This paper addresses the nature and significance of productive engagement in cognitive activity and metacognitive regulation in collaborative learning tasks that involve complex scientific knowledge. A situative framework, combining the constructs of social regulation and content processing, provided the theoretical basis for the development of a comprehensive coding scheme for interactive data analysis. An empirical study was conducted with two groups of university students working on two science-learning tasks. It examined the function of metacognitive regulation to control the flow of cognitive activity, and the extent to which group differences in cognitive and metacognitive regulation processes during collaborative learning could explain differences in the groups' learning outcomes. The findings provide validation of the framework and its derived coding scheme. An example of a way in which a group engages in socially shared metacognitive regulation is presented to demonstrate how the coding scheme was applied to the data. Theoretical and empirical implications of the findings are discussed.