Objective: The statistical method of motor unit number estimation (MUNE) assumes that all motor unit potentials (MUPs) have the same size. The present study aims to evaluate the consequences of this assumption as well as its implications for the validity of statistical MUNEs. Methods: We performed statistical and multiple point stimulation (MPS) MUNE with an array of 120 electrodes on the thenar muscles of 15 healthy subjects. These recordings allow isolation and quantification of the effect of non-uniform MUP size on MUNE, because the differences in submaximal CMAP size (and, hence, in MUNE) between electrodes are due almost entirely to differences in (summed) MUP size. Results: We found no correlation between statistical and MPS MUNEs. Statistical MUNEs proved very sensitive to small variations in the "bandwidth" (variance) of the response series; MUNEs from electrodes only 8 mm apart could deviate by as much as 60%. This variation in bandwidth resulted from spatial (and, hence, size) differences between the contributing MUPs. Conclusions: Statistical MUNEs are very sensitive to violation of the uniform MUP-size assumption, to an extent that blurs any correlation with MPS MUNE in healthy subjects. Significance: Statistical MUNE cannot be used to detect mild to moderate motor unit losses. © 2010 International Federation of Clinical Neurophysiology.