This paper introduces a framework based on EMCA and semiotics related to practices concerned with perception. This is subsumed under the title: semiotics of distributed perception. It will be argued in this presentation, that perception is not just a private, cognitive experience based on one’s own motor skills and interpretation of sensory input, but achieved in and through practical communicative actions in co-operation with other agents. Based on years of video ethnographic research conducted among visually impaired people and an EMCA and semiotic framework, this paper introduces the concept of perception-related actions and shows how these are observably provided by other agents as resources that constitutes a distributed perceptual field, which enables blind people to perceive and construct actions. The kind of semiotics that will be brought into the EMCA framework is the triadic Peircean (1955) tradition with a reference to the works of Kockelman (2005), Enfield (2017) and Goodwin ; i.e. with a focus on the interactional accountability of the sign production. In this presentation I will particular focus on different forms of semiotic agency. One reason EMCA scholars arguably could to turn to semiotics is in order to get a more vivid terminology and understanding of the agents involved in interactions. EMCA do not recognize ANT’s heterogenous approach to agents but maintains the ability of accountability. Not in any juridical sense, but in the ethnomethodological understanding of member’s procedures for making settings accountable and in some way to hold agents responsible. To make actions accountable involves also producing them in ways that make them recognizable for what they are (Koschmann, 2012). Other agents than human beings may produce accountable actions as we clearly dismiss cognition as relevant. In this presentation I will show how distributed perception is achieved in and through actions and assistance from a human agent, a non-human (dog) agent and a technological AI agent. Although a dog and an AI hardly can be said to have “commonsense knowledge” (Garfinkel & Sacks, 1986, p. 178) as members of a society, they can surely act in ways “that are sensible in that society and in the situations in which they find themselves” (Have, 2002, p. 6).