Making global sense has long been seen as one of the most important criteria for judging the success of a given interpretation. For consecutive in particular, special emphasis is placed on the coherence and structure of the rendition. This study addresses the question of how to investigate coherence in interpreting and observe its development in trainees. We propose Rhetorical Structure Theory (RST), as a framework for exploring how coherence is realised in interpretations produced by professional as well as trainee interpreters. A corpus of 66 consecutive interpretations, by eight novice and three professional interpreters, of three Chinese and three English speeches, was transcribed, segmented into functional units, and mapped into a tree-like RST description. The analyses and results reveal that novices tend to focus on local cohesion while professionals tend to emphasise the global structure of the discourse. This difference can usefully be addressed in training.