is written separately; Drabbe’s notation is followed here. Note that the Aghu future tense marker -e is not present in the negated future tense forms. Hence the negated future tense paradigm looks like a negated semi- finite irrealis paradigm. However, the semi-finite irrealis verb forms are negated by adding kuoxo to the verb, resulting in a prohibitive (81) or a negative wish (82).

(81) Amse xo büshü aghinang-guoxo. children TOP house go.II[NON1SG]-NEG ‘Do not let the children go home.’ Aghu (Drabbe 1957:14) 9.1. Awyu Negation 129

Table 9.1: Aghu negated paradigms

semi-ﬁnite realis future

1SG (fede) da-d-e-de xo (fede) ade-de xo (NEG) come.I-REAL.1-1SG-NEGNEG .COP (NEG) come.II[1SG]-NEGNEG .COP I did not come recently I will not come NON1SG (fede) da-x-e-de xo (fede) ade-n-de xo (NEG) come.I-REAL.NON1-NON1SG-NEGNEG .COP (NEG) come.II-NON1SG-NEGNEG .COP he/she/it did not come recently he/she/it will not come 1PL (fede) da-d-oa-de xo (fede) ad-oan-de xo (NEG) come.I-REAL.1-1PL-NEGNEG .COP (NEG) come.II-1PL-NEGNEG .COP we did not come recently we will not come NON1PL (fede) da-x-enan-de xo (fede) ad-enan-de xo (NEG) come.I-REAL.NON1-NON1PL-NEGNEG .COP (NEG) come.II-NON1PL-NEGNEG .COP you(pl)/they did not come recently you(pl)/they will not come

distant past historical past

1SG (fede) da-d-k-e-de xo (fede) da-dia-de xo (NEG) come.I-REAL.1-DIST.PST-1SG-NEGNEG .COP (NEG) come.I-HIST[1SG]-NEGNEG .COP I did not come a while ago I did not come a long time ago NON1SG (fede) da-ak-i-de xo (fede) da-dia-de xo (NEG) come.I-DIST.PST-NON1SG-NEGNEG .COP (NEG) come.I-HIST[NON1SG]-NEGNEG .COP he/she/it did not come a while ago he/she/it did not come a long time ago 1PL (fede) da-d-k-oan-de xo (fede) da-dia-oan-de xo (NEG) come.I-REAL.1-DIST.PST-1PL-NEGNEG .COP (NEG) come.I-HIST-1PL-NEGNEG .COP we did not come a while ago we did not come a long time ago NON1PL (fede) da-ak-enan-de xo (fede) da-dia-nan-de xo (NEG) come.I-DIST.PST-NON1PL-NEGNEG .COP (NEG) come.I-HIST-NON1PL-NEGNEG .COP you(pl)/they did not come a while ago you(pl)/they did not come a long time ago

Source: Drabbe(1957:17)

(82) Nu atosunu akume-kuoxo. me care.for die.II[1SG]-NEG ‘Care for me lest I die.’ Aghu (Drabbe 1957:14)

A negated future tense verb form (83) has a negative declarative meaning.

(83) ... axu fede baxe-nan-de xo...... people NEG stay.II-NON1PL-NEG NEG.COP... ‘...no people will stay (here)...’ Aghu (Drabbe 1957:80)

9.1.2 Pisa negation Pisa has double negation, placing a negative element fa or fana before the verb and a negator de after the verb. Both negators are obligatory. The element de also appears in Pisa as an interrogative marker and functions as an emphatic marker with copular meaning as well. As in Aghu, the negator de might have its origin in the emphatic 130 9. Negation de. Table 9.2 contains negated semi-ﬁnite paradigms, while Table 9.3 contains Pisa ﬁnite past negated paradigms.1

Table 9.2: Pisa negated semi-ﬁnite paradigms

realis semi-ﬁnite irrealis semi-ﬁnite

1SG fa(na) wu-d-i de fa(na) awu de NEG enter.I-REAL-1SGNEGNEG enter.II[1SG] NEG NON1SG fa(na) wu-k-i de fa(na) awu˜ de NEG enter.I-REAL-NON1SGNEGNEG enter.II.NON1SGNEG 1PL fa(na) wu-d-a de fa(na) awu-a de NEG enter.I-REAL-1PLNEGNEG enter.II-1PLNEG NON1PL fa(na) wu-k-inã de fa(na) awu-nã de NEG enter.I-REAL-NON1PLNEGNEG enter.II-NON1PLNEG

Source: Drabbe(1947:21)

Table 9.3: Pisa negated ﬁnite paradigms

hesternal past distant past

1SG fa(na) wu-di-ra de fa(na) wu-d-axa-ri de NEG enter.I-REAL-1SG-HESTNEGNEG enter.I-REAL-1SGNEG NON1SG fa(na) wu-k-i-ra de fa(na) wu-ak-i de NEG enter.I-REAL-NON1SG-HESTNEGNEG enter.I-DIST.PST-NON1SGNEG 1PL fa(na) wu-d-a-ra2 de fa(na) wu-d-axa-ya de NEG enter.I-REAL-1PL-HESTNEGNEG enter-REAL-1PLNEG NON1PL fa(na) wu-k-ina-ra de fa(na) wu-ak-inã de NEG enter-REAL-NON1PL-HESTNEGNEG enter.I-DIST-NON1PLNEG

Source: Drabbe(1947:21)

9.1.3 Shiaxa negation Shiaxa negative forms are made by always adding fa or fana in front of the verb and optionally adding de after the verb. Drabbe gives a total of two negated examples in Shiaxa, one in the future tense (84) and one using a semi-ﬁnite realis form with present meaning (85).

(84) Fan agoxo-ne (de). NEG go.II-NON1SG (NEG) ‘He will not go.’ Shiaxa (Drabbe 1950:118)

1Recall from Chapter7 on Tense that Pisa does not have a separate future tense; it expresses future meaning using its semi-ﬁnite irrealis verb forms. 2Drabbe(1947:21) lists wudaxaya as the hesternal 1PL form, but it is the distant past 1PL form. The correct 1PL hesternal form is wudara. 9.1. Awyu Negation 131

(85) Fana fete-d. NEG see-REAL[NON1SG] ‘I do not see (it).’ Shiaxa (Drabbe 1950:118)

9.1.4 Reconstruction of Proto Awyu negation Aghu, Pisa and Shiaxa all have double negation. The diachronic development of double negation has been studied and is often explained in terms of the Jespersen cycle. In 1917, Jespersen noted explained the cycle as follows:

The history of negative expressions in various languages makes us wit- ness the following curious fluctuation: the original negative adverb is first weakened, then found insufficient and therefore strengthened, generally through some additional word, and this in its turn may be felt as the negative proper and may then in course of time be subject to the same development as the original word. (Jespersen 1917:4)

In other words, there is an original pre-verbal negator whose meaning bleaches and which is then strengthened by a new post-verbal negator, resulting in double negation. Eventually the first negator will pass out of use and only the final negator will remain. At different stages of this cycle either negator might be optional or obligatory. The term ‘Jespersen cycle’ was coined by Dahl, who stated that “Since we are dealing with a cyclical process in the sense that we go from a single particle to a double and back again, we may refer to this kind of development as ‘Jespersen’s Cycle’ ” (Dahl 1979:88). van der Auwera(2009:38) gives a summary of the Jespersen cycle in five stages using the French negators ne and pas:

1. neNEG

2. neNEG (...pasNEG)

3. neNEG ... pasNEG

4.( neNEG...) pasNEG

5. pasNEG

The cycle begins with a single negator (step 1), which is then strengthened by a second negator. The second negator is optional at ﬁrst (step 2) but later becomes obligatory (step 3). The next step is for the ﬁrst negator to start fading out of use (step 4), eventually leaving the second negator as the sole negator (step 5). The source of the second negator is often an emphatic marker. In his article entitled The Evolution of Negation, Croft notes that negators often originate in emphatic particles:

The primary source of verbal negation that has been observed in prior research is via the employment of emphatic particles. This has been noted, for example, through direct historical evidence in the evolution of the 132 9. Negation

French emphatic forms pas, point, etc. in addition to the original negator ne (Meillet 1921:140). Eventually, the original negator is fused with the emphatic marker, or, as in modern colloquial French, drops out. (Croft 1991:5)

The second negator de in Awyu languages has its source in the emphatic marker de; de is an emphatic marker with copular function in Aghu, Pisa and Shiaxa. For Aghu, Drabbe notes that the negator de should not be confused with the predicating, emphatic de (Drabbe 1957:8), which indicates that in Aghu emphatic de has fully grammaticalized as a negative element. In Pisa, emphatic de (allomorph di) can follow a noun (86) or a verb (87). In fact, Drabbe notes that emphatic de can also follow a negative de (88) in Pisa, showing that emphatic de and negative de are two distinct elements in Pisa as well.

(86) Na nabo di. 1SG.POSS father EMPH ‘It is my father.’ Pisa (Drabbe 1950:98)

(87) Emo-xoy de? do.I-REAL.NON1SG EMPH Did he do it? Pisa (Drabbe 1950:98)

(88) Fa emo-xoy de di. NEG do.I-REAL.NON1SG NEG EMPH Indeed he did not do it. Pisa (Drabbe 1950:98)

The fact that the origin of the negator de can be traced to an element that also retains its original function of emphatic marker, provides an additional indication that de is a younger negator than pre-verbal fa(na). The Jespersen cycle is represented in Awyu languages in its various stages. Shi- axa represents the oldest phase of double negation development in Awyu languages, where the pre-verbal negator is still obligatory but where a second post-verbal negator has already entered the scene and occurs optionally after the verb. Pisa represents the next stage, where both negators are obligatory. Aghu has gone even further, making the pre-verbal negator optional and strengthening the post-verbal negator de with a new copular morpheme, namely oxo.

For Proto Awyu, *fa/*feNEG is reconstructed as an obligatory pre-verbal negator. Aghu then added an element de to the Proto Awyu preverbal *fe, while Pisa and Shiaxa (optionally) added an element -na to Proto Awyu *fa; the origin of these two elements cannot be traced. An optional post-verbal negator *deNEG is reconstructed for Proto Awyu as well; this *de also functioned as an emphatic marker in Proto Awyu. The Jespersen cycle predicts that, at one point in time, *fa/*fe was the sole negator, but because it is not attested as a sole negator in any daughter Awyu languages, it is not reconstructed as such. 9.2. Dumut Negation 133

9.2 Dumut Negation

The three Dumut languages each employ a number of negation strategies, although none of them make use of double negation in the way Awyu languages do. The following sections describe how nominals, non-finite verbs, semi-finite verbs and finite verbs are negated in Mandobo (Section 9.2.1), Yonggom Wambon (Section 9.2.2) and Digul Wambon (Section 9.2.3). In Section 9.2.4, a Proto Dumut negator is reconstructed. For the reader’s convenience, an overview of the Dumut negation strategies is presented in Table 9.4.

Table 9.4: Dumut negation strategies

9.2.1 Mandobo negation The negator in Mandobo is nda. It can occur after a noun, for example after kerewatop ‘face’ (89) or ok ‘water’ (90). (89) Ro itigio-gen do, kerewatop nda. form(SS) see-REAL[non1SG] CONN, face NEG ‘He looks and they do not have faces.’ Mandobo (Drabbe 1959:11) (90) Ok mbo nda taro kater e ndu ri-gin-on. water TOP NEG because spit TOP sago stand-REAL-NON1PL. ‘There is no water, therefore they stand (and make) sago with (their) spit.’ Mandobo (Drabbe 1959:39)

The Mandobo negator nda may also follow a non-ﬁnite verb form. A negated non- ﬁnite form in Mandobo is always followed by an imperative verb form in the next clause (91). (91) ŋgo mbo kea nda, ok tiritiop mberemo tarap ti mbegi-ro you TOP go[IRR] NEG, river high.wall stay hut build live-NON.CLOSE mbutup ti-nok. house.on.poles build-IMP. ‘Do not go, stay and build a hut on the river’s high wall, build a house on poles!’ Mandobo (Drabbe 1959:54) 134 9. Negation

The connective -o may attach to the negator nda, yielding ndayo, as occurs in (92).

(92) No mandep itiwa to komo-nda-y-o, nomandep wonden doro My son earth dig put.in-NEG-LIG-CONN, my.son good CONN mbutüp korup teendükmo-nog-i-nin ne-gen. house inside put.straight.up-IMP-LIG-NON1PL say-REAL[NON1SG] ‘Do not put my son in earth, put my son (standing) straight up in a good house, she says.’ Mandobo (Drabbe 1959:82)

Thus, Mandobo negative non-finites only occur in sentences that also contain an imperative. Mandobo negated non-finites can be interpreted as a prohibitive or as a negative imperative, and sentences that begin with a negated non-finite form mean something to the effect of ‘don’t do X, do Y!’ Mandobo semi-finite realis -ken forms are negated by adding an element ne after the verb stem and the negator nda after the person-number marker. The element ne consists of transitional nasal /n/ and the support verb -e, which Drabbe mentions is a support verb that means the same thing as ke-, namely ‘to be’ (Drabbe 1959:28).3 Example (93) contains a negated semi-finite realis -ken form.

(93) Matero engot keemo itigio-gen do yombutop Get.up torch light see-REAL[NON1SG] CONN door tokmo-n-e-gen-nda te ran-gen. open-tr.nasal-SUPP-REAL[NON1SG]-NEG CONN lie-REAL[NON1SG] ‘She gets up, lights a torch and sees that the door is not open and lies back down.’ Mandobo (Drabbe 1959:36)

Drabbe does not explain how past tense forms are negated in Mandobo. For future tense forms, however, he notes that the verb stem is followed by an element -notü or -notürü, which is then followed by a person-number marker and negator nda. According to Drabbe, -notü/-notürü might come from notüt ü, although he is unable to give the meaning of notüt ü (Drabbe 1959:38). The Mandobo future tense marker -en does not occur in these forms.

9.2.2 Yonggom Wambon negation The most used negator in Yonggom Wambon is -nok, although a negator tomba (allomorph domba) is used to negate nominal predicates (94), while locative and existential predicates are negated by the negative element ndoi (95, 96). According to Drabbe’s wordlist, ndoi, like Mandobo nda, means ‘nothing’ (Drabbe 1959:178).

(94) Sowen domba. mosquito NEG ‘It is not a mosquito.’ Yonggom Wambon (Drabbe 1959:124)

3It must be noted that Drabbe does not analyze the element ne in negated -ken forms as /n/ plus -e. 9.2. Dumut Negation 135

Table 9.5: Mandobo negated future tense paradigm

SGPL

1 ndare-notü-p-nda ndare-notürü-won-nda hear-?-1SG-NEG hear-?-1PL-NEG I do not want to hear we do not want to hear NON1 ndare-notü-n-nda ndare-notür-on-nda hear-?-NON1SG-NEG hear-?-NON1PL-NEG he/she/it does not want you(PL)/they do not want to hear to hear

Source: Drabbe(1959:38)

(95) Mitik ke-t ye ya magomŋgui mindi-r-in-a, night be-REAL[NON1SG] and her grandchildren come-REAL-NON1PL-SEQ etaga-r-in de ra-mbari e ndoi. see-REAL-NON1PL CONN old-woman TOP NEG ‘It became night and her grandchildren came (home) and they saw that the old woman (their grandmother) was not there.’ Yonggom Wambon (Drabbe 1959:155) (96) Menep e kagup ndoi. now TOP men NEG ‘Now there are no men.’ Yonggom Wambon (Drabbe 1959:147) The Yonggom Wambon negator -nok follows the verb stem in negated semi-finite and finite realis forms. According to Drabbe, adding -nok to a verb stem nominalizes it. Therefore a verbalizing support verb ke ‘to be’ or mo ‘to do’ follows -nok. The support verb is conjugated. Example (97) contains a negated semi-finite realis -ken form, (98) a negated semi-finite realis -t form and (99) a negated past tense form. (97) Rap-nok mo-gon-ep. take-NEG do-REAL-1SG ‘I did not take (it).’ Yonggom Wambon (Drabbe 1959:140) (98) Mitik ke-t te ndari-r-in de roman kup night be-REAL[NON1SG] CONN hear-REAL-NON1PL CONN crying with me-nok ke-t. come-NEG do-REAL[NON1SG] When it has become night, when they listen, (the sound of) crying does not come to them.’ Yonggom Wambon (Drabbe 1959:157) (99) Etok-nok ke-r-ew-an see-NEG be-REAL-1SG-PST ‘I did not see (it).’ Yonggom Wambon (Drabbe 1959:140) The finite future tense form is also negated by placing the negator -nok after the verb stem, followed by the support verb mo ‘to do’ (100). The support verb mo is conjugated with the future -ti suffix. The future tense marker -ti only occurs in negated 136 9. Negation

ﬁnite future forms with negative indicative meaning; the Yonggom Wambon future marker -in does then not occur.

(100) Me-nok mo-ti-p. come-NEG do-FUT-1SG ‘I will not come.’ Yonggom Wambon (Drabbe 1959:140)

9.2.3 Digul Wambon negation In order to negate nominal predicates, Digul Wambon uses the negative element ndiknde (101). In addition, either one of the negators mbumba (102) or tembet (103) is used to negate locative or existential predicates.

(101) Ev-o kav-e na-mbap-ndiknde. that-CONN man-TOP my-father-NEG ‘That man is not my father.’ Digul Wambon (de Vries & Wiersma 1992:37)

(102) Ande-n-e mbumba. food-tr.nasal-TOP NEG ‘There is no food.’ Digul Wambon (de Vries & Wiersma 1992:38)

(103) Ahituv-e av-e tembet. Ahitup-TOP home-TOP NEG ‘Ahitup is not at home.’ Digul Wambon (de Vries & Wiersma 1992:38)

In their paragraph on negation, (de Vries & Wiersma 1992) note that the negator ndoi is sometimes found with non-ﬁnite verb forms, giving (104).

(104) taximo-ndoi buy(SS)-NEG ‘not buying (same subject following)’ Digul Wambon (de Vries & Wiersma 1992:36)

The exact meaning of such a negated non-finite, or its context, remains a mystery in de Vries & Wiersma’s (1992) description; (104) is the only negated non-finite found in their Digul Wambon grammar description. However, Jang’s (2008) description of Digul Wambon sheds light on the matter. He states that a non-finite verb followed by ndoi expresses a past impossibility, with meaning ‘could not’ (Jang 2008:53), as illustrated in (105).

(105) Sit indup=e aphinotopke-l=o Banana seedling=TOP be.dry.season-REAL[NON1SG]=CONN sokmboghe-l=o lo-ndoi. become.dry-REAL[NON1SG]=CONN plant-NEG. ‘As it was dry season, banana seedlings had withered and (people) could not plant (them).’ Digul Wambon (Jang 2008:53) 9.2. Dumut Negation 137

Digul Wambon semi-ﬁnite realis ke(nd) forms are negated by adding the negator -nok, which Digul Wambon shares with Yonggom Wambon, after the verb stem and before the realis marker -ke(nd), as in (106). No support verb occurs in Digul Wambon semi-ﬁnite realis negation.

(106) lap-nok-kend-ep take-NEG-REAL-1SG ‘I do/did not take’ Digul Wambon (de Vries & Wiersma 1992:35)

Although semi-ﬁnite realis -t forms are just as frequent as realis -ke(nd) forms in Digul Wambon, there is no information in Digul Wambon grammar descriptions about how they are negated. Semi-ﬁnite irrealis forms are negated by adding the negator -nok after the verb stem, and a second negator -si follows -nok, while a person-number marker comes at the end of the negated verb form (107). De Vries & Wiersma(1992:35) note that the negator -si might have its origin in the verb soi ‘to refuse’.

(107) e-nok-si-t eat-NEG-NEG-3SG ‘he does not want to eat’ Digul Wambon (de Vries & Wiersma 1992:35)

Example (107) is the only example of a negated semi-ﬁnite irrealis form given by de Vries & Wiersma(1992). Jang(2008) notes that the -si form encodes ‘present impossibility’, as evident in (108).

(108) Kixuw=e nda-ke-ŋgot=o Kolopkam-sixi Digul=TOP come-REAL[NON1SG]-CAUS=CONN Kolopkam-for ko-nok-si-t. go-NEG-NEG-3SG ‘Because the Digul river is ﬂooded, he cannot go to Kolopkam.’ Digul Wambon (Jang 2008:52)

Jang(2008) goes on to give another paradigm of negated semi-finite irrealis forms, which he states express negative wishes or intentions. They consist of a semi-finite irrealis form followed by the negator ndoi. The paradigm is given in Table 9.6. Thus, between de Vries & Wiersma(1992) and Jang(2008), two different ways of negating semi-finite irrealis forms in Digul Wambon are described, either by a combination of negators -nok and -si, or by negator ndoi. In Digul Wambon finite past tense forms, the negator nok follows the verb stem, and a support verb mo ‘to do’ conjugated in past tense follows nok, resulting in a form like the one in (109). In Digul Wambon the support verb mo ‘to do’ occurs in negative constructions, rather than the support verb ke ‘to be’ .

(109) e-nok-ma-l-ep-o4 eat-NEG-do-REAL-1SG-PST ‘I did not eat’ Digul Wambon (de Vries & Wiersma 1992:35) 138 9. Negation

Table 9.6: Digul Wambon negated irrealis semi-ﬁnite paradigm

SGPL

1 atki-p-ndoi atki-wa-ndoi wrap-1SG-NEG wrap-1PL-NEG I do not want to wrap we do not want to wrap NON1 atki-ndoi atki-na-ndoi wrap[NON1SG]-NEG wrap-NON1PL-NEG you/he/she/it does not want to wrap you(pl)/they do not want to wrap

Source: Jang(2008:55)

A negative indicative future action, with meaning ‘shall not’, is expressed in Digul Wambon by placing xet ndoi after a semi-ﬁnite irrealis form consisting of a verb stem and a person-number marker. The element xet is analyzed as a negative marker by de Vries & Wiersma(1992:35) but is better analyzed as a (petriﬁed) form of the support verb ke ‘to be’. A literal translation of taximo-xet-ndoi is then ‘he shall buy do it not’, freely translating as ‘he shall not buy it’. A paradigm of negative indicative future forms is given in Table 9.7. Note that in the 1SG form, xet does not occur; rather, the 1SG person-number marker -ep is repeated.

Table 9.7: Digul Wambon negated irrealis semi-ﬁnite

SGPL

1 taximo-ep-ep-ndoi taximo-eva-xet-ndoi buy-1SG-1SG-NEG buy-1PL-SUPP-NEG I shall not buy we shall not buy NON1 taximo-xet-ndoi taximo-na-xet-ndoi buy[NON1SG]-SUPP-NEG buy-NON1PL-SUPP-NEG you/he/she/it shall not buy you(pl)/they shall not buy

Source: de Vries & Wiersma(1992:36)

9.2.4 Summary and reconstruction of Dumut negation A plethora of negation strategies are used by the three Dumut languages, with only some forms being the same across multiple Dumut languages. Table 9.8 provides an overview of the various negation strategies Dumut languages employ. Yonggom Wambon and Digul Wambon stand out together because they have a negator nok that does not occur in any other Awyu-Dumut language. Drabbe describes nok as a nominalizer, which is then followed by a support verb with verbalizing properties. The nok negation strategy is a shared innovation of Digul Wambon and Yonggom Wambon, not reconstructible for Proto Dumut. Rather, *nda/*ndoi is

4Due to morphophonemic rules, the /p/ of 1SG-ep and the /mb/ of past tense marker -mbo combine to yield /p/. 9.3. Kombai Negation 139

Table 9.8: Dumut negation strategies

Mandobo Yonggom Wambon Digul Wambon nominals nda tomba ndiknde locatives/existentials – ndoi mbumba, tembet non-finites nda + IMP – ndoi semi-finite realis ne...nda nok supp.verb nok semi-finite irrealis – – -nok+-si, ndoi finite realis – nok+supp.verb nok+supp.verb finite irrealis notü...nda -nok supp.verb+-ti xe-t ndoi reconstructed as the Proto Dumut negator. This negator, which is a grammaticalized form of a lexical item meaning “nothing” in both Yonggom Wambon and Mandobo, is found in all three Dumut languages. In Yonggom Wambon it occurs with locative and existential predicates, in Digul Wambon with non-finites, negated irrealis semi- finites and in negative future indicative forms. In Mandobo the negator nda occurs in all negation constructions. The exact construction in which *nda/*ndoi occurred in Proto Dumut cannot be reconstructed, except that the negator occurred after the constituent it negated.

9.3 Kombai Negation

Kombai, like Awyu languages, has double negation. In negated expressions that encode present or past meaning, the negative element fe- optionally comes before the verb, while the negative element -do is obligatory after the verb. The mood marker and person-number markers found in afﬁrmative semi-ﬁnite realis forms (110) are not found in their negated counterparts. Rather, the two negators appear around a verb stem (111 and 112). Hence the negated verb form is the same for all person-number conjugations in Kombai.

(110) Nu ai fera-d-e. I pig see-REAL-1SG ‘I see a pig.’ Kombai (de Vries 1993:30)

(111) Nu ai fe-fera-do. I pig NEG-see-NEG ‘I do not see a pig.’ Kombai (de Vries 1993:30)

(112) Ya ai fe-fera-do. They pig NEG-see-NEG ‘They do not see a pig.’ Kombai (de Vries 1993:30)

In negated future tense forms, there are also no mood or person-number markers, but the future tense marker -i does appear after the verb stem. The post-verbal negator in future forms is madü rather than do (113). 140 9. Negation

(113) Nu fe-ami-n-i-madü. I NEG-drink-tr.nasal-FUT-NEG ‘I will not drink.’ Kombai (de Vries 1993:30)

The post-verbal negative element -do can also occur in nominal predicates (114 and 115).

(114) Mene a-do. This house-NEG ‘This is not a house.’ Kombai (de Vries 1993:30)

(115) Xo mene n-are-do. Person this my-father-NEG ‘This person is not my father.’ Kombai (de Vries 1993:32)

9.4 Reconstruction

In the reconstruction of Proto Awyu, it was shown that Jespersen’s cycle for double negation was at work in Awyu languages, and that the oldest form that could be reconstructed for PA was *fa/*feNEG...(*deNEG/EMPH). As double negation is also found in Kombai, in which the pre-verbal negator is optional and the post-verbal negator obligatory, a double negation can be reconstructed for Proto Awyu Dumut. The same stage of the Jespersen cycle is reconstructed for Proto Awyu-Dumut as for Proto Awyu, namely the oldest stage found in any of the daughter languages. Hence an obligatory pre-verbal negator *pe co-occurred with an optional post-verbal negator *nde in Proto Awyu-Dumut. The pre-verbal negator is reconstructed with a *p because Proto Awyu-Dumut *p fricativized to *f in both PA and Kombai. The vowel in the pre-verbal negator is /e/ rather than /a/ because /e/ is found in both Kombai and Proto Awyu, while the /d/ is prenasalized, as it always is in Proto Awyu-Dumut. That Proto Awyu-Dumut had double negation is further supported by Korowai and Tsaukwambo, two Greater Awyu languages that also have double negation. Korowai has (ba)...da as a negative construction (van Enk & de Vries 1997:106-107), while Tsaukwambo uses (bo)...nda (de Vries 2012a:174). For Proto Dumut, the post-verbal negator *nda/*ndoi was reconstructed, and this occurrence of a single post-verbal negator could represent the ﬁnal stage of the Jes- persen cycle. However, PD *nda/*ndoi is clearly the grammaticalized form of a lexical item meaning ‘nothing’, while the post-verbal negator *(n)de found in Proto Awyu- Dumut, Proto Awyu and Kombai originated as an emphatic marker. Hence Proto Dumut *nda/*ndoi cannot be related to Proto Awyu-Dumut *nde, although the similarity in form remains striking. 10

Deictics and Demonstratives

Spatial deictics in Awyu-Dumut languages can be described in terms of a three- way distinction between a speaker-related deictic (here, close to speaker), a hearer- related deictic (there, close to hearer) and a third-person, neither speaker-related nor hearer-related deictic (over there at greater distance, away from both speaker and hearer). Each Awyu-Dumut language makes this distinction between ‘here’, ‘there’ and ‘over there’ and has three basic roots to express these concepts. These three basic deictic terms are described for all Awyu-Dumut languages in Section 10.1, while the demonstratives derived from them, namely ‘this’, ‘that’ and ‘that over there’, are discussed in Section 10.2. A well-known grammaticalization path is the development of spatial deictics into textual deictics, that is, where spatial deictics are no longer used to point to concrete objects or locations but rather to point back to what has been said earlier in the discourse or to refer to what the speaker assumes the hearer knows. The ways in which Awyu-Dumut spatial deictics function as textual deictic elements are discussed in Section 10.3.

10.1 Awyu-Dumut Basic Deictics

All Awyu-Dumut languages have three deictic terms for ‘here’ (near deictic), ‘there’ (far deictic) and ‘over there’ (distant deictic), as listed in Table 10.1.1 Awyu-Dumut deictic terms rarely occur in their bare form as found in Table 10.1; rather, they combine with a number of morphemes. Below, deictic elements in each Awyu-Dumut language will be described before a reconstruction of Proto Aywu, Proto Dumut and Proto Awyu-Dumut deictic elements is presented in Section 10.1.7.

10.1.1 Deictics in Yonggom Wambon For Yonggom Wambon, Drabbe notes that me, ep and kop have derived forms mene, ewe and kowe, a combination of the deictic elements with topic marker =e (Drabbe

1No information or data are available on deictic elements in Yenimu. 142 10. Deictics and Demonstratives

Table 10.1: Awyu-Dumut deictics

near deictic far deictic distant deictic Dumut MAN me mbe mbego YWB me e(p) ko(p) DWB ne ep kop Awyu AXU nego wüo (wo), i- xo PSA ne u xate SHI ne ewe xage Ndeiram KOM me mofe maxo

Sources: Yonggom Wambon (Drabbe 1959:121); Digul Wambon (de Vries & Wiersma 1992:50-52); Mandobo (Drabbe 1959:19, 30); Aghu (Drabbe 1957:25); Pisa (Drabbe 1950:102-103); Shiaxa (Drabbe 1950:102-103); Kombai (de Vries 1993:37-38)

1959:121).2 Instead of kowe, kore is a more frequent derived form of kop ‘over there’; -re is a nominalizer that also occurs with secondary deictic elements such as koture ‘the one up there’ (Drabbe 1959:121-122), which is a combination of ko ‘over there’, tu- ‘to go up’ and -re. Drabbe goes on to note that the ﬁnal p of ep is optional and that a ﬁnal p may also appear on the derived forms of me and kore when they precede a noun, yielding menep and korep. When Yonggom Wambon deictic elements occur as locative adverbials, the ergative marker ŋga occurs between the deictic element and the verb, as in (116).3 When a deictic precedes a motion verb or a position verb in Yonggom Wambon, ŋga does not occur, as in (117).

(116) Enop kore ŋga ri-r-an. tree over.there ERG chop-REAL[NON1SG]-PST ‘He chopped down a tree over there.’ Yonggom Wambon (Drabbe 1959:121)

(117) Mene me-gen-ep. here come-REAL-1SG ‘I came to here.’ Yonggom Wambon (Drabbe 1959:121)

10.1.2 Deictics in Digul Wambon Like Yonggom Wambon deictic elements, the Digul Wambon far deictic ep can co- occur with circumstantial marker ka when it functions as a locative adverbial, as in (118).

2For further information and explanation of the topic marker =e, see Section 10.3.2. 3A description of ergative marker ŋga and its functions is presented in Section 11.3.2. 10.1. Awyu-Dumut Basic Deictics 143

(118) Kutip-ke-lo Kui mata-l-eva-mbo. Ep=ka night-SUPP-NON.CLOSE Kou arrive-REAL-1PL-PST. There=ERG la-l-eva-mbo-n=o... sleep-REAL-1PL-PST-tr.nasal=CONN... ‘At nightfall we arrived at Kouh. There we slept and...’ Digul Wambon (de Vries & Wiersma 1992:51)

The Digul Wambon near deictic ne always combines with an element mbo, yielding nombo, according to de Vries & Wiersma(1992). De Vries & Wiersma(1992) analyze this mbo as a predicative element originating in the verb mba ‘to sit’. Jang(2008:91) notes that ne can also occur independently or together with ka, citing (119), while he states that nombo has a temporal meaning ‘now’.4

(119) Nombo=ŋga noxop=e ko-no ambamtalok ko etu-wa-ketoi. here=ERG we=TOP go-SIM destination there enter-1PL-FUT.NEG ‘Now/at this time, if we continue, we will not make it to the destination.’ Digul Wambon (Jang 2008:91)

Both de Vries & Wiersma(1992) and Jang(2008) note that mbo also combines with the distant deictic kop ‘over there’, resulting in kopo. The distant deictic may occur without mbo and then is often shortened to ko (de Vries & Wiersma 1992:51).

10.1.3 Deictics in Mandobo The three Mandobo deictic elements are me ‘here’, mbe ‘there’ and mbego ‘over there’. The far deictic mbe also functions as a topic marker, a point which will be further explored in Section 10.3. For the near deictic, a derived form mene is more frequent than me. Like Yonggom Wambon mene, Mandobo mene is a combination of the near deictic me and topic marker =e, with a transitional nasal occurring intervocalically. The topic marker =e does not combine with the far and distant deictics in Mandobo. However, the far deictic mbe does combine with a nominalizer -re, cognate to the Yonggom Wambon nominalizer -re, yielding mbere ‘the one there’. The distant deictic mbogo also combines with -re: mbogore ‘the one over there’. The distant deictic mbogo consists of the far deictic mbe and the verb ko ‘to go’, where the ﬁnal vowel of mbe harmonizes with the /o/ of ko, and the /k/ of ko changes to /g/ intervocalically. The /e/ of mbe nearly always harmonizes with the nearest following vowel. The locative adverbial use of Mandobo deictic elements mene, mbe and mbogo are illustrated in (120) and (121).

(120) ...nen etot mene makmo mbegi-won ne-gen. 1SG.POSS etot here stay.together sit-1PL say-REAL[NON1SG] ‘...let us stay here together and sit in my etot he said.’ Mandobo (Drabbe 1959:60)

4De Vries & Wiersma(1992) list the near deictic as being no, while Jang describes it as ne. Jang’s analysis is correct, while no results from vowel harmony, which occurs when ne combines with mbo. 144 10. Deictics and Demonstratives

(121) Koro torogo Ngou n’etot torogo-gen doro, küaw go go.uphill Ngou POSS’etot go.uphill-REAL[NON1SG] CONN man agöw-agöp teriw agöw-agöp mba raŋ-gen. many-many woman many-many there lie-REAL ‘When they go uphill to the etot of Ngou, many men and many women are lying there.’ Mandobo (Drabbe 1959:77) (122) itigio-gen do riwop mbogo see-REAL[NON1SG] CONN middle.of.river over.there önö-gen. come.above.water-REAL[NON1SG] ‘He saw that it (the tree) came above water over there in the middle of the river’ Mandobo (Drabbe 1959:52)

10.1.4 Deictics in Aghu The Aghu deictic elements are nego ‘here’, wüo or wo ‘there’ and xo ‘over there’. Aghu deictic elements are always followed by the topic marker ke or kemu, often shortened to k’, when used adverbially (123-124). (123) Pani-di-k o-x-e: Kiawi, nego k’ ioxe come.up-SS-CONN say-REAL.NON1-NON1SG Kiawi here TOP be.NON1SG numo-x-e. say-REAL.NON1-NON1SG ‘He came up and said: Kiawi, will you stay here? he said’ Aghu (Drabbe 1957:67) (124) Ogsu-k-e. Bibimu wo k’ Go.up.small.way-REAL.NON1-NON1SG crouched there TOP ba-x-e. stay-REAL.NON1-NON1SG ‘He went up a little bit. He sat there crouched.’ Aghu (Drabbe 1957:70) In Drabbe’s texts, another Aghu verbal deictic with meaning ‘there’ is found, which Drabbe does not mention in his paragraphs on deictics and demonstratives, but which is much more frequent than wüo ‘there’. This element is ikemu (125). (125) Kiawi o-x-e: kesaxe dabudotobu Kiawi say-REAL.NON1-NON1SG tree across.water i-g-e, ikemu onu-k-enã lie-REAL.NON1-NON1SG there go.across.stream-REAL.NON1-NON1PL numo-x-e. say-REAL.NON1-NON1SG ‘Kiawi said: “The tree lying across the river, there they went across the river,” he said.’ Aghu (Drabbe 1957:66) 10.1. Awyu-Dumut Basic Deictics 145

The morpheme ikemu is mentioned by Drabbe in paragraph 77 of his 1957 Aghu grammar as an element meaning ‘there’. I analyze ikemu as consisting of i ‘there’ and ke/kemu, the topic marker that also combines with all other deictic adverbials. Thus, i- is another Aghu deictic root meaning ‘there’.

10.1.5 Deictics in Pisa and Shiaxa The other two Awyu languages, Pisa and Shiaxa use u (Pisa) and ewe (Shiaxa) to express the concept of ‘there’. Shiaxa ewe is clearly related to the far deictic ep found in Digul Wambon and Yonggom Wambon. The other two deictic elements are ne ‘here’ and xate ‘over there’ in Pisa, and ne ‘here’ and xage ‘over there’ in Shiaxa. In Pisa and Shiaxa, deictic elements combine respectively with topic marker ke (Shiaxa)5 or locative element ma (Pisa, (126)) when they occur as adverbials.

(126) U ma ayõ. there LOC lie.FUT[NON1SG] ‘He will lie there.’ Pisa (Drabbe 1947:10)

Before position verbs, adding ke is not obligatory in Shiaxa, while in Pisa, ma does not occur after deictics before the verb ba ‘to sit, to stay’. In addition to adding ke after deictic elements, Shiaxa deictic elements are often preceded by da, an element that might come from the verb ‘to come’, which is mode in Shiaxa but de and da in Pisa and Aghu, respectively. Thus da ne bax means ‘he is here’ in Shiaxa.

10.1.6 Deictics in Kombai Kombai has deictic elements me ‘here’, mofe ‘there’ and maxo ‘over there’. The near and far deictics me and mofe combine with the element -e when occurring adverbially. A transitional nasal occurs, yielding mene ‘here’ and mofene ‘there’. The distant deictic maxo does not combine with -e (127).

(127) Xumo-ra maxo Romalü bürü xa-no. die-CONN over.there Romalü place go.REAL-NON1PL ‘They die and go there to Romalü’s place.’ Kombai (de Vries 1993:38)

Besides mofene, mofexo also occurs, a combination of mofe ‘there’ and xo ‘to go’. The far deictic maxo might be analyzed as also containing xo ‘to go’ after ma, which means ‘to come’ in Kombai.

5ke is analyzed as a topic marker in Shiaxa because it occurs in similar positions as Aghu topic marker ke; Drabbe’s description of Shiaxa does not, in and of itself, provide enough information for this analysis. 146 10. Deictics and Demonstratives

10.1.7 Reconstruction of basic deictics Near, far and distant deictic elements can be reconstructed for Proto Awyu, Proto Dumut and Proto Awyu-Dumut. Table 10.2 lists the basic deictic elements and their derived forms found in the eight Awyu-Dumut languages, as well as their reconstructions.

Table 10.2: Reconstructed Awyu-Dumut deictics

near deictic far deictic distant deictic Dumut MAN me, mene mbe, mbere mbego, mbegero YWB me, mene(p) e(p), ewe kop, kowe, kore(p) DWB ne, nombo ep ko(p), kopo Awyu AXU nego (kemu) wüo (wo), i- (kemu) xo (kemu) PSA ne u xate SHI ne ewe xage Ndeiram KOM me, mene mofe, mofene maxo

PA *ne *(e)wV *xa/*xo PD *me, *ne *ep *kop PAD *me, *ne *ep *kop

Sources: Yonggom Wambon (Drabbe 1959:121), Digul Wambon (de Vries & Wiersma 1992:50-52), Mandobo (Drabbe 1959:19, 30), Aghu (Drabbe 1957:25), Pisa (Drabbe 1950:102-103), Shiaxa (Drabbe 1950:102-103), Kombai (de Vries 1993:37-38)

The near deictic is reconstructed as *ne in Proto Awyu, while Proto Dumut and Proto Awyu-Dumut each have two reconstructed forms, *me and *ne. There is no regular sound change established that would account for a possible /m/ to /n/ change or vice versa, and hence two forms are reconstructed. The form *me ‘here’ could be related to the verb ‘to come’, which is reconstructed as *me.6 The element mbo, which combines with the Digul Wambon near and distant deictics and which was hypothesized in Section 10.1.2 to have its origin in the position verb mba ‘to stay’, also occurs as the far deictic in Mandobo. The use of the verb mba in deictics is a change undergone by both Digul Wambon and Mandobo after splitting off from Proto Dumut; for Proto Dumut, only *ep is reconstructed as the far deictic. Mandobo replaced Proto Dumut *ep with mbe. In Proto Awyu, the far deictic is difﬁcult to reconstruct because Aghu and Pisa both have far deictics that consist of only one vowel. Nevertheless, the Proto Awyu far deictic is tentatively reconstructed as *(e)wV, with reﬂexes ewe in Shiaxa and w(ü)o in Aghu. This element is clearly cognate to Proto Dumut *ep, and thus for Proto Awyu-Dumut *ep is reconstructed as the far deictic meaning ‘there’.

6Yonggom Wambon, Mandobo, Shiaxa, Yenimu and Kombai have a reﬂex of *me for ‘to come’, while Digul Wambon, Aghu and Pisa have a different realization of the verb ‘to come’. 10.2. Awyu-Dumut Demonstratives 147

The distant deictic element is reconstructed as *kop in Proto Dumut and Proto Awyu-Dumut and as *xa/*xo in Proto Awyu. There are four Awyu-Dumut languages that combine their reﬂex of Proto Awyu-Dumut *kop with another morpheme: Man- dobo combines kop with its far deictic mbe, Kombai precedes xo with its near deictic me, while Pisa and Shiaxa add pragmatic markers te and ke to the distant deictic. In Yonggom Wambon, kop may occur independently, but it may also combine with topic marker =e. The reconstructed distant deictic *kop might have its origin in the Proto Awyu-Dumut motion verb *ko ‘to go’. Lastly, a nominalizer *-re and a circumstantial marker *-ka are reconstructed as co-occurring with deictic elements in Proto Dumut, as both of these elements occur in two Dumut languages. Deictic elements are thought to also have combined with topic markers such as *=e and *=ke in Proto Awyu, Proto Dumut and Proto Awyu- Dumut when they occurred before verbs that were not motion or position verbs.

10.2 Awyu-Dumut Demonstratives

Awyu-Dumut demonstratives, like Awyu-Dumut deictics, are three in number and roughly correspond to the English translations ‘this one here’, ‘that one over there’ and ‘that one far away’. They are given for all Awyu-Dumut languages except Ye- nimu in Table 10.3. It is not feasible to reconstruct Awyu-Dumut demonstratives, but the demonstratives are included in this chapter because they derive from Awyu- Dumut deictic elements and in turn grammaticalize into topic markers, theme markers and possibly into deﬁnite articles.

Table 10.3: Awyu-Dumut demonstratives

near demonstrative far demonstrative distant demonstrative Dumut MAN mene mbo – – YWB mene ewe kowe DWB nombone, nombono eve, evo – Awyu AXU nego wüo (wo), i- xo PSA nego ugo xatego SHI nere ewere xagere Ndeiram KOM mene mofene –

Awyu-Dumut demonstratives are derived from Awyu-Dumut deictics. Dumut languages derive their demonstratives from their deictic elements by adding a topical element, which is also derived from a deictic. Yonggom Wambon and Digul Wambon add topic marker =e to the deictic stem to get a demonstrative. This =e 148 10. Deictics and Demonstratives is derived from the far deictic ep, as shown in Section 10.3.2. Instead of =e, Digul Wambon may also add =o to a deictic element, yielding nombono ‘this one here’ and evo ‘that one there’; nombono and evo may only occur pre-nominally, functioning as demonstrative modiﬁers. Mandobo does not add either =e or =o, but rather a combination of the near deictic mene with the far deictic mbo yields a near demonstrative ‘this one here’. The Mandobo far deictic mbo, like =e, also functions as a topic marker, as illustrated in Section 10.3. An example of the Mandobo near demonstrative is given in (128).7

(128) Ngo-ano kurow ande-r-an na u you-SUBJ stealing eat-REAL[NON1SG]-PST my pig kurürümo-r-an; omba mene mbo ŋgo yo ge-ro katomo run.off.IT-REAL[NON1PL]-PST other here TOP you call be-SIM seek ge-ro ni-ge... be-SIM IMP-go... ‘your stealing and eating (=adultery) caused my pigs to run off; this other (pig) you go call and seek and...’ Mandobo (Drabbe 1959:46)

In Awyu languages, three different strategies are employed to form demonstratives. In Shiaxa, the suffix -re is added to the deictic, yielding nere, ewere and xagere. This element -re might be cognate to the nominalizer found with Mandobo and Yonggom Wambon deictic elements. In Pisa, the suffix go is added to deictic forms to yield demonstratives, for example ugo ‘that one over there’. Aghu employs a strategy different from that used by Pisa and Shiaxa. It does not add a suffix to get demonstratives; rather the ‘plain’ forms have demonstrative meaning, while if the topic marker ke follows nego, wüo or xo, they function as adverbial deictic elements. Before turning to how Awyu-Dumut demonstratives can develop a textual deictic function, I would like to shed light on another grammaticalization path that demonstratives often follow, namely that of turning into (definite) articles. This path was first noted by Greenberg(1978) and further explored in Nikolaus Him- melmann’s 1997 doctoral thesis entitled Deiktikon, Artikel, Nominalphrase: Zur Emer- genz syntaktischer Struktur. In the International Handbook for Language Typology

7In paragraph 46 of his Mandobo grammar description, Drabbe also describes another way in which demonstratives are formed in Mandobo, namely by adding the possessive (ne) or the possessive plus the word for ‘human’ (ne guap) to the deictic elements mene and mbere, for example:

(1) mene na ran here POSS woman ‘this woman’ Mandobo (Drabbe 1959:20)

(2) mbua mene nu guap clothes here POSS man ‘these clothes’ Mandobo (Drabbe 1959:20)

However, except in paragraph 46, no sentences containing a deictic element followed by ne or ne guap can be found. Rather, the combination mene mbo ‘this one here’ is a far more frequent demonstrative. 10.3. Awyu-Dumut Textual Deixis 149 and Language Universals, Himmelmann lists two criteria that can be used to distinguish articles from demonstratives, that is, to determine when a demonstrative has grammaticalized into an article. First, articles only occur in nominal expressions, while demonstratives occur in all manner of phrases, and second, the article’s position with regard to the noun is ﬁxed, whereas demonstratives are freer in their placement (Himmelmann 2001:832). However, the number of occurrences of demonstratives in Awyu-Dumut texts and grammar descriptions are too few to determine whether these criteria are met. Hence Awyu-Dumut demonstratives are not analyzed as having grammaticalized into deﬁnite articles, but the reader must bear in mind that this could - perhaps under certain circumstances - nevertheless be the case.

10.3 Awyu-Dumut Textual Deixis

Spatial deictics that are at ﬁrst used to refer to the world around the speaker can develop into anaphoric markers that refer to what has previously been said in a discourse. As Heeschen puts it, deictic elements “assume discourse functions and no longer refer to points in concrete space but to items previously mentioned in the linguistic context” (Heeschen 1997:177), and as Reesink has noted: “Many Papuan languages extend the use of a deictic to mark textual relationships such as topic and relative clause” (Reesink 1987:233). In Dumut languages, but not in Awyu languages, demonstratives have grammaticalized as markers of textual deixis. Dumut demonstratives function as topic markers. The deﬁnition of topic employed here is provided by Lambrecht(1994):

A referent is interpreted as the topic of a proposition if in a given situation the proposition is construed as being about this referent, i.e. as expressing information which is relevant to and which increases the addressee’s knowledge of this referent. (Lambrecht 1994:131)

Topic is here defined in terms of ‘aboutness’, rather than ‘givenness’. It can be summarized as “that which the speaker wants to talk about”(Reesink 1987:208-209). A topic most often has been mentioned before in the discourse, but not necessarily so. Lambrecht(1994:118) notes that others (Chafe, Dik) make a distinction between ‘topic’ and ‘theme’, where ‘topic’ is intra-clausal and defined, as Lambrecht does, in terms of ‘aboutness’, while a ‘theme’ is extra-clausal and “sets a spatial, temporal or individual framework within which the main predication holds”(Chafe 1976:50). I will also make this distinction, first talking about topics in Section 10.3.1 before talking about themes in Section 10.3.3.

10.3.1 topics The topic in a Dumut sentence, namely that element of the sentence about which the speaker wants to give information, is often marked by a topic marker that originates 150 10. Deictics and Demonstratives as a demonstrative. After nouns, Mandobo far demonstrative mbe, often realized as mbo, functions as a topic marker (129), as do Yonggom Wambon far demonstrative ewe (130) and Digul Wambon far demonstrative eve (131). Section 10.3.2 illustrates how the element =e found in Digul Wambon and Yonggom Wambon also functions as a topic marker. These topic markers occur after nouns or pronouns. In Mandobo example (129), the agent, who does the stealing of the tubers while his in-laws sleep, is topicalized using the Mandobo far deictic mbe, here realized as mbo. The speaker wants to talk about this person – who is half human and half pig in the myth – and what he is doing.

(129) Wemin ke-gen doro, e-anemo, night be-REAL[NON1SG] CONN 3SG.POSS-father.in.law e-ombut, e-gonöp, e-angen kinum 3SG.POSS-brother.in.law 3SG.POSS-mother.in.law 3SG.POSS-wife sleep koŋgon riŋgio-gen. Ege mbo kouyap ma ko-gen. fast lie.down-REAL[NON1SG] 3SG TOP again get.up go-REAL[NON1SG] Ma go irandöp kurop kouyap ŋgen. get.up go tubers steal again eat.REAL[NON1SG] ‘When it is night, his father-in-law, his brother-in-law, his mother-in-law and his wife lie fast asleep. He gets up again and goes. He gets up, goes, steals tubers again and eats them.’ Mandobo (Drabbe 1959:33)

In the following Yonggom Wambon example, the wife Kukyar is the most topical participant. She is introduced at the beginning of the sentence as the topic, and then the rest of the sentence talks about her and her offspring. Notice that the Yonggom Wambon demonstrative topic ewe combines with the ergative marker ŋga, which it often does.

(130) Mbinmarumyap Kapan, Kukyar ewe ŋga ŋgapmo-gen Mbinmarumyap Kapan, Kukyar TOP ERG bring.forth-REAL[NON1SG] de ui ke-r-an, itit ke-r-an. CONN pig be-REAL[NON1SG]-PST, cassowary be-REAL[NON1SG]-PST ‘Mbinmarumyap and Kapan, about (their wife) Kukyar, she brought forth (several offspring) and (these) become pigs and cassowaries.’8 Yonggom Wambon (Drabbe 1959:154)

In Digul Wambon example (131) the elder brother is the topic, but because the context of the utterance is unknown, it is not clear whether the elder brother occurred earlier in the story and hence constitutes ‘known information’. It is clear, however, that the elder brother is a topic in the sense that the speaker apparently wants to say something about his brother.

8Mbinmarumjap and Kapan are two males introduced at the beginning of this tale, and Kukyar is their shared wife. Kukyar is marked as the topic. 10.3. Awyu-Dumut Textual Deixis 151

(131) Na-net-eve mbap sinim-ŋga mba-ke. POSS-elder.brother-TOP father next-ERG sit-REAL.NON1SG ‘My elder brother sits beside father.’ Digul Wambon (Jang 2008:85)

10.3.2 topic marker =e Yonggom Wambon and Digul Wambon have an element =e, which is a shorter form of the far deictic ep ‘over there’ and that functions as both a topic marker and a deﬁnite article. Drabbe notes that “the deictic element ep is actually e...; the other primary deictic forms also occur with -p, namely menep, korep...”9 Thus, e is the far deictic in Yonggom Wambon, an analysis that may be extended to Digul Wambon. This far deictic functions as a topic marker when it occurs after a verbal clause, much like the demonstrative ewe in Digul Wambon and Yonggom Wambon. It then results in a thematic clause, as explained in Section 10.3.3 and illustrated by (132). (132) Mando-n-in e rogo-y-i-p. come-NON1SG-FUT TOP say-LIG-FUT-1SG ‘If he will come, I will say so.’ Yonggom Wambon (Drabbe 1959:142) When =e occurs after nouns in Digul Wambon and Yonggom Wambon, both de Vries & Wiersma(1992) and Drabbe(1959) analyze it as a connective (‘relator’ and ‘schakelele- ment’ in their terms, respectively). Another possible analysis is that in these instances =e is a topic marker, as it is after verbal clauses. Drabbe does note that, in Yonggom Wambon, =e can function as a subject marker (Drabbe 1959:119), and subjects are often the topic of an utterance. In Yonggom Wambon example (133) and Digul Wambon example (134), =e marks the topic of the sentence. (133) Ndokma-r-in de ok=e kojake-r-an. block-REAL-NON1PL CONN river-TOP get.full-REAL[NON1SG]-PST ‘When they blocked (it), the river ﬂooded.’ Yonggom Wambon (Drabbe 1959:146) (134) Ndako koma-l=o odo nux=e uto... and die-REAL[NON1SG]=CONN CONN 1SG-TOP go.in... ‘And when it died, I went in...’ Digul Wambon (de Vries & Wiersma 1992:84)

10.3.3 thematic clauses When Dumut topic markers follow verbal clauses, they form thematic clauses, which can also be referred to as ‘domain-creating constructions’ (Reesink 1994). Thematic clauses are background clauses that are subordinate and serve as a grounding point for the rest of the clause. They are the most common way to express subordination in Dumut languages. De Vries(2006:814) notes that thematic clauses can consist of all sorts of con- stituents and that they function to produce thematically coherent speech. Haiman

9“Het element ep is eigenlijk e...; zo komen ook de andere primaire [deiktische] elementen voor met p, namelijk menep, korep...” (Drabbe 1959:121). 152 10. Deictics and Demonstratives points out in his classic 1978 paper ‘Conditionals are Topics’ that what are often described as conditionals in grammars are actually topics, as both “are givens which constitute the frame of reference with respect to which the main clause is either true (if a proposition), or felicitous (if not)” (Haiman 1978:564). Of course, conditionals are not topics in the sense of being what the sentence is about (Reesink 2014:21), but the point made here is that clauses marked with a topic marker can have a conditional interpretation. Indeed, thematic clauses can have many interpretations/translations, for example as relative clauses, adverbials, temporal clauses, causatives or conditionals (Reesink 1994). Perhaps the translation that captures thematic clauses the best is “given the state of affairs x, y” (de Vries 2006:815). As in many Papuan languages, in Dumut languages the demonstratives that function as topic markers are employed to form thematic clauses. Thus Mandobo mbo and Yonggom Wambon and Digul Wambon ewe and =e can follow a verbal clause, nominalizing it and turning it into a thematic background clause. The following two examples illustrate the occurrence of thematic clauses in Dumut languages, with square brackets enclosing each thematic clause.

(135) E-nou mbo tagaŋ-gen: na-y-o no metarep toro 3SG.POSS-mother TOP say-REAL[NON1SG] son-LIG-VOC I hungry CAUS [etot ne terip ndu mbogo rigio-gen mbo], no go makmo ndu hut of women sago there stand-REAL[NON1SG] TOP I go join sago rigi-w-o ne-gen. stand[IRR]-1SG-CONN say-REAL[NON1SG] His mother says: “I am hungry, the hut of the women making sago is over there, I am going to join (them) and make sago,” she says. Mandobo (Drabbe 1959:78)

In example (135), mbo marks the clause ‘hut of women making sago’ as given information that forms the background or setting of the rest of the utterance.

(136) Ndare, matero [ndu mbi ri-gen mbo], matero motu-ndümu hear get.up sago DUR stand-REAL[NON1SG] TOP get.up sack-sago tömo te amkamo ŋgunop ndarakmo me-gen. pull.out CONN turn.inside.out head put.on come-REAL[NON1SG] ‘She hears, she gets up - she was making sago - she gets up, pulls out a small sago sack and turns it inside out, puts it on her head and comes.’ Mandobo (Drabbe 1959:81)

In example (136), mbo marks the clause she was making sago as background information that is essential to place the rest of the utterance in the right context (because she was making sago, she had a sago sack with her). In Yonggom Wambon thematic clauses, ewe tends to co-occur with ergative marker ŋga, as illustrated in (137). The topic marker e occurs withoutŋga (138) when it marks a thematic clause. 10.4. Summary 153

(137) [Iŋ-gin-in ewe] ŋga ut-ken-ep. eat-REAL-NON1PL TOP ERG go.in-REAL-1SG ‘When they were eating, I went in.’ Yonggom Wambon (Drabbe 1959:134)

(138) [Mirip kup maya-r-an e,] kigip ko rain COMIT come.down-REAL[NON1SG]-PST TOP other go ti-r-an. build-REAL[NON1SG]-PST ‘When it rained (given that it rained), she built another (hut).’ Yonggom Wambon (Drabbe 1959:156)

Lastly, two examples from Digul Wambon are provided, with eve marking a subordinate thematic clause.

(139) [Simson=e nde-t-mbo-n=eve] nexo salip=e Simson=TOP come-REAL[NON1SG]-PST-tr.nasal=TOP 3SG.POSS wife=TOP wasi ande-t-mbo. already eat-REAL[NON1SG]-PST ‘When Simson arrived, his wife had already eaten.’ Digul Wambon (Jang 2008:86)

(140) Mbonop-ŋgambun=ka mbaxe-mbel=o ŋgerkayi lavo-va Mbonop-maelstrom=ERG stay-SEQ=CONN saw take[IRR]-NON1PL ne-mbel=o ep=ka mba-l-eva-mbo. [Ep=ka say-SEQ=CONN there=ERG stay-REAL-NON1PL-PST there=ERG mba-l-eva-mbo-n=eve] sanov=e ilo ka-l-eva-mbo. stay-REAL-NON1PL-PST=TOP Monday=TOP go.down go-REAL-NON1PL-PST ‘We stayed at the Mbonop maelstrom, in order to saw we stayed there. Given that/when we stayed there, on Monday we went down (there).’ Digul Wambon (de Vries & Wiersma 1992:87)

10.4 Summary

This chapter offered a description and reconstruction of Awyu-Dumut spatial deictic elements and then focused on how Awyu-Dumut demonstratives derive from Awyu-Dumut deictic elements. Two grammaticalization pathways were then discussed:

1. demonstratives that grammaticalize into deﬁnite articles/markers of deﬁnite- ness 2. deictics and demonstratives that grammaticalize into topic markers

For the ﬁrst grammaticalization path, this chapter demonstrates that there is too little Awyu-Dumut data to determine whether demonstratives (and deictic-derived 154 10. Deictics and Demonstratives element =e) have indeed turned into deﬁnite markers, although it is likely that at times they do have this function. The grammaticalized use of deictic-based topic markers was only found in Dumut languages, where demonstratives are employed as topic markers after nouns and after entire clauses. When they cliticize to a clause, they form a thematic clause or domain-creating construction. 11

Clause Linkage

Languages can link clauses together into sentences in multiple ways. The most common ways to link clauses together cross-linguistically are through either coordination or subordination, which form two ends of a continuum (called the continuum of hierarchical downgrading by Lehmann(1988)). As in many Papuan languages, clause chaining rather than either coordination or subordination is the main clause linking strategy in Awyu-Dumut languages. Foley notes that clause chaining is “probably the most distinctive feature of Papuan languages” (Foley 1986:175), going on to describe chained clauses as containing ‘co-subordinate’ verbs and as falling in between coordination and subordination on the coordination-subordination continuum. Awyu-Dumut clause chaining is discussed in Section 11.2, after Awyu- Dumut clause coordination is described in Section 11.1. At the subordination end of the continuum Awyu-Dumut thematic clauses are found, which, along with other types of subordinate clauses, are described in Section 11.3. Important to remember in this chapter is that Awyu-Dumut languages have three types of verbs: non-finites, semi-finites and finites. Non-finite verbs consist of a verb stem, semi-finite verbs consist of a verb stem, a mood marker and person- number marker, and finite verbs consist of a verb stem, a mood marker, a person- number marker and a tense marker. Non-finite verbs are dependent verbs, while both semi-finite and finite verbs are independent verbs. ‘Independent’ is here defined as ‘not needing another verb in the clause or sentence in order to function’; an independent verb form can hence be the only verb form in a complete sentence. Dependent verb forms always need another verb in order to function. These distinc- tions will be important in distinguishing between coordination and clause chaining, and also play a major role in Awyu-Dumut’s switch reference systems, described in Section 11.5.

11.1 Coordination

Coordinated clauses are syntactically equal and may also occur independently, meaning that a coordinated clause can be a sentence by itself. As Foley(2010:27) notes: 156 11. Clause Linkage

“[c]lauses linked in a coordinate nexus are not in an asymmetrical relationship of embedded versus matrix clause, but are joined at the same level, strung along rather like beads on a string.” Awyu-Dumut languages have two strategies to coordinate clauses: juxtaposing them (Section 11.1.1), or placing a coordinator between them (Section 11.1.2). As will become apparent in Section 11.2, clause chaining can also be considered a type of coordination.

11.1.1 Coordination by juxtaposition In Awyu-Dumut languages, most coordinated clauses are juxtaposed, together forming a sentence. Coordinated verbal clauses are found in Yonggom Wambon, Digul Wambon, Mandobo, Aghu and Kombai. Unfortunately, Drabbe gives no description of how clauses are joined together in Pisa, Shiaxa or Yenimu, nor is there information available on how juxtaposed clauses differ from separate sentences prosodically. Awyu-Dumut languages, like many Papuan languages, have medial verb forms and final verb forms, that is, verbs that may occur clause-finally and those that may only occur clause-medially. The final verb forms are more finite than the medial verb forms. In Awyu-Dumut languages, both semi-finite forms and finite verb forms may occur clause-finally, while non-finite verb forms only occur clause-medially. A (non-final) clause is coordinated to the next clause if it ends in a semi-finite or finite verb. Thus Aghu example (141) consists of two coordinated clauses, both ending in a semi-finite realis verb form.

(141) [Gi-k-enã] [efe n’amoko axe saxa-k-e]. scrape-REAL-NON1PL 3SG POSS’son ﬁsh shoot-REAL-NON1SG 1 ‘They scrape (poison into water) and her son shoots ﬁsh. ’ Aghu (Drabbe 1957:57)

Likewise, the following examples from Mandobo (142) and Yonggom Wambon (143) consist of multiple coordinated clauses ending in semi-finite or finite verb forms, with a semi-finite or finite verb form at the end of the sentence.

(142) [Tokmo ndigi-r-an] [ro me-re divide give-REAL[NON1SG]-PST have.with come-NON.CLOSE rurukma-gen-on]. put.together-REAL-NON1PL. ‘He divides and gives (the pigs), and they come put them together.’ Mandobo (Drabbe 1959:22)

(143) [Naerop taemba-r-in] [kima-r-an]. Naerop shoot-REAL-NON1PL die-REAL[NON1SG]-PST ‘They shoot Naerop and he dies.’ Yonggom Wambon (Drabbe 1959:147)

1Drabbe notes that, in Aghu culture, women catch fish by scraping a poisonous root that drugs the fish into the water, while men catch fish using bow and arrow. 11.2. Clause Chaining 157

11.1.2 Coordination by coordinator In Digul Wambon, whenever a semi-finite -kend form or a finite past or future form occurs non-finally, it is coordinated to the next clause. The coordinating connective =o occurs on these clauses but also on all other medial clauses that are part of clause chains rather than fully coordinated. An example of two coordinated clauses, the first of which ends in a semi-finite realis form, is given in (144).

(144) Mulo kilim-ke-knd-ev=o hetaxa-l-ep-mbo-n=e... go.down pursue-be-REAL-1SG-CONN see-REAL-1SG-PST-tr.nasal=TOP ‘I went down, pursued it and I saw that...’ Digul Wambon (de Vries & Wiersma 1992:84)

Besides Digul Wambon, Kombai is the only other Awyu-Dumut language that employs an explicit coordinating connective, namely -a, after medially-occurring semi- ﬁnite verb forms, as in (145).

(145) Xalufo nu bone-d-ef-a gu o mi. yesterday 1SG DUR.eat-REAL-1SG-CONN 2SG water drink[REAL.NON1SG] ‘Yesterday I ate and you drank water.’ Kombai (de Vries 1993:57)

In Yonggom Wambon and Mandobo, =o does occur as a linking element, but it is unclear whether it is a coordinator in these languages. It usually appears between a verb of speaking and that which was said, as in example (146).

(146) Sinam ŋga nataem-i-non=o nde-t. bow ERG shoot-[IRR]NON1PL-IMP=CONN say-REAL[NON1SG] “‘You must shoot me with a bow” he said.’ Yonggom Wambon (Drabbe 1959:149)

De Vries(1986, 2010) claims that Digul Wambon ndi and its cognate in its two Dumut sister languages, namely te in Mandobo and Yonggom Wambon, are also clausal coordinators. However, see Section 11.3.4 for an argumentation explaining why Dumut te is better analyzed as a subordinating connective.

11.2 Clause Chaining

Besides fully coordinated clauses, Awyu-Dumut languages have clauses that are chained together. I will adhere to Longacre(2007:399) for a definition of a clause chain; he defines clause chains as consisting of “a final clause that has a verb of distinctive structure that occurs but once in the entire chain” and “non-final clauses which have verbs of different and more restricted features.” Longacre goes on to note that “the final clause is like an engine that pulls a string of cars.” There may be one or more non-final clauses. The main feature of clause chains is that the medial verbs are dependent upon the last verb for the interpretation of one or more of their inflectional categories. On the coordination-subordination continuum, clause chaining is best classified as a type of coordination, one which involves dependent, but not embedded, verbs. As Foley notes, clause chains “do not function as embedded 158 11. Clause Linkage parts within a whole, but are linked to a fully inflected verb in a linear string, much like beads on a necklace. Because the linking of such clauses is at the same structural level, rather than as part within a whole, I regard such clauses as coordinate and verbs of such clauses as ‘coordinate-dependent’” (Foley 1986:177). In Awyu-Dumut languages, the medial verbs used in clause chains are non-finite verbs that consist of a verb stem or a verb stem plus temporality marker, while semi-finite or finite verb forms can occur clause-chain finally and are independent.2 The non-finite verbs are dependent on the final verb for mood and person-number interpretation. If the final verb expresses tense in addition to mood and person-number, this category is also inherited by the non-finite verb(s) that precede(s) it. A Digul Wambon sentence consisting of three clauses, of which two are clause chains ending in a semi- finite verb, is found in (147), while (148) contains a Yonggom Wambon clause chain and (149) an Aghu clause chain. (147) [Yandit wamip=ka oy=e hetak-mbel=o tambe-l-ev=o] road inside=ERG pig=TOP see-SEQ=CONN shoot-REAL-1SG=CONN [liki-lap-elo ka-l=o] [kilim-ke-knd-ev=o...] break-take-SIM go-REAL[NON1SG]=CONN pursue-be-REAL-1SG=CONN... ‘I saw a pig inside the road and I shot it, and the pig took the arrow, went and I pursued it and...’ Digul Wambon (de Vries & Wiersma 1992:83) (148) [Ngo nanŋgui e, ŋgo nenŋgui e, Matiram ŋga 2SG.POSS younger.brother TOP, 2SG.POSS elder.brother TOP, Matiram ERG ŋgotonde e-no mir e top me agumo mba-gen-op]. kill.IT eat-SIM bones TOP pit come put.in stay-REAL[NON1SG]-HABIT ‘Matiram killed and ate your younger and elder brothers (habitually/iteratively), coming and putting their bones in a pit.’ Yonggom Wambon (Drabbe 1959:154) (149) [Nu ni da-de-k syü isiom a-de-k büsyü kem 3SG 1SG come.I-SS-CONN banana all take.I-SS-CONN home TOP xo-do-k ab-su-k-e]. go.I-SS-CONN take.I-go.up.I-REAL-NON1SG ‘He came to me, took all the banana’s, went home and took them up (into the house).’ Aghu (Drabbe 1957:19)

11.3 Subordination

Clauses that are not coordinated or chained, are subordinate. There are multiple types of subordinate sentences in Awyu-Dumut languages, and in the following sec-

2Digul Wambon semi-finite -t forms are described by de Vries & Wiersma(1992) as not occurring sentence-finally and hence not being independent, but see Section 11.5.4 for new data which show that Digul Wambon -t forms are much like other Awyu-Dumut semi-finites: independent and sentence-final. 11.3. Subordination 159 tions, six different subordination strategies employed by Awyu-Dumut languages are discussed. All Awyu-Dumut subordinate clauses are linked to the rest of the sentence by a subordinating connective; the six subordination strategies make use of different subordinating connectives. What all Awyu-Dumut subordinating connectives have in common is that they nominalize the subordinate clause that they follow. Awyu- Dumut subordinate clauses are either hypotactic or embedded. Hypotactic subordinate clauses are not dependent on the main clause, nor do they have a syntactic function within the main clause (Lehmann 1988:182, 185). Embedded subordinate clauses, on the other hand, are dependent on the main clause and are governed by it. Another important feature of Awyu-Dumut subordinate clauses is that they present background information. Concerning subordinate clauses in Papuan languages in general Foley states: “Subordinate clauses in Papuan languages always function to background given information, and correspond to two different constructions in more familiar languages, adverbial clauses and relative clauses” (Foley 1986:201). Awyu-Dumut subordinate clauses fall within this generalization; they either present information usually found in adverbial clauses, for example information about the timing, location, and/or manner of the main action, or they present further information about the main topic of the sentence.

11.3.1 Subordination: thematic clauses with demonstrative topic markers Thematic clauses formed with demonstrative-based topic markers are a very common way to form subordinate clauses in Dumut languages and were described in Section 10.3.3. Thematic clauses were there deﬁned as “background clauses that are subordinate and serve as a grounding point for the rest of the clause. The topic marker nominalizes the clause that it follows.” Thematic clauses are extra-clausal, most often appear in a left-dislocated position and can best be translated as “given the state of affairs x, y” (de Vries 2006:815), although many other translations are possible. What many Papuan thematic clauses have in common is that thematic clause markers are markers that occur on both noun phrases and verb clauses, and in both instances render the phrase or clause they follow (more) presuppositional (Reesink 1983:223-224). In many languages, including Dumut languages, thematic clause markers also function as topic markers after nouns. Examples of thematic clauses formed with demonstrative topic markers mbo (Mandobo), ewe and e (Digul Wambon and Yonggom Wambon) were given in Section 10.3.3 and are not repeated here. 160 11. Clause Linkage

11.3.2 Subordinator ŋga

The ergative marker ŋga,3 found in Digul Wambon and Yonggom Wambon, functions as a subordinator when it attaches to verbal clauses. The ergative marker optionally marks the agent of an action in Yonggom Wambon, as ergative markers do in other Papuan languages such as Dani and Hua (Dixon 1994:58). Its function is then to either disambiguate who the agent is or to mark less prototypical agents. Thus ŋga in (150) makes it clear that Matiram is the cannibal, rather than the brothers, while in (151), ŋga marks a snake pulling on a sago sack, an action usually performed by humans. (150) Ngo nanŋgui e, ŋgo nenŋgui e Matiram ŋga 2SG.POSS younger.brothers TOP, 2SG.POSS elder.brothers TOP Matiram ERG ŋgotonde e-no mir e top me agumo kill.IT eat-NON.CLOSE bones TOP hole come put.in mba-gen-op. stay-REAL[NON1SG]-HAB ‘Your younger brothers, your older brothers, Matiram used to kill (them), eat (them) and come put the bones in a hole.’ Yonggom Wambon (Drabbe 1959:153-154) (151) Kinum ya-r-a aŋgun ŋga matui mbima-t sleep lie-REAL[NON1SG]-SEQ snake ERG sago.sack pull-REAL[NON1SG] mborotke-t... break-REAL[NON1SG] ‘(she) lies asleep and the snake pulls (on) the sago sack and it (the sago sack) breaks...’ Yonggom Wambon (Drabbe 1959:155) The ergative marker originates in what de Vries & Wiersma(1992:58) call a ‘circumstantial marker’, which may express instrumental (152), locative (153), manner (154) or time reference (155) meaning. Foley notes that it is common in Papuan languages for (peripheral) case markers that mark ablative, causal and/or instrumental case to develop into (core) ergative markers which mark the agent. He explains that “[t]he extension of the instrumental and causal case to the actor is not surprising: both are causes of the event, either primary or secondary, and identity between the ergative and instrumental case markers is a common feature of languages, including other Papuan languages such as Enga (Lang 1973), Kewa (Franklin 1971), Kate (Pilhofer 1933) and Selepet (McElhanon 1972)” (Foley 1986:107). In his book on ergativity, Dixon(1994:57) further notes that cross-linguistically, the ergative case is known to also function as a locative, instrumental or genitive case marker. (152) Mayum tombon ŋga ragae turuma-t. mayum.sago leaf.vein ERG ﬁsh shoot-REAL[NON1SG] ‘He shot ﬁsh with mayum sago leaf veins.’ Yonggom Wambon (Drabbe 1959:146) 3ŋga has an allomorph =ka in Digul Wambon. In Digul Wambon, it is a clitic, while Drabbe writes ŋga separately in Yonggom Wambon. 11.3. Subordination 161

(153) Kixup=ka okima-l-eva-mbo. Digul=ERG take.bath-REAL-1PL-PST ‘We took a bath in the Digul river.’ Digul Wambon (de Vries & Wiersma 1992:58) (154) Wasi=ŋga nda-xe quick=ERG come-REAL[NON1SG] ‘He comes quickly.’ Digul Wambon (de Vries & Wiersma 1992:59) (155) Ariw-amin ŋga mende-r-a mberemo ra early-morning ERG come-REAL[NON1SG]-SEQ gather take ku-r-an. go-REAL[NON1SG]-PST ‘In the early morning she came and gathered [the fish traps], took them and went.’ Yonggom Wambon (Drabbe 1959:155) The ergative marker ŋ ga functions as a subordinator, or thematic clause marker, when it follows a verbal clause, much like the Dani ergative marker (n)en, as observed by Reesink(1994:110), who cites (156) from Bromley (1981). 4 In this case, the clause ‘we finished thatching’ is marked as a subordinate clause by the ergative marker -en, which “signals a kind of causality...which shades into temporal succes- sion” (Reesink 1994:110). (156) Wa’lekka saloko hel-uk-en hupakko mottok ki ak-y. thatch covering put-PERF-ERG night altogether enter come-I ‘After we finished the thatching, when it was completely dark, I came in here.’ Dani, Bromley 1981 Drabbe notes that ŋga is a marker of time both after nouns and after verbal clauses. He contrasts ŋga after an NP, namely miti ŋga kuran ‘night ERG he.went’ (he went at night) with ŋga after a clause: mitik ken ŋga kajip literally ‘night will.be ERG I.will.go’ (when it is night, I will go) (Drabbe 1959:144). Thus subordinate clauses marked by the ergative ŋga in Yonggom Wambon and Digul Wambon always have a temporal meaning, which can be translated with ‘while’, ‘when’ or ‘after’. The example given in (157) illustrates ŋga functioning as a subordinator in Digul Wambon, while (158) provides a similar example for Yonggom Wambon. (157) Nux=e ande-l-ep-o=ŋga nexep nde-t-mbo. I=TOP eat-REAL-1SG-PST=ERG he come-REAL[NON1SG]-PST When I ate, he came.’ Digul Wambon (de Vries & Wiersma 1992:60) (158) Ndun rino mbage-ro ŋga eto-gen-ep te sago make stay-non.close ERG see-REAL-1SG CONN mende-r-an. come-REAL[NON1SG]-PST ‘While I was preparing sago, I saw that he came.’ Yonggom Wambon (Drabbe 1959:134)

4Bromley (198:266) did not analyze (n)en as an ergative marker, but called it the generalized source relator -nen. Foley(1986:107) is probably the source of the reanalysis of Dani -nen. 162 11. Clause Linkage

For Digul Wambon, de Vries claims that “since all medial verb clauses are coordinated with the next clause, the postpositions/subordinators never occur with medial verb clauses” (de Vries & Wiersma 1992:60). While this is true for Digul Wambon, where indeed a subordinator never follows a non-ﬁnite SS verb form, in Yonggom Wambon ŋga frequently follows non-ﬁnite SS verb forms, as is the case in (158) above. This is problematic because, as de Vries noted, medial verb forms stand in a (chaining) coordinate relationship to the next verb, so how can they occur in a subordinate clause? Reesink has found a similar puzzling construction in Usan and notes that although “medial verbs basically constitute asymmetrical coordinate relationships...this basically coordinate relationship may be changed into a subordinate one in order to clearly express conditionality. Once this is done, again with the [Usan] ‘given’ demonstrative eng, the preceding clause falls outside the scope of negative, imperative and interrogative, as they apply to the following clause” (Reesink 1987:239). The difference between an Usan medial verb in a subordinate clause and a medial verb in a coordinated clause is illustrated by (159), showing a subordinate clause, and (160), showing a coordinate clause.

(159) Yarab eng ye-nipat qur big-ar. come.SS the me-step.over.SS money put-pl.IMP ‘If you come, step over me and put your money (in the basket).’ Usan (Reesink 1987:239)

(160) Yarab qoasi-qoasi qur big-ar. come.SS miss-miss.SS money put-pl.IMP ‘Come, go beside me and put your money (in the basket).’ Usan (Reesink 1987:239)

In the first sentence, the imperative of the final verb does not have scope over the medial SS verb yarab ‘to come’, only over the medial SS verb nipat. In the second sentence, the imperative has scope over both of the medial SS verbs yarab ‘to come’ and qoasi-qoasi ‘to go beside’. Thus the Usan subordinator eng blocks the scope of the imperative. In his article on domain-creating constructions, Reesink notes that “[m]edial (SS and DS) clauses followed by -eng...are shielded from various operators on the verb of the matrix clause: negation, illocution, nominalization. The categories of tense and subject person-number, however, are subjected to the general constraints exerted by the final verb on the preceding ones in a sentence” (Reesink 1994:104-105). Thus because the final verb bigar in both (159) and (160) has a plural person-number interpretation, all medial SS verbs, including yarab in the subordinate eng clause, have plural person-number interpretation. The subordinator eng blocks certain but not all operators and categories of the (final) matrix clause verb when it occurs after a medial verb. Like Usan -eng, Yonggom Wambon ŋga does not block switch reference. There are too few occurrences of ŋga in the data to determine whether or not it blocks the scope of imperatives, interrogatives or negation. 11.3. Subordination 163

11.3.3 Subordination with existential verb ke ‘to be’ The support verb ke ‘to be’ appears as a marker of subordinate sentences with conditional meaning in both Awyu and Dumut languages. In Digul Wambon, what de Vries & Wiersma(1992) describe as a conditional marker, namely kexet, can be analyzed as being a fossilized non-ﬁnite form of the verb ke followed by a third person non-singular realis form of the same verb, as illustrated in (161). Digul Wambon kexet only occurs in conditional sentences and is always followed by either of the topical subordinators =e or =eve, discussed in Sections 10.3.3 and 11.3.1, yielding kexeleve or kexele.

(161) ke-xe-t be-be-REAL[NON1SG] ‘being, it is/was’

In (162), kexel=eve marks the subordinate protasis of the sentence ‘if the river rises, we will not go.’

(162) Kixuv=e nde-t-ke-xe-l=eve noxov=e Digul=TOP come-REAL[NON1SG]-be-be-REAL[NON1SG]=TOP we=TOP Manggelum ko-nok-siva. Manggelum go-neg-intent.NEG.1PL ‘If the Digul River rises, we do not want to go to Manggelum.’ Digul Wambon (de Vries & Wiersma 1992:31)

In Yonggom Wambon, a construction ket kinin ŋga occurs, which can be analyzed as consisting of a non-ﬁrst person realis form of the verb ke followed by a non-ﬁrst person future form of ke, which in turn is followed by the ergative marker ŋga (163). Like in Digul Wambon kexel=eve, the verb ke occurs twice in the Yonggom Wambon ket kinin ŋga construction.

(163) ket ki-n-in ŋga be-REAL[NON1SG] be-NON1SG-FUT ERG ‘it is, when it will be’ Yonggom Wambon (Drabbe 1959:134)

Drabbe always translates Yonggom Wambon clauses containing ket kinin ŋga with ‘when’, rather than with a conditional ‘if’, and states that they talk about future simultaneous events, such as those expressed in (164), in which the giving of food is not dependent on whether the people are working, but on when they are working.

(164) Awoŋ rap-kin-in ke-t ki-n-in ŋga, ko work take-REAL-NON1PL be-REAL[NON1SG] be-NON1SG-FUT ERG go nayok ande-nan. give.IMP eat-NON1PL ‘When they are working (when it is the case that they are working), go give (them food), let them eat.’ Yonggom Wambon (Drabbe 1959:134) 164 11. Clause Linkage

In all three Dumut languages, the verb ke in subordinate sentences with conditional meaning is always conjugated in non-first person singular form because “the entire first clause is the subject of get”(Drabbe 1959:105).5 That is, the ke verb does not agree with the subject within the first clause, but rather with the clause as a whole.6 For Mandobo, Drabbe notes that ke-t only occurs when the protasis contains a finite past tense verb or a negated verb (Drabbe 1959:105), giving (165) and (166) as examples. (165) Ngoano kurop timo-r-an ke-t, no tagamo-nok. you secret take-REAL[NON1SG]-PST be-REAL[NON1SG] 1SG tell-IMP ‘If you stole it, tell me.’ Mandobo (Drabbe 1959:105) (166) Nöp mbegi-ne-gen-e-mda ge-t, rogoa nanet 1SG be-NEG-REAL-1SG-NEG be-REAL[NON1SG] fish my.older.brother ndigio-nok. give-IMP ‘If I am not there, give the fish to my elder brother.’ Mandobo (Drabbe 1959:105) Similarly, in Aghu, ki ‘to be’ only appears in conditionals if the protasis is negative, as shown in (167). The verb ki appears as k˜ı, the non-first person optative form of the verb. (167) Fiko afi-nde k˜ı, sumke edaxe-nde. work take-NEG be[NON1SG] tobacco give-NEG. ‘If you do not work, I will not give you tobacco’ Aghu (Drabbe 1959:23) Pisa is the last Awyu-Dumut language in which ke is attested in sentences with conditional meaning. In Pisa, the protasis starts with a non-first person singular form of the verb rigirö, a verb no longer in use that Drabbe translates as ‘to happen’. Af- ter this verb, the rest of the subordinate clause follows, and the subordinate clause always ends with a non-first person future form of the verb gi ‘to be, to happen’ (the Pisa cognate of ke). Thus rigirö dafina giro, edaxay literally means ‘if it happens (that) they will come, if (that) will be, I will give it” (Drabbe 1950:127). It can be concluded that the verb ke ‘to be’ occurs in sentences with conditional meaning in many Awyu-Dumut languages, and can in those instances best be translated as ‘it being the case that...’ The verb has the whole preceding clause as its argument and always appears in non-first person singular (that is, third person singular) form. In Digul Wambon and Yonggom Wambon, ke appears twice in a fossilized form. Due to its widespread occurrence in Awyu-Dumut sentences with conditional meaning, it is likely that ke appeared in sentences with conditional meanings in Proto Awyu, Proto Dumut and Proto Awyu-Dumut.

5“De gehele voorzin is subject van get”(Drabbe 1959:105); get is the non-ﬁrst person singular realis -t form of ke in Mandobo after a vowel. 6Note that Reesink describes something similar for Usan existential verbs which follow ﬁnal verbs: “the existential verb seems to have the total preceding event as its argument, much as sensory verbs have the event that is expressed by the medial verb as their complement” (Reesink 1987:92). 11.3. Subordination 165

11.3.4 Subordinator te

All three Dumut languages have a focus element te that links clauses together.7 On the subordination-coordination continuum, the connective focus marker te is hard to classify as either coordinating or subordinating. The arguments presented in the following paragraphs indicate that an analysis of te as a subordinating connective is most plausible. Between verbal clauses, te will simply be glossed as CONN ‘connective’. Drabbe describes clause-ﬁnal te as a linking element or a conjunction,8 but he does not indicate whether it is a coordinating or a subordinating conjunction. The translations Drabbe gives for te clauses are numerous. Thus, for example, Drabbe notes that the Mandobo sentence

(168) Ne-neti taga-ŋgen oto me-gen-ep. 1SG.POSS-father say-REAL[NON1SG] CONN come-REAL-1SG Mandobo (Drabbe 1959:103) can mean ‘my father said it and then I came’, or ‘my father said it and therefore I came’ or ‘when my father said it, I came’, but also simply ‘my father said it and I came.’9 Similarly, he notes that koŋgenep to tagamoran can be translated as either ‘after I wrote, he read it’ or ‘I wrote it and he read it’. Thus Drabbe ascribes both coordinating meaning (‘and’) and subordinating meaning (‘after, when, because’) to the linking element te, although he does not comment further on te’s syntactic role in the sentence except to say that it is a linking element. Unlike Drabbe, who leaves undecided whether te is a coordinating or a subordinating conjunction, de Vries (2010) analyzes Dumut te as a coordinator. The scope that polarity (negation) markers or illocutionary force (imperative, interrogative) markers have within a sentence consisting of multiple clauses is often used as a test to determine whether one is dealing with subordinate clauses. Haiman shows that subordinate clauses in Hua are impervious to the mood, tense or negation of the ﬁnal clause (Haiman 1976). Similarly, Reesink(1987) describes the scope of negation, interrogatives and imperatives as not extending to subordinate -eng clauses. Thus if negation, imperative and interrogative modalities marked in the ﬁnal clause do not extend over any preceding te clauses, those clauses are likely to be subordinate. The scope of negation and imperative do not extend over the clause marked by te in examples (169) and (170). Unfortunately there are too few

7Allomorphs ndi, oto (DWB), to, do, de, odo, oto (MAN, YWB). The Dumut focus marker te was also discussed in Chapter3 on Nouns. It was argued that te is always a focus marker when it occurs after a noun, although it may have a copular interpretation. Only when te is repeated after each element of an enumeration can it be analyzed and glossed as COORD (coordinator). This analysis of te as a focus marker in most instances after nouns was a reanalysis of both Drabbe and de Vries, who alternately call te a copula, an emphatic marker, a linking element or a coordinator. 8“In al die gevallen functioneert te min of meer als voegwoord” (Drabbe 1959:134). 9“Met neneti tagaŋgen oto megenep kan bedoeld zijn : mijn vader zei het en toen ben ik gekomen, of : hij zei het en daarom kwam ik, of : toen hij het zei ben ik gekomen, maar ook eenvoudig : hij zei het en ik kwam” (Drabbe 1959:103). 166 11. Clause Linkage occurrences of negation, imperatives and interrogatives in Dumut texts to solidify the claim that their scope does not extend over te clauses, although the few examples there are do indicate that Awyu-Dumut te clauses are similar to the subordinate clauses described by Reesink and Haiman.

(169) Matero eŋgot keemo itigio-gen do yombutop get.up torch light see-REAL[NON1SG] CONN door tokmo-ne-gen-da te raŋ-gen open-NEG-REALNON1SG-NEG CONN lie.down-REAL[NON1SG] ‘She gets up, lights a torch and sees: the door is not open, and she lies down.’ Mandobo (Drabbe 1959:§96)

(170) Ndu orü rigi-non do, küap tak sago cut.down stand-[IRR]NON1PL CONN people invitation kea-nog-i-nin ne-gen. walk-IMP-LIG-NON1PL say-REAL[NON1SG] ‘While/when the sago is being cut, go invite people, he said.’ Mandobo (Drabbe 1959:§109)

Another indication that Dumut te is not a coordinator is its occurrence at the end of sentences. Sentence-final te cannot be a coordinating linking element, as there is no following clause to which te links the previous clause. Sentence-final te clauses are like the construction that Evans refers to as ‘insubordination’, that is, “conven- tionalized main clause use of what, on prima facie grounds, appear to be formally subordinate clauses” (Evans 2007:367). According to Schapper & SanRoque(2011) and Reesink(2014), such insubordinate clauses express the speaker’s epistemic attitude. Reesink notes that Usan sentences that end in -eng “convey the speaker’s attitude of assertiveness, indignation, or surprise” (Reesink 2014:249). The examples in (171) and (172) illustrate the occurrence of te sentence-finally in Yonggom Wambon and Digul Wambon; in (171) the sentence-final te adds assertiveness to the crocodile’s demand, while in (172), sentence-final te marks a clause which is added to the main clause as an afterthought, strengthening it.

(171) Maturu te nu mene ŋga ok natin-in nde-t te. Ok come.up CONN 1SG here ERG river row-IMP[SG] say-REAL CONN river ti-no mboke-t, ran ya mimirop kup row-SIM stay-REAL[NON1SG] woman 3SG.POSS menstruation.blood with undu-t. go.across-REAL[NON1SG] “‘After he came up, he said: row me across the river here!’ While it (the canoe/crocodile) was being rowed across the river, a woman is menstruating, it (the canoe/crocodile) goes across.’ Yonggom Wambon (Drabbe 1959:§114) 11.3. Subordination 167

(172) Malin loxa-t-po; ap=we malin.bird speak-REAL[NON1SG]-PST; house=TOP wasatmo-ni=nde. tomorrow.be[IRR.NON1SG]-INTENT=FOC ‘A malin bird had spoken; a day was about to break.’ Digul Wambon (Jang 2008:127)

Although de Vries(1986) shows how focus marker te became a nominal coordinator, he does not describe well how he thinks a focus marker became a coordinator after Awyu-Dumut verbal clauses. Chapter3 contains a summary of de Vries’ ﬁndings concerning te after nominals. It was described how a focus marker can develop into a coordinator after noun phrases; because (coordinated) noun phrases are rare in Awyu-Dumut languages, they are always followed by a (salient) focus marker, which is then associated with the coordination of nouns. From this analysis de Vries then comes to the conclusion that te is a coordinator when it occurs after Awyu- Dumut verbal clauses as well, labeling it a case of further grammaticalization.10 However, as was pointed out in Chapter3, the analysis of te as a nominal coordinator is problematic. Furthermore, even if the analysis of te as a nominal coordinator is deemed plausible, the ‘saliency argument’ which applies to nominal phrases does not apply to coordinated verbal clauses in Awyu-Dumut languages; they are quite frequent, and so it is unclear why te should be analyzed as a coordinator in this context. Neither is it clear how a verbal coordinator would be a further, more abstract, grammaticalization of a nominal coordinator. Rather, the data contained in Drabbe’s and de Vries’ grammars support an analysis of te as a focus marker and do not necessitate hypothesizing about any diachronic development. To summarize, te is best analyzed as a subordinating conjunction rather than a coordinating conjunction because (1) the scope of negation and imperatives probably do not extend over clauses marked by te, (2) te occurs sentence-ﬁnally and (3) it is not clear how a focus marker would develop into a clausal coordinator in Awyu-Dumut languages. Furthermore, and very importantly, Drabbe notes that the addition of te to a verbal clause renders a it into a nominal.11 Nominalization is a characteristic shared by all Awyu-Dumut subordinating morphemes. One question remains: how can a focus marker function a subordinator, render- ing the clause that it follows into a dependent background clause that supports the main clause? Concerning morphemes which, like Dumut te, occur on both nominals and clauses, Reesink notes:

In many languages a medial verb desinence12 or following connector is

10The term ‘further grammaticalization’ is used by Heine et. al. to denote the development of a lexical item into a grammatical item, after which that grammatical item attains yet another, more abstract grammatical function. 11“There is no meaningful difference between the construction with and without te, but now one has two sentences, the ﬁrst is nominal, the second verbal” in Dutch: “Er is geen verschil van betekenis tussen de constructies zonder, en die met te, maar men heeft nu een voor- en een nazin, de eerste nominaal, de tweede verbaal” (Drabbe 1959:134). 12‘Medial verb’ here means ‘non-ﬁnal verb’, not necessarily ‘dependent verb.’ 168 11. Clause Linkage

used that can also appear on nominals. This is the case with -ma in Fore, and se in Waskia. The impression one gets from such nominal markings on clauses is that these clauses have more of a presuppositional status than other medial clauses. This has actually been proven by Haiman for Hua, and it would be more than likely that this would hold for the other languages as well. (Reesink 1983:223-224)

The morphemes that Reesink refers to here are, however, not focus markers like te. It seems strange at ﬁrst sight that, like these nominal markers in Usan, Hua, Waskia and Fore, a focus marker could mark clauses as presupposed or subordinate, since clauses that are in focus usually present salient, new information. However, evidence from multiple languages shows that focus markers can be used in structuring clauses. Bril(2007, 2010) writes about both topic markers and focus markers that are used as subordinators. She states the following:

The topicalised frame clause is the anchor for the propositional content of the other clause, which contains the main information and assertion; by contrast, the focused restrictor clause contains the highly salient information and assertion, which restricts the truth value of the other clause and its propositional content. (Bril 2007:224-225)

In other words, clauses that are subordinated by a focus marker restrict the circumstances or conditions under which the main assertion can be made/uttered/held as true. These focused restrictor clauses are subordinate because they are not a part of the main assertion, although they are necessary to understanding it, providing essential context. Thus, the Dumut focus marker te, when attached to a verbal clause, can be analyzed as functioning as a restricting subordinator. In other words, te clauses present a restricting context or background setting in which the main assertion, that is the assertion made by the ﬁnal clause, is situated.

11.3.5 Relative clauses Within the grammar descriptions and texts on Awyu-Dumut languages, only the two grammars on Digul Wambon and Kombai by Lourens de Vries contain any information on, or examples of, relative clauses. They are discussed here. In Digul Wambon, relative clauses look much like the thematic clauses discussed in Section 11.3.1, ending in a demonstrative topic marker =e or =eve. In addition, what de Vries & Wiersma(1992) call a relative clause delineator, -a, appears on the ﬁrst constituent of any Digul Wambon relative clause. The examples given in (173) and (174) illustrate Digul Wambon relative clauses, with the relative clause indicated in square brackets.

(173) Nuk [oy-a temke-n-o kav=e] hetak-nok-knd-ep. I pig-RC shoot-tr.nasal-CONN man-TOP see-NEG-REAL-1SG. ‘I do not see the man who shoots the pig’ Digul Wambon (de Vries & Wiersma 1992:56) 11.3. Subordination 169

(174) [Kaw-a ŋguw=e alip=ka ŋgaluma-t-po-n=ewe] man-RC you-TOP yesterday=ERG meet-REAL-PST-tr.nasal-TOP na-net-ndi. 1SG.POSS-older.brother-FOC ‘The man who you met yesterday is my older brother.’ Digul Wambon (Jang 2008:105) In Kombai, relative clauses end in ro, the lexical word for ‘thing’, or romu, the lexical word for ‘person’, depending on whether one is talking about a thing such as a bush knife (175) or a person, such as one’s uncle (176). (175) Gana [gu fali-x-a ro] na-gana-y-a. bush.knife 2SG carry-go-REAL-NON1SG thing 1SG.POSS-bush.knife-LIG-FOC ‘The bush knife that you took away, is my bush knife.’ Kombai (de Vries 1993:77) (176) Yare [gamo xereya bogi-n-o rumu] old.man join work DUR.do.REAL.NON1SG-tr.nasal-CONN person na-momof-a. 1SG.POSS-uncle-FOC ‘The old man, the person who is joining the work, is my uncle.’ Kombai (de Vries 1993:77) Besides occurring with relative clauses, Kombai ro also occurs after adverbial clauses, functioning as a subordinator (177, 178). (177) Uni berino-n-o ro, na-büwogo gamo Uni DUR.make.REAL.NON1PL-CONN thing 1SG.POSS-parent join rino. make.REAL.NON1PL ‘When they built Uni, my parents joined the work’ Kombai (de Vries 1993:82) (178) Mü bo-xe-re13 nu meda-n-e-madü. rain DUR-be.REAL.NON1SG-thing I come-tr.nasal-FUT-NEG ‘If it rains, I shall not come’ Kombai (de Vries 1993:58) Kombai conﬁrms Foley’s remark that Papuan languages tend to have one subordination strategy that covers both relative clauses and adverbial clauses (Foley 1986:201). Unlike Dumut languages, Kombai does not use demonstratives frequently to form subordinate clauses, although the near deictic mene may follow an adverbial clause ending in ro, as in (179) and (180). (179) Gu gwari muno gu ugi-ro˜ mene gu 2SG snake Patola 2SG bite.REAL.NON1SG-thing here 2SG fe-xumalene-madü. NEG-die.FUT.2SG-NEG ‘If a Patola snake has bitten you, you will not die’ Kombai (de Vries 1993:55) 13In this example, ro has cliticized to the verb boxe and vowel harmony has occurred, resulting in re. 170 11. Clause Linkage

(180) Amaxalo xumelei-ro mene xwaimigi waluwano:... Amaxalo die.REAL.NON1SG-thing this foreigners say.REAL.NON1PL ‘Concerning the death of Amaxalo, the foreigners had already said...’ Kombai (de Vries 1993:57)

When mene occurs after ro, de Vries glosses it as FRM ‘frame’, which he explains expresses the meaning ‘given that...’ (de Vries 1993:55). Thematic clauses formed with demonstratives in Dumut languages can also be translated as ‘given that’. Kombai adverbial clauses with ro mene are very similar to these Dumut thematic clauses, and both are examples of subordination strategies also found in many other Papuan languages.

11.3.6 ‘Conditionals’ In Mandobo, Aghu and Shiaxa, sentences with conditional meaning occur without a subordinator, constructions which Drabbe terms ‘conditionalis’. In Mandobo, these ‘conditionals’ consist of a realis -t form in the protasis, and an optative (181) or imperative verb (182) in the apodosis.

(181) Murüp ŋgwande-t kea-on. rain stop-REAL[NON1SG] go-1PL ‘If it stops raining, let us go.’ Mandobo (Drabbe 1959:32)

(182) Ro menda-r-on noa. thing bring-REAL-NON1PL give.IMP ‘If you brought something, give it (to me).’ Mandobo (Drabbe 1959:32)

A similar construction exists in Aghu; in a sentence with conditional meaning the protasis contains an optative verb, while the apodosis contains a future tense form (183, 184).

(183) Fiko af˜ı, sumke edaxe-y-e. work take.NON1SG tobacco give[1SG]-LIG-FUT ‘If you work, I will give you tobacco.’ Mandobo (Drabbe 1959:23)

(184) Dü fü baxe˜ geme-y-e. sago some stay.NON1SG buy[1SG]-LIG-FUT ‘If there is some sago, I shall buy it.’ Mandobo (Drabbe 1959:23)

For Shiaxa, Drabbe states that conditional sentences consist of an apodosis containing a future tense form and a protasis containing a future tense form.14 However, Drabbe does not provide any Shiaxa examples. The assertion that these Mandobo and Aghu constructions contain subordinate clauses should be questioned. Their form is unlike that of all other subordinate

14“In Sjiagha vinden we in conditionele zinnen de gewone futurumvorm zonder meer in voor- en nazin.” (Drabbe 1950:127) 11.4. Tail-head Linkage 171 clauses discussed in the sections above because they lack an explicit subordinator. Instead they have the same form as juxtaposed coordinate clauses. Reesink (1987:239-240) notes that some Usan coordinate clauses can be translated as having a conditional interpretation, but that that does not make them subordinate; they are syntactically different from subordinate clauses. Thus these Mandobo and Aghu constructions consisting of juxtaposed clauses are better analyzed as coordinate rather than subordinate. However, since Drabbe labels these constructions as ‘conditionalis’, I have included them here in the section on subordinate clauses.

11.4 Tail-head Linkage

In Awyu-Dumut narratives, sentences are commonly linked by tail-head linkage, in which the final verb of the first sentence is repeated at the beginning of the second sentence. By repeating the verb, tail-head linkage gives speakers the time to process the new sentence and hearers the opportunity to process the information in the previous sentence (de Vries 2006:817). In addition to processing ease, three other functions of tail-head linkage are referential coherence, thematic continuity and thematic discontinuity (de Vries 2005:363). Within Awyu-Dumut languages there are two types of tail-head linkage: chained tail-head linkage and thematized tail-head linkage, the first occurring with coordinated or clause-chained clauses in the head, and the second with thematic clauses in the head. Three instances of chained tail-head linkage from Kombai, Aghu and Mandobo are illustrated in (185-187). In the Kombai example (185), the head consists of a semi- finite verb followed by a coordinating connective -a, while the Aghu and Mandobo heads each contain a non-finite SS verb. In addition, the Mandobo head contains a noun that was part of the tail of the previous sentence. Chained tail-head linkages carry referential coherence mechanisms and event sequencing mechanisms across sentence boundaries (de Vries 2005:363). The story line is continued, but the speaker and hearer are given some time to process what is being communicated.

(185) Lenamalino-n-a büwogo xede mene marofora descend.REAL.NON1PL-tr.nasal-CONN old.people other TOP next Fiyabo-xu Ola-xu arabumano. Fiyabo-COORD Ola-COORD stand.in.between.REAL.NON1PL [Arabumano-n-a] stand.in.between.REAL.NON1PL-tr.nasal-CONN bo-lomõ-g-eno-n-a DUR-refuse.REAL.NON1PL-tr.nasal-CONN... ‘They descended and other older people stood between Fiyabo and Oba. They stood in between but they refused...’ Kombai (de Vries & Wiersma 1992:109) 172 11. Clause Linkage

(186) Idi küo-x-enã. [Küo-do-k] a-de-k büsyü tuber dig-REAL.NON1-NON1PL dig-SS-CONN take-SS-CONN house da-x-enã. come-REAL.NON1-NON1PL ‘They dig tubers. They dig (tubers) and take them and come home.’ Aghu (Drabbe 1957:62)

(187) Orat ke-r-an do, korü ko ŋgetap ti-gen. light be-REAL[NON1SG]-PST CONN go.down go hut build-REAL[NON1SG] [Ngetap ti] womo mbegi-ri itigio-gen do, u hut build guard sit-NON.CLOSE see-REAL[NON1SG] CONN pig me-gen. come-REAL[NON1SG] ‘When it was light, he descended, went and built a hut. He build a hut, guarded it and saw that a pig came.’ Mandobo (Drabbe 1959:17)

In thematic discontinuous tail-head linkage, the head is a thematic clause that marks a break in the story line and precedes a clause with background information. The theme or story line is then picked up again in the next sentence. The head is marked by a thematic clause marker te (188) or =eve (189). The verb in the head of a thematic discontinuous tail-head linkage is always a semi-ﬁnite or a ﬁnite verb form. The- matic discontinuous tail-head linkage does not occur in Aghu, as it does not have thematic clause markers. Thematic discontinuous tail head linkages are also not attested in Kombai.

(188) Omboitagumop taeŋgamo ra mari ndokma-r-in-an. type.of.plant cut.down take come.down block-REAL-NON1PL-PST [Ndokma-r-in de] og e kojake-r-an. block-REAL-NON1PL CONN river TOP get.full-REAL-PST Ema-t te ragae kup ke-r-an. thus.do-REAL[NON1SG] CONN ﬁsh with become-REAL-PST ‘They cut omboi-tagumop, bring it down (to the river) and block (the river). Because they block it, the river rises. That happened and ﬁsh came (into the river).’ Yonggom Wambon (Drabbe 1959:146)

This example contains two instances of tail-head linkage. The ﬁrst instance is ndokma- r-in, a recapitulated verb. A more generic verb with meaning ‘thus doing’ or ‘thus happened’ may appear in tail-head linkages instead of a repetition of the previous sentence’s ﬁnal verb, as illustrated by ema-t in the above example. Both heads in this example are subordinate thematic clauses marked by the subordinating focal connective te. 11.5. Awyu-Dumut Switch Reference Systems 173

(189) Koiv=o talom=o mben=o waxol=eve last=CONN year=CONN seven=CONN month=TOP Mbonop-ŋgambun=ka mbaxe-mbel=o ŋgerkayi lavo-va Mbonop-maelstrom=ERG stay-SEQ=CONN saw take-[IRR]1PL ne-mbel=o ep=ka mba-l-eva-mbo. [Ep=ka say-SEQ=CONN there-ERG stay-REAL-1PL-PST there=ERG mba-l-eva-mbo-n=eve] sanov=e ilo ka-l-eva-mbo. stay-REAL-1PL-PST-tr.nasal=TOP Monday=TOP go.down go-REAL-1PL-PST ‘Last year in the seventh month we stayed at the Mbonop maelstrom to saw, there we stayed. Given that we stayed there, on Monday we went down there.’ Digul Wambon (de Vries 1993:87)

Tail-head linkage is a widespread linking strategy in Awyu-Dumut languages, but also in Papuan languages in general. Hence its occurrence in Awyu-Dumut languages can best be explained as an instance of areal spread, rather than as a shared construction reﬂecting genealogical relatedness. Having said that, tail-head linkage might have already spread at the points in time when Proto Awyu, Proto Dumut and Proto Awyu-Dumut were spoken, and might, therefore, have occurred in these proto languages as well.

11.5 Awyu-Dumut Switch Reference Systems

Switch reference is a common feature of Papuan languages, particularly of Trans New Guinea languages, and can be defined as “an inflectional category of the verb, which indicates whether or not its subject is identical with the subject of some other verb” (Haiman & Munro 1983:x). Subject in Awyu-Dumut languages is defined as that element of reality that is referred to by the person-number marker on the semi- finite or finite verb.15 Reesink(1983) argues that, in Papuan languages, it is the topic or most salient NP that is traced by switch reference, rather than the subject per se. This is true for Awyu-Dumut languages as well. Roberts(1997) offers an extensive yet lucid description of Papuan switch reference systems, based on a sample of 169 Papuan languages. He found seven different strategies for marking switch reference in Papuan languages (Roberts 1997:136). Awyu-Dumut languages more or less employ what Roberts calls Strategy (iv), in which the absence or presence of a subject (person-number) marker marks the distinction between same subject verbs and different subject verbs: same subject verbs do not have a subject marker, while different subject verb forms do. In Roberts’ sample, which did not include any Awyu-Dumut languages, ten other languages were

15Recall that Awyu-Dumut languages have three types of verbs: non-finite verbs consisting of only a verb stem, semi-finite verb forms consisting of a verb stem, a mood marker and a person- number marker, and finite forms, which have a tense marker in addition to a mood marker and a person-number marker. 174 11. Clause Linkage found that use Strategy (iv). This switch reference system can be viewed as a basic or rudimentary switch reference system, as no specific switch reference morphology exists. The reality of Awyu-Dumut switch reference marking is slightly more complex than a simple distinction in degree of finiteness marking a difference between SS and DS verbs. Awyu-Dumut SS verb forms are straightforward: they are non-finite verb forms consisting of a verb stem and an optional temporality marker. They occur in clause chains and are dependent for their person-number interpretation on the next more finite verb. Awyu-Dumut DS forms are semi-finite and finite verb forms occurring sentence-medially. A clause which ends in a semi-finite or finite verb form is never chained to the next clause, because they (semi)-finite verb is an independent verb form. Thus the marking of switch reference is not restricted to clause chains in Awyu-Dumut languages, although clause chaining is a prerequisite for developing a switch reference system (Roberts 1997:159, Foley 2000:384). Furthermore, Awyu-Dumut semi-finite and finite verbs do not always have a different subject than the next verb, but only in specific contexts may they have the same subject. Awyu-Dumut (semi)-finites may only have the same subject as the verb in a following clause if one of two conditions apply:

1. when one event is fully completed before the next event 2. when a clause contains an explanation or a further speciﬁcation of the previous clause

These two conditions are further explained and illustrated in Section 11.5.2. Before these semi-ﬁnite same-subject conditions are considered, Awyu-Dumut non-ﬁnite same subject verb forms are extensively discussed in Section 11.5.1. I will then show in 11.5.3 that switch reference even occurs in Awyu-Dumut languages when one clause is subordinate to the next clause. Switch reference functions somewhat differently in Digul Wambon than in other Awyu-Dumut languages, and therefore Digul Wambon switch reference is discussed separately in Section 11.5.4. Section 11.5.5 offers a summary of Awyu-Dumut switch reference systems, as well as some thoughts on the diachronic development of these systems, or the lack thereof.16

11.5.1 Same subject verb forms

All Awyu-Dumut languages have SS verb forms, which consist of a verb stem with no morphological expression of mood, person-number or tense. Awyu SS verb stems are always marked with a SS marker, either -di or -bu. Awyu-Dumut SS forms occur in contexts in which the following verb has the same subject. Concerning these forms in Aghu, Drabbe notes that they occur “in one sentence in which two or more processes are expressed which have the same subject; the last process is then expressed by a verb form which indicates the tense of all the processes, and which has

16Note that nothing is known about the full switch reference systems of Pisa, Shiaxa and Yenimu, as no texts are available for these languages. 11.5. Awyu-Dumut Switch Reference Systems 175 a subject marker.”17 Thus the final verb has the tense and subject marking on which the preceding SS form or forms (called ‘deelwoorden’ or participles by Drabbe) depend. Although SS forms are dependent on a following more finite verb form, Drabbe notes that they occur in coordinate sentences: “Such forms occur where we in Dutch have two sentences with one subject, which are connected with the coordinator ‘and.’ ” 18 These are sharp observations by Drabbe about Awyu-Dumut SS forms, as the phenomenon now known as switch reference was only first described by William Jacobsen 1967, long after Drabbe’s time. An example of a Yonggom Wambon SS form is sumo in (190), which depends on the semi-finite verb form kogen for its mood and subject person-number interpretation.

(190) Sumo ko-gen. pick.up go-REAL[NON1SG] ‘He picked it up and went.’ Yonggom Wambon (Drabbe 1959:132)

In (190) tense is not marked on the final verb, as it often is not in Awyu-Dumut languages. Rather, the expression of mood is far more important, and SS verb forms are dependent on the following verb for mood interpretation. In Chapter6 it was described how Awyu languages and Kombai have separate realis and irrealis verb stems, which means that a distinction in mood is marked on the verb stem, and so the SS form would not be dependent on the next verb for mood. However, Drabbe carefully notes that only the realis verb stems are used in SS verbs in Aghu, Pisa, Shiaxa and Yenimu, even if the final verb is in irrealis form. The same is true for Kombai, for which de Vries(1993:33) notes that the difference between dependent SS verb forms and infinitive verb forms is that dependent verb forms do not express mood, whereas infinitive verb forms mark the contrast between realis and irrealis mood.19 However, mood is realized to some degree in Pisa and Shiaxa SS verb forms, as the two obligatory SS markers in these languages (-di and -bu in Pisa, -d and -b in Shiaxa), mark a distinction in mood; -di/-d is used when the final verb is in realis mood, while -bu/-b is used when the final verb is in irrealis mood. The Shiaxa uses of SS markers -di and -bu are contrasted in (191) and (192).

(191) Mode-d, do raru-d, yoxoba ede-mak-ere. come.I-SS.REAL sago bake.I-SS.REAL 3PL give.I-DIST.PST-NON1SG ‘He came, baked sago, and gave them (some of it).’ Shiaxa (Drabbe 1950:110)

17 “De ‘deelwoorden’ worden gebruikt wanneer in één zin twee of meer processen worden uitgedrukt die hetzelfde subject hebben; het laatst genoemde proces wordt dan uitgedrukt door een werk- woordsvorm die de tijd aangeeft van het geheel der processen, en met een subjectwijzer” (Drabbe 1957:19). 18“Zulke vormen treden op wanneer wij in het Nederlands twee zinnen hebben met één subject, die verbonden zijn door het voegwoord ‘en’ ” (Drabbe 1959:132) and “de deelwoorden in Shiaxa treden vooreerst op waar wij nevenschikkende zinnen gebruiken” (Drabbe 1950:110). 19Note that de Vries(1993:33) actually does not talk about mood but about tense. However, remember that in Chapter6 on Mood, Kombai non-future vs. future contrast was reanalyzed as a realis vs. irrealis contrast. 176 11. Clause Linkage

(192) Mode-b, do raru-b, yoxoba edexo-ne. come.I-SS.IRR sago bake.I-SS.IRR 3PL give.II-[IRR]NON1SG ‘He will come, bake sago, and give them (some of it).’ Shiaxa (Drabbe 1950:110)

In Yenimu, only the SS marker -d occurs, while for Aghu, Drabbe notes that the distinction in function between di and bu found in Shiaxa and Pisa has faded, with no difference in meaning found between SS forms which occurring with -d and those occurring with -b (Drabbe 1957:18). Thus it can be concluded that Shiaxa and Pisa SS forms are dependent on the ﬁnal verb for only their person-number interpretation, while Aghu, Yenimu and Kombai SS forms, like Dumut SS forms, are also dependent for the category of mood.

Temporality markers on SS verbs

Besides SS markers occurring in Awyu languages, Awyu-Dumut SS forms can be marked by temporality markers and markers that express conceptual closeness or distance. For Pisa, Drabbe notes that “we only ﬁnd the participles [=SS forms] when two processes have the same subject, happen simultaneously and belong closely together.”20 As Pisa SS only occur in simultaneous and conceptually close context, they do not need morphemes to express these concepts. We could say that Pisa SS forms are quite restricted in their use and represent a stage that SS forms in other Awyu-Dumut languages have already passed. In all other Awyu-Dumut languages, SS forms occur in both simultaneous and sequential contexts. When the action expressed by a SS verb occurs simultaneously with the action of the following clause in Yonggom Wambon and Digul Wambon, the SS form is marked with a simultaneous marker -no, as illustrated in (193).21

(193) Yaxov=e ko-no ŋgom li-knd-e. they=TOP go-SIM song sing-REAL-NON1PL ‘While walking they are singing a song.’ Digul Wambon (de Vries & Wiersma 1992:20)

(194) ketmom i-no mandonanin. dance dance-SIM come.FUT.NON1PL ‘They will come here (while) dancing a dance.’ Yonggom Wambon (Drabbe 1959:133)

Other Awyu-Dumut languages do not have simultaneity markers, but sequence markers do occur in some of them. Mandobo has a sequence marker -ra, which occurs only with verbs of motion and posture (Drabbe 1959:14). Digul Wambon has a sequence marker -mbet (de Vries & Wiersma 1992:20), while Kombai has a sequence marker -fo (de Vries 1993:21) which may occur after SS verb forms. Furthermore,

20“In Pisa treffen we de deelwoorden alleen aan als twee processen hetzelfde onderwerp hebben, tegelijkertijd geschieden en nauw samenhoren” (Drabbe 1950:110). 21Note that the simultaneous marker -no is analyzed by de Vries & Wiersma(1992:20) as consisting of the coordinator -o and a transitional nasal. While noting that the linking element -o might indeed be the origin of the simultaneous form, I choose to extend Drabbe’s analysis of the simultaneous marker as being -no to Digul Wambon. 11.5. Awyu-Dumut Switch Reference Systems 177

Yonggom Wambon, Mandobo and Pisa have sequence markers that occur only with semi-finite verb forms, not with non-finite SS forms; these sequence markers are discussed in Section 11.5.2. In Aghu, most SS forms have an element -ke suffixed to them, often shortened to -k. Whenever two processes occur sequentially, this ke occurs. It is not a sequence marker, however, because it also attaches to SS forms that occur simultaneously with the next verb, as in (195).

(195) Enigomo-do-k da-x-enã. eat.iterative-SS-CONN come-REAL-NON1PL ‘(While) eating they came here.’ Aghu (Drabbe 1957:22)

Drabbe glosses -ke after SS forms as SCH ‘linking element’ and notes that it is the same element as the emphatic ke which can occur after nouns and which occurs with deictics.22 I gloss it as CONN ‘connective’, noting that it most frequently occurs in sequential settings.

Conceptual non-close marker Besides the temporality markers just described, in the Dumut languages and Kom- bai a morpheme can attach to non-ﬁnite SS verb forms that, for a lack of a better term, I will call a ‘conceptually non-close marker’. The term comes from de Vries & Wiersma(1992:20), who in turn borrowed it from Haiman(1983), citing Haiman’s deﬁnition of conceptual distance as follows:

First, two concepts are conceptually close to the extent that they share semantic properties (e.g., two verbs are closer if they share a common tense, mood, subject, object, or topic); second, two concepts are close to the extent that one is thought to affect the other (e.g. the conceptual closeness between a verb and its object varies with the transitivity of the verb); ﬁnally, two concepts are close to the extent that they are perceived as inseparable (e.g. there is a closer conceptual link between a possessor and an inalienably possessed object than between a possessor and an alienably possessed object.) (Haiman 1983:783)

Haiman further states that conceptual distance equals formal distance; if two linguistic forms are further apart (that is, obligatorily separated by morphemes), they are conceptually more distant than when they are juxtaposed. I admit that the term ‘conceptually non-close’ covers morphemes whose function is not entirely clear. In Digul Wambon, the conceptually non-close marker is -lo, occurring directly after a bare verb stem (de Vries & Wiersma 1992:19). In Yonggom Wambon, the cognate

22Aghu ke might best be analyzed as a topic marker, but Drabbe provides too little information for such an analysis to be argued for. 178 11. Clause Linkage equivalent is -ro (Drabbe 1959:132).23 Drabbe notes that -ro does not occur when the SS verb has an adverbial interpretation, as wagaemo does in (196). (196) Wagaemo rap-ken. do.good hold-REAL[NON1SG] ‘He held it well.’ Yonggom Wambon (Drabbe 1959:132) In Mandobo, the conceptual non-close marker is -re or -ro; it only occurs with posture and motion verbs, and Drabbe notes that it can be viewed as always marking sequentiality. Yet Mandobo -re/-ro is different from the Mandobo sequence marker -ra discussed above; Drabbe consistently glosses -ra as ANT (anteriority marker, the same as a sequence marker), whereas he glosses -re/-ro as IV ‘inﬁnitive’ consistently, even though it does not appear on all inﬁnitive forms (Drabbe 1959:14-15). He also glosses Yonggom Wambon -ro as IV, which indicates that it is cognate to Mandobo -re/-ro. Lastly, Kombai has a morpheme -ra that can be added to SS verb forms. De Vries(1993:21) describes Kombai -ra as a general coordinator which implies conceptual non-closeness. Since Kombai -ra is so similar to the Dumut non-close markers in form and function, I choose to reanalyze it as a non-close marker.

Reconstruction of SS morphology Table 11.1 contains an overview of the temporality markers and conceptual non- close markers that occur after SS verb forms in Dumut languages and Kombai; in Awyu languages, these categories are not marked on SS verb forms.

Table 11.1: Temporality and conceptual non-close markers found on SS verbs

SIMSEQNON-CLOSE

MAN – -ra -re/-ro YWB -no – -ro DWB -no -mbet -ro

KOM – -fo -ra

PD *-no – *-ro PAD – – *-rV

A simultaneity marker *-no can be reconstructed for Proto Dumut, which then was dropped by Mandobo (or at least not attested in Mandobo data). The sequence markers are too dissimilar to allow for reconstruction. A conceptual non-close marker *-rV can be reconstructed for both Proto Dumut and Proto Awyu-Dumut; it is not possible to resolve which vowel the original non-close marker had. It remains remarkable that the Mandobo sequence marker -ra is so similar in form to the Man- dobo conceptual non-close marker, and even exactly the same in form as the Kom- bai non-close marker. There is a similarity in function also: non-close markers will

23Note that Digul Wambon /l/ is cognate to Yonggom Wambon /r/; both can be traced back to Proto Dumut *t. 11.5. Awyu-Dumut Switch Reference Systems 179 never occur in contexts in which actions are temporally close and take place simultaneously. Thus it is quite plausible that a non-close marker came to be used as a sequence marker. Note, however, that this is not what happens in Digul Wambon and Kombai, which have each pressed another morpheme into use as a sequence marker; the origins of Digul Wambon -mbet and Kombai -fo are unknown.

11.5.2 Semi-finite same subject conditions When semi-finite and finite verbs occur clause-medially in Awyu-Dumut languages, they usually have a different subject than the (semi-)finite verb that follows them. Illustrated in (197-199) are the appearance of Aghu, Kombai and Yonggom Wambon semi-finites in their common DS context. Aghu has a separate DS marker, -ne, which may occur after semi-finite DS forms.24

(197) Xo-do-k i ting-g-e. Ti-ni-k a-bu go-SS-CONN bird shoot-REAL-NON1SG shoot-SS-CONN take-SS da-de-k büsyü osu-du-k efe n’apigi yoxo come-SS-CONN house go.up-SS-CONN 3SG POSS’boy 3PL ed-ox-e-ne e-ŋg-enã. give-REAL-NON1SG-DS eat-REAL-NON1PL. ‘He goes and shoots a bird. He shoots (it), carries it home, goes up, and their son gives it to them, and they eat (it).’ Aghu (Drabbe 1957:57)

(198) Bo-yadema-no-n-a kepalahansife xorabo marofo-ra DUR-meet-[REAL]NON1PL-tr.nasal-CONN headman wife go.up-SEQ xe-yale bo-luwa. 2SG.POSS-husband DUR-be.angry.with[REAL.NON1SG] ‘They held a meeting and the wife of the headman started to speak angrily to her husband.’ Kombai (de Vries 1993:110)

(199) Naerop taemba-r-in kima-r-an. Naerop shoot-REAL-NON1PL die-REAL[NON1SG]-PST ‘They shoot Naerop and he dies.’ Yonggom Wambon (Drabbe 1959:147)

However, sometimes clause-medial Awyu-Dumut semi-ﬁnite forms have the same subject as the next verb. This can only occur when one of two conditions are met: (1) the verbs must express sequential actions that have no temporal overlap or (2) the second verb must occur in a clause that further explains the clause in which the ﬁrst verb occurs. The following two paragraphs present detailed illustrations of these two conditions.

24Note, however, that this example is the only occurrence of ne in Aghu texts, so it is not widely used. 180 11. Clause Linkage

Two non-close, sequential actions Drabbe notes that when two processes ‘cannot be considered as one’, that is, if they are separate actions that occur sequentially with no overlap in time, then two semi- ﬁnite verb forms with the same person-number marking may occur one after the other.25 In (200), for example, there are four actions that occur one after another; a new action does not begin until the previous one has been concluded.

(200) [Wokife afamo-x-e], [sumke bamboo.pipe open-NON1.REAL-NON1SG tobacco bomoku-ŋg-e], [sibomo-x-e], put.in-NON1.REAL-NON1SG close-NON1REAL-NON1SG [kiﬁ-ox-e]. put.away-NON1REAL-NON1SG ‘He opened a bamboo pipe, put tobacco inside, closed (it) and put it down.’ Aghu (Drabbe 1957:19)

Like Aghu, Yonggom Wambon and Mandobo also use multiple coordinated semi- finite or finite verbs with the same subject to express processes that occur sequentially without overlapping. In Mandobo example (201), a pig is divided. Various parts of the pig are given to different people in order; first the upper body, then the lower body and then the two back legs are given away. The subject, which is the giving party referred to by the NON1PL marking on the three verbs, remains the same.

(201) [Tirambö mbo küap ndigio-gen-on]; [kegeman mbo, upper.body TOP human give-REAL-NON1PLlower.body TOP back.piece mün mbo, kondog ei mbo, küap ndigio-gen-on], [kondog TOP leg one.side TOP man give-REAL-NON1PL leg one.side, ei, omba mi küap ndigio-gen-on]. other again man give-REAL-NON1PL ‘They give the upper body to a man; they give the lower body, the back piece, one back leg to a man, another man again they give the other back leg.’ Mandobo (Drabbe 1959:42)

In Yonggom Wambon example (202), the subject first takes hold of some bark and after that gets some banana ashes. In a summing up of similar actions like this, semi-finite verbs can occur medially while the next semi-finite verb has the same subject.

25“In de gevallen dat twee processen niet als één kunnen beschouwd worden vinden we ook wel twee werkwoorden achter elkaar die in een tijdvorm met subjectwijzer staan” (Drabbe 1957:19). 11.5. Awyu-Dumut Switch Reference Systems 181

(202) Yuw=e kirigit watek rawa-t, tot wanin He=TOP type.of.sago bark take.hold-REAL[NON1SG] banana ashes rawa-t. take.hold-REAL[NON1SG] ‘He takes some kirigit bark, he takes some banana ashes.’ Yonggom Wambon (Drabbe 1959:151)

In Yonggom Wambon example (203), the main character takes a vine, attaches it to a tree and swings from one area to another multiple times. Each clause ends in a ﬁnite past tense form.

(203) [Ku-r-a torom ŋgama-r-a, ra go-REAL[NON1SG]-SEQ vine chop.off-REAL[NON1SG]-SEQ take me-ro mbonmo oro-ro wamkarok kurugut come-NON.CLOSE put.away put-NON.CLOSE type.of.tree up atigo-ro Wambon ku-r-a bind-NON.CLOSE Wambon go-REAL[NON1SG]-SEQ mende-r-an]; [Sagit ku-r-a come-REAL[NON1SG]-PST Sagit go-REAL[NON1SG]-SEQ mende-r-an]; [Kambon mende-r-a come-REAL[NON1SG]-PST Kambon come-REAL[NON1SG]-SEQ ku-r-an]; [Morop ku-r-a go-REAL[NON1SG]-PST Morop go-REAL[NON1SG]-SEQ mende-r-an]; [Suruk ku-r-a come-REAL[NON1SG]-PST Suruk go-REAL[NON1SG]-SEQ mende-ran]. come-REAL[NON1SG]-PST ‘He goes and cuts a vine and binds it in the top of a Wamkorok tree and goes and comes (swings) to Wambon, he goes and comes to Sagit, he goes and comes to Kambon, he goes and comes to Morop, he goes and comes to Suruk.’ Yonggom Wambon (Drabbe 1959:151)

Besides ﬁnite past forms, semi-ﬁnite -t forms marked by a sequence marker -a are found in (203).26 Yonggom Wambon verbs marked by this subject marker -a have the same subject as the next verb if there is no overlap in time between the two verbs, as is the case with the person swinging from place to place in (203). The occurrence of Yonggom Wambon -a is further illustrated in (204).

26The sequence marker -a never attaches to Yonggom Wambon semi-finite -ken forms, nor does it attach to non-finite or finite forms. 182 11. Clause Linkage

(204) Ema-t te ra ko mbukma-r-in-a thus.do-REAL[NON1SG] CONN take go cut.up-REAL-NON1PL-SEQ andi-r-in-an. eat-REAL-NON1PL-PST. ‘After having done thus, they cut up (the crocodile) and eat (it).’ Yonggom Wambon (Drabbe 1959:149)

Mandobo, like Yonggom Wambon, also has a sequence marker that only occurs after semi-finites which have the same subject as the next verb. Drabbe finds the occurrence of semi-finites in this context remarkable; he would have expected non-finites (Drabbe 1959:18). The Mandobo semi-finite sequence marker is -aro27 and only occurs in between (semi)-finite verbs that have the same subject, as in (205).

(205) Terep mbio-gen-aro, e-metip n’itŋgombün ﬁbre twist-REAL[NON1SG]-SEQ 3SG.POSS-daughter POSS’upper.arm kagö-ŋgen. bind-REAL[NON1SG] ‘He twists (a rope) from ﬁbre and he binds it to his daughter’s upper arm.’ Mandobo (Drabbe 1959:57)

The fact that two semi-finites can have the same subject when they occur in a sequential condition may be explained by the fact that Awyu-Dumut non-finite SS forms originally only occurred in simultaneity (SIM) conditions. This is still the case in Pisa where, as noted in Section 11.5.1, SS verb forms may only occur in clauses in which (a) the subjects of the two verbs are the same, (b) the actions occur simultaneously and (c) the actions are closely related (Drabbe 1950:110). Sequence conditions may hence be a context in which semi-finite SS verb forms in the Awyu-Dumut switch reference sytem ‘linger’. In Pisa, whenever two actions are sequential, two semi-finite verbs are used. If the subject of the two semi-finite verbs is the same, an element kidi (if the verbs are in realis mood) or kibu (if the verbs are in irrealis mood) appears in between the two verbs. This kidi/kibu element is a combination of a verb stem ki- ‘to be’ with the Pisa SS markers -di or -bu. The use of kidi in Pisa is illustrated in (206).

(206) de-x-i ki-di ro-xo-y come-REAL-NON1SG be-SS say-REAL-NON1SG ‘he came and said’ Pisa (Drabbe 1950:110)

Thus, the first condition in which semi-finites with a same subject may follow each other is when they express sequential actions. A sequential marker may appear on the first semi-finite verb. A comma after a semi-finite in Drabbe’s texts is an (intonational) indication that the semi-finite finds itself in a sequential SS condition.

27With non-ﬁnites, the sequence marker -ra occurs, as noted above. However, -ra only occurs with motion verbs and posture verbs; -aro is used with verbs that are not motion or posture verbs. 11.5. Awyu-Dumut Switch Reference Systems 183

Specification or explanation If a clause in a sentence is a further specification of, or explanation of, a clause directly preceding it, then semi-finite verbs with the same subject may occur in both clauses. The second clause can function as a separate sentence but is conjoined to the previous clause because the clauses are closely tied together in meaning; Drabbe uses a semi-colon to indicate the coordination of the two clauses. Semi-finite verbs occur in this SS context in Aghu, Yonggom Wambon and Mandobo. Drabbe, when giving example (207), writes that the second clause (xabã tümoxe) can be seen as a new sentence, in which the storyteller explains the previous clause or sentence.28 In the case of (207), the second clause specifies which part of his enemy the main character cut off.

(207) [Bodo kutoﬁ-di-k tüm-ox-e], [xabã tree.branch put.down-SS-CONN cut.off-NON1.REAL-NON1SG head tüm-ox-e]. cut.off-NON1.REAL-NON1SG ‘He put (his enemy) on a tree branch and cut off, the head he cut off.’ Aghu (Drabbe 1957:19)

In Yonggom Wambon example (208), the second clause speciﬁes the time duration of the ﬁrst clause, emphasizing that the chopping of the canoe took a long time.

(208) [Nde, konoi e ririmo-gon-in;] [tuma-r-in-a woŋgopon Thus, canoe TOP chop-REAL-NON1PL chop-REAL-NON1PL-SEQ long.time kegemo-gon-in.] be.iter-REAL-NON1PL ‘Thus, they chop the canoe; they chop the canoe and stay(work on it) for a long time.’ Yonggom Wambon (Drabbe 1959:149)

The Yonggom Wambon example (209) is like the Aghu example (207) in that it further specifies an act of violence. In this example, both clauses end in a past finite verb form rather than in a semi-finite verb form.

(209) Sagot ŋga me-ro kigum ŋga yan ani big.rat ERG come-NON.CLOSE club ERG his elder.sister i-r-an; itop rira-r-an. hit-REAL[NON1SG]-PST ground hit-REAL[NON1SG]-PST ‘A big rat came and with a club hit his elder sister; it hit (her into) the ground.’ Yonggom Wambon (Drabbe 1959:153)

Multiple Mandobo semi-ﬁnite -ken forms with the same subject can also appear in specifying contexts, like Aghu and Yonggom Wambon semi-ﬁnites. In (210), a

28“Xaba tümoxe kan men beschouwen als een nieuwe zin, waarin de verteller uitleg geeft van het voorafgaande” (Drabbe 1959:19). 184 11. Clause Linkage tree trunk that fell into the river has just turned into a crocodile. The main character of the story observes how the tree trunk/crocodile first comes above water, then moves towards the bank, and then comes out of the river. Each independent clause gives new specifics regarding the movement of the tree trunk/crocodile. The whole assertion means something like ‘it came above the water, in fact, it was coming towards the bank, in fact, it came on land’. (210) Riwop mbogo önö-gen; middle.of.river there.distant come.above.water-REAL[NON1SG] önö muŋ-gen; e göp mba come.above.water come.across.other.side-REAL[NON1SG] him self towards ro me-re kima-gen. take come-NON.CLOSE come.on.land-REAL[NON1SG] ‘it (a tree trunk turned crocodile) comes up in the distance in the middle of the river; it comes up and goes across to the other side, it comes towards him and comes on land.’ Mandobo (Drabbe 1959:53) Note that, in Kombai texts, the speaker often repeats himself, such as in (211). The second clause then does not offer an explanation or a specification of what was said earlier. Rather, the repetition appears to be a speech strategy employed by the speaker to allow himself more time to think. As Drabbe’s stories were clearly not recordings of natural everyday speech but of origin myths recounted slowly, his data misses such instances of such repetition. (211) Unafano-n-a unafano-n-a hit.REAL.NON1PL-tr.nasal-CONN hit.REAL.NON1PL-tr.nasal-CONN lenamalino. descend.REAL.NON1PL ‘Theyi hit themj and theyj went away’ Kombai (de Vries 1993:109) Thus the second condition under which semi-finites with the same subject may occur is when the second semi-finite verb occurs in an independent clause that further specifies the clause in which the first semi-finite verb occurs. Drabbe often sepa- rates these independent clauses with a semi-colon, noting that they can be seen as separate sentences.

11.5.3 Switch reference and subordinate clauses

In Section 11.3 it was discussed how the ergative marker ŋga (YWB, DWB) and the focus marker te (YWB, DWB, MAN) function as subordinators.29 It was shown that

29Note that Digul Wambon ka is not a part of the discussion in this paragraph because it does not occur after non-finites. For the same reason, Digul Wambon ndi (cognate of te) is also not a part of this discussion; it only occurs after semi-finites and finites and hence there is no evidence that it does not block switch reference. Therefore it may be assumed that like all other subordinators, Digul Wambon ka and te block switch reference. 11.5. Awyu-Dumut Switch Reference Systems 185

ŋga does not block switch reference; when it occurs after a non-finite SS form, that form is still dependent on the next (semi)finite verb form for its person-number interpretation. Thus etogoro in (212) gets its 1SG interpretation from the finite form raŋgandiyip, even though etogoro appears in a subordinate clause. (212) Endom etogo-ro ŋga raŋgandi-y-ip. enemy see-NON.CLOSE ERG shout-LIG-FUT.1SG ‘When I see the enemy, I will shout.’ Yonggom Wambon (Drabbe 1959:134) Reesink, who came across similar examples of the Usan subordinator -eng following dependent medial verb forms, notes that “medial (SS and DS) clauses followed by -eng... are shielded from various operators on the verb of the matrix clause: negation, illocution, nominalization. The categories of tense and subject person-number, however, are subjected to the general constraints exerted by the final verb on the preceding ones in a sentence” (Reesink 1994:104-105). Thus a subordinator does not necessarily exclude switch reference from occurring between a subordinate clause and its matrix clause. The same can be argued for the subordinator te; like Yonggom Wambon ŋga, it may occur after non-finite SS forms in both Yonggom Wambon (213) and Mandobo (214). (213) Matik te mende-r-a Ramut ogirit mba-gen. get.up CONN come-REAL[NON1SG]-SEQ Ramut waterfall sit-REAL[NON1SG] ‘After he gets up, he comes and sits (at) the Ramut waterfall.’ Yonggom Wambon (Drabbe 1959:148)

(214) Aŋge Ngou meto u e-nou mbökma-gen. Immediately Ngou TOP pig 3SG.POSS-mother cut.up-REAL[NON1SG] Mbökmo to ro-go to mbegi-ra to ŋ-gen-on. cut.up CONN take-go CONN stay-SEQ CONN eat-REAL-NON1PL ‘Immediately Ngou cuts up the mother pig. After they cut it up, carry it and sit down, they eat (the pig).’ Mandobo (Drabbe 1959:§45) In these instances, the non-finite verbs matik and mbökmo, rogo and mbegira are dependent for their person-number interpretation on the semi-finite final verbs ke-t and ŋgenon, respectively. The difference between te andŋga is that te also occurs after semi-finite verb forms, whereas Yonggom Wambonŋga only appears after non-finite SS forms. The question which then needs to be asked for te, which need not be asked for ŋga, is whether the semi-finite verbs preceding it have the same subject as, or a different subject than, the (semi-)finite verb in the matrix clause. Semi-finite verbs that occur before the subordinator te in Mandobo and Yonggom Wambon can occur in either SS or DS contexts. In Yonggom Wambon example (215) and Mandobo example (216), the semi-finite verb before te has the same subject as the conjugated verb in the matrix clause. However, examples such as Yonggom Wambon example (217) and Mandobo example (218), where the semi-finite verb preceding te has a different subject than the (semi-)finite verb in the matrix clause, are far more frequent. 186 11. Clause Linkage

(215) Kagup tawok ndugupma-r-an. Memema-r-in de people message send-REAL[NON1SG]-PST. Come.IT-REAL-NON1PL CONN mbukma-r-in-an. cut.up-REAL-NON1PL-PST. ‘Hei sent a message to the people. When they came, they cut (himj) up.’ Yonggom Wambon (Drabbe 1959:154) (216) Yemik tima-gen do kare e-gen do, irandöp trap make-REAL[NON1SG] CONN ready be-REAL[NON1SG] CONN tuber tötöma-gen-aro, yemik korup ro-gen. pull.out-REAL[NON1SG]-SEQ trap centre put-REAL[NON1SG] ‘After he makes a trap, after it is finished, he pulls tubers out (of the ground) and puts them in the middle of the trap.’ Mandobo (Drabbe 1959:34) (217) Ndokma-r-in de og e kayake-r-an. block-REAL-NON1PL CONN river TOP rise-REAL[NON1SG]-PST ‘When they block (the river), it rises.’ Yonggom Wambon (Drabbe 1959:146) (218) Kou mba-gen do, noŋgun’irandöp u meto ane just sit-REAL[NON1SG] CONN POSS.2PL’tuber pig TOP eat ndamo-gen ne-gen-on. finish-REAL[NON1SG] say-REAL-NON1PL “‘While you are just sitting (there), the pig is eating all our tubers” they say.’ Mandobo (Drabbe 1959:26) Thus another context in which semi-finite verb forms in Mandobo and Yonggom Wambon can have both the same subject as, or a different subject from, the next semi-finite or finite verb form is before the subordinator te. The te SS context is a special instance of the sequential SS context sketched in Section 11.5.2; the action described in the subordinate clause always precedes, and has no temporal overlap with, the action described in the matrix clause.

11.5.4 Switch reference and Digul Wambon Digul Wambon was mentioned little in Section 11.5.2 on semi-finites and same subject switch reference marking. This is because its semi-finite -kend and -t forms function differently from those in other Dumut languages. They by and large mark DS when they occur non-finally and outside of subordinate clauses. Digul Wambon -kend forms may only have the same subject as the following verb if they occur before a perception verb such as ‘see’ or ‘hear’, as in (219). (219) Mulo kilim-ke-knd-ev=o hetaxa-l-ep-o-n=e go.down pursue-SUPP-REAL-1SG=CONN see-REAL-1SG-PST-tr.nasal=TOP ko la-t-mbo-ŋgelo... there lay-REAL[NON1SG]-PST-CONN... 11.5. Awyu-Dumut Switch Reference Systems 187

‘I went down, pursued (the pig) and I saw: it lay there and...’ Digul Wambon (de Vries & Wiersma 1992:84)

Digul Wambon semi-finite realis -t forms always have a different subject from the following verb when they occur non-finally. Indeed, de Vries & Wiersma(1992) and de Vries(2010) claim that Digul Wambon -t forms are dedicated DS forms that never occur in other contexts than medially marking DS. However, data by Jang(2008) show that Digul Wambon -t forms may also occur sentence-finally, as in (220) and (221).

(220) Ewo aŋgai na-n-ayam=o amit=e in-ande-t. that dog POSS-tr.nasal-chicken=CONN young=TOP kill-eat-REAL[NON1SG] ‘That dog killed and ate my chicken’s young’ Digul Wambon (Jang 2008:57)

(221) Andemop-sixi laxop ka-t. food-PURP ﬁeld go-REAL[NON1SG] ‘He went to the ﬁeld for (in order to get) food’ Digul Wambon (Jang 2008:79)

This means that Digul Wambon semi-finite realis -t forms are independent, just like all other semi-finite verb forms in Awyu-Dumut languages. They are not dependent on a finite verb form. It is true that only past tense finite forms, not future tense finite forms, may follow realis -t verbs, but that is simply because speakers do not switch mood in the middle of a sentence, not because the semi-finite form is dependent for its tense interpretation. Thus Digul Wambon -t forms are not dedicated DS forms, but are much like other Awyu-Dumut semi-finites. They have in common with Kombai semi-finites that they are only attested in DS conditions. However, I do not exclude the possibility that more Digul Wambon data might reveal that Digul Wambon realis -t forms may also occur in one of the two semi-finite SS conditions described in Section 11.5.2, especially since the Digul Wambon corpus is smaller than the Kombai, Aghu, Man- dobo and Yonggom Wambon corpuses. Nevertheless, a hypothesis stating that switch reference is an emergent system in Awyu-Dumut languages and that the process is most complete in Digul Wambon, as set forth by de Vries(2010), cannot be entirely discounted. The fact that Digul Wambon realis -t forms do not occur when the following verb has the same subject must not be lightly bypassed. Digul Wambon could mark the endpoint of a diachronic path wherein Awyu-Dumut languages replace all semi-finites in SS conditions with non-finite verb forms.

11.5.5 Summary of Awyu-Dumut switch reference systems Awyu-Dumut languages have rudimentary switch reference systems in which the presence or absence of a person-number marker makes the difference between a SS verb and a DS verb. The DS verbs may occur outside DS contexts as well, when specific conditions are met. The occurrence of DS verb forms when the next verb 188 11. Clause Linkage has a same subject can diachronically be viewed from two angles. Either the Awyu- Dumut switch reference systems are constant, and DS verbs have always been allowed to occur in these two specific same subject conditions, or Awyu-Dumut switch reference systems are emergent, in which case non-finite verb forms are on their way to marking SS in all contexts, while two SS contexts remain in which their semi-finite and finite counterparts may still occur. I prefer the first angle, as I see too few differences in the switch reference systems of Awyu-Dumut languages to support an emergent hypothesis. However, the reader must keep in mind that an emergent hypothesis, as presented by de Vries(2010), could also account for the synchronic reality of Awyu-Dumut switch reference systems. Certainly the two main processes noted by de Vries(2010) as being responsible for the shaping of Awyu-Dumut switch reference systems, namely coordination reduction and frequency association, cannot be denied a place in a constant switch reference hypothesis. It is likely that coordination reduction, as described by Haiman & Munro(1983), led to the birth of SS verb forms, whereas frequency association then allowed those SS non-finite verbs to spread throughout the language system. These processes must, however, have taken place before Proto Awyu-Dumut split off from its Trans New Guinea sisters, as a switch reference system with SS and DS verb forms is reconstructed for Proto Awyu-Dumut. Switch reference systems are widespread in Trans New Guinea languages and are often seen as a typical feature of a TNG language. It is unlikely that (proto) Awyu-Dumut languages developed switch reference systems independently.

11.6 Summary

Two types of clause linkage exist in Awyu-Dumut languages: coordination and subordination. These are two ends of a continuum. There are multiple types of coordination, as well as of subordination. Types of coordination are (a) juxtaposition, (b) the use of a coordinator (-o in DWB and -a in Kombai) and (c) clause chaining. Types of subordination are (a) thematic clauses with demonstrative topic markers, (b) with subordinators ŋga (YWB, DWB) or te (Dumut languages), (c) relative clauses and (d) the existential verb ‘to be’ functioning as a subordinator. All of these types of clause linkage, in most instances with explicit coordinators or subordinators, can be reconstructed for Proto Awyu, Proto Dumut and Proto Awyu-Dumut. In addition, tail-head linkage constructions were described. Awyu-Dumut languages have a rudimentary switch reference system. Non- finite SS verbs occur throughout all types of clauses, whether they are linked through coordination or subordination. Semi-finite and finite DS verb forms may also occur when the next verb has the same subject, but only if (1) the two verbs describe fully sequential events, or if (2) one clause is a further specification or explanation of the previous clause. Coordination reduction and frequency association are processes that can account for the existence of switch reference in Awyu-Dumut languages, although Proto Awyu-Dumut is thought to have already had a switch reference sys- 11.6. Summary 189 tem resembling that of its daughter languages, in which a distinction in verb finiteness equals a distinction between SS and DS verb forms.

Conclusion

Awyu-Dumut languages had already been spoken by generations of people in the Digul River basin by the time Petrus Drabbe started his study of them in 1940. The current study, published nearly 75 years later, builds on Drabbe’s rich research and subsequent research on Awyu-Dumut languages undertaken by Alan Healey(1970), Bert Voorhoeve(2001, 2005) and Lourens de Vries (1992, 1993, 1997). The proto phonologies reconstructed by Healey(1970) and Voorhoeve(2001) provided the basis for the proto phonology and regular sound changes presented in Chapter2. The same chapter showed that Korowai cannot be considered an Awyu- Dumut language, as it is quite different from the four Awyu languages, the three Dumut languages and Kombai. However, Korowai is still hypothesized to be distantly related to Awyu-Dumut languages; it is not a language isolate. The regular sound changes that were established confirmed the subgrouping of Awyu-Dumut languages into four Awyu languages and three Dumut languages, as proposed by both Healey and Voorhoeve. Furthermore, on the basis of phylogenetic methods applied to a 430-item lexical word list, Kombai was subgrouped together with Proto Dumut, although a critical note will be made below concerning this subgrouping. Having a solidly reconstructed proto phonology allowed for the next step in unraveling Awyu-Dumut’s linguistic past; the reconstruction of its proto morphology. A rigorous, bottom-up application of the comparative method, while following the form-function constraint, paradigmaticity constraint and systematicity constraint, resulted in an extensive proto morphology of Awyu-Dumut languages, presented in chapters 3-11. The table below contains an overview of all the proto morphemes reconstructed for Proto Awyu, Proto Dumut and Proto Awyu-Dumut. I will highlight some of them. Reconstruction was at times not possible due to the languages having diversified greatly. For example, past tense markers were difficult to reconstruct for Awyu- Dumut languages due to the great variation in number and kind of past tenses in Awyu languages. The four Awyu languages each have two, three or four past tenses, and for none of these tenses was a morpheme reconstructible for Proto Awyu, let alone for Proto Awyu-Dumut. The three Dumut languages each have one past tense, but Digul Wambon has a different past tense marker than Yonggom Wambon and Mandobo; Yonggom Wambon and Mandobo reflect the past tense marker found in 192 Conclusion

Awyu-Dumut proto morphology

Proto Awyu Proto Dumut Kombai Proto Awyu-Dumut

nominal morphology nominal *ku *kup, *te xu *kup coordinator kinship plural *-gi *-Ngu(i) – *NgV marker possessive *-nV *-nV – *-nV marker pronouns 1SG *nu *nup nu(f) *nup 2SG * gu *Ngup gu(f) *Ngup 3SG **eke *yup, *eke xe *yup,*eke 1PL *nüku *nakup nagu *nakup 2PL *güku *Ngakup, *nakip nage *Ngakup, *nakip 3PL *yaku *yakup ya *yakup subject person-number 1SG *-e(fe) *-ep -e(f) *-ep NON1SG *-en *-n -n, -e *-en 1PL *-efan *-ewan -efo *-epan NON1PL *-enan *-enan -eno *-enan deictics near deictic *ne *me, *ne me(ne) *me, *ne far deictic *(e)wV *ep mofe(ne) *ep distant deictic *xa/*xo *kop maxo *kop tense and mood future tense *-e *-en/*-in -i, -e *-en/*-in past tense – *-an – – realis mood *-d (first person), *-t, *-ken -d (first person), *-t *-k (non-first -x (non-first person) person) irrealis mood *-∅ *-∅ -∅ *-∅ negation double negation *fa/*fe...(*de) – (fe)...do *pa...(*nde) single negation – *ndV – switch reference same subject *-bu (irrealis), ––– marker *-di (realis) simultaneity – *-no – – marker ‘non-close’ – *-ro -ra *-rV marker 193

Proto Dumut, a marker that could not be reconstructed to Proto Awyu-Dumut. Kombai simply has no past tense at all. Although it was not possible to trace how the Awyu languages acquired their multiple past tenses – or perhaps how the Dumut languages lost multiple past tenses – they tell an essential part of Awyu-Dumut’s linguistic history, as does all other morphological diversity found in these languages. A diachronic path occurs when various languages are at different stages of a grammaticalization path, and because Awyu-Dumut languages are so close, I expected at the outset of this research project to ﬁnd quite a few diachronic paths. However, just one diachronic path was discovered, namely the development of double negation in Awyu languages. Kombai and the four Awyu languages each represent a stage of the Jespersen cycle, which describes how languages develop double negation. Depending on which stage of the Jespersen cycle a language is in, it has obligatory or optional pre-verbal and/or post-verbal negators. The data used by de Vries(2010) to support another diachronic path, namely the emergence of switch reference in Dumut languages, were found to be open to another interpretation, namely that Awyu-Dumut languages have stable switch reference systems. It must be noted that both claims about Awyu-Dumut switch reference systems are defend- able using the data available, and further intonational data and research into the operation of switch reference systems in Papuan languages is needed to determine which claim is most valid. Although few diachronic paths across languages were found, many grammaticalization paths – such as the development of the far deictic ep into both a pronoun and a topic marker – were found in individual languages. Realis mood markers, and their reconstruction, provide yet another intriguing insight into the shared history of Awyu-Dumut languages, and reveal a shared innovation between Kombai and Proto Awyu. The fact that Proto Awyu and Kombai both have alternating mood markers -d and -k, while Dumut languages have two non-alternating mood markers -t and -ken, while Proto Awyu-Dumut has one realis mood marker *-t, should lead to a subgrouping of Kombai and Proto Awyu versus Proto Dumut. However, in Chapter2, Kombai was subgrouped together with Dumut languages based on phylogenetic methods applied to Awyu-Dumut lexical data. One way to account for this discrepancy in subgrouping is to say that Kombai’s lexicon followed a different path than its morphology; the Kombai lexicon might have become more similar to the (Proto) Dumut lexicon through borrowing, or it might differ signiﬁcantly from the (Proto) Awyu lexicon because Proto Awyu renewed a large part of its vocabulary, again through borrowing. Whether Kom- bai is better subgrouped with the Awyu languages or with the Dumut languages, or better left alone as an ‘in-between’ language, the different results from phylogenetics and the comparative method show the vital importance of applying multiple methods in unravelling a language family’s linguistic past, because only then can its full history be discovered.

Appendix A: Word List

The following pages contain a word list of 430 lexical items for all Awyu-Dumut languages and Korowai. The sources for this word list are Drabbe(1959) for Shi- axa, Yenimu, Pisa, Aghu, Yonggom Wambon and Mandobo, van Enk & de Vries (1997) for Korowai and personal communication with HongTae Jang for the Digul Wambon list, and Sun-Kyu Chi for Kombai. Additionally, the Digul Wambon and Kombai word lists given by Lourens de Vries to Voorhoeve for his 2001 Proto Awyu- Dumut phonology were consulted and incorporated. The word list is nearly complete, but a - indicates a speciﬁc lexical item is missing for a particular language. The word list also contains my cognacy judgments, based on the regular sound correspondences established in the chapter on phonology, and used for the phylogenetic analyses presented in the same chapter. An ‘L’ instead of a number indicates that the word is a known borrowing or loan word. Lastly, the orthography in this word list is different at times from the orthography found in the sources; in this word list, I follow the orthography as set out in Section 2.2 in the phonology chapter. Shiaxa Yenimu Pisa Aghu Mandobo Yonggom Digul Wambon Kombai Korowai Wambon

1. a few togo 1 tugu 1 ago 2 toxopomu 1 kene 3 kenae 3 ŋgalundei 4 (mo)folumo 5 pol 6 2. above syiri 1 xato 2 siri 1 si 1 mbotörö 3 koture 4 talaxop 5 ma(ku)runi 6 - 3. Adam’s apple muxoroboge 1 muroxomabegi 1 mi 2 müsobo 3 karomorop 4 ogarurop-mit 5 kalum 6 wanduxlo 7 - 4. afterbirth gutoxo 1 - togoro 2 amtütüne 3 mun tiŋgorop 4 ŋgum 5 amitlo nap 6 gou 7 - 5. alang reed bote 1 sekino 2 seso 3 - mböp 4 mbup 4 mbup 4 bafo 4 fiyo, kamEn 5 6. all monoxo 1 munoxo 1 inimu˜ 2 isiomu 3 mep 4 amukmo 5 mitkap 6 imimo 7 xo(l)funE 8 7. animal tail mini 1 wobu 2 wobu 2 wobugo 2 wombüt 2 wambit 2 wabit 2 minü/wabu 1, 2 xendep 3 8. ankle keno-boge 1 kino-begi 1 kito k˜ı 2 kito wako 3 kondok kerop 4 wan wogoi 5 kenop kelop 6 kudoklo 7 - 9. anus oto 1 oto 1 o to 1 o to 1 ten top 2 osop 1 lombatop 3 blukalu 4 - 10. armpit togo 1 togo 1 toxõ 1 bodo toŋgõ 1 taŋgon top 1 taŋgo top 1 taŋgotop 1 rogo 1 lenan 2 11. arrow arEbo 1 - tamin˜ı 2 dibisi 3 ŋgop 4 ŋgop 4 mesan 5 yagu 6 püx 7 12. ascend oto-d 1 oto- 1 su- 1 osu-k 1 törö- 2 turu- 2 matut- 3 oro- 1 lu 1 13. ash sena goto 1 sinaro 1 taki 2 sinako 1 kotep 2 kosep 2 ŋgandop 3 rinoxware/bura bulanom 2 1, 2 14. ask for roxo-d 1 roxo- 1 ro- 1 o-x 1 keŋgwa- 2 namepmo- 3 nukup ne- 4 luka 1 nExmo 5 15. ask question roxo-d 1 roxo- 1 ro- 1 o-x 1 tomamo- 2 rambamo- 3 loxo- 1 luka 1 laifo 4 16. avenge goto eme-d 1 gutu eme- 1 nasi emo- 2 kutagamu ame- 3 keaton karemo- 4 ŋgotap karemo- 5 maton lap- 6 kuro 1 - 17. axe gEtaxa 1 gitaxa 1 yama 2 - - - kambak 3 koba 3 xamba 3 18. back mu-boge 1 mu-begi 1 mik˜ı 2 mübã 3 mbüman 4 mimit 5 mimit 5 buma 4 banun 6 19. bake, roast raru-d 1 yunu- 2 du- 3 tomü-g 4 ondü- 5 undo- 5 nanu- 6 adü 5 indo 5 20. bald xeiba puni 1 pudia (L) xaibã- bagide 2 xabumu fede 3 - ŋgin saŋ nderan 4 mixip 5 muba 6 mux 7 21. bamboo sumbE 1 suba 1 fio 2 woki 3 ndüam 4 egop 5 hexop 5 feyo 2 fu 2 22. banana tu 1 tu 1 su 1 syü 1 tetkei, tyut 1 tit 1 sit 1 rü(l) 1 dendü 2 23. bark yi a xa 1 yi xa 1 wasoko 2 xa 3 kota 3 kotae 3 enop kat 4 molalu 5 keyo 6 24. bat piriri (L) piriri (L) besoxori 1 sumapio 2 mbumbwop 3 tomin 4 mbet 5 emu 6 loxetix 7 25. bathe oxo ki-nd 1 oxo ki- 1 e k˜ı- 1 oxo k˜ı- 1 og gin- 1 ok kim- 1 okimo- 1 aximofo 1 dadü 2 26. be awake wend e-d 1 wed e- 1 kadaxai gi- 2 xagide i- 3 ketamo mba- 4 kondan mba- 5 matutmo- 6 muro 6 xafen 7 27. be cooked ma ke-d 1 paxani ki- (L) mado- 2 pani-k (L) ando- 3 ndo- 3 kitmo- 4 wado-/boda 3,5 - 28. be full toxo bo-d 1 toko be- 1 tarE-ma ba- 2 tape-ki (L) tome- 3 koya ke- 4 hayakde- 5 ga- 6 - 29. be hungry ye ke-d 1 ya ki- 1 du ye gi- 1 ode ki-k 2 metarep ke- 3 menkok ke- 4 otwakmo- 5 dou finage 6 - 30. be pregnant amutu mudu metexi modu dasoxo 3 kanike ki-k 4 mun mba- 5 3 kumui ke- 6 kutmo- 7 kugu 8 xondulmeŋga 9 bo- 1 bo- 2 31. be satisfied mudu wu-d 1 modu wu- 1 taxa gi- 2 kokü ba ki-k 3 wendin timo- 4 wandin ke- 4 kitke- 5 fururabo- 6 - 32. be shy xa xoxomo-d 1 xasifu gi- 2 fofa gi- 3 e ki-k 4 tereŋget 5 tereŋget 5 awoke- 6 adura 7 - 33. be sick xa go re-d 1 xa go re- 1 kuxa gi- 2 kumaŋ gi-k 3 watek ŋgwamo- 4 yayu yan- 5 yamboloke- 6 yamblo 6 lep 7 Shiaxa Yenimu Pisa Aghu Mandobo Yonggom Digul Wambon Kombai Korowai Wambon

34. be silent fagate-nd 1 bodo ki- 2 faki ba- 1 siŋgi-k 3 ogage- 4 irukmo- 5 luk andimxe- 6 reakeneanena 7 pofuleli 8 35. be sleepy kero kono ŋge-d 1 kiru kunu gi- 1 kunu˜ gi- 1 kunuŋ gi-k 1 kinum eren ge- 2 kinum kok ke- 3 kinimxexe- 1 xunu-ge˜ 1 lenup 4 36. be slippery woforo 1 tatafuru 2 wugin˜ı 3 ifia 4 tendarak 5 ndaragae 6 ndalatke- 7 weinagane 8 fep 9 37. be thirsty mu bo-d 1 oxo ya ki- 2 e ye gi- 3 oxo ya ki-d 2 ok kurutmo- 4 ok yerep ke- 2 oktaxe- 5 oko finage 6 - 38. be unwilling mo-nd 1 mu- 1 gamu-˜ 2 mu-˜ 1 mbonemo/a- 3 ayukmo- 4 soeke- 5 bidoge 6 - 39. beard gando-boge- tere-begi-mo 2 masi-rõ 3 masü 3 matit 3 matit 3 maŋgot lon 4 maru 3 - moxo 1 40. belly mudu 1 modu 1 kaku 2 kokü 2 wendin 3 wandin 3 ot 4 a 4 xondul 5 41. below moka 1 moka 1 fu 2 maka 1 mbogorü 3 korire 4 hiptop 5 makaluni 1 - 42. big binigE 1 binigE 1 tEŋgaxai 2 tadi 3 koneni 4 kamae 5 hiwin/indip 6 muyano 7 tale 8 43. bile etimu 1 etimu 1 isimu˜ 1 isimu 1 men 2 men 2 mesen 2 gabugi 3 melun 2 44. bind ado-d 1 ado- 1 ada- 1 ada-k 1 rün- 2 andap- 1 andowomo- 1 ada-fede-fo 1 lamelo 3 45. bird yi 1 yi 1 yi 1 ˜ı 1 et 1 yet 1 et 1 el 1 delamol 2 46. bird of fodo 1 fodo 1 ife 2 efe 2 wando 1 andoi 1 suxum 3 fodo 1 beni 4 paradise 47. bird’s tail otoro 1 otoro 1 osirõ 1 osia 1 ataron 1 oi tet 1 teklon 2 wabu 3 xendep 4 48. bite ati-d 1 ati- 1 asu- 1 asi-k 1 tereni- 2 atigo- 1 asiŋgamo- 1 bogi- 3 - 49. black boxo 1 boxo 1 siri 2 sisi 2 ngürüp 3 kui 4 mimop 5 gunüharu 6 xofi(lun) 7 50. blind kero so 1 kiro so 1 kiro peyo 2 kiokone de 3 kerop tegop 4 kerop mitpan 5 kelop kolofu 7 - wandimop 6 51. blood goŋ 1 go 1 go 1 gõ 1 ŋgom 1 ŋgom 1 ŋgom 1 gom 1 büŋga 2 52. blunt ta boxoda 1 - bio 2 müsiã 3 - ngoŋ 4 toxot 5 flixi 6 peli 7 53. body adu 1 adu 1 adu-wasoxo 1 wusoxo 2 ondü 1 kotae 3 kalandit 4 da 1 loxül 5 54. bone boge 1 begi 1 bagi 1 bigi 1 mirap 2 mit 2 mit 2 fia 3 xolol 4 55. bow di 1 du 1 du˜ 1 du˜ 1 mbendit 2 sinam 3 mandit 4 badi 2 ati 5 56. brains gu 1 gu 1 gu˜ 1 gu˜ 1 mök 2 ŋgin-kawae 3 ŋgisiwop 4 mukokei 5 loxul 6 57. branch moto 1 kebi 2 kab˜ı 2 kabiã 2 rogonop 3 kitup 4 nandop 5 adia 6 dedamol 7 58. breadfruit yuato 1 yowo 1 yawo 1 xã 2 raot 3 rawot 3 lawot 3 aluwo 3 yawol 1 59. break rope, gene-d 1 geni- 1 geni- 1 geni- 1 ŋgwane- 1 ŋgande- 2 ŋgande- 2 - bamo 3 intrans. 60. break rope, rigamo-d 1 wugeme- 2 iŋgemo- 3 ageme- 3 roŋgwamo- 1 ra ŋgamo- 1 lap ŋgamo- 4 - - trans. 61. break stone, xaxa ne-d 1 - xarani- 2 poni- (L) karone- 2 karonde- 2 toke- 4 - - intrans. 62. break stone, waxamo-d 1 waxamo- 1 waxaramo- 1 üpomo- 2 - ra karomo- 3 lap tokmo- 4 - - trans. Shiaxa Yenimu Pisa Aghu Mandobo Yonggom Digul Wambon Kombai Korowai Wambon

63. break wood, xoxoi-d 1 - xa˜ı 1 gaxa-k 1 ŋgoŋgwan- 2 kande- 3 kapde- 4 - - intrans. 64. break wood, raxawu-d 1 axafu- 1 axafu-˜ 1 axafu-˜ 1 ragamo- 2 ra kaende- 3 lap kapde- 4 - - trans. 65. breast, o-me 1 o-me 1 õ 1 nõ 2 am 1 om 1 om 1 am 1 am 1 female 66. breathe fife-d 1 fifire- 1 sif˜ı fero- 2 f˜ı fi-ox 1 utünün mo- 3 upneŋ ke- 4 xinop- 5 finafo- 6 - 67. bring along red mode-d 1 - radi de- 1 ade da-x 1 timo me- 2 rap me- 1 lap ndok- 4 lefa me 1 - 68. broad atat 1 ata 1 aromaferã 2 tã 1 korok 3 arugup 4 indip 5 belu 6 - 69. build a fire regete-d 1 agotere- 2 y˜ı ngafu-˜ 3 yã abuku-˜ 4 in kön- 5 enoptenow enop humo- 7 e fukamo 8 melil fu 9 ugum- 6 70. burn raru-d 1 yinu- 2 yi du- 3 fi-ox 4 ondü- 5 undo- 5 enop sapmo- 6 adü- 5 do 5 71. burp begu ru-d 1 bagi ru- 1 mur˜ı gi- 2 megomo- 3 üŋgun orü- 4 mbaroŋgoi in- 5 mbiŋguimo- 6 - - 72. butterfly apa (L) apapipa (L) apero (L) apo (L) apap (L) awomburui 1 awonamayop 1 awoyo 1 - 73. buttocks obo 1 obo 1 o s˜ı 1 bubute 2 andüon 3 mbaem 4 kut-kut 5 bobo 2 - 74. calf kodo 1 kodo 1 kito-pi 2 kut baxamgo 3 wanomun 4 wan mom 5 amit halo- 6 mukafe 7 - 75. call ri-d 1 ri- 1 ri- 1 üfüaku-˜ 2 yo- 3 yo- 3 yo- 3 ya 3 dodepo 4 76. canoe xaya 1 xeya 1 yefu˜ (L) yofu˜ (L) yoün (L) konoi 2 alep 3 yafu (L) alEp 3 77. carry rakako-nd 1 akeku- 1 keku-˜ 1 keku-˜ 1 tombeni 2 sumo- 3 lapko- 4 lefa 5 abi 6 78. cassowary sawari 1 sawari 1 karuwe 2 woküe 3 mbaeŋgap 4 itit 5 sanip 6 muwayo(f) 7 sanip 6 79. centipede xaxa 1 xafe re 2 xae 3 xa 3 iŋgiarop 4 mamuririop 5 ŋgaluyop 6 amia 7 bebal 8 80. cheek moxo-boge 1 moxo-begi 1 moxõ-bagi 1 moxope 1 amoga 1 koyop 2 ayam amit 3 ma 4 maŋgum 1 81. chest mata-boge 1 mata-begi 1 b˜ı 2 besame 3 mbeman 4 mbemit 5 wamot 6 bema 4 ban 7 82. child, ka 1 ka 1 ofo 2 amoko 3 mun 4 mun 4 amit 5 kenodufo 6 mbam 7 descendant 83. child, young amutu 1 metexi modu ofo 3 amoko 4 mun 5 mun 5 amit kewet 6 miyo mogo 7 mbam 8 one bo- 2 84. chin gando-boge 1 tere-begi 2 makabi 3 te-bigi 2 ŋgandöm 4 maŋgor-ip 5 maŋgot 5 gadu 1 - 85. chop tumu-d 1 tumu- 1 tumo- 1 ü-k 2 orü- 3 taeŋgamo- 4 samo- 1 xalo 5 di 6 86. chop down ru-d 1 ru- 1 ru- 1 ü-k 1 orü- 1 ri- 1 ŋgapmo- 2 klumo 3 ü(l) 1 87. close eyes ker’abu-nd 1 kenagabu- 2 kinabu- 1 kio bu-˜ 1 kerop kerop mitke- 3 kimbikmo- 4 kloro kubuni 5 - kumbugne- 2 88. cockroach gonogoŋ 1 tubiri 2 ambusumi 3 xabgibo 4 kagerak (L) kakarak (L) tambat 2 lio 5 labul 2 89. coconut yaŋgata 1 pe begi (L) beraxai 2 peyo bigi (L) mbian mirap 3 mbian ndok 5 - biyo 3 mbiyon 3 90. cold toru 1 toru 1 taru 1 tü 1 ŋgit 2 sarui 1 salon 1 lo-xe 1 xagil 3 91. come mode- 1 mede- 1 de 2 da-x 2 me- 1 me- 1 nda- 2 me- 1 lai 3 Shiaxa Yenimu Pisa Aghu Mandobo Yonggom Digul Wambon Kombai Korowai Wambon

92. corpse taxa 1 - yuxu su 2 saxu akümã 3 küap kümüran 4 kagup kimiran 5 tuk 6 xowaru 7 loxül 8 93. correct, true kaxamo 1 aga xeya 2 agani 3 xamiki 4 kogoröm 5 kareop 6 mop 7 maxano 8 mofexup 9 94. cough ow ti-d 1 uwo ti- 1 eso si- 2 adimo si-k 3 oüok tü- 4 ayak ti- 4 apyokmo- 5 ahüo 4 lemu 6 95. crocodile saxambo 1 aya 2 xaiambu 3 kiambu 3 kiambut 3 mbanep 4 mbuayop 5 romoi 6 semail 7 96. crooked boka 1 wimku 2 burud˜i 3 bubuni-k 4 ŋgombüt 5 ŋgoi 6 ŋgui-ŋgui 6 welene 7 fale 8 97. cross-eyed kero - kiro burud˜i 2 kiogo mete-ox 3 kerop kerop ŋgoi ke- 5 kelop ŋgui 5 kololigi 6 - robomoko-nd 1 regemburü e- 4 98. crowned moxosy 1 moxow 1 kutE 2 üküte 2 kutea 2 kotim 2 yawoe 3 feruwo 4 aülem 5 pidgeon 99. dark asyu 1 asu 1 gugunu˜ 2 asü 1 weminarep 3 ndembeŋ 4 mim 5 rekema 6 xofi(lul) 7 100. daughter otobra 1 otoba 1 subrã 1 subã 1 metip 2 matip 2 tombalan 1 lanomu 3 lal 4 101. daylight sera 1 madi 2 sawE 3 sowo 3 orat 1 waran 4 sat 1 rei 1 alüp 5 102. deaf toro boxoda 1 - suru peyo 2 suto de 3 keretop 4 mburümop 5 naŋgom 6 bulugi 7 - 103. deep koro 1 kuru 1 buru-˜ 2 bubogo 3 ŋguruop 4 ŋguru 4 ndinop 5 wake 6 maxol 7 104. defecate oto tomo-d 1 oto tomo- 1 asa su- 1 o tefi-ox 2 a ŋgo- 3 oi aŋgo- 3 lombat ŋgan 4 bara 5 - 105. descend mui-d 1 mui- 1 ru- 2 osü-k 2 korü- 3 ri- 2 lomo- 2 moyona 4 nai 5 106. die ku-nd 1 ku- 1 ku-˜ 1 ku-˜ 1 kün- 1 kim- 1 kim- 1 xumo- 1 xomu 1 107. dig ku-d 1 kuo- 1 ko- 1 kü-ox 1 to- 2 so- 2 aso- 2 xüho- 1 yaxü 3 108. do, make eme-d 1 eme- 1 emo- 1 ame- 1 ra- 2 onoŋnemo- 3 si- 4 fla- 5 mo, amo 1 109. dog se 1 se 1 aŋgi 2 yaŋgi 2 aŋga 2 aŋgae 2 agai 2 maxu 3 xendep 4 110. door abita 1 auti 2 afito 1 büshü to 3 yombutop 4 ahap piri 5 tiak 6 rafira 7 abüok 8 111. dream kono maxete axatu fite- 2 asã iku- 3 asü ete-ox 4 ö itigio- 4 ui eto- 4 asixitmo- 3 u ferao 5 - fete-d 1 112. drink mi-nd 1 mi- 1 mi- 1 mi-k 1 emi- 1 mi- 1 ami- 1 mi 1 mi 1 113. drum woto 1 wisi 1 wu˜ 1 kine 2 kinda 3 kendet 4 kinde 3 wei 1 - 114. dry kera 1 soxo 2 se 2 so 2 tomap 3 kerewet 4 sok 2 roxe 2 gololfobo 5 115. ear toro 1 turu 1 suru˜ 1 suketo 2 keretop 3 turutop 1 silutop 1 ruro 1 xotop 4 116. earth moka 1 moka 1 mokã 1 soxo ku-˜ 2 itiwa 3 itop 3 ip 3 i(f) 3 me 4 117. earthquake eriri 1 iriri 1 iriri 1 yei 2 ondön 3 indum 3 hendem 3 fedu 3 - 118. earwax toro-dEge 1 - surua 2 su-naŋge 3 keretow a 4 turutow-oi 5 silunot 6 rurukeya 7 - 119. eat e-nd 1 e- 1 n˜ı- 2 e˜ 1 ande- 3 en- 1 en/ade- 1,3 ade-fo, en- 1,3 le 4 120. egg gena 1 gana 1 mugo 2 mügo 2 wendin 3 wagot 4 wandin 3 indi 3 loxesux 5 121. elbow mo boge 1 mo begi 1 bidE-bu˜ 2 bodo mu 3 i ŋgombün 4 win ŋgambin 4 mbisit ŋgambim 5 i-gabü 4 laxa 6 122. empty boxoda 1 bEigida 1 bagide 1 fede ki-k 2 nda 3 ndoi 3 mbumba 4 deyalu 5 - Shiaxa Yenimu Pisa Aghu Mandobo Yonggom Digul Wambon Kombai Korowai Wambon

123. enemy kowE 1 kufE 1 waki gu˜ 2 kufe 1 endon 3 endom 3 awumxomop 4 kofe 1 xoxuxanop 5 124. enough yaxaro 1 - xarini- 2 xaini-k 2 kare- 2 kare 2 kit 2 mofeafene 3 - 125. evil kie 1 kia 1 yama 2 xad˜ı 3 kuyom 4 kuyom 4 koyom 4 gabürabo 5 lembul 6 126. extinguish wumbumu- 1 wubumu- 1 ibuma ku-˜ 1 yã übumu ku-˜ 1 in utegimo- 2 eoptenow litomo- 4 erakimona 5 - a fire itiguimo- 3 127. extinguishing buni-d 1 buni- 1 buni- 1 bünü-k 1 kün- 2 kim- 2 litoman 3 e rakimona 4 lemile 5 of fire 128. eye kEro 1 kiro 1 krimogo 1 kiomogo 1 kerop 1 kerop 1 kelop 1 xoro 1 lulop 2 129. eyebrows kinise-boge- kinti-begi-mo 1 kirimogo rõ 2 kio-buto 3 - kerop-toŋgot 4 kelop-lon 5 koblo 6 lulxolol 7 moxo 1 130. eyelashes kero-moxo 1 kiro-mo 1 kiro-bi 2 kio-b˜ı 2 kere-ron 3 kerop-ron 3 kelop-wambip 4 kloblonolo 5 - 131. face kato 1 kiroto 1 kEto 1 kioto 1 kerewã top 2 ŋgin-o-kerok 3 kelop-kalit 4 xeda 1 lulgelip 5 132. faeces or 1 o 1 o 1 o paxa 1 a 1 oi 1 lombat 2 a 1 ol(exi) 1 133. fart o syaxamo-d 1 o syamo- 1 o-ru ro- 2 o pani-k 3 ombüt 4 oi mbuk ke- 5 lombot ablema 7 - mbixitmo- 6 134. fast fadamo 1 xagei 2 xayo 3 xaxamu 2 oget 4 sukmae 5 wasin 6 wei 7 diof 8 135. fat (adj.) kobete kubi toxo- 1 tEŋgaxai 2 fofüko 3 tögö göp 4 kutugut 5 tuxutkup 4 iya 6 - toxomo-d 1 136. fat (noun) kepe 1 kepe 1 duwo 2 xadimie 3 tögö 4 tugui 4 tuxut 4 xia 5 yaben 6 137. father ayE 1 ayE 1 bo 2 eto 3 neti 4 nati 4 mbap 5 nare/are 3,4 ate 3 138. feathers yi moxo 1 yi mo 1 rõ 2 xa xõ 3 ron 4 ron 4 lon 4 ruwo 5 daya 6 139. feminine su 1 si 1 rã 2 syü 1 roŋgu 2 raŋgui 2 laŋgui 2 kuri 3 - 140. fence xEiti 1 xEisi 1 xaisi 1 boxi 2 ndara 3 kuk 4 alun 5 ba 6 - 141. find fete-d 1 fite- 1 feto- 1 a-x 2 timo/a- 3 rap- 2 ŋgalumo- 4 koloma 5 - 142. finger bedo toro 1 toro 2 bidE-torõ 1 bodo butõ 3 it kurugut 4 wit ketop 5 mbisit-talon 6 ilo 7 melol 8 143. fire yi ndo 1 yi do 1 y˜ı 2 yã 2 in ndumbut 3 enop-tenop 4 enop 5 e 6 melil 7 144. firefly se 1 si 1 m˜ı 2 m˜ı 2 minap 2 mindui 3 minop 2 mi 2 belil 4 145. firewood yi 1 yi 1 yi kadE 1 yã 1 in 2 enop kok 3 enop 4 ewodo 5 melil 6 146. fish axae 1 axae 1 axae 1 axe 1 rogoa 2 ragae 2 laxai 2 duwo 3 xele 4 147. five bed-afi-bumo 1 bed-afi-komE 2 bedaxafimã 3 bidikumu 4 oŋgugup 5 aŋgu(kup) 5 ambalop 6 abalo 6 wayo 7 148. flat, even tetede 1 fira 2 imuferã 3 tã 4 metap 5 mandon 6 kukmo 7 - - 149. flower kete 1 ki 1 ki 1 ki 1 ket 1 ket 1 ketek 2 xe 1 xel 1 150. fly (insect) obusiri 1 obose 1 ebu˜ 2 obeka 3 umonop 4 imonop 4 mbuluk 5 luru 6 lotup 6 151. fly (verb) boro-nd 1 buru- 1 buru-˜ 1 bu˜ xo-x 2 mberene go 3 ururuk ko- 4 mbuluk buxa 6 bedi(fo) 7 ko-/-nda 5 152. flying fox xonge 1 xoge 1 ŋgoŋgi 2 xoŋgi 2 yayap 3 towoi 4 tawot 5 emu 6 yemül 7 Shiaxa Yenimu Pisa Aghu Mandobo Yonggom Digul Wambon Kombai Korowai Wambon

153. foot,leg kitu 1 kito 1 kito 1 kito 1 kondok 2 kondok/wan 2,3 kenop 4 xudo/xino 2,4 bel 5 154. footprints kia 1 kia 1 kito bu 2 kuto isipomo- 3 ot 4 yugut 5 ixun 6 kudo bobou 7 betom 8 155. forehead mi-boge 1 mika-begi 2 kirobagi 3 mika 1 kimbin 4 kombisop 5 kimbin 4 kumiri 6 loxul 7 156. forest ado 1 ado 1 gasi 2 - toŋgwap 3 yagok 4 tamot 5 xuro 6 lop 7 157. forget goba-nd 1 gaba- 1 fimomo- 2 omumu 3 koŋgomo- 4 katoni ke- 5 amokmo- 6 emu 7 enontelo 8 158. four okom-okomo 1 okum-okum 1 soxondemã 2 sigiane 3 kurugutköp 4 kurugut(kup) 4 hitulop 5 woromibogo 6 wayafül 7 159. friend xabe 1 xabe 1 buri 2 apofe 3 namonop 4 kae 5 yaxe 6 na 7 noxo 8 160. frog bono 1 bono 1 kokoxo 2 naŋge 3 koak 4 kak 4 waknup 5 - - 161. fruit ro 1 ro 1 yindero 2 edeko 3 rop 1 rop 1 lop 1 dodo-lo 1 op 1 162. gaba-gaba yarEs 1 tagia 2 tawoxo 3 ditaŋge 4 mbeŋgi 5 mbaŋgi 5 tomak 6 dubuga 7 iŋgenun 8 163. garden matoxo 1 gia 2 giõ 2 xü 2 yogüp 3 yagip 3 laxop 4 yarimo 5 yasim 5 164. gather gigiko-nd 1 gigiku- 1 gigio ku-˜ 1 mumu- 2 rurukmo/a- 3 yaŋgumo- 4 yaŋgumo- 4 yademano 5 kelioxmo 6 165. get up dato-d 1 date- 1 doto- 1 doto-x 1 matere- 2 matigo- 2 matutke- 3 marule 2 melu 4 166. ghost, soul bona 1 bana 1 bana 1 bono 1 aterop 2 ndimndop 3 hinop 4 kwai 5 laleo 6 167. ghost, xogoi 1 xoxEi 1 xoxoi 1 xoxi 1 kugu 1 kagui 1 kaxui 1 xaixwo, xwai 1 xayan 1 spirit 168. give ede-d 1 ede- 1 edo- 1 ede-ox 1 ndigio- 2 yo- 3 ndak- 4 adia 1 fedo 5 169. give birth xeire-d 1 xei- 1 xai- 1 xaifi-ox 1 kap- 2 ŋgamo- 3 halo- 4 rimofa 5 fu 6 170. go away ago-d 1 age- 1 xo- 2 xo-x 2 ko- 2 ko- 2 ko- 2 ka 2 gelilfo 3 171. go get agod re-d 1 - xodi ra- 2 xod’abu da-x 3 ko timo/a- 4 ko rap- 5 ko lap- 5 - - 172. good xagoto 1 maxeya 2 popora (L) yafi 3 wonden 4 wagae 5 matet 6 yafe 3 manop 7 173. grandchild totoxo 1 totoxo 1 xab˜ı 2 anisi 3 magop 4 magop 4 maxop 4 lono 5 xafun 2 174. grandfather nimu 1 yeneti 2 si 3 anisi 4 ambai 5 mbae 5 ambai 5 aduwy 6 andüop 6 175. grandmother anE 1 anE 1 wi 2 itime 3 aye 4 nomben 5 asek 6 ama 7 max 8 176. grasshopper woto 1 pesu (L) bagisoro 2 boki 3 tere 4 tet kondok 5 wamuyop 6 role 7 daŋgup 8 177. grated payo (L) pe (L) peyo (L) peyo bigi (L) mbian 1 mbian 1 mon 2 biyo 1 mbiyon 1 coconut 178. ground soxo 1 soxo 1 soxo 1 soxo 1 itiwa 2 itop 2 ip 2 i(f) 2 fium 3 179. guard edind e-d 1 - xaibã ri- 2 kiaxaime- 3 womo- 4 womo- 4 wambemo- 5 wabeba 6 - 180. hair noxo 1 mo 2 rõ 3 xa xõ 4 ron 2 ron 2 lon/muk 2,5 xalo 2 mux 5 (on body) 181. hair moxo 1 mo 2 xabi-ro 3 xabumu 4 mögö-ron 5 ŋgi-ron 6 amba-lon 7 xabialo 3 mux 1 (on head) 182. hand, arm bedo 1 bido 1 bidE 1 bedo 1 it 2 wit 2 mbisit 3 il 2 mel 4 183. hard tetEra boge 1 bogeyEfe 2 xaxarE 3 buto 4 - sondot 5 mbatop 6 awage 7 xoxux 8 Shiaxa Yenimu Pisa Aghu Mandobo Yonggom Digul Wambon Kombai Korowai Wambon

184. head xEiba 1 xEiba 1 xaibã 1 xabã 1 kembian 1 ŋgin 2 ambat 3 xabia 1 xabean 1 185. hear da-d 1 da- 1 da- 1 da-k 1 ndot- 1 ndat- 1 ndat- 1 xaxe 2 dai 1 186. heart dibo 1 dibo 1 duburo 2 dübo 1 ndümarop 3 ndimndop 3 dimlop 3 dümo 3 debop 1 187. heavy bodi ke-d 1 bodu gi- 1 yandoxu 2 bodü 3 kagön 4 kagun 4 kuxun 4 inuxe 5 baul 6 188. heel bo 1 bu 1 kito mu 2 kito bu 3 wandunu 4 wan ŋgambui 5 kenop mum 6 kudoramu 7 - 189. hide rotomo-d 1 atoromo- 2 asurumo- 2 akosu-˜ 3 törökmo- 4 turukmo- 4 sulukmo- 4 yeruma 5 bai(l)mo 6 190. high syiri 1 arumE 2 siri 1 sisi 1 ŋguruop 3 ndawan 4 kukulop 5 muku 6 ilax 7 191. hit wiE- 1 wure- 2 ufuro- 3 üfü-ox 3 ü- 4 in- 4 in- 4 oruwo 5 ülmexo 6 192. house afoxo 1 afoxei 1 xa˜ı 2 büshü 3 mbutüp 3 mbitip 3 ap 1 a(f) 1 xaü 4 193. human noxo 1 noxo 1 yuxu 2 axu 2 küap 3 kagup 4 kap 5 xo 6 yanop 7 194. hurt xaxafE ke-d 1 xaxafi gi- 1 xarimo- 2 xaime- 2 ko e- 3 kat kokmo- 4 koxe- 3 ufla 5 - 195. husband xogo 1 xago 1 nuxu 2 amo 3 anogomberi 4 kop pari 5 nomok/kap 6,7 yale 8 yalen 8 196. inside wumu 1 womu 1 womu 1 womu 1 korup 2 kop 3 wamip 1 xalü 2 - 197. intestines yiro 1 iro 1 o 2 o mutu 3 omunop 4 oi 2 otmin 5 akuklayo 6 xul 7 198. island mu 1 mu 1 dubi 2 dübi 2 og ndun 3 oŋndum 3 ok ndum 3 iduwa 4 xandul 5 199. jump koxosomo- 1 doged oto- 2 bunoxo ru- 3 pinoxo ü-k 4 mbororükne- 5 soke- 6 seŋgamo- 7 boblagama 8 - 200. kangaroo koso ware 1 kiso weri 1 sibaxai 2 kuso baxi 3 - - ŋgetop 4 robai 5 - 201. kill wu-d 1 wu- 1 wu- 1 - u-˜ 1 iro uru- 2 in- 1 bogi 3 xedi 4 202. knee bokin 1 bu-begi 2 bagibu˜ 3 bümo 4 wambüon 5 kondok werop 6 ŋgot kelop 7 bouklo 8 gabün 9 203. knife sogodam 1 sugudam 1 waki 2 woki 2 wegi 2 wagi 2 waxi 2 waxi 2 fix 2 204. know fimi-d 1 fimi- 1 nafã gi- 2 dafaŋ gi- 3 noam ge- 4 ŋgoropmo- 5 katke- 6 xumima 7 daibo 9 205. lame taxas 1 - bagi burud˜ı 2 kuto wa 3 kondok wa 4 wan mbetat 5 katkut 6 - - 206. lance xati 1 xeiti 1 xasi 1 xasi 1 kiatö 1 arat 2 yaŋai 3 yogui 4 mal 5 207. laugh meaxa somo-d 1 me wu- 2 ab gi- 3 abe gi-k 3 kan- 4 aritow in- 5 amben lap- 6 giama 7 lesifu 8 208. leaf anamo 1 anamo 1 rõ 2 õ 2 ron 2 ron 2 enop-lon 2 enowalo 2 mux 3 209. leech tese 1 tisi 1 karia 2 sisi 2 teren 2 teren 2 selen 2 reyo 3 layo 3 210. lie down re-d 1 ri- 1 ri- 1 i-g 1 ran- 1 yan- 2 la-/le- 1 le/lei 1 ibo- 3 211. light ara 1 - kawa 2 keni 3 orat 1 rat 1 wesat 1 kwani 3 dalun 4 (bright) 212. light (not kone-d 1 kakani- 2 wofota 3 ta 4 taŋganden 5 rat 4 wawut 3 wafuke 5 dalun 6 heavy) 213. lighten bawu-d 1 bawu- 1 bã 1 wimi- 2 mbomba 3 warawae 4 watwoe lap- 5 boba 1 faxbeax 6 214. lip bondumu 1 bonaxadi 2 bunaxa-me 2 bomga 2 mbaŋga 2 mbaraŋgat 3 mbaŋgat 2 mogoro 4 - 215. live kadaxae ke-d 1 kedige gi- 1 kadaxai ba- 1 xagide ki-k 1 are mba- 2 ariok mba- 2 mba- 3 ba 3 bo 3 Shiaxa Yenimu Pisa Aghu Mandobo Yonggom Digul Wambon Kombai Korowai Wambon

216. liver toro 1 wu 2 wu˜ 2 u˜ 2 ndugon 3 ahum 4 ndimlop 5 afina 6 üm 7 217. long pere (L) piri (L) kowE 1 pi (L) ŋguruop 2 ŋgurup 2 kolamop 3 reyo 4 dal 5 218. louse go 1 gu 1 aŋgu 1 aŋgu 1 ŋgut 1 ŋgut 1 ŋgut 1 gu 1 meli 2 219. love ya ke-d 1 ya ki- 1 ye gi- 1 mumuŋ gi-k 2 ŋgamen ge- 3 wagae nde- 4 kunumo- 5 - - 220. low mokan 1 moka 1 bago 2 makã 1 oŋgok 3 kutuk 4 kutkutdoe 4 bobogo 5 - 221. lung xateta 1 xatipa 2 - nomügo 3 kanatöt 4 uk 5 huloplon 6 fufu 7 falip 8 222. machete patak (L) bagisi 1 farŋga 2 - kararan 3 kararam 3 kalanam 3 glona 4 pal 223. make war wuki ko-nd 1 wEki ku- 1 iku- 2 kotom ü-k 3 taŋgat ke- 4 up- 5 awimo- 6 - afü 7 224. man xo-butu 1 xo-baxati 2 xo bis˜ı 1 xobas˜ı 1 küap 3 kagup 4 kap 5 wafi 6 abül 7 225. marry ra re-d 1 ra re- 1 du n˜ı 2 ã a-x 1 timo- 3 sumo- 4 (lan/kap) lap- 1 lafa 5 defol fo 6 226. marsupial koso 1 kiso 1 kiso 1 kuso baxi 1 ndao 2 kunop 3 amitlonap 4 - füon 5 227. masculine xo 1 xo 1 xo bis˜ı 1 xobas˜ı 1 mbetin 2 kombatim 2 mbasin 2 malea 3 - 228. mat kapi 1 kapte 2 kiripã 3 bigio 4 yom 5 yom 5 - lika 6 - 229. meat kodo 1 kodo 1 kadu 1 kudu 1 kandö 1 yom 2 kandu 1 xudo 1 xal 1 (on body) 230. medicine mamuke 1 mumu 1 yi xa 2 - in ndumbut 4 iŋgamaŋgat 5 - - - 231. milk o-xoŋ 1 ome-xu 1 õ-xu 1 nõ-xu 2 am kun 1 om kun 1 om 1 a 1 am gü 1 232. molar tooth ta-keko 1 kike 2 maga 3 maga-guxa 4 imban-yariwa 5 tenep 6 esop taŋgue 7 - - 233. moon afE 1 arE 2 bid˜ı 3 bid˜ı 3 ogo 4 wogoi 4 waxot 4 maga 5 waxol 4 234. morning mimi 1 mimi 1 mim˜ı 1 mim˜ı 1 aget köp 2 ariw amin 3 awaektop 4 rekudo 5 - 235. mosquito syimpere 1 negi 2 ise 3 isinigi 4 taenop 5 sowen 6 etenop 5 gegemo 6 letün 7 236. mother wini 1 wini 1 ni 2 api 3 nou 2 noi 2 ni 2 nani 2 ni 2 237. mountain abiE 1 abugE 2 xabE 3 xaibie 3 mburündün 4 amgon 5 halim 6 fani 7 fanip 7 238. mouse tema 1 - semese 2 kuso naŋge 3 tugunap 4 sogonap 4 amun 5 fira 6 duo 7 239. mouth xate-to 1 xate-to 1 xatE 1 xato 1 maŋgot 2 maŋgot 2 maŋgotop 2 mogoro 2 bonggol 3 (inside) 240. mouth bonaxa 1 bonaxa 1 bunaxa 1 bomga 1 mbaŋga 1 mbonop 2 mbonop/esop xana 4 bonggol 1 (outside) maŋgot 2,3 241. much, naxaba 1 naxaba 1 wemã 2 weaxa 2 agöp 3 kutok 4 ndaxit 5 biduma 6 fiüm 7 many 242. mucus syi-nifu 1 si-nifo 1 sinifo 1 si-nifo 1 wogüra 2 warigae 3 waliok 4 - xim 5 243. mute roxo boxoda 1 - mimi 2 u de 3 arek ndap 4 ruk ndoiowop 5 ŋgilek 6 - dofu 7 244. nail doxo 1 dosE 2 kia 3 bedi moxo 4 mbetit 5 muk 6 bisit 5 idodo 7 - 245. name fi 1 fi 1 fi 1 fi 1 üp 1 ip 1 (k)hit 2 fi 1 fi 1 Shiaxa Yenimu Pisa Aghu Mandobo Yonggom Digul Wambon Kombai Korowai Wambon

246. navel mudu-gu 1 guro 2 ogu˜ 3 modü bogo 4 aŋgun 5 oi ŋgum 5 ŋgambin 6 goumuro 7 manütul 8 247. near kataxaya 1 kExeya 2 kinaxã 3 akiã 4 kirop 5 sindik 6 sinim 7 deima 8 xalu 9 248. neck mu-boge 1 tu-begi 2 gub˜ı 3 ekõ 4 ŋgemben 5 awut-mit 6 ŋgait/ŋgip 7 gubi 8 balebol 9 249. neck mu 1 mu-to 1 muto 1 adena 2 koman 3 ŋgomben 4 ŋgaip 5 xuma 3 xomofexolol 3 (throat) 250. nest yira 1 - yi a 1 a 2 ŋgerim 3 ap 4 elap 1 goglo 5 amül 6 251. net yaxama 1 yaxapa 1 in˜ı 2 kapüo 3 mbeŋgiat 4 ok kit 5 umap 6 eno 7 - 252. new noxoŋgo 1 - kadaxai 2 xagide 3 are 4 ariok 4 aluyok 4 alu 4 sol 5 253. night asyu 1 asu 1 asu 1 wom˜ı 2 wemin 2 mitik 3 kitip 4 fim 2 gülnanggaup 5 254. nippa palm ku 1 ku 1 - sedi 2 - - - - lambil 3 255. nipple o-tebo 1 omE-tobo 2 õ-subu 1 nõ subu 3 am toŋgut 4 om muk 5 omlolop 6 amolei 7 antenül 8 256. nose syi 1 si 1 sinto 1 si-payo 1 toŋgut 2 ambotop 3 kalit 4 ranggü 5 gelip 6 257. nostril syi-to 1 si-to 1 sinto-to 1 simto-to 1 toŋgut-top 2 ŋgiritop 3 kalitop 4 raguro 5 gelitop 6 258. nothing boxoda 1 bEigida 1 karE 2 fede 3 nda 4 ndoi 4 tembet 5 domo 6 mafem 7 259. now ne ke 1 nogo ki 1 togoro 2 sowo nego 3 koandep 4 kowandut 5 ndo nombo 6 menemanaru 7 imonE 8 260. old man yaxa 1 yaxo 1 yuxu xa 2 xu iwi 3 komberi 4 kop pari 4 kap yalen 5 yale 6 yalen 6 261. old (age) yaxa 1 yaxo 1 yoxoxoro 2 iwi 3 mberi 4 mbari 4 yalen 5 yale 5 nggel 6 262. old pusu (L) patu (L) semebe 1 posyü (L) wandop 2 woŋgopon 3 sinop 4 muno 5 - (duration) 263. old woman nayo 1 nayo 1 yoxoxoro 2 an iwi 3 ran ŋgeŋget 4 ra mbari 5 lan yalen 6 yale 6 lebaxop 7 264. older brother yaxa 1 yaxo 1 kapu 2 exo 1 anet 3 net 3 net 3 nai 3 afe 4 265. older sister ni 1 neni 1 de 2 eni 1 neni 1 nani 1 non 1 nani 1 aw 3 266. one esya 1 isyo 1 sikirE 2 fasike 3 ome 4 omae 4 sanop 5 ragla 6 lidop 7 267. outside bEtaxa 1 bitaxa 1 butaxa 1 betaxa 1 kat 2 kat 2 mbesan 3 a xuxuni 4 - 268. paddle, oar kafe 1 kefi 1 kafi 1 kefi 1 igio 2 suguyaŋ 3 ndayoŋ 4 iko 5 igo 5 269. palate xokoro 1 inigo-begi 2 xosopa 3 xaŋgadu 4 kanan 5 kanaŋgit 6 maŋgiŋ 7 - - 270. penis tege 1 tigi 1 sigi 1 segi 1 teŋget 1 teŋget 1 teget 1 reŋge 1 dul 2 271. phlegm ow uru 1 - Eso gu˜ 2 adimo 3 ŋgonöm 4 ayak 5 kayom 6 irimu 7 - 272. pig wi 1 wi 1 ui 1 wi 1 u 1 ui 1 oi 1 ai 1 gol 2 273. pinkie, se 1 risa 2 sio 3 sigiã 4 keteget 5 seget 6 kenop talon 7 kudololo 8 senan 9 small toe 274. plant ko-nd 1 ku- 1 koxõ- 2 soxo ku-˜ 3 ro- 4 ro- 4 halo- 5 lo 4 ao 6 275. play xaxafe re-d 1 xafe re- 2 xafe ra- 2 xabax’a-x 4 kukmo/a- 5 worow in- 6 hilik lap- 7 yama 8 gümo 8 276. rain a 1 a 1 aa 1 a 1 murüp 2 mirip 2 kaip 3 mulü/mlü 2 maun 4 277. rainbow piris (L) o 1 ir˜ı 2 wisi ai 3 regugut 4 erenayop 5 wandak 6 kwarilrei 7 anol 8 Shiaxa Yenimu Pisa Aghu Mandobo Yonggom Digul Wambon Kombai Korowai Wambon

278. rat otoru 1 - xayo 2 kuso yowõ 3 tomon 4 temon 4 kosot 3 fira 5 duo 6 279. rattan yoro 1 abuno 2 yoxomo 3 yoxu 3 tik 4 tik 4 sik 4 ri 4 yebun 2 280. red kaŋgo 1 gofo 2 kaŋgo 1 kaŋgõ 1 tenop 2 tenop 2 xeliop 3 re-rabo 4 xafümengga 5 281. rib ri-mboge 2 ri-begi 2 rimbigi 2 ini-bigi 2 eriman 3 rin 4 lin 4 lina 4 menaxolol 5 282. ripe waxama 1 bedi 2 waxamã 1 yomu˜ 3 omu 3 amui 3 hiliop 4 yamu 3 xayo 5 283. river oxo 1 ui 2 wadi 3 widi 3 ok 1 ok 1 ok 1 wodei 3 ax 1 284. root yi te 1 yi tagae 2 tatE 3 tete 3 i gumbüt 4 ndit 5 ŋgelek 6 kubu 7 dedil 8 285. rope tere 1 tiri kodo 2 kik˜ı 3 kik˜ı 3 tik 4 tik 4 sik 4 ri 4 nan 5 286. sago do 1 du 1 du 1 dü 1 ndu 1 ndun 1 ndu 1 doü 1 daü 1 287. saliva xate 1 xatipa 2 mase 3 xasi 1 katet 1 katet 1 walok 4 bedo, lua 5 mexesim 6 288. sand gEte poporo (L) poporo (L) gir˜ı 1 giŋge 2 ŋgirim 3 ok yaman 4 ok senap 5 abü 6 nenim 7 289. say roxo-d 1 roxo- 1 ro- 1 o-x 1 arek tagamo- 2 rogo- 1 loxo-/ne- 1 uma 3 di 4 290. scabies xa koko 1 - yam˜ı 2 yom˜ı 2 imbut 3 irombut 4 ayek 5 liubu 6 xambap 7 291. scar oru 1 uw 2 pEso ru 3 u˜ 2 kogu ut 4 ateram 5 atet 5 firo 6 - 292. scold oto-nd 1 ot u- 1 kEsema ro- 2 bobomo 3 eikmo/a- 4 tagapmo- 5 otixopmo- 6 luka 7 lebidi 8 293. see fete-d 1 fite- 1 feto- 1 ete-ox 2 itigio- 1 eto- 1 hetak- 1 fera 1 imo 3 294. seek edidimi- 1 - agu 2 agu-k 2 katomo’/a 3 kagaende- 4 mbulumo- 5 xloma 6 bi- 7 295. shallow asa 1 sa 1 asege 1 gono 2 mbarap 3 tut 4 ndindoe 5 fewakerambodo 6 - 296. sharp tareba 1 - kis˜ı 2 kis˜ı 2 ŋgeten 3 ŋget 3 keten 3 gerege 4 fül 5 297. shin bi-mboge 1 bi-bEgi 1 biniro 2 bino 2 mbiton 2 wan mbin 3 mbimit 4 wago fiya 5 - 298. shiver xa bobo-nd 1 xa bobo- 1 bob˜ı- 2 kudu itimi- 3 mbombanden mbambariri samo- 6 roralima 7 - ge- 4 ke- 5 299. shoot piemo (L) - pemo (L) t˜ı- 1 teen- 1 taem- 1 tamya- 1 rabiye-ne 2 ülmexo 3 300. short xoro 1 bagedi 2 bago 3 bago 3 oŋgok 4 kutuk 5 atuk 6 bogo 3 gembenul 7 301. shoulder yakere-boge 1 yekero begi 1 makiE 2 ge bigi 3 mak 4 witmak 5 malin 6 dodou 7 main 8 302. side keŋge-puru 1 woro 2 kik˜ı 3 moditi 4 aua 5 mereŋ 6 selem 7 - leam 8 303. sing yi ri-d 1 yeri ri- 1 itiro ri- 2 ada i-k 3 ŋgom orü- 4 ŋgom ri- 4 ŋgom ali- 4 gobo lü 5 Epo- 6 304. sit bo-d 1 be- 1 ba- 1 ba-x 1 mba- 1 mba- 1 ba- 1 ba 1 bo 1 305. skin xa 1 xa 1 xa 1 xa 1 kota 1 kotae 1 kotai 1 xa 1 xal 1 306. skull ogo-boge 1 afu-begi 2 xaibã-bagi 3 mika-bigi 4 kembian-mirap 5 ŋgin-mit 6 ambat mit 7 muko fiya 8 - 307. sky xotu 1 xuto 1 xou 2 xuito 1 kut 1 kumut 3 kut 1 ramo-xoü 1 dali 4 308. sleep kono re-d 1 kunu re- 1 kunu˜ ri- 1 kumun i-g 1 kinum ran- 1 kinum yan- 1 kinum (ran)- 1 xünu lei 1 ibo 2 309. slow soxomo 1 sukum 1 sukumã 1 teteme- 2 ndügüman 3 mbonmo- 4 menopmo- 5 agumo 6 aŋgox 7 310. small kasede 1 kasede 1 monoxo 2 patoxo 3 mberon 4 mberon 4 ambunop 5 muragaya 6 hiEn 7 Shiaxa Yenimu Pisa Aghu Mandobo Yonggom Digul Wambon Kombai Korowai Wambon

311. smell fugu-nd 1 fugu- 1 fi mi- 2 fumi-k 3 umo- 4 ipmo- 5 hip- 5 mo-foruma 6 sumo 7 312. smoke oru 1 afuwo 2 aku 3 yãn üku 4 in oruk 5 enow uruk 6 iruk 7 emarü 8 lemül 9 313. snake wuti 1 witi 1 wasi 1 wisi 1 ŋgweti 1 aŋgun 2 ŋgusin 1 guari, gwari 1 anol 3 314. sneeze syi ti-d 1 si ti- 1 as˜ı si- 1 asinaŋ gi-k 2 tatanan tü- 3 asiganae ti- 4 asiŋmo- 5 rena 6 lambimatimo 7 315. soft foxotena 1 xaxuwo 2 paxad˜ı 3 puxu 4 menap 5 awoi 6 aŋgonom 7 gouke 8 - 316. sole of foot kitu-xato 1 - kito atã 2 kito babu 3 kondog a 4 wan-tat 5 kenop atat 6 kudoako 7 - 317. son amutu 1 amu 2 mu 3 amoko 4 mandep 5 mandup 5 tum 6 miyo 7 abül 8 318. sour ta toŋgi 1 te tugi 1 xae 2 xoŋge 3 rugup 4 rowot 5 esop ndom 6 rougi 4 - 319. speak roxo roxo-d 1 ro roxo- 1 ru ro- 1 u o-x 1 arek tagamo- 2 rug in- 3 loxo- 4 uma 5 u(mo) 5 320. spider abusE 1 obose 1 abusã 1 süã 2 otugan 3 suwan 2 asowan 2 gamu, buluko 3 - 321. spine se-boge 1 si-begi 1 si-begi 1 mü-bigi 2 mbüman mirap 3 mimit ketop 4 mimitmit 4 buma fiya 5 müfexolol 6 322. spit xate somo-d 1 me wu- 2 mase si- 3 xasi si-k 4 katet tü- 5 katet tiomo- 5 walok si- 6 bedo 7 - 323. stairs furi 1 atiko 2 afir˜ı 1 fike 3 werik 1 ŋgwerop 4 halik 1 fali 1 yafin 5 324. stand e-d 1 e- 1 maseri- 2 e-k 1 ri- 3 re- 3 la-, lo- 3 le-, la- 3 alo 4 325. star mi 1 mi 1 m˜ı 1 m˜ı 1 minap 1 mindui 3 minop 1 mi 1 belil 4 326. steal kui re-d 1 ki re- 1 ki ra- 1 ki a-x 1 kurop timo- 2 kambir rap- 3 kambit lap- 3 xwui fa- 4 - 327. stink foxomo-d 1 foxomo- 1 paxa fu foxomo- 1 pafu˜ me- 2 ŋgenen ge- 3 kosip ke- 4 hipke- 5 furuma- 6 - 328. stomach mudu go wu-d 1 modu go wu- 1 kaku xarimo- 2 okiame- 3 a gogop tomo- 4 oi rewerep ke- 5 ot ŋgamek 6 idikafeke 7 - ache 329. stone sEŋge boge 1 egiro 2 iro 3 io 3 iwan 4 irop 3 ŋgai 5 riga 6 ilol 3 330. stone axe kosi 1 kose 1 xabu 2 xabu 2 kurü 1 kori 1 kowei 3 fafi 4 xul 1 331. straight mode 1 yo 2 imidi˜ı 3 kid˜ı 4 ndindip 5 yani 6 kuk 7 bumio 8 godax 9 332. strong bini 1 begi 2 xaiku 3 buto 4 eŋgoan 5 mbarewen 6 met 7 wale 8 beben 9 333. suck mi-nd 1 mi- 1 soxõ- 2 umu- 1 emi- 1 mi- 1 mesepmo- 3 fuamo-fo 4 depo 5 334. suck at ome min-nd 1 ome mi- 1 õ mi- 1 nõ mi-k 2 am emi- 1 om mi- 1 om mi- 1 ami bomi 1 am gü 3 breast 335. suckle om ede-d 1 om ede- 1 õ edo- 1 nõ ede-ox 2 am ndigio- 3 om yo- 4 om ndak- 5 adi 6 - 336. sugarcane ku mboge 1 ku 1 ge 1 ku˜ 1 küm 1 kin 1 wamboi 2 kumo 1 bahüom, baliam 3 337. sun sera 1 sera 1 sawE 2 sowo 2 teet 3 sat 3 sat 3 rei(l) 3 lup 4 338. swallow miko-nd 1 miko- 1 muku-˜ 1 miku-˜ 1 min kön- 1 mi-agumo- 2 mixopmo- 2 rumogeoma 3 - 339. sweat kote 1 kuti 1 kas˜ı 1 kes˜ı 1 koten 1 koten 1 setelep 2 kware 1 xelünax, xün 3 340. sweet yam dubisi 1 go 2 kabubura 3 kinabe 4 tomborop 5 mbondeŋ 6 mbondeŋ 6 leblu 7 dolum 8 341. swim oxo ri ki-nd 1 oxo ri ki- 1 k˜ı xo- 2 oxo müa-x 3 ögö uŋgu- 4 kimra ko- 5 kimo- 6 fuku 7 dadü 8 Shiaxa Yenimu Pisa Aghu Mandobo Yonggom Digul Wambon Kombai Korowai Wambon

342. take along red ago-d 1 - radi xo 1 ade xo-x 1 timo go- 2 rap ko- 3 lap- 3 lefa- 4 - 343. take, grab re-d 1 re- 1 ra- 1 a-x 1 timo- 1 rap- 1 lap- 1 lofa 3 ati(lo) 4 344. taro xorifi 1 madi 2 wi 3 fike 4 irandöp 5 wirop 6 luk 7 ludi 8 simbelu 9 345. tears efe ken-oxo 1 efe kino 1 kin-E 2 kin’oxo 3 kerow-ok 3 kerow-ok 3 kelop ok 3 okorofono 4 - 346. tell tega-nd 1 taga- 1 kiru si- 2 sia si-k 3 maton tokmo- 4 matom tokmo- 4 loxo- 5 uma 6 u(mo) 6 347. tell a lie yaxamo wege roxo- 2 yaxama ro- 3 fimb’o-x 4 arekpon ge- 5 koyapmo- 6 koyapmo- 6 balike 7 - semewu-d 1 348. tendon me 1 me 1 afusu 2 me 1 temet 3 met 1 met-met 1 melo 1 - 349. testicle xaxa-ro 1 xoxo-ro 1 wobu-ro 2 wobio 2 wambirap 2 norop 3 wambilin 2 wablo 2 loxesuxop 4 350. thick flat kuduwugu 1 kaduwugu 1 womukumã 2 wogide 3 mbotüt 4 igit 5 hinim 6 fini 6 fanip 7 objects 351. thigh midi 1 midi 1 mid˜ı 1 midi bigi 1 eŋgin 2 kitup 3 mindin 1 kinomidi 4 - 352. thin paŋge (L) pagi (L) seskadi 1 bigi da 2 eret 3 waguwop 4 helewet 5 maxo 6 afop 7 353. thin flat aseke 1 aska 1 asexe 1 asiga 1 mbarap 2 ahak 3 hinindoe 4 lolake 5 afop 6 objects 354. think fimi-d 1 fimi- 1 fimo- 1 fimi- 1 imbimo- 2 opkon ke- 3 hinopko halo- 4 fibima 5 xul duo 6 355. thorn komo 1 yomo 2 yamoxõ 3 yomo 2 orün 4 arin 4 alin 4 alü 4 aün 4 356. three okom-esya 1 okum-isyo 1 kumandimã 2 okuom-asike 3 ititmo 4 itipmo 5 taxem 6 woromi 7 piŋgup 8 357. throat roxodi-boge 1 mu-roxoma- musubo 3 u-müto 4 togorop 5 ogarurop 6 ok kalum 7 waniklo 8 - begi 2 358. throw rasomo-d 1 rasumu 1 kosomo- 2 kusumu- 2 ta- 3 somo- 3 samo- 3 fia mone 4 püxmo 5 away 359. thumb, or 1 u 1 wado 2 wodo 2 oŋgu 3 aŋgu 3 kenop ambalop 4 ilolo 5 wayo 6 large toe 360. thunder xo ru-d 1 xu ru- 1 xou ru- 1 xü ü-k 1 komöt 2 kumut 2 kumut 2 xumü 2 balalmo 3 361. tired finu 1 fino 1 taxã 2 taxã 2 ketewa 3 tet 4 soe 5 bu 6 bu 6 362. to braid ti-d 1 ti- 1 si- 1 si- 1 ti- 1 ti- 1 ipke- 2 bubloma 3 - 363. to roxo-d 1 roxo- 1 ro- 1 o-x 1 tagan- 2 rogo- 1 loxo- 1 luka 1 - command 364. to cry efe ro-nd 1 efe ro- 1 ifE ru- 1 ifiõ- 1 run- 2 rom- 2 lomo- 2 lo 2 Exmo 3 365. to have sex yo-nd 1 yo- 1 yõ- 1 yõ- 1 reŋgendemo- 2 mbarok rap- 3 yenem mo- 4 gawono 5 gomo 6 366. tobacco somgE 1 nomoxo 2 samkaxae 1 sumke 1 ogoyap 3 sapuk (L) sawuk (L) wania 4 depon 5 367. toe kitu toro 1 toro 2 kito torõ 1 kuto butõ 1 kondok kurugut 3 wan ketop 4 tilin 2 kudolo 5 - 368. tomorrow syefE 1 odofE 2 rigiro 3 bagidi 4 yanen do 5 yan ŋga 6 wamot 7 berekina 8 walelelexu 9 369. tongue fagE 1 fagE 1 fagE 1 fage 1 oŋgat 1 anop 2 hanop/hat-gat faŋga 1 lef 3 1,2 Shiaxa Yenimu Pisa Aghu Mandobo Yonggom Digul Wambon Kombai Korowai Wambon

370. tooth tare 1 maga 2 maga 2 maga 2 imban 3 inim 4 esop 5 imba 3 lebil 6 371. torch xEino 1 sia 2 sia 2 xasisia 3 eŋgot 4 yaŋgot 4 yaŋgot 4 kwaiyeria 5 wel 6 372. tree, wood yi 1 yi 1 kasero 2 kesaxe 3 in 4 enop 5 enop 5 dodo 6 du 7 373. turtle sato 1 sabu 2 waxabi 3 mügo 4 mbirip 5 ambum 6 mbilip 5 falu 7 abeap 8 374. twins koroxo 1 kuroxo 1 - - - mun koyop 2 mnyep 3 - - 375. twist rope ebi-nd 1 ebi- 1 r˜ı- 2 ˜ı- 2 mbio- 3 ip- 4 lawa lilimo- 5 gobo- 6 - 376. two okomo 1 okum 1 kurumã 2 okuomu 1 rumo 3 irumo 3 ilumo 3 ragla ragu 4 pol 5 377. ulcer ruŋ 1 - ru˜ 1 üne 1 orün 1 yun 2 yunop 3 lü 1 gun 4 378. umbilical gu 1 gu 1 ogu˜ 2 amoko gu˜ 3 aŋgun 4 oi ŋgum 4 ot ŋgambiŋ 5 gou 1 - cord 379. uncooked kadaxae ke-d 1 kedige 2 kadaxai 2 xagide 3 are 4 ariok 5 alun 6 wafedado 7 - 380. unripe kadaxae 1 kedige 1 kuwo 2 kaxi 3 ketot 4 mbatop 5 alun 6 fedado 7 - 381. upper arm bedo midi 1 bido midi 1 bidE bagi 2 bodo bigi 2 taet top 3 wit tun 4 yawet 5 imidi 6 labul 7 382. urinate yiti ti- 1 yi tu- 1 yi su- 1 isi fi-ox 2 erok tü- 2 yetok ti- 2 etoksi- 2 beru 3 dalalü 4 383. urine yiti ti- 1 yitu 1 yi su- 1 isi 1 erok 1 yetok 1 etok 1 erono 2 dulax 3 384. vagina ato 1 ato 1 ato 1 ato 1 atop 1 atop 1 atop 1 aro 1 lil 2 385. voice roxo 1 ro 1 ru 1 u 1 arek 2 ruk 1 luk 1 lu 1 aup 3 386. vomit ko eseme-d 1 kodi kefsomo- 2 kandi gi- 3 kumukan˜ı gi-k 4 agön- 5 aerap- 6 wayun lap- 7 baxugi/ba-xu- waxalimo 9 ge 8 387. wait ko de-d 1 ku de- 1 da ri- 2 de-k 3 ŋgirok ri- 4 irukmo re- 5 min- 6 fakele 7 imba 8 388. walk xo-d 1 xo- 1 xo- 1 xo-x 1 ko- 1 ko- 1 ko-, ka- 1 xa 1 xai 1 389. wall xEiri 1 rifa 2 rikirE 3 kiko 4 kök 5 ndayaŋ 6 kim 7 ba 8 damon 9 390. warm apato (L) apa (L) apa (L) apufo (L) ndam 1 mamin 2 mamin 2 mamü-ge 3 xox 4 391. wash oxo ko-d 1 oxo ku- 1 e ku-˜ 2 ünoxo ku-˜ 3 tombümo- 4 agum mbiamo- 5 mbutmo- 6 rünü 7 dadü 8 392. water oxo 1 oxo 1 e 2 oxo 1 ok 1 ok 1 ok 1 o(x) 1 ax 1 393. weak ma ke-d 1 ma ke-d 1 xai karE- 2 puxu 3 kömböt 4 awoi 5 katkut 6 faloni 7 - 394. wet oxo da 1 - kake 2 paxa (L) tögöp 3 ŋgaenak 4 oxoyak 5 guxe 6 fex 7 395. white xayo 1 ara 2 xaya 1 bo 3 koat 4 kuk 5 kowalop 6 xwalu/xuwaru 7 xoxolun 8 396. white ant tomogaxa 1 efige 2 karibubu 3 doxosi 4 toma 5 tomae 5 tomae 5 duari 6 kembaxi 7 397. widow xamose 1 xamsE 1 xamE 2 ã se˜ 3 ran kamok 4 ran sarip 5 lan salip 5 kamo 1 - 398. widower xosE 1 xosE 1 sEŋgo 2 xu se˜ 3 küap kamok 4 kagup sarip 5 kap salip 5 kamo 6 - 399. wife ra 1 ra 1 ra pasu 2 agã 3 aŋgen 3 sarip 4 salip 4 la 1 defol 5 400. wind kifi 1 kifi 1 kifi 1 kifi 1 kiou 1 kiwui 1 kiwin 1 xifei 1 fup 2 401. wing yaxa boge 1 yaxa begi 1 ba 2 tefio 3 mborõ 4 mburui 4 mbulun 4 yagiya 5 baul 4 Shiaxa Yenimu Pisa Aghu Mandobo Yonggom Digul Wambon Kombai Korowai Wambon

402. woman finigi 1 finigi 1 rã 2 ã 2 ran 2 ran 2 lan 2 la(n) 2 lal 2 403. work toto re-d 1 toto re- 1 firio emo- 2 fiko a-x 3 ra- 4 awoŋ rap- 5 kelepmo- 6 aya 7 kelaya ati- 8 404. worthless yefe 1 yefe 1 bEtero 2 wa 3 wagarep 4 mbetat 5 asat 6 - xalox 7 405. wound kEwi 1 kefi 1 pEso (L) peso (L) kogu 2 mbom 3 hitop 4 ruruko 5 yaxesel 6 406. wrong goba 1 gaba 1 bEtero 2 omumu 3 ŋgomban 4 ŋgamban 4 ŋgombam 4 gobage 1 lembul 5 407. yawn bo ru-d 1 tere atoxomo- 2 xufamo- 3 xufame- 3 mbanan tokmo- 4 andunow in- 5 maŋgolam lalo- 6 brabuma 7 - 408. yesterday syefE 1 odofE 2 asüwe 3 bagidi 4 orüwa 3 wamin 4 alip 3 kaklo 5 xaxul 6 409. young afo 1 afo 1 kadaxai 2 xagide 3 moŋgop 4 mamae 5 ketot 6 muno 7 dax 8 410. young syiboŋ 1 siba 1 subã 1 xosübã 1 moŋgop 2 mamae 3 kap kewet 4 lüwobadi 5 xofel 6 person 411. young raramo 1 raromo 1 amgi 2 amigi 2 meŋgeet 3 maŋgat 3 lan maŋgat 3 yamondei yo menel 5 woman mongo 4 412. younger kuda 1 kuda 1 kudE 1 küda 1 anet 2 net 2 nana(n) 3 wabü 4 mofexa 5 brother 413. younger neme 1 kuda 2 kudE 2 küda 2 embiat 3 mbiat 3 lan aŋgot 4 wabü 5 modol 6 sister 414. 1SG no 1 nu 1 nu 1 nu 1 nöp 1 nup 1 nuk 1 nuf 1 ne, nu(p) 1 415. 2SG go 1 gu 1 gu 1 gu 1 ŋgöp 1 ŋgup 1 ŋgup 1 gu 1 gu(p) 1 416. 3SG ewe, ege 1,2 ewi, egi 1,2 eki 2 efe 1 ege 2 yup 3 nexep 2 xe 2 ye, yu(p) 3 417. 1PL noxo 1 nugu 1 nugu 1 nügu 1 noŋgüp 1 naŋgup 1 noxop 1 nagu 1 noxu(p) 1 418. 2PL goxo 1 gugu 1 gugu 1 gügu 1 neŋgip 2 ŋgaŋgup 1 ŋgoxop 1 nage 2 naŋge 2 419. 3PL yoxo 1 yoxo 1 yoxo 1 yoxo 1 yeŋgip 1 yaŋgup 1 yaŋgup 1 yaxop 1 yexene- 1 420. 1SG.POSS na 1 na 1 na 1 na 1 nene 1 na 1 na 1 na- 1 n, nV 1 421. 2SG.POSS ga 1 ga 1 ga 1 ga 1 ŋgone 1 ŋgo 1 ŋga 1 gu 1 g-, gV- 1 422. 3SG.POSS wa 1 wa 1 ena, ewa 1,2 efe 2 ene 1 ya 3 nexo 4 xe 5 y-, yV- 3 423. 1PL.POSS naxa 1 niga 1 nuna 2 nügu 1 noŋgüne 1 naŋgo 1 noxo 1 nagu 1 noxu- 1 424. 2PL.POSS gaxa 1 giga 1 guna 2 gügu 1 neŋgine 3 ŋgaŋgo 1 ŋgoxo 1 nage 3 gexene- 1 425. 3PL.POSS yaxa 1 yaxa 1 yoxona 2 yoxo 1 yeŋgine 3 yaŋgo 1 yaxo 1 yaxop- 1 yexene- 1 426. near kataxaya 1 kExeya 2 kinaxã 3 akiã 4 kirp 5 sindik 6 - - xalu 7 427. far off xaxa 1 xaxa 1 xamã 2 mekese 3 koman 2 kamam 2 komamop 2 xiado 4 lexiŋga 3 428. yonder xagere 1 - xatego 2 xo 3 mbogo 4 kore 5 - - - 429. what kenaxa-de 1 kenaxafaxa 1 memã 2 makeaxe 3 kenemop 4 agaeop 5 keno 4 malufa 6 mbaxa 7 430. who ekaxa 1 koxo 2 monoxoi 3 meoxo 4 koap 5 agap 6 avop 7 yaforumu 8 yaxop 9

Bibliography

van der Auwera, Johan. 2009. The Jespersen Cycles. In Cyclical change, ed. E. van Gelderen, 155–180. Amsterdam/Philadelphia: John Benjamins.

Baas, Peter. 1981. Tsaukambo ﬁeld notes. Unpublished.

Boelaars, Johannes H.M.C. 1970. Mandobo’s tussen de Digoel en de Kao: bijdragen tot een etnograﬁe. Assen: Van Gorcum.

Boelaars, Johannes H.M.C. 1995. Samen met Papoea’s op weg, deel 2. Kampen: Kok Lyra.

Bril, Isabelle. 2007. Coordination, information hierarchy and subordination in some Austronesian languages. Frankfurter Afrikanistische Blätter 19:213–243.

Bril, Isabelle. 2010. Informational and referential hierarchy: Clause-linking stategies in Austronesian-Oceanic languages. In Clause Linking and Clause Hierarchy, ed. I. Bril, 269–312. Amsterdam: John Benjamins.

Chafe, Wallace. 1976. Givenness, contrastiveness, deﬁniteness, subjects, topics and point of view. In Subject and Topic, ed. C. Li, 25–56. New York: Academic Press.

Comrie, Bernard. 1976. Aspect. Cambridge Textbooks in Linguistics. Cambridge: Cambridge University Press.

Comrie, Bernard. 1985. Tense. Cambridge Textbooks in Linguistics. Cambridge: Cambridge University Press.

Croft, William. 1991. The Evolution of Negation. Journal of Linguistics 27:1–27.

Cysouw, Michael. 2003. The Paradigmatic Structure of Person Marking. Oxford Studies in Typology and Linguistic Theory. Oxford: Oxford University Press.

Dahl, Östen. 1979. Typology of sentence negation. Linguistics 17:79–106.

Dixon, Robert. 1982. Where have all the adjectives gone. In Where have all the adjectives gone? and other essays in semantics and syntax, ed. Robert Dixon, 1–68. New York: Mouton Publishers. 212 Bibliography

Dixon, Robert. 1994. Ergativity. Cambridge: Cambridge University Press.

Drabbe, Petrus. 1947. Spraakkunst van het Pisa-dialect der Awju-taal. In KITLV Inventaris 157. Collectie Petrus Drabbe M.S.C.

Drabbe, Petrus. 1950. Twee dialecten van de Awju-taal (met kaartje). Bijdragen tot de Taal-, Land- en Volkenkunde 106:92–147.

Drabbe, Petrus. 1954. Talen en dialecten van Zuid-West Nieuw-Guinea. Micro- Bibliotheca Anthopos. Posieux: Anthropos-Institut.

Drabbe, Petrus. 1957. Spraakkunst van het Aghu dialect van de Awyu-taal. ‘s Graven- hage: Martinus Nijhof.

Drabbe, Petrus. 1959. Kaeti en Wambon, twee Awyu dialecten. ‘s Gravenhage: Martinus Nijhof.

Drabbe, Petrus. 1962. Mijn 45 jaren taalstudie. Text of a radio address for the Ra- diorubriek Nederlands Nieuw Guinea, the Netherlands.

Dunn, Michael, Robert Foley, Steven Levinson, Ger Reesink, and Angela Terrill. 2005. Structural phylogenetics and the reconstruction of ancient language history. Science 309:2072–2075.

Durie, Mark, and Malcolm Ross. 1996. The comparative method reviewed. Regularity and irregularity in language change. Oxford: Oxford University Press. van Enk, Gerrit J., and Lourens de Vries. 1997. The Korowai of Irian Jaya. Oxford: Oxford University Press.

Evans, Nicholas. 2007. Insubordination and its uses. In Finiteness: Theoretical and empirical foundations, ed. I. Nikolaeva, 366–431. Oxford: Oxford University Press.

Foley, William A. 1986. The Papuan Languages of New Guinea. Cambridge: Cambridge University Press.

Foley, William A. 1998. Toward understanding Papuan languages. In Perspectives on the Bird’s Head of Irian Jaya, Indonesia: Proceedings of the Conference, Leiden, 13-17 October 1997, ed. J. Miedema, C. Odé, and R.A.C. Dam, 503–518.

Foley, William A. 2000. The Languages of New Guinea. Annual Review of Anthropo- logy 29:357–404.

Foley, William A. 2005. Linguistic prehistory in the Sepik-Ramu basin. In Papuan Pasts, cultural, linguistic and biological histories of Papuan-speaking peoples, ed. A. Pawley, R. Attenborough, J. Golson, and R. Hide, 109–144. Canberra: Paciﬁc Linguistics.

Foley, William A. 2010. Clause linkage and Nexus in Papuan languages. In Clause Linking and Clause Hierarchy, ed. I. Bril, 27–50. Amsterdam: John Benjamins. Bibliography 213

Foley, William A., and Robert D. Van Valin. 1984. Functional Syntax and Universal Grammar. Cambridge: Cambridge University Press.

Gonda, J., and J. Anceaux. 1970. Korte Mededelingen: Pater Petrus Drabbe. Bijdragen tot Taal-, Land- en Volkenkunde 126:459–462.

Gray, Russell, and Quentin Atkinson. 2003. Language-tree divergence times support the Anatolian theory of Indo-European origins. Nature 426:435–439.

Gray, Russell, David Bryant, and Simon Greenhill. 2010. On the shape and fabric of human history. Philosophical transactions of the Royal Society B 365:3923–3933.

Gray, Russell, A.J. Drummond, and Simon Greenhill. 2009. Language Phylogenies Reveal Expansion Pulses and Pauses in Paciﬁc Settlement. Science 323:479–483.

Gray, Russell, and Simon Greenhill. 2005. Testing population dispersal hypotheses: Paciﬁc settlement, phylogenetic trees and Austronesian languages. In The evolution of cultural diversity: phylogenetic approaches, ed. C. Holden R. Mace and S. Shennan, 31–52. London: UCL Press.

Greenberg, Joseph H. 1978. How does a language acquire gender markers? In Universals of Human Languages, ed. Joseph H. Greenberg, Charles A. Ferguson, and Edith Moravcsik, 47–82. Stanford: Stanford University Press.

Haiman, John. 1976. Presuppositions in Hua. In Chicago Linguistic Society 12, 258– 270.

Haiman, John. 1978. Conditionals are Topics. Language 54:564–589.

Haiman, John. 1983. On some origins of switch-reference marking. In Switch Re- ference and Universal Grammar, ed. J. Haiman and P. Munro, 105–128. Amsterdam: John Benjamins.

Haiman, John, and P. Munro. 1983. Switch Reference and Universal Grammar. Amster- dam: John Benjamins.

Halliday, Michael A.K. 1967. Notes on transitivity and theme in English, Part II. Journal of Linguistics 3:199–244.

Healey, Alan. 1970. Proto-Awyu-Dumut Phonology. In Paciﬁc Linguistic Studies in Honour of Arthur Capell, ed. S.A. Wurm and D.C. Laycock, 997–1061. Canberra: Paciﬁc Linguistics.

Heeschen, Volkert. 1997. Relativities: Use and Non-use of Spatial Reference among the Yale Speakers in Irian Jaya. In Referring to Space: studies in Austronesian and Papuan languages, ed. G. Senft, 175–196. Oxford: Clarendon Press.

Himmelmann, Nikolaus P. 1997. Deiktikon, Artikel, Nominalphrase: Zur Emergenz syntaktischer Struktur. Tübingen: Niemeyer. 214 Bibliography

Himmelmann, Nikolaus P. 2001. Articles. In Language Typology and Language Uni- versals: an International Handbook, ed. Martin Haspelmath, Ekkehard König, Wulf Oesterreicher, and Wolfgang Raible, 831–341. Berlin: Walter de Gruyter.

Holden, Clare, and Russell Gray. 2006. Rapid radiation, borrowing and dialect continua in the Bantu languages. In Phylogenetic methods and the prehistory of languages, ed. Peter Foster and Colin Renfrew, 19–31. Cambridge: MacDonald In- stitute Press, University of Cambridge.

Holton, Gary, Marian Klamer, Frantisek Kratochvil, Laura C. Robinson, and An- toinette Schapper. 2012. The Historical Relations of the Papuan Languages of Alor and Pantar. Oceanic Linguistics 51:86–122.

Holton, Gary, and Laura C. Robinson. 2012. Internal Classiﬁcation of the Alor- Pantar Language Family Using Computational Methods Applied to the Lexicon. Language Dynamics and Change 2:123–149.

Hughes, Jock. 2009. Upper Digul Survey. SIL International, Dallas.

Huson, Daniel H., and David Bryant. 2006. Application of phylogenetic networks in evolutionary studies. Molecular Biology and Evolution 23:254–267.

Jacobsen, William. 1967. Switch-reference in Hokan-Coahuiltecan. In Studies in Southwestern Ethnolinguistics, ed. Dell Hymes and William Bittle, 238–263. The Hague: Mouton.

Jang, HongTae. 2003. Survey Report on Languages of Southeastern Foothills in Papua Merauke Regency of Papua, Indonesia. SIL International, Dallas.

Jang, HongTae. 2008. Morphology and Syntax of Wambon, a grammar sketch. Indonesia: SIL unpublished.

Jespersen, Otto. 1917. Negation in English and Other Languages. Copenhagen: Høst & Søn.

Kass, Robert E., and Adrian E. Raftery. 1995. Bayes Factors. Journal of the American Statistical Association 90:773–795.

Koch, Harald. 1996. Reconstruction in morphology. In The comparative method reviewed: regularity and irregularity in language change, ed. Malcolm Ross and Mark Durie, 218–263. New York: Oxford University Press.

Kriens, R, and R. Lebold. 2010. Report on the Wildeman River Survey in Papua, Indone- sia. SIL International, Dallas.

Kriens, R, R Lebold, and Y Susanto. 2013. A Report on the Bamgi, Kia, and Lower Digul River Language Survey in Papua, Indonesia. SIL International, Dallas. Bibliography 215

Lambrecht, Knud. 1994. Information structure and sentence form. Cambridge: Cam- bridge University Press.

Lehmann, Christian. 1988. Towards a Typology of Clause Linkage. In Clause Com- bining in Grammar and Discourse, ed. J. Haiman and S. Thompson, 181–226. Ams- terdam: John Benjamins.

Longacre, Robert. 2007. Sentences as combinations of clauses. In Language Typology and Syntactic Description, ed. Timothy Shopen, 372–449. Cambridge: Cambridge University Press.

Mithun, Marianne. 1988. The grammaticization of coordination. In Clause Combining in Grammar and Discourse, ed. J. Haiman and S. Thompson, 331–360. Amsterdam: John Benjamins Publishing Company.

Mithun, Marianne. 1999. The languages of Native North America. Cambridge: Cam- bridge University Press.

Nichols, Johanna. 1996. The Comparative Method as Heuristic. In The comparative method reviewed. Regularity and irregularity in language change, ed. Mark Durie and Malcolm Ross, 39–70. Oxford: Oxford University Press.

Nichols, Johanna, and Tandy Warnow. 2008. Tutorial on Computational Linguistic Phylogeny. Language and Linguistic Compass 5:760–820.

Palmer, Frank Robert. 2001. Mood and Modality. Cambridge Textbooks in Linguistics. Cambridge: Cambridge University Press.

Pawley, Andrew. 2005a. Introduction to the chapters on historical linguistics. In Papuan Pasts, cultural, linguistic and biological histories of Papuan-speaking peoples, ed. A. Pawley, R. Attenborough, J. Golson, and R. Hide, 1–14. Canberra: Paciﬁc Linguistics.

Pawley, Andrew. 2005b. The chequered career of the Trans New Guinea hypothesis. In Papuan Pasts, cultural, linguistic and biological histories of Papuan-speaking peoples, ed. A. Pawley, R. Attenborough, J. Golson, and R. Hide, 67–107. Canberra: Paciﬁc Linguistics.

Pawley, Andrew, Robert Attenborough, Jack Golson, and Robin Hide, ed. 2005. Papuan Pasts, cultural, linguistic and biological histories of Papuan-speaking peoples. Canberra: Paciﬁc Linguistics.

Payne, Thomas. 1998. Describing Morphosyntax: a guide for ﬁeld linguists. Cambridge: Cambridge University Press.

Reesink, Ger. 1983. Switch Reference and Topicality Hierarchies. Studies in Language 7:215–246. 216 Bibliography

Reesink, Ger. 1987. Structures and their functions in Usan. Amsterdam: John Ben- jamins.

Reesink, Ger. 1994. Domain-creating constructions in Papuan languages. In Topics in descriptive Papuan Linguistics, ed. G. Reesink, 98–121. Leiden: Semaian 10. Leiden: Vakgroep Talen en Culturen van Zuidoost Azië en Oceanië.

Reesink, Ger. 2014. Topic management and clause combination in the Papuan language Usan. In Information structure and reference tracking in complex sentences, ed. Rik van Gijn, Dejan Matic, Jeremy Hammond, Saskia van Putten and Ana Vilacy Galucio, 231–262. Amsterdam: John Benjamins.

Reesink, Ger, Ruth Singer, and Michael Dunn. 2009. Explaining the Linguistic Diver- sity of Sahul Using Population Models. PLoS Biology 7(11):e1000241. doi:10.1371/ journal.pbio.1000241.

Roberts, John. 1997. Switch Reference in Papua New Guinea, a preliminary survey. Papers in Papuan Linguistics 3:101–241.

Ronquist, Fredrik, and John P. Huelsenbeck. 2003. MRBAYES 3: Bayesian phylogenetic inference under mixed models. Bioinformatics 19:1572–1574.

Ross, Malcolm. 2005. Pronouns as a preliminary diagnostic for grouping Papuan languages. In Papuan Pasts, cultural, linguistic and biological histories of Papuan- speaking peoples, ed. A. Pawley, R. Attenborough, J. Golson, and R. Hide, 15–66. Canberra: Paciﬁc Linguistics.

Ross, Malcolm. 2007. Calquing and Metatypy. Journal of Language Contact THEMA 1:116–143.

Ross, Malcolm. forthcoming. Deﬁning the Trans New Guinea Family: preliminary evidence from pronouns. In Papuan Languages and the Trans New Guinea Phylum, ed. Malcolm Ross Andrew Pawley and Meredith Osmond, forthcoming. Can- berra: Paciﬁc Linguistics.

Schachter, Paul, and Timothy Shopen. 2007. Parts-of-speech systems. In Language Typology and Syntactic Description Volume II, ed. T. Shopen, 1–60. Cambridge: Cam- bridge University Press.

Schapper, Antoinette, and Lila SanRoque. 2011. Demonstratives and non-embedded nominalisations in three Papuan languages of the Timor-Alor-Pantar family. Stud- ies in Language 35(2):380–408.

Shopen, Timothy. 2007. Language Typology and Syntactic Description: grammatical categories and the lexicon. Cambridge: Cambridge University Press.

Simmons, R.T., D.C. Gajdusek, and M.K. Nicholson. 1967. Blood group genetic vari- eties in inhabitants of West New Guinea. American Journal of Physical Anthropology 277–302. Bibliography 217

Stasch, Rupert. 2009. Society of Others. Berkeley: University of California Press.

Stassen, Leon. 2000. AND-languages and WITH-languages. Linguistic Typology 4:1– 54. Susanto, Y. 2004. Report on the Mapi River survey, South Coast of Irian Jaya, Indonesia. SIL International, Dallas. Versteeg, Hans. 1983. Zijn stam en taal. In Een open plek in het oerwoud: evangelie- verkondiging aan het volk van Irian Jaya, ed. Tjerk Sijbe de Vries, 21–25. Groningen: Vuurbaak. Voorhoeve, Clemens L. 1971. Miscellaneous notes on languages in West Irian, New Guinea. In Papers in New Guinea Linguistics No. 14, ed. T. Dutton, C.L. Voorhoeve, and S.A. Wurm, 47–114. Canberra: Pacific Linguistics. Voorhoeve, Clemens. L. 1975. Central and Western Trans-New Guinea Phylum Lan- guages. In New Guinea area languages, vol.1: Papuan languages and the New Guinea linguistic scene, ed. Stephen A. Wurm, 345–459. Canberra: Pacific Linguistics. Voorhoeve, Clemens L. 2001. Proto-Awyu-Dumut Phonology II. In The boy from Bundaberg: studies in Melanesian linguistics in honour of Tom Dutton, ed. A. Pawley, M. Ross, and D. Tryon, 361–381. Canberra: Pacific Linguistics. Voorhoeve, Clemens L. 2005. Asmat-Kamoro, Awyu-Dumut and Ok: an enquiry into their linguistic relationship. In Papuan Pasts, cultural, linguistic and biological histories of Papuan-speaking peoples, ed. A. Pawley, R. Attenborough, J. Golson, and R. Hide, 145–166. Canberra: Pacific Linguistics. de Vries, Lourens. 1986. The Wambon relator system. Working Papers in Functional Grammar 17:1–62. de Vries, Lourens. 1987. Kombai kinship terminology. Irian, Bulletin of Irian Jaya XV:105–118. de Vries, Lourens. 1989. Studies in Wambon and Kombai: aspects of two Papuan languages of Irian Jaya. Doctoral Dissertation, University of Amsterdam. de Vries, Lourens. 1993. Forms and Functions in Kombai, an Awyu language of Irian Jaya. Canberra: Pacific Linguistics (Series B-108). de Vries, Lourens. 1994. Numeral systems of the Awyu language family of Irian Jaya. Bijdragen tot Taal-, Land- en Volkenkunde 150:539–567. de Vries, Lourens. 2005. Towards a typology of tail-head linkage in Papuan languages. Studies in Language 29(2):363–394. de Vries, Lourens. 2006. Areal pragmatics of New Guinea: Thematization, distribution and recapitulative linkage in Papuan languages. Journal of Pragmatics 38(11):811–828. 218 Bibliography de Vries, Lourens. 2007. De grondtaal van Nieuw Guinea. Amsterdam: VU Uitgeverij. de Vries, Lourens. 2010. From clause conjoining to clause chaining in Dumut languages of New Guinea. Studies in Language 34(2):327–349. de Vries, Lourens. 2012a. Some notes on the Tsaukambo language of West Papua. Language and Linguistics in Melanesia Special Issue 2012 Part I:165–193. de Vries, Lourens. 2012b. Speaking of clans: language in Awyu-Ndumut commu- nities of Indonesian West Papua. International Journal of the Sociology of Language 214:5–26. de Vries, Lourens, Ruth Wester, and Wilco van den Heuvel. 2012. The Greater Awyu language family of West Papua. Language and Linguistics in Melanesia Special Issue 2012 Part I:269–312. de Vries, Lourens, and Robinia Wiersma. 1992. The Morphology of Wambon of the Irian Jaya Upper Digul Area, with an introduction to its phonology. Leiden: KITLV Press.

Wurm, Stephen A. (ed.). 1975. New Guinea area languages, vol.1: Papuan languages and the New Guinea linguistic scene. Canberra: Paciﬁc Linguistics. Samenvatting

Een linguïstische geschiedenis van Awyu-Dumut – morfologische studie en reconstructie van een Papoease taalfamilie Alles heeft een geschiedenis, een verhaal over waar het vandaan komt en hoe het geworden is wat het is. Elke taal, gesproken van generatie op generatie, heeft een geschiedenis. Dit boek is een weergave van de linguïstische geschiedenis van de Awyu-Dumut taalfamilie. Deze Papoease taalfamilie bestaat, zover nu bekend, uit acht talen, en wordt gesproken in West-Papoea, Indonesië. De Awyu-Dumut taalfamilie behoort tot de grotere Trans New Guinea taalfamilie. De focus van dit boek ligt op de herkomst van de morfologie van de Awyu- Dumut talen aangezien morfologie één van de meest stabiele factoren in een taal is, en daarmee zeer bruikbaar om de geschiedenis van niet-geschreven talen te tra- ceren. Door een strikte toepassing van de comparatieve methode worden proto- morfemen gereconstrueerd waardoor een beeld ontstaat van de proto-taal waar Awyu-Dumut talen van afstammen. Met het reconstrueren van proto-vormen is echter nog niet de hele linguïstische geschiedenis van een taal verteld. Om inzicht te krijgen in hoe Awyu-Dumut talen door de tijd heen zijn veranderd, traceer ik ook diachrone en grammaticale paden. Hoewel de focus ligt op de gedeelde geschiedenis van Awyu-Dumut talen is er in dit boek ook aandacht voor morfologische diversiteit en het eigen verhaal van elke taal. Zodoende is het ook een naslagwerk voor wie meer wil weten over de morfologie en syntaxis van individuele Awyu-Dumut talen. Hoofdstuk 1 geeft zowel achtergrondinformatie over de verschillende Awyu- Dumut talen als over de databronnen die geraadpleegd zijn. Voor Awyu-Dumut talen zijn rijke descriptieve data beschikbaar, grotendeels te danken aan het werk van Petrus Drabbe, een katholiek priester die werkzaam was als missielinguïst in Nieuw Guinea van 1935 tot 1960. Voor vijf van de acht Awyu-Dumut talen is een volledige grammatica met teksten beschikbaar, terwijl Drabbe voor nog drie andere Awyu-Dumut talen een grammaticaschets heeft gepubliceerd. Daarnaast is er een woordenlijst van 430 woorden beschikbaar in alle Awyu-Dumut talen. Hoofdstuk 1 gaat ook in op de methodologie die gebruikt is om Awyu-Dumut proto-morfologie te reconstrueren. 220 Samenvatting (Dutch Summary)

Hoofdstuk 2 richt zich op de proto-fonologie van Awyu-Dumut talen. Eerdere Awyu-Dumut proto-fonologieën gereconstrueerd door Alan Healey(1970) en Bert Voorhoeve(2001) vormen de basis voor de hernieuwde proto-fonologie die in hoofdstuk twee gepresenteerd wordt. De comparatieve methode wordt toegepast op lexicale data. Aan de hand van de klankovereenkomsten en klankverschillen kan geconcludeerd worden dat het Korowai geen Awyu-Dumut taal is, maar wel een aanverwante taal. Een toepassing van fylogenetische methodes bevestigt de interne verdeling van de taalfamilie in vier Awyu-talen en drie Dumut-talen. De Kombai- taal wordt door fylogenetische methodes ingedeeld bij Dumut-talen, maar door de comparatieve methode bij Awyu-talen. Dit conﬂict is in dit boek opgelost door Kombai te behandelen als een eigen subgroep. Het hebben van een solide gereconstrueerde proto-fonologie, evenals een interne subgroepering van de taalfamilie, maakt de reconstructie van Awyu-Dumut proto-morfologie mogelijk. Alhoewel Awyu-Dumut talen weinig nominale morfologie hebben, wordt er in hoofdstuk 3 toch een kleine hoeveelheid nominale proto-morfologie gereconstrueerd. In alle acht Awyu-Dumut talen volgt het adjectief het naamwoord, en drie strategieën om samengestelde naamwoorden te vormen worden door alle Awyu- Dumut talen gedeeld. Juxtapositie is de meest voorkomende manier om bezit uit te drukken, terwijl alle Awyu-Dumut talen ook gebruik kunnen maken van bezittelijke voornaamwoorden. Het meervoud van naamwoorden wordt in Awyu-Dumut talen gerealiseerd door het redupliceren van (de eerste syllabe van) het naamwoord. Daar- naast hebben verwantschapstermen hun eigen meervoudsmarkeerder, gereconstrueerd als *-gi voor Proto Awyu, *-ŋgu(i) voor Proto Dumut and *-ŋgV voor Proto Awyu- Dumut. Het hoofdstuk sluit af met een omschrijving van de coördinatie van naamwoorden in Awyu-Dumut talen, en de twee nominale coördinatoren *ku(p) en *te werden gereconstrueerd. In hoofdstuk 4 komen Awyu-Dumut pronomina aan de orde. Zowel persoonlijke als bezittelijke voornaamwoorden worden gereconstrueerd. De bezittelijke voornaamwoorden staan dichterbij de voornaamwoorden die voor het Proto Trans New Guinea zijn gereconstrueerd, en zijn daarmee ouder. Daarnaast is aangetoond dat het derde persoon enkelvoudig voornaamwoord in de meeste Awyu-Dumut talen is afgeleid van het demonstratieve element ep, wat ‘daar’ betekent. Als laatste wordt in dit hoofdstuk geïllustreerd dat persoonlijke voornaamwoorden weinig frequent zijn in Awyu-Dumut talen, en dat als ze voorkomen, ze hun nadrukleggende vorm aannemen. Hoofdstuk 5 behandeld de vier persoons- en getalsmarkeerders die Awyu-Dumut talen rijk zijn. De persoons- en getalsmarkeerders zijn onderdeel van het werkwoord. In Awyu-Dumut talen zijn de tweede en derde persoonsmarkeerders hetzelfde, zowel in het enkelvoud als in het meervoud, een samenvoeging die typologisch gezien niet vaak voorkomt. De vier persoons- en getalsmarkeerders die gereconstrueerd zijn voor Proto Awyu-Dumut zijn *-ep (1SG), *-en (NON1SG), *-epan (1PL) en *-enan. Dezelfde vormen worden gereconstrueerd als *-e(fe), *-en, *-efan en *-enan voor Proto Awyu, terwijl de corresponderende Proto Dumut vormen *-ep, *-n, *-ewan en *-enan zijn. 221

In de hoofdstukken 6 tot en met 8 komt het zogenaamde TAM (Tense Aspect Mood) systeem van de Awyu-Dumut talen aan bod. Hoofdstuk 6 gaat in op mood in Awyu-Dumut talen, waarbij de focus ligt op de vele verschillende manieren om realis mood te realiseren. In hoofdstuk 7 komen alle Awyu-Dumut tijdsvormen aan de orde, en wederom is er sprake van grote diversiteit. Twee van de vier Awyu-talen kennen vier verschillende verleden tijden, de zogenaamde historicum hodiernum, historicum hesternum, historicum distans en de historisch verleden tijd. De andere twee Awyu-talen hebben drie verleden tijden, terwijl Dumut-talen één verleden tijdsvorm kennen, en de Kombai-taal zelfs geen. Hoofdstuk 8 belicht de aspectuele betekenissen die door Awyu-Dumut werkwoorden uitgedrukt kunnen worden, zoals duratieve, habituele en voltooide acties. In Awyu-Dumut talen wordt net als in het Nederlands een positiewerkwoord gebruikt om aan te geven dat een actie over een langere tijd plaatsvindt, zoals in bijvoorbeeld ‘zij staat kofﬁe te drinken’ of ‘hij ligt een boek te lezen.’ Andere Awyu-Dumut aspecten zijn al even typologisch wijdverbreid of semantisch voor de hand liggend, en zijn daarmee geen indicatie van genealogische verwantschap. Hoofdstuk 9 behandelt Awyu-Dumut negatie. Een uitgebreid diachroon pad is zichtbaar in Awyu talen dat illustreert hoe dubbele negatie tot stand kan komen. Elk van de vier Awyu talen, evenals Kombai, representeren ieder een fase van de Jespersen cyclus; afhankelijk van de fase van de Jespersen cyclus waarbinnen een taal zich bevindt heeft het optionele of verplichte pre-verbale en/of post-verbale negatiemarkeerders. Dumut talen hebben heel andere negatiestrategieën dan Awyu- talen en Kombai, en daarom is het reconstrueren van negatie op het niveau van Proto Awyu-Dumut niet mogelijk. In hoofdstuk 10 wordt de morfologie die de plaats van objecten en acties aan- duidt besproken. Deze deiktische elementen, die vertaald kunnen worden met ‘hier’, ‘daar’ en ‘ginds’, evolueren in Awyu-Dumut talen tot aanwijswoorden die vertaald kunnen worden met ‘deze’, ‘die’ en ‘die gindse’. Op hun beurt evolueren deze aanwijswoorden door tot markeerders van topicaliteit. Hoofstuk 11 bespreekt hoe Awyu-Dumut talen zinsdelen aan elkaar verbinden, namelijk door nevenschikking of onderschikking. Nevenschikking en onderschikking zijn twee uiteinden van een continuüm, en meerdere typen nevenschikking en onderschikking komen voor in Awyu-Dumut talen. De meest karakteristieke manier om zinsdelen aan elkaar te verbinden is clause chaining, waarbij het eerste zinsdeel deels afhankelijk is van het tweede zinsdeel. Ook komt switch reference in dit hoofdstuk aan bod. Awyu-Dumut talen hebben een rudimentair switch reference systeem, waarbij een onderscheid in de ﬁnietheid van werkwoordsvormen grotendeels overeenkomt met een onderscheid in werkwoordsvormen die aangeven of het eerstvolgende werkwoord eenzelfde of een ander subject heeft. Het boek sluit af met een terugblik op alle gereconstrueerde vormen, conclude- rend dat de geschiedenis van een taalfamilie het beste ontrafeld kan worden door het combineren van verschillende methodes en invalshoeken, en dat ook onderde- len van de morfologie die niet reconstrueerbaar zijn een essentieel onderdeel zijn van het van verhaal van een taalfamilie.