We present main effects of treatment and moderating effects of student background on science and engineering outcomes for grades 3-5 children participating in an engineering intervention. Data is from a large-scale randomized-controlled trial of two matched pairs (intervention and comparison) of engineering units. The intervention curriculum utilizes methods of project-based learning with support to apply science concepts to engineering design, while the comparison curriculum relies primarily on direct instruction. Both groups were taught the same science content prior to treatment. Intervention students in one of two curriculum units showed greater improvement on science content outcomes, as compared to control. The white/minority science achievement gap shrank after the intervention. Results suggest that the project-based intervention improves science achievement, particularly for minorities.