We evaluated possible methods of normalising EMG measured during running. MVC, Sprint and 70% Peak Running Speed methods were evaluated and their repeatability, reliability and sensitivity to incremental running speed were compared. Twelve runners performed the same experimental protocol on three separate occasions. Each day, subjects firstly performed MVCs, followed by a 20 m maximal sprint (with a 20–30 m run-up). Following this, they performed the peak running speed (PRS) test until exhaustion. After which they ran at 70% of PRS for 5 laps. Results indicated that normalising EMG data to MVC and Sprint methods are more repeatable for VM, BF, MG and RF, VL, LG, respectively, with the average ICC > 0.80. The 70% PRS demonstrated poor to fair levels of repeatability ranging between ICC 0.27 and 0.70. Whereas the 70% PRS method had the least intra-subject variability and the greatest sensitivity to increasing running speeds. More specifically, demonstrating significant changes in muscle activity in VM with increasing running speed while MVC and Sprint methods were unable to detect these changes. The dynamic methods were the most appropriate for EMG normalisation showing repeatability, better intra-subject reliability and better sensitivity during running over different days and for once-off measurements.