A correction method for DC drift artifacts.

In order to cope with the problem of drift artifacts in ERP research a new off-line correction method is described. It estimates the DC drift from all prestimulus baselines of an experiment. For this end, an amplifier reset is performed at regular intervals and the DC off-set preceding any reset is stored. A regression is calculated between the prestimulus baseline amplitudes of consecutive trials and the time that has passed by. The amplitude trend which can be explained by either a linear or non-linear regression model is then subtracted from all data points. The power of the method is illustrated by two examples. The first shows that detrending increases the signal-to-noise ratio in ANOVA designs. The second shows that amplitude differences of event-related slow potentials, which appeared between the first and the second half of an experiment, could be explained by a non-linear drift component.