This article targets action formation in multimodal sequences. It shows how nonpolar interrogatives in Estonian are used for noticing breaches in others’ embodied conduct, focusing on kes (“who”)-interrogatives. In contrast to information questions with kes, a “noticing of a breach” does not seek an informative answer, which would be an identification of the grammatical actor of the action depicted in the interrogative. The actor is instead the addressee of the turn, often called by name, and thus clear to everyone present. These “rhetorical” kes-interrogatives formulate a just-observed conduct as problematic, and attribute responsibility for it. Since they call for either a remedy of the (embodied) conduct or a contesting of the blame as the next action, noticing breaches marginally qualify as directive actions. At the same time, they do not explicitly provide any guidelines for the future. The study argues that to determine function in language, it is necessary to study grammatical structures in their temporally emerging and embodied activity contexts. The data are Estonian with English translation.