AC amplifiers can introduce significant distortions into the low frequency and DC components of recorded electrophysiological data such as event-related potentials (ERPs). Methods for correcting such distortions (i.e., estimating the waveform of the original data) after the data have been amplified and recorded rely on an accurate estimate of the amplifier's time constant (TC). We show that the filter characteristics of AC amplifiers in at least some commercially available ERP recording instruments may vary considerably across individual channels, even when each houses an identical AC amplifier circuit. Clearly, distortion correction methods must take this variability into account. We propose an empirical means of estimating the correct TC value. This approach yields more accurate correction than those based on TCs calculated analytically.