One Step at a Time

[X99Lynx at aol.com]

In a message dated 7/20/2001 8:29:01 AM, alrivera at southern.edu writes:

<< Suppose the parent of a "half-a-dozen languages" was a language that had two
different classes of verbs:  those from a language X that conjugated one way
and those from a language Y that conjugated another way.  (I think there was a
Greek dialect of Arabic that had a feature like this mentioned on the list
earlier?)>>

In a message dated 7/20/2001 9:05:14 AM, larryt at cogs.susx.ac.uk replied:

<<What Muke is asking about is a *single* language, with two classes of
verbs, which gives rise to some daughters.>>

Then why did Muke write "those from a language X that conjugated one way and
those from a language Y that conjugated another way?"  The point I believe of
the hypothetical was two lines of genetic descent.

And the key here is the assumption that there is only one ancestor language
somewhere along the line.  That assumption is NOT justified by the
comparative method, which yields only ancestor forms.  The idea that there
has to be one "parent" somewhere comes from a different theory.

If two languages can each contribute genetically distinct classes of verbs to
one "daughter" language, then they can each contribute those classes to
multiple "daughter" languages.  There does not have to be a single language
intervening, only original unitary ancestor forms coming from different
earlier languages.

<< The comparative method has nothing to say about the origin of a
proto-language, unless there are yet further languages that can be compared
with it normally.>>

But that's not true, for a number of reasons.  One is that the "further
languages" may be forced into a single parent based on a particular method of
family tree construction.

In fact, when the comparative method has "nothing to say" about the origin of
certain forms, the tree model has often taken over and assigned those forms
to the parent in any case.  So that Prof Trask's "further language" will
actually not be recognized as evidence of "the origins of the
proto-language."

If some of the "daughter" languages don't have one or the other verb classes,
the assumption is often that those cases were "lost."  All you need is three
or four IE languages with both verb cases to assume that *PIE had both.  So
those "further languages" - which might only have had one class and point to
multiple ancestry - may be of no consequence.

(I once counted in Ringe's sample grid of IE "nodes" that nearly a third of
all the individual forms he used to construct his tree were marked "lost" -
meaning that they were not attested but were assumed to be in *PIE.  I'm not
positive but I think that if this assumption were not made Ringe may have
gotten a very different result.  I'm pretty sure that "parsimony" - the best
match for the data - would have created a tree that looked more like a street
map of lower Manhattan than the neat Christmas fir he got.  And this even
despite Ringe's strict screening for only genetic etymologies that were
"certain.")

What is interesting again here is that Prof Trask assumes that the "further
languages" he mentions come later in the analysis.  This is again the cart
before the horse.  Unless one knows the conclusions ahead of time, there
should be no different treatment between the original set and the "further
languages."  They don't come in afterwards - logically they are part of the
original analysis.

In that case, when the "further languages" are included right from the start,
multiple ancestors - according to Prof Trask - could be identified.  But not,
of course, if a single parent is expected.  Once a single parent is assumed,
all indications of multiple ancestry are considered innovations, borrowing,
lost, etc.

It is the assumption of single parentage that turns "other" genetic strains
into non-genetic forms.  This is the circular effect that can erase the
possibility of multple ancestors, even if it is there.  But the comparative
method itself might possibly support EITHER conclusion.

Regards,
Steve Long