We propose a new model for learning that relates video watching behavior and engagement to quiz performance. In our model, a learner's knowledge gain from watching a lecture video is treated as proportional to their latent engagement level, and the learner's engagement is in turn dictated by a set of behavioral features we propose that quantify the learner's interaction with the lecture video. A learner's latent concept knowledge is assumed to dictate their observed performance on in-video quiz questions. One of the advantages of our method for determining engagement is that it can be done entirely within standard online learning platforms, serving as a more universal and less invasive alternative to existing measures of engagement that require the use of external devices. We evaluate our method on a real-world massive open online course (MOOC) dataset, from which we find that it achieves high quality in terms of predicting unobserved first-attempt quiz responses, outperforming two state-of-the-art baseline algorithms on all metrics and dataset partitions tested. We also find that our model enables the identification of key behavioral features (e.g., larger numbers of pauses and rewinds, and smaller numbers of fast forwards) that are correlated with higher learner engagement. [For the full proceedings, see ED596512.]