In this study we examined the effectiveness of a computer-based geometry learning tool with 4th grade students. In both an experimental and control condition children constructed a series of common four-sided figures to fit a set of visual constraints. Additionally, children in the experimental condition were required to validate the presence of parallel segments, congruent segments, or right angles embedded in their figures with the assistance of a visual depiction of the property’s spatial significance. Following training we conducted six identification tasks in which participants attempted to discriminate two valid members of a given polygon class from four displayed polygons. In all six instances, varying polygon, instruction, and feedback type, children in the experimental condition were more likely to correctly identify both class members than children in the control condition.