To assess the effectiveness of ability grouping in a fall prevention structured exercise program for elderly people. We enrolled 124 subjects from among 2,582 elderly people aged 70 to 84 years living in the Tsurugaya district in Sendai City. Exclusion criteria were 1) motor fitness scale (MFS) score 9 points or more, 2) severe sensory, cognitive, or 3) physical disorders, and 4) nursing care grade 2 or more. Those ranked in the lower fourth and in the upper 3 fourths of the timed up and go test (TUGT) were each randomly assigned to 3 groups. Subjects in groups A and B had an exercise program for lower and higher fitness subjects separately, whereas all subjects in group C underwent a single exercise program. The exercise program, once a week for 12 weeks, consisted of strength and stability training. TUGT, lateral reach (LR), leg power per body weight and MFS were measured after the intervention and compared with the baseline values. There were no difference in the baseline characteristics among the groups. For group A, no significant changes in physical fitness measures, for group B a small but significant deterioration in LR, and for group C a small but significant deterioration in LR and TUGT were observed. MFS score improved significantly in all groups. Ability grouping appeared to be effective for a short-term exercise program in maintaining the physical ability, but the effectiveness did not reach statistical significance in the randomized controlled design.