LINGUIST List 5.945

Fri 02 Sep 1994

Disc: Altaic, Binary comparison

Editor for this issue: <>


  1. "J. Marshall Unger", Re: 5.935 Altaic
  2. Alexis Manaster Ramer, Re: 5.939 Binary Comparison?
  3. Richard M. Alderson III, Re: 5.939 Binary Comparison?

Message 1: Re: 5.935 Altaic

Date: Wed, 31 Aug 1994 16:50:31 Re: 5.935 Altaic
From: "J. Marshall Unger" <>
Subject: Re: 5.935 Altaic

 For the record: (1) I don't think the claims of
Rona-Tas or Larry Clark should be dismissed out of hand; (2)
there was no conspiracy going on at Baldi's 1987 workshop;
(3) I don't think Miller's work is reliable; (4) I don't see
what harm is done by taking I Kimun's advice and looking at
Korean and Japanese in relation to Tungusic languages before
bringing in other witnesses.
 Those who want to pick a fight: please beat up on
Nichols and leave me alone. Better yet: instead of
spending effort negatively, how about doing something
positive, like systematizing and evaluating the etymologies
in the Tungusic etymological dictionary so that they can be
used by others? I recall Austerlitz saying, in our
discussions, that he found the dictionary to be very uneven in
quality: it would be most helpful to have a short list of
really secure etymologies and reconstructions for Tungusic
to work with.
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 2: Re: 5.939 Binary Comparison?

Date: Thu, 1 Sep 94 08:37:21 -04Re: 5.939 Binary Comparison?
From: Alexis Manaster Ramer <>
Subject: Re: 5.939 Binary Comparison?

In response to Scott DeLancey's query, some of the so-called
anti-Altaicists HAVE stated that only binary comparison should
be allowed (e.g., Turkic and Mongolic but not Turkic, Mongolic,
and (Manchu-)Tungusic all at once). I am sure that Doerfer has
said this although I have no reference at hand, and I am also
sure that no basis for this has been given (much as he gave no
basis for his assertion vis-a-vis Nostratic that you cannot
compare reconstructed (proto-)languages to each other). As I
read Ringe's little book and his exchange with Greenberg in
the Procs of the Am. Philosophical Society (see my forthcoming
review in Diachronic, he seems to be saying that (a) language
classifications should be done on the basis of overtly probabilistic
arguments about the distributions of sounds in certain positions
in certain lists of words with fixed meanings, and (b) when doing
this, you should only do two languages at a time rather than more
than two, because it is, informally speaking, easier to get
a spurious resemblence between more than two than between two.

Neither point seems acceptable to me, neither the first, since
as the discussion between Ringe and Greenberg (if I can call
that mud-fest a discussion) seems to have concluded with one
point being agreed on and one only, namely, that the kind of
method Ringe has in mind could not show the relatedness of
certain pairs of Indo-European languages, nor the second, since
the only situation Ringe evaluates carefully in his attack on
n-ary vs. binary comparison is one where you compare a whole
bunch of languages (15 I think in his examp) and only require
a small subset of the total to agree on any given word (I forget
how many in his work, but it is certainly less than 13 or 14).
Now, in reality, the odds of a chance resemblence can be much
higher or much lower with binary than with n-ary comparison
depending on how large the total set of languages is, how many
of these are required to agree, and what the probabilities are
of the various outcomes (e.g., are there two possible phonemes
for a given position in each language and are they equally
frequent or are there twenty).

So, yes, Scott, there is a Santa Claus, and there are published
arguments that assume that binary comparison is better than
n-ary or indeed the only valid method to use. (I will look
for specific references in the anti-Altaic literature, unless
Sasha Vovin beats me to it).

Manaster Ramer
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 3: Re: 5.939 Binary Comparison?

Date: Wed, 31 Aug 1994 15:19:19 Re: 5.939 Binary Comparison?
From: Richard M. Alderson III <>
Subject: Re: 5.939 Binary Comparison?

Like Scott DeLancey, I was intrigued by AMR's references to binary comparison.
I have in the past asked for references for this proposed methodology, because
the only place I have seen it referred to is in the writings of Greenberg and
Ruhlen, where it is characterized as the way "professional linguists" (G & R's
scare quotes, not mine) do things.

Haas, in _The Prehistory of Languages_, makes reference to Bloomfield's dictum
regarding the best number of languages with which to start a comparison; this
is drawn from both his _Language_ and his Algonquian paper in Hoijer et al.
Dyen took it so greatly to heart that in his course on Comapartive Method he
insisted on allowing students no fewer, and no more, than four languages for
their term project.

Anttila and Lehmann at least tacitly assume non-binary comparison in the
examples presented in their textbooks; I assume that Hock does as well.

As an Indo-Europeanist, I have always assumed a large number of comparanda to
be desirable, as did my teachers in any comparative study I undertook.

So, once again I must ask: Who are these linguists whom G & R see as the enemy
and the anti-Altaicists see as supports? And how have they concluded that a
binary comparison is reasonable?

 Rich Alderson
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue