Force (peak torque) of m. quadriceps femoris was measured during 60 repeated, voluntary dynamic knee extensions in 10 men before and after a 4-week training regimen of one-legged cycle exercise. Biopsies for histochemical analysis were obtained from the lateral vastus muscle after the training period. One leg was trained with the blood flow to the leg muscles reduced by local supra-atmospheric external pressure of 50 mmHg ('Ischaemic leg, I-leg'). Employing the same work-load profile the other leg was trained at normal atmospheric pressure ('Non-restricted-flow leg, N-leg'). In response to I-training, Maximum Peak Torqued (MPT; the highest torque produced in any contraction) and Initial Peak Torque (IPT; the average peak torque of the initial 12 contractions) decreased by 8% (P less than 0.01) and 9% (P less than 0.001), respectively. Final Peak Torque (FPT; the average peak torque of the final 12 contractions) increased by 13% (P less than 0.05) after I-training. No changes in MPT, IPT or FPT occurred following N-training. After training the proportion of slow-twitch fibres was higher (P less than 0.05) and the mean slow-twitch fibre area was larger (P less than 0.05) in the I-than in the N-trained leg. The results indicate that blood flow-restricted training, in contrast to non-restricted-flow training, decreases maximum voluntary dynamic force, possibly by inducing an increase in the share of the muscle cross-sectional area consisting of slow-twitch fibres. That flow-restricted training improves maintenance of force during short-term local exercise may reflect ischaemically induced changes in the metabolic characteristics of skeletal muscle.