Jerk-locked averaging (JLA) is used to record the timing and distribution of brain activity preceding brisk involuntary movements such as those observed in patients with myoclonus. JLA is capable of revealing a premyoclonus spike in the absence of paroxysmal activity in the routine EEG. The technique of JLA is described together with a consideration of the way in which the presence or absence of premyoclonus activity and the somatosensory evoked potential (SEP) can help to classify patients with involuntary movement disorders. The relationship between the premyoclonus spike and the N25-P33 complex of the SEP is discussed. Although the most common finding in patients with myoclonus of cortical origin is a large, well-defined positive-negative spike with the positive peak occurring about 20 ms before electromyographic onset, a wide variety of time relationships and topographies has been observed in other disorders of involuntary movement and some of these are described. The lack of correlation between the apparent origin of the premyoclonus spike in motor cortex and pathology in patients with myoclonus is discussed.