This paper presents a corpus-based study of deictic non-verbal expressions produced by Greek television interview participants and their relation to the co-uttered speech. Deictic non-verbal expressions were extracted from a multimodally annotated multimedia corpus and they were studied in terms of typology, modalities employed, frequency, and distribution per participant. Furthermore, in order to identify the meaningful content they acquire, they were classified into categories according to their referent, expressed or not through speech, be it a concrete object, person or location or an abstract concept related to time and space.