Recent studies of MOOCs demonstrate their ability to reach a large number of users, but also caution against the high rate of dropout. Some have looked closely at MOOC participation in order to better understand how and when users start to disengage, and, if they remain engaged, in what activities they participate. Most of this prior work relies heavily on descriptive statistics or clustering methodologies to highlight basic user participation characteristics. In this paper, we adapt NMF to provide a multi-dimensional view of user participation. We use log data to create a bottom-up understanding of user participation, and identify five basic behaviors associated with participants' use of content and their engagement with assessment. Furthermore, we do a cross-course analysis across four courses and find that these five behaviors are present in all courses. Interestingly, users' participation patterns - how they engage in these five behaviors - vary across courses even when the course topics are similar. Our methodology can be applied to other data sets, and findings from this work can assist in interventions to help users successfully accomplish their learning goals. [For the full proceedings, see ED592609.]