The article considers how ethnomethodology (EM) studies visuality. Historically, there were four approaches to visuality in EM, those of visuality as observable activity, as images, as practices of vision, and as language constructions. Harold Garfinkel’s idea of witnessability implies that the phenomena of order exist in observable methods of their production. Observability, or witnessability, at this point can be an equivalent for visuality because it suggests an opportunity to be studied with visual methods. In such a case, any ethnomethodological study is a visual study because it implies a description of observable methods of order production. Besides this idea of observability, three separate projects of visual research were developed in EM from 1980s to the 1990s by Michel Lynch, Charles Goodwin, and Jeff Coulter. They all tried to present the practical approach to visual perception (in contrast with perception as an individual psychological process), but found solutions in studying different aspects of visuality, which were images, practices of seeing, and language constructions describing different modes of perception. This text considers the relationships between the three conceptions and Garfinkel’s initial idea of observability. It analyzes ideas which Lynch, Goodwin, and Coulter added to Garfinkel’s ethnomethodological program, and shows how other ethnomethodologists use these additions. The article demonstrates that together they produced EM’s approach to visuality, although none of these projects were completely realized inside EM. It is based on Garfinkel’s idea of observability, supplemented by the opportunity to study perception as a practical social achievement situated in local interactional contexts.