Focus and Contrastive Topic in a Complex Model of Discourse
			Daniel Büring, UCSC

     In this talk I will present a novel theory of 'contrastive
topic' -- as marked by pitch accents in languages like
English, German, and Dutch -- and its relation to Focus. I
submit that both focus and contrastive topic are best
characterized by means of pragmatic conditions on their use,
and that these conditions are crucially different for both
categories.

     There is a broad consensus that the grammatical category
focus is related to the pragmatic notion of 'newness' (or, as
I will assume, following Schwarzschild t.a., the other way
around: non-focus is related to givenness). No such consensus
exists w.r.t. contrastive topics. While some authors regard
them as essentially old, e.g. sub-sets of the background, old
foci or links to the discourse file, others take them to be
essentially new, e.g. shifters of topics, markers of deviance
from a question, or focus within the topic/theme.

     The theory I propose holds that the distinction between
new and old (or between the corresponding grammatical
categories focus and background) is irrelevant to characterize
contrastive topics. The function of contrastive topics is to
indicate a certain abstract configuration of discourse
structure, a 'strategy'. Strategies are structured sets of
questions and answers, devoted to resolving a common super-
question. Contrastive topic may be found in virtually any
assertive move within a strategy: the last, the first, some
intermediary; it therefore may relate to previous moves,
following moves, or both at the same time. This is what lead
to the apparently contradictory positions alluded to above.

     Building on the framework proposed in Roberts (1996) I
will introduce 'discourse trees' as a particularly well-suited
way of representing the kind of abstract structure needed to
make the description of contrastive topic precise. Both a
Schwarzschildian notion of focus and one of contrastive topic
along the lines proposed can be formulated w.r.t. discourse
trees and be shown to interact in interesting ways.