The objective of the present study was to evaluate the variability of the surface EMG signal of the same muscle in healthy subjects, because of lack of reproducibility of the EMG signal for the same subject and muscle in different trials of maximal isometric voluntary contraction. The results showed an EMG coefficient of variability of 21.61%, indicating that this variability must be considered in experiments with an inappropriate condition for normalization procedures, such as EMG biofeedback in rehabilitation sessions, or normalization procedures by the maximal isometric voluntary contraction.