On Wed, 15 Jul 2009 05:51:53 -0700, Lee <waywardwretch@...> wrote:

>The more I think about whether individual phonemes are
over-/underrepresented in conlangs, the more I wonder whether a better gauge
to measure against would be usage-based, rather natlang-count-based. IOW,
rather than treating each natlang as though each has an equal number of
speakers, weight each according to number of speakers.

I'm afraid I don't understand the point of that, unless you had auxlang-type
criteria and were after a set of phones that the most people were
comfortable with. As typology it seems, well, misguided. As I see things,
it's essentially an axiom of typological studies that each individual
language is a sample picked from some single hypothetical probability
distribution of human languages which we're trying to figure out. From that
perspective it only biases your estimators to weight by number of speakers,
much as it would to, say, weight people by net worth when trying to compute
the mean human height or something.

Alex