This study has important implications for microworlds such as Logo, HyperGami, and Newton's World, which use interaction to learn spatial mental models for science, math, geometry, etc. This study tested the hypothesis that interactively rotating (dragging) virtual shapes primes mental rotation. The independent variable was observation vs. interaction: a) watching an animation of a shape rotating, versus b) manually rotating a shape on the computer. The dependent variable was mental rotation of the same shape. Two age groups, 9-year-olds and college undergraduates participated. For 9-year-olds, the interactive group mentally rotated significantly more accurately and faster than the observational. Therefore, interaction primed mental rotation. For the college undergraduates, the interactive group mentally rotated significantly more accurately, but significantly slower than the observational group. This suggests that the interaction disrupted a routine process, causing undergraduates to switch strategies. Results from both age groups reinforce the educational value of more naturalistic interaction with virtual shapes, i.e., dragging is better than clicking.