In this paper, we study the role of emotions and expressive behaviour in socially interactive characters employed in educational games. More specifically, on how we can use such emotional behaviour to help users to better understand the game state. An emotion model for these characters, which is mainly influenced by the current state of the game and is based on the emotivector anticipatory mechanism, was developed. We implemented the model in a social robot named iCat, using chess as the game scenario. The results of a preliminary evaluation suggested that the emotional behaviour embedded in the character indeed helped the users to have a better perception of the game. Categories and Subject Descriptors D.3.3 [Artficial Intelligence]: Robotics