344 Dik Bakker and Anna Siewierska
mechanisms, and of the parallel, incremental nature of the process of
speech production in general.
Example (10), from Chafe (1987: 23), is a case in point. In (10), a hy-
phen indicates lengthening of the preceding segment; two dots indicate a
brief temporal break; three dots indicate a pause; the acute accent shows an
intonation peak. Each line would be an intonation unit in Chafe's terms.
(10) a. ... A-nd .. he would come into cláss
b. ... a-t .. uh- you know three or f
c. ... precísely one minute after the hóur,
d. or something like thát
e. ... a-nd he- .. wou-ld .. immédiately open his ... nótes up
The proposal in Mackenzie (2000) for an incremental model of FG
(IFG) should be seen in the light of the above remarks.
A last point we would like to make here is related to the nature of the
representations of the information given at the respective levels in
Hengeveld (this volume). For the RL level the usual Underlying Clause
(UC) structure is assumed, cut off at the propositional layer. For EL we
propose the structures introduced in Section 2; the link with the other levels
of representation will be worked out further in the next section. However,
some aspects of the IL representation remain unclear to us. The outer lay-
ers – Move, Act and Illocution – are formally of the Underlying Clause
(UC) type. Their contents are variables, operators, and references to
speech-act participants, i.e. they are formal in nature. The inner layer, un-
der variable C, has referring acts (with the variable R) and ascription acts
(with the variable T) as its contents. These acts get their semantic content
only after projection on the RL level, in the form of predicates of the lan-
guage, plus the corresponding variables, operators and functions. The point
now is how this predicational content should be determined. This problem
relates directly to the way in which mental representations are made in se-
mantic memory in the first place, both of the knowledge to be reported in
general and more specifically of the communicative context. If linguistic
relativism is assumed, as is suggested by the stepwise lexical decomposi-
tion philosophy proposed within FG ever since Dik (1978), then the
predicates of the language concerned make a considerable contribution to
the ‘stuff which thoughts are made of’. In addition, Dik (1989) proposes to
take UCs as the representational formalism for those types of knowledge
which are non-pictorial in nature. In other words, FG promotes a linguisti-
cally motivated type of knowledge representation. If this were indeed valid,