The discrimination of fibrillation potentials and endplate potentials based on the conventional EMG methods is difficult if there is only discrete denervation. The reliability of discrimination can be highly improved if the frequency behaviour of the potentials is taken into account. The average proportional consecutive interval difference is the best discrimination variable with 90% correct findings per analysed potential sequence. When three analyses per muscle are made, the accuracy increases to 97.2%. The second best variable is the mean consecutive difference followed by interval bandwidth, standard deviation of interval, minimum interval and finally maximum interval. The mean duration of intervals does not allow of any differentiation. The frequency analysis can be restricted to spontaneous activity sequences of ten seconds. It is immaterial which part of a prolonged sequence is analysed for ten seconds. The conventional evaluation of registered spontaneous activity either alone or with observation at the monitor under simultaneous acoustic control is inferior to the frequency analysis. The results do not allow a statement of the probability of wrong diagnosis in clinical routine work.