Pan-Americanisms

[Miguel Carrasquer Vidal]

----------------------------Original message----------------------------
Lyle Campbell <l.campbell at ling.canterbury.ac.nz> wrote:
 
>        Let me start by repeating that the term is not mine; I read it
>somewhere as a student (I thought in writings of Swadesh's), but have never
>been able to find it again.  I'd be grateful to anyone who could (re)locate
>the source of the term.
 
Yes, sorry, I got that wrong in my short resume (and I *had* read the
footnote where you say so).
 
>        You ask if a block of 23 "pan-Indo-Europeanisms" found in Germanic
>as a group would be significant (for demonstrating a closer relationship)?
>The answer, I would say, has to do with the fact that in IE and in the
>Americas we are on rather different ground, or, better said, there are more
>straightforward considerations and more convoluted considerations, and each
>is important to the overall stance that should be taken in answering a
>question such as yours.  The answer would depend on the nature of the 23
>things compared and on the nature of the languages in which they are
>compared (though probably "no" in both instances).  Let's take up the
>nature of the items compared first.
>        The so-called pan-Americanisms are a pretty messy, shaggy, muddled
>lot (probably of something between 25 and 100 forms, depending on who is
>looking);
 
There was a tacit assumption in my first message, that I'm glad to
see confirmed here, namely that the number of "pan-Americanisms" is a
rather small one.  As a matter of fact, I had originally ended my
message saying "But what about a block of 23 (or rather: 230)
"pan-Indo-Europeanisms" found as a group in Germanic and Germanic
only?", but in the end I changed my mind and deleted the
parenthetical remark because it sounded cryptic without further
explanation, and because I didn't know for a fact whether the number
of pan-Americanisms was indeed roughly a factor 10 smaller than the
number of IE roots (1000+ in Pokorny).
 
If we pretend for a minute that "pan-Americanisms" can be treated the
same way as "pan-Indo-Europeanisms", i.e. if we pretend they are
cognates at some level, then I think it would be statistically
justified to compare a block of 23 pan-Americanism in Hokan with a
block of 230 pan-Indo-Europeanisms in Germanic.  I am aware of the
rule which says that "shared retentions are not valid evidence for
subgrouping; only shared innovations provide support" ("American
Indian Languages", p. 258 [which I should have quoted in my resume,
because it's a central point in the argument]), and which is also the
point Benji Wald was making in his reaction to my message.  In my
view, this is only a rule of thumb, and in exceptional circumstances
it can safely be ignored.  I think that if we did find a group of
languages which has retained the same 25% or more of the original
vocabulary, that should be taken as strong indication that the group
is closely related (in the same sub-group).  Of course, there is
usually no need to do so, because languages that closely related will
show a large amount of other evidence for their close relationship.
 
But, as Mark Rosenfelder says in his excellent Web article "How
likely are chance resemblances between languages?"
<http://www.tezcat.com/~markrose/chance.htm>: "The brain is no good
at probabilities" (under the heading: "Why are we so easy to fool?").
I know my brain is no good at them, so maybe I'm totally wrong about
my equation of the 23 Hokan with 230 Germanic look-alikes.
 
 
As to the nature of the pan-Americanisms themselves, Lyle's
characterisation of them as a "messy, shaggy, muddled lot" (and
further explanation, deleted for brevity) casts doubts on whether we
*can* compare them with "pan-Indo-Europeanisms" (i.e. cognates) even
in principle.  Maybe it would be better to avoid the term until
somebody figures what they really are, which is probably best done
anyway by treating each item individually and not worry about them as
a whole.
 
In that sense, I'm still not happy with "pan-Americanism" as a reason
for dismissal of items in Haas' list: if (and only if) regular sound
correspondences can be established between two or more languages,
based on sufficient and valid data [and that's what I gather is the
real weakness of Haas' list: too few illustrations of the proposed
sound-correspondences, and some of them clearly onomatopoeic or
borrowed], then the fact that for some of the words wide-spread
look-alikes exist is not in itself reason to reject those forms.
After all, we don't reject PIE *swek^s and *septm just because they
are Wanderwoerter (or possibly borrowings into PIE itself), occurring
all over the place (Semitic, Egyptian, Berber, Kartvelian, Etruscan,
Basque, i.e. "pan-Mediterraneanisms").  [I should add that I feel the
same in principle about another reason for dismissal in that
paragraph, viz. "little phonetic similarity": if the sound law is
real, it shouldn't matter how dissimilar the reflexes are (I feel
tradition-bound to insert a reference to Armenian <erku> here)].
 
 
=======================
Miguel Carrasquer Vidal
mcv at wxs.nl
Amsterdam