102 John Connolly
which an utterance is produced. According to Devlin, a description of this
nature will include the following information: (i) who is addressing whom,
when and where, (ii) what is being uttered and (iii) which individuals are
being referred to by means of which lexical items; see Devlin (1991: 237).
Full details of the relevant ST-style notation need not be considered here,
but a rough indication of what is involved in the description of an assertive
speech act is the following:
(9) << speaking-to, Speaker, Addressee, Location, Time, true >>
and
<< utters, Speaker, Proposition, Location, Time, true >>
and
<< refers-to, Speaker, word 1, Referent 1, word 2, Referent 2, ...,
Location, Time, true >>.
Clearly, this is a more elaborate characterisation of a speech act than is
conventional in FG, and thus merits our attention in the present context.
- A framework for discourse representation
From what has been said so far, it is clear that the representation of dis-
course involves coping with a large amount of information. We are also
faced with two conflicting goals. On the one hand, we would wish our rep-
resentation to provide an adequate coverage of discourse phenomena. On
the other hand, we would want our notation to be easy to read, as well as
being straightforward and quick to produce. What is presented below is a
compromise between these desiderata.
The problem of representation is complicated by the fact that discourse
has more than two dimensions. We need to accommodate temporal (or else
spatial) sequencing, hierarchical structure, different levels of analysis and
also relational phenomena (in particular adjacency sequences and rhetorical
relations). All these need to be projected onto the two-dimensional page.
It is fairly natural to represent temporal sequencing as progressing verti-
cally down the page, and it is convenient to organize the levels of analysis
on the horizontal dimension across the page, with each level being con-
tained in a separate column, while maintaining a horizontal alignment
between entries which correspond to one another. In fact, it is proposed
that we employ a series of columns, to accommodate each of the following: