Median method for detecting endogenous event-related brain potentials.

Most studies of event-related brain potentials (ERPs) use the averaging method. This procedure is assumed to increase signal/noise ratio in proportion to the square root of the number of trials if the signal is invariant and the noise is stationary. In addition, the averaged value is the appropriate measure of central tendency if the measurements have a Gaussian distribution. For endogenous ERPs, however, application of the averaging method encounters 3 serious problems. First, the invariance of signal cannot always be assumed because the endogenous signal reflects psychological processes which may be missing in some trials. Secondly, the Gaussian distribution cannot be assumed when a small number of trials is averaged. Thirdly, averaging a limited number of trials is sensitive to inclusion of occasional, non-random, noise. The characteristics of the median ERP can address all 3 problems. These characteristics are demonstrated by an original equation.