The data analyzed in this paper is classroom interaction in a class of 3rd graders working in pairs to solve a mathematical puzzle game. The research approach is practice-based design research, informed by ethnographic methods and multimodal interaction analysis. The aim of the interaction analysis is to inform further design through the creation of a collection of learning designs embodying principles of observational learning and social motivation, along with observations from their deployment in the classroom. Data is collected by using a wide-angle video-camera in the ceiling, which is recording nonstop, in combination with handheld video cameras recording over the shoulder of pairs engaged in the activity. Two pairs of learners are followed as they solve a complex logical puzzle. I will look at how the pair members construct each other as more or less “knowledgeable”, in relation to the process of the game and in relation to the task they have to solve. A second issue is to what extent the notion that the activity is a game contributes to structuring interaction between pair members in the two pairs followed. Since my approach concerns design of learning activities, both issues will be related to how opportunities for gameplay, collaboration and rivalry relate to the design of the learning activity.