The objectives of this study were to (a) evaluate electromyography (EMG) repeatability during maximal voluntary isometric knee extension (MKE) and (b) develop a protocol yielding the most reproducible EMG. Subjects, 11 volunteers, performed 3 MKE (1.05 rad flexion) twice, separated by at least 5 days. EMG was recorded from vastus lateralis, vastus medialis, and rectus femoris. Root-mean-square (RMS) and integrated rectified (IEMG) values were determined across 4 averaging windows (W = 100,200,500,1000 ms). Coefficients of variation (CV) and intraclass correlations (ICC) described variability and reproducibility among trials and days. Error between days was assessed by the technical error of measurement (TEM) statistic. MKE was highly reproducible (ICC=0.93). For EMG, trial-to-trial variability was lower than day-to-day for all W, and wider W (i.e., 500 or 1000 ms) minimized variability. TEM was consistently greatest for W=100 ms, indicating that more error was introduced at narrow W. No differences in repeatability were found between RMS-EMG and IEMG. To summarize, EMG variability is minimized within days by increasing W. Day-to-day variability exceeds trial-to-trial and should be considered when reporting treatment effects. © 1997 National Strength & Conditioning Association.