Cognition and FDG 83which stays as it would have been, without any indication of the difference
in relative social power between the interlocutors:
(3) [pi: [[PastProg ei: matsuv (x 1 )Ag (x 2 )Go] (ei)] (pi)]
This means that the request has to go down to the next layer (= level in
FDG), where it produces some changes at the expression level. More con-
cretely, the form o-machi shite is employed instead of matte ‘waiting’
(from matsu ‘to wait’); likewise, the form orimashita is used instead of
imashita (‘was’ or ‘had been’ in this example). It is at this level (the
equivalent to the physical layer in the computer communication model)
that the form is transmitted. Then, in decoding, the process is repeated, but
this time it goes upward from the most ‘physical’ level to the actual level at
which the request is processed. The ‘package’ is fetched and ‘unpacked’ at
the expression level. The data obtained from processing the expression
then goes to the representation level. At this level, the representation can be
correct from a propositional point of view, but there is still a package of
data to be processed which cannot be interpreted there. The package then
goes up to the interpersonal level, where it gets the interpretation it needs,
which is important, because the addressee needs to evaluate the adequacy
of the interpersonal settings that the speaker is using for the current com-
municative context. At all stages, of course, the receiver has needed to
consult the long-term information stored in the cognitive component in or-
der to be able to interpret the data packages.
This approach is also compatible with discourse processing, since both
macro-planning (concerned with the preparation of large stretches of
speech) and micro-planning (concerned with grammatical and lexical
choices; cf. Butterworth 1980: 159) can be encoded in some way in the
data packages involved. There are some marks at the expression level that
clearly encode information relevant to the interpersonal level (e.g. dis-
course markers, which show relations in discourse). This can also be
applied to pauses, which are significant as indicators of processing time.
As Butterworth (1980: 155) explains, “(...) the more the delays, the more
cognitive operations are required by the output”. As it can be assumed that
long pauses reflect more complicated cognitive processes than short
pauses, they can be signals of discourse planning and show relevant land-
marks of the cognitive processes involved (cf. Schilperoord 1998). As they
are involuntary, it is questionable whether we should think of pauses as
some kind of unit analysable at the expression level, but there is no doubt
that the addressee can use them to reconstruct discourse structure.