Alexis' methodological points (was: Linguistic classification)

[Larry Trask]

----------------------------Original message----------------------------
Stefan Georg writes:
 
> NOW: before Alexis slaughters me: of course, a scenario is
> thinkable, where original relationship (i.e. *identity*) first led
> to dissolution of an original parent language (*increasing*
> "differences"), and *later* convergent processes might have
> happened, which have led to the observable processes (after all, the
> Balkans area is made up of languages, which are related).
 
Indeed, and this scenario is not rare at all.
 
Iceland was chiefly settled by Norwegians, and so we might expect that
a family tree should show Norwegian as closest to Icelandic.  But most
family trees don't: instead, they show Norwegian as most closely
related to Danish and Swedish.  That's because, since the settlement,
Norwegian has both diverged from Icelandic and converged strongly with
its continental neighbors.  Hence a typical family trees displays the
result of convergence, and not the historical relationships.
 
Likewise, the Valencian variety of Catalan has been converging
strongly upon Castilian Spanish for generations, and in many respects
is now more similar to Castilian than to its historically closer
relative, the Catalan of Barcelona.  A few more generations of this,
and we might start revising our family trees to reflect the new
reality.
 
[snip]
 
> But, *given* all those methodological/epistemological precautions -
> and I feel they are pretty trivial - should we not be allowed to
> call A and B "unrelated" (i.e. of course *heuristically* unrelated
> *for the time being*) as long as we are not convinced of the merits
> of any claim to the contrary which happens to be on the market ?
 
I for one say "yes".  Hock and Joseph prefer "unrelatable" to
"unrelated", and I myself often write "not discoverably related".  But
why shouldn't we just say "unrelated" and be aware of what we mean?
 
> When some non-linguist asks me what Basque might be related to, I
> invariably answer "to nothing else", since that is the state of the
> art as of now. I'm not claiming at the same time that this is
> *impossible* to change some day. Should I really enumerate all the
> hypotheses re: Basque, labelling them as "yet unproven", if asked
> such a question ?
 
Amen.
 
[snip]
 
Larry Trask
COGS
University of Sussex
Brighton BN1 9QH
UK
 
larryt at cogs.susx.ac.uk