Print

Print


On 6 Jun 2015 17:27, "Jeffrey Brown" <[log in to unmask]> wrote:
>
> Alex,
> By "history" I meant "political history."
> What I am saying is that the statistics compiled by WALS are too
confounded
> by political factors, as well as geographical/topographical factors, to be
> a useful indicator for language construction. What I would like to have is
> an understanding of linguistic change mediated *solely* by
> anatomical/physiological/neurological factors. In the meantime, the
> confounding of the sources of diachrony makes me believe that the best one
> can do with the WALS data is simple binary ascertainment: Does this
feature
> exists in a natlang (yes/no)? It is, as you say, a very coarse measure. I
> do not believe, though, that the data is usable as a metric for more.

I don't see significant confounding political factors, as I explained in a
previous message. By geographical/topographical factors are you thinking of
areal influence, and thinking that it will tend to have homogenizing
effects that mean that a random sample of actual languages does bit
actually constitute a random sample of possible natlang grammars? That's
probably true.

A further well-known problem with drawing conclusions about language from
cross-linguistic surveys of known languages is that extant languages
comprise only a small portion of languages that have existed. Perhaps that
doesn't matter when trying to assess the characteristics of the typical
natlang, since once areal effects have been controlled for there remains a
large enough pool of languages to base the generalizations on. But in the
present discussion of naturalism in conlangs we were wanting to distinguish
between the impossible and the merely rare: it's that distinction that I
think for many features we can never be in a position to make using
statistical methods.

And