36 Matthew P. Anstey
structure as semantic relations, thereby eliminating the need for an
abstract NP structure;^17
(c) deferment – sentential and clause structure, including linearization
information, is reified into word-order templates, which, although
‘existing’ per se, are prohibited from contributing their structural in-
formation until after the UR is fully specified with functional
information.
In other words, nouns and adjectives are assembled into ‘term frames’
(that is, term structures), which are assembled into verb-bearing ‘predicate
frames’, which are assembled (with other ‘term frames’ as satellites) into
‘predication frames’ (that is, URs), which are mapped onto ‘sentence
frames’ (that is, word-order templates). The genius of Functional Gram-
mar is that this approach creates the illusion, in my opinion, that there is
no autonomous constituent structure, when in fact the four frames have an
a priori status in the model.^18 The presence of covert structure in the
model explains FG’s ability to describe certain (often distributionally
dominant) types of language data and its difficulties with other types of
language data.
For the sake of argument, however, let us assume that my ‘reconstruc-
tion’ is incorrect and that there is genuinely no constituent structure in
Functional Grammar. In this case, the success of Dik’s approach in de-
scribing linguistic data raises the obvious question: why does it work at all?
Why does Functional Grammar demonstrate that (a large majority of) syn-
tactic structure is apparently epiphenomenal, a mimicry of semantic
structure? And why can systematic syntactic deviations from this isomor-
phism so often be tied to pragmatic features of the communicative context?
Finding thorough answers to such questions is the quintessential task of the
functional linguist, as it would demonstrate that the fundamental functional
hierarchy of influence – pragmatics > semantics > syntax – is not just an a
priori belief but a persuasively substantiated linguistic theorem.^19
5.4.2. The problem of the underlying representations (PR2)
According to Dik, the URs of FG 1 supposedly only contain lexical items in
functional relationships with one another. Such a view facilitated Dik’s
shift to understanding the URs as bearers of meaning. No additional
semantic interpretative module is necessary, since the UR contains all the
information necessary to deduce the meaning of the sentence. As Dik