152 Michael Fortescue
stantiation’) that enter into the act of reference, not variables. The basic
cognitive processes as I see them are rather Whiteheadian ‘prehensions’,
the grasping and assimilation of both perceptual and conceptual ‘data’,
specifically ‘propositional prehensions’, which link the prehension of a
predicate to the prehension of a set of logical subjects (the totality of rele-
vant ‘entities’ bound by the predicate – cf. Fortescue 2001: §1.2.5). At this
level of cognition we are not yet talking of linear chains of linguistic signs:
the processual ‘concrescences’ whereby multiple data-to-be-expressed are
integrated on the way towards a determinate linguistic output should be
taken as having their own internal complexity (unless purely automatic re-
sponse is involved).
Now Dik’s term variable ‘x’ in underlying predications simply stands
for a first-order entity (as opposed to a proposition ‘X’, for example). In a
nuclear predication variables and their first restrictors are replaced by
terms. ‘Firstness’ is all that determines which predicate restrictor is to be
treated as term head by binding with a preceding variable. This kind of no-
tation is innocuous and does indeed serve a useful purpose, namely in
relating underlying clause structure to the lexicon, where individual predi-
cates (including those defining terms) can be represented as a predicate
frame relating directly to a particular kind of State of Affairs. However,
this is a matter of pattern, not of process. There is no guarantee that the
straightforward linkage between predicates and SoAs has any direct rele-
vance for cognitive processes as such.
I would like now to suggest that we may need two versions of the FG
model, a Process and a Pattern version, rather than just one (Dik’s original
one, say) serving both purposes – or, if there is agreement on a monolithic
framework again, then it should at least be of such a nature as to be inter-
pretable in two clearly distinct ways and not remain indeterminately
‘hybrid’. This has been the subject of much debate in recent years within
FG circles, in the form of discussions of the upward limit of layering of the
theory. Dik himself proposed that his model should be envisaged as em-
bedded in – or side by side with – a distinct discourse grammar (Dik 1997:
409ff.), and this idea has been followed up in two principal ways (as re-
flected in the various articles in Bolkestein and Hannay 1998),^3 either by
advocating the extension of the original model to include such a ‘discourse
grammar’ or by calling for their treatment as two separate, interacting
modules. I would suggest that neither of these options is really what we
need if the original functional model is to be seen in cognitive terms. For
you cannot simply embed pattern within process (or vice versa). The model
as it is (as pattern) needs a process interpretation – where by ‘process’ I re-