The WWW is now in widespread use for delivering on-line learning content in many large-scale education settings. Given such widespread usage, it is feasible to accumulate data concerning the most useful learning experiences of past students and share them with future students. Browsing events that depict how past students utilized the learning content to accomplish higher levels of achievement are especially valuable. This paper presents a new method for identifying potentially effective browsing events based on a contextualized browsing model built through association mining and statistical techniques. The model annotates browsing events with several contextual factors, including educational ones (group relevance and performance relevance) and non-educational ones (support and confidence). Based on this model, a personalized content recommender was implemented in a Web-based learning content management system, called IDEAL, to deliver personalized learning content based on a student's browsing history. An experiment was conducted to compare the user feedback concerning the recommendations provided through different recommendation models. The results show that students with different levels of achievement prefer different types of contextualization information. Finally, another performance experiment demonstrated that the contextualized browsing model is more effective in improving learning performance than the pure association mining model. (Contains 6 tables and 10 figures.)