Yeniseian and Na-Dene

Thu Nov 12 01:28:08 UTC 1998

----------------------------Original message----------------------------
Steven Schaufele writes:

...  a Reuters story (posted on www.cnn.com)
about an article by Merritt Ruhlen appearing this week in the
Proceedings of the National Academy of Sciences, claiming an affiliation
between Ket, the sole surviving Yeniseian language, and the Na-Dene
family of North America.
...

I'm wanting to know, does anybody subscribing to this EBB know anything
about this?  Is there anything to this proposed affiliation?  And if
not, is anybody doing anything about clarifying the issue for the
general public?

- - - - - - - - - - - - - - - - - - - - - - - - -

I have seen the article and discussed it with a couple of journalists who
contacted me about it.  Here is the gist of what I said:

The burden of proof is on Ruhlen to demonstrate that his resemblances
exceed what is expected by chance.  There are well-known tests, standards,
criteria, etc. for doing this in linguistics and in statistics.  He has not
invoked or applied any such criteria, so his findings must be assumed to be
chance resemblances until and unless he shows otherwise.

Ruhlen lists 36 word sets that he considers resemblant and indicative of
genetic relatedness between Yeniseian and Na-Dene.  In fact these 36 sets
are not numerous enough and not closely enough resemblant to exceed the
range of chance.

I've worked out the chances of finding words with two similar consonants in
the same order, with similar but not necessarily identical meanings, in two
languages.  Out of a fixed list of 100 meanings chosen in advance (this is
analogous to asking "what is the probability that similar forms will mean
'water' in both languages?", and so on for another 99 glosses), this is how
many resemblant sets it takes to exceed the range of chance and show that
relatedness is likely:

2-consonant words with the very same meaning:   7
2-consonant words with similar meanings (modeling this as a search that
allows up to 5 senses' leeway, e.g. for 'fly' also 'flee', 'wing', or
whatever; these must also be specified in advance):  25
1-consonant words (or 2-consonant words with one resemblant consonant and
one non-resemblant one) with the same meaning:  27
1-consonant words with 5 senses' leeway each:  over 50

Ruhlen has:

15 words with two resemblant consonants in the same order
2 words with two resemblant consonants but in different orders or with
other consonants intervening ('children' and 'foot')
18 words with one resemblant consonant
1 word with perhaps no resemblant consonant ('name', where the initial
glottal stop in the Proto-Yeniseian form is found in no daughter language
and must there be phonetic detail of the pronunciation of initial vowels
rather than a true structural consonant; and where similarly there doesn't
seem to be any evidence for reconstructing a glottal stop as initial, or
even as non-initial but the first consonant in the root, for Na-Dene).

For simplicity and to give Ruhlen the benefit of the doubt these can be
described as 17 two-consonant resemblances and 19 one-consonant
resemblances.

11 of Ruhlen's sets have considerable semantic leeway; most have some
leeway; for only a few is the sense really the same.  So this can be
modeled as a search of up to perhaps five senses.  In practice this means
the researcher finds a word in one of the languages and gets to cast about
in the other language looking for a word of similar form in a meaning
plausibly connected to that of the first word.  Occasionally s/he is lucky
and finds a resemblance in the very same sense; usually s/he has to search
through a few close senses; and sometimes s/he has to search longer.

So even if Ruhlen had started out with a strict list of 100 glosses and a
list of 5 senses' worth of leeway for each, he would need about 50 sets of
the kind he has.  (I'm not calculating this with any precision.  I figure
he has about half of the 25 2-consonant words he needs, so he also needs to
have about half of the 50+ 1-consonant words he needs, and the sum of about
half of 27 and about half of over 50 is somewhere around 50.)

But he didn't set up a closed list of glosses.  His wordlist is not 100 but
some open-ended number.  At least the total number of words found in the
source on Yeniseian he draws on, or maybe the total of those plus any
additional meanings found in his various Na-Dene sources.  Or maybe it's
the average total number of root words per language on earth.  I don't know
quite how to model this, but he clearly has a wordlist of well over 100.
So he needs correspondingly more than the about 50 words he would need if
he used a pre-specified closed list of 100.

My calculation is for comparisons of actual daughter languages, while
Ruhlen compares protolanguages.  A protoform has a greater chance of
occuring in the protolanguage than in any daughter language.  (This is
because a protoform occurs in the protolanguage by definition, while for
any cognate set it's likely that one or another daughter language will
happen to lack a cognate.)  Therefore, for comparisons of protolanguages
larger numbers of resemblant words will be needed than for daughter
languages.

For instance, Ruhlen's set 'hunger' has Haida as its only Na-Dene
representative, and 'river' has Haida as the only form with a two-consonant
resemblance.  Some of the sets have no Haida cognate, e.g. 'foot', 'birch
bark', 'boat'.  From the point of view of methodology, this means he
allowed himself to adduce evidence from either Haida or from
Eyak-Athabaskan-Tlingit, and the chances of finding a resemblant in one or
another family are greater than finding a resemblant in one particular one.
Again, this can be modeled as a search in which if the researcher doesn't
find a resemblant in Eyak-Athabaskan-Tlingit s/he gets to look for one in
Haida.

In some sets an ejective consonant in Na-Dene corresponds to a sequence of
consonant plus vowel plus glottal stop in Yeniseian (e.g. 'birch bark'); in
others (e.g. 'clay') an ejective just corresponds to a Yeniseian consonant
with no glottal stop elsewhere in the word; in still others there is a
Yeniseian glottal stop but no ejective anywhere in Na-Dene (e.g. 'lake',
'word').  Thus Ruhlen seems to have allowed himself to scan consonants in
whatever way maximized resemblances.

For all of these reasons, Ruhlen's resemblances are no more numerous and no
closer than would be expected to come up by chance.  They are exactly the
sort of thing one finds when looking through dictionaries and casting about
to find resemblances with few constraints on how the search is to be
conducted.

This is all assuming that the linguistic analysis and the protoforms are
sound, matters I cannot judge.  Any errors in data, analysis, or
reconstruction increase the number of resemblances that must be adduced if
they are to be regarded as non-accidental.

The paper is not up to date on human genetics and archeology.  It mentions
only the Greenberg-Turner-Zegura (1986) theory of three American
settlements, not mentioning all the work on mitochondrial DNA which has
been done since 1987.  The claim that "the first migration of the Amerinds
[occurred] about 11,000 years ago" is no longer current; the well-dated,
well-accepted Monte Verde site in southern Chile at ca. 12,500 years ago
shows that the first Americans entered well over 11,000 years ago.

Johanna Nichols

* * * * * * * * * * * * * * * * * * * * *
Johanna Nichols
Professor
Department of Slavic Languages
Mailcode 2979
University of California, Berkeley
Berkeley, CA 94720, USA

Phone:  (1) (510) 642-1097 (direct)
        (1) (510) 642-2979 (messages)
Fax:    (1) (510) 642-6220 (departmental)
* * * * * * * * * * * * * * * * * * * * *