162 Michael Fortescue
As regards ‘phrasal suffixes’, the situation is much the same as I have
described for West Greenlandic (Fortescue 1992: 131), namely that if this
is a matter of derivation in the Fund (a perfectly reasonable assumption),
then such a derivational process must be able to ‘dip down’ (or rather ‘up’,
depending on the kind of diagramming of the FG model one has in mind)
into layers of clause structure where not just extended predications but
even whole propositions have been elaborated as input for the derivational
process concerned. The bound derivational suffixes of Nootka and West
Greenlandic need in any case to be treated independently of the predi-
cate/term formation processes into which they might enter and belong in
the lexicon with their own predicate frames. As the examples from Japa-
nese and English in note 5 show, this cross-level recursivity of derivation is
not limited to polysynthetic languages.
Finally, as regards the ‘zero copula’ construction, if one chooses to fol-
low Dik’s approach to copular constructions, a conversion of a lexical
‘term’ to a predicate expression is assumed, but if all lexical items are am-
biguous as to term or verbal status in Nootka, there is no need for such a
conversion at all – unless one first produces a term by term formation and
then converts it back to a predicate by term-predicate formation, a seem-
ingly unmotivated extra complication for this type of language. If, on the
other hand, one follows Hengeveld’s approach mentioned above, one
avoids having to postulate ad hoc predicate formation rules, and can treat
non-verbal predicates in terms of predicate variables (f 1 ), just like verbal
ones. However, there is very little evidence in Nootka for distinct predicate
frames (with predetermined argument sets) being associated with particular
non-verbal predicate construction types any more than with verbal predi-
cates, so it is hard to tell if this is an advantage or not. I shall return to this
matter below.
In sum, the nub of the problem is the relationship between the lexicon –
with its predicate frames corresponding to discrete SoAs – and the rest of
the model (in particular the choice of elements in the basic predication). A
given lexical morpheme (or “predicate”) in Nootka may enter into various
constructional “frames”, with the valency even of a specific verbal frame
being rather vague compared to the situation in more familiar language
types. In Nootka, lexical morphemes are not anchored in the lexicon in
quite the same way as in more familiar languages: it would appear to be a
typological fact one must live with (and FG claims typological adequacy)
that some languages have a much closer correspondence between States of
Affairs (SoAs) and lexical items than do others (compare Fortescue 1992
for another polysynthetic language – Koyukon – which does, unlike