A Historical Dictionary of Yukaghir 9783110892840, 9783110186895

The Historical Dictionary of Yukaghir has two main purposes. First, it is intended as a relatively complete source of

239 47 20MB

English Pages 513 [516] Year 2006

Report DMCA / Copyright


Table of contents :
1. Organization of the dictionary
1.1. The structure of an entry
1.2. Sources
1.3. Presentation of the data and translations
1.4. Transcription and transliteration
1.5. Grammatical information
1.6. Reconstructions, etymologies, and comments
2. Sources of the Yukaghir material
2.1. Old Yukaghir
2.2. End of the 19th century and 20th century
3. Basics of Kolyma Yukaghir phonology
3.1. Phonemic inventory
3.2. Positional restrictions on consonants
3.3. Vowel harmony
3.4. Alternations of vowels
3.5. Alternations of consonants
3.6. Epenthesis
4. A reconstruction of Proto-Yukaghir phonology
4.1. Vocalism
4.2. Consonantism
4.3. The structure of the stems
5. List of affixes
Index of meanings
Language index
Recommend Papers

A Historical Dictionary of Yukaghir
 9783110892840, 9783110186895

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

A Historical Dictionary of Yukaghir


Trends in Linguistics Documentation 25


Walter Bisang Hans Henrich Hock Werner Winter

Mouton de Gruyter Berlin · New York

A Historical Dictionary of Yukaghir


Irina Nikolaeva

Mouton de Gruyter Berlin · New York

M o u t o n de Gruyter (formerly M o u t o n , The Hague) is a Division of Walter de Gruyter & Co., Berlin. The publication of this volume was m a d e possible by subsidies f r o m the Finno-Ugric Society a n d the Foundation for Endangered Languages.

© Printed on acid-free paper which falls within the guidelines of the A N S I to ensure permanence a n d durability.


of Congress Cataloging-in-Publication


Nikolaeva, I. A. (Irina Alekseevna) A historical dictionary of Yukaghir / by Irina Nikolaeva. p. cm. - (Trends in linguistics. D o c u m e n t a t i o n ; 25) Includes bibliographical references a n d index. ISBN-13: 978-3-11-018689-5 (hardcover : alk. paper) ISBN-10: 3-11-018689-6 (hardcover : alk. paper) 1. Yukaghir l a n g u a g e - H i s t o r y - D i c t i o n a r i e s . I. Title. II. Series. PM20.N55 2006 494'.6—dc22 2006012647




by Die Deutsche


Die Deutsche Bibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic d a t a is available in the Internet at < h t t p : / / d n b . d d b . d e > .

ISBN-13: 978-3-11-018689-5 ISBN-10: 3-11-018689-6 ISSN: 0179-8251 © Copyright 2006 by Walter de Gruyter G m b H & Co. K G , D-10785 Berlin All rights reserved, including those of translation into foreign languages. N o part of this book may be reproduced or transmitted in any f o r m or by any means, electronic or mechanical, including photocopy, recording or any information storage a n d retrieval system, without permission in writing f r o m the publisher. Printed in Germany.


I would like to express my gratitude to all my Yukaghir consultants, and especially my principle guide to the language, the late Vasilij Salugin. I am deeply indebted to Tapani Salminen for his help in converting my database into a format suitable for publication. Without his assistance and encouragement this volume would not have appeared. I also thank Aleksandr Anikin, U w e Bläsing, Michael Fortescue, Eugene Helimski, and Tapani Salminen for discussing Yukaghir materials and related questions with me. M a n y special thanks go to Simon Carne for patiently editing the English text. T h e work on the book has been partly supported by the Spinoza Prize awarded to Frederick Kortlandt by the Netherlands Organization for Scientific Research, by the Finno-Ugric Society (Helsinki), and by Deutsche Forschungsgemeinschaft through Sonderforschungsbereich 471 at the University of Konstanz. 1 wish to thank Professor Kortland for his kind support of my work. 1 am very grateful to the Foundation of Endangered Languages for sponsoring this publication. The book is dedicated to the m e m o r y of my father, Aleksej Vasil'evic Nikolaev.


Yukaghir is spoken in the extreme North-East of Siberia. Several centuries ago there used to be several Yukaghir idioms, but by the end of the 19th centuiy only two varieties had survived. They are referred to as Tundra (or Northern) Yukaghir and Kolyma (or Southern) Yukaghir. These languages exhibit fairly significant differences, especially in vocabulary. Although traditionally they have been termed dialects, some scholars prefer to view them as separate languages and speak about "a Yukaghir language family" rather than "the Yukaghir language" (Kurilov 2001; Maslova 2003). According to estimates from the late 1980s, Tundra Yukaghir is spoken fluently by about 150 people and Kolyma Yukaghir by about 50 people, but these numbers seem to be rapidly decreasing. This book has two main purposes. First, it is intended as a relatively complete source of information on the Yukaghir lexicon. No comprehensive lexicographic description of Yukaghir is available for the international linguistic community. In addition to several relatively small glossaries (Veenker 1989; Endo 1997, 2001; Nyikolajeva 2000; Nikolaeva and Salugin 2003; Maslova 2001, 2003), two major dictionaries have been published so far: Angere (1957) and Kurilov (2001). The former is based on Jochelson (1900) and mostly presents the data from Kolyma Yukaghir as spoken at the end of the 19th century. The transcription and morphological segmentation used in this dictionary are seriously flawed. Kurilov (2001), together with its shorter version Kurilov (1990), is a very rich collection of the modern Tundra Yukaghir lexicon, but it uses the Cyrillic transcription with Russian translations and copies are rare outside Russia. Most importantly, each of these sources only describes one Yukaghir idiom. In contrast, this dictionary presents the different varieties of Yukaghir in comparative format. The modern Tundra Yukaghir materials are taken from published sources, while the modern Kolyma data were obtained through my own fieldwork conducted in the 1980s-1990s. Although some lacunae are inevitable, the book presents the first fairly comprehensive lexicographic description of Kolyma Yukaghir. In this respect it can be viewed as an attempt to document the lexicon of this highly endangered language. The dictionary also contains earlier materials starting from the late 17th century, most of which reflect extinct varieties of Yukaghir.



Second, and most importantly, the dictionary provides a reconstruction of Proto-Yukaghir, which has not previously been attempted (but see Nikolaeva 1988). Proto-Yukaghir is understood as the common ancestor of all known Yukaghir varieties. Proto-Yukaghir stems are established based on internal reconstruction and comparison between various Yukaghir idioms, as well as the external data. In some cases I cite possible cognates of Yukaghir stems from other languages, mainly Uralic and Altaic. Yet this dictionary is not etymological in the usual sense, i.e. it is not meant to provide etymologies for all Yukaghir words. Our present knowledge of the history of Yukaghir is insufficient for this purpose, but this dictionary may be a first step in this direction. The Yukaghir family is usually classified as isolated, although its possible distant genetic relationship with Uralic has been much discussed (Collinder 1940, 1957, 1965; Bouda 1940; Angere 1956; Tailleur 1959b; Nikolaeva 1988; Fortescue 1998). There is no consensus on this matter, mainly because the history of Yukaghir has remained little known. Since in this volume Yukaghir forms are not only cited in their modern shape, but are reconstructed for the first time, my hope is that it will provide a foundation for future etymological work. The dictionary will assist scholars to establish the affiliation of Yukaghir and to reconstruct the ethnic and linguistic situation in prehistoric northern Asia, as well as to investigate the possibility of distant genetic relations between language families in general.

Table of contents

Acknowledgements Preface Abbreviations Introduction 1. 1.1. 1.2. 1.3. 1.4. 1.5. 1.6. 2. 2.1. 2.1.1. 2.1.2. 2.1.3. 2.1.4. 2.1.5. 2.1.6. 2.2. 2.2.1. 2.2.2. 2.2.3. 2.2.4. 2.2.5. 3. 3.1. 3.1.1. 3.1.2. 3.1.3. 3.2. 3.3. 3.3.1. 3.3.2.

Organization of the dictionary The structure of an entry Sources Presentation of the data and translations Transcription and transliteration Grammatical information Reconstructions, etymologies, and comments Sources of the Yukaghir material Old Yukaghir Witsen Mueller/Lindenau Klitschka Boensing Billings/Merk Matjuskin (Cuvan and Omok) End of the 19th century and 20th century Kurilov Krejnovic Spiridonov Jochelson Schiefner Basics of Kolyma Yukaghir phonology Phonemic inventory Vocalism Consonantism Some non-phonemic variations Positional restrictions on consonants Vowel harmony Harmony in the first foot Harmony of short vowels after the first foot

ν vii xi 1 1 1 2 3 5 7 8 10 10 10 12 15 17 18 24 25 25 25 25 26 27 29 29 29 31 33 34 35 36 37


Table of contents

3.3.3. 3.3.4. 3.4. 3.4.1 3.4.2. 3.4.3. 3.4.4. 3.5. 3.5.1. 3.5.2. 3.5.3. 3.5.4. 3.6. 3.6.1. 3.6.2. 3.6.3. 4. 4.1. 4.1.1. 4.1.2. 4.1.3. 4.1.4. 4.2. 4.2.1. 4.2.2. 4.2.3. 4.2.4. 4.2.5. 4.3. 4.3.1. 4.3.2. 4.3.3. 5.

Harmony of long vowels Synharmonism Alternations of vowels Quantitative alternations Deletion Assimilation of mid vowels Reduction of the stem-final vowel Alternations of consonants "Sonorant ~ obstruent" alternation Assimilation of voiced consonants Voicing and palatal assimilation Assimilation of sibilants and j Epenthesis Consonantal epenthesis Vocalic epenthesis in word edge clusters Word-medial vocalic epenthesis A reconstruction of Proto-Yukaghir phonology Vocalism Inventory and vowel harmony Non-high vowels High vowels Long vowels Consonantism Inventory Sibilants and affricates Fricatives Sonorants and glides Consonantal clusters The structure of the stems Monosyllabic stems Bisyllabic stems Trisyllabic stems List of affixes

Notes References Dictionary Index of meanings Language index

39 40 41 41 43 45 46 47 47 48 50 51 52 53 53 54 57 57 57 57 61 64 65 65 66 68 69 71 74 74 75 77 79 84 87 95 465 499


Languages Chuk. Esk. Ev. Evk. Κ Kor. Mong. Nen. Ngan. Ost. Rus. Selk. Τ Yak.

Chukchi Eskimo Even Evenki Kolyma Yukaghir Koryak Modern Mongolian Nenets Nganasan Ostjak Russian Selkup Tundra Yukaghir Yakut


Finno-Permic Finno-Ugric Finno-Volgaic Proto-Mongolian Northern Samoyed Northern Tungus Ob-Ugric Old Yukaghir Altaic Proto-Eskimo Proto-Yukaghir Samoyed Tungus Uralic

Sources of Yukaghir materials Β BO KD KJ KK KL Μ MC ME MK MU MO RS SD SU TD

Billings (1787) Boensing (1781) Kolyma Yukaghir from Jochelson's manuscript dictionary Kolyma Yukaghir (Jochelson 1898, 1900) Kolyma Yukaghir (Krejnovic 1982) Klitschka (1781) May de 11 (Schiefner 1871b) Chuvan materials of Matjuskin (Wrangel 1841) Merk (1787) Kolyma materials of Mueller/Lindenau (1741) Ust'-Janskoe materials of Mueller/Lindenau (1741) Omok materials of Matjuskin (Wrangel 1841) Rajskij/Stubendorf (Schiefner 1871 a) Kolyma Yukaghir (Spiridonov 2003) Suvorov (Schiefner 1871a) Tundra Yukaghir from Jochelson' manuscript dictionary




Tundra Yukaghir (Jochelson 1900) Tundra Yukaghir (Kxejnovic 1958, 1982) Witsen (1692)

Sources of etymologies CED DEWOS EDAL ESRD ESRZ

Fortescue et al. (1994) Steinitz (1966-1993) Starostin et al. (2003) Anikin (2000) Anikin (2003)


Slepcov (1972) Collinder (1940) Collinder (1965) Fortescue (1998) Collinder (1957)


iterative habitual hortative lative locative multiplicative nominal derivation negation nominative non-iterative object focus ordinal participle perfective plural possessive postposition predicative prohibitive prolative pronominal qualitative resultative subject focus singular


ablative accusative adjectival derivation adverbial action nominal applicative attributive augmentative body part causative collective comitative conditional connective converb dative delimitative desiderative diminutive directive different-subject converb emphatic evidential frequentative future


genitive imperative imperfective inchoative indefinite instrumental intensive interrogative intransitive


subject nominal same-subject converb stative supine temporal transitive translative interjection verbal derivation

Other abbreviations dial. id. lit. P· pr.

dialectal idem literally page proper noun

rect. rev. smb smth

rectus reverse somebody something


1. Organization of the dictionary The dictionaiy part of the volume is preceded by this Introduction which explains the principles of the phonological transcription adopted in the book, describes the sources of the data, and discusses the grounds for the suggested reconstruction, as well as the relationship between different varieties of Yukaghir. The dictionary is followed by an English index intended to facilitate the search for Yukaghir equivalents of English words, and a language index.

1.1. The structure of an entry The data in the dictionary are organized alphabetically according to reconstructed stems preceded by an asterisk. In the alphabetical order the diacritics such as the palatalization sign ('), the length sign (:) and the hachek are disregarded. The letters ο and ο; η, Λ and g and γ; I and /; d, d' and δ, as well as e and a are treated as one letter each. For recent borrowings into Yukaghir, mainly from Russian, Yakut and Even, the source forms are cited without an asterisk. Russian verbs are cited in the infinitival form, although Yukaghir normally borrows the bare stem. Each reconstruction or source word is provided with a consecutive number from 1 to 2659, with which the entry is cited in the reverse index. Unfortunately, for technical reasons the following 36 numbers are missing from the numeration: 113, 196, 207, 318, 416, 428, 437, 588, 690, 799, 1114, 1143, 1144, 1222, 1244, 1296, 1559, 1578, 1706, 1736, 1873, 1922, 1923, 2001, 2080, 2119, 2175, 2216, 2259, 2266, 2460, 2461, 2462, 2463, 2494, 2503. Thus, the dictionary includes 2623 entries. In each entry the reconstruction or the source word is followed by the Yukaghir data. The source of the data is indicated by an abbreviation in capital letters before the word. For most entries the data are organized in several groups, each of which is written on a separate line. A group is formed by at least two words from different sources which exhibit more or less identical morphological structure in an uninflected form. For example, verbs have the same aspectual affixes, and nouns have the same derivational affixes or are parts of the same compound. When deciding whether words belong to the



same group I have ignored transcriptional differences between the sources, as well as regular phonological variations. The data within a group is cited in the following order: data from the 20th century Kolyma Yukaghir - data from the 20th century Tundra Yukaghir - data from the earlier sources. The words in a group normally have the same meaning and it is only cited once, after the first word. So when the following words in a group are not accompanied by a translation, it should be assumed that they have the same meaning as the first word in this group. Where the meaning of a word in a group differs, a translation immediately follows. If a word has an additional meaning together with the meaning it shares with other words, this is indicated with a plus sign (+). Words from different sources within a group are separated by a semi-colon. As an example I cite below part of the entry for the stem * köy-,

(1) *köqK körje:- to chop; K K köijie-, korjie-; KJ korjie-; K D korjie-; Τ köyie-


undo, to unrip; to cut; TJ karjie-; TD konie- to cut a skin or a fish Κ körju: split, crack; KK kurju: + scratch; KD koiju: KJ koijda- to take off |TK köqiere- to cut

This fragment includes two groups, one with the basic meaning 'to chop' and another 'split, crack'. Both go back to the reconstructed proto-stem *köij-. Some words have meanings that differ from the basic meaning, e.g. TD konie'to cut a skin or a fish'. KK kurju: means 'split, crack' and additionally 'scratch', as indicated by the plus sign. The entry contains also the words KJ korjda- 'to take off and TK körjiere- 'to cut' derived from the same stem but having no exact correspondence in any other available sources. Such words are not members of a group. They are presented on a separate line and divided by the sign |. Translations are given after each word. When it is unclear whether the word belongs to the relevant entry, it is preceded by a question mark. Some entries also contain brief etymological remarks and comments (see 1.6 for details). These are written on separate lines.

1.2. Sources The modern Kolyma data come from my field materials recorded among the Kolyma Yukaghirs in the settlements Nelemnoe and Zyrjanka, as well as in Jakutsk, in 1986-1991. Some of these materials remain unpublished, others



have been published as text collections (Nikolaeva 1989, 1997, 2000) and a school dictionary (Nikolaeva and Salugin 2003). I have also used other sources of modern Kolyma Yukaghir for consultation, such as Spiridonov (1997), Endo (1997, 2001), and Maslova (2001, 2003). However, I do not indicate them because they mostly correspond to my field data.1 It is not my goal to present all Kolyma Yukaghir derivations here, especially the numerous aspectual forms of verbs, which exhibit various degrees of productivity. Normally, verbal derivations are cited in three cases: (i) if the derived form has a direct correspondence in at least one other source and therefore is part of a group, (ii) if its meaning is unpredictable and substantially differs from the base form, and (iii) if it provides some phonological information relevant for the reconstruction. So if a verbal derivational form is not cited, the reader should not automatically assume it to be non-existent. Relevant generalizations and a list of attested verbal derivations can be found in Maslova (2003). On the other hand, I have tried to cite all or most attested nominal derivates and compounds, as well as some idiomatic expressions. Words of Russian origin in Kolyma Yukaghir are only cited if they are recorded several times from different informants and so can be considered more or less established loanwords. But I cite all Russian words that occur in other sources. The modern Tundra Yukaghir data are exclusively taken from the works of Kurilov (1990, 2001). Other data from the 20th centuty and the late 19th century come from the works of Jochelson, Spiridonov and Krejnovic. My intention was to cover these sources as fully as possible, but I have omitted many expressions whose status seems to be in between lexicalized compounds and free collocations, as well as some predictable derivational forms. Earlier materials are represented by the data from Schieffner (1859, 1871a, 1871b) from the second half of the 19th century, and by what I refer to as Old Yukaghir, i.e. data from all known Yukaghir sources starting from the late 17th centuiy until the first third of the 19th century (on Old Yukaghir see 2.1). They are taken either from published sources or unpublished archive collections. See Abbreviations for explanations of the labels, and Section 2 of the Introduction for descriptions of the sources.

1.3. Presentation of the data and translations Although Kolyma Yukaghir data are unified on phonological grounds, in some cases I also cite free and idiolectal variants; they are separated with a comma. Verbs are cited as uninflected stems followed by a hyphen, as are



some postpositional stems that can host an inflection. Nouns are given in the basic Nominative Singular form. The words from published and unpublished sources are normally cited as they appear in the source, while Cyrillic-based data are transliterated (see 1.4). In some cases I have chosen to provide corrected versions of records that are obviously erroneous in square brackets after the abbreviation rect. Variations that occur in the sources are separated by commas. The presentation of the data from the 20th century sources basically follows the same format as for modern Kolyma Yukaghir. Nouns are written without a hyphen, unless they only occur in an inflected form in the source in question. Verbal stems are given without inflectional affixes and are followed by a hyphen. Because of fusional processes on the morpheme boundaries, such truncated forms may differ from the actual stem. For example, KJ 'strange, odd' is represented by the participle pailice, where -ce is a participial affix. In fact the verbal stem is pajlu:-, while -u: changes into -i before certain affixes. But as the stem pajlu:- is not attested in KJ in this shape, I cite it as pajli-, merely omitting the participial suffix. In contrast, pre-20th century data are given as they occur in the source, i.e. in inflected forms. For example, the stem mon- 'to say' is represented in SU as mo'ny and monka. These are inflected forms of the 3rd person Singular and Plural, respectively, cf. Κ mon-i and moij-ψ. Translations and grammatical comments for such forms are not provided. Hyphenation, spacing and capitalization that occurs in the sources is not preserved. However, I provide a morphological analysis for compounds and separate their components with a hyphen. Hyphenation concerns both nominal and verbal compounds. By verbal compounds I mean lexicalized expressions which normally include a verb and a noun in a possessive, case or postpositional form. Proclitics are also separated with a hyphen, while suffixes are not. In some cases it is difficult to see whether we are dealing with a compound or a suffix, e.g. -ed'ilil in Τ arjded'ilil 'waist' (cf. Κ aijd-il 'waist'). Such elements are not separated with a hyphen. If a compound is attested in more than one source, it is cited as a group of words in the entry that corresponds to its first component. Consider for example a fragment of the entry *αηα. (2)Κ αηαη-pugelbe: beard, moustache [lit. mouth fur]; KJ αηα-pugolbie; αηαη-bugut'e; angen-bugulbie


K D aya(n)-pugelbie, arjan-pudelie; Τ aya-buguce; T D an'an-bugoce, an'an-pugoce; Β anghen-bugu'lbv,




As can be seen in (2), literal translations may be provided to facilitate the understanding of the morphological structure of a compound. Note that in most nominal compounds the first component stands in the Genitive in -n/-d, which does not belong to the stem. If the second component of a compound does not exist independently, it is cited as a separate entry. If a compound is represented in only one source, for modern Kolyma and Tundra Yukaghir it is given as a whole. Isolated compounds from other sources are not cited, but their components are treated separately under their respective entries. Consider the following example. (3) Τ morqile broken country with many hills and holes; TD -morxile The TD -morxile is only attested as the second component of an isolated compound, i.e. a compound which does not have a direct correspondence in other sources. This compound is not given, but TD -morxile is cited together with the independently existing Τ morqile and is preceded by a hyphen. Where the translations provided in the sources were in a language other than English, I have given English equivalents. However, I shortened and simplified some of the translations from Kurilov (2001), which gives several meanings for each entry. For fuller information on the semantics of Tundra Yukaghir words the reader should consult Kurilov's original works. Closely related meanings are separated by commas, while more distantly related meanings are separated by semi-colons. Obviously, the decisions I had to make here are somewhat individual. Proper nouns are translated with the abbreviation pr. and further explanations are given in brackets (e.g. a man, woman, river, etc). For most birds and fish Latin names are provided in addition to English translations.

1.4. Transcription and transliteration The modern Kolyma Yukaghir data are cited in phonological transcription, as explained in Section 3 of this Introduction. Latin-based sources are TD, KD, Β, ME, Μ, R, SU, ST, W, MU, and MC. These data are cited essentially as in the source, but I transliterated ϊ as y, denoted vowel length with a colon rather than a macron, and have ignored some diacritics (for example, the signs \ " and ° on vowels). Some small changes in the citation of individual sources are also possible. They are explained in the respective subsections of Section 2.



Sources based on Cyrillic are T, KJ, TJ, TK, KK, SD, BO, KL, MC, and MO. The data from these sources are given in transliteration. In (4) I present the basic rules of transliteration applicable to all sources cited in this volume. (4)

Cyrillic a 6 Β Γ

s Λ e e >K 3 Η Η κ J1 Μ Η H ο


Transliteration a b W, ν g γ d e,jejo, 'o ζ ζ i j k 1 m η 0 ö



π Ρ c τ y

Φ χ α Η UI

m t bl b 3 fO Ά e, " j a f '

Ρ r s t u f q, χ c c s CS 0 y f e ju, 'u ja, 'a e,je-

This system differs slightly from the standard transliteration used to render Cyrillic in that β is rendered as w rather than ν and χ is rendered as q rather than x. Such a transliteration brings the records of the Yukaghir words closer to phonemic transcription. Note that when transliterating β and χ in Russian words I use the letters ν (rather than w) and λ; (rather than q), respectively. The so-called jotacized letters (e, to and Ά) are transliterated in two ways depending on their position. Word-initially or after a vowel they are transliterated as the combination of j and the corresponding non-jotacized vowel (o, u or a). After a consonant they are transliterated as a non-jotacized vowel, while the preceding consonant takes a palatalization sign (')· Again the purpose of this system is to make the form closer to the Yukaghir equivalent. The jotacized vowel e, as well as the Old Russian letters e and "jat"', occur only in Old Yukaghir materials, since modern Cyrillic-based sources use 3



instead. These letters do not normally involve palatalization of the preceding consonant. However, in Matjuskin's materials (MO and MC) the initial e- may denote je- as well as e, cf. MC eme ~ Κ ece: 'father' but MC eooii ~ Τ jabo:j 'dead'. For these sources I write je- when it is confirmed from other sources and (j)e- when the word is otherwise unknown. The word-final "hard sign" τ> which was used after a final consonant in the old Russian orthography does not denote any sound and has been left out in the transliteration. The "soft sign" b denotes palatalization of the preceding consonant, but if it occurs after η ( = c) palatalization is not marked because c is already palatalized. When this sign is used after η, I, and d, the whole string is transliterated as ή, /', and d\ respectively. Boensing (BO) and Klitschka (KL) additionally use the sign whose meaning is not entirely clear. On vowels it is likely to denote stress, while on consonants palatalization. This sign is preserved in transliteration, but at the beginning of the word before a vowel it is rendered as j. Some Cyrillic letters, e.g. φ, ij and tif, occur in Yukaghir words only by mistake, since the corresponding sounds do not exist in Yukaghir.

1.5. Grammatical information The dictionary is not intended as an introduction to Yukaghir grammar and provides only a minimum of grammatical information, mostly indications of word classes. Yukaghir distinguishes the following grammatical classes: (i) nouns, (ii) verbs, (iii) adjectives, (iv) adverbs, (v) postpositions, (vi) numerals, (vii) pronouns, (viii) conjunctions, (ix) particles, and (x) interjections. Verbs and nouns are easily distinguished by morphological criteria. As mentioned above, nouns are uninflected in the Nominative Singular and are cited in this form. Verbs rarely appear uninflected, and verbal stems are followed by a hyphen. The category of verbs includes qualitative and quantificative stems translated with adjectives and numerals, respectively, e.g. Κ emba- 'black' and ataql- 'two'. Transitive and intransitive verbs have different inflectional paradigms. In some cases I indicate whether the Yukaghir verb is transitive or intransitive by the abbreviations TR and INTR after the translation. This is done, first, if the English translation is ambiguous, as e.g. for the verb 'to open', and second, when the transitivity of the Yukaghir verb does not correspond to its English equivalent. For example, Κ ege:ra- 'to tread on, to step on (TR)' is transitive, unlike its English counterpart.



Adjectives in Yukaghir form a small closed class, which includes at least the following words: Κ narica 'big', juku 'small', como 'big', joloqs 'last', ponqa ' w h i t e ' , p ö m L · 'round', cicks 'long, tall', co:I'd 'old, ancient', and il'l'a 'new, fresh, (an)other'. The grammatical class of these words is not indicated in the dictionary. Adverbs, numerals, conjunctions and pronouns are not indicated either, but their category should normally be clear from the translation provided. Postpositions in Kolyma Yukaghir are marked as PP. Interjections are marked as INTJ and in some cases an approximate translation or an explanation of use is provided in brackets. Particles either express a certain grammatical meaning or serve discourse purposes. Unlike adverbs or interjections they cannot form an independent utterance, not even an elliptical one. From a morphological viewpoint they rather form an heterogeneous class. Some particles are free standing words, others are bound words, while yet others are clitics. I do not provide details of their morphological behavior here. Some particles are translated with their English equivalents, while for others I provide a description using the general word "marker". For example, the approximate meaning of the Κ qata is something like 'let us', but in the dictionary it is translated as 'hortative marker'. Proper nouns and interjections, as well as most particles, are not included in the reverse index. Section 5 of this Introduction contains a list of Yukaghir inflectional and derivational morphemes which is meant to facilitate the morphological analysis of the cited materials. Grammatical labels used in the dictionary are largely based on Maslova (2003) with a few deviations discussed in Nikolaeva (2005).

1.6. Reconstructions, etymologies, and comments The principles of the Proto-Yukaghir reconstruction are explained in Section 4 of this Introduction, which should be consulted before using the dictionary. Alternative reconstructions are indicated with a slash or brackets. Homonymous reconstructions are provided with numbers, e.g. *aj- 1 and *aj- 2. As with all sources, if a reconstructed stem does not normally occur in an uninflected form, it is followed by a hyphen. This is characteristic of all verbal stems, as well as some non-verbal stems. No attempt is made to reconstruct the Proto-Yukaghir meaning, but in most cases it can be easily deduced from the meanings of the daughter words. If the word in only attested in one Old Yukaghir source, the reconstruction may only be a very rough approximation, since these sources



are not very reliable. Such reconstructions are marked with a question mark. Thus, the meaning of the question mark differs: in the presentation of the data it indicates that it is questionable whether the word is related to the relevant stem, while in a reconstruction it indicates that it is approximate. In most cases I have tried to reconstruct a stem alone and so omitted easily recognizable and fairly productive affixes even if they are present in all daughter words. For example, the aspectual and valence-changing verbal affixes are excluded from the reconstruction. On the other hand, reconstructions may include some non-productive affixes, such as for example the nominal suffixes -r/-t, -ms, -I'd, and -ks-f-qd. Etymological comments are very brief. I mostly cite potential cognates from Uralic and Tungus-Manchu languages, but in a few cases I also mention Turkic, Mongolian, Chukchi and Eskimo parallels. For the most part potentially related words are cited as they appear in the cited sources. In some instances small transcriptional changes are made, but without any substantial revisions. If the comparison with Yukaghir is highly problematic, it is preceded by a question mark. If a comparison has been suggested in the previous literature, references are given after the double slash sign //. When such references are missing, the comparison is suggested here for the first time. Translations are provided for the cognate words, unless their meanings are basically identical to the meaning of the Yukaghir words and are easily detected. In the comment line I indicate whether the phonological shape of the word is irregular and comment on individual phonological and semantic changes. Irregularity either suggests that the word looks like a recent borrowing but the source is unknown to me, or that the word is morphologically complex. The comment line may also contain information on whether the Yukaghir word was borrowed into another language.



2. Sources of the Yukaghir material This section of the Introduction describes the sources from which the Yukaghir data are cited, other than my own field notes on Kolyma Yukaghir, which will be addressed in Section 3.

2.1. Old Yukaghir In this section I outline the sources of what I refer to as Old Yukaghir. This term describes the varieties of Yukaghir spoken between the first known record of the language (in the second half of the 17th century) and the first third of the 19th century. The upper "border" is marked by the time when an important phonological change, the simplification of the intervocalic clusters, took place (see 4.2.5). The Old Yukaghir data come from published and unpublished sources. The unpublished sources are kept in the following archives: the archive of the Academy of Sciences, St Petersburg, abbreviated as AAN, the Manuscripts Department of the State Public Library, St Petersburg, abbreviated as OR GPB, and the Russian State Archive of Ancient Acts, Moscow, abbreviated as RGADA. Most of these data were collected by the expeditions of the Russian Academy of Sciences, such as the Second Kamchatkan Expedition (1733— 1743), the North-Eastern Expedition of Joseph Billings (1785-1793), and the expedition of Ferdinand von Wrangel (1821-1824). The work of these expeditions is addressed in detail in Gnuceva (1940), Vdovin (1954) and Sirina (1983), among others. Other materials were collected by Petr Pallas via correspondence with the local administration and later systematized by Friedrich Adelung, as part of the Russian Academy of Science's programme for the compilation of comparative dictionaries of world languages. Pallas published his materials twice, in slightly different versions (Pallas 1786, 1789), while Adelung's compilations remain in the archives. 1 All these sources represent slighly different idioms, but I leave open the question of whether they should be classified as dialects or separate languages. The philological analysis of Old Yukaghir sources presented here was first suggested in my dissertation (Nikolaeva 1988a).

2.1.1. Witsen The first known record of Yukaghir is actually a published text. It appeared in a book by the Dutch diplomat and scholar Nicolaas Witsen (1641-1717) first

Introduction 11 published in 1692. Witsen himself did not visit the Yukaghir land. He received the Yukaghir materials by post sometime after his visit to Moscow in 1664— 1665, probably through his cousin Andrea Vinius, who later headed the Siberian Chancellery in Moscow (Gebhard 1881: 44; Wladimiroff 1997). The text is a translation into Yukaghir of the Lord's prayer. Nothing is known about its origin. It is essentially a word-to-word translation and in some respects preserves the Russian syntax, for example, by postposing a possessive pronoun, which is not typical of Yukaghir. The text is written in Dutch orthography; for instance, the digraph oe denotes u. Witsen's text was published by Adelung (1806: 562), Schiefner (1871a: 393-394), and Collinder (1940: 14), however these publications omit some diacritic signs and do not provide explanations for some words. Therefore I cite it here as a whole after Witsen (1785: 687). The first line presents Witsen's text, the second line gives possible modern Kolyma Yukaghir correspondences, while the third line provides glosses. Otje ece: father

Mitsje, mit-l'a we-POS

Totlie, tst-l's thou-POS

Legatei lVqi-ta-j be-PL-FUT-3

Laetiot, l'a-t-i at be-FUT-3SUBJ Je i and

kandi qadi which

koendsjoenga, kuzu:-ga sky-LOC poegandallanpoh, pugu-d-alb-bs sun-GEN-boss-N

t'sjemol al como:l-ol will-N

levianh: lebe:-ga earth-LOC

kaltei, kel-to-j come-FUT-3

Lünliagel i:-legul smoke-food

keyck kej-k give-IMP

Mitin mit-in we-DAT

Taldelpon tadul-ps debt-PL

M'ltlaepoel, mit-Fa-pul we-POS-PL

Miltje mit-l'a we-POS

telaman, 7-ma here-TEMP Mi't mit we

Je i and

kondan qodo how


? ?

Nim, liu: name

Totlie tat-l'a thou-POS konda qodo how

koed Zjuga, kuzu: -ga sky-LOC

Monidetjelash, moni-ds-ts-ls stomach-TR-FUT-? ponkatsj poha:s leave. IMP

Mitin, mit-in we-DAT

poniatsjok tannevinol pona:s-o:k tahha-?-qin leave- INTER. 1 PL owe-?-DAT

12 Introduction Mitlaepoel, mit-l'o-pul we-POS-PL

Je i and

Olo Oimi'k, 9l-?-mik NEG-7-TR.2SG kimda annelan; kimdains-? deceive-? tonbanck, tönbo-s-k strong-V-IMP

kondo qodo how kondo qodo how

Olgonilask ol-qaiji:-bk NEG-chase-PROH moliak moli by.no.means

Le dot l'a-t at be-FUT SUBJ Je i and

tasndaslov, ? ?

Mitel mit-u-1 we-O-PRON.ACC

Mitel mit-u-1 we-O-PRON.ACC

poegoedal Lenpoh, pugu-d-fallaj-ba sun-GEN-boss-N

Je i and

koendejanck. kude-js-k become-?-IMP

As can be seen above, most words from this text have Kolyma Yukaghir equivalents, so we may be dealing with an Old-Yukaghir idiom that was very close, if not identical, to Kolyma Yukaghir. If this is so, the text is likely to have been recorded on the upper Kolyma. In this dictionary Witsen's materials are indicated as W. The original transcription is preserved, except that, as for the other sources, I write y instead of Ϊ.



Chronologically the next data on Old Yukaghir were recorded by Jakob Lindenau, a participant of the Second Kamchatkan expedition organized by Gerhard Friedrich Mueller. Lindenau's task was the ethnographic description of the Siberian peoples, including their languages. His Yukaghir materials consist of two wordlists recorded in 1741, when Lindenau traveled from Jakutsk to Oxotsk. Both are translations of the standard Latin list compiled by Mueller, which contains about 300 words from the basic vocabulary. The lists reflect two different Old Yukaghir idioms. The list I refer to as Mueller/Kolyma (MK) is kept in RGADA in the socalled Portfeli Millera (fund 199, list 2, part 513, Ν« 13, pp. 20-23 rev.). Page 20 contains the title: Jukagirisch (Kolymskije Yukagiri) - Korjakish Korjakisch II - Kurilisch, and consequently some Chukotkan-Kamchatkan words. There are no references to Yukaghirs in the comments provided on page 23, so the exact place of the recording is unknown. It is however clear that we



are dealing with Yukaghir from the upper or middle Kolyma. The list contains about 220 lexical entries, as well as the incomplete verbal paradigm. As far as I know, it has never been published before. · 3 This source is likely to reflect the language of the Kolym-cy tribe that lived in the middle basin of the Kolyma, as well as on the rivers Jasacnaja and Korkodon (Dolgix 1960). Kolym-cy is the Russian word, while the Yukaghirs used the word kögims to refer to this tribe, cf. modern Τ kö:jma 'Kolyma Yukaghir'. According to Sauer (1802), Yukaghirs of the upper Kolyma considered themselves the descendants of the people called konghini, which suggests the PY *köijkim3, cf. also the hydronym koygiina found in Jochelson's materials. According to Jochelson (1900: 209), Kolyma Yukaghirs of the end of the 19th century regarded the kögima as an extinct archaic tribe. In the 18th century ten kögims clans merged together on the upper Kolyma into three clans called in Russian Rybnikovskij (Fish clan), Nartennyj (Sledge clan), and Uskanskij (Hare clan) (Dolgix 1960: 417). The former later joined the lower Kolyma Yukaghirs and mixed with other tribes (Soromba and Omok-i), while the latter two became ancestors of the modern Kolyma Yukaghirs. The idiom reflected in MK was apparently spoken by a Northern group of kögimd, possibly by members of the Rybnikovskij clan. Lindenau's second wordlist contains about 280 words. It was recorded in Ust'-Janskoe, a settlement on the mouth of the Jana in the middle of the Northwestern Yukaghir territory and is abbreviated here as MU (Mueller/UstJanskoe). According to Dolgix (1960), in the 17th century this territory was occupied by several Yukaghir tribes: Jandin-cy (on the right bank of the Lena), Xoromo-i (in the upper basin of the Yana), Jangin-cy (around the middle of the Indigirka), Onojdi (on the right bank of the Jana), and Oljubenzi (in the lower basin of the Indigirka). Yeenker (1984) attributed the language of MU to the Xoromo-i tribe. However, by the time it was recorded (1741), the Northwestern Old Yukaghir tribes had been already greatly mixed. The Jandin-cy were a mixed Even-Yukaghir tribe to start with (Dolgix 1960: 382). According to Gurvic (1982), they later mixed with the Xoromo-x and Jangin-cy, while the Jangin-cy mixed with the Evens, in their turn. The tribe Onojdi was divided into two clans, Petajskij/Betil'skij and Del'janskij/Zerjanskij. The former was apparently Tungus by origin, and Gurvic (1966: 11), contrary to Dolgix, did not even consider it Yukaghir. So the idiom reflected in MU is likely to be a mixture of several idioms spoken by North-Western tribes, and also had an Even component. This list is known in several copies. The main copy, arguably written down by Lindenau himself, is preserved in RGADA (fund 199, list 2, part 513, N° 10, pp. 5-8 rev.) and contains the following title on page 5: Lexikon burjatskij, tungusskij, lamutskij [The Buriat, Even and Evenki vocabularies]. The first

14 Introduction sheet has the additional title: Yukaghiri Ust-Janskoe [The Yukaghirs from Ust'Janskoe]. It is this copy that is published in the present volume and denoted as MU. In addition, there are several copies made by scribes. (i) ORGPB, fund 7, JVb 111, pp. 15-20 rev. (ii) AAN, fond 94, list. 1. Jfe 170, pp. 1-14 (iii) OR GPB, Erm/nem., JVb 41/1, pp. 95-100 (iv) OR GPB, Erm., Ne 577 (v) OR GPB, fond 7, Jte 142 (vi) OR GPB, fond 7, N° 111, pp. 21-22 The relationship between these copies can be represented in the following way: RGADA fond 199, list 2, part 513, Jte 10, pp. 5-8 rev.

(ϋ) I (iii) (iv)



Copy (i) has the following title on page 9: Tungusike (Werchna Angara) Tungusice (Jakutz'k) - Yukagiri (Ust-Janskoe) [Evenki (Upper Angara) Evenki (Jakutsk) - Yukaghir (Ust'-Janskoe)]. Page 16 rev. contains two additional words absent from the other copies, which I have included in the dictionary: kaka 'penis' and papa 'vagina'. Copy (i) and consequently all copies derived from it, i.e. (iv), (v) and (vi), demonstrate numerous inconsistencies in rendering the diacritic signs, as well as the following copying mistakes (in order of appearance). Table 1. MU

copy (i)

numä tschuwöndschä jongjongla angga tschanngdschamä öllega

memä tschiwoondschä jongjongla angya tschangdscham olloga

modem Yukaghir Knumö Κ äubed'a Kjoyul KTaija -

Τ al'ya

house heart nose mouth wolverine fish



Table 1 cont. tschul Ijängdschandö lengde-bündsche lengdetsche imobujen ängdsche ongetschele ongenit namdätschit lutsche -jelaxlon -andaklon jömbon (irische vaitagä noxtscha endschdsche -tschomo jehoti

tschal gängdschando lerigderbundsched lengdesche imobyjen anydsche ongetschell ongenet namdütschit tiitsche -jelaklon -andklon jämboon ousche baitagä noktscha endschöldsche -tschonv jehoti

Κ Τ öu:l Κ jaijzsd-ö: Τ leqdabud'a

meat, flesh (goose) egg want to eat (INTR. 1SG)

Κ Τ leqdstca Κ immuijs

eat(FUT.lSG) get drunk (INTR. 1SG) cry stand (FUT.INTR.1PL) stand (FUT.INTR.3PL) low smoke four two dead (INTR.3SG) water woman (AUGM) sable living (IMPF.PART) big arrow


Κ Τ oyo:ti:li Κ Τ oyo^itqj -

Kjuxa Kjelekun Κ ataqun Tjabo:ri Κ o:zi: Κ pa:jt9ga Κ noqsa Κ end'a Κ como:Kjoyoti:

Copy (i) or one of its derivatives served as the source of Pallas's publications (1786, 1789),4 although Pallas transliterated it into Cyrillic. His materials reflect the deviations from the main copy cited above; for example, he writes MCMH for memä, instead of numä (numo) 'house'. Notably, Pallas omitted some words from the original list. A few words from copies (ii) or (iii) were published in Schiefner (1859, 1871a) and Veenker (1984).5 In this volume the MU list is published in full for the first time.

2.1.3. Klitschka The materials of the Irkutsk governor Franz (Fedor) Klitschka include the translation of the 22 so-called Bachmeister sentences (Bachmeistersche Sprachproben) and numerals. They were sent by Klichka to Pallas in a letter received by the addressee on 10 July 1781. The manuscript is kept in OR GPB (fund 7, N° 11, pp. 3 rev. - 5). Page 5 rev. contains the following note: Requ le 10 Juillet 1781 de Mr de Klitschka Gouvernier d'Irkutsk, while page 3rev. contains the title: Reci perevedennye srednekovymskogo jukagirskogo rodu ot knjazca Afanasija Ostrjakova i procix inorodcev logo ze roda [Sentences translated by the princeling of the Yukaghir Srednekolymsk tribe Afanasij



Ostrjakov and other members of the same tribe]. The so-called Schögren's fund (AAN fund 94, list 2, Ν» 60, p. 5) has a copy of the numerals from Klitschka's materials with the following title: Perevod na jukagirskoj jazyk, perevodcikov zdes ne slucilos', a nizepisannye slova najdeny ν preznix delax [A translation into Yukaghir. There are no translators here, but the words below were found in the earlier documents]. So the translation was made in the vicinity of Srednekolymsk where the Northern group of Kolyma Yukaghirs (Kolym-cy) lived. Indeed, from a lexical viewpoint, Klitschka's materials stand very close to Mueller/Kolyma (MK), see 2.1.2. Klitschka's materials were originally written down in Cyrillic, but published by Schiefner (1871a) in a Latin transliteration. This transliteration deviates from the original at several points, partly because Schiefner attempted to correct supposed errors in the Cyrillic text. The deviations are shown below together with the transliteration used in the present volume and, where possible, approximate equivalents from modern Yukaghir. Table 2. Schiefher


el jennili



my transliteration el-(j)ennylyi

marxilduek ale amtan

MapXHilJIHCK-b ajieaMflam.

marxildiek ale-amdan

marqil'dö:k sl-amdoj

cemut keilani

neMy KefijiaHiH MaHan'a

chemu kejlanii manal'a oil'a elejuc (j)ejulec p'andaj (j)ejunkul qoneii comon ju'oa ljukon ipimzailo modoni tynn adyi

cumu kejlarii majlo öjl'3 jo:bj jo:lsj pe:daj joyul qoni como:j jö: juko:j memzajate modoni tii) adi


oilä elejun jojulen pändai jonkul xoni comoc juoa lukoc pimzai modoi tin adi

orai'a 3JieiOH element n'aHflaft aoHKyjrb XOHij-bHl HOMOHb

ro'oa JIIOKOHT> HnuMJKai-Jio MOflOHH TblHHt a^bin


do not want (INTR.1PL) little girl immortal (INTR.3SG) everybody red (INTR.3SG) hair there is/are no ill (FNTR.3SG) ill (INTR.3SG) burn (INTR.3SG) nose go (INTR.3SG) big (INTR.3SG) see (TR. 1SG) small flame (ACC) sit (INTR.3PL) this firm, hard (INTR.3SG)



Klitschka's materials were published in Jochelson (1900: 229-230), where they were essentially translated into contemporary Yukaghir. In this dictionary they are reproduced in my transliteration after the manuscript and denoted as KL.

2.1.4. Boensing The materials of the assessor Boensing include the translation of Bachmeister's sentences and some numerals. The original manuscript is kept in OR GPB (fund 7, Ne 132, pp. 45 rev. - 46 rev). As follows from the note on page 46 rev., the data were sent to Pallas by Boensing in a letter of 24 March 1781 received by the addressee on 5 September 1781. In addition, the State Public Library in St Petersburg (OR GPB) has several secondary copies in fund 7: (i) N° 132, pp. 26-31, (ii) No 132, pp. 32-36, (iii) Ns 132, pp. 37-41, (iv) Ne 137, pp. 9-12, (v) •N° 137, pp. 13-16, and (vi) Nb 137, pp. 17-20. These contain many copying mistakes and are not considered here. Copies (i), (ii) and (v) as well as the main copy have the following title (the English translation is mine): Translation of Russian sentences into Koryak, Chukchi and Yukaghir with a separate indication of each language made in the fortress of Giziga according to the printed book sent by Mr. Timofej Stnalev. Copies (iii) and (iv) have a different title: Translation from Russian into Koryak Chukchi and Yukaghir made by the commander of the Okhotsk port Captain Leutenant Zubov via interpreters. This indicates that the translation was made in or near Giziga, i.e. not in a Yukaghir territory. 6 As already suggested by Tailleur (1962: 56-58), the idiom recorded by Boensing is particularly close or even nearly identical to the Cuvan idiom recoded by Matjuskin (on which see 2.1.6 below). This idiom was spoken by the Cuvan-cy tribe, which lived on the Anadyr' and underwent strong assimilation by the Chukchi in the 18th century. The Cuvan-cy were constantly attacked by the Chukchi, especially after 1747 when the Chukchi defeated the local Russian troops and moved closer to Anadyr' (Wrangel 1841: 82; Argentov 1886: 12; Antropova 1957: 179; Okladnikov 1975: 17; Gurvic 1982: 173-74). A part of the Cuvan-cy escaped to Nizhnekolymsk and in the 19th century was assimilated by the Tundra Yukaghirs. Some of them still preserved their ethnic identity at the end of the 19th century, but did not know their language (Nejman 1872: 40; Maydell 1894: 61; Jochelson 1894: 27). Another part of the Cuvan-cy was assimilated by the reindeer breeding Koriaks (Maydell 1925: 24). Finally, the third part settled on the Anjuj in Markovo and underwent russification (see 2.2.5). 7 Boensing's materials were first published in Schiefner (1871a), while Tailleur (1959a) later reproduced this edition. Apart from omitting many



diacritic signs, the following deviations from the original copy can be seen in Schiefher's publication. Table 3. Schiefner ili umboti


my transliteration ili-jumboty

modern Yukaghir Τ ol-jabataj

imonxabo tyngi titengec onomn-kondo jüxcon

Κ lis-yanbo:] Ttegi


imoxanbo tingii titungec onomikondo juxkon makagat te jaizee inza elookongat uotle-

HMOH'xabi) Tbi'urH TH'TljHre'Hb OHO'MHKOH/IO

ro'xHOHb Ma'Hararb


mänagat to jänzeje izhielo onongat





Τ -köde Tjuko:n Κ monoya -

Κ o:zi:b Κ unuqgst -

immortal (INTR.3SG) five these run (INTR.3SG) man little (INTR.3SG) how eyes smoke water (ACC) from the river (ABL) more

In this dictionary Boensing's materials are cited as in the original copy and abbreviated as BO.

2.1.5. Billings/Merk A list of 500 words was recorded during the North-Western expedition directed by Captain Joseph Billings. The place, time and circumstances of the recording are well known from the published diary of the expedition (Sauer 1802). The recording took place on 15 January 1787 in the settlement Nunkimtung-nagel (K nungad-an-αψΐ) situated in the mouth of the river Nelemnaja, a tributary of the Jasacnaja. These materials also reflect the language of the Kolym-cy tribe (see 2.1.2). All words found in Billings' materials are known in modern Kolyma Yukaghir or the earlier variety of Kolyma Yukaghir recorded by Jochelson. Several copies of the list have survived, but they show significant differences. 8 Three copies are based on the English orthography: (i) Billings' copy (OR GPB, fund 7, N° 137, pp. 23-30 rev.); page 30 rev. contains a note: Aus Billings Papieren, (ii) the publication of Sauer (1802), and (iii) the



publication of Schiefner (1871a). Further, there are t w o publications in Cyrillic, Saiycev (1811: 9 3 - 1 0 2 ) and Pallas (1789). T h e y both represent the transliteration of the English-based list, but lack several entries. T h e transliteration of the digraphs tsh, zsh, sh, dzsh as well as gh by Sarycev is inconsistent and there are other mistakes that m a y have appeared w h e n transliterating f r o m Latin to Cyrillic script. These are shown below. Table 4. Sarycev arniiy jiapKynb HHHeHHe OHMaHHCbl MHJinriajixjiofTb aubHi'ÖMa HeiOHMOHryjlb XyHHIHKH ejienff/KicHt yMaT iaicb 3TX5I biaay ΗΒΗ ΗΟΗΛΡΗ Moiioxaia K)KeHb

Sauer antoo larkul natshennee onmannee malgialachlon angnuma nejunmolgul chuniirkiellendzshien umat i-ak etchea yädoo ivi nondri moinchaija irken

Κ ad-ö: larqul naxarii önmsni malyi-jalo:j aqnums na-molyil kunir-kil'd'o:j jö:mst aja:k ece -

ajbi: nonzo:j -


son root 5Η3φ (INTR.3SG) clever (INTR.3SG) eight (INTR.3SG) earlier year nine (INTR.3SG) see (TR.2PL) glad (IMP) father husband shadow weak (INTR.3SG) wave one

B o t h S a r y c e v ' s copy and S a u e r ' s list contain c o m m o n copying mistakes absent f r o m other sources, cf.: Table 5. Sarycev jryKVH/iaii MaHaiuia' HMeaHjiaMeH nyHjiajiBOJie jneyuia appaHrw loaHrymejib

Sauer lukundae manalla'e itsheendamey poondalvolle lie'usha arra'ngia i-oanguitshel

Κ jukud-ö: majls ixad-amun poid-olbsl'a arannsj jo:n-kicil

boy hair elbow widower soul, spirit light (1NTR.3SG) forehead



Table 5 cont. Billings lukundä


Κ jukud-ö: majta i:cod-amun poid-olbsPa ju:cs aramisj jo:n-kicil


monolä itscheendami poondolvole lewsha arannei ionguitshel


itsheendami poondolvole lewsha arrangya i-onguitshel

boy hair elbow widower soul, spirit light (INTR.3SG) forehead

But the following words show that Sarycev did not copy directly from Sauer, since Sauer's list contains copying mistakes not followed by Sarycev. Table 6. Sarycev Mapxejrb jiareTaK nyaanMaH OHMaHHCbl nOHHIOJieHL· Maarmn. KVHÄLUy

Sauer marchet sagetak pudanniai onmanneig poinjuletsk määjil kunshu

Κ marqil' legitak pudenme:j onmani psn-joulac mayil kuzhu:

girl feed (IMP) tall (3SG) clever evening coat sky

This indicates that Sarycev and Sauer used a common source based on English orthography. Pallas's publication contains the same Cyrillic transliterations as Sarycev's list, but also shows the individual mistakes illustrated below. Table 7. Pallas cyyTa aanejib arannia HTMaK aHbiOHMa



Billings tshukta aktshel alangnae

arraK aHbHiOMa




Κ sukts ekcil' eleqhajaqtak aqnumo

city; Jakutsk boat lazy sing (IMP) earlier

Some of these mistakes could have emerged as a result of the misunderstanding of the Cyrillic form, which suggests that Sarychev and Pallas used a Cyrillicbased source.



On the other hand, Billings' and Schiefner's materials exhibit common features and contain considerably more words than those of Sarycev and Sauer, although they lack a number of words present in the latter. The following table demonstrates copying mistakes shared by Billings and Schiefner. Table 8. Sarycev Te KaHÖyHHOH




Sauer tia kanbunnoi







cha'nnin nunbur keivey

channen nunbar keivy

cha'nnen nunbar keivy


ondzshil ondzshi

onzshil onzshi

onzhil onzshi

ud'il' o:zi:

Κ ti: kenbunaj


here wide (INTR.3SG) when dew thin (INTR.3SG) nail water

Billings' list also contains some individual features not shared by Schieiner. Table 9. Billings markloo oniak nengashe pomzshole'ni angunma tshayoia

Schiefrier marchloo aniak nenganshe pomzsholene angnuma -

Κ marql'-ö: atinsk nigiza pömzstani aqnums coyoja

daughter speak (IMP) yesterday round (INTR.3SG) earlier knife

This suggests that Billings' and Schiefner's lists both go back to another English-based source, presumably different from the one used by Sarycev and Sauer (see Veenker 1984: 573 for a similar conclusion about Schiefner's and Pallas's publications). In other words, the inheritance schema for this group of sources can be reconstructed in the following way.


Introduction English-based proto-source

English-based source


English-based source / -


Cyrillic-based source




In this dictionary I cite the reconstructed forms of the English-based protosource and indicate them as B. The reconstructed forms are cited with an asterisk if they do not coincide with any of the actual sources. Their relationship to the individual sources is shown in Table 10. Table 10. reconstruction *aimaivi *artshetshiingzsha *arränei *etcheä *illeyennee *i-onguitshel *kailey *ke-ick *koikil *ledemnee *poldshitsha *po0ndolvolle *yakandaitsha *yälon *yärka *yelaklon *yelondzshendigia *yelongedukshimba *yomüel *yöatsh *yonrul

black sorrow light father wind forehead raven give stump low leaf widow horse three ice four sun's ray east neck pain mow

Billings aimaivi arlshetshunzsha arrangya etschea illeyennee i-onguitshel bailey ke-ich

Sauer aimäibi artshetshunzsha arrängiä etchea illejennie i-oanguitshel -


keick koikel ledemnie paldshitsha poondalvolle



yalon yarka yelaklon yelonzshendigia yelongidukshimba yomuel yoatsh yourul

iälon iärka ielahlon jelondshent digia jelongedukshimba jomüel joatsh

ledemnee -


Introduction 23 Table 10 cont. reconstruction *aimaivi *artshetshiingzsha *arranei *etcheä *illeyennee *i-onguitshel *kailey *ke-ick *koikil *ledemnee *poldshitsha *po0ndolvolle *yakandaitsha *yälon *yärka *yelaklon *yelondzshendigia *yelongedukshimba *yöatsh *yomüel *yonrul

black sorrow light father wind forehead raven give stump low leaf widow horse three ice four sun's ray east pain neck mow

SaryCev auMaHBH apnenyHHca appamifl 3TX5I Hjuie-bHbe ioaHryftHejiL·

Schiefiier aimaivi artscha tschungzsha arannei etcheä illevennee ionguitel



Kerne K0HKHJ1T> jieaeMHe nojiauima nynaanBOJie -

HJIOFTh iapna, apxa ajiarjioHB -

iejiOHri ioant FOMyejrh -


ledemnee peldshitsha poondolvolle yakandaitscha yalon yarka yelaklon yelonzshendigia jolongodukshimba joatsh jomuel jonrul

Finally, three wordlists that resulted from the work of the same expedition are based on German orthography: (i) OR GPB, fund 7, N° 137, pp. 2 rev. - 8; page 2 rev. contains the title: Bey de Billingschen Expedition ven Dr. Merk gesammelt? (ii) AAN, fund 94, list 2, Ne 59, pp. 1-8; page 8 contains the folllowing note: Captain Billings and some explanation about the Yukaghir calendar in English, as well as the title: Vocabulary of the Dialect of the Kovima Ukagers, and (iii) the publication of Rohbeck (Billings 1802: 91-129). Copies (i) and (ii) are almost fully identical, except that in (ii) all words begin with a capital letter, as was usually the case when copies were made by a scribe. Copy (iii) is a transliteration from English into German orthography. Its source was either Sarycev or a Cyrillic-based list that served as Sarycev's source. This can be seen from the fact that the word OHMÜHHCU 'wise' in Sarycev's list appeared as a result of the miscopying of the word onmannei, while Rohbeck re-transliterated it as onmannsy. The German-based list (i) is cited in the dictionary as ME. Since none of the mentioned sources contains all the words found in the other source, the

24 Introduction present volume is the first publication of Billings and Merk's materials in full. Whether Β and ME had a common source remains unclear. The words in these sources are written in the same order, but they are likely to result from parallel recordings of the same speaker by two different members of the expedition. This could account for the divergences in citing grammatical forms. For example, the entry 'to see' is represented as the 2nd person Plural in Β (:umat) and as the Imperative and the 2nd person Plural in ME (juk and jumat); the entry 'thou' is translated as the Nominative in Β {tat) and the Predicative in ME (tattak), and the entry 'to eat' is represented as the Action Nominal in Β (,landgal) and the Imperative in ME (langdak, lagk). Such divergences could have appeared if the informant repeated the word several times in different grammatical forms and each recorder wrote down a different form.10

2.1.6. Matjuskin (Cuvan and Omok) Further Old Yukaghir materials were recorded by sub-lieutenant Fedor Matjuskin, a participant of Wrangel's expedition (1821-1824). The recording was made in 1821 on the river Malyj Anjuj (Wrangel 1841: 81). Matjuskin's materials represent two Cyrillic-based word lists, the so-called Cuvan list (about 210 words) and the Omok list (134 words), plus five Cuvan sentences. Compared to other Old Yukaghir materials, the quality of the transcription is very poor and both lists, especially the Omok, contain numerous mistakes. A large number of words cannot be identified, probably due to fundamental mishearings or misspellings. The original manuscript seems to be absent from the materials of Wrangel's expedition kept in the State Navy Archive in St Petersburg. The lists we have at our disposal were first published in the diary of the expedition (Wrangel 1841: 115-125) and reproduced in Starcevskij (1889: 4 2 1 ^ 3 7 , 461-464. 639). The latter version contains many copying errors, but served as the basis of some phonological conclusions in Angere (1956), see Tailleur (1962: 57). The Cuvan list reflects the language of the Markovo Cuvan-cy (see 2.1.4). The identification of the Omok materials is difficult because the term omo is present in all varieties of Yukaghir. 11 Both in modern Kolyma and Tundra Yukaghir it generally means 'people, tribe', cf. also Κ om-ni: 'people'. It remains unclear to which of the Yukaghir tribes Matjuskin referred as Omok. What is certain is that the Omok materials exhibit phonetic parallels with Tundra Yukaghir not shared by other varieties of Yukaghir (see 4.2.3 and 4.2.4). However, there are no reasons to believe that the Omoks were direct ancestors of all the modern Tundra Yukaghirs; rather they may constitute



one of their components (Nikolaeva, forthc.b). In fact, Tundra Yukaghir exhibits more Even loanwords than the language recorded by Matjuskin and differs from it in several other respects. Tailleur (1959a, 1962) was the first to analyze Matjuskin's materials. We owe him the identification of many Cuvan and Omok words, although in many cases his solutions differ from mine. In the present volume I cite the Cuvan and Omok materials in transliteration after Wrangel (1841), and indicate them as MC and MO, respectively.

2.2. End of the 19th century and 20th century 2.2.1. Kurilov The modern Tundra Yukaghir data denoted here as Τ are cited after the works of Gavril Kurilov, mostly his Yukaghir-Russian dictionary (Kurilov 2001) and its shorter version (Kurilov 1990). Lexical entries included in the dictionaries are taken from Kurilov's large collection of texts, some of which have been published in Latin-based transliteration in Maslova (2001) and in Cyrillic in Kurilov (2005). These data are transliterated as described in 1.4. I have tried to cover these sources as fully as possible, but did not include all productive derivational forms cited in Kurilov's dictionaries. Sometimes my English translations differ from his Russian ones. Tundra Yukaghir has (or until recently had) two local idioms, Qaijaji: and Alaji:, which exhibit slight phonetic differences (Kurilov 1987: 5-8; Krejnovic 1958: 24-25). Kurilov mostly recorded the Alaji: variety originally typical of the so-called First Alazeja tribe on the lower Alazeja. More precisely, his materials reflect Tundra Yukaghir as spoken in the last third of the 20th century in the settlement Andrjuskino. Qaijaji: Yukaghir was spoken by the clans who nomadized between the lower Kolyma and Alazeja and is primarily known from the works of Jochelson (2.2.4).

2.2.2. Krejnovic Eruxim/Evgenij Krejnovic's materials are indicated in this dictionary as KK for Kolyma Yukaghir (Krejnovic 1982) and KT for Tundra Yukaghir (Krejnovic 1958, 1982). The Cyrillic transcription is transliterated on a regular basis. However I do not indicate some non-phonemic



pronunciational variants such as f, which, according to Krejnovic (1982: 12), represents the stop variant of the uvular fricative ^(i.e. γ). Krejnovic made his recordings approximately between 1940 and 1980. They mainly reflect the Alaji: variety of Tundra Yukaghir.

2.2.3. Spiridonov The materials of Nikolaj Spiridonov are denoted as SD and reproduced after Spiridonov (2003). This work is a publication of Spiridonov's manuscript Yukaghir-Russian and Even-Russian dictionaries edited by Aleksej Burykin. The original manuscripts have not been preserved. The dictionaries are known from the copies made by the ethnographer E. Orlova, which are currently kept in the Magadan Regional museum. Spiridonov himself was a native speaker of Kolyma Yukaghir. The Yukaghir part of the dictionary reflects the language of the Kolyma Yukaghirs of 1930-1931. According to the editor of the published version, the dictionary remained unfinished: there are considerably more words beginning with the letter a than those beginning with the last letters in the alphabet. The dictionary includes about 750 entries, and many of them are toponyms and hydronyms. The published version contains many misspellings, as the text was copied several times by people who did not have a good knowledge of the language, first of all Orlova herself. Spiridonov used a mixed orthography based both on the Cyrillic and Latin writing systems, as had been customary in some publications on Northern languages in the early 1930s before the unified writing system was adopted. The 2003 publication omits the diacritics that apparently indicated stress. In this volume the Cyrillic letters are transliterated.

2.2.4. Jochelson Woldemar Jochelson's materials reflect the varieties of Yukaghir spoken at the end of the 19th century and the very beginning of the 20th century. He recorded Kolyma Yukaghir and the Qayaji: variety of Tundra Yukaghir. Jochelson's materials are represented by two types of data. First, I have used his published works (Jochelson 1898, 1900). They are marked as KJ and TJ for Kolyma and Tundra Yukaghir, respectively. Jochelson's other published works (e.g. Jochelson 1905) are not used, as they hardly contain any words absent from the publications mentioned above. In rendering these Cyrillic data I have followed my usual transliteration practice with the following amendments:



Jochelson's ä is rendered as e, d, which optionally represents z, is rendered as z, μ is rendered as d\ and the sign indicating the optional palatalization on c is omitted. Stress is not shown. Second, I have used Jochelson's manuscript dictionaries, which remain largely unpublished. They are kept in the form of card files in the Archive of St Petersburg's Institute of Oriental Studies in the so-called Jochelson's Fund.12 The Kolyma Yukaghir dictionary is found in fund 23, part 1, N° 17-22 under the title Yukaghir-Russian dictionary. Part of this dictionary was published in Jochelson (1926: 317-327), but the manuscript is much more substantial. In the present volume these materials are denoted as KD. The Tundra Yukaghir dictionary is found in fund 23, part 1, N° 13-16 under the title Khangai-Russian dictionary. A few Qaijaji: words are cited in Jochelson (1926: 317-327). In the present volume these materials are marked as TD. I also cite Yukaghir words from the text of Jochelson (1926) and have included them in KD and TD. Although my publication of the manuscript dictionaries is more complete than Jochelson (1926), I have not reproduced them in full. A separate edition of these would be an interesting project, though not strictly required for the purpose of this dictionary. In particular, I have omitted many idiomatic expressions and sentence examples, as well as Jochelson's comments on the origin and usage of some words. In the manuscripts verbs are cited in the Imperative and in the 3rd person Singular forms, while I only cite verbal stems. In the Khangai dictionary Jochelson cites nouns in the emphatic form in -η, but I cite them uninfected. The dictionaries are written down in Latin-based transcription. While rendering it I made the following additional changes: stress is not indicated, / stands for / and /' stands for /, ή stands for η , h for g. (both h and g. represent non-phonemic variants of γ in the original), η for n, d' or /, ζ for d, and m for m . In the original, m" indicates a non-phonemic palatalization before the diphthongoid e: [ie],

2.2.5. Schieftier Anton Schiefner was the first to undertake a linguistic study of Yukaghir, in the second half of the 19th century. In his three articles (Schiefner 1859, 1871a, 1871b) he provided short grammatical notes and published word lists compiled from previous archival sources, as well as the new data he himself collected through correspondence with the local administration. On Schiefner's publications of earlier sources see 2.1.2, 2.1.3, 2.1.4 and 2.1.5. Schiefner's materials come from the following sources. First, in 1859 he received two short Yukaghir texts from the governor of Jakutsk, Dr. Julius



Stubendorff, and one text from the teacher in a Jakutsk primary school, Fedor Rajskij. These texts were recorded from an Omolon Yukaghir who visited Jakutsk in 1858 and represent the same idiom. Schiefher published them twice, in Schiefher (1859) and, with the improved transcription and translation, in Schiefher (1871a). In this dictionary I cite these materials after the latter publication and denote them as RS. Second, in 1861 Schiefher received a list of 2289 words, 50 sentences and one short text from Petr Suvorov, the Chukchi missionary in Srednekolymsk. Suvorov's materials are published in Schiefner (1871a). In this dictionary they are denoted as SU. RS and SU reflect the idioms spoken on the Omolon and in Srednekolymsk, respectively. They represent geographically and linguistically close varieties of Kolyma Yukaghir. Third, Schiefner (1871b) presents 111 sentences recorded by the General Governor of Eastern Siberia, Baron Gerhard von Maydell. These are indicated here as M. Maydell himself was involved in geographic and ethnographic work in North-Eastern Siberia in 1868-1870 (see Maydell 1894-1896, 1925). He recorded his Yukaghir materials in 1870 on the river Anadyr' in the settlement Markovo. The basin of Anadyr' and Markovo in particular was then inhabited by the Yukaghir Cuvan-cy tribe, see 2.1.4. However, Maydell noted that most Cuvan-cy could not speak their language when he visited them. Some had begun to speak Russian, while others had been assimilated by the Kolyma Yukaghirs (Maydell 1894-1896; Jochelson 1905; Gurvic 1966: 142). Maydell found it difficult to find speakers of the language in Markovo. Only one elderly lady was able to provide him with translations of Russian sentences into Yukaghir, and the influence of Russian syntax is evident in these translations. These materials do not show significant linguistic differences from Kolyma Yukaghir and hardly represent the language of the Cuvan-cy. Most of Schiefner's materials were re-published in Jochelson (1900: 227 229, 231-237), however Jochelson made many changes and corrections, essentially adapting them to the language spoken at the end of the 19th century. I cite Schiefner's materials after his publications, with the following changes in transcription: η instead of n, d'instead of j ( ' ) , and γ instead of g.



3. Basics of Kolyma Yukaghir phonology In the dictionary I have used my own phonological transcription for modern Kolyma Yukaghir. Here I explain this transcription and provide a phonological description of the language, which serves as the basis for the Proto-Yukaghir reconstruction in Section 4. This description should not be viewed as comprehensive: it only concentrates on certain relevant topics. Note that it deviates from previous descriptions (Krejnovic 1982; Maslova 2003) on a number of points. In some cases I have chosen to compare the Kolyma Yukaghir data with modern Tundra Yukaghir. In this Introduction I have rendered the Tundra Yukaghir in my phonological transcription, although in the text of the dictionary I have left all data other than the Kolyma Yukaghir as they appear in the existing sources. Where in this Introduction there is no indication of the source, the examples are taken from Kolyma Yukaghir.

3.1. Phonemic inventory 3.1.1.


Kolyma and Tundra Yukaghir have identical vocalic systems as below. (1)

front unrounded rounded high non-high

i i: e e:




back unrounded rounded

a a:

u u: ο ο:

Short non-high vowels are not specified for place of articulation when they occur after the first bimoraic foot, i.e. after the initial structures CVCV, CVC or CV:. As was first suggested in Nikolaeva (1986), in this position they are represented as a neutral vowel a, which may partly harmonize to a full vowel (see 3.3.2). The vowel d is not included in the phonemic inventory, because its distribution is fully predictable from the syllabic structure. Therefore it does not contrast with full vowels. However, I find it useful to indicate it in the transcription because, as shown below in 3.4.1 and 3.4.2, a has distinct phonological properties not shared by other vowels. The vowel a is also present in a few monosyllabic functional words that normally form a single phonological phrase with the adjacent lexical word,



such as the pronouns mat Ί ' , tdt 'thou', and possibly some forms of the auxiliary verb I'd- 'to be'. For Tundra Yukaghir Kurilov usually writes diphthongs instead of long mid vowels, but not always consistently, cf. Τ xoolew- 'to kill' (Kurilov 1990: 307) and xuole.w- id. (Kurilov 1990: 308). There are variations in Jochelson's and Krejnovic's materials, too, e.g. KJ xamluo— xamlo- 'how many, how much' and KK juo- ~ jo:- 'to see'. Krejnovic (1958: 9, 1982: 9 10) explicitly mentioned the diphthongs ie and uo, but did not discuss their phonological status. However, there are no minimal or quasi-minimal pairs that can distinguish diphthongs from the corresponding long mid vowels. Instead, the falling diphthongs are the non-phonological variants of long mid vowels, i.e. e: can be realized as ie, ο: as uo, and ö as uö (or wo).2 Diphthongs are generally preferred in the stressed position, especially in monosyllabic words, but this distribution is a tendency rather than a strict rule. There are only two words in Kolyma Yukaghir where the long e: never diphthongizes: me:me: 'bear' and emme: 'mummy'. The former is apparently an emphatic taboo word which appeared as a result of reduplication (see 4.3), while the latter is an affectionate nursery word. Τ ne:nuke: 'riddle' and eke: 'elder sister' do not seem to allow diphthongs either, but both are recent Even borrowings. In any case, these few words do not provide sufficient evidence for the phonological status of falling diphthongs. Maslova (2003) analyzes the glides j and w in the syllable-final position as consonants. In this Introduction a different solution is adopted: the glides j and w before a consonant or a pause are analyzed as vocalic components of rising diphthongs and are transcribed as j and u, respectively. There are no prevocalic rising diphthongs, but the glides are realized as the consonantal j or Κ b ~ Τ w, cf. moj-m 'holds (TR.3SG)' vs. moj-o:-j 'is held (RES-3SG)', kou-ds- (optionally [köwds-]) 'to drive out' ~ keb-ej- 'to leave', and Τ lau-m 'eats (TR.3SG)' (optionally [lawm]) ~ law-i-td- 'to feed'. The following arguments support this solution. First, it follows from the phonotactic properties of the language. According to Maslova (2003: 34), consonantal clusters are disallowed in the syllable-final position with the exception of y'-initial clusters, e.g. qojl ' G o d ' and qojl-gdt (ABL). The diphthong analysis strengthens the generalization: if j is analyzed here as vocalic, such words do not present exceptions. Similarly, in Kolyma Yukaghir the Imperative morpheme -k attaches after a vowel-final stem, but is absent after a consonant, since a word cannot end in two consonants (see 3.6), cf. mada:-k 'sit down (IMP)' but söribs 'write (IMP)'. But -k occurs after j, e.g. kej-k 'give (IMP)', apparently because the latter is vocalic. Second, the glide can be followed by the epenthetic /, which only occurs

Introduction 31 between two vowels (3.6.1), cf. tadi:-l-a:- 'to start giving' (INCH) and uj-l-a:- 'to start working' (INCH). Third, as discussed in 3.4.1, in monosyllabic nouns the vowel is always long. This is because such words must be bimoraic, but the final consonant does not provide a mora. However, monosyllabic nouns ending in u or j do not contain long vowels, e.g.jou 'illness' and köj 'man'. This indicates that the final segment here is not a consonant but a moraic vowel. Finally, the consonant j does not occur at the end of a consonantal cluster (see 3.2). As noticed in Maslova (2003: 34), the only exception is the cluster wj, e.g. sewjemet 'came in (INTR.2PL)' (in Maslova's transcription). With the analysis of w as vocalic, this case does not present an exception: j here functions as a syllable onset. I transcribe this word as söujamat. The rising diphthongs in the first foot can contain any vowel, e.g. köj 'man', pajps 'women', pujm 'blows (TR.3SG)', tolou 'wild reindeer', and leu 'eat (TR.1SG)'. After the first foot only two diphthongs involving non-high vowels seem to be allowed in non-compound words, ej and aj. Their distribution depends on the harmonic quality of the stem in the same way as the distribution of harmonizing long vowels a: and e\ (see 3.3): ej occurs in front stems and aj occurs in back stems. For example, these diphthongs are present in certain verbal affixes such as the Perfective -ej-/aj-, -dej-/-daj-, -rej-/-raj-, -tej-/-taj-, and -sej-/-saj-, e.g. locil-daj- 'to make fire' vs. sel'ga-dej- 'to break', coly-aj- 'to push' vs.jed-ej- 'to appear'. Note that although the only short non-high vowel allowed after the first foot is a, these affixes do not contain a. As is shown in 3.3.2, a optionally harmonizes in rounding to the vowels of the first foot, so if they contained a we would expect the form *locil-doj-, at least in some idiolects. However, such forms do not exist. This provides an additional argument for analyzing ej and aj as diphthongs, rather than the combination "short vowel + / ' .



The consonantal systems of Kolyma and Tundra Yukaghir differ slightly. The table below marks with the symbols Τ or Κ the phonemes that are particular to either language.





voiceless stops voiced stops voiceless fricatives voiced fricatives voiceless affricates voiced affricates nasals laterals trills

Ρ b


Τ w





t d Ts, K s Kz


velar uvular q

g γ c d' ή

r J

Krejnovic (1982) records the palatal stop t' instead of c, however in modern Kolyma Yukaghir it is pronounced as a palatal voiceless affricate and is so transcribed by most other sources. According to Maslova (2003), b is not a phoneme in Kolyma Yukaghir, which instead has the phonemic w. She takes b to be a variant of w, either a free variant in the intervocalic position or a positional variant after a sonorant. However, the alternations shown by b are similar to the alternations of other voiced stops. It is realized as b after a sonorant and as b or w (in some idiolects) intervocalically. In the coda position b undergoes assimilative processes typical of other voiced consonants (see 3.5.2). For this reason I have treated b as a voiced stop in this volume. So, unlike in Maslova's description, all voiceless stops have a voiced counterpart: ρ ~ b, t ~ d, k ~ g, c ~ d\ s ~ z, and q ~ γ (in the latter case there is also a difference in the mode of articulation). As for w, I have argued in the previous subsection that in Kolyma Yukaghir it does not have a phonemic status. Rather it is an optional pronunciational variant of u, which occurs before a consonant as the second component of a diphthong. Before a vowel, w may optionally replace b in some idiolects. In contrast, in Tundra Yukaghir w is phonemic: it occurs word-initially and contrasts with b in the intervocalic and syllable-initial position, e.g. Τ qabaija: 'bald' vs. qawarcp 'pit', ilwi:- 'to graze' vs. tilba:'to trample down'. The phonemes w and b in Tundra Yukaghir have a different origin, see 4.2. The distribution of k/q, on the one hand, and g/y, on the other, is not free but depends on the harmonic quality of the stem (see 3.3.4). But since the domain of synharmonism does not include inflectional suffixes, one can easily find minimal and quasiminimal pairs distinguished by velar vs. uvular

Introduction 33 consonants, e.g. monoyd 'mountain sheep' vs. nono-ga 'thorn (LOC)'. This suggests that k, q, g, and γ are distinct phonemes.

3.1.3. Some non-phonemic


Most allophonic variations are described in Krejnovic (1982) and Maslova (2003); here I will only discuss the variations relevant for my transcription. Voiced stops tend to undergo fricativization in the intervocalic position. For example, Krejnovic and Maslova normally write s as the allophonic variant of c. s occurs intervocalically or word-finally (3a) in contrast to the c available word-initially or after some consonants (3b). But I write c in both cases, on phonological grounds. (3)a. ece: kebec

[esie] [kebes]

father went (3SG)

b. erca cobul

[erca] [cobul]

bad sea

Another example is the inter-speaker variation b ~ w mentioned in the previous subsection, which is reflected in Maslova's transcription. Such variations are fully predictable and are not recorded in my dictionary. The previous sources have not mentioned that the vowels a and ο undergo optional fronting after the palatal consonants j, c, l\ and ή, especially if they are unstressed, e.g.janmdgej- ~ jenmagej- 'to run into' and najdo:— nejdo:- 'spare'. These variations are reflected in my data. The initial prevocalic j can be deleted, which leads to various changes, e.g.jomil ~ imil 'neck', ju:-legul ~ i:-legul 'flour', jurgud'e:ja ~ irgud'e:jd 'star', jöuluga ~ i:lugs 'sorrow', and joijzo:— itjzo:— urjzo:- 'to sleep'. In other instances the initial j, /'or ή assimilates to the second syllable, e.g ,jel'o:d'd ~ l'el'o:d'd 'sun', numusej- ~ mumusej- 'to press', l'a:jd ~ja:ja 'spleen', jouluc~ louluc- 'to ask', and ja.zi:- ~ na.zi:- (< *nancd-) 'to scratch'. Other non-predictable optional variations that may be reflected in my transcription are rql' ~ si' (e.g. maslo: < marqlo: 'daughter'), ql ~ qs (e.g. ataqhsta ~ ataqsasts 'second'), m ~ b (e.g. mi. bd ~ mi:ma 'basis, character, custom'), ö ~ u (e.g. lukil ~ lökil 'blunt arrow'), i ~ u (e.g. mimil ~ mumul 'young antlers'), and i ~ e (e.g. mi:d'i: ~ me:d'i: 'sledge').



3.2. Positional restrictions on consonants Since the syllabification constraints do not allow more than one consonant in the coda or onset position (see 3.6), Kolyma Yukaghir only exhibits two-consonantal clusters. A table of attested clusters is presented in Krejnovic (1982: 31), but it appears to miss some generalizations, namely, that consonants fall into several natural classes based on their combinatorial properties: voiceless obstruents (stops, affricates, and fricatives), voiced obstruents, the vibrant r, other sonorants, and j. It also fails to mention that the morpheme boundary plays a certain role in the well-formedness of clusters either. Table (1) represents possible combinations across the morpheme boundary. Table (2) shows the co-occurrence of consonants within a morpheme. The symbol # stands for the word boundary. Table 1.

voiceless obstruents voiced obstruents sonorants j #

voiceless obstruents

voiced obstruents










+ +







+ +




# +

+ +

Table 2. voiceless obstruents

voiced obstruents


other sonorants

voiceless obstruents





voiced obstruents r other sonorants











j -

_ + +


As follows from the tables, clusters ending in j are totally forbidden. The syllable-initial j assimilates to the previous consonant both across the morpheme boundary and inside a morpheme (3.5.4). Voiced obstruents only occur intervocalically, after a sonorant or a glide, and alternate with other consonants elsewhere (3.5.1 and 3.5.2). They are disallowed word-initially with the exception of a few words, but in Tundra Yukaghir word-initial voiceless consonants may undergo voicing (3.5.3).



Generally speaking, the combinatorial potential of a consonant within a morpheme is lower than across a morpheme boundary. Sonorants do not normally precede voiceless obstruents within a morpheme due to the historical process of voicening (see 4.2.5). However, there are no restrictions at the morpheme boundary, e.g. ludul-ta- 'to provide with iron (iron-TR)' and urun-ts- 'to provide with a bed (bed-TR)'. The consonant r typically co-occurs with tautomorphemic voiceless obstruents, but in some cases it may precede a voiced obstruent. There are additional restrictions on the individual consonants not reflected in the tables. The consonant η does not occur word-initially and ή is not allowed word-finally, although it is possible in Tundra Yukaghir. In Kolyma Yukaghir ή is depalatalized if it appears at the end of a word, e.g. kejban 'make thinner (IMP)' from kejbsn- 'to make thinner'. The clusters In and jl usually assimilate into ήή and j j , respectively. This creates geminates, which are otherwise rare. The clusters nl and ηΐ are generally disallowed, at least at the morpheme boundary. In such clusters η assimilates to / with further palatalization before the palatal /'. Examples are kil'-L· 'whose' < kin 'who' + I'd (Possessive) and mol'-M 'said (EV.INTR.3SG)' < mon- 'to say' + Μ (Evidential). But when a sonorant-final stem is followed by an /-initial affix, / usually falls out, e.g. nugens < nugen-b 'hand (ACC)' and qa.rdk < qa:r-bk 'skin (PRED)'. The geminated c is found in the word occa (evidently, from *otcd) 'vessel made of birch bark' and may optionally be created as a result of morphological derivation, e.g. terikad-jd 'married (INTR. 1SG)' > teriksd-cd

> terikst-cs



3.3. Vowel harmony Vowel harmony operates in two domains, the first foot and the uninflected word, and the rules differ. The first foot is bimoraic, and so has the following structures: (C)VCV, (C)V: or (C)VC (on the foot structure see 3.4.1 and 4.3). Obviously, the harmony is only observed in the (C)VCV type. Note that the distinction between two harmonic domains has not been mentioned in previous work. All previous studies (Jochelson 1905; Collinder 1940; Krejnovic 1982; Maslova 2003) describe one type of harmony and extrapolate it onto the word as a whole.



3.3.1. Harmony in the first foot Stems are either harmonically back or harmonically front. The distribution of vowels into harmonic classes according to frontness is shown in (4). (4)

front back

e a

ö ο

u (u)

i i

If the first foot has two vowels, i.e. has the structure (C)VCV, these vowels must share the value for the frontness feature. The front and back vowels do not co-occur within the first foot. The value of the frontness feature of the stem determines the harmonic value of some long vowels in further syllables (3.3.3), as well as the synharmonism (3.3.4). As follows from (4), e and ö are front, and a and ο are back. The vowel u is harmonically front but phonetically back (for a historical explanation of this fact see 4.1.3). This means that it is compatible with e and ö (e.g. Κ pure, Τ pure 'on, outside', Κ cugö 'road, trace'), but typically not with a and o. There are only a few words where u and a co-occur. Some are recent borrowings, e.g. Τ pura.qil' 'jaeger (Stercorarius)' ~ Ev. prakil 'kind of bird' and Τ qularqd 'seagull' < Ev. kular 'gull'. Similarly, Κ sule- 'to stand still' is a back stem. The vowel i occurs both with front and back vowels, as demonstrated in (5a) and (5b), respectively. (5) a. kiseb. ibor citata: Τ cibaya-


Κ Τ ig er


wart mido-, Τ miraground squirrel Κ Τ imol to peep (of a mouse)

to roam saddle reindeer

Although the latter cases are less frequent, they are by no means uncommon. For a historical explanation of this see 4.1.3. The vowels of the first foot also share the roundness features, but the roundness harmony only operates on a subset of vowels, namely, on non-front non-high vowels. In other words, the back vowels ο and a cannot co-occur within the first foot, so the structures *CaCo or *CoCa are generally forbidden. In (6a) I present examples of roots with non-front vowels, and in (6b) of roots with front vowels. (6) a. Κ Τ aqa Κ Τ nono

mouth loop

b. ögö mere-

fish trap to fly

Introduction 37 A number of exceptions such as toyal 'scum' or Τ mojarqs 'softened (fish)' are insignificant. They may result from erroneous recording. Vowels in front stems do not necessarily agree in roundness, e.g. Τ köde 'man', Τ köne 'friend'. Such stems normally have the structure CöCe (but not *CeCö), while the first consonant is the labial ρ or m, sometimes also k or s/s. This suggests that originally the stem was CeCe, but the first vowel was labialized under the influence of the word-initial consonant (see 4.1.2). In other words, the lack of rounding harmony is here secondary.

3.3.2. Harmony of short vowels after the first foot After the first foot the harmony is mainly restricted to short non-high vowels. All high vowels, most of which are epenthetic (see 3.6), are harmony-neutral. As was mentioned in 3.1.1, the only short non-high vowel after the first foot is a. In roots and derivational suffixes this vowel can harmonize to the vowels of the first foot. Unlike in the first foot, this kind of harmony is optional. Some speakers always pronounce a, while in the pronunciation of other speakers this vowel shows partial harmony in frontness and roundness to the vowel(s) of the first foot in the following manner. (7)

first foot a ο e



a ο e

[saqala] or [saqala] [monoyo] or [monoya] [terike] or [teriks]

fox moufflon wife

As shown in Nikolaeva (1988a), harmony is more likely to occur in the stressed position, while the non-stressed a normally remains phonetically neutral. This kind of optional assimilation of a has been described as harmony in previous studies on Yukaghir and is the main reason for the transcriptional inconsistencies in the existing sources. The vowel a may be rendered as a, o, or e, even within the same word and when recorded by the same author. This is shown by the following examples. (8)

a. a:c9 'domestic reindeer' KK a:t'e, at'e, a:c'a, a:c'e; KJ a:ce; KD a:ce; SD aca, ase-, ace-; SU -aca; RS -aca; Μ äca, aace, aaca; MO amd'e [rect. atce]; Β aaitsha; ME ointscha; MK äatsche, aatschen-


Introduction b. touka'dog' KK touke, tewke-, towke-, touka; KJ toboko, toboke; SD tobuka; RS toweka; Μ towoka, towoka, towoko-, towoka-, touka; MC tawala [rect. tawaka]; Β tabaka; ME tavacke c. unema 'ear' KK uneme-; KJ uneme; SD unume; SU unema; RS unama; MC inama; BO inemo-; KL unemo; Β oonomma; ME unoma; MK unjäma d. soroma 'man' KK soromo; KJ soromo, romo; KD coromo, coro, romo-; SD soromo; TD -soromo ; SU coroma; RS -soroma; Μ soroma, soromaga; KL soroma, soromo; Β toromma, -toromma; ME torroma, -schoromok; MK tschrroma

Such variations do not normally occur with short vowels within the first foot, which are rendered in the same way by all authors, cf. (9)

a. aqa 'mouth' KK aqa; KJ aqa; KD aqa; SU äga; RS angä; MC aigga [rect. angga], anga; MO aigg [rect. angg]; BO anäriga; KL angaga; Β angä; ME anga; MU anggä; MK änga b. moyo 'hat' KJ moyo; SD mogo; Β mogo; MU mongo, but ME mochga

Inflectional morphemes are normally excluded from the harmony. By inflectional I mean case and number suffixes for nouns and tense/mood and agreement suffixes for verbs, as well as suffixes of non-finite verbs. In contrast, verbal aspectual and voice-like affixes do harmonize. In inflectional morphemes all non-high short vowels are pronounced as a or (stressed) e (stressed a in Tundra Yukaghir). For example, moyo-g3 'hat (LOC)' may be pronounced as [moyoge] or [moyogd] and αηα-h 'mouth (INSTR)' may be pronounced as [αηαΙέ\ or [ayah].

Introduction 39 3.3.3. Harmony of long vowels There are two types of long vowels, harmonizing and non-harmonizing. Affixes that contain the long o: do not harmonize, probably due to the fact that in most cases o: has appeared as a result of the contraction of a short vowel and a consonant (see 4.1.4). Some examples of non-harmonizing affixes with long ο: are Κ -bo:-, Τ -bo\l- (Qualitative) and Κ Τ -(η)ο\(Stative/Resultative). They are compatible with both front and back stems, cf. the Resultative forms kuded-o:- 'killed' and amd-o:- 'dead'. 3 In addition, there is a group of non-harmonizing affixes containing the long a\ and e: such as Κ Τ -e:- (Causative), Κ Τ -de: (Diminutive), and Τ -na:(Inchoative). Other affixes with long non-round vowels do harmonize. Harmonizing affixes are represented by two variants, one of which co-occurs with back stems and contains the vowel a:, and another that co-occurs with front stems and contains the vowel e:. Roundness harmony does not operate in this case. Some examples of harmonizing affixes with long vowels are Κ Τ -a\-l-e:(Inchoative), Κ Τ -qcr.-l-ke\- (Inchoative), Κ Τ -ccr.l-ce\ (nominal derivational suffix), and Κ Τ -d'a.-l-d'e\- (Intransitive). This is demonstrated by the following Inchoative derivations (the consonant / is epenthetic, see 3.6.1). a. anyi:qoyi:im-


to scratch to pierce to sit down

b. jugibjö> mere-


to to to to

kiss suck see fly

The same behavior is observed for affixes containing the diphthongs ej/aj (see 3.1.1), for example: (11)

a. jatayajoq-na-


to straighten to open

b. selgsju:-ha-


to break to smoke



If a harmonizing affix follows the long o:, it harmonizes to the first vowel of the word in bisyllabic words (12a). In trisyllabic words both front and back variants of the affixes are available (12b). (12)

a. kizo:jero:-


light shallow

b. nigejo:ceginmo:-

nigejo:-l-e:nigejo:-l-a:heavy ceginmo:-l-e:- ceginmo:-l-a:- deep

As discussed below in 3.4.1, the vowel a is lengthened in certain phonological environments, in which case it harmonizes to the vowels of the first foot.

3.3.4. Synharmonism As was mentioned in 3.3.1, every stem in Yukaghir is either back or front. This distinction conditions the synharmonism, i.e. the distribution of velars and uvular consonants. Front stems may only contain k and g, while back stems are only compatible with q and y. The domain of synharmonism is the root augmented by derivational affixes, for example: (13)

qolyut kigi:-

mammoth to prick


to pierce (ITER) to roll (ITER)

Synharmonism is not observed in recent borrowings, e.g. kal'ip 'iron pincers for casting shot, clay mould for casting bullets' < Rus. kalibr and Τ qune 'two year old male reindeer' < Ev. kma:- 'to gallop (of a reindeer)'. In inflectional morphemes consonants fail to harmonize. In this case Kolyma Yukaghir has generalized the front variants with g and k, and Tundra Yukaghir has generalized the back variants with q and y, see the following examples with the Locative affix Κ -ga ~ Τ -ys rendered in the phonetic transcription. (14)

[locil-ge] [labut-ke]

in the fire in the snag

Τ [nime-ya] Τ [mat-qa]

in the house by me

The generalization of one harmonic variant seems to be a rather recent process. Jochelson's materials (from the end of the 19th century) show that



at least in certain cases synharmonism operated in inflectional morphemes as well, e.g. KJ mada:-q 'sit down (IMP)' ~ mada:-k, KJ jo-yole 'head (ACC)' ~jo:-gsh. Vowel harmony was also more widespread, especially in the now extinct Korkodon variety of Kolyma Yukaghir.

3.4. Alternations of vowels Sections 3.4 and 3.5 address only those alternations that are relevant for the transcription and reconstruction suggested in this volume.

3.4.1 Quantitative


There are two basic types of vowel lengthening in Kolyma Yukaghir. First, a full short vowel (i.e. not a) can be lengthened for prosodic reasons. Monosyllabic nouns have the structure (C)V: or (C)V:C, while the structures *(C)VC or *CV are impossible. This distribution can be explained with the assumptions that (i) the minimal lexical word must correspond to a foot (McCarthy and Prince 1995); (ii) the foot in Yukaghir is bimoraic, and (iii) the word-final consonant does not project a mora. In other words, bimoraicity is satisfied by vowel length. In most monosyllabic words the vowel is genuinely long and does not alternate with short vowels, cf. o: 'trousers' and ο:-ήα:-Μ 'apparently had trousers (trousers-PROPR-EV.INTR.3SG)'. Such vowels are reconstructed for Proto-Yukaghir as long. Other stems exhibit alternations with short vowels, e.g. e:r > je:r 'furkation' but ere-l'a- 'to spread (legs)', and l'e:-j 'is (INTR.3SG)' but te-t-i 'will be (FUT-INTR.3SG)'. The vowel here is underlyingly short, but lengthened when the word is monosyllabic to maintain bimoraicity. For such stems I reconstruct a short vowel. Second, before the bimoraic affixes CV: or CVC the vowel a lengthens into e: or a:, depending on the harmonic quality of the stem. As was shown above, rounding harmony does not operate in this case. Short vowels other than 9 do not lengthen in this context. Some examples of bimoraic morphemes that cause this kind of lengthening are Κ Τ -de: (Diminutive), Κ Τ -Μ- (Evidential), Κ Τ -nun-( Habitual), -ji:-/-ci:-/-d'i:- (Iterative), Τ -ci\(Causative), -d'a:-/-d'e:- (Habitual), Κ -bo:-, Τ -Z>o.7-(Qualitative), and Κ -saj-/-sej-, -daj-l-dej-, Τ -raj-l-rej-, -taj/-tej- (Perfective Transitive). (15a) illustrates the Diminutive, and (15b) the Evidential.

42 (15)

Introduction a.

a:ca po:re önms ejma

aca:-de: pora:-de: önme:-de: ejme:-de:

reindeer spring mind price



joyortaM'd jaqta:-l'ol egctc:-l"ol ejre:-l'3l

to to to to

wound sing put, to raise walk

This lengthening is caused by the need to parse the word into bimoraic feet (see Nikolaeva 1998 for details). The bimoraic affixes that take the final position in the word, e.g. the Ablative -gat, the 2 nd person Plural transitive -mat, the 3 rd person Plural transitive -ηα:, the Predicative -lak, and the Dative -ηίη, do not cause lengthening. Vowel shortening occurs in (at least some) (C)V:Cs stems, e.g.: (16)

ha:ca ha:cs

edge face


to hit to return (INTR)

Alternations take place when there is a long vowel or a diphthong in subsequent syllables. So the vowel length depends on the phonological environment, namely, the presence of bimoraic syllables (except CVC) later in the word. This suggests that it may be related to stress. Stress is not addressed in detail here. In short, it is quantity-sensitive, although the rules of stress assignment differ slightly from those described in Maslova (2003: 58-59) after Nikolaeva (1988a). The stress falls on the rightmost syllable with a long vowel or a diphthong. Otherwise it falls on the right-most CVC syllable, and in the absence of the latter on the last syllable. 4 So, for example, in na.cs the stress falls on the first vowel and in nacayadaj- the stress falls on the diphthong aj. The stem shows the alternation naca- (where the first vowel is unstressed) ~ na:cd (where the first vowel is stressed). The question is whether a short or a long vowel is underlying, and how to formulate the condition for the alternation. There are various solutions to this problem, but in this work I adopt the analysis under which the vowel length is primary. This is reflected in the transcription of such stems as (C)V:Cs throughout the dictionary. The underlying long vowel may be unstressed and shortened if there is a stress-attracting syllable further in the word. See 4.3 for more information on the structure of the stems and their historical origin.

Introduction 3.4.2.



Vowels are deleted in several environments. First, deletion is conditioned by the ban on vowel clusters (see 3.6). If a vowel-initial affix follows a vowel-final stem, this creates the potential context for a hiatus. The most frequent vowel-initial affixes are the following: Κ Τ -ο:- (Resultative), Κ Τ -o:k (1st person Plural Interrogative), Κ Τ -a:-/-e:- (Inchoative), Κ Τ -aj-/-ej(Perfective), Κ -ο:ΐ- (Desiderative), Κ Τ -i:- (Causative/Transitive), Κ Τ- -i: (Noun), Κ Τ -e: (Causative/Transitive), Κ Τ -u: (Noun), and Κ Τ -u:(Intransitive). To avoid a hiatus, the vowel a is deleted before such affixes, as is shown by the following Resultative examples. (17)



to kill to play to curse

Deletion does not apply to long vowels; instead an epenthetic consonant is inserted to avoid the violation of syllabification constraints (3.6.1). Full short vowels demonstrate mixed behavior: normally they are preceded by a consonant epenthesis, but in some cases they are deleted like a. As argued in Nikolaeva (1998), the choice of a strategy (vowel deletion vs. consonant epenthesis) is determined by prosodic considerations. Second, the vowel a is deleted in the sequence of two underlying light syllables CaCa. The deletion applies to every second Ca syllable with the exception of the last syllable, so that the following strings are possible (V here is not equal to a). (18)

a + C + Ca + C ... + Ca# V/C + Ca + C + Ca ... + Ca#

Some of the affixes that show the alternation Ca/C are Κ Τ -t3-/-t- (Future), Κ Τ -ri3-/-n- (Proprietive), Κ -S3-/-S-, Τ -sd-/-s- (Causative/Transitive), Κ -dd-/-d-, Τ -rd-/-r- (Causative/Transitive), Κ Τ -bd/-b- (Noun), and Κ Τ

-pd-/-p- (Plural). It is also observed in CaCa affixes such as Κ Τ -mahf-mh (3rd person Singular Perfective Participle and Object Focus) and Κ Τ -tdgaZ-tka (Augmentative). (19)

a. ulta-t-mala jelda-t-mala

tied (FUT-PERF.PART.3SG) pulled out (FUT-PERF.PART.3SG)


Introduction b. aji:-ta-mla lek-ta-mla


The alternation a ~ 0 can in principle be described either as epenthesis or as the deletion of a. Here the second approach is chosen because, as argued in Nikolaeva (1998), there is a phonological contrast between the a-final and consonant-final words. Roots (C)V:C and (C)V:Ca clearly contrast in the bare form, e.g. a:ra 'baby's cloth' and qa:r 'skin'. In suffixes a phonological contrast exists, for example, between the word-final Perfective Participle suffix -ma and the 3rd person Singular suffix of transitive verbs -m, cf. qarta-ma 'divided (PERF.PART)' vs. qarta-m 'divided (TR.3SG)'. The epenthesis analysis fails to explain why epenthesis applies to some consonant-final forms, but not others. In other words, the conditions for epenthesis cannot be stated in phonological terms. So 3 is underlying and is included in the representation of affixes in Section 5 of this Introduction. Certain aspects of the Ca/C alternation were first described in Krejnovic (1978, 1982), although he did not consider strings containing several alternating affixes. Under his account, nominal and verbal stems fall into two classes depending on the quality of the final segment, so that Ca affixes are combined with one class, and C affixes with another. In Nikolaeva (1998) I argue against this analysis and show that the deletion of a is motivated prosodically, namely by the interaction of syllable and foot constraints. In brief, Kolyma Yukaghir exhibits a strong, though not exclusive, tendency for the underlying material to be parsed into bimoraic monosyllabic feet, i.e. the structures CVC or CV:. 5 This implies that unlike many other languages Yukaghir prefers close syllables to open syllables. The sequences CaCa are transformed into CaC because they better satisfy this requirement. Some affixes Ca(Ca) never show the alternations in question, i.e. they do not lose a when following a a-final stem. These are the affixes that always take a word-final position such as Κ Τ -la (Accusative), Κ Τ -l'a (Possessive), Κ -ga, Τ -ya- (Hortative), Κ -gala, Τ -yana (Definite Accusative), Κ -na (Comitative), Κ Τ -ma (Temporal), Κ -ga, Τ -ya (Locative), e.g. m/rfa-ma 'in autumn (TEMP)' and Τ sespa-la 'door (ACC)'. In addition, there are a number of non-final Ca suffixes that never show alternations, such as Κ Τ -da- (3rd person Possessive), Κ Τ -d'a(Intransitive), Κ Τ -da- (Intransitive), Κ Τ -ga-/-ya- (Iterative), Κ -za(Transitive), and Κ -ba- (Inchoative), e.g. pul'd'a-ga- 'to be loose (ITER)' and nanma-da-ga, Τ nanma-da-ya 'in his/her shrubs (POS-LOC)'. The reason these affixes do not alternate is that they contain voiced obstruents



that go back to consonantal clusters. At some stage of the history of the language the structure of the suffix was *CCo and perhaps it still has to be represented as such at the synchronic level (see 3.5.1, 4.2.5 and the list of affixes in Section 5). When it follows a Cs syllable, the sequence *C9-CCa is well-formed from the point of view of foot constraints, therefore deletion does not apply.

3.4.3. Assimilation

of mid


The mid vowels in the first syllable may assimilate fully or partially to the vowel of the second syllable. When the second syllable has the vowel a: or the diphthong aj, ο in the first syllable may change into a. The following examples show either the Inchoative -a:- or the Perfective -aj-. coypol-dsmodoorp-o:qonpoj-e:-

to to to to to to

cut take out sit hang go carry a burden


to to to to to to

cross the river escape sit down raise set off lift

In a similar manner, ο in some words becomes e when there is e: or i: in the second syllable, e.g. ροή-ο:- 'to stay, to remain' ~ pen-i:- 'to leave' and poj-e:— pej(j)e:~ 'to carry a burden'. The vowel ö alternates with ο and e before o: and e:, respectively. In (21) -o:- is the Resultative affix, and -e:- is the Inchoative affix. (21)


to enter standing fish trap


to be inside to stand to stand up

In some cases ö may be pronounced as u before a high vowel in the following syllable, cf. ugujd 'morning' derived from the same stem 'to stand'. These assimilative processes are not productive and must be specified lexically. Alternations are never observed in some stems with mid vowels such as, for example, lolya- 'to boil' and ono- 'to steal', cf. their Inchoative forms loly-a:- and on-a:-, respectively.



3.4.4. Reduction of the stem-final vowel Some affixes of the structure Ca do not show the Co/C alternation described in 3.4.2, but trigger the deletion of the stem-final vowel in verbs instead. This process has not been addressed in previous descriptions of Yukaghir, but it is important for reconstructing the structure of the stem. The vowels a, e, ο and sometimes a can undergo deletion at the end of the verbal stem. Examples (22a) are from Kolyma Yukaghir and (22b) from Tundra Yukaghir. a. pöme-gacice-gamodocinepöd'e-1

to turn to lengthen to sit to cut down smell


round long to seat to cut down smelling

b. loqo-yssila-

to protrude dry


protruded dry

In some cases the stem-final vowel is deleted before consonant-initial affixes that have another structure, and each case has to be lexically specified. (23a) illustrates Kolyma Yukaghir and (23b) illustrates Tundra Yukaghir. (23)

a. olokile-

to steal to wade


to steal vagrant

b. ile-tejlaja-ysra

to push back side


to push backward

In the absence of a stem-final vowel the epenthetic u ~ i may be inserted before the suffix, see 3.6.3. In some stems the final vowel is deleted when there is no further derivational affix. When the stem is followed by an affix, the vowel is preserved, e.g. coy- 'to cut' but coyo-jd 'knife', je:r < e:r 'fork' but ere-gs'to spread'.

Introduction 47 3.5. Alternations of consonants 3.5.1. "Sonorant ~ obstruent"


The voiced obstruents d, b, g/y and Κ ζ alternate with sonorants of the same place of articulation. Unlike in previous description of Yukaghir, all these alternations are analysed here as realizations of the same historically motivated process (see 4.2.5). Obstruents occur before a vowel, while sonorants occur before a consonant or a pause with the following distribution. (24)

before V d d' b g/γ Kz, T r

before C or # η ή m 0 η

Examples of the "sonorant ~ obstruent" alternations are presented below; (25a) exemplifies Kolyma Yukaghir and (25b) Tundra Yukaghir. before V kud-u-1 lud-u-1 ed'-u-l oy-u-nbani:zated-u-1

mud iron life to get wet to squeal, to squall wealth

before C kun-talun-buge en-d'a oq-d'a nen-yizanen-da-

to make dirty cauldron living wet to growl, to snarl to get rich

jab-a:mub-eganira-yasab-ayaod'e say-u-ss-

to die to shorten to grin to stretch drop, dew to lose


to be ill short to growl flat dampish to lose (ITER)

The consonants η and ή alternating with d and d' may undergo further assimilation if they occur before / or ΐ (see 3.2), e.g. mid'-u-m 'took (TR.3SG)' ~ min-td-m 'will take (FUT-TR.3SG)' ~ mil'-M-u-m < *min-M-u-m 'took (EV-TR.3SG)'.



The question that arises in a synchronic description is which consonant is underlying, a sonorant or a voiced obstruent. As will be shown in 4.2.5, the alternating consonants originate as homorganic clusters, i.e. neither of them can be treated as a historical source of the other. In principle it should be possible to analyze the alternating consonants as clusters at a synchronic level as well. This is supported by the fact that some affixes Ca containing a voiced obstruent behave as if their underlying structure were CCs with respect to vowel deletion (see 3.4.2). However, there are also arguments against this analysis. In particular, the structures (C)VCV, where the second consonant goes back to a cluster, end in a full short vowel rather than a, as is typical of (C)VCCo structures. In this they pattern together with bisyllabic stems without consonantal clusters (see 4.1.4). I therefore leave the question of the underlying representation of alternating sonorants and obstruents open.

3.5.2. Assimilation

of voiced


Apart from the frequent "sonorant ~ obstruent" alternation described in the previous subsection, Yukaghir exhibits the second type of alternation for voiced obstruents. In some stems and affixes voiced obstruents occur in their basic form before a vowel, but undergo various assimilative processes before a consonant or a pause. 6 These alternations differ from the "sonorant ~ obstruent" alternation in two respects. First, the quality of the resulting consonant depends on the quality of the following segment, i.e. whether it is voiced, voiceless or a sonorant. Second, the resulting consonant is not necessarily homorganic with the voiced obstruent which occurs before a vowel. Assimilative alternations of voiced obstruents are shown in the following table, where the first line indicates what follows the alternating consonant. Table 3. vowel d g Y b

voiceless obstruent t k q P

# t u~w u~w u~w

voiced obstruent u~ w u~w~η u~w u~w




n m m~r) u~w


1 u~w u~w u~w

ί) η

Introduction 49 Some examples follow. (26)

a. leg- 'to eat' leg-u-m (TR.3SG), lek-to-m (FUT-TR.3SG), leu-jo (IMPF.PART), leq-qa: (TR.3PL), lem-mok (TR.2SG), leu-dollo (SS.PERF), leu (TR.1SG) b. coy- 'to cut' coy-u-m (TR.3SG), coq-to-m (FUT-TR.3SG), cou-jo (IMPF. Ρ ART), coq-qa: (TR.3PL), com-mok (TR.2SG), cou-dollo (SS.PERF), cou (TR.1SG) c. töb- 'to cover' töb-u-m (TR.3SG), törj-qa: (TR.3PL), töu-mok (TR.2SG), töu-l'ol(EV), töu (TR.1SG), töp-ci:- (ITER) d. kebe- 'to leave' kebe-j- (PERF), köu-do- (TR) e. terikod- 'to marry' terikod-i (INTR.3SG), terikot-co (FUT.INTR. 1SG), terikot (IMP), terikol-l'ol- (EV)

The consonants d' and ζ do not show such alternations. Previous studies of Yukaghir (e.g. Maslova 2003: 3 9 ^ 2 ) describe some alternations discussed in this subsection and the previous subsection, but do not mention that they systematically fall into two types: the "sonorant ~ obstruent" alternation and the assimilation of voiced obstruents. The latter is typical of some stems and affixes, and the former of the others. As will be shown in Section 4.2, they have different origins: while voiced consonants alternating with sonorants go back to clusters, voiced consonants showing assimilation go back to single consonants. Consonants d' and ζ do not participate in the assimilative alternations because they cannot go back to a single consonant. In a synchronic description of Yukaghir these facts have to be lexically specified. In this dictionary stems exhibiting the "sonorant ~ obstruent" alternation are represented with a slash, e.g. jan-fjad- 'to send', whereas stems exhibiting assimilation are represented with a single consonant, e.g. köd- 'to gather, to tighten (a rope)'. The two processes in question prevent voiced obstruents from appearing in the syllable-final position (see 3.2). If a voiced obstruent occurs before another consonant, it



either assimilates partly or fully to it or alternates with a homorganic sonorant.

3.5.3. Voicing and palatal


Consonants in clusters usually agree in voicing and palatalization. The consonants η, ή, l, Γ assimilate in palatalization to the following consonant. The non-palatalized η and I are palatalized before Λ, l\ c or d\ although this process is optional and is not always reflected in the transcription adopted in the dictionary. Some examples were presented in Section 3.2. Depalatalization is observed for ή and /' , when they occur before a non-palatal consonant, e.g. an-t3- 'to answer' vs. an-na- 'to talk' and Τ pun-na:'to kill (INCH)' vs. pm-i-m 'killed (TR.3SG)'. In Tundra Yukaghir this process also applies across word boundaries, e.g. Τ tmt abuce:n (< abuce:-n) dadi:rj Ί gave(it)to my grandmother', where -Λ is the suffix of the Dative changed here into -n before the non-palatal consonant beginning the verbal form dadi:rj (< tadi:q). Further, as described in Maslova (2003: 43), voiced obstruents (stops or affricates) show progressive devoicing after voiceless obstruents, cf. the following examples involving the 3rd person Possessive suffix -gi and the Locative suffix -gd. (27)

lukul pulut qa:rbas

arrow old man Russian boat

lukul-gi pulut-ki qa:rbss-ki

lukul-ga pulut-ka qairbas-ka

Regressive devoicing takes place when a voiced obstruent occurs before a c which has emerged as a result of the assimilation j > c (see 3.5.4), e.g. terikdd-jak 'married (INTR.2SG)' > terikdd-cdk > terikat-cak ( > teriksc-cdk). In Tundra Yukaghir stops and affricates are voiced after a sonorant across a word boundary, while the sonorant may be optionally deleted. As shown in Nikolaeva (forthc.), this process applies within a syntactic phrase, mostly NP, PP or a Focus Phrase. For example, Τ lacil barih 'spark of the fire' < lacil 'fire' + parih 'spark', caljsdawur dite < caljadawur tite 'like the mittens', and caj-h barul < caj-hy (PRED) parul 'we made tea'. Similarly, the Negative proclitic a/- triggers voicing in Tundra Yukaghir, e.g. sl-bund'a < al-pund's Ί didn't kill'.

Introduction 51 3.5.4. Assimilation

of sibilants


The sibilants (K s, Τ s) alternate with r in the position after η and η both in Tundra and Kolyma Yukaghir and sometimes with d in Tundra Yukaghir, although these processes seem to apply selectively. They normally occur in compounds where the first component takes the Genitive affix -n. This nasal may fall out, e.g. ju:cd-ra: 'fire for smoking out mosquitoes' < ju.cd-n 'smoke' + sa:l 'tree, stick' and Τ ja:-rawa 'birch bark' < ja:-n 'birch tree' + saw a 'skin'. Examples of the alternation s ~ d in Tundra Yukaghir are the following: Τ egi-n-dukun 'collar' < egi-l 'back of the head' + sukun 'cloth', co.-n-duskd 'mug' < co: 'iron' + suskd 'bowl', and cumu-n-daburqa 'flat place on the hill' < cumu-r 'hill' + saburqs 'flat place'. In Tundra Yukaghir s changes into r in other environments as well, sometimes even across word boundaries, e.g. Τ köde rusej < ködetj susej 'the man threw (it)'. As shown in Tables 4 and 5, the consonant j never occurs as the second component of a cluster. It assimilates to the preceding consonant both across the morpheme boundary and inside a morpheme. The rules of assimilation are basically the same for Tundra and Kolyma Yukaghir and are as follows: j j > c (phonetically [i] in Kolyma Yukaghir), Cj > Cc, and Nj > Nd\ where C is an obstruent and Ν is a sonorant. The consonant r demonstrates mixed behavior: in some forms it behaves like an obstruent, while in other forms it behaves like other sonorants. These assimilative processes are very common in, although not restricted to, the paradigm of intransitive verbs, as most intransitive agreement affixes begin with j . Examples (28) are for 1st person Singular intransitive forms. modomalajpe:d9terqonoij-

modo-ja mala-ca pe:dat-c3 er-ca qon-d'a οη-d'a

to sit to fall asleep to burn bad to go wet

The process j j > c could have occurred through the intermediate stage jc, as is shown by the following variations found in Jochelson's materials: KJ malac ~ malaic < malaj-j 'fell asleep (INTR.3SG)'. Other frequent contexts for the assimilation of j are the suffix of the Subject Nominals -jo:n/-jo:d and the Iterative suffix -ji:, which may take the forms -ci: or -d'i:, e.g. tadi:-ji:- 'to give (ITER)' but jar-d'i:- 'to swim (ITER)' and sek-ci:- 'to enter (ITER)'. After the negative clitic a/- the stem-initial j does not change



into d\ Instead the following alternations are observed: a/ + j > ΐΐ ~ jj, e.g. al-jö:j > djjö.j ~ al'lo:] 'did not see (3SG)' and Τ al-jawaj > al'l'awaj 'did not die (3SG)'. In the word-final position assimilation of j after a consonant does not apply. Instead the glide is realized vocalically as i. For example, this occurs in the 3rd person Singular intransitive inflection, cf. modo-j 'sits (INTR.3SG)' but qon-i 'goes (INTR.3SG)' and ajhd-i 'washes (INTR.3SG)'. After u: the word-final j is realized as c [i], while the vowel changes into i, e.g. eru:- 'bad' but eri-c 'bad (INTR.3SG)' < eri:-j < eru:-j. Verbal stems ending in the Intransitive affix -u: are numerous and are represented with u: in the dictionary, although u: is absent in most inflectional forms. After i: the word-final j either falls out or changes into c, while the vowel is shortened, e.g.pöm-c ~ροήί: titi-me:-d'd (INTR.1SG).

3.6. Epenthesis The syllabification rules in Kolyma Yukaghir forbid vowel-initial syllables, except at the beginning of the word. This means that vocalic clusters are ruled out.7 A possible hiatus is prevented either by vowel deletion (see 3.4.2) or by consonant epenthesis. Further, more than one consonant in the onset or coda



position are ruled out. This means that consonantal clusters are disallowed at the word edges, and no word-medial cluster can contain more than two consonants. 8 To prevent impossible clusters epenthetic vowels are inserted.

3.6.1. Consonantal


In Kolyma Yukaghir the epenthetic consonant is /. the same vowel-initial suffixes that cause vowel 3.4.2. Normally an epenthetic I is inserted after vowel, but in some cases it can also follow a short examples (29), which involve the Resultative -o:-. (29)

a. monsög-


b. jaqaiii:d'i-

jaqa-l-o:to arrive ni:d'i:-l-o:- to tell

It generally occurs before deletion, as described in a stem ending in a long vowel not equal to a. See

to say to come in

Maslova (2003: 56) notices that the epenthetic / is also inserted after j, e.g. ukej-l-o:- 'to exit (RES)'. As was argued in 3.1.1, this demonstrates that j in such cases counts as vocalic. In Tundra Yukaghir other epenthetic consonants may apply. For example, j functions epenthetically as in Τ ki-j-o:- 'two' vs. ki-n 'two (ATTR)' and Τ memda-j-o:l- 'to be prepared (RES)' vs. memda- 'to prepare'. The consonants r and η are inserted after the proclitics md- and Λί-, respectively, if the latter precede a vowel-initial stem. Examples can be found in the respective entries of the dictionary. In contrast, in Kolyma Yukaghir proclitics are not accompanied by epenthesis when they precede a vowel-initial stem. Instead, either the first vowel of the cluster is deleted or the cluster is preserved in violation of the constraint that prohibits vocalic clusters.

3.6.2. Vocalic epenthesis in word edge clusters According to Maslova (2003), the only epenthetic vowel in Kolyma Yukaghir is u. Indeed, this is very frequently the case, but sometimes i also functions epenthetically, although this has to be lexically specified. For example, i occurs between the stem-final d and the nominal derivational



suffix -/ in the word ad-i-l 'youth', cf. ad-ö: 'son' (lit. 'youth child'), as well as before the Dative Κ -η, Τ -ή in the pronominal forms mat-i-n ~ Τ nidt-i-ή 'to me (DAT)' and the like. In older Russian loanwords i was inserted at the beginning of the word to prevent initial clusters, e.g. istena: < Rus. stena 'wall' and KK istakan < Rus. stakan 'glass'. The variations u ~ i are also possible, e.g. sar-u-m ~ sar-i-m 'covered (TR.3SG)'. The epenthetic vowel is systematically inserted after a consonant-final stem before a suffix of the structure -C which takes the final position in the word. These are, for example, the following suffixes: K T -m (3rd person Transitive), K T - / (nominal derivational suffix and Action Nominal), K T - / (Pronominal Accusative), and Κ -t, Τ -r (Same-Subject Converb). Examples (30a) illustrate the Action Nominal, while examples (30b) show the 3rd person Singular Transitive forms. (30)

a. mid'-u-l (T med'-i-l) kel-u-1 mon-u-1

to take to come to say

b. mid'-u-m leg-u-m pad-u-m

to take to eat to cook

Some word-final suffixes of the structure -C never trigger epenthesis for various reasons. The Imperative -k falls out after a consonant in Kolyma Yukaghir (but not in Tundra Yukaghir), the Genitive -n/-d replaces the stem-final consonant, the 3rd person Intransitive -j turns into -i (see 3.5.4), and the 3rd person Hortative -n always occurs after the Hortative morpheme -gs, so the need for epenthesis never arises. Kolyma Yukaghir also employs the epenthetic d, which occurs only beween the consonant-final stem and the Predicative morpheme -k, e.g. mat-d-k Ί (PRED)' and pulut-d-k 'old man (PRED)'.

3.6.3. Word-medial vocalic


Word-medial vocalic epenthesis is addressed neither in Krejnovic (1982) nor in Maslova (2003), but it is in fact very frequent. Epenthetic vowels that occur word-internally are u or i, and variations are possible, e.g. izilbd ~ izulbd 'tiredness'. Word-medial epenthesis serves to prevent potential three-consonantal clusters. For instance, an epenthetic vowel must precede the complex Inchoative affix -Ibd- (from / + ba) when it follows a consonant-final stem, e.g. er- 'bad' > er-u-lbs- 'to become bad' and ad'strong' > ad-i-lbd- 'to become strong'.



In other cases epenthesis occurs in phonotactically ill-formed biconsonantal clusters. Consonant-initial affixes immediately follow a consonant-final stem if the resulting cluster is allowed by the phonotactic constraints, but are preceded by epenthesis otherwise. This concerns such affixes as Κ Τ -m(u)- (Inchoative), Κ -zu- (Iterative), Κ Τ -ji:- (Iterative), Κ Τ -td-, -sä- (Causative/Transitive), and Κ Τ -n(u)- (Imperfective), and some others. For instance, -n(u)- follows a i-final stem in sörihs-n(u)- 'to write (IMPF)' and -td- follows q in joq-td- 'to bring (TR)' because the clusters sn and qt are well-formed; -ji:- follows an r-final stem in jar-d'i:- < jar-ji:- 'to swim (ITER)' because it assimilates to d' in this position; and so on. On the other hand, if the potential cluster is bad, epenthesis applies, e.g.: adej-reabudjedlegsoy-e:ede:s -

strong to walk to pull to be seen to eat to lose one's way to call


to to to to to to to

become stronger walk pull out be seen feed lose call

As described in 3.4.4, the stem-final vowel may fall out. In some words an epenthetic vowel can be inserted instead of the deleted vowel of the stem. This process is extremely frequent and may apply even if the potential cluster is well-formed. Examples (32a) and (32b) illustrate the vowel deletion and epenthesis insertion before various affixes in Kolyma and Tundra Yukaghir, respectively. jaqaligejodomoromodoamla-dajporqa-ja jömga-

jaq-u-ji:to come lig-u-muold jod-u-tsto bind mor-u-sato dress mod-i-bs to sit to swallow aml-u-jba curved riverbank porq-u-sajjömg-i-ja to spin


to to to to

stay become soft frighten cover


to come to become old to twist to dress seat sunset to bend whirlpool to be late soft frightening extra blanket



Generally speaking, the conditions on word-internal epenthesis are not entirely clear in the sense that many instances of epenthesis seem to have become lexicalized. But most occurrences of short high vowels in non-first syllables appear to be epenthetic or at least have an epenthetic origin. For example, the Habitual morpheme -nun- is likely to go back to reduplication of the Imperfective affix -n(u)- with the epenthetic u between its two occurrences. The vowel has now become part of the Habitual affix. The same is true for the Supine affix -din (< dn < da-n) and some other affixes. Such affixes are represented with a vowel in the list of morphemes in Section 5. On the other hand, although some affixes are frequently accompanied by epenthesis, this does not apply in all contexts. For example, the Inchoative -m- and the Imperfective -n- are normally followed by u, but the latter is absent before a long vowel and in some other cases. Such affixes are cited without a μ in Section 5.



4. A reconstruction of Proto-Yukaghir phonology This section suggests a reconstruction of Proto-Yukaghir phonology. By Proto-Yukaghir I mean a common ancestor of all known modern and extinct varieties of Yukaghir.

4.1. Vocalism 4.1.1. Inventory and vowel harmony The following short vowels are reconstructed for Proto-Yukaghir. Note that *ii and *y are not present in modern Tundra and Kolyma Yukaghir. (1)

front vowels back vowels

i y

e a

ö ο

(ii) u

As in modern Yukaghir, non-high vowels were reduced and realised as a in the position after the first foot. This vowel was not phonemic, as its distribution is fully predictable from its position. In some environments d could undergo lengthening, while in other contexts it could be deleted. The vocalic system in (1) exhibits frontness harmony. The distribution of stems into harmonic classes remained practically unchanged. This is because, as I show below, the modern u goes back to *ii and therefore behaves as a front vowel, while *y has changed into i with back harmonic properties. So the only difference between (1) and the system described above for modern Kolyma Yukaghir is that in the latter there is no front correlate to u. It is also possible that Proto-Yukaghir had roundness harmony similar to that present in Kolyma Yukaghir (see 3.3.1). As all non-high short vowels were realised as a after the first foot, roundness harmony was only limited to the initial structures (C)VCV. The correspondences of Tundra and Kolyma Yukaghir vowels are presented below. I do not discuss here irregular correspondences that are only represented by one or two examples. They are usually mentioned in the text of the dictionary and alternative reconstructions are provided.

4.1.2. Non-high vowels PY *a > Κ Τ a This trivial correspondence is illustrated by the following examples: Κ Τ αηα 'mouth', Κ aqca ~ Τ waqcd 'angle', and Κ Τ qaml 'eagle'.



PY *o > Κ ο, M;TO,Ö Examples of the correspondence Κ ο ~ Τ ο are Κ porjdd ~ Τ poyodd 'money', Κ toyal 'scum' ~ Τ toyul 'used tea leaves', and Κ Τ omo 'people'. On the other hand, Κ ο can correspond to Τ a, cf.: Κ omducobul ozu: tolou odul jomil morosouya

Τ amducawul warulu: talaw wadul hamil marasawya

to hurry sea root wild reindeer Yukaghir neck to put on wooden bowl

Not all varieties of Tundra Yukaghir exhibit a in these cases. Jochelson mostly records o, Krejnovic ο or a, and Kurilov a, cf. TJ wodul ~ Τ wadul 'Yukaghir', TD wodurci— Τ wadurci- 'to try, to stand firm', and TK wolod'e ~ Τ waluod'd 'prize'. In some words the variation ο ~ a is attested in Kurilov's materials. According to Kurilov (1987: 7), this reflects two local idioms of Tundra Yukaghir, Qayaji: and Alaji: (on them see 2.2.1). The vowel a is typical of the Alaji: variety, and the vowel ο of the Qaqaji: variety. Other Yukaghir idioms, including Old Yukaghir, show ο in these cases. Even the Omok idiom, which phonologically stands the closest to Tundra Yukaghir (see 4.2.3 and 4.2.4), exhibits ο in the place of Τ a, e.g. MO wogo ~ Τ waya 'face'. Observe also the following example: MU olloga, MK oljogci ~ Τ al'yd (< *al'ayd) 'fish'. This suggests that Τ a goes back to o, as is also supported by the following Even words: Τ lalima ~ Ev. nolima 'sledge' and Τ abuka in abuka-mayil 'fur coat without a traditional triangle gusset' ~ Ev. obuka 'man's clothes'. While the direction of borrowing is unclear in the former case, the latter word has Tungus correspondences (see TMS 2 4) and so is borrowed from Even into Yukaghir. This indicates that in some Tundra Yukaghir words the change *o > a took place. This change was fairly recent and typical of the Alaji: variety, but has now become the norm. There are also Even loanwords where the Even a corresponds to Τ a, e.g. Τ ama: < Ev. atna: 'father', Τ qabal's- 'to grow bald' < Ev. qabata 'bald'. This raises the question of the phonological environment of the change *o > a. There is no clear answer to this question. The examples in (2) suggest that the change occurred in the context of the labial consonants m, b or w, probably as a result of dissimilation. This seems to be true for most instances



of the correspondence Κ ο ~ Τ a. However, there are exceptions. For example, in (3a) this correspondence does not occur in the context of labials, while in (3b) the presence of a labial does not trigger the change of ο into a in Tundra Yukaghir. Κ nodo locil jonod'sjotni:-

bird fire to tidy up to drive

Τ nada lacil janud'i:jatni:-

wolf fire to collect one's belongings to drive

omo qobo

people down

omo qobo

people on the surface

I reconstruct PY *o in stems with the correspondence Τ a ~ Κ ο. If we only have Tundra Yukaghir forms with a at our disposal but no Kolyma Yukaghir correspondences, it is impossible to tell whether *a or *o should be reconstructed. I reconstruct *a in such cases, but this is merely conventional. For example, for Τ cabuu 'tin (box)', which has no corresponding Kolyma word, the reconstruction is *campu\, but in principle it could also be *compu\.

In some infrequent cases, the reverse correspondence is also observed, i.e. Κ Ü — T o . Again it mostly occurs in the context of the labials, e.g. (4)

Κ parcaysarpoalma

Τ porcsysworpswo I ma

to chatter cautious shaman

This indicates that either Kolyma Yukaghir showed the same dissimilation process *o > a as Tundra Yukaghir, albeit to a lesser extent and in different words, or that the vowel here should be reconstructed as *a, in which case Τ a must have changed into ο when next to the labials. In the absence of clear evidence I have left this question open and have provided alternative reconstructions for this case. For example, the first stem in (4) is reconstructed as

*parcd-/*porc9-. PY *ö > Κ ö, ιι; Τ ö

The reconstruction of the PY *o is based on the trivial correspondence Κ ö ~ Τ ö, cf. Κ Τ köd'e 'worm', Κ Τ p ö m m - 'round', Κ Τ örd'd 'middle', and Κ Τ pög'to gallop'. In Old Yukaghir and Schiefner's materials ö was mostly rendered



as ο or e, but this is likely to represent an auditory error. Jochelson, Krejnovic, Kurilov, and Maslova are not always consistent is rendering ö either. Jochelson systematically records ö only before n. The correspondence Κ u ~ Τ ö results from the change of Κ ö into u under the influence of the high vowel of the following syllable. This change is relatively recent, since Jochelson records ö, e or ο in this case, cf.: Κ unmut unuq pulut ud'il'

KJ önmuunuq, onpolut od'il', ed'iP

Τ önmur önu: pölur öd'il'

antler, horn river old man nail, claw

The change *ö > u motivates the morphonological alternations in some stems, cf. Κ öwja 'real, true' ~ ubuj 'true, truth'. It is also observed for long vowels, e.g. Κ cö:l'd 'ancient' ~ cuild'i: 'tale' ( < KJ cuoled'i). PY *e > Κ e, ö, i; Τ e, ö This phoneme is pronounced as e in modern Yukaghir, but in Old Yukaghir sources e is often rendered as a or o, which may indicate that its pronunciation was closer to the low ä, e.g. Β amea, ME aime, MU omei, KL amei ~ Κ emej 'mother'. Jochelson consistently writes ä in the place of the modern e. PY *e can be illustrated by the following correspondences: Κ Τ epe: 'grandmother', Κ Τ en-/ed- 'live', and Κ ejrs- ~ Τ ewrs- 'to walk'. The vowel *e could labialize after the initial labial consonants ρ and m, e.g. Κ mödi:-, Τ möri:- 'to hear, to understand' (< PY *meö-) ~ TU *mede'to feel, to notice' and Κ pömns-, Τ pomna- 'round' (< PY *peme-) ~ TU *pem-/*pim- 'to wind, to be twisted'. Some words show the ö ~ e variations within one idiom or between idioms, e.g. Κ medin ~ mödin 'as soon as', Κ memz3j3 ~ Τ mömdija 'flame', and Κ pöril' 'toe' ~ Τ peril 'kick', pörinds'to kick'. The process of labialization could have taken place after the nonlabial consonants k and s/s as well, as shown by the following words: Κ söbul ( > subul) 'branch of the larch tree; bedding made of larch branches' ~ NT *seg-l*sew- 'to lay branches in a yurt' < TU *seg- 'to lay branches in a yurt' and Κ sere- ~ söre- ~ TU *sere:- 'to embroider'. See also the following correspondence and variations in the modern languages: Κ könma ~ kenma 'friend', Κ sog- ~ Τ seg- 'to enter', and Κ SöjI ~ sejl 'small stone, pebble'. In such cases I provide alternative reconstructions with *e and *ö. In some words the quality of the first vowel in Proto-Yukaghir can be reconstructed from the quality of the vowel of the second syllable. For example, the change *ö > e has probably taken place in Τ körel 'devil', Τ mörelwo:-



'melted' and Τ köne. These words have e in the second syllable and are likely to have had it in the first syllable as well, if we assume roundness harmony. The respective Proto-Yukaghir reconstructions are *kere~, *merelwo:- and *kene. The labialization *e > ö explains the violations of the roundness harmony in front stems in modern Yukaghir addressed in 3.3.1. Kolyma Yukaghir exhibits the change *e > i before a high vowel, parallel to the change *ö > u observed above, e.g. Κ mi:d'i: ~ KJ med'i 'sledge' and Κ ne.d'i.l, m:d'il ~ K K med'uol

4.1.3. High



PY *w> Κ ii, i; Τ u, i; PY *i> Κ i, u\ Τ i, u

Most Proto-Yukaghir high vowels are represented by the two trivial correspondences: PY *u > Κ Τ u (e.g. Κ Τ murga 'thick forest', Κ Τ kukul 'devil', Κ Τ nug- 'find') and PY */ > Κ Τ i (e.g. Κ igeja ~ Τ igijs 'rope', Κ Τ kind's 'moon', Κ Τ cilga 'branch', Κ ib— Τ iw- 'to suck'). Other words demonstrate the opposite distribution in the sense that i in one language corresponds to u in another language. In (6a) I cite examples for the correspondence Κ i ~ Τ u and in (6b) for Κ u ~ Τ i. Κ a. lige:mided'a iiigers mimil

Τ luge: murid's huqura mumul

old needle side young antlers

b. kuce: numö

kice: nime

mosquito house

Old-Yukaghir materials show either u or i in such cases, cf. MK niimu, MU numa

'house', M K


~ Τ umujd

'pike', and M K



midendscha 'needle'. There are reasons to think that changes of the high vowels occurred under the influence of the neighbouring consonants. In particular, i could develop into u when adjacent to a labial. This is confirmed by the word 'house' related to TU *nimä. Τ nime preserved the original vowel i, while in Kolyma Yukaghir i changed into u, apparently under the assimilative influence of m. Also observe the potential relation between Κ cumuci: 'fishing rod' and NT *ciye- 'to fish with a fishing rod', where Κ u could have developed from i next to the labial m. The opposite process *u > i was likely to occur in the context of a palatal



consonant, cf. Κ nugers, mgera 'side' related to TU *ήιιηηί: 'thigh', Κ nugen ~ Τ niijin, nugen-, ήιιηιιη- 'arm' related to TU *nurja 'muscles; elbow', and Τ mir 'brook, source' related to TU *uni/*iim id. If the assimilation explanation is true, we can expect that the words 'young antlers' and 'needle' contained */ and should be reconstructed as *mimil and *miSencd, respectively. The change *i> u occurred in Tundra Yukaghir. The same can be assumed for Τ umuja 'pike' with respect to OY *imojs (in modern Kolyma Yukaghir this word is absent). Assimilation to another vowel also played a role here. In the modern Yukaghir languages the alternation u ~ i may depend on the quality of the vowel of the second syllable, cf. Τ niqe: 'stomach, mood, soul' but Τ nuqurukun 'apron' (literally 'stomach thing'), and Κ pige 'box' but lun-buge 'cauldron' from lud-ul 'iron' (< *lunt-) and pige 'box', where the following development is assumed: *lun + *pige > limbige > lunbuge. However, several words such as Κ lige- ~ Τ luge- 'old' still remain unexplained and are provided with alternative reconstructions in the dictionary. PY *w > Κ Τ u The reconstruction of the vowel *ü is problematic, as it is not represented in any modern Yukaghir variety. This vowel is not used in the reconstructions provided in the dictionary and *u is written instead. Yet, there are two reasons to think that *ii could have been present in Proto-Yukaghir. First, *ü can be reconstructed based on systemic considerations as a front counterpart to PY *u. Second, in both modern varieties of Yukaghir u behaves as a harmonically front vowel, although its phonetic quality is back (see 3.3.1). As was first suggested in Nikolaeva (1992), this mismatch can be explained if the source of Κ Τ u is the harmonically front *ii and the harmonic quality of the stem remained intact after the change *ii > u. In the modern languages u is harmonically back in only a few words, mostly from Tundra Yukaghir. Such words are normally fairly recent borrowings from Tungus. Examples follow. (7)

Τ culal 'ermine' < TU colci 'pole-cat' Τ culd'dyd- 'to twitter' < TU *co:li:- 'to gossip, to babble' Κ Τ quno: < Εν. kunaw- 'to make noises (of a gull)' Τ qme 'two-year old male reindeer' < Ev. kuna:- 'to gallop (of a reindeer or a horse)' Τ cubayaj- 'to dive' < T M *cup-/*cop- 'to dive; deep place' Τ curqd 'two-year old female reindeer' < T M *cur-/*cir- 'three-year old wild deer'



In such cases u has a source different from *w, normally ο or the harmonically back u. This explains its back quality in modern Yukaghir. As there are no or veiy few cases of the original back w, we might suspect that *u changed into another back vowel before the change *ü > u took place. This matter is still unresolved, but at least in some words the change u > ο could have occurred. Observe the following words probably borrowed from Tungus. (8)

Τ payad'i: (with the further change into a in Tundra Yukaghir) < TU *puyu 'awl' Κ Τ moli modal particle < TU *mulri- 'doubt, not be able' Κ pol- 'gum' < TU *pul- id. Τ ma.ro:- 'happy' (probably from *mars-o:-) < TU *murun 'happiness, mind' Κ ρ ο ΐ - 'to slip' < TU *belu-/*bul 'flat slippery ice surface'

Such loanwords are older than most of those that demonstrate the correspondence TU/Ev. u ~ Yuk. u, e.g. Κ kuru:k < Ev. or Yak. kuru:k 'always' and Τ muktija 'knife with the tip cut off < Ev. mukti 'blunt'. However, Τ kokota: 'closed end of a sleeve in a child's overalls' < Ev. kukatan 'gloves' may be quite recent. PY *y > Κ i, ja-, α; Τ/, e The PY *y is reconstructed based on the following evidence. First, as shown in 3.3.1, both Kolyma and Tundra Yukaghir have words where i is harmonically back. I suggest that the corresponding stems in Proto-Yukaghir contained the back *y. Second, in some cases the back i in Tundra Yukaghir corresponds to Κ a or ja- (in the word-initial position), e.g. Κ sasayaj— Τ sisayej- 'to become torn' and Κ sarqul 'paw' ~ Τ sisqa 'finger', cf. also KJ qid'ilbe- ~ Κ qad'ilbd'laugh'. If the last word is not an erroneous recording of Jochelson, we are dealing here with PY *y which changed into a in Kolyma Yukaghir. Some Tundra Yukaghir words have a back e which can also go back to *y, e.g. Τ eyaba 'waist'. Also observe the following correspondences: Κ jayil ~ Τ eyil, ayil 'edge' and Κ qaqss-, kiqss- 'to choke' ~ Τ qeqsa- 'to gurgle'. Finally, Tungus back stems with i are usually borrowed in Yukaghir as back, e.g. Τ tilba:- 'to trample down' < TU *tilpa- 'to smear, to knead, to squash' and Τ sirba:- 'to squeeze' < TU *sir- 'to press'.



4.1.4. Long


In the dictionary the long vowels are reconstructed in monosyllabic stems and in (C)V:Co stems, although in both these types they were probably secondary and caused by prosodic reasons (see 3.4.1 and 4.3). Since quantitative alternations of vowels are basically identical in Tundra and Kolyma Yukaghir (see 3.4.1 and 3.4.2), they are reconstructed for Proto-Yukaghir. Most of the original *(C)VCC9 stems where the medial cluster was simplified into a single consonant (4.2.5) do not contain a long vowel in modern Yukaghir. So they look just like (C)VCV stems in which the medial consonant does not go back to a cluster and the vowel of the first syllable is short. Examples of such stems follow. PY *somp9l'3*qompa *nonts "Toqqs*noqqa

Κ sobol'aqobo nodo joyodajηογο

Τ sabal'aqobo nada jogotajηογο

to lose bottom bird; animal to open sand; ashes

In a few stems where the second consonant goes back to a cluster the first syllable is long, i.e. they pattern together with (C)V:Cs stems, e.g. Τ li:daj' t o c r u m b l e ' (< *li:nts-),

Τ qa.dah

' a r m p i t ' ( < *qantah),

Τ ki:d'd ' d e m o n ' (

< *kinc'a), and Κ o:zd- 'to drink' ( < *onc3~). I do not have an explanation for why they behave differently from the stems listed in (9). Some *(C)VCCa stems show variations in length, e.g. Τ cid'e ~ ci:d'a 'bottom part of a woman's coat trimmed with dog fur and tassels'. Many long vowels in Kolyma Yukaghir have appeared as a result of contraction of a short (typically stressed) vowel and the following spirant or glide, namely *j, *w or *y. The contraction normally took place if the spirant or glide was syllable-final, as illustrated by the following examples. (10)

*aw > o: *aj > a: *ow > o: *uw > u: *uy > u: *ej > i: *9j > i: *9w > u:

Κ aro:jo ~ KJ aravje Κ cajsils, ca:sib ~ KJ caisile Κ jo:y9, jouya ~ KJ joboye Κ jo: - K J j o b o Κ hu: ~ Β newe, ME niiv, S niv Kju:lugul ~ S juglugul Κ ni:n-qa:r ~ S nejin ( > nejn)-kar Κ -ji:li ~ KJ -jeili Κ arnu:ja: ~ MK arimnewäja, S arinuaja

lenok bullfinch back disease name shoulder cloud INTR.1PL wolverine

Introduction *öy>ö: *ΐγ>ί:

Kjö:l'9tl'3- ~ Β joglerim Κ i:da- ~ S igda


to love to catch

Observe also the following correspondences: Κ i:cd ~ Τ oje? 'point, edge' (< *öj-) and Κ je:d'a ~ Τ jeijd'a 'excrement' {rj has contracted in K). In Tundra Yukaghir contraction is less frequent, but also takes place, e.g. Τ i:sd- 'suck' < iwsd-. The contraction is still a live process, as shown by the following free variations attested in modern Kolyma Yukaghir: (11)

ujni: ~ u:ni: pe:d'a ~ pejd'a kerpi: ~ kerpsjs ( Κ Τ ρ, PY *t > Κ Τ ΡΥ *k/q > Κ Τ k/q, ΡΥ *m > Κ Τ m, ΡΥ *π > Κ Τ n, ΡΥ > Κ Τ η, ΡΥ */ > Κ Τ /, ΡΥ *Γ> Κ Τ l\ and PY *r > Κ Τ r. Examples can be easily found in the dictionary. Below only non-trivial correspondences are addressed.

4.2.2. Sibilants and affricates The opposition of the palatal and non-palatal affricates *c and *c' is only reconstructed for the position after a sonorant (4.2.4). Even if it was originally present in other positions, reflexes do not differ in the modern languages: there is only one non-voiced affricate c (optionally s in Kolyma Yukaghir, see 3.1.3). Jochelson sometimes records a palatal, but it rather corresponds to Κ s, i.e. a positional variant of the non-palatal affricate. So there are no reasons to assume two affricates other than after a sonorant, although it is not excluded either that their reflexes have merged elsewhere. In Nikolaeva (1988a, 1988b) I reconstruct the palatalization opposition for the sibilants *s and s. Under this analysis, the palatal *s is the source of the regular correspondence Κ s ~ Τ s, e.g. Κ sög- ~ Τ seg- 'to enter' and Κ sure 'body' ~ Τ sure 'skin (of a hand)'. In Tundra Yukaghir *s lost palatalization, while in Kolyma it became s, possibly through the intermediate stage of *£' or * 0 \ Note that languages spoken in the Yukaghir area, such as Yakut, Even, Evenki, and the old Russian dialect of Kolyma (Bogoraz 1901), do not have palatalization opposition for sibilants. The only sibilant normally has the interdental pronunciation. This could be also true for Κ s and Τ λ\ They are both recorded as c in Jochelson's manuscript dictionaries



(KD and TD), which suggests that the pronunciational difference between them at this stage was insignificant. I further argued that *s changed into I or 0 , possibly through the stage of *(). This reconstruction is based on external comparison. In a number of words the Uralic or Altaic *s corresponds to / or 0 in Yukaghir. (13)a. U *seykV/*säijkV 'upper arm; forearm' ~ Τ egi(:)l 'back of the head' FU *särV 'root; vein' ~ Κ larq- 'root' FU *sewe-/*seye- ~ K T leg- 'to eat' U *sala ~ Κ olo- 'to steal' b. TU *saija(r) ~ Κ irjer 'pit, hole' PA *sbjri ~ Κ ujd'i: 'sinew' There are two instances where MC shows s in the place of j or I in the other Yukaghir varieties: MC sogote ~ Κ Τ joyoti: 'arrow' and MC sogondonde ~ Β logundintshina 'salt'. These words are likely be the result of an erroneous recording, as suggested in Tailleur (1962: 91, 93). So the following processes can be assumed for the history of Yukaghir. (14)

a. *s > θ > 1/0 b. *s > (s/θ') > Κ s, Τ s

Similar changes are reconstructed for the Proto-Uralic sibilants in Ugric and Samoyed languages (e.g. Xelimskij 1982). However, they are hypothetical in Yukaghir, as (14) crucially depends on external correspondences. Therefore in this dictionary I only use one sibilant and write it as *s, assuming that it developed into Κ s and Τ s. In other words, in the convention used in this dictionary, *s corresponds to the presumed *s in (14), while for the words in (11) */ or 0 are reconstructed. This system may in principle reflect a more recent stage of Proto-Yukaghir. It is worth noticing that the Old Yukaghir sources hardly differentiate between affricates, on the one hand, and sibilants, on the other, cf. Β tshall, ME tschal, MU tschal, MK tscheel ~ Κ sa:l, Τ sa:l 'tree' and Β tshul, ME tschul, MU tschul, MK tschul ~ Κ Τ cu:l 'meat'. Therefore Collinder (1940: 168) and Jochelson (1898: 154) suggested that sibilants and affricates had the same source. Yet, their reflexes in the modern languages clearly differ. In addition, the comparison with Tungus shows that affricates and sibilants were originally distinct, cf.:




a. TU *cab- 'noise, noisy' ~ Κ cabo:- 'loud' TU *cöle- 'to jump on one foot' ~ Τ cule- id. b. TU *saja 'interval (between fingers)' ~ Κ saj-, Τ saj- 'across' TU *seri 'decking, covering' ~ Κ seril, Τ siril 'bottom edge of the chamois that covers a yurt'

I believe the reason why affricates and sibilants were often conflated in earlier sources lies in the difference between male and female pronunciations. According to Jochelson (1898: 153-154), in earlier Kolyma Yukaghir men pronounced both c and s as c, while in the female pronunciation affricates and sibilants were always distinguished. It is reasonable to assume that most Old Yukaghir sources reflect the male pronunciation. In fact, there is documentary evidence for that: Billings (1811: 25) refers to his consultant as an 'old man', and so does Matjuskin. For this reason most Old Yukaghir sources did not distinguish between s and c. The difference between the male and the female pronunciation was later lost and the female norm was generalized.

4.2.3. Fricatives The reconstruction of the Proto-Yukaghir interdental *δ is based on the correspondence Τ -r ~ Κ -d-, -t. The word-final t in Kolyma Yukaghir probably appeared due to the devoicing of d in the word-final position, see 3.2 on the phonotactic restrictions. The PY *δ only occurred intervocalically or wordfinally. (16)

a. Κ j o : d d - , Β lioda, ME liota ~ Τ l'o:ra- 'to play' Κ mödi:-, MC mody, KL moriza, mozyi, Β moedik, mo.e.dik, ME moktyk, moitik ~ Τ möri:- 'to hear' Κ p u d e , MU buden 'up', Βpudanmoi, M E p u d a n g m a , MCpadanmei 'tall' ~ Τ pure 'up' Κ köde.l, B M E M K kodel ~ Τ köre:I 'wolf b. Κ p u l u t ~ Τ p ö l u r 'old man' Κ unmut, ME onmut ~ Τ önmur 'horn, antler'

Most Old Yukaghir sources record d/t in the place of PY *δ. The only Old Yukaghir source that renders *δ as r, as in Tundra Yukaghir, is MO, cf. the following words:

Introduction (17)


Κ meint, B / M E melud, M K melut, M C malyt ~ M O melur 'breast' Κ abut, Β crwoot, M E auut, M K -awut ~ Τ awur 'nest, container', M O

abor 'brain' Κ qodol'-

M O χoral' 'boat'

This shows that different reflexes of *δ (d/t vs. r) were present even before the Old Yukaghir period. When a Tundra Yukaghir word has r in the absence of a Kolyma counterpart, strictly speaking it is impossible to tell whether *r or *δ is to be reconstructed. However, I do not provide alternative reconstructions in this case and always write *r. The reconstruction of the PY * /'). However, this possibility is hypothetical and is not adopted in the dictionary. In the modern languages γ only occurs as the back counterpart of g. Synharmonic variations of velar and uvular consonants can in principle go back to the Proto-Yukaghir stage. But voiced stops were altogether absent in ProtoYukaghir. So the pair g/γ is likely to go back to the PY *γ. In the modern languages this phoneme is represented by a number of alternations described in 3.4.2: as g/γ before a vowel; k/q before a voiceless obstruent; u ~ w before a voiced obstruent, / or a pause; η before η\ and m before m. So in some cases the PY *y could become a glide component of preconsonantal diphthongs, as demonstrated by the following word: Κ ejrs-, Τ ewrd— KL agra, S agreje < *ey-rd- 'to walk'. The Old Yukaghir sources show g or h in the place of *y, e.g. Β -igia, M E -ikeie, M K -igee ~ Κ igejs, Τ igijd 'rope' and Β ighil, M E jehyll,

MC ege- ~ Κ jayil, Τ iyil 'edge', whereas Jochelson sometimes records r, e.g. KJ joroto-

~ joyoto-

'to w o u n d ' .

4.2.4. Sonor ants and glides

The bilabial approximant *w is preserved in Tundra Yukaghir in all positions except for the initial position before /'. There are no words that begin in *wi-. In Kolyma Yukaghir the initial *w- was lost, cf. the following correspondences:




Κ a:daonor e:n

Τ wa: rewan ar


to pull tongue another

The only Old Yukaghir idiom that preserves the word-initial *w- as in Tundra Yukaghir is MO, e.g. MO wogo ~ Τ waya 'face', MO wener ~ Τ wanar 'tongue', and MO woldik 'to laugh' ~ MC ol'cin 'joy'. The reconstruction of the initial *w- is supported by external parallels such as Κ ozu: 'thin root used as a thread for fastening boats', Τ waruluu 'root' ~ U *wacV/*wancV 'root' and Τ wentd- 'to stretch out, to expand' ~ FU *wene- 'to stretch out, to expand'. When a Kolyma Yukaghir word begins with a vowel, except /, and there is no corresponding Tundra word, I provide alternative reconstructions, with and without the initial *w-. The word-medial *-w- was preserved in Tundra Yukaghir and changed into b in Kolyma Yukaghir, e.g. Κ abut ~ Τ awur 'next, container' and Κ ibi- ~ Τ iwi- 'to suck'. The word-medial w is normally written as w or u in Old Yukaghir, e.g. Κ abut ~ KL awim-, MK -awut, Β awoot, ME auut 'nest, container' and Κ lebe: ~ BO /ewe-, KL /ewe-, MK lewe, W levia-, Β levyβ, ME leviya 'earth'. This shows that the change w > b in Kolyma Yukaghir is a fairly recent process. The word-medial *-«- is reflected as ή in both Kolyma and Tundra Yukaghir. However, in the other positions it can undergo various changes. For example, in some words the initial *Λ- is reflected as j in Kolyma Yukaghir, cf. Kjomil~T namil 'neck'. The same concerns the initial */- which is sometimes reflected as j in the modern languages, but the Old Yukaghir data clearly show that we are dealing with */-. Observe the following words: MC Ijangjaja, Β landzsha, ME landscha, MU Ijä'ngdscha, Ijä'ngdschand-, MK Ijdangdscha, lljdngdscha ~ Κ jayzd, Τ jayds,jayr3 'goose' and Β lieril, ME lyril, MU lirir\ MK liril, liril— Κ iril, Τ l'iril 'stomach, belly'. As was mentioned in 3.2, there is no final ή in Kolyma Yukaghir. In particular, it exhibits j instead of the Τ ή in the 3rd person intransitive agreement marker of some verbs, normally those whose stems ends in ο: (Τ ο: ~ a:). Most Old Yukaghir idioms demonstrate ή (or η as its transcriptional variant) in the place of K j , e.g. (19)

Κ ataqlo:-j ~ KL adaklon, MK αηάάΜοή, MU ändaklon 'two' Κ juko:-j ~ MK ljukon, BO juxcon 'small' Κ como:-j ~ KL como-ή, MK tscho/ηάή, MU tschomon 'big' Κ ja:lo:-j ~ MU jälon, MC jalon 'three'



B/ME shows j in some words, like the modern Kolyma Yukaghir, e.g. Β tchomoi 'big', and η {ή) in other words, e.g. Β inglon, ME inglong 'thick' (K iijlo:-j), Β antachlon, ME antaglon 'two', and Β *ya:lon, ME jalon 'three'. As the idiom of B/ME can be considered a more or less direct ancestor of modern Kolyma Yukaghir (2.1.5) the change *-ή > j is likely to be a recent Kolyma Yukaghir innovation. 4.2.5. Consonantal clusters Proto-Yukaghir exhibited word-internal consonantal clusters of the type "sonorant + voiceless obstruent", which have undergone various changes in the modern language. Potential clusters of this structure are shown in (20), although not necessarily all of them were present in practice. *mt *mp *mk/q *mc *mc'

*nt *np *nk/q v Λ *nc *nc'

*iit *ήρ *iik/q *nc *nc'

*r)t *qp *qk/q *qc *r)c'

*lt *lp *lk/q *lc *lc'

*l't *l'p *11c/q *l'c *l'c'

Such clusters are impossible in the modern languages inside a morpheme (see 3.2), however the clusters "sonorant + voiced obstruent" are fairly frequent. I take this to indicate that in such clusters the obstruent was voiced. This stage is attested in Old Yukaghir, as shown below. (21)

a. PY *-mc- > ΟΥ -Π13MC memze- ~ Κ memzd-, Τ mömdd- 'to burn' b. PY *mt-> OY-mdB omda-, ME omtu- ~ Κ omdu, Τ amdu- 'to hurry' c. PY *-np- > OY -nbB tonbo-, ME tonbe- ~ Κ tönba- 'strong' d. PY *-rjc'-> OY -rjd'B indshi ~ Κ iijd'i: 'sinew' e. PY *-ήρ- > OY -nbB xänbo, MC qatöa, BO qänbo ~ Κ qaüba 'palm'


Introduction f. PY *-lk-> OY-lgB oolgala ~ Κ ulguh 'urine' g. PY *-qc- > OY -q3B landzsha, ME langdscha, MU Ijangdscha, MC Ijangjaja ~ Κ jaijzd, Τ]αηά3,]αψ3 'goose' h. PY *-qt-> OY-qdMC landygik, Β longdok, ME longtok ~ Κ lotjdd- 'to dance' i. PY *-lc- > OY *-l3MCpol'woroie 'green', Β *poldshitsha,podrie,poldshasha-, poltschitscha,podrie ~ Κpolzicd 'leaf


Neither the Old-Yukaghir sources nor Jochelson systematically record the palatalization opposition for affricates. In other words, both *d' and what I reconstruct as * j (voiced non-palatal affricate) were often transcribed in the same way. Jochelson also used the same symbol for d' and what corresponds to Κ ζ, although older sources sometimes record d or I in the place of * j This shows that the phonetic difference between palatal and non-palatal affricates was minimal until recent times. Still, their reflexes are clearly opposed in modern languages: whereas the PY *c' after a sonorant is represented as d\ the PY *c (> OY j ) has changed into Κ ζ and Τ r ~ d. The situation is more complicated in clusters that begin with r. In most words in the modern languages the following obstruent is voiceless. Moreover, some loanwords show devoicing after r, e.g. Κ mo.rta 'fish trap' < Rus. dial. morda and Τ lerpu: 'shaggy dog' < TU *ler(be)- 'shaggy'. However, there are a number of words with a voiced obstruent after r. (22)

Κ jurgu: moryulTun nörgumo

slot minnow notch

Τ serbad-amun purga se:rdi:d-ile

shoulder bone knot reindeer rejected from slaughter

This shows that r has a dual status: sometimes it behaves like a sononant and sometimes like a voiceless stop. But since the clusters "r + voiceless obstruent" are more frequent, I reconstruct them for the Proto-Yukaghir stage, unless there is positive evidence against such a solution.



In the examples in (21) the consonants do not have a homorganic articulation. When they do, further changes were observed in the clusters with nasals. The clusters changed into a single consonant: a nasal before a consonant or a pause, or a voiced obstruent in the prevocalic position. The resulting alternations are described above in 3.5.1. This diachronic process of the simplification of clusters took place in Old Yukaghir.1 It may have occurred at varying speeds in the different Old Yukaghir idioms, but in general was completed by the mid 19th century, i.e. at the end of the Old Yukaghir period. This is evident because all the later sources, beginning with Schiefner (1859), record single consonants in the place of the Old Yukaghir homorganic clusters. Examples of homorganic clusters with nasals follow. (23)

a. PY * n t > O Y n d > n ( + C ) ~ d ( + V ) Β ME tandi- ~ Κ Τ tadi:- 'to give' MC lundylo, Β -lundal, ME -london, -lundol, MK lundul, Κ ludiil 'iron', lun-bugö, 'cauldron' (lit. 'iron box')


b. PY *nc'/*nc' > OY nd'/hd' > ή (+C) ~ d' (+V) KL kinica, kinind'a, Β kininsha, ME kinindscha, MU kininsche, MK kinedsche ~ Κ Τ kind's 'month, moon' MC -konty, Β -kondzsha, ME -kontscha ~ Κ Τ köd'e 'worm, larva' MU endschit, MK endschi, ME endschi ~ K T ed'i 'alive', Κ enna'alive, vivid' c. PY *nc > OY n3 > η (+C) ~ Κ ζ, Τ r (+V) MC andrele, andzhe, MU andschub ~ Κ cäu: 'word, language' Β kundshu, ME kuntschu, MC kuntu-, kunzup ~ Κ kuztr., Τ kurul 'sky' Β onzshi, ME ondschy, MU onsche, MC inzha, MO onde ~ Κ o:zi: 'water' Β pandalitsh 'to scratch', ME pantalytsch 'to scratch' ~ Κ pazili:Tparali:- 'to tickle' d. PY *qk/*qq > OY rjg/ηγ > rj (+C) ~ g/γ (+V) Β angitak, ME angietak~KT ayits- 'to hide' Β iongul, ME jonkool, MU jongjnogha, BO jungol, MO niongol' ~ Κ Τ joyul 'nose' Β ongai, ME ongen, MK ongoi ~ Κ οη-, Τ οηιι- 'damp, wet', Κ oyunbs- 'to get wet'


Introduction e. PY *mp > OY mb > m (+C) ~ b (+V) MC iumbo, MU jämboon ~ Τ jaba- 'to d i e \ j a m d ' i - 'ill', Κ jou < KJ jobo 'disease' MC jembend- ~ Κ j o b o y s > jouya 'back' Β M E nunbur ~ Κ nabur 'foam'

Thus, there are two sources of voiced obstruents in modern languages: voiced fricatives and glides (4.2.3 and 4.2.4) and homorganic clusters. When the source of a voiced obstruent is unclear, I provide alternative reconstructions throughout the dictionary. A number of words demonstrate irregular clusters that violate the phonotactic constraints. This may indicate either that the word is a recent borrowing or that there used to be a morpheme boundary between the components of the cluster. This is observed, first, if a voiceless obstruent follows a sonorant, e.g. Τ samqdj 'tea pot', Τ ilka 'cross-bar in a sledge', Κ Τ monqd 'hill', and Κ qantu: 'crop'. Second, some words exhibit homorganic clusters with nasals which should have changed into a single consonant, e.g. Τ qandaba 'inside the lower part of fur trousers', Τ timbdCdn- 'swollen from absorbed water', and Τ lirjgacz 'scum'. In clusters "sonorant + d'" the affricate has two sources, *c' and */. The latter is due to the assimilative process j > d' after a sonorant described in 3.5.4. I therefore provide alternative reconstructions for such stems, e.g. PY *semc'-/*semj- > Τ semd'i- 'to sneak up' and PY *culc 's-/*culj3- > Τ culd'dya- 'to twitter, to chirp'. If a modern Yukaghir language exhibits nd' I reconstruct *nj, e.g. PY *penjija: > Τ pend'ijs 'stream'. Such clusters cannot go back to *nc' because *nc' has developed into η ~ d\ Other changes in clusters are the same as for single consonants, i.e. PY *w > Κ b and PY (*v >) > Τ s, Κ ί .

4.3. The structure of the stems 4.3.1. Monosyllabic


As mentioned in 3.4.1, in modern Yukaghir the words are minimally bimoraic and the word-final coda consonant does not project a mora. This was also typical of Proto-Yukaghir. It had two types of non-derived monosyllabic nominal stems both of which contained a long vowel: (C)V:C and (C)V:, while the structure *(C)V(C) was impossible. Examples are *ö:, *jo: and *qa:r. There are no restrictions on vowel length for monosyllabic nominal stems that were always accompanied by a derivational affix because the affix provided an



additional mora. Such stems could contain a short vowel and have the structure (C)VC- or (C)VCC-, e.g. *nol- and *marq-. Similarly, monosyllabic verbs could have a short vowel because they do not normally occur in a bare form, e.g. *ιηοη-, *1βγ-, *soijq-,

4.3.2. Bisyllabic



In both modern Yukaghir languages the native stems *(C)V:CV and *(C)VCs are ruled out, but the types (C)V:Ca and (C)VCV are frequent. They exhibit different stress patterns: in the stems CVCV the final syllable is stressed, whereas the stems (C)V:Co bear the stress on the first syllable, as follows from the usual rules of stress assignment. As was discussed in 3.4.1, the structure (C)V:Ca may alternate with CVCV if it is followed by stress-attracting syllables. This situation is synchronically analyzed as alternations in vowel length, depending on the position of the stress. From a historical viewpoint, two analyses of this situation are possible. Under the first, Proto-Yukaghir had two types of stems, which contained short vowels in the first syllable but differed in the original position of the stress: CVCV' and C V C V . At some stage the first stressed vowel in the latter type was lengthened and the second unstressed vowel was reduced. Reduction does not take place if the stress is located further in the word. Under this analysis the difference between the two types originates in the independent lexical specification of short vowels as either stressed or unstressed. Under the second analysis, the situation is basically the same as in modern Yukaghir. That is, the distinction in vowel length existed in Proto-Yukaghir and determined the position of stress. Although there is no etymological evidence that helps us to choose between the two analyses, I adopt here the second solution, for the following reason. The type (C)V:Co patterns together with another bisyllabic type CVCCa in two respects. In both cases the stress falls on the first syllable and both types have a in the second syllable, while the full short vowel is impossible (i.e. *(C)VCCV). Examples of the CVCCa type are *jaqta, *marqa/*morq3 and *tönpd-. As was argued in 4.2.5, some consonantal clusters have undergone simplification. In particular, this applied to the stems (C)VCCa, which originally had a word-internal cluster. Such stems had the stress on the first syllable. If we assume that the position of stress determined the vowel length, we would expect the following development: (C)V'CCa > (C)V'Co (after the simplification of the cluster) > CV:'Co (due to the lengthening of the stressed vowel). However, as was shown in 4.1.4, in most cases the resulting stems belong to the (C)VCV' type. This rather implies the following development:



(CyV'CCs > (C)V'Ca > CVCV'. In other words, the stress did not cause lengthening of the first vowel. Instead, after the simplification of the cluster it shifted to the last syllable following the regular rules of stress assignment, which caused a change of the final vowel. Thus, no lexically specified stress distinctions are assumed for Proto-Yukaghir. The two types of stems differed in the length of vowels, as, for example, is reflected in the following reconstructions: *no:ja, *ca:qa and *o:rs- v.v. *jara-, *pude and *sölö. As in modern Yukaghir, the position of the stress was fully determined by the vowel length. Other types of bisyllabic stems are usually formed by means of adding an extra consonant or consonantal cluster to the three basic bisyllabic types (C)VCV, (C)V:Ca and ( Q V C C s . In other words, they could have the structure (C)V:CsC(C)-, (C)VCVC(C)- or (C)VCC3C(C)-, e.g. *ant9q-/*wantdq-, *il'imp- and *kimer. There are no non-derived bisyllabic stems *(C)V:C(C)V:, which would only contain high vowels, and no stems ending in a short high vowel. Long high vowels occur in the second syllable of bisyllabic stems in the structures (C)VC(C)V: and (C)V:C(C)V:, but are likely to be secondary. The long i: and u: in this position probably result from the contraction of a short vowel with the word-final glides *j and *w, respectively. For instance, Κ o:zi: 'water' may go back to a derivative of the verb 'to drink' *o:zg-j(d), and so on. Another very distinct, although infrequent, type of bisyllabic stems is CV:CV: with identical non-high long vowels in both syllables. Such stems could have originated through reduplication of a monosyllabic stem CV:. The reduplication is limited to a few expressive words (ideophons and nursery or taboo words). It is apparent in Κ pa.ba: 'elder sister', which is probably related to pa: 'woman' (with further voicing of the middle consonant), and Κ qa.qa: 'bear, grandfather' (with contraction), cf. Τ qajce: id. Other words that belong to this group are cited below. (24)

Κ me:me: bear ca:ca: + elder brother, uncle tu:tu: trumpet made of birch bark

Τ ma:ma:- to eat (of a child) pa:pa:- to urinate qa:qa: (child's) excrement

Note that non-derived stems with a long vowel in a closed syllable are infrequent. Normally they are either recent borrowings or are morphologically complex. For instance, the type (C)V:CCa- may go back to a non-derived stem (C)V:C augmented by a derivational consonant-initial affix -Ca-. But the strings Vj and Vw can precede a consonant because they have the status of diphthongs (see 4.2.1).



The reconstruction of stem-final short vowels in verbal stems is a complicated matter. As was shown in 3.6.3, the stem-final vowel may fall out before some suffixes and is replaced by an epenthetic vowel. In this dictionary the final vowel of the stem (a, o, e, ö or a, subject to the usual distribution) is reconstructed if it is attested in at least one derivative of the relevant stem. For example, I reconstruct *aija because it is the source of the Κ Τ αηα 'mouth', among other words. On the other hand, if a stem-final vowel is not overtly present in any of the attested forms, the stem is reconstructed as consonant-final. This decision is based on mere convention, since even in the latter case there is a theoretical possibility that the stem could end in a vowel that was syncopated in all attested forms. I conventionally reconstruct a consonant-final stem if in all descendant forms it is followed by a vowel-initial affix that causes vowel deletion (see 3.4.2 for the list of such affixes). For instance, the source of the Τ eri:- 'to answer a call', where the final -/.·- is probably a Transitive/Causative verbal affix, is reconstructed as *er-. However, in principle this stem could look like *ere-, the final vowel having been lost before the vocalic affix. The stem *a:l- is represented by the following reflexes: Τ a:li:l, TD alii, 'order', Τ a:le:ca 'person who gives orders', TK a:liece 'order', and Τ a:le:-, TK alie-, a:lie-, a:li- 'to make an order'. It is reconstructed as *a:l- but it could look like *a:h-, while the vowel d may be lacking from attested forms because of the following vowel.

4.3.3. Trisyllabic


Trisyllabic stems are usually formed by adding a syllable Ca, CV: or Ci/uC to bisyllabic stems, e.g. *üuyör3l*mqkar3, *qantdh, and *solqarqa:. At least in some cases this third syllable may be an ancient suffix, but in other instances it could be part of the stem. In the ( Q V C V C s stems the vowel of the second syllable often falls out, e.g. Κ munzv., Τ mundi: ~ Β moonendshi, KJ muned'idie, M E munentschy, MK mumndschi 'lower j a w ' . For this word I reconstruct *munenci: with the syncope of the vowel e in the second syllable. Such syncope is especially typical of Kolyma Yukaghir and is evident if we compare modern Kolyma Yukaghir with older materials. (25)


kind's ~ Β kininsha, ME kinindscha 'moon; month' anfijd'a ~ Β alnindsha, ME alnintscha 'princeling, head, chieftain' irde: ~ KD iridic 'new-born reindeer calf qajn, qanin ~ KJ qanin 'when'



Κ kund'dd- ~ KJ kuned'i- 'to become sticky' Κ -dejhl-deuh

~ KJ -degele ( P O S - A C C )

Some words show alternations in the modern language, e.g. Κ erd'e:- ~ erid'e:'to w i s h ' , Κ murud-u:

~ murd-u:

' f u r stocking', and Κ iqil'o:—


'squint'. Observe also that the consonant preceding the syncopated vowel may change into a glide and further contract with the vowel of the first syllable, as in the following variants: Κ majfe ~ Τ mcmah 'hair', Κ ta:yih ~ ta:jls 'then, so', Κ so:ma ~ soums

~ soroms

' m a n ' , and Κ to:ks ~ toukd (