%675$&7The presentation of information using multiple modalities influences the perception of users, their comfort, and their performance in using a computer-based information system. This paper presents a user study investigating the effects of different output modality-combinations on the effectiveness to transport information and on the user's acceptance of the system. We chose a tourist information system as a test environment and conducted the study on three different devices (PDA, TV set, and desktop computer) to investigate whether the best modality-combination depends on the used device. It turned out that the modality-combination of spoken text in connection with a picture was the most effective regarding recall-performance. This effect was strongest for users working with PDAs, which can be explained by the cognitive load theory. In contrast to this, participants ranked different modality combinations as most appealing, namely those with written text..H\ZRUGV user studies, user interfaces, multi-modal, device adaptation, cognitive load theory.,1752'8&7,21The design of intelligent user interfaces becomes increasingly important for software designers and also for the design of consumer electronic devices. This process is due to the fact that software becomes more complex and systems become integrated into a ubiquitous information infrastructure. The paradigm of "one device -one functionality" is over and today we can access mutually any service through any device. A TV set for instance is nowadays a control center for various applications, e.g., video programming, e-commerce, Web-browsing and more. A mobile PDA can be used as a telephone, tourist guide, remote control, calendar, or Web-access device. In the EMBASSI project [2], we are working on multi-modal user interaction in a private household scenario where users access information through different terminal devices such as TV sets, desktop computers, and PDA. The question emerges, how to present the same information on these different device types in order to assist and also to please the user in the best way.Insights to human processing of multi-modal information can be gained from so-called dual task experiments, in which two different tasks have to be solved at the same time. In general, people perform better when the two tasks are rather different than similar (e.g.[5]). A neuropsychological account for this could be the separation of different modality-specific areas (e.g. for auditory presented and visually presented texts) in the working memory of the brain (e.g.[5]).The cognitive load theory of Sweller et al.[6] explains this using Baddeley's model [7] where two separate sub-systems for visual and auditory memory work relatively independent. The load can be reduced when both subsystems are active compared to processing all information in a single sub-system. Due to this reduced load, more resources are available for processing the information in more depth and thus for storing in long-term memory. Therefore, they propose to use different moda...