The classification of coherence relations and their linguistic markers: An exploration of two languages

It has become popular among discourse linguists to explain a text's coherence by identifying 'coherence relations' which apply at various levels between its component spans. However, there is currently no overall agreement about how to define a standard set of coherence relations, and even about what the coherence relations in a text are intended to represent. In this paper, both questions are addressed: we outline a conception of relations as modelling psychological constructs used by readers and writers, and suggest how a limited set of categories of coherence relations can be identified. We relate two independent methods for investigating relations, one drawing mainly on psycholinguistic experiments on Dutch speaking subjects, the other starting from a study of the 'cue phrases' used to signal relations in English text. Both approaches lead to classifications of relations and cue phrases. We examine to what extent these classifications converge - and to what extent they accord with the psychologically motivated classification - in a comparative study of a set of cue phrases in English and Dutch. Interesting similarities are noted on both counts.