- Frame structuring. The main clause functions as a frame for its subclause, therefore the processing of a complex sentence is easier if the main clause precedes the subclause. We expect this order to be more frequent than the opposite order.
- Referential coherence. As given information tends to precede new information, we expect that an expression referring to events that are given (because they have been mentioned in previous discourse) will be sentence-initial, while an expression referring to a new event will be sentence-final.
The production of complex sentences in context
This project deals with a specific case of the linearization problem (Levelt, 1989): the production of complex sentences in natural language discourse. More specifically, it focuses on the position of temporal and causal expressions in Dutch complex sentences. Suppose a speaker wants to express two acts or events in one sentence, in that case he may choose to express one of the acts by the verb of the embedding clause, and the other in a subclause (with a tensed verb) or another expression headed by a nontensed form of a verb (an infinitive, or a present or a past participle). As those expressions may occupy several positions in the independent clause, the question is: what positions do they have in texts? Is there a dominant order, and if there is, how can we explain for it? In the literature, two possible constraints on ordering can be identified:
It is unclear how both constraints interact in actual language use. However, there must be some principle governing the choice between frame structuring and referential coherence, because language producers are able to solve this problem in an implicit way. Otherwise they could not choose a particular production order in a given situation. Hence, the central question is: Which constraint dominates under which conditions? We propose that the choice depends on another (implicit) choice, viz. between speaker/writer versus listener/reader-oriented principles of cognitive economy. By means of a speaker/writer-oriented principle of cognitive economy, for example frame structuring the language user minimalizes his own text production costs. We expect that this kind of principle is favored, when writing activities have to be accomplished under time pressure. When there is no time pressure, the speaker/writer can afford to use a listener/reader-oriented principle, in order to minimalize the recipient's cost. In that case, we expect frame structuring to be the dominant constraint in text for listeners, but referential coherence in text genre meant for reading. These specific hypotheses are tested in corpus studies, for which four natural language corpora are selected that allow for the testing of contrastive hypotheses. Corpora vary on three dimensions: spoken vs written, spontaneous vs edited and text function (instructive, informative and argumentative).
The ultimate aim of the project is to contribute to a theory of the production of complex sentences. Such a theory predicts in which linguistic context which order of expressions is produced, on the basis of speaker vs. reader economy principles. Such an account would be compatible with production models like Levelt's, which is underdetermined at the level of 'message generation'.