In this work we present an approach to emotional reasoning for believable agents, by introducing a mechanism to progressively build a map of knowledge for reasoning. We present the notion of inference graph for progressive reasoning in an emotional context. In this model, knowledge is partially highlighted and noticed by the agent.