This study investigated changes in elderly muscle joint angle–torque relation induced by resistance training. Older adults were assigned to either training (n= 9, age 74.3 ± 3.5 years; mean ±s.d.) or to control groups (n= 9, age 67.1 ± 2 years). Leg-extension and leg-press exercises were performed three times per week for 14 weeks. Maximal isometric knee extension torque was measured across the knee joint angle range of movement. Vastus lateralis muscle architecture was examined in vivo using ultrasonography. The vastus lateralis muscle fascicle force was estimated from the measured joint torque, enabling construction of the fascicle length–force relation. Electromyographic (EMG) activity was measured from representative agonist and antagonist muscles. Training altered the angle–torque relation: (a) displacing it by 9–31% towards higher torque values (P < 0.05); and (b) shifting the optimal angle from 70 deg (corresponding torque: 121.4 ± 61 N m) before to 60 deg (134.2 ± 57.2 N m; P < 0.05) after training. Training also altered the fascicle length–force relation: (a) displacing it by 11–35% towards higher force values; and (b) shifting the optimal fascicle length from 83.7 ± 8 mm (corresponding force: 847.9 ± 365.3 N) before to 93.2 ± 12.5 mm (939.3 ± 347.8 N; P < 0.01) after training. The upward displacement of the angle–torque relation was mainly due to a training-induced increase in agonist activation, whilst the shift in the optimal angle was associated with changes in muscle-tendon properties.