FDG and language production 181else being equal, conforms with LIPOC. Note that there is nothing anti-
functional about this understanding of ‘choice’. On the contrary, automatic
processing can itself be seen as functional, since it certainly contributes to
efficiency, freeing the speaker’s and hearer’s minds to concentrate on the
subject-matter of the discourse.
Applying these insights to the construction of a model of FG, we will
need to regard the various generalizations made within FG as constraints
upon the cognitively impenetrable processes of language production. What
would then distinguish a functional grammar from a non-functional one
would be that the constraints are not arbitrary. Rather, they are understand-
able a posteriori in terms of efficient interpersonal communication. The
position that suggests itself, then, is one that is compatible with Jackend-
off’s second option: an FG is declarative and takes the form of a set of
constraints on processing. Across billions of communicative acts, these
constraints have proved their worth. It is the functionalist’s task to identify
them and to provide explanations for their existence.
- FG as a model of language production
As mentioned in Section 1, some recent work has been moving FG in the
direction of Jackendoff’s third position. The main exception to the domi-
nant interpretation of FG as simply a declarative grammar is to be found in
the work of Jan Nuyts, who has forcefully presented his Functional Proce-
dural Grammar (Nuyts 1992). He has argued that a number of matters
traditionally regarded in FG as being part of grammar proper should be as-
signed to non-grammatical cognition. Questions such as the assignment of
pragmatic functions should in his view be handled in conceptual and/or
textual components, and in any case fall outside the grammar. Another im-
portant consequence of Nuyts’s work concerns matters such as the
distinction between operators and satellites, or the distribution of informa-
tion over main and subordinate clauses. For both phenomena, their
treatment in the grammar is entirely dependent upon their ultimate formal
properties: operators correlate with grammatical, and satellites with lexical
expression; similarly, there is no simple or direct correlation between cog-
nitively nuclear information and main clauses, or between cognitively
subsidiary information and satellite clauses. Nuyts argues that these
distinctions are also more plausibly understood as resulting from cognitive
(pre-linguistic) operations. He takes Jackendoff’s third position in propos-