280 Jan Nuyts
tem of qualificational expressions. It obviously also implies that there is a
discrepancy between the two. One may take this to mean that there must be
two layered systems, which differ in more or less substantial ways: one in
grammar and one in conceptualization. However, the observations rather
force one to rethink the role of layered representation in grammar in more
fundamental ways. As argued in Nuyts (1998, 2001a), there is no need for
a layered system in the cognitive procedures dealing with linguistic struc-
tures at all. Layering as such is a purely semantic issue, viz. a matter of
semantic scope. But grammar deals with forms, not with meanings. And
where forms respond to semantic properties, as is often the case in the be-
haviour of qualificational expressions, this should be handled in terms of
the procedural interaction between linguistic structure and the layered sys-
tem in conceptualization, rather than by duplicating the latter in grammar. I
am not arguing that layering plays no role in grammar: quite on the con-
trary, its effects on grammatical/syntactic structure and processing are
tremendous. But this is precisely the right way to state the issue: we are
dealing with effects, not with an inherent feature. To put it metaphorically:
layering is an ‘emergent’ property of linguistic structure (more or less in
Hopper’s 1998 sense).
Some effects of the semantics of layering certainly do become inherent
in grammar, i.e. when they are structuralized or grammaticalized in linguis-
tic form. But this process leads to different results for different
qualificational expression types, so that it cannot be stated in generalized
terms anyway (contrary to what is implied by the postulation of one global
layered system in grammar). Thus, if we stick to our example of epistemic
modality, in languages such as English, Dutch or German there are consid-
erable differences between the grammatical (auxiliary), lexical (adverbial)
and predicational (adjectival, verbal) expressions. This is true in terms of
which distinctions they code: the adverbs and adjectives allow a fairly sub-
tle rendering of all positions on the epistemic scale (even more so when
combined with modifiers), but the auxiliaries and verbal predicates only
allow a vague reference to a few areas on the scale (and they hardly allow
additional modifiers). And this is equally true of the behaviour of the forms
in these expression types, for example in the range of word-ordering possi-
bilities. Moreover, all these features can be stated in terms of the traditional
dimensions of grammatical description, i.e. which forms are available, how
can they be combined, and how can they be ordered in an utterance, in
view of the basic properties of the ‘parts of speech’ involved. There is no
need for any additional constructs such as a layered system. Once again,
there is no doubt that the linguistic properties of qualificational forms are