Herbert H. Clark’s Joint Action Theory (JAT) has been groundbreaking for understanding the social and cognitive mechanisms that allow people to effectively coordinate joint actions in conversational, face-to-face settings. Using a method we call “Pair Analytics,” we have extended the application of JAT to the study of analytical reasoning in computer-mediated, human-to-human interactions. Pair analytics (PA) sets a naturalistic scenario in which the social and cognitive role of human-human and human-computer interactions can be studied. In this paper, we support the claim that coupling JAT and PA is an effective research strategy to capture and study three socio-cognitive phenomena in collaborative visual analytics: (1) structuring and navigation of joint analysis; (2) management of joint attention; (3) and signaling of cognitively demanding tasks.