Skip to content

Chapter 3: Preprocessing

Continuing from Chapter 2 — Importing and Inspecting Data.

The steps in this chapter clean up the continuous recording before it's cut into epochs — the same goal as the MNE-Python track's Preprocessing chapter, reached through EEGLAB's Tools menu instead of code.

Filtering

Tools → Filter the data → Basic FIR filter (new, default) applies EEGLAB's standard filter (pop_eegfiltnew under the hood) — enter a low cutoff, a high cutoff, or both, exactly like setting l_freq/h_freq in raw.filter() on the Python side. A separate notch filter option in the same menu removes electrical line noise (50 or 60 Hz depending on where the data was recorded).

Re-referencing

Tools → Re-reference the data lets you switch the recording's reference: average reference (every channel's mean, subtracted from every channel), a single channel, or the average of a specific pair (commonly the two mastoids). This is the direct GUI equivalent of MNE's raw.set_eeg_reference().

Bad channels

A channel that's flat, saturated, or dominated by noise should usually be excluded rather than dragging down every analysis downstream:

  • Tools → Reject data using... → Reject channels for manual, visual selection, or an automated criterion (e.g. Clean Rigor, installed as a plugin) that flags outlier channels statistically.
  • Once a channel is marked bad, Tools → Interpolate electrodes reconstructs it from its neighbors — useful when you'd rather keep a full, regular channel set for later group averaging than simply drop the channel.

Resampling

Tools → Change sampling rate downsamples the recording — smaller files and faster processing, at the cost of the very highest frequencies. A common choice is resampling to somewhere around 250–500 Hz for standard EEG/ERP work, well above where most signal of interest lives.

A note on reproducibility: EEG.history

Every GUI operation you run is silently logged as an equivalent line of MATLAB code in EEG.history. File → Save history (or Edit history) exports that log as a runnable .m script — meaning a pipeline you clicked together in the GUI once can be replayed exactly, or adapted into a script that runs unattended over many subjects. This is EEGLAB's built-in bridge between "exploring by hand" and "scripted and reproducible," worth keeping in mind heading into Chapter 7.

Next: Chapter 4 — Artifact Removal with ICA