The aim of this process-oriented video-observation study is to explore how groups that perform differently differ in terms of the number, quality, and temporal variation of their content-level (knowledge co-construction) and meta-level (monitoring) activities. Five groups of teacher education students (n = 22) were observed throughout a 3-month course. Video recordings (33 hours) of face-to-face group interaction (n = 12,931 speech turns) and pre- and post-tests of students' knowledge were collected. The results show that the well-performing group was more engaged in high-level knowledge co-construction and monitoring activities. The well-performing group was also capable of maintaining a higher level throughout the tasks, whereas the lower performing groups' knowledge co-construction and monitoring activities was reduced during the course.