Bridging the gap between coherence and cohesion: which cohesive devices are really textual?

Abstract

The purpose of this paper is to show that only a small number of the cohesive ties normally accounted for in cohesion analyses do actually contribute to the reader's perception of relevance and coherence. To this aim, a comment article from Guardian Unlimited consisting of 60 coherence units and representing typical argumentative written text is analysed from a discourse-as-process viewpoint. First, the relation between each pair of consecutive sentences is discussed with 25 participants in order to arrive at a consensus about the type of connection that helps to perceive the relevance and coherence of the text at each juncture. Next, the cohesive ties contributing to the participants' perception of relevance and coherence are determined and analysed. Finally, all kinds of cohesive ties are identified from a text-as-product perspective and analysed. A comparison is made between the features of the cohesive items identified under the two approaches. The results show that only those cohesive resources dealing with whole sentences, larger fragments, or certain simple clauses linked paratactically, can be regarded as textual in nature, i.e. as contributing to the readers’ perception of text relevance, and therefore, coherence. Thus Sinclair's (1993) hypothesis about written text structure is confirmed.