Modality Choice for Generation of Referring Acts: Pointing versus Describing

The main aim of this paper is to challenge two commonly held assumptions regarding modality selection in the generation of referring acts: the assumption that non-verbal means of referring are secondary to verbal ones, and the assumption that there is a single strategy that speakers follow for generating referring acts. Our evidence is drawn from a corpus of task-oriented dialogues that was obtained through an observational study. We propose two alternative strategies for modality selection based on correlation data from the observational study. Speakers that follow the first strategy simply abstain from pointing. Speakers that follow the other strategy make the decision whether to point dependent on whether the intended referent is in focus and/or important. This decision precedes the selection of verbal means (i.e., words) for referring