132 Francis Cornish
to them, there is no systematic coding in that language which would single
out expressions bearing the Topic function in relation to other, non-topical
expressions (but see Hannay 1991: 142, fn. 9 for an opposing view). I be-
lieve that this over-reliance on systematic, overt coding as the main
condition determining the existence of particular pragmatic functions in a
language is a reflection of the fact that the assignment of such functions is
standardly assumed to take place within the grammar, as the final specifi-
cation before the complete underlying clause structure representation is
transferred to the expression component for its formal realization. A num-
ber of linguists within FG have argued against this locating of pragmatic
function assignment^16 within the grammar, for a variety of reasons (e.g.
Vet 1998, Bolkestein 1998, Van den Berg 1998). More recently the type of
model presented by Hengeveld (this volume) is conducive to representing
the looser kind of relationship between a pragmatic module and a gram-
matical one which these authors argue for. See also Hannay’s (1991) five
proposed message management modes (to be discussed in Section 4 below)
and their relevance for grammatical coding.
Now, as Siewierska (1991: 148) points out, the subdivision of the Topic
function within FG was designed to relate the specification of clause struc-
ture to the wider discourse setting in which the clause under analysis may
occur. The four Topic statuses recognized in FG (Given, Sub, Resumed
and New) would appear to subsume the distinction in CS between FOCUS
and DEIXIS. Given, Sub and Resumed Topics would realize CS DEIXIS,
and New Topics, FOCUS (i.e. they would correspond to the CS value NOT
IN-FOCUS). FG GivTops, SubTops and ResTops clearly form a scale of
deicticity, in the CS sense of the term, GivTops retrieving obvious refer-
ents, where no competition is involved between referents of the same type:
examples used so far in this chapter are it in the second sentence of exam-
ple (10), and elle in (11) (where this pronoun is unaccented).
SubTops^17 require the drawing of a minimal inference from a contextu-
ally given set of entities to the existence of a particular member of that set,
or from a whole to one of its parts, and so on – thus, encoding a slightly
higher degree of DEIXIS than that associated with GivTops: examples
would be the preverbal subject terms the cathedral and the rest of Antibes
in examples (6a) and (6b) above. Both these definite terms manifest a part-
whole relation within the immediate macro-topic of their containing dis-
course segment, namely the city of Antibes.
Finally, ResTops encode the highest degree of deicticity of the three
subtypes, the referent being assumed to exist within the current discourse
model, but no longer active or present in the addressee’s current con-