Electromyograms (EMGs) need to be normalized if comparisons are sought between trials when electrodes are reapplied, as well as between different muscles and individuals. The methods used to normalize EMGs recorded from healthy individuals have been appraised for more than a quarter of a century. Eight methods were identified and reviewed based on criteria relating to their ability to facilitate the comparison of EMGs. Such criteria included the magnitude and pattern of the normalized EMG, reliability, and inter-individual variability. If the aim is to reduce inter-individual variability, then the peak or mean EMG from the task under investigation should be used as the normalization reference value. However, the ability of such normalization methods to facilitate comparisons of EMGs is questionable. EMGs from MVCs can be as reliable as those from submaximal contractions, and do not appear to be affected by contraction mode or joint kinematics, particularly for the elbow flexors. Thus, the EMG from an isometric MVC is endorsed as a normalization reference value. Alternatively the EMG from a dynamic MVC can be used, although it is recognized that neither method is guaranteed to be able to reveal how active a muscle is in relation to its maximal activation capacity. Copyright © 2010 Elsevier Ltd. All rights reserved.