ECOLING at aol.com
ECOLING at aol.com
Tue Aug 31 03:56:19 UTC 1999
Concerning John McLaughlin's post today
on Algonquian and Wiyot-Yurok.
I was interested in a particular stage in the process,
not the current state of affairs today (see below).
John addresses mostly the question whether
the Comparative Method has now succeeded
in establishing the relationship of these languages.
When I wrote my previous message,
I believed it had by now done so to the satisfaction
of at least most, but since I had not recently heard
of the views of some previous doubters,
I at least hope I was precise in my previous
statements, not committing as to all current views.
Give the matching irregularities mentioned by John,
I would assume that there are no serious doubters today.
I certainly know of none.
In my previous message, however, my point
was quite a different one. I was NOT AT ALL
interested in whether Algonquian and Wiyot-Yurok
were demonstrably relatable by the Comparative
Method today. That is the eventual result of a long
process. I was interested in this as a case study,
specifically a case study of an INTERMEDIATE
PERIOD in discovery or proof (whichever way you
wish to think about it).
The statement I made was that AT ONE TIME
at least some conservative comparativists
did NOT accept that the Comparative Method could establish
or had established the relationships in this case (I do not
remember enough to distinguish those two phrasings).
But the SAME people believed firmly that the
languages were indeed related. It necessarily
follows from this that they believed this based
on something other than a complete proof using the
comparative method. So I was saying that this
potentially is a way to treat the Comparative Method
empirically, to discover WHAT would cause sane
and conservative people to become convinced
of a language relationship BEFORE there had been
a proof, or at least a complete proof, via the
Therefore: convincing evidence of relationship
NEED NOT BE a complete proof by the Comparative
Method, though it may perhaps necessarily involve some
contribution by the Comparative Method (that is
the narrowest possible conclusion I can draw).
Discovering the answer to THAT question,
how someone could be convinced not entirely by the C.M.,
could lead to extensions of the Comparative Method,
which some might regard as secure tools, others
might regard as heuristics, and we need not be
concerned about what they choose to call it,
as long as we regard it as a legitimate step in the
process from inkling to discovery to hypothesis to proof.
THAT process, as a continuum not a dichotomy,
was my point.
When I mentioned this sort of paradox in another context
some years back, Karl Teeter very kindly sent me his
work on the subject. Of course HE was quite convinced
already then that the Comparative Method had solved
the problem. My statements were about an earlier stage.
(If I did not name names, there was a reason,
it might embarrass someone or cause a denial,
when viewed now many years later in retrospect.
I can however safely say that
one of the highest placed and most respected and most
conservative people in the Amerindianist field was the
doubter whose views I most clearly remember
as being under discussion at that time, and
that another of the highest placed and most respected
people in the Amerindianist field, though not such an
extreme conservative, was the person who first observed
the paradox of being convinced of the relationship yet
not by a complete Comparative Method proof,
and verbalized it explicitly, bringing it to my attention.)
No one example matters,
so let's go to the other one I mentioned to see the analogy:
Albanian and IndoEuropean.
There are I would assume none today who doubt that
Albanian is proven by the Comparative Method to be
an Indo-European language.
Yet only recently one correspondent stated that Ringe and the
Philadelphia group excluded Albanian from their algorithmic
approach to the family tree of Indo-European because their
algorithms would not work on it. (Their algorithms use less
information than does the Comparative Method, it is safe to say?)
Was there perhaps a similar situation with Albanian,
that the relationship was AT SOME STAGE accepted as
certain by a large proportion of those interested, yet they
would have said that at that time there was as yet no proof by
the Comparative Method? If so, it is another example of what
I am referring to. In distinction to the Algonquian-Wiyot/Yurok
example, I do not have the benefit of capturing intermediate
stages as moments in time, I was not present to observe
intermediate stages in the linking of Albanian to IE.
Just for consistency, since we are concerned in these discusions
with the legitimacy of logical arguments more than whether
particular languages are or are not related, if we are trying to
strengthen our tools...
I need to point out an oddity of reasoning.
John McLaughlin writes:
>For example, a -t- is inserted in Proto-Algonkian
>between a possessive pronominal prefix and a vowel-initial root,
>whereas in Wiyot a -t- is inserted
>between possessive prefixes and a root beginning with hV
>(with the subsequent loss of the h).
>Algonkian *ne + *ehkw- = *netehkw- 'my louse',
>Wiyot du- + hikw = dutikw 'my louse'.
Do not misunderstand, I am not going to argue that this is NOT
evidence. I am merely going to point out, for consistency,
that it is WEAK evidence. The -t- is about the weakest possible
consonant which could substitute for hiatus or glottal stop,
so it COULD HAVE arisen independently in two cases,
and not be an inherited relic at all. Because other evidence
seems consistent with a genetic relation, it seems reasonable
to assume that this also is a genetic inheritance in this case.
Now consider Greenberg's attempt to find a morphological
irregularity to bolster his claims of genetic relationship among
some of the languages of his Amerind. He adduced some alternation
I think it was in 3rd sg. pronominals between y- and t-.
This is WEAK evidence, for exactly the same reasons that it is
weak evidence in the Wiyot-Algonquian case. Because we do NOT
(yet?) have evidence to convince most that all of these "Amerind"
languages can be demonstrated to be related, it is argued against
Greenberg that this could have arisen by chance, and is therefore
"not" evidence. But evidence is evidence, whether or not it is part of
a successful proof (or legal case). It is simply not logically consistent
to accept in one case but not in another this kind of very weak
resemblance, involving the weakest, least information-rich sounds.
(Yes, I could try to argue along with the rest of you that the Wiyot-
Algonquian case might be better if it is specific to pronominal prefixes;
and of course because much more effort has been spent on it too!
But that would be cheating, based on the final result much later.)
To have a SCIENTIFIC algorithm or method for judging relatedness,
we must assign a very small weight to a "lookalike" of this kind,
but not a zero weight, and we must do it consistently,
not fudging our examples by pretending it is there only in cases
where we "know" the ultimate answer from some additional data.
More information about the Indo-european