On 2008-12-04 23:10, Arnaud Fournet wrote:
> What is this model capable of predicting what is not already known by other
> means ?
It makes quite a few non-trivial predictions. For example, that
high-frequency words will be universally conserved (resistant to lexical
replacement) but at the same time they will evolve phonetically much
faster than the bulk of the lexicon. I don't know of any other model
that can reconcile the general regularity of sound change (i.e. the
group behaviour of lexical sets) with the observation that "chaque mot a
son histoire" in a natural way. Those individual histories are the more
unique, the greater the success of the word as a replicator (because it
has more opportunities to "mutate").
It can also be predicted that traditional lexicostatistic calculations,
and especially glottochronological conclusions based on them, will
consistently prove worthless, since the rate of lexical replacement in
the "core" vocabulary is _extremely_ uneven (the expected lifetimes,
even among the "Swadesh words", depend strongly on their frequency of
occurrence and typically differ by at least two orders of magnitude).
Come to think of it, the very fact that linguistic variation and change
are inevitable falls out from the model. Of course that's something
everybody knows, but how do other models account for this "already
known" observation?
Piotr