Use of the Median Method to Enhance Detection of the Mismatch Negativity in the Responses of Individual Listeners

Abstract

The median method was evaluated as an alternative way of expressing the mismatch negativity (MMN). Traditionally, signal averaging has been used to extract these event-related potentials from unwanted background noise. However, mean values are biased by unrejected artifact that skews the relatively small distribution of values on which the MMN is based. Because the median is a more valid measure of central tendency in asymmetric distributions, it may describe MMN data more accurately. Better representation of the signal in the median waveform might enhance detection of the MMN in the responses of individual listeners. Mean and median waveforms were computed from previously recorded MMN data. Visually identified MMNs were validated using area and onset latency criteria. Detectability of the MMN was not improved using median waveforms. Despite this result, a theoretical argument for use of the median is presented.