Building on Buxton’s foreground/background model, we discuss the importance of explicitly considering both foreground interaction and background interaction, as well as transitions between foreground and background, in the design and implementation of sensing techniques for sensor-enhanced mobile devices. Our view is that the foreground concerns deliberate user activity where the user is attending to the device, while the background is the realm of inattention or split attention, using naturally occurring user activity as an input that allows the device to infer or anticipate user needs. The five questions for sensing systems of Bellotti et al.  proposed as a framework for this special issue, primarily address the foreground, but neglect critical issues with background sensing. To support our perspective, we discuss a variety of foreground and background sensing techniques that we have implemented for sensor-enhanced mobile devices, such as powering on the device when the user picks it up, sensing when the user is holding the device to his ear, automatically switching between portrait and landscape display orientations depending on how the user is holding the device, and scrolling the display using tilt. We also contribute system architecture issues, such as using the foreground/background model to handle cross-talk between multiple sensor-based interaction techniques, and theoretical perspectives, such as a classification of recognition errors based on explicitly considering transitions between the foreground and background. Based on our experiences, we propose design issues and lessons learned for foreground/background sensing systems.