Peripheral promotes "common innovation" ?

[ECOLING at aol.com]

[Sent to IE list on 25 January, 2000]

I have quite a number of questions regarding what may be
relatively unquestioned assumptions underlying 
decisons about what qualify as "common innovations".

***

One of these questions I raised in a separate message sent today
on what Sapir's concept of "Drift" does for us,
if major changes are taken as going through a number of stages
in the process of "drift".  Can we distinguish clearly between
dialect areas and branches in family trees (not confusing dialect areas
which have been constantly contiguous with Sprachbünde which
result from later contact, or contact of languages not closely related).

As a consequence of that perspective...

Is it perhaps the case that innovations which are restricted
to a very narrow dialect area are more likely to be considered
common innovations justifying a branch on a family tree,
while innovations which covered a much wider area,
and are perhaps much more fundamental (!),
*BECAUSE WE CAN DETECT THE TIME THEY TAKE
FROM THEIR ORIGIN AT THEIR CORE UNTIL THEY 
REACH THEIR FARTHEST EXTENTS*
will be classed as borrowings of tendencies between related
neighboring languages, and thus excluded from consideration for
family-tree groupings?

***

Another has to do with how geographic position, 
central vs. peripheral in a dialect-area network,
may possibly affect judgements about common innovations.

I'm wondering whether, methodologically, the effects of contiguity
in dialect areas tend to bias the Penn. approach, and perhaps many others,
to indicating an early separation (separate branches) 
of the dialect groups which are at the geographical extremes of the total 
composite dialect space?  (Thinking perhaps of Italic & Celtic.)

That is, because the groups at the edges do not have genetically
related dialects on all sides of them, matters which were really
dialect-area phenomena, interpreted as common innovations on
nodes on a tree, will tend to segment off the outermost areas first?
That is not necessarily a wrong result, from some point of view,
but it may be a much too discrete result, as opposed to a more
accurate result which might have to be gradient.  

More generally, I would love to receive references to the most recent
productive thinking on how to combine dialect-area perspectives
with branching-tree perspectives.

To me, branching is most fully secure only when LATER borrowings
clearly postdate some internal change, and do not show the effect
of that change.  That is asking more of the evidence than merely
common innovation, I think, but not sure how to state this
with absolute precision.

***

I think these are each important questions of how method interacts 
with the content which the method is applied to.  
(In statistics, at least, it is known
that all methods are appropriate or inappropriate depending on the 
structure of the data they are applied to, etc. etc.,
that they may give wrong results when applied to the wrong 
kind of data.)

***

Lloyd Anderson
Ecological Linguistics