This investigation tracked changes in categorical bias (i.e., placing objects belonging to the same spatial group closer together than they really are) while 7-, 9-, and 11-year-olds and adults were learning a set of locations. Participants learned the locations of 20 objects marked by dots on the floor of an open square box divided into quadrants. At test, participants attempted to place the objects in the correct locations without the dots and boundaries. In Experiment 1, we probed categorical bias during learning by alternating learning and test trials. Categorical bias was high during the first test trial and decreased over the second and third test trials. In Experiment 2, we manipulated opportunities for learning by providing participants with either one, two, three, or four learning trials prior to test. Participants who experienced one or two learning trials exhibited more bias at test than did those who experienced four learning trials. The discussion focuses on how categorical bias emerges through interactions between the cognitive system and task structure.