Towards a speaker model of FG 357- Conclusions
In our contribution to this volume we have done three things. First, we
gave a short introduction to our dynamic model for the FG expression
rules. Then we looked critically at the Functional Discourse Grammar
model as proposed by Hengeveld (this volume). We made some sugges-
tions for the further elaboration of FDG. Finally, we tried to amalgamate
our slightly adapted version of FDG and the expression component into
one dynamic model of a discourse-oriented FG grammar of the Speaker. A
short example from spoken language served as a rough sketch of the work-
ings of this model.
Most of the problems that we ran into have been solved, at least to some
extent. Many aspects should be worked out in much greater detail, and in
the light of more empirical material, preferably recorded and annotated
versions of informal spoken conversation. We mention just a few of the
major problematic issues that spring to mind.
There is the – old – point of keeping track of the material that has been
expressed, and the related question of which underlying elements of the IL
and RL levels may be expressed more than once. This is of much greater
importance in a dynamic model such as the one given above than in the
traditional grammar model which deals with the ideal speaker-hearer. Re-
lated to this point is the requirement that the model should provide an
answer to what may, or must be repeated after hesitations and self-
corrections, and how corrections affect the eventual representations.
A second point is the precise nature of the interaction between the re-
spective kinds of memory, and the way information is stored there, what is
retrievable, how it may be extended, and under what conditions. A sub-
problem is the way discourse is represented in the speaker’s memory. This,
in turn, calls for a deeper insight into the way speakers monitor their own
speech process. Underlying this is the fundamental, and extremely complex
problem of finding an acceptable way of representing abstract concepts.
Undoubtedly, there are many problems that have been overlooked by us
altogether. These can not be solved without state-of-the-art knowledge of
the intricacies of human mental models of information processing. Fortu-
nately, within the Cognitive Linguistics (CL) enterprise there are many
hundreds of linguists, psychologists and computer scientists working in this
area. CL seems to be one of the obvious scientific arenas FG should turn to
for discussion, suggestions and solutions about these and other matters.
A dynamic speaker model is not only a research goal in its own right.
We are confident that the combined work on such a construct will put the