The research presented in this paper examines how collaborative learning manifests in different environments of emergent play. Emergent games are interesting objects of study from a serious games perspective as their non-linear and open-endednature canalleviate issues caused by impersonal and inflexible content. But, in order for them to be useful in learning contexts, methods for assessment of player actions and participation in emergent games need to be improved. Our approach to this issue was to device a methodology to track individual group members’ work contributions during different types of group exercises. Groups of middle-school children, ages 6-9,were tasked to build structures out of LEGOs and in the game Minecraft and, through the devised tracking method, data from the different exercises were compared in order to determine how the collaborative patterns within the groups varied depending on what type of exercise they were performing. The results of the study indicate that the computer based emergent system was experienced as more engaging and immersive than the face-to-face one, and that it fostered continuous discovery, experimentation and problem solving throughout the game session.The devised methodology resulted in some good indicators regarding collaborative behavior, but more parameters need to be added for it to be usable for effective and meaningful player assessments.