Re: About methodology...

Regarding Piotr and Glen's debate on the value of "tree"
versus "tangled bush". I came to recognise the value of the bush
model in the Highlands of Papua New Guinea. The Wiru Language, for
instance had a 55% West-Central Highlands Family feature, and a 45%
Medlpa Family Feature. Was it originally a West Central Highlands
language in which Medlpa arrived as an superstrata or a substrata
Medlpa with a West Central Highlands superstrata.

I suspect there are a lot of languages in such a situation. I know
it is certainly true of Latin America where comparitive linguists are
having huge difficulties in sorting out whether a language is Macro
Ge, Andean-Equatorial or Macro-Carib. I suspect Japanese is a
classic example in another guise. I have seen people who put
Japanese as the first split of Proto-Austronesian. Others consider
it a type of strange Altaic (as does Glen). This kind of thing is
more common than one would consider, especially after a long and
complex history.

For instance English as a huge Latinate adstratum. If we were in the
absence of any other Germanic language surviving, wouldn't it be
possible that a CompLx somewhere would be quite likely to put Modern
English into the Romance Languages as a distant cousin?

Such effects accumulate across the millennia, and the amount of
static noise that would accumulate in any language as a result would
be huge. It is really a question as to where this static becomes so
large that any "underlying message" becomes lost. Piotr argues that
the static comes very soon to the present, Glen is close to the
Greenbergians in arguing it only comes very early. In either case,
the "noise to signal" ratio, is something Glen, that can be best
determined mathematically ;-)

Hope this helps

Regards

John