Intelligent multimodal presentation (IMMP) systems are able to generate multimodal presentations adaptively, based on the run-time requirements of user-computer interaction. Modality allocation in IMMP system needs to adapt the modality choice to changes in various relevant factors, such as the type of information to be conveyed, the presentation goal, the characteristics of the available modalities, the user profile, the condition of the environment, and the type of user task. In this study, we emphasize that modality allocation in IMMP systems should also take into account the cognitive impacts of modality on human information processing. We first describe several modality-related cognitive and neuropsychological findings. Then a user study is presented to demonstrate the effects of modality on performance, cognitive load and stress, using a high-load and time-critical user task. Finally, we show a possible way to integrate relevant cognitive theories into a computational model that can systematically predict the suitability of a modality choice for a given presentation task.