High-pass filters and baseline correction in M/EEG analysis. Commentary on: "How inappropriate high-pass filters can produce artefacts and incorrect conclusions in ERP studies of language and cognition"

J Neurosci Methods. 2016 Jun 15;266:164-5. doi: 10.1016/j.jneumeth.2015.12.003. Epub 2016 Jan 8.


Tanner et al. (2015. Psychophysiology, 52(8), 1009. doi: 10.1111/psyp.12437) convincingly demonstrate how a late deflection like the N400 or the P600 is reflected into both earlier and later latencies by the application of high-pass filters with cutoff frequencies higher than 0.1Hz. It nicely underlines the importance of test-wise application of filters with different parameters to electrophysiological data to identify such unwanted filter effects. In general, we agree with their approach and conclusions, particularly with the notions that the application of a high-pass filter is reasonable if it improves the signal-to-noise ratio (SNR) of the signal of interest, and that low frequency signals may carry important information. However, we disagree in two aspects: First, the test data of Tanner et al. are not optimally suited to demonstrate the benefits of high-pass filtering as they are only minimally contaminated by low frequency noise, and second, the standard baseline correction for particular applications in M/EEG data analysis should be replaced with high-pass filtering-as recommended by Widmann et al. (2015. J Neurosci Methods, 250, 46. doi: 10.1016/j.jneumeth.2014.08.002).

Keywords: Baseline correction; EEG; ERF; ERP; High-pass filter; MEG.

Publication types

  • Comment

MeSH terms

  • Artifacts*
  • Cognition
  • Electroencephalography
  • Evoked Potentials
  • Female
  • Humans
  • Language*
  • Male
  • Signal Processing, Computer-Assisted