288 Jan Nuyts
cursive or interactional issue, relating to how the speaker organizes his/her
discourse in view of his/her communicative purposes.
Still, it is quite obvious that such clause connections do relate to the hi-
erarchy of qualifications, though probably in very complex ways. For
example, temporal connections are unavoidably based on the temporal
situation of each of the related SoAs. Or consider conditionals. Performa-
tive epistemic expressions do not occur in a conditional protasis (Nuyts
2001a: Sections 2.3, 3.3 and 4.3). The explanation may be similar to that
for the absence of performative epistemic forms in questions. In the prota-
sis, a speaker brings up a hypothesis about how some conceptual gap might
be filled. But (s)he is not conjecturing how (s)he expects the gap to be
filled; in that case an epistemic assessment would be to the point. Rather,
(s)he is stating one possible way the gap might be filled, with the aim of
pointing out (in the apodosis) what consequences that might have, irrespec-
tive of its probability. Hence, there is no use for an epistemic expression in
the protasis. This analysis does show that conditionals have an intimate
link with the domain of epistemic qualification. Nevertheless, this and
other dimensions of SoA linking are not part of the layered system as
such.^18 How combining SoAs in relation to the layered system in concep-
tualization does work in a language production system is a matter that goes
beyond the scope of the present chapter.
With reference to the current concerns in FG with discourse structure,
the foregoing obviously implies that in the present cognitive-functional
framework, discourse organization as such does not belong in grammar ei-
ther, i.e. in the cognitive system for processing linguistic structures (I am
using the term ‘grammar’ in this chapter in this narrow sense, obviously). I
am not certain where or how in Hengeveld’s (this volume) new proposal
the lexicon comes in. But assuming that at least the ‘expression level’ is
fully lexical, since a complete unit there is a paragraph, I take it that dis-
course organization is ultimately considered to be entirely part of linguistic
structure and processing. In the current cognitive-functional framework,
however, grammar is a device solely for utterance processing. Surely,
grammatical processing is critically determined by discourse factors, but
that is a matter of processing the effects of discourse on the internal consti-
tution of a single utterance. Discourse organization as such is handled in
several steps in Functional Procedural Grammar. But central to it is a sys-
tem (the ‘situational network’) which performs a first adaptation of long-
term conceptual knowledge to the actual communicative situation. This
system still works in terms of conceptual structures, unlike grammar
proper. But unlike the central conceptual system, it uses mid- to short-term