Random Noise in Multilateral Comparison.

[ECOLING at aol.com]

In a discussion about Random Noise,
I was making a point that it is much less serious a problem
in Multilateral Comparison than it is when one is trying to make
an argument that two particular selected languages
ARE GENETICALLY RELATED.

>> The point I made about random noise had NOT to do with whether
>> particular languages are ultimately related, but whether a given language
>> or family W is more closely related to others X or Y.
>> In that context, why should "random" (by definition) noise in the data
>> selectively favor W to X rather than W to Y?  No possible reason that
>> I can imagine.

To this Larry Trask replied:

>But this only holds good if you assume in advance that all languages are
>related.  This is exactly the assumption which I chided you for earlier,
>and it is also exactly the assumption which you have just told me in an
>off-list posting that you do not hold.  So what is going on?

There is a severe gap in Trask's logic here.
There is no inconsistency on my part.

Saying that W is more closely related to X than to Y
INCLUDES the possibility that it is related to X and not to Y.
Saying that W is more closely related to Y than to X
INCLUDES the possibility that it is related to Y and not to X.
Of course all three may be ultimately related,
or none of the three may be, as well.
So there is absolutely no necessary assumption about
absolute statements of relationship needed in Multilateral Comparison.

***

Greenberg assumes that all languages are ultimately related,
or at least within the domains to which he has applied
Multilateral Comparison.

I do NOT make that assumption.
(But I have no difficulty reading the work of someone
who makes the assumption, because it affects almost
nothing except preferred choice of words.)

Multilateral Comparison works the same way either
with or without that assumption.

With the assumption of ultimate relatedness,
we get a single family tree.

Without the assumption,
one simply concludes that the languages which Multilateral
Comparison groups as closer are more likely to be related,
or are likely to be more closely related; this includes of course
the converse that those which Multilateral Comparison groups
least closely are the ones most likely to be unrelated.

IT DOES NOT MATTER whether Greenberg takes all languages
in his studies to be related,
as most of his statements seem to imply,
or whether he is simply making a working assumption
(which he much too rarely says).  Whichever he actually believes,
what he has actually accomplished by
doing Multilateral Comparison is the same in either case.
It yields the same kinds of results, just expressed in different words.

Within a context like all of the Americas, Multilateral Comparison
method essentially gave the results that Na-Dene and
Eskimo-Aleut were UNRELATED (note: WITHIN that
domain, whatever they are ultimately).

Good linguists should have no difficulty whatsoever
in using data sets from people whose assumptions differ
from theirs in this structurally trivial way.  The kind of
information yielded by Greenberg's use of Multilateral
Comparison is structurally the same whatever the reader's
assumptions about ultimate relationships.

There is no elaborate translation required for understanding.

There is merely a gracious tolerance required for
conversations with other linguists with whom one does
not agree about some things.

Best wishes,
Lloyd Anderson