Trivial Truths and Genetic "Patterns"

Rich Alderson alderson+mail at panix.com
Fri Jul 13 22:49:53 UTC 2001


On Sat, 7 Jul 2001 04:38:26 EDT, Steve Long wrote:

> In a message dated 7/6/01 11:18:00 PM, alderson+mail at panix.com writes:

>> The total set of reconstructed forms, based on all the forms attested in
>> the group of languages under examination, is considered to be a single
>> language.  Period.  By definition.  Period.

> Well, I'm sure if you punctuate it enough, it will surely end up being true.

Not a matter of truth by punctuation, but of emphasizing the nature of the
definition.  No exceptions, no averments.

> Actually, you put yourself in a logical bind here that I'm sure you will
> understand if you consider it carefully and unemotionally.

Actually, it occurred to me that you might take this line of argument or one
very like it after I wrote the message to which you are responding.

> The Lehmann quote very prudently avoids this equation between "forms" and
> "languages."  ("In using the comparative method, we contrast forms of two or
> more related languages to determine the precise relationship between THESE
> FORMS.  We indicate this relationship most simply by reconstructing THE FORMS
> from which they developed." Caps mine.)

> The reason this is wise is that it avoids a conclusion that the comparative
> method in and of does not support in the very example you described earlier.

> In fact, you give a clear example of where the comparative method - applied
> to "a larger set of languages" - will reconstruct the total set of related
> forms into not one, but three different languages.

Not, of course, in the general case.

> In a message titled "Re: The Single Parent Question" dated 7/5/01 3:21:49 PM,
> alderson+mail at panix.com writes:

>> ...if we found a group of languages descended from a prehistoric Michif, we
>> could not reconstruct the two parents thereof.  You are probably correct....
>> from the point of view of the comparative method, these languages do *not*
>> have two parents, but only one, the paleo-Michif.

>> If, on the other hand, we were dealing with a larger set of languages, some
>> descended from the left-hand parent of paleo-Michif; some, the right-parent;
>> and some from paleo-Michif itself, we could (if the events in question were
>> not so far in the past as to prevent our dull tools from working at all) in
>> principle work out the relations of all and sundry.

> Please observe carefully where you've arrived in this last paragraph.  You
> have used the comparative method on languages linked by various systematic
> correspondences.  And the result you forecast is that with the application of
> the method in this situation you will have ended up reconstructing three
> languages - not one.  And one of those reconstructed languages will contain
> what you know are TWO genetically distinct sets of forms.

Please observe very carefully how I arrived at the conclusion in which you take
so much joy:  A very special occurrence, to wit, absolute language mixture,
with recognizable descendants of both parents of this special creole available
for comparison.

I stated that *IN* *THAT* *CASE*, and in that case only, we could proceed back
from the mixed language.  Otherwise, we could only reconstruct back as far as
the one language in question.

> In other words, repeating your words above: << The total set of reconstructed
> forms, based on all the forms attested in the group of languages under
> examination, is considered to be a single language.>> is not the case in the
> example you give.

You've misunderstood.

>From the point of view of the descendants of the mixed language, and them only,
there is only a single parent.

You have missed the entire point of the exercise, which is that the kind of
thing you want to be commonplace is only possible in a very specific instance,
one that is so extremely rare as to make professional linguists disagree on its
reality.

> In fact, the "total set of reconstructed forms" yielded by the application of
> the method in the example you give MUST be three reconstructed languages.
> Because you CANNOT reconstruct the two ancestors accurately UNLESS you
> recognize that there are two distinct sets of systematic correspondences in
> the already reconstructed language "paleo-Mistif."  This means that you are
> using less than "the total set of reconstructed forms" that you've already
> identified in "paleo-Mischif" to also reconstruct the left-handed parent.
> And less than the "total set" to also reconstruct the right-handed parent.

Wrong.  Go back and read the thought experiment again.  The only way to recog-
nize the mixed nature of the one protolanguage is to have reconstructed the two
contributor languages based on their other unmixed descendants.  Without that,
you have no business looking for multiple parents.

> The comparative method triangulates back to a single set of reconstructed
> forms.  It does not triangulate back to a whole language.  And, in the
> example you give, "the total set of reconstructed forms" - multiple
> triangulations - yields three different reconstructed languages.

Not at the same chronological level it doesn't.  The three reconstructions must
proceed from different sets of forms drawn from different languages, or it all
falls apart.

> I understand the comparative method enough to know that you cannot count
> apples and end up with oranges.  What in fact the method does is compare
> forms and find correspondences that show common descent, not whole languages
> or all the possible genetic aspects of a reconstructed language.  In the
> example you give, multiple applications of the method to the same data yield
> multiple reconstructions, each using only a subset of the total reconstructed
> forms.  And so the "total set of reconstructed forms" does not equal one
> language.  The definition you give is not coherent.

There is no "multiple applications of the method to the same data" in this
example.  The only possible way to connect the three protolanguages that are
reconstructed is an after-the-fact examination of the reconstructions, not an
_a priori_ decision that they will turn up if you just hold your mouth right.

> Of course, equating a reconstructed set of forms to a "language" does not
> only create a logical incoherence.  If two languages share noun morphology
> alone (as in the Niger-Kordofanian example), it is patently absurd to call
> the resulting reconstructions a "language."  The processes that created the
> correspondences do not all call for the existence of a third language, even
> if they do establish descent from a common form.  And the comparative method
> establishes descent from a common set of forms, but it can say nothing else
> about the relationship of the remainder of those two languages.  Inferences
> may be drawn.  But that is what they are at best - inferences, not part of
> the method itself.

This only holds true if the "forms" in question are drawn only from a single
subsystem of the languages under consideration, which is an extremely limited
and limiting thing to do.  Suppose that you are only, for whatever reason,
going to deal with "nouns", and one of the languages from which you are drawing
forms for comparison has developed a very strong set of denominative verbs, but
otherwise lost most of its inherited nouns due to influence from an areally
influential language.  You will lose the data from this language by restricting
yourself artificially.

> If this equation of forms to a single language is by definition, a careful
> look at the operation as you yourself described it shows that the definition
> is logically faulty.  And that has nothing to do with me, so personal remarks
> and brow-beating won't change it.

The logic seems to me to hold.  All of the logic.

								Rich Alderson



More information about the Indo-european mailing list