Three components determine the power spectrum of the surface EMG signal: the auto- and cross-power spectra of the firing processes and the power spectra of the motor unit action potential (MUAP). To clarify the relative contribution of these components to the median frequency (MF) of the power spectrum, a stochastic simulation model was used in which most input parameters [e.g., MUAP peak-peak time (PPT), mean interpulse interval time, and synchronization parameters] were described in terms of distribution functions. Simulation clearly predicts that MF is especially sensitive to variations in MUAP shape, the MUAP PPT, and synchronization. The influence of the firing process parameters was predicted to be marginal. To obtain values for the MUAP parameters, a needle-triggered averaging technique was used to gather surface MUAPs from the m. biceps brachii. With use of these MUAPs as input for the model, it was found that intrasubject variability of MF is caused by variations in both MUAP PPT and MUAP shape, whereas intersubject variability in MF is caused primarily by variations in PPT.