Students (n = 797) from 36 4th-grade classrooms were taught the control of variables strategy for designing experiments. In the instruct condition, classes were taught in an interactive lecture format. In the manipulate condition, students worked in groups to design and run experiments to determine the effects of four variables. In the both condition, classes received the interactive lecture and also designed and ran experiments. We assessed students' understanding using a written test of their ability to distinguish valid from invalid experimental comparisons. Performance on this test improved from the pretest to the immediate posttest in all conditions, and gains were maintained at a 5-month delay. For students from both higher and lower achieving schools, gains ordered as follows: both greater than instruct greater than manipulate. However, students from higher achieving schools showed greater gains in all conditions. Item analyses showed that the interactive lecture improved students' understanding of the need to control irrelevant variables, and experimentation improved students' understanding of the need to vary the focal variable. (Contains 4 tables, 2 figures and 1 footnote.)