AbstractIn fields such as rehabilitation and sports science, it is important to evaluate the state of muscle fatigue exactly in order to realize the maximum training effect while avoiding damage due to overtraining. Since the surface EMG is noninvasive and measurement is easy, the EMG evaluation parameters calculated from EMG are widely used for evaluation. It should be noted, however, that the behavior of the evaluation parameters is similar in true muscle fatigue and in cases where muscle fatigue is simulated by intentionally decreasing the muscle tension. This makes discrimination of the two cases difficult. To recognize the difference between the states of muscle activity, the authors proposed a method in which the time series of the EMG evaluation parameters is represented by a Karhunen–Loève (KL) expansion, but the physiological condition which is expressed by the result of KL expansion analysis has not been sufficiently investigated. This study uses a surface EMG simulation model in which the firing pattern of the motor units (MU) can be varied flexibly, and attempts to clarify how the MU firing pattern affects the results of analysis by KL expansion. It is found that the same analytical results as for the state of muscle fatigue can be obtained by considering the fluctuation of the firing probability in multiple MU and the accompanying intermittent MU recruitment. © 2003 Wiley Periodicals, Inc. Syst Comp Jpn, 34(12): 45–55, 2003; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/scj.10326