ABSTRACTAs collaborative learning is actualized through evolvingdialogues, temporality inevitably matters for the analysis ofcollaborative learning. This study attempts to uncoversequential patterns that distinguish ``productive” threads ofknowledge-building discourse. A database of Grade 1–6knowledge-building discourse was first coded for the posts’contribution types and discussion threads’ productivity. Twodistinctive temporal analysis techniques – Lag-sequentialAnalysis (LsA) and Frequent Sequence Mining (FSM) – weresubsequently applied to detecting sequential patterns amongcontribution types that distinguish productive threads. Thefindings of LsA indicated that threads that were characterizedby mere opinion-giving did not achieve much progress, whilethreads having more transitions among questioning, obtaininginformation, working with information, and theorizing were moreproductive. FSM further uncovered from productive threadsdistinguishing frequent sequences involving sustainedtheorizing, integrated use of evidence, and problematization ofproposed theories. Based on the significance of studyingtemporality in collaborative learning revealed in the study, weadvocate for more analytics tapping into temporality oflearning.