Relations that are seldom or never signaled

[Alistair Knott]

Here's a response to the message Bill Mann sent on Friday.

1) Bill suggests that it's possible that (a) an RST relation holds between
two text spans, (b) that a connective also occurs between these spans, and
yet that (c) the connective doesn't signal the relation. The example he
gives is the one given by John Bateman:

> (1)     Look, you'll just have to go with him to the store.
> (2)     Since his foot is broken, he can't go by himself.

John and Bill both suggest that the relation here is BACKGROUND; Bill
explicitly notes that there is a relation of logical consequence being
signalled by `since' too; I guess that John has a similar story in mind.

This seems to cut to the heart of the matter. The idea being suggested is
a familiar one: that relations apply between text spans at different
levels, and that BACKGROUND holds at one level, and logical consequence at
another level.  You could imagine fleshing this idea out in different
ways, for instance using the ideational, interpersonal and textual levels
advocated in systemic linguistics, or using Moore and Pollack's
informational and intentional levels. The idea of levels isn't explicit in
classical RST, as far as I can tell: there is a distinction between
subject-matter and presentational relations, but these are set up as
alternatives to one another, not as relations that can apply
simultaneously. (I think Moore and Pollack's point is roughly that they
should apply simultaneously.)

I agree that as soon as multiple levels of relations are introduced, the
use of connectives as a diagnostic for the presence of relations becomes
much more complicated; we have to ask ourselves which level the connective
is signalling a relation at, and I guess it might even be that it's
signalling a complex situation that relates to more than one level. I
still feel that parsimony compels us to explore the idea that we can
account for the coherence of a text without having to posit multiple
levels: the idea of multiple levels is a very large and expensive piece of
theoretical machinery, and we should be very sure we actually need it. In
the above example, for instance, I really don't feel it's necessary. I
think we can imagine a rich definition of the relation of justification
that nets in both the idea of logical consequence and the idea of
`sufficient comprehension' that's at the heart of the BACKGROUND relation.
Of course, one example isn't going to prove the case. But I think that
the debate we've been having on the relationship between connectives and
coherence relations leads naturally into the debate about the need for
multiple levels. Which is another can of worms, but one worth opening!

Ali