In this paper, we take the approach of carefully defining and manipulating information spaces (see Chapter 11 of [20] for an introduction) to address the task of tracking hidden targets in shadows. We use the abbreviations I-space for information space and I-state for information state. Since the shadow sequence is extracted from an observation history, a choice is made: Almost all of the sensor data is discarded; only a much condensed combinatorial sequence is left. The choice is not arbitrary; the general task of tracking unpredictable targets beyond the sensor range induces an equivalence relation over the workspace-time space that yields the space of shadows. In this section, we characterize how I-spaces, tasks, and filters are closely related.