Research in game-based learning environments aims to recognise and show emotion. This chapter describes the main approaches and challenges involved in achieving these goals. In addition, we propose an emotional student model that can reason about students’ emotions using observable behaviour and responses to questions. Our model uses Control-Value Theory (Pekrun et al., The control value theory of achievement emotions. An integrative approach to emotions in education. In: Schutz, P.A., Pekrun, R. (eds.) Emotion in Education, pp. 13–36. Elsevier, London, 2007) as a basis for representing behaviour and was designed and evaluated using Probabilistic Relational Models (PRMs), Dynamic Bayesian Networks (DBNs) and Multinomial Logistic Regression. Olympia, a game-based learning architecture, was enhanced to incorporate affect and was used to develop PlayPhysics, an emotional game-based learning environment for teaching Physics. PlayPhysics’ design and emotional student model was evaluated with 79 students of Engineering at Tecnológico de Monterrey, Mexico City campus (ITESM-CCM). Results are presented and discussed. Future work will focus on conducting tests with a larger population of students, implementing additional game challenges and incorporating physiological signals to increase the accuracy of classification.