Agree on Multilateral Comparison (w John McLaughlin)

[ECOLING at aol.com]

I thank John McLaughlin for his very balanced statement.

I agree probably with even a bit more of it than he thinks.

I would only change SLIGHTLY the wording of a couple
of statements following:

> [Multilateral Comparison] is not a
>conclusive proof of anything in and of itself because it ignores factors
>of chance, fails to properly identify borrowing, fails to take into account
>historical and other nonlinguistic factors, and is dependent on the quality
>and quantity of available materials for the languages involved.  It can
>be suggestive of potential areas for further research.

I agree that MC does not provide proof, but it does not "ignore" chance,
rather, it aims to gain useful results despite the vagaries of chance
and noise in the data.  This difference in phrasing is important,
because it removes the imputation of incompetence among those
who know perfectly well what its uses are, and that it cannot do
what the Comparative Method can do when sufficient information
is available and when the languages are close enough (approximate
conditions, I'm not concerned about excessive exactness just at
this point).

WHEN USED properly as a heuristic,
I am most concerned about a weakness of Multilateral Comparison
in being influenced too much by massive borrowing.
Given the limited goals for which Multilateral Comparison is appropriate,
perhaps we should then think in terms of the wisdom
that when borrowing is massive enough,
it is hard to say which is the "true" ancestor language.
The biases of most of us linguists in favor of grammar as carrying
the "true" genetic line can be considered as just that, biases,
in some points of view, even though there is also truth in that
point of view.

>Misused, it can give
>the appearance of a final proof of genetic relationship that may or may
>not actually be there.

It is not Multilateral Comparison which can give the appearance
of final proof.  It is improper claims made by some practitioners.
That is little different from any other method,
which can equally be abused by claiming more for it than it can do.
I do not consider Multilateral Comparison to be more subject to this
than any other method.  (And I do not consider this a variant of
the view that we should not blame the guns for the murder rate,
it is human beings that fire them.)  The availability of Multilateral
Comparison in the tool kit does NOT itself contribute
to misuse of Multilateral Comparison.  Rather, clear teaching
in our introductory and graduate linguistics departments of what
the technique of Multilateral Comparison really is, independent
of whatever claims may have been made for it at one time or another,
is the best antidote.  Celebrate its successes, point out that it can
be used to suggest
ultimate non-relationship (the less close resemblances)
just as much as ultimate relationship (the closer resemblances).
Provide inoculation against misuse that does not go overboard
by globally condemning its use and disregarding how useful
it has in fact been in the history of work on both preliminary
classifications globally and preliminary classifications within
many language families.  Teaching of historical linguistics
should not ignore heuristics, they are just as important,
even if less "final" (in mature fields), as are the methods used
to refine details when languages are known to be related.
It is the LACK of teaching of the proper use of Multilateral
Comparison which is most likely to give rise to abuses of it.

I even agree that Greenberg has most often overstated his case.
I think this is because he is attempting to use
Multilateral Comparison at a presumed depth of separation
much greater than that in most of its demonstrated
successful uses.
Some of the deepest of the successful cases
probably include the internal classifications within the
major highest-level families of African languages.

However, I was a direct witness to early conversations by
Greenberg in which he admitted that all the method could do
was show which languages or families were relatively
closer or more distant.  I wish he had maintained that
approach more generally.  I also wish his critics had used
their knowledge to correct his errors AND THEN to evaluate
whether their corrected data caused his way of applying
Multilateral Comparison to yield different results,
rather than (illogically) arguing that the existence of error
made all aspects of what he did worthless.  There is error
in virtually everything human beings do.  Even massive
error needs to be evaluated for its effect on the outcome,
simply calling it massive does not remove the need to deal
with the actual facts.

I still believe that the subtle study of both
similarities and differences between the African and
American contexts may effectively teach us what can and
cannot be accomplished by Multilateral Comparison,
and may permit us to refine various versions of that method.

Sincrerely,
Lloyd Anderson
Ecological Linguistics