The effect of four amplitude normalization methods on intersubject variability of electromyographic (EMG) profiles in normal gait was examined. Bipolar silver/silver chloride surface electrodes were applied to the rectus femoris, vastus lateralis, biceps femoris, tibialis anterior, and soleus muscles of the right lower extremity, in 11 healthy subjects. The myoelectric signals were telemetered via an FM multichannel biotelemetry system, full-wave rectified and low-pass filtered, then A/D converted together with the footswitch signal. Within-subject ensemble-average patterns were generated from the linear envelope EMG of at least six strides for each subject. Each subject's ensemble average was then normalized to the following: (a) the average EMG over three 50% isometric maximum voluntary contractions (MVC), (b) the EMG per unit isometric moment of force, (c) the peak of the subject ensemble average, (d) the mean of the subject ensemble average. Intersubject variability was quantified for each of the normalization methods by the coefficient of variation (CV). The normalization to either the peak ensemble or the mean ensemble drastically reduced intersubject variability, by 12%-73%. In contrast, normalization to the average EMG during 50% MVC or to the EMG per unit moment increased intersubject variability. It was concluded that the reduction of intersubject variability by appropriate amplitude normalization is possible, thereby increasing the sensitivity of surface EMG as a diagnostic tool in gait analysis.