The electroencephalogram (EEG) measures tiny voltage fluctuations at the scalp — on the order of tens of microvolts — produced by the synchronized post-synaptic activity of large populations of cortical neurons. The raw signal is dominated by rhythmic activity, and by convention we divide that rhythmic activity into five frequency bands.
The bands are useful because different physiological and cognitive states tend to change the power in specific frequency ranges. They are not, however, discrete biological entities. They are windows we draw on a continuous spectrum.
The five conventional bands
Delta (~0.5–4 Hz)
The slowest and highest-amplitude rhythms. Dominant during deep, non-REM (slow-wave) sleep in healthy adults. Prominent delta in a waking adult record is unusual and often reflects artifact (movement, sweat, drift) rather than brain activity.
Theta (~4–8 Hz)
Associated with drowsiness, light sleep, some meditative states, and in children, normal waking activity. Frontal midline theta is a well-studied correlate of sustained attention and working-memory load in adults.
Alpha (~8–13 Hz)
The classic "relaxed wakefulness" rhythm, first described by Hans Berger. Strongest over posterior (occipital and parietal) sites, and strongest with the eyes closed — opening the eyes typically attenuates alpha, a phenomenon called alpha blocking or alpha desynchronization.
Beta (~13–30 Hz)
Associated with active thinking, focused attention, and motor activity. Beta is normally lower in amplitude than alpha. Certain medications (benzodiazepines, barbiturates) elevate beta.
Gamma (~30–50 Hz and above)
The fastest conventional band, tied to high-level perceptual and cognitive binding. Gamma is genuinely interesting — and genuinely hard to measure at the scalp. Scalp EMG (muscle activity from the face and neck) sits in the same frequency range, and low-amplitude cortical gamma is easily swamped by it. Any gamma result from standard scalp EEG needs artifact rejection you can defend.
There is no biological line between "alpha" and "beta." The boundaries are historical conventions, and they vary between textbooks, labs, and software packages. You will see alpha reported as 8–12 Hz in one paper and 8–13 Hz in another. Some pipelines split beta into low, mid, and high sub-bands. Others treat 12–15 Hz as a sensorimotor rhythm (SMR) of its own.
This matters more than it sounds. If your configuration is Alpha 8–12 and Beta 13–30, the slice from 12 to 13 Hz belongs to no band. It is silently excluded from your "total power" sum, so every relative band power percentage is computed against an incomplete denominator. Nothing in the output flags it. The numbers look reasonable and are quietly wrong.
The same happens in reverse if adjacent bands overlap — the shared Hz gets counted twice. Before you trust a band-power table, confirm that your bands tile the range you care about exactly once.
Individual differences matter too
The individual alpha peak frequency (IAF) — the frequency where a person's alpha rhythm is strongest — varies from roughly 8 to 13 Hz across healthy adults, and shifts with age, arousal, and cognitive load. Using fixed band edges for everyone smears real subject-to-subject differences into noise. Some analyses instead define bands relative to each subject's IAF (e.g. "lower alpha" = IAF−4 to IAF−2 Hz). It is more work and often worth it.
What to take away
- The five bands are useful shorthand, not physiology.
- Whatever band edges you use, write them down and use them consistently across every recording you compare.
- Check that your bands tile the frequency range you're summing without gaps or overlaps.
- Treat scalp gamma with suspicion until you've ruled out muscle artifact.
The next article — EEG Band Power — covers how the numbers in each band are actually computed, and why the denominator problem above ripples through every relative-power result.
NeuroTrace computes absolute and relative band power per channel from an uploaded EDF, and lets you configure the band edges so you can match your lab's convention exactly.