The study investigated whether (1) instruction providing for experimental manipulation by students was superior to instruction in which teachers perform the manipulations and children observe, and (2) spatial ability was related to mode of instruction. Participating in the 8-week study were 141 students aged 12-15 in seven intact eighth-grade classes. To measure spatial ability, the Card Rotations Test and the Group Embedded Figures Test were used. Cognitive level for proportional logic was determined from a balance-beam task. A 42-item Geometry Achievement Test was constructed. The units studied focused on elementary concepts in geometry approached through transformations. The control classes did not study geometry until after the study ended. On the geometry posttest, the student-centered group had 80% of the items correct, compared to 68.6% for the teacher-centered group and 25.5% for the control group. Regression analysis indicated that students with high proportional logic ability achieved 5.5% more correct answers. Children with low spatial scores benefited more from the student-centered treatment. Other results were also presented. It was concluded that an inductive approach including active manipulation of objects helps foster understanding of transformations. (MNS)