<b><i>Background/Aims:</i></b> Walking speed (WS) is an objective measure of physical capacity and a modifiable risk factor of morbidity and mortality in the elderly. In this study, we (i) determined effects of 3-month supervised aerobic-strength training on WS, muscle strength, and habitual physical activity; (ii) evaluated capacity of long-term (21 months) training to sustain higher WS; and (iii) identified determinants of WS in the elderly. <b><i>Methods:</i></b> Volunteers (F 48/M 14, 68.4 ± 7.1 years) completed either 3-month aerobic-strength (3 × 1 h/week, <i>n</i> = 48) or stretching (active control, <i>n</i> = 14) intervention (study A). Thirty-one individuals (F 24/M 7) from study A continued in supervised aerobic-strength training (2 × 1 h/week, 21 months) and 6 (F 5/M 1) became nonexercising controls. <b><i>Results:</i></b> Three-month aerobic-strength training increased preferred and maximal WS (10-m walk test, <i>p</i> < 0.01), muscle strength (<i>p</i> < 0.01) and torque (<i>p</i> < 0.01) at knee extension, and 24-h habitual physical activity (<i>p</i> < 0.001), while stretching increased only preferred WS (<i>p</i> < 0.03). Effect of training on maximal WS was most prominent in individuals with baseline WS between 1.85 and 2.30 m·s<sup>−1</sup>. Maximal WS measured before intervention correlated negatively with age (<i>r</i> = −0.339, <i>p</i> = 0.007), but this correlation was weakened by the intervention (<i>r</i> = −0.238, <i>p</i> = 0.06). WS progressively increased within the first 9 months of aerobic-strength training (<i>p</i> < 0.001) and remained elevated during 21-month intervention (<i>p</i> < 0.01). Cerebellar gray matter volume (MRI) was positively associated with maximal (<i>r</i> = 0.54; <i>p</i> < 0.0001) but not preferred WS and explained >26% of its variability, while age had only minor effect. <b><i>Conclusions:</i></b> Supervised aerobic-strength training increased WS, strength, and dynamics of voluntary knee extension as well as habitual physical activity in older individuals. Favorable changes in WS were sustainable over the 21-month period by a lower dose of aerobic-strength training. Training effects on WS were not limited by age, and cerebellar cortex volume was the key determinant of WS.