In order to understand and model the role of emotion in cognitive processes we attempt to integrate theoretical approaches originating from different disciplines in an implemented cognitive architecture for embodied agents. Our virtual humanoid agent Max employs this architecture to generate believable human-like behaviors in a variety of situational contexts. In this article, we describe how we go about endowing Max’s architecture with increasingly elaborated kinds of emotions – from primary emotions like happiness and fear, toward secondary emotions like hope and relief.