Alexis on classification

[manaster at umich.edu]

----------------------------Original message----------------------------
 
Larry Trask writes:
 
>
> I therefore see no great obstacle to reasoning as follows: "Look --
> you've spent years scouring these languages for evidence of
> relatedness, and all you've been able to come up with is this
> miserable collection of junk.  Therefore, there are good grounds for
> concluding that the languages are not discoverably related at all."
>
 
There is in my view a subtle fallacy here (and also I think in
Scott Delancey's reply to my posting).  It is true that if someone's
arguments/data are exploded, they have no case left for their theory.
But this does not justify concluding that there IS no "discoverable"
case for the same theory.  For example, I showed that Sapir did not
so much have bad arguments as had no arguments at all for relating the
"Coahuiltecan" languages to each other, only for relating one of them
(Tonkawa) to Hokan (I am oversimplifying slightly).  But there is
lots of other evidence for relating some of tehse languages (esp.
what I call the Pakawan group) to each other. I dont know how many
of you have read my work (in Anthro Lg) or agree with it, but that
is not germane to the point at issue here, namely, that even the
complete failure of someone's arguments for a given lg family
only means that THAT case has failed.  I would of course argue that
Altaic is a good example of this too.  The early work on relating
these languages (e.g., Halevy) was often a joke, but I would argue
that the more recent work establishes their relatedness.  An even
clearer example of the same is Uto-Aztecan, which is clearer simply
because no one questions it now (not even the people like Doerfer or
Campbell who deny Altaic or Pakawan).
 
AMR