Abstract
This study falls into the bigger framework of the CHROME project, addressing the definition and testing of a methodology of collecting, analyzing and modeling multimodal data for the design of virtual agents serving in museums. The paper analyses three tourist guides’ speech, focusing on silent pauses and voiced pauses (filled pauses and segmental prolongations). In this regard, a description of phonetic-acoustic and functional features of pauses and a classification of concomitant gestures have been performed. Results show a) speakers’ idiosyncratic linguistic behaviours; b) a clear distinction between silent pauses, mainly used for grammatical and intentional reasons, and voiced pauses that instead occur as ungrammatical and hesitation devices; c) such a distinction is confirmed by concomitant gestures, they are semantically loaded in silent pauses and semantically empty in voiced pauses.