Instructional videos have become an important site of everyday learning. This paper explores how these videos are used to complete practical tasks, analyzing video-recorded interactions between pairs of users. Users need to repeatedly pause their videos to be able to follow the instructions, and we document how pausing is used to coordinate and interweave watching and doing. We describe four purposes and types of pausing: finding task objects, turning to action, keeping up, and fixing problems. Building on these results, we discuss how video players could better support following instructions, and the role of basic user interface functions in complex tasks involving different forms of engagement with the physical world and with screen-based activity.