Relations that are seldom or never signale

[Bill Mann]

Dear RST listers:

In this message I just want to respond to a couple of ideas from Daniel Marcu's
interesting message.

I agree that the notion of multiple senses applies to "discourse markers."
Beyond that, the function of discourse marking can certainly also be done using
the open class items, phrases or even clauses that convey  relational status.

Focusing seems to affect ordering of spans, which RST does not restrict.

I don't think RST's past concentration on small spans is anything more than
conveince and management of effort.  In the early development, Barbara Fox
analyzed whole Scientific American articles.  The website has analyses where the
initial units are orthographic *paragraphs,* and it has analyses of texts of
over 30 units, with no wall in sight.

The reason for concentrating on small things in the early work was partly to
establish the idea that the interclausal relations are, largely or entirely, the
discourse strucural relations of this sort.  So it is more history than
principle.

Bill Mann