This study sought to determine whether aerobic exercise training affects cognitive functioning among semi-independent (> 75 years) and independent elderly people. Seventy-six semi-independent and independent elderly subjects were divided into the following groups: semi-independent control (n = 16); semi-independent training (n = 13); independent control (n = 22); independent training (n = 25). During the 12-week intervention, subjects in both training groups performed a bench-stepping exercise at the intensity of the lactate threshold. Subjects were assessed at baseline and post-intervention using the Frontal Assessment Battery test (FAB), the Modified Mini-Mental State Exam (3MS), and a bench-stepping test of aerobic capacity. During the intervention, FAB scores improved only in the semi-independent training group (p = .002), while 3MS and MMSE scores improved in both the semi-independent training (p = .032 and p = .004) and independent training groups (p = .001 and p = .013). FAB and 3MS scores were higher in the semi-independent training (p = .001 and p = .002) compared with the semi-independent control group after the intervention. Finally, post-intervention scores for FAB, and 3MS in the semi-independent training group almost reached the baseline levels of the independent control and independent training groups. These results indicate that moderate intensity bench-stepping exercise training can improve cognitive and frontal lobe functioning in semi-independent elderly people, almost to the level exhibited by independent elderly people.