The present study examined whether the strength capability of knee extensor muscles is associated with global cognitive function, assessed by Mini-Mental State Examination (MMSE), in non-disabled elderly men. Isometric torques during maximal voluntary knee extension, plantar flexion, and elbow flexion and MMSE scores were determined in 39 non-disabled men aged 61-79 years and used for the cross-sectional analysis examining the associations among the measured variables. Moreover, 27 of the subjects participated in a training program consisted of body mass-based exercises (sitting down onto and standing up from a chair, hip joint extension and flexion, calf raises, side leg raises, and trunk flexion and extension) 6 days a week for 3 months. Isometric torques and MMSE scores were determined after the intervention. Among the data before intervention, only knee extension torque (KET) and KET relative to body mass (KET/BM) significantly correlated to the MMSE scores: r = 0.579 (P < 0.0001) for KET and r = 0.520 (P < 0.001) for KET/BM. After the intervention, KET and KET/BM increased significantly, but MMSE score did not. However, the absolute change in MMSE scores was significantly associated with that in KET (r = 0.381, P < 0.05) and KET/BM (r = 0.422, P < 0.05). These findings indicate that the strength capability of knee extensors is associated with global cognitive function in non-disabled elderly men, and provide a new perspective to a general concept that exercises strengthening knee extensor muscles should be included in resistance training programs for elderly individuals.