Alterations of the electromyographic power spectrum have been studied extensively to assess fatigue development in the neuromuscular system. Usually, a data reduction has been applied to create an index based on the mean power frequency or the median frequency. The physiological origin of the spectrum alterations has been (and to some extent still is) incompletely known. However, during the 1980s, substantial progress has been made in this field. The factors affecting the electromyographic power spectrum discussed in this review are action potential velocity decrease, firing statistics alterations, action potential modification, muscle temperature, additional recruitment at fatigue, and force level. Their impact on three commonly used fatigue indexes, mean power frequency, median frequency, and zero crossing rate, is also reviewed.