This paper describes a user-oriented approach to a multimodal onboardcomputer system for use in cars. The intention of a prototype design was to demonstrate multimodality under “real” conditions, so most of the user-tests took place in a driving-simulator environment. The system features speech, gesture and manual input as well as a GUI and speech output and is developed in five steps with systematic user-tests. All the subjects’ experiences were taken into account when redesigns were accomplished.

The basic principle of the system is that the user can change the input modality at almost any time, but cannot use different channels simultaneously ("serial redundant multimodality").