Process and pattern interpretations 163Nootka, have a very close correspondence between the two). Perhaps the
mistake is to take the SoAs of the original model to be truly universal.
An attempt to incorporate a typology of languages with varying degrees
of rigidity of word-class distinctions within FG has been made by
Hengeveld (1992: 45ff.). At first sight Nootka would appear to go with
languages like Tongan, representing the highly ‘flexible’ type, in which
predicates can be used in practically any function. This contrasts with
highly ‘rigid’ languages like Cayuga, which may have only one part of
speech, the predicate, but where this must be morphologically derived to
function as other parts of speech than the verb. In fact Hengeveld’s ex-
tremes meet, since in ‘rigid’ languages any one morpheme may be
extremely flexible as regards morphosyntactic context, and the real contrast
is with ‘specialized’ languages like English with a full array of morphosyn-
tactically distinct parts of speech. The problem with fitting Nootka into this
typology is that the distinct morphological marking of predicate words in
the one or the other sentential or phrasal function is not obligatory, so that
it partakes of both the Tongan and the Cayugan type depending on whether
there is suffixation present or not (and, as we have seen, suffixes tend to be
shifted to the first word of constructions, whatever it is). Hengeveld (pers.
comm.) suggests that Nootka might represent a ‘rigid’ language in which
the relevant morphology has relatively recently lost its obligatory status.
It seems then that some languages, like Nootka, are simply more con-
trolled by discourse pragmatics than others which are more lexically
controlled and in which there is a closer link to SoAs.^15 Nootka will, for
instance, split what in other languages correspond to unitary predications
into two clauses (of varying degrees of tightness) to emphasize the salience
of the undergoer argument. If this is so, we must look for another place in
the model (or an extension of it) where such matters as discourse-
determined choice of initial predicate can be accommodated. Short of put-
ting virtually all of the morphosyntax of Nootka within a ‘discourse
module’ (on the nature and categories of which there is no consensus
within FG today), the only obvious possibility would seem to be to intro-
duce discourse factors determining the choice of main clause predicate at
the lowest level of clause structure (e.g. by a feature ‘most newsworthy’),
obviously a severe, and undesirable, disruption of the model. This can be
overcome, I would suggest, by clearly distinguishing between a ‘process’
and a ‘pattern’ interpretation of the FG model. It is the former that must be
reassessed in the light of the kind of evidence I have presented. Let me
elaborate the point.