From: Brian M. Scott
Message: 66957
Date: 2010-12-14
>>> As for how to do it, if you're at a loss of methodology:See, for example, Ann Kumar & Phil Rose, 'Lexical Evidence
>>> you could demonstrate regular derivational relationships
>>> among these words. Or failing this, you could calculate
>>> how many different words with semantics as similar as
>>> this, and a shared structure as similar as
>>> *l-and-medial-nasal-or-labial, we statistically expect
>>> to find in the languages you are taking into account,
>>> and to sho that that number is much smaller than the
>>> number of forms you listed considered to be separate.
>> Nobody does that, so I won't.
> All historical linguistics is based on regularly deriving
> words from one another. Inference from statistical
> properties of lexicons is indeed rarer (I've seen some
> examples, none of the precise scope I suggested there),
> but that doesn't render it an invalid method.