The power spectrum of eye-related component

Hi Mike,

One reviewer wants me to check if an observed ERP effect is related to noise due to eye blinks. He/She suggested checking the power spectrum of the effect and if the effect is an eye-related component, we should see the power spectrum concentrated at a low frequency below 5 Hz.
I’m confused about this suggestion. Isn’t the power spectrum supposed to concentrate at low frequencies because of the 1/f law? Should I compare the power spectrum of the effect to that of something else (such that the 1/f trend is subtracted out)? I wish I could get your input. Thank you

The brain’s power spectrum does concentrate at low frequencies because of 1/f, but the eye-blink component does as well and is much stronger. The eye-blink component can be removed using ICA and/or using LOC/ROC electrodes.

Agreed – eyeblinks do have a low-frequency spectrum, but the presence of energy at low frequencies doesn’t mean eyeblinks. Also, ERPs are trivially dominated by low frequencies because averaging is a low-pass filter.

@Chao perhaps an idea is to run ICA on your cleaned data, see if you find an eyeblink component, and then plot its ERP. Alternatively (or additionally), if you recorded EOG channels, then compute the VEOG bipolar and show its ERP.

1 Like

Thank you, Steve and Mike. Your inputs are quite hopeful!