This paper presents a first attempt to score the relevance of multimodal referring expressions within a task oriented environment. It is based upon the application of the ecological approach to multimodal system design, which in particular implies that perception has to play a central role in the understanding of a gestural designation. The results of an experimental work is presented, together with the scores obtained by combining the linguistic characteristic of the referring expression and the properties of the corresponding gestures (if any). Even if the calculus that we present has to be refined it seems to be already suited to validate our approach regarding the importance group salience and access type in the choice of a referring mode.