The artifact bestiary

You are skimming: the title, the first figure, and the short version. Switch to Read in the header for the full page, or Deep to open every deep dive.

The specimens

Blink. A smooth, rounded deflection, 100 to 300 µV, about 300 milliseconds wide, largest at the forehead (Fp1, Fp2) and shrinking toward the back, same sign in both frontal channels. It is the eyeball rotating upward under the closing lid, which moves the eye’s standing dipole. Nobody blinks less than about ten times a minute, so a recording without blinks is a recording with a bad frontal electrode. At the source: ask people to blink freely between trials and hold still during them; never ask them not to blink, which produces a tense person who blinks anyway.

Eye movement. A step, not a bump, on the channels beside the eyes, opposite signs left and right for a horizontal look. Held gaze holds the step. Slow drifts of gaze look like slow drifts of baseline. At the source: a fixation point.

Jaw and face muscle. A burst of dense, spiky, high-frequency activity, mostly above 20 Hz, largest at the temples and back of the head where the temporalis and neck muscles are. Chewing, talking, swallowing, frowning, and the tense clench of someone trying too hard. It overlaps the beta and gamma bands and cannot be filtered away without losing them. At the source: relax the jaw, rest the head, and do not record a person who is speaking.

Heart. A small, sharp spike about once a second, regular, most visible near the ears and on the reference, tiny at the vertex. It is the ECG’s QRS complex, a millivolt on the chest leaking to the head as a few microvolts. In an averaged ERP it can appear as a fake component if the heartbeat happens to line up with the stimulus. At the source: mostly you cannot; you learn to see it, and ICA removes it in Phase 2.

Sweat and drift. A slow wander of the whole baseline over seconds to minutes, on one or all channels. Sweat changes the skin’s chemistry under the electrode and shifts its offset; so does the paste slowly settling. At the source: a cool room, time for the electrodes to settle, and a 0.5 Hz high-pass.

Electrode pop. A sudden step or spike on one channel, followed by a slow exponential recovery, with nothing on the neighbours. The electrode’s contact changed: a cup lifted, the paste cracked, a lead was tugged. At the source: fresh paste, tape, and leads that do not pull.

Hum. A steady sinusoid at 60 Hz (50 outside North America), on every channel, that changes size when you move a lamp cord. See why the hum for the ordered list.

Movement. Everything at once: a large low-frequency swing plus muscle plus a pop, as the head moves and the electrodes shift. Recognizable because it is on all channels and ugly. At the source: a comfortable chair and a short recording.

Pulse. A slow rhythmic wobble at the heart rate, on one channel, when an electrode sits on an artery (the temples are the usual place). It is the vessel moving the electrode. Distinguish it from the heart spike by its shape: smooth, not sharp. At the source: move the electrode a centimetre.

Reference contamination. Every channel shows the same artifact at the same time, whatever it is. That means it is on the reference electrode, which is subtracted from every channel. At the source: a reference somewhere quiet, and a linked-ears or average reference in Phase 2.

How to tell brain from not-brain

Three questions. Does it have the right shape? Brain rhythms wax and wane; artifacts are steps, spikes, or steady tones. Does it have the right geography? Alpha is at the back, mu is at the top, blinks are at the front, muscle is at the edges, and an artifact on one channel alone is an electrode. Does it happen at the right time? Anything locked to a blink, a heartbeat, a swallow, or the moment the subject moved is not the response you were looking for.