Towards a speaker model of FG 337user. A grammar model represents the respective components a theory dis-
tinguishes in the grammar of a language and the relations between these
components. Furthermore, it provides formal representations for the re-
spective rules and structures. It abstracts away from any speech situation
and perceives the language and its grammar as an abstract, formal con-
struct. The grammar model shows in what way the data structures (the
static elements of the grammar, such as the lexical items and morphologi-
cal and syntactic rules) lead to underlying semantic and syntactic
representations and these to actual linguistic expressions. Usually, these
models are restricted to the sentence as the highest level of representation
and production. There may be a fixed order in the flow of information
through the respective grammar components. Often, the ‘later’ components
not only expand the structures generated by the ‘earlier’ components but
also operate as filters on ill-formed structures. In principle, such models are
bidirectional in the sense that they relate the lexicon and underlying repre-
sentations to expressions and vice versa via ‘left-to-right’ or ‘right-to-left’
applications of the rules of the grammar. However, the traditional interpre-
tation of such models is from underlying representations to expressions; for
this reason they are often called generative. The grammar model is usually
thought to represent the linguistic competence of the ‘ideal speaker-hearer’.
It does not deal with typical performance aspects such as speech errors and
hesitations. This, indeed rather abstract model is the most common one
found in the linguistic literature. The model of FG presented in Dik (1997:
58) is an example of this type. Sells (1985) presents such models for Gov-
ernment-Binding Theory (1985: 24), Generalized Phrase Structure
Grammar (1985: 79) and Lexical Functional Grammar (1985: 137), respec-
tively. Although, as a model of the theory, such models should have
descriptive adequacy, they are not required to be psycholinguistically ade-
quate.
In Dik (1997: 1) the ultimate goal of a linguistic theory is formulated as
the construction of a model of the language user, a model that can ap-
proximate the communicative performance of human beings when using
their linguistic competence. It might therefore be assessed with regard to its
psychological adequacy. Such a model should not only contain linguistic
knowledge. It should also deal with all kinds of other knowledge, with be-
liefs, prejudices, etc. Furthermore, it crucially takes into consideration the
various types of memory: working memory, semantic memory and epi-
sodic memory. While a grammar model is neutral, a model of the language
user should distinguish between speaking and hearing, given that these two
modes of linguistic behaviour differ in many respects. A model of the