Modeling and predicting students’ knowledge and behavior is an important task in an adaptive computer-based learning environment. The teaching decisions in such a system are based on the predictions of the so-called student model. A lot of work has focused on constructing models that are able to accurately represent and predict the knowledge of the student. In this talk, I will first give an overview of the most popular approaches to student modeling. I will present two widely used approaches based on probabilistic models and item response theory. I will then introduce more complex models with higher representational power. Finally, I will demonstrate, how general features can be integrated into a student model.