Reproducibility of the spike-triggered averaging technique of motor unit estimation (MUE) was assessed in biceps-brachialis muscle in 10 normal subjects and 15 subjects with amyotrophic lateral sclerosis (ALS). MUE was calculated by dividing the compound muscle action potential by the mean amplitude of 15 surface motor unit potentials (S-MUPs) of low recruitment threshold. Averaged MUE values in normal subjects were higher than in ALS subjects, with few values overlapping. Differences between test and retest MUE values were not significant for either subject group. The relative differences between test-retest values were 45.3% for normal subjects and 32.6% for ALS subjects. Correlation coefficients between test and retest values were low (r = 0.07) for normal subjects when influential outlying points were removed, and higher ( r = 0.65) for ALS subjects when individuals with MUE values within the normal range were removed. The higher correlation of testretest MUE values in ALS subjects compared to normal subjects may be due to a greater probability of resampling among the smaller number of motor units in ALS subjects. In summary, the reproducibility and technical aspects of the spike-triggered averaging technique are similar to those reported for other MUE techniques.