"Pan-Americanisms"

[bwald]

----------------------------Original message----------------------------
I cannot comment definitively on Miguel Carrasquer Vidal's question about
Lyle Campbell's
>"American Indian Languages.  The Historical Linguistics of Native
>America" (Oxford University Press, 1997), since I have not even seen, much
>less read that book.  Therefore, I don't know how coherently its entire
>argument hangs together.
 
However, I thought there was an interesting problem in logic involved, and
as far as Miguel's discussion went, I thought his question was worth asking
and thinking about.  He ends up by asking:
 
>... Certainly, the
>"pan-Americanism" argument, as explained above, can be used against
>each and every one of these 23 forms individually, but can it be used
>against the 23 as a whole?  That is my worry.   English "two" and
>German "zwei" cannot individually be used as an argument for a close
>relationship between these two languages, because we are clearly
>dealing with "pan-Indo-Europeanisms".  But what about a block of 23
>"pan-Indo-Europeanisms" found as a group in Germanic and Germanic
>only?
 
This is in regard to a *set* of 23 Yana-Karuk (Hokan) comparisons by Mary
Haas, which Miguel reports Campbell as rejecting as evidence of a *special*
relationship between Yana and Karuk, on the basis that the forms (whatever
they are) are widespread enough among Amerind languages outside of these
groups to be set aside as "Pan-Americanisms", and therefore not evidence of
any more of a special relationship between Yana and Karuk than of either of
these two and any other Amerind groups.
 
As best as I can understand, Miguel is supposing a situation in which these
23 Pan-Americanisms are scattered amongst numerous Amerind groups, so that,
say, only Yana and Karuk share the SAME 23, even though other far-flung
Amerind groups show that each of them is more widespread, and *if* all
relevant Amerind languages were indeed genetically related, then only Yana
and Karuk preserve these particular 23 *as a set* ... whereas another group
A maybe preserves 12 of them, and another group B 15 of them (maybe
overlapping with A for 6 of them, or whatever).  I think this is what
Miguel means, but his parallels in Germanic are not well chosen.  The
"two"/"zwei" set for a *special* German/English relationship (i.e.,
Germanic) is indeed found just about everywhere in Indo-European.  He wants
something like German and English have 23 examples attributable to
Indo-European, but, say, only 4 of them are found in Tokharian, only 7 are
found in Greek, and only 12 in Albanian, and so on for various permutations
and combinations, until we are sure that all 23 are "Proto-IE".
 
The general problem then seems to be the likelihood of a relatively large
number of conservative (NOT innovative) forms remaining in two languages
(or groups) WITHOUT those groups having a special historical relationship
to each other.  The alternative to "special relation" (i.e., innovation,
which is precluded by the problem) is that it is just an accident that such
a set remains in those languages/groups, but not in any other (related)
languages/groups.
 
In principle (or in the *absence* of further principles of word loss), it
does not seem UNlikely to me that such a thing might happen.  That is, it
could happen accidentally.  For example, let's take 4 language groups, A,
B, C and D related to each other though descent from Proto-Z, along with F
through X.  So at the Proto-Z stage there is a single lexicon.  Then C and
D innovate for 23 words (over time or what-have-you).  A and B don't.  So A
and B now share 23 Proto-Z words by virtue of not being in the "loop" of C
and D.  But F-X also contain those 23 words so nothing special about A and
B.  But over more time F-X do their own thing, and lose some of those
words, F loses 3 of them, G 4 of them, etc.  F-X just preserve enough of
them (a few each) to insure that we can recognize the set of 23 in A and B
as Proto-Z, i.e., "pan-Zisms".  Of course, A and B are also losing (or
replacing) pan-Zisms too, but other ones, not these 23.
 
In the eventual outcome I would expect A and B to not only share 23
pan-Zisms, but also for A and C, but not B, to share a comparable number of
pan-Zisms, and so on.  There should not be a unique special relation
between A and B if what they share is accidental; there should also be a
comparable "special relation" between A and some other group to the
exclusion of B, and so on.  Thus, there is no unambiguous special
relationship between A and B, because if there were, there would also be a
special relationship between A and C to the exclusion of B.  A tree model
cannot allow that, and Miguel seems to be assuming a tree model.  I assume,
but may be wrong, that Campbell is criticising such a tree model for
Yana-Karuk relationship (on the basis of such evidence).  If so, he should
be able to complete his argument by showing that Yana but not Karuk (or
vice-versa) shares another large number of pan-Americanisms with another
Amerind group.  Maybe he can't do it for Amerind because there are not
enough pan-Americanisms identified at this point.  Then can he do it for
three branches of Indo-European?  I would suppose so.  I would find it very
interesting if it turns out it cannot be done, and would wonder why.  After
all, English has lost IE words that have remained in German, and no doubt
the other way around too.  What is there to prevent such things from
happening?