Urheimat in Lithuania? (was Re: the Wheel and Dating PIE or NW-IE)

[Robert Whiting]

On Tue, 28 Mar, petegray <petegray at btinternet.com> wrote:

[quoting Brian M. Scott]

>> I fell into the same trap at first, but Bob's right, assuming
>> that one can always classify one branch as innovating and the
>> other as non-innovating.  Start at the root, and at each node
>> follow the non-innovating branch; you *must* end up at some
>> leaf of the tree.  It is equally true that if you always follow
>> the innovating branch you will arrive at some leaf.  Which leaves
>> these are is of interest; that there are such leaves is not.

>Yes - you're right!

>Does this raise questions about the validity of this particular
>tree structure?    How far does this mirror reality - are we
>really saying that in any dialect group there must of necessity
>be one dialect that never innovates?   That's what the tree - in
>this form - implies, and it is clearly untrue of real life.   Or
>is it only a result of the fact that we select certain
>innovations (those that create distinctions) and ignore others
>(those within a group already distinct)?

Yes, he is right, but he is also not right (and since he was
defending a position that I had taken originally, I was also both
right and not right).  He is right if you only consider the
behavior of the tree at the nodes.  But he is not right if you
cansider what may happen along the branches as well.  In my
original statement of the process, I conflated two steps.

While a tree node typically represents a binary opposition
between innovation and non-innovation, there is nothing that
requires that the non-innovating branch coming out of the node
has to be the least innovative branch by the time that the next
node is reached or even at the end of the branch if it never
branches again.  I think that this is the point that you were
trying to raise with your original post on this matter.  So you
(meaning I) cannot simply use the non-innovating branch out of a
node to determine which is ultimately the least innovative branch
and this is the point that I did not make clear originally.  The
example of Lithuanian itself exemplifies this since it was at one
point on the innovative branch of a node.  Now it would be
possible to design a tree so that Lithuanian would always be on
the non-innovating branch (this would mean that you couldn't use
palatal assibilation or RUKI palatization to define a node), but
this would be manipulating the data to produce the desired result;
the exact equivalent of using binary choices to force a card on
the subject in a magic trick.

Therefore it is not true that there must of necessity be one
dialect that never innovates.  In fact, it is extemely unlikely
that there would ever be a dialect that never innovates.  If
there is, it is not a living language.  But unless there is some
law that says that all languages have to innovate exactly the
same amount over time, there must be a minimally innovated
dialect.  I do not think that there is such a law; if there were,
glottochronolgy would work.  In the absence of such a law then
dialects are free to innovate as they will, and, barring an
astronomical amount of coincidence, some dialect will innovate
less than all the others.

It is my contention that this least innovative dialect will be at
the end of the path through the tree from the top to the bottom
that has the lowest total number of innovations on it.  Again, it
is my contention that, unless all languages must innovate to
exactly the same extent, there must be such a minimal path
through the tree.  Saying that this path can be found by
following the non-innovating path from each node is not strictly
speaking true (It might be, but it doesn't have to be).  It may
not even be on the least innovative branch between one node and
the next.  But if you add up all the innovations (perhaps as
number of rule applications necessary to derive new forms from the
old) for each path through the tree, one of these paths will be
minimal.

Yes, trees do not reflect real life.  Only part of our
information can be displayed with a tree.  The tree will look
different depending on which part of our information we choose
to display.  Nodes are customarily defined by innovations found
on one branch out of the node and not on the other since it is
only shared innovations that define linguistic groupings within
a larger group.  Innovations that do not result in branching are
generally not accounted for in trees.  But trees are still useful
for visualizing certain relationships even if they do not
correspond to reality (rather like the planetary model of the
atom).

Bob Whiting
whiting at cc.helsinki.fi