Research on children’s categorization presents seemingly paradoxical results: Presenting exemplars at the same time (simultaneously) and presenting exemplars apart in time (spaced) have both been argued to support learning. This research was designed to explain these results by examining the visual attention and forgetting dynamics underlying various presentation schedules. Across three experiments, preschool-aged children ( N = 292) were presented with science category exemplars on simultaneous, massed, and spaced schedules. The first experiment revealed that children had the strongest generalization performance in the spaced condition at the delayed post-test. In subsequent experiments, children visually attended less and forgot more during spaced learning. These results are discussed in the context of several theoretical accounts in cognitive science and applied implications for science education.