We address the problem of interpreting sensor data under uncertainty, using temporal and spatial context to facilitate the identification of objects. We seek to identify the type of an object presented in an ambiguous image by reasoning about conditional probabilities and the possible movements objects can make. A conditional probability (that an object is of a certain type given that some of its properties have been recognized) is used in conflict resolution, and an object is assigned an alternative type when an impossible movement is detected. Think of a map as being a frame and a sequence of frames as being a film. The idea is to construct a consistent and plausible (coherent and highly probable) film in which an object of one type does not mysteriously change into an object of another type.