Trivial Truths and Genetic "Patterns"

[Rich Alderson]

On 23 Jun 2001, Steve Long wrote:

> In a message dated 6/22/2001 12:44:17 AM, hstahlke at gw.bsu.edu writes:

>> There's an almost trivial sense in which Steve has to be right, but I think
>> he's taken his case beyond that point.  Having worked on Niger-Congo
>> languages, especially those of the eastern half of West Africa, I've been
>> faced with the question of where to begin....  As Bill Welmers used to say,
>> "You get to the point where you know that these language can't be
>> unrelated."  Of course, he would also add that you then start using the
>> comparative method to work out the relationships and make sure they're
>> there.  We're dealing with different orders of hypothesis.  Using some
>> careful lexicostatistics gives you a reasonable hypothesis, but then
>> applying the comparative method takes you to a much stronger one.  As I
>> said, this is an almost trivially obvious point.

> Well, of course, my question was, at that point, did you ever consider the
> hypothesis that the language had two ancestral language groups to whom it
> "can't be unrelated."

You aren't reading what Herb Stahlke wrote.

The hypothesis is not that "this language is related to ancestral group X", but
rather "these two/three/four/dozen languages are related to each other".  That
makes a very big difference.

Hypotheses of relationship are made among existing languages (or reconstructed
protolanguages once those are available), not between a single language and a
possible ancestor.  Once the comparative method has been employed on data drawn
from a group of languages--and the canonical number is *NOT* 2--and a proto-
language reconstructed (even in part), the latter can be used as a *shortcut*
in examining data from other languages that it is thought may be related, but
it *IS* a *SHORTCUT* for comparing the new data from these languages to all the
data from all the previous languages used to build the protolanguage.

And once again, the hypothesis is not that "this language is related to the
reconstructed ancestral group X", but rather "these two/three/four/dozen
languages are related to each other, and to the two/three/four/dozen already
examined in the same way previously".

So that, I think, is the pithy answer to your continued asking of why any
language can't have more than one ancestor under the comparative method:  The
method does not address single languages, but groups of languages, and as was
pointed out by Larry Trask, if we can't build a protolanguage with the compara-
tive method, the languages in question are *JUST* *NOT* *RELATED*.

> As you were kind enough to concede, you have to see patterns BEFORE you even
> apply the comparative method.  Heck, borrowing creates patterns.  "Patterns"
> therefore CANNOT explain the difference between genetic and non-genetic
> relationships.

Borrowing produces patterns that are qualitatively different from those that
arise from genetic relationship.  The concept of patterning alone is necessary
but not sufficient to define genetic relationship; the *kinds* of pattern are
the determining factor.

								Rich Alderson