In my syntax class this year I encouraged students to write essays comparing
WG with some other formal theory.  One suggestion I made was the work of the
mediaeval Arabic grammarians, and the suggestion was followed up in a nice
piece of work by one of the students (perhaps it is not a coincidence that
she is the daughter of a philosopher here).  There are three facets to
this work (done in the Middle Ages remember!) which I think will be of
interest to WGers:

(1) The theory is essentially a dependency grammar with three basic elements
distinguished: the governor, the governed element, and the effect induced by
the governor on the governed element.  In the case of a typical VSO sentence,
the governor is the verb and subject and object are the governed elements.
The induced effects are nominative and accusative case inflections on the
S and O respectively.  The order of dependents is held to be (strictly) 
governor > governed

(2) A problem is posed for this theory of head-dependent order by what are
called nominal sentences.  These are copula-less sentences with order
subject > predicate nominal.  The theoretical solution is to posit a _null
governor_, topic, which is positioned at the head of the sentence.  Since
in nominal sentences, the subject is inflected in the nominative case and
the predicate nominal in the accusative case, the result is the desired
VSO order.

(3) Even more interesting is the following problem and its solution: 
transitive verbal sentences have an alternate order VOS.  This is not
a problem since both dependents are positioned following the governor.
However, nominal sentences occur in inverted order (with the predicate
nominal initial and the subject final).  This is explained by invoking
the null governor of (2) _and_ by developing a formal theory of analogy
in which the inverted order is explained by saying that it is analogous
to the inversion which occurs with VSO sentences (which allow VOS order).

Ignoring the question of whether this is reasonable, it is at the very
least intriguing.  Do any of the currently available syntactic theories
have anything analogous to this notion of sanctioning one construction
type on the basis of its similarity to another which is already sanctioned?
Notice that the Arabic theorists are not saying that [Topic]-PredNom-Subj
sentences are instances of VOS sentences.  They are saying that such
sentences are sanctioned on the basis of their resemblance to VOS sentences.

The only way of doing this in WG that comes to mind is to use multiple
inheritance to allow an "inverted" order.  VOS order and [Topic}-
PredNom-Subj orders would then be instances of inverted sentences.  But
this is ontologically very different from the Arabic grammarians did.

Chet



%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%