188 J. Lachlan Mackenzie
such as English put, which in its literal sense calls for specification of
Agent, Goal and Location. In Levelt’s (1999: 95) formulation, “[i]n a way
grammatical encoding is like solving a set of simultaneous equations. Each
lemma [in FG, ‘predicate’ JLM] requires particular syntactic constraints
from its environment and the emerging syntactic structure should simulta-
neously satisfy all these constraints”.
If the representational component is seen as a set of constraints on proc-
essing, it is no longer necessary to assume that every utterance has a full
propositional form. The grammar need simply be consulted for what is nec-
essary to the task at hand, or as Hengeveld (this volume) has it, “it is no
longer necessary to assume that every discourse act contains a proposi-
tional content”. Thus an act such as A 3 in (x) above, containing only one
referential subact, undergoes only those constraints that relate to term
structure, including, in those languages that require this, a semantic or syn-
tactic function that determines its case form. In this way, it is possible to
treat so-called fragmentary utterances without recourse to rules of deletion.
The complete clause then emerges as a system for supporting the Focus,
providing a context within which the addressee can understand it as well as
possibly satisfying other communicative goals of the speaker’s.
The interpretation of Hengeveld’s model that I am proposing is a com-
bination of Jackendoff’s second and third options: the overall grammar is
procedural (position 3), but in the construction of complex acts, it needs to
consult, indeed conform to, a declarative grammar (position 2). The advan-
tage of a procedural approach is that FG will accord better with the
“psychological models of linguistic competence and linguistic behaviour”
which Dik (1997a: 13) wishes the practitioners of FG to bear in mind. In-
cremental processing in the production of utterances works on the
assumption that the concepts that lie behind a complex act are not all avail-
able at once. Speakers characteristically do not wait until all the concepts
are “in” before starting to speak. The various levels of activity, conceptual-
izing, formulating and articulating, as Levelt (1989) has it, work in parallel,
with conceptualization having a head start over formulating, and formulat-
ing over articulating. However, speakers are not free to express their ideas
exactly as they occur to them: the declarative grammar, with its conven-
tional principles, will often mean that particular parts of a message are put
into a buffer before they can be expressed.
Clausal structure appears to have exactly this buffering function. At least,
this seems to be plausible for English, a language with rather fixed word or-
der, by which I mean that the syntactic template plays an important role in
expressing the relations imposed by the declarative grammar. But what of a