Electromyography (EMG) is a commonly used tool that can be plagued with poor signal-to-noise ratios. One result of poor signal-to-noise ratios is increased within- and between-subject variability of quantified EMG variables, for example, the integrated EMG. Methods that reduce within- and between-subject variability of quantified EMG variables can increase the statistical power of an experimental design and aid in the functional interpretation of experimental results. The purpose of this investigation was to determine the effectiveness of spatially averaging the surface EMG signal to reduce the variability of the quantified EMG obtained during maximum voluntary contractions (MVC). The present study extends the work of earlier investigators describing the enhanced signal characteristics obtained by spatially averaging the surface EMG measured during submaximum voluntary isometric contractions and stretch reflexes. Ten subjects performed maximum voluntary isometric and anisometric (concentric and eccentric) contractions of the elbow flexors. Four electrodes, forming two pairs of bipolar electrodes were placed over both the biceps brachii and brachioradialis muscles. Four rectified and integrated EMG signals from the electrode array were compared. Data from each subject's contraction condition and from each muscle were used to compute a coefficient of variation that was considered representative of the within-subject variability. These data were analysed with a multifactorial repeated measures analysis of variance (ANOVA). The results revealed a muscle-specific, statistically significant superiority of one of the methods in reducing the variability of the rectified and integrated EMG signal. Summing the rectified and integrate signals from each bipolar pair of electrodes in the array was shown to reduce significantly the within-subject variability.