This thesis addresses two aspects related to enabling systems to recognize the affective state of people and respond sensibly to it. First, the issue of representing affective states and unambiguously assigning physiological measurements to those is addressed by suggesting a new approach based on the dimensional emotion model of valence and arousal. Second, the issue of sensing affect-related physiological data is addressed by suggesting a concept for physiological sensor systems that live up to the requirements of adaptive, user-centred systems.