Training designed to support and strengthen higher-order mental abilities now often involves immersion in Virtual Reality (VR) where dangerous real world scenarios can be safely replicated. However, despite the growing popularity of VR to train cognitive skills such as decision-making and situation awareness, methods for evaluating their use rely heavily on subjective measures or analysis of final outcomes. Without strong performance measures the outcome of training in terms of both content knowledge and ability to transfer newly acquired skills to the real world is unknown. This paper discusses research directions in affective computing which suggest the relationship between affect and cognitive learning provides potential new approaches to ensure the processing of cognitions which occur prior to final outcomes are adequately evaluated. This review was undertaken in the wake of recent studies reporting the higher the level of affective intensity experienced in Virtual Reality the more effective the learning outcomes, indicating that affective state is a key variable for investigation in evaluation of training in VR.