358 Dik Bakker and Anna Siewierska
FG researcher in a better position to test all kinds of hypotheses about spo-
ken language. A rather concrete aspect of the integrated model presented
above, one which could introduce an interesting line of research to the the-
ory is the selection of the right subcategorization for the pragmatic and
semantic material offered to the expression component, i.e. the semantics-
syntax interface. The constraints the model presents may help us see better
how cognitively or semantically coherent stretches of input to the expres-
sion rules control and shape morphosyntax, with the obvious connections
to grammaticalization. We might also find out how certain grammatical-
ized patterns might in their turn ‘pull’ on semantics via the feedback
mechanisms, thus to some extent shaping the rest of the planning process.
Eventually, this should lead to a better understanding of the behaviour of
the language user, and to a better explanation of why languages are the way
they are.
Notes
- The authors wish to thank Matthew Anstey for a number of valuable com-
ments and interesting points of discussion with regard to an earlier version
of the text. We hope we have done justice to them. - Actually, outside-in expansion of layers (or left-to-right as it is called by
Hengeveld) is only assumed for the interpersonal level. Nothing is said
about the coming into existence of semantic representations. Although we
think that here, too, the hierarchy among the respective layers has some
predictive value in the outside-in sense (cf. Bakker 1994 for more on this),
there can be no direct correspondence between the universal order in which
underlying clauses are structured and language-specific surface orders of
constituents. - Of course we could, if only for the sake of argument, take an extreme func-
tional perspective, denying all autonomy to morphosyntax. It might then turn
out to be possible in the grammar of individual languages to determine the
combination of all underlying factors which will eventually lead to certain
order patterns. These factors could be made available as μ operators to the
Phase I expression rule that determines the form of the finite verb. Unless we
see the form of the agreement marker as the expression of this conglomerate
of factors, i.e. not only of person/number aspects of the Subject, but also as
information about its pragmatic status, this would indeed boil down to pre-
running the expression rules in a more or less disguised way. - This requirement can be made explicit by formulating the role of the ex-
pression component such that a grammar of language L is descriptively