Along with colleagues at Penn (Aravind Joshi), Otago (Ali Knott),and
Rutgers (Matthew Stone), I have been investigating a weak, low-level
syntax for discourse that closely resembles clause-level syntax.
The scope of this level of discourse are multi-clause descriptions
of entities such as eventualities and situations (but also individuals
and sets).
A notable feature of this work is that the mapping of ``discourse
syntax'' to discourse semantics uses only the same mechanisms as are
needed at the clause-level syntactic-semantic interface -- primarily,
* A straight-forward compositional semantics defined on the elements
of this ``bare bones'' discourse structure and the operations used to
assemble them together;
* Inference based on world knowledge, usage conventions, etc., that
makes defeasible contributions to discourse interpretation that go
beyond the non-defeasible propositions contributed by compositional
semantics.
* Non-structural anaphoric presuppositions that can provide (a)
arguments for asserted first-order and higher-order semantic
relations, and (b) modal operators on propositions and eventualities,
based on information that speaker and hearer are taken to share.
Having these three mechanisms means that a rather simple ``syntactic
structure'' for discourse can nevertheless give rise to a much
more complex semantics (and semantic structure). This relieves
the structural analysis of a discourse from the full burden of
conveying discourse relations, by seeing many of them as arising
non-structurally from the grounding of anaphoric
presuppositions
********************************************************