Spectral Analysis (Spectrogram and Coherence)

Spectral analysis describes a field potential by frequency: a power spectrum gives the power in each band, a spectrogram shows how that power changes over time, and coherence measures how strongly two signals are correlated at each frequency, from 0 to 1. The conventional bands are delta (1 to 4 Hz), theta (4 to 8), alpha (8 to 13), beta (13 to 30) and gamma (above 30), and a spectrogram's window length, typically 50 to 500 ms, sets its time resolution against its frequency resolution. Coherence is the cross-spectrum divided by the two auto-spectra and needs trial averaging to be reliable.

What the electrode changes is what a channel's spectrum means. Power is summed over the tissue a site averages, so a large site or a surface grid gives a cleaner, more regional spectrum and a small site a more local one; and a distant reference ↗ adds the reference's own rhythms to every channel, so coherence between two channels that share a reference can be the reference rather than the brain. Bregma-aligned grids, fixed session structure and staging from an independent signal are what make spectral and coherence differences comparable across animals, as the rat EEG study ↗ shows. Spectrograms are interactive in Radiens Videre.