Universality of GR's

[Avery Andrews]

Following up on Chris Manning's point, I think another question about the
languages that seem to lack GR's is whether they also seem to lack anything
like a `core/oblique' distinction as presented for example in my 1985
paper on the functions of NPs in the Shopen typology volumes.

In some recent work in LFG-like frameworks (Alsina, Manning, Kroeger,
Wechsler & Arka) the main divisions are between core (or term) on one
hand and oblique on the other, with a `pivot' picking out one of the
terms for special treatment.  If some languages just lack the pivot, then
that might produce a lot of big differences in overt typology without
the formal architecture actually being so different (semantic roles
are for example used to discriminate between multiple non-pivot
terms in these newer approaches, so it's not really that different
if they are discriminating between all terms, in the absence of a pivot).

But if there's no term/oblique distinction, then these languages really
do look much more profoundly different, if one takes their present
descriptions at face-value.  Of course identifying this distinction and
justifying its presence might well be a challenge.  An interesting case,
if I recall it correctly, is Acehnese, as described by Mark Durie.
Here there does seem to be a strong term/oblique distinction, with the
terms able to appear in an optional preverbal position (which looks sort
of like a pivot but doesn't seem to trigger any interesting grammatical
behavior), but the morphosyntactic organization of the language seems
to be `Active', with the big distinction being between Agentlike
arguments (Dixonian A/S_a) and Patient-like ones (Dixonian O/S_a).
This distinction is crystal clear in the morphology, while the syntax
basically seems to be `semantic' as described for the putative languages
without grammatical relations.

  Avery Andrews