We investigate how to reveal an initial mid-air gesture on interactive public displays. This initial gesture can serve as gesture registration for advanced operations. We propose three strategies to reveal the initial gesture: spatial division, temporal division, and integration. Spatial division permanently shows the gesture on a dedicated screen area. Temporal division interrupts the application to reveal the gesture. Integration embeds gesture hints directly in the application. We also propose a novel initial gesture called Teapot to illustrate our strategies. Our main findings from a laboratory and field study are: A large percentage of all users execute the gesture, especially with spatial division (56%). Users intuitively discover a gesture vocabulary by exploring variations of the Teapot gesture by themselves, as well as by imitating and ex-
tending other users’ variations.