vajda-2008.pdf

Dene-Yeniseic Symposium, Fairbanks; Ed Vajda Draft March 3.08 Comments welcome (eddievajda@yahoo.com ) 1

A Siberian link with Na-Dene languages 1

Edward Vajda (Western Washington University)

1. Introduction

The Yeniseic (Yeniseian) microfamily of central Siberia (upper and middle reaches

of the Yenisei River basin) is genetically unrelated to other families of the Old World.

Yeniseic includes the extinct Kott, Assan, Arin, Pumpokol, and Yugh languages, as well

as the highly endangered Ket, now with fewer than 200 speakers, most over the age of

50. Only Ket and Yugh were documented in modern times. Fortunately, excellent

materials were collected from the last Kott speakers by Finnish linguist M. A. Castrén

(1858). Assan, Arin and Pumpokol disappeared before 1800 and were only sparsely

documented by travelers and explorers. Nevertheless, the extant documentation offers

valuable lexical comparanda that testify eloquently to the value of documenting even the

most obscure of the world's endangered languages before they disappear. Werner (2005)

contains a complete description of all 18 th century documentation of Yeniseic languages.

Monograph-length descriptions of Modern Ket phonology and grammar include Werner

(1997), Vajda (2004), and Georg (2007).

The prefixing verb structure of Ket differs strikingly from the surrounding Uralic,

Turkic, Mongolic, and Tungusic languages of Inner Asia and Siberia. During the past two

centuries, linguists have attempted to link Yeniseic to other Northern Hemisphere

families with a prefixing verb, notably Burushaski, Abkhaz-Adygh (Northwest

Caucasian), Nakh-Dagestanian (Northeast Caucasian), Sumerian, and Na-Dene

(Athabaskan-Eyak-Tlingit). It is no exaggeration to say that the position of Ket in Inner

Eurasia has up until now remained as enigmatic as that of Basque in Europe, Zuni is the

American Southwest, or Burushaski in South Asia.

The full history of published and unpublished speculation on the external relations of

Yeniseic can be found in Vajda (2001). The earliest suggestion that Ket has a special

historical connection with Native American languages dates back to Adriaan Reeland in

1708 (cf. Vajda 2001:2). The first person to claim a genetic link specifically between

Yeniseic and Athabaskan-Tlingit (Eyak was then unrecognized as a Na-Dene language)

was the Italian linguist Alfredo Trombetti (1923). Since that time, many other linguists,

notably Merritt Ruhlen (1998) have repeated the same suggestion, though typically

including Haida in Na-Dene). No one has produced anything to support this claim beyond

random look-alike words or general typological resemblances. The typical inclusion of

1 I thank Jim Kari, Mike Krauss, and Jeff Leer for valuable comments on earlier drafts. More generally, I

owe a deep debt of gratitude to my Athabaskanist colleagues, without whose decades of contribution to Na-

Dene historical linguistics it would hardly be possible to undertake the comparisons I am attempting here. I

own a similar debt to Siberian Ketologists, particularly the indefatigable Heinrich Werner, for their crucial

documentation of Yeniseic dialects and languages. Werner's work on the now extinct Yugh is similar in

importance to that of Michael Krauss on Eyak, Jeff Leer on Tlingit, or Jim Kari on Ahtna. I deeply thank

Bernard Comrie, director of the Linguistics Divison of the Max Planck Institute for Evolutionary

Anthropology, Leipzig, for his extensive support and particularly for hosting the Na-Dene Workshop in

August, 2006, where some of these ideas were first publicly presented; and I thank our colleagues at

Leipzig, particularly Juliette Blevins and Don Stilo, as well as Eric Hamp, for their enthusiasm and

encouragement in support of my work. I also thank all of my colleagues from Tomsk Pedagogical

University, where I began my serious work on Ket. Finally, I especially thank Jim Kari and his colleagues

at the University of Alaska, Fairbanks, for organizing the Dene-Yeniseic Symposium (Feb. 2008), for

which this article was prepared. See also the postscript at the end of this article.

Dene-Yeniseic Symposium, Fairbanks; Ed Vajda Draft March 3.08 Comments welcome (eddievajda@yahoo.com ) 2

Haida in such linkages is indicative of their extremely speculative nature, as there is no

demonstrable genetic relationship between Haida and Na-Dene. But if speculation counts

as discovery, then the beginnings of "Dene-Yeniseic" linguistics belong to Trombetti,

whose initial conclusion has proven to be completely correct 2 .

Random similarities in basic vocabulary are insufficient to demonstrate language

relatedness. A list of look-alike words can be compiled, even using basic vocabulary,

between any human languages. Nor are typological similarities, even involving relatively

uncommon traits such as a rigid prefixing verb structure, a reliable diagnostic for genetic

relatedness in the absence of a system of cognate morphology. The only accepted way of

demonstrating the existence of a language family is to identify a sufficient number of

cognates in basic vocabulary to establish interlocking sound correspondences that are

reflected in the language's grammatical systems, as well; cf. Campbell (1997) for a

thorough, state-of-the-art treatment on the issue of demonstrating genetic relatedness. All

accepted language families share this combination of homologies to an extent that

permits at least partial phonological and morphological reconstruction of an ancestral

proto-language. Though generally not stressed by historical linguists, true evidence of

genetic relationship also provides, by default, external comparative data useful for tracing

the internal historical development of each member language or group of languages.

Word lists or typological comparisons cannot be used in this way. A linguistic debate

about Haida's membership in Na-Dene is no more useful to veterinary science than a

debate about whether unicorns exist. Haida comparisons have failed to shed any light

whatsoever on the historical development of Athabaskan-Eyak-Tlingit, outside the realm

of contact phenomena. The same could be said of the still undemonstrated Altaic

Hypothesis, which is useless for understanding the internal structure of Modern Halh

Mongolian. A Slavic linguist who refuses to accept Indo-European, on the other hand,

would be more like a traveler who denies the existence of the automobile. Many facets of

Slavic linguistic prehistory simply cannot be fully appraised without acknowledging the

demonstrable relationship of Slavic to Baltic, Latin, Iranic, and its other Indo-European

relatives. The unavoidable usefulness of a proven genetic connection between languages

is the best confirmation of its validity.

If Yeniseic is demonstrably relatable to Na-Dene, the evidence should be able to

help solve Na-Dene internal problems by providing hitherto unknown external

2 In this paper I have endeavored to credit other scholars for contributions to what could be called "Dene-

Yeniseic" studies. Although I reject conclusions of genetic relationship made solely on the basis of look-

alike vocabulary or typological resemblances, four earlier publications should be mentioned here. Alfredo

Trombetti (1923), the first to suggest the connection, may also have proposed the first Dene-Yeniseic

cognate: Ket de'N and Athabaskan dine 'people' (though cf 3.3.4). Merritt Ruhlen's (1998) proposed

cognate sets contain several genuine cognates, among over 75% coincidental look-alikes. These are

Ruhlen's comparisons for: head, stone, foot, breast, shoulder/arm, birch/birchbark, old, and burn/cook, and

possibly a few others. The correct identification of cognate words for "birch/birchbark" is particularly

noteworthy, as this basic vocabulary item is specific to families of the northern latitudes. The finding of

these cognates, though it was impossible to confirm them as such in the absence of much more

investigation, represents an important contribution, in retrospect. Also important here is Johanna Nichols'

(1992) Linguistic diversity in space and time , which offered an innovative typological/geographical

perspective on where to look for possible genetic links. In the same vain, Michael Fortescue's (1998)

pioneering book Language relations across Bering Strait is significant for offering a preliminary inventory

of unusual morphological features shared by Yeniseic and Na-Dene, but not by other Northern Hemisphere

languages.

Dene-Yeniseic Symposium, Fairbanks; Ed Vajda Draft March 3.08 Comments welcome (eddievajda@yahoo.com ) 3

comparative data. Similarly, Na-Dene comparanda should provide answers to questions

in the development of Yeniseic languages that defy resolution based on Yeniseic-internal

data alone. One of the main aims of this article is to demonstrate that this is indeed the

case - first by examining the prefixing verb morphology of both families, then moving on

to explain systematic sound correspondences based on the cognate vocabulary.

The Yeniseic verb complex shows a striking system of morphological homologies

with the oldest layer of Athabaskan-Eyak-Tlingit verb prefixes. Section 2 begins by

demonstrating that Yeniseic verb morphology does not resemble other Old World

prefixing languages. Section 2.2 provides an overview of specific morphological

homologies between Yeniseic and Athabaskan, Eyak, and Tlingit. Successive

subsections describe homologies in tense/mood/aspect affixes (2.2.1), spatial prefixes and

incorporated body part nouns (2.2.2), pronominal elements (2.2.3) and in the pre-root

"classifier" prefixes (2.2.4). Section 3 discusses cognates in basic vocabulary and the

system of sound correspondences they reveal. This section begins with patterns of coda

reduction (3.1) and the emergence of phonemic tone in three types of Yeniseic syllables

(3.2 – 3.5), then turns to onset correspondences (3.6). Section 4 briefly summarizes the

evidence that Yeniseic and Na-Dene share a common linguistic origin. The value of the

comparanda in helping understand the historical development of both Yeniseic and Na-

Dene will be highlighted along the way. My conclusion is that this body of data offers the

first meaningful indication of a genetic relationship between the two language families

and also proves that Na-Dene and Haida cannot possibly form a genetic unit.

In the sections that follow, I will make extensive use of the Proto-Na-Dene

(PND), Proto-Athabaskan-Eyak (PAE) and Proto-Athabaskan (PA) reconstructions that

have been worked out over many decades. The symbol ~ before a reconstructed proto-

form indicates that the author regards it as approximate; this is generally the case with

Leer's (2008) reconstructions linking Tlingit with Athabaskan-Eyak. I will not, however,

generally attempt to offer Proto-Yeniseic forms, which have not yet been worked out

systematically; in fact, at present we still lack a firm reconstruction of the basic Proto-

Yeniseic sound inventory. Instead, I will use cognate forms in the attested Yeniseic

daughter languages to illustrate the points I am trying to make.

2. Verb affixes and pronominal elements

2.1. Verb prefixes, typology, and genetic linguistics

This section compares verb structure in a number of geographically disparate

language families, including Yeniseic and Na-Dene, noted for possessing a rigid series of

verb prefixal classes. It will be shown that a wide variety exists in the types of prefixing

verb structures found throughout the world, with Yeniseic and Na-Dene sharing a unique

core of morphological traits. It will be argued that this set of homologies is not due to

typology or coincidence but rather derives from a common genetic origin. The notion that

Yeniseic shows any special linguistic affinity to Southern Eurasia is not born out by the

details of its verb structure.

Yeniseic languages are famous for having a strongly prefixing verb in an area of

the world dominated by suffixation. The Modern Ket verb template is shown in Table 1.

Dene-Yeniseic Symposium, Fairbanks; Ed Vajda Draft March 3.08 Comments welcome (eddievajda@yahoo.com ) 4

Table 1. Modern Ket verb structure

suffix

new

subj.

person

agr

clitic

incor-

porated

noun or

adj. root,

or new

verb base

obj

agr.

thematic

con-

sonants

tense/mood

marker

( s or ƒa )

inanimate

marker b

tense/mood/

aspect

consonant

( l or n )

old 1p,

subject

agr

verb

base

animate

subject

plural

Vajda (2007) shows that the Modern Ket verb innovated a new lexical root position

toward the left of the verb complex (position 7) in imitation of the suffixing languages

spoken around it. The new position 8 subject markers in Modern Ket are clitics rather

than true prefixes and often encliticize to the preceding word rather than attaching to the

verb form itself. All productive patterns of Ket verbs thus begin phonologically with a

lexical root, in much the same way as the neighboring Samoyedic, Turkic and Tungusic

languages. The oldest layer of verbs, however, is invariably prefixing and root-final, as is

the case in all documented Yeniseic languages.

Comparison with verbs from extinct Yeniseic languages shown that the Modern

Ket verb grew out of the structural model shown in Table 2.

Table 2. The Proto-Yeniseic verb complex

morphemes outside the

phonological verb

verb base

verbal

complements

(adverb,

object NP, etc.)

shape prefix,

incorporated

body part

noun

animacy

classifier

d - anim.

b - inan.

tense, mood,

aspect

(combination of

two markers)

( s or ƒa ) + ( l or n )

subject

agreement

(1 or 2 p)

or stative

resultative

prefix

The prefixing structures of both Modern Ket and Proto-Yeniseic are quite different from

other prefixing verbs of the Old World. Below are verb templates from three languages

(Sumerian, Burushaski and Abkhaz) that have sometimes been linked to Yeniseic as part

of previous speculations about the family's external genetic connections.

Table 3. Sumerian verb structure (based on Rubio 2007: 1139)

modal

prefixes

connective

prefix

(possibly

meaning

'also, 'then')

conjugation

(tense/aspect)

prefixes

dative

(indirect

object

pronominal

agreement)

dimensional

prefixes

(locative,

ablative,

etc.)

2, 3p

agreement

(non-

agent)

root

pronominal

suffixes

(agent

agreement)

Table 4. Burushaski verb structure (based on Tikkanen 1995:91)

neg-

ative

thematic

prefix

person/

number/

class

agr.

prefix

causative,

benefactive

root distrib-

utive

pl.

subject/

action

suffix

subject

agr.

suffix

dur-

ative

non

past

optative,

participle,

converb,

infinitive

2,3p or

plural

subject

agr.

suffix

inter-

rogative

Dene-Yeniseic Symposium, Fairbanks; Ed Vajda Draft March 3.08 Comments welcome (eddievajda@yahoo.com ) 5

Table 5. Partial template of Abkhaz verb affixes, based on examples in Hewitt (1979)

direct obj

prefix

subject

prefix

caus-

ative

root

Suffixes expressing: negation,

tense/mood/aspect, stative/dynamic,

finite/non-finite verb form

The simple presence of an unusual typological feature, such as prefixes arranged in a

rigid series of position classes, does not by itself indicate genetic relationship. To further

illustrate how fundamentally different templatic prefixing languages can be from one

another, Table 6 shows another well-known prefixing position-class verb, that of Bantu.

The interdigitation of pronominal and tense/mood/aspect prefixes in Bantu, though

likewise completely distinct from either Yeniseic or Na-Dene, is nevertheless closer to

both typologically than either is to the other prefixing languages of Eurasia or the

Americas.

Table 6. Ha (Bantu, Tanzania) verb morphology (Harjula 2004:86)

locative

clitic

subject

prefix

several

slots for

TAM

distinctions

object

prefix

root

derivational

suffix

TAM

locative

clitic

As can be seen, prefixing verb morphologies are by no means monolithic typologically.

Yeniseic no more belongs to the southern Eurasian typological area than it does to Africa.

The next section demonstrates that, among the world's prefixing verb systems, the

affinity shared by Yeniseic and Na-Dene goes far beyond mere typology, extending to

intricate specific systems of morphological homologies.

2.2. Na-Dene and Yeniseic prefixal verb morphology compared

All Na-Dene languages likewise have a strongly prefixing verb structure. This is

one of the morphological hallmarks of the family that distinguishes it from other New

World families. Other North American languages with strongly prefixing verbs, such as

Algonqian and Caddoan (Melnar 2004), show a completely different arrangement of

prefixes. These and other prefixing languages in the New World are as different from

Athabaskan, Eyak and Tlingit as the prefixing languages of southern Eurasia are from

Ket. Modern Navajo verb structure, shown in Table 7, provides an illustration of the type

of prefixation found in Na-Dene languages.

Table 7. Position classes in the Modern Navajo verb

( for published descriptions, with right-to-left numbering, cf. Young & Morgan 1987:37-8, and Young 2000:18-26)

10 9 8 7 6 5 4 3 + 2 1 0

"outer objects"

(incorporated

postpositional

construction,

indirect obj.

reciprocal, etc.)

outer

lexical

prefixes

distrib-

utive

plural

/da/

(more

than 2)

direct

object

3 rd

person

subject

inner

prefixes

(shape

prefixes,

etc.)

" conjugation "

prefix

relating to

tense/mood/

aspect

( si, ni, ƒ i )

1 st , 2 nd

subject

" classifier "

(=valence

change

prefix)

¬, d

l, Ø

indirect obj.

prefix

TAM

verb stem

= set of

root + suffix

allomorphs

expressing

tense, mood,

aspect

disjunct prefixes | conjunct prefixes | conjugation+subject | classifier + stem

iter-

ative

prefix

Plik z chomika:

Inne pliki z tego folderu:

Inne foldery tego chomika: