In the field of artificial intelligence the production of believable emotions are vital to be able to produce believable behavior of virtual agents. This is done with a process called affective appraisal, which means that events and situations are appraised and emotions are produced accordingly. The Artificial Intelligence and Computer Graphics (AICG) lab at Linköpings University has been devel- oping an AI architecture for virtual agents. This architecture had an appraisal module in need of improvement. This M.Sc. thesis had the purpose of doing this. Several approaches to affective appraisal are discussed and compared and finally one approach, called the OCC model, is chosen for implementation. This model is suitable for a real-time AI architecture as it is simple, easy to implement and can produce a wide range of emotions. The implementation of the OCC model is described in terms of how its different parts are incorporated into the previously existing AI architecture. Three extensions to the OCC model are also implemented to improve the results: emotional memories, the appraisal of unexpected events and interaction between the produced emotions. Finally the implementation is tested and the results of the tests are discussed. It is found that the implementation produces sufficient results for the scope of the thesis and for the requirements of the AI architecture into which it is incorporated.