Unfinished turns can be considered problematic instances according to traditional syntactic frameworks. The multimodal interactional approach makes it possible to study these realizations and, in particular, their "non-canonical" completions, i.e. when they are accomplished in an embodied way by the same speaker. In this paper, relying on the video analysis of naturally occurring interactions, I propose a sequential and syntactic account of three main configurations of multimodal completions in Italian. The analysis of three types of syntactically unfinished turns and the embodied resources that complete these realizations lead me to discuss the multimodal organization of talk-in-interaction, namely the sequential status of iconic and pointing gestures integrating syntactic structures, as a primordial feature for the definition of interactional units across situational specificities. Extracts are issued from an eight-hours corpus documenting different settings, such as interactions in fair trades, ordinary interactions, and guided visits with native Italian speakers.