The user's intention is reflected in not only the actual input action but the ones immediately before it as well. "P-Recognition" recognizes the preceding actions, and predicts the intention just when the actual action starts. This paper tests P-Recognition in a pen-based map navigation interface as an example, where the map is panned by user's dragging strokes and zoomed by user's enclosure by a circle. The feasibility of the proposal is confirmed in an experiment. We find that dragging and circling actions are distinguishable before the pen touches the screen. Moreover, for some users we can recognize their intention to write text. It is confirmed that the user's intention is present in the preceding actions and so is detectable.