Six healthy women (21.8 +/- 0.4 y) did isometric strength training of the left plantarflexors at an ankle joint angle of 90 degrees. Training sessions, done 3 times per week for 6 weeks, consisted of 2 sets of ten 5 s maximal voluntary contractions. Prior to and following the training, and in random order, voluntary and evoked isometric contraction strength was measured at the training angle and at additional angles: 5 degrees, 10 degrees, 15 degrees, and 20 degrees intervals in the plantarflexion and dorsiflexion directions. Evoked contraction strength was measured as the peak torque of maximal twitch contractions of triceps surae. Training increased voluntary strength at the training angle and the two adjacent angles only (p less than 0.05). Time to peak twitch torque was not affected by training. Twitch half relaxation time increased after training (p = 0.013), but the increase was not specific to the training angle. There was a small (1.1%, p less than 0.05) increase in calf circumference after training. Evoked twitch torque did not increase significantly at any joint angle. It was therefore concluded that a neural mechanism is responsible for the specificity of joint angle observed in isometric training.