This study determined the e€ects of a short period of knee isometric training on the quadriceps muscles accessible to surface electromyography (EMG). For this purpose, a training (n ˆ 9) and a control (n ˆ 7) group were tested on ®ve identical occasions at 1 week intervals during 4 weeks. The training group exercised three times a week by making isometric knee extensions at 80% of the maximal voluntary contraction (MVC). During the test sessions, maximal and submaximal torque and associated activations of the rectus femoris (RF), vastus lateralis (VL) and vastus medialis (VM) muscles were analysed. As a result of training, di€er-ences between MVC values of the two groups were highly signi®cant (P < 0.001), whereas only RF-EMG showed signi®cant di€erences (P < 0.05). The VL and VM did not present any signi®cant changes in maximal activation. The EMG±torque relationships were analysed individually before and after the training period. For the control subjects, EMG±torque relationships did not present signi®cant changes while for the training group, these relationships showed a signi®cant increase in RF, VL, and VM maximal activation in 6, 6 and 4 subjects, respectively, and a signi®cant decrease in 1, 2 and 5 subjects, respectively. In almost all cases, a sig-ni®cant downward shift of the relationship was observed. This study con®rmed that the parts of the quadriceps muscle tested present di€erent adaptation capacities and demonstrate inter-individual variability in the strategies used to enhance muscle strength. In conclusion , to analyse the neural e€ects resulting from training in a large and compartmentalized muscle like the quadriceps femoris, it is desirable to take into account each muscle independently. Moreover, we suggest that overall results obtained from the experiment population should be completed by an analysis on individuals .