Features (was: empirical basis, typology, and gradience)

[Mike Maxwell]

Raul Aranovich writes:

>...if one can introduce a new feature to account
>for the data 'on demand', so to speak,  without
>a clear conception of what is language-particular
>and what is language-universal, then the perception
>may be that the analysis is only aimed at describing
>the phenomenon.

Speaking of this--It seems obvious to me that one has to be able to
introduce new features (e.g. to account for different classifier systems in
Mayan languages, Amazonian languages, etc.).  But I would also be surprised
if there weren't universals.  There's been work on universals of
morphosyntactic features in other frameworks: Rolf Noyer's work on person
and number marking features, for instance.  Halliday also did some stuff a
long time ago, although I don't recall whether he claimed universality for
the features he proposed, or whether that was even an issue for him.

Can anyone point me to work on universals (or non-) of morphosyntactic
feature systems in the HPGS (or GPSG) framework?  Or in any other
*generative* framework, for that matter.

Of course typology work on tense, aspect, etc. has a bearing on this, but
I'm really looking for s.t. more "generative", to use that term in its broad
sense.

                                         Mike Maxwell
                                         Summer Institute of Linguistics
                                         Mike_Maxwell at sil.org