326 Dik Bakker and Anna Siewierska
semantics, and semantics before expression, which is an improvement from
the perspective of the principle of functional explanation (Dik 1986). Fur-
thermore, the scope between the layers is respected in the sense that
choices are determined on the basis of scope relations.^2 The relevant part of
the information on the interpersonal level, mainly reference and ascription,
is mapped onto the representational level, and is then expressed from there.
Purely pragmatic information is expressed directly from IL, without inter-
mediate semantic mapping. Although Hengeveld’s model does not discuss
the role of the lexicon in any detail, we assume that both the interpersonal
and the representational levels have access to it via the Cognition module.
As opposed to the pragmatic and semantic levels, the FDG model does
not contain any new proposals for the organization of the expression rules
(ER) in a direct sense; they simply map IL and RL entities onto the Expres-
sion Level (EL), i.e. the morphophonological string that will eventually
trigger articulation. However, several aspects of the model suggest that the
overall structure of the standard expression component should be reconsid-
ered in order to fit the rest of the model. Firstly, the expression rules now
take their input from two levels rather than just from the traditional fully
specified underlying representation, as is the case in the standard model.
This considerably complicates the interaction between the respective com-
ponents of the grammar. Secondly, the expression level communicates with
both the Cognition module and the Communicative Context. The Cognition
module contains the linguistic competence, i.e. the lexical and grammatical
knowledge of the language user, which is also represented in the standard
model, although there the grammar is mainly presupposed. However, this
module also represents communicative competence and all other informa-
tion needed during the formulation and encoding of the speaker’s intentions.
The interaction with the Communicative Context, not represented in the
standard model at all, should be considered vis-à-vis expression in more de-
tail as well. Hengeveld’s model suggests that the Communicative Context
module is one outlet for expressions. Thus, after having been uttered, ex-
pressions also go there in their raw form, i.e. in the shape of the output of
the expression rules. This is to take care of metalinguistic operations, where
anaphoric reference is made to forms rather than meanings. Finally, the
model seems to assume a hierarchical organization for the (pre-phonetic)
output of the expression rules, going from paragraph to sentence to clause to
the different types of constituents and further down. The traditional expres-
sion rules and their output lack any form of constituency at all.
In this chapter, we wish to consider in some detail the implications of
the grammar model proposed by Hengeveld, henceforth FDG, for the ex-