LINGUIST List 5.1317

Sat 19 Nov 1994

Disc: Comparative method

Editor for this issue: <>


  1. , Re: 5.1288 Comparative Method
  2. Ecological Linguistics,Anderson,PRT, Re: 5.1288 Comparative Method

Message 1: Re: 5.1288 Comparative Method

Date: Sun, 13 Nov 94 19:30:14 ESRe: 5.1288 Comparative Method
From: <>
Subject: Re: 5.1288 Comparative Method

In think that it is me that Bill Poser is responding to when he
says that he can find no evidence for the assertion that those who
maintain (without any factual basis, as we have already established)
a supposed 8 or 10 millenium ceiling on the comparative typically
seem to be, as I said, "gunning after Nostratic or Amerind". I don't
see why Poser thinks that this was "ad hominem" argument, but in any
case one person who seems to me to make the connection between
the alleged ceiling and the refusal to accept (or even seriously
consider) the Nostratic hypothesis is Johanna Nichols on p. 6 of
her book Linguistic Diversity in Space and Time. Despite all my
respect for both Bill and Johanna (which I only feel it necessary
to affirm because of the startling ad-hominemity charge, since normally
I would just assume that it is taken for granted), I believe quite
simply that (a) there is no such ceiling, (b) Nostratic is a very
powerful theory and probably right, (c) people DO connect the
two issues all the time.

Alexis MR
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 2: Re: 5.1288 Comparative Method

Date: 15 Nov 94 15:23 GMT
From: Ecological Linguistics,Anderson,PRT <>
Subject: Re: 5.1288 Comparative Method

Comparative method

Mr. William Poser notes his disbelief that supposed limitations on the
comparative method have been used to block consideration of hypotheses about
deep genetic relationships.

I would like to let these things lie, but the lack of knowledge of some of the
true history, or the reluctance to believe it, is very much like the folks who
deny the Holocaust, fortunately not in severity but in kind. The fact of
professional misbehavior in the resistance to Greenberg's hypotheses is a fact.
It of course in no way argues in favor of Greenberg's hypotheses, but it does
impugn the value of the critics as witnesses to fact (Mr. Poser's last point).

The attempt to deny this misbehavior is in the long run a discredit to
legitimate practitioners in the field, because it lumps the misbehaving with
the very good work that even some of the same people do. We do need to
distinguish between these, and rap knuckles of those who misbehave whoever they
may be.

Mr. Poser's wording of his question underwent two transformations from his
first version to his third, so I will try to deal with each of them precisely.

)The recent discussion of limitations on the comparative method
)contained several assertions that such limitations had been and
)were being used to justify resistance to proposals of remote
)relationships, including Amerind and Nostratic. To my knowledge
)this is absolutely false.

***To my knowledge this is quite true. At the Albuquerque LSA Summer
***Institute many years ago, Ives Goddard did exactly that. He said
***that languages are ***known*** to change so rapidly that discovering
***links at that depth was ***impossible***. This is not the only time.

)In every instance that I am aware of
)in which perceived temporal limitations on the comparative method
)have been mentioned, it is by way of explaining why it is there are
)no relationships beyond a certain degree of remoteness on which
)there is consensus, or by way of making predictions as to what
)historical linguistics will ultimately be able to achieve.

This was not at all the tone of Ives Goddard's remarks. He was not merely
predicting, he was exemplifying ridiculing and promoting ridiculing. He was
indeed using the question of time depth (which he referred to in the many tens
of thousands of years, by the way, not the 12,000 mentioned by Poser) to
justify "resistance to [consideration of] proposals of remote relationships",
even without regard to the errors in the data. Though of course he did *also*
point out what he considered errors in the data.
 [In my quote from Posner, above, I inserted the words [consideration of]
before "proposals", which I hope readers of this will consider not a change of
the sense but a specification of what it Mr. Poser legitimately meant, given
the careful distinction he was trying to draw.]

)why it is there are no relationships beyond a certain degree
)of remoteness on which there is consensus

There has been a very strong tone of this also (slightly different from the
phrasing in Mr. Poser's initial paragraph). That is, there does come up
repeatedly the innuendo or invited inference that if it has not been discovered
by competent specialists already, it cannot be. This is not ever asserted
quite outright, of course, but in pussyfooting around it, one can see one
manifestation of it in the discussions of what ***historical-comparative method
can achieve*** substituting for discussions of what ***a particular set of
methods or rules of evidence can achieve***. These are two quite different
things, because defining comparative-historical method circularly as the set of
methods currently used, with no extensions ever to be allowed, no refinements,
no methods of handling data which can better sort out noise than the ones we
now use. Such an assertion in any field of science is laughable, in
comparative-historical linguistics no less than in any other. Of course it is
not outright asserted. But some discussion treats it as an assumption.

One of my favorite examples is the rule against using less than three-consonant
matches for putative cognates, because otherwise too many chance resemblances
can creep in, etc. If this rule is enforced, then we can never prove (or even
consider!!!) a relationship of Siouxan-Yuchi, for example, since one of the
members has only very short roots available for comparison (I have forgotten
some of the details of this), and this impossibility is enforced even ***
independent of the time depth***. I would like to see the comparative-
historical stalwarts being just as critical of such supposed rules, and
ridiculing them just as loudly as they do Greenberg. But of course they do
not. They do not in general take sufficient distance to look back at current
supposed rules of comparative-historical linguistics and actually attempt to
prove some of them inapplicable if taken as absolutes.

This is of course not to deny that having three-consonant comparisons is better
than having only two-consonat comparisons, and indeed precisely for the reason
asserted, namely that it is better at eliminating noise, that is, chance
resemblances. I here use the term "noise" because I would like for
comparative-historical linguists to see their science as merely one among many
which deals with noisy data, one which can achieve more when methods are
improved for identifying and eliminating the effects of noise on our
judgements. The problem is the use of the supposed "rule" as an absolute, and
the failure to rap the knuckles of those who do so just as loudly as rapping
the knuckles of Greenberg and associates.

 [My personal interpretation is that some of these partisans take defeat of
Greenberg to be such an overriding concern, or they simply are so appalled,
that they either cannot conceive of any vulnerabilities of current comparative-
historical practice, or else they do not want any public perception of such
vulnerabilities to interfere with the campaign against Greenberg. The basis
for this interpretation on my part is hearing overt statements that they are
concerned with student enrolments in traditional courses and that is why they
are fighting Greenberg. They might do better to keep the excitement of
interesting hypotheses, and pose the challenge of how to deal with them
rigorously. That would be a challenge which could attract a new generation of
students, to improve the ability of comparative-historical techniques to deal
with noisy data and distant time depths.]

Concerning the matter of errors, the treatment of them can be evaluated for its
forthrightness. I remember during the early years of the debate, when Goddard,
Chafe, and Campbell pointed out the many errors making Greenberg's work
"worthless", that I repeatedly communicated with each of them asking for a list
of all the errors they had identified (so that I could try to estimate whether
correcting them would change the direction of the relationships which Greenberg
claimed was closer). No such lists were forthcoming, nor any notification to
me of later publication of them. I believe the individuals in question
sincerely believed there were so many errors that ... but did not feel a
responsibility to provide a list of those errors when making the public
condemnation. That is in my view a lack of respect for the right of all to
judge the evidence, rather than merely assertions by authority, which is in
fact what we were faced with for a considerable time and still to a
considerable degree.

Robert Rankin did published his list for areas of his specialty (in IJAL). For
this I certainly commend him. I find his discussion odd for two reasons. One
is because I am a pattern-finder, and it seems that Greenberg did indeed switch
columns at some stage in recopying his data, mixing up a couple of the
languages to which words were attributed. This would be I think the conclusion
of standard studies of manuscript transmissions. (Rankin notes other errors
which I would not attribute to this kind of column-switch, but the expanation
of switching columns is I believe still valid for a substantial portion of the
The second oddity is that Rankin himself says that perhaps correcting these
errors would not affect the conclusions one would draw for grouping these
languages (I am not referring back to the article at this point so please
forgive a slight inaccuracy here). But then Rankin fails to follow up on his
own comment, by evaluating the possibility that Greenberg's conclusions might
not be affected by correcting this particular data (most of which was a mixup
on which language the data came from within a given well recognized family).

This is very important, because if the results would not be substantially
changed, it will invalidate a prominent claim made ***many times*** by
Greenberg's critics that the errors make the entire study "worthless" or
"seriously flawed" (use of this last term is a dead giveaway for an academic
posing as the ultimate higher authority judge of someone else who is supposedly

One of Greenberg's points has been that the methods he is using, selecting the
closest from among innumerable potential language matchings rather than simply
trying to prove two particular languages to be related, has some degree of
resistance to noise in the data. His own errors will be one type of such
noise. Since this has been a major point of debate, it is surprising to me
that Greenberg's detractors do not face the issue purely as a factual matter.
If they do, they will actually let their corrections of his errors be tested
for whether such corrections substantially change which languages come out as
more closely related, by his methods. (Note, by the way, that the standard of
evaluation of this question is ***not*** how much evidence there remains for a
particular two-language connection, since that is not Greenberg's method, and
since any identification of errors in anything will always reduce the remainind
data, circularly by definition.)

So let's indeed deal with the factual questions about how resistant to noise
Greenberg's methods actually are.

And discover as many other ways to increase resistance to noise as we can,
since after all Greenberg's is merely one tool, along with the *other* tools
used in historical and comparative linguistics, in a larger task of gaining
more understanding of language and human history.

Back to the third of Mr. Poser's three phrasings:

)I do not
)know of a single instance in which someone has argued:
) Such and such a proposed relationship is associated
) with a time-depth of X years. This exceeds the
) temporal limits of the comparative method. Therefore
) the proposal must be wrong.
)If anyone can provide evidence of such an argument being made I would
)be most interested.

The wording above is quite different from the wording Mr. Poser used earlier,
if we are being exact. "The proposal must be wrong" is quite different from
the claim "It is wrong to make the proposal or to consider it." Only the
second, or something like "we cannot ever know if a proposal at such a time
depth could be right", could follow purely logically from an assertion that
relationships cannot be proved beyond a certain time depth. I do not
specifically recall anyone juxtaposing the two statements which are NOT
logically connected so blatantly as Mr. Posner does here. But in any case this
is not the point. It is the earlier forms of Mr. Poser's question which are
relevant to our discussion, and no rewording such as this can make the
misbehavers innocent by converting the question into a straw claim which of
course they were not so silly as to assert overtly.

)In sum, whatever the validity of proposed temporal limits on the
)comparative method, and I agree that such limits are far from exact,
)the view that this has anything to do with reactions to Greenberg's
)work on Amerind and similar work is a red-herring.

No, it is not. Now Mr. Poser is asserting a conclusion based on his assumption
of lack of misbehavior, rather than, as he began, saying that he was not aware
of any and asking to be notified if others knew differently. I do know
differently, and actually, I think, so does he, if he took more time to examine
from various perspectives, but it probably never occurred to him that some of
the "rules" being asserted as absolutes were forms of misbehavior because they
would indeed rule out even the consideration of hypotheses, even such as

)To evaluate such proposals, look at the data and look at the methodology,
)not at the alleged (and generally unknowable) motivations of the
)critics. You'd think that the irrelevance of ad hominem arguments
)except in matters of credibility of witnesses would not need to be
)repeated constantly.

)Bill Poser

I agree with this entirely. In this case, of course, we ***are*** concerned
with the credibility of some professionals as witnesses to what is actually
done in comparative-historical linguistics, and as to the effect of noise in
data on the possibility of drawing conclusions or on particular conclusions.
That is, they are both misreporting what they actually do (unconsciously in
most instances, I think), and holding defeat of Greenberg's approach to be so
important that they have been slipshod both in their logic and in their respect
for the right of other professionals to judge based on the facts themselves
rather than assertions of authority.

I have responded to Mr. Posner's message at length because his difficulty in
believing that there has been misbehavior coloring these debates shades over so
far into a rewriting of history. Such assertions must not go unchallenged by
those who have been witnesses to relevant facts and who care very deeply about
historical linguistics being an open discipline responding to actual data,
rather than to personalities or vested interests. The value of recovering
parts of the history of the human mind is simply too great to tolerate blocking
further progress for illegitimate reasons.

As an aside, a propos of recent discussions of typology, I should note that
there also been a long-standing antipathy towards linguistic typology among
comparative-historical linguists, based I believe on excesses of the era of
Schmidt's Die Sprachkreisen der Erde and on "purely superficial comparisons"
(i.e. those which do not make sufficient use of, nor show sufficient respect
for, the special skills of the historical linguists). It has led to a
difficulty in seeing that typology can be a typology of paths of change of
particular constructions. Typological-historical linguistics is just as
crucial to comparative-historical linguists, by augmenting their unconscious
judgements of the likelihood that there could be an A from which B and C both
descend, as the results of comparative-historical linguistics are to
typological historical linguists, in providing basic data on known or at least
highly plausible cases of historical changes, the necessary basis for any
typology of known possibilities. The two are inseparable parts of the same
effort at understanding our world. The field of typological historical
linguistics only makes more explicit what all comparative-historical linguists
must do, whether consciously or implicitly. We need to keep all the parts of
linguistics in touch with each other, not allow any one part to become isolated
and develop too many merely in-group conventional customs about what kind of
cookies are allowed with one's tea, customs which may easily lose their
connection with their empirical basis fostering reliable conclusions.

Just as a personal note, even at the Albuquerque meeting mentioned, I indicated
why I thought some of Greenberg's conclusions about pronouns *might be wrong*,
from work I had done. But, unlike Goddard, I was not questioning the value of
his raising specific hypotheses, and certainly was also explicit that he might
be right. How could I possibly "know" otherwise without time to examine his
reasoning and data in greater detail than was possible on a new presentation
(and even after such an opportunity, I would probably not know the answer; that
is for a future generation, for whom we can provide some better patterned and
cleaner data).

I am one of the more ***conservative*** in what I consider sufficient evidence
for a claim, but one of the more ***liberal*** in what I consider legitimate
hypotheses to consider, precisely because I know enough of history of science
to know that we cannot predict in advance what kinds of evidence will be
forthcoming or will turn out to be probative. Some are not so modest about the
absoluteness of their knowledge. The most recent example of my work in this
field is a book on the undeciphered writing system of La Mojarra, Veracruz,
Mexico, in which I provide the latest in concordances to glyphic phrasings and
potential parallels as aids to decipherment, with only very few interpretations
claimed as have some substantial support in the data. In contrast to some
other workers on this ***still undeciphered*** text, I avoided grandiose claims
to substantial decipherments of the vast majority of the text.

Sincerely, Lloyd Anderson

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