We propose to mine temporal patterns in Intelligent Tutoring Systems (ITSs) to uncover useful knowledge that can enhance their ability to provide assistance. To discover patterns, we suggest using a custom, sequential pattern-mining algorithm. Two ways of applying the algorithm to enhance an ITS's capabilities are addressed. The first is to extract patterns from user solutions to problem-solving exercises for automatically learning a task model that can then be used to provide assistance. The second way is to extract temporal patterns from a tutoring agent's own behavior when interacting with learner(s). In this case, the tutoring agent reuses patterns that brought states of "self-satisfaction." Real applications are presented to illustrate the two proposals. (Contains 2 tables and 6 figures.)